linux/drivers/ata/libata-core.c

6941 lines
178 KiB
C
Raw Normal View History

/*
* libata-core.c - helper library for ATA
*
* Maintained by: Tejun Heo <tj@kernel.org>
* Please ALWAYS copy linux-ide@vger.kernel.org
* on emails.
*
* Copyright 2003-2004 Red Hat, Inc. All rights reserved.
* Copyright 2003-2004 Jeff Garzik
*
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2, or (at your option)
* any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*
*
* libata documentation is available via 'make {ps|pdf}docs',
* as Documentation/DocBook/libata.*
*
* Hardware documentation available from http://www.t13.org/ and
* http://www.sata-io.org/
*
* Standards documents from:
* http://www.t13.org (ATA standards, PCI DMA IDE spec)
* http://www.t10.org (SCSI MMC - for ATAPI MMC)
* http://www.sata-io.org (SATA)
* http://www.compactflash.org (CF)
* http://www.qic.org (QIC157 - Tape and DSC)
* http://www.ce-ata.org (CE-ATA: not supported)
*
*/
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/pci.h>
#include <linux/init.h>
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/spinlock.h>
#include <linux/blkdev.h>
#include <linux/delay.h>
#include <linux/timer.h>
#include <linux/interrupt.h>
#include <linux/completion.h>
#include <linux/suspend.h>
#include <linux/workqueue.h>
#include <linux/scatterlist.h>
#include <linux/io.h>
#include <linux/async.h>
#include <linux/log2.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/slab.h>
#include <linux/glob.h>
#include <scsi/scsi.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_host.h>
#include <linux/libata.h>
#include <asm/byteorder.h>
#include <linux/cdrom.h>
#include <linux/ratelimit.h>
#include <linux/pm_runtime.h>
#include <linux/platform_device.h>
#include "libata.h"
#include "libata-transport.h"
/* debounce timing parameters in msecs { interval, duration, timeout } */
const unsigned long sata_deb_timing_normal[] = { 5, 100, 2000 };
const unsigned long sata_deb_timing_hotplug[] = { 25, 500, 2000 };
const unsigned long sata_deb_timing_long[] = { 100, 2000, 5000 };
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
const struct ata_port_operations ata_base_port_ops = {
.prereset = ata_std_prereset,
.postreset = ata_std_postreset,
libata: make reset related methods proper port operations Currently reset methods are not specified directly in the ata_port_operations table. If a LLD wants to use custom reset methods, it should construct and use a error_handler which uses those reset methods. It's done this way for two reasons. First, the ops table already contained too many methods and adding four more of them would noticeably increase the amount of necessary boilerplate code all over low level drivers. Second, as ->error_handler uses those reset methods, it can get confusing. ie. By overriding ->error_handler, those reset ops can be made useless making layering a bit hazy. Now that ops table uses inheritance, the first problem doesn't exist anymore. The second isn't completely solved but is relieved by providing default values - most drivers can just override what it has implemented and don't have to concern itself about higher level callbacks. In fact, there currently is no driver which actually modifies error handling behavior. Drivers which override ->error_handler just wraps the standard error handler only to prepare the controller for EH. I don't think making ops layering strict has any noticeable benefit. This patch makes ->prereset, ->softreset, ->hardreset, ->postreset and their PMP counterparts propoer ops. Default ops are provided in the base ops tables and drivers are converted to override individual reset methods instead of creating custom error_handler. * ata_std_error_handler() doesn't use sata_std_hardreset() if SCRs aren't accessible. sata_promise doesn't need to use separate error_handlers for PATA and SATA anymore. * softreset is broken for sata_inic162x and sata_sx4. As libata now always prefers hardreset, this doesn't really matter but the ops are forced to NULL using ATA_OP_NULL for documentation purpose. * pata_hpt374 needs to use different prereset for the first and second PCI functions. This used to be done by branching from hpt374_error_handler(). The proper way to do this is to use separate ops and port_info tables for each function. Converted. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:50 +08:00
.error_handler = ata_std_error_handler,
.sched_eh = ata_std_sched_eh,
.end_eh = ata_std_end_eh,
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
};
const struct ata_port_operations sata_port_ops = {
.inherits = &ata_base_port_ops,
.qc_defer = ata_std_qc_defer,
.hardreset = sata_std_hardreset,
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
};
static unsigned int ata_dev_init_params(struct ata_device *dev,
u16 heads, u16 sectors);
static unsigned int ata_dev_set_xfermode(struct ata_device *dev);
static void ata_dev_xfermask(struct ata_device *dev);
static unsigned long ata_dev_blacklisted(const struct ata_device *dev);
atomic_t ata_print_id = ATOMIC_INIT(0);
struct ata_force_param {
const char *name;
unsigned int cbl;
int spd_limit;
unsigned long xfer_mask;
unsigned int horkage_on;
unsigned int horkage_off;
unsigned int lflags;
};
struct ata_force_ent {
int port;
int device;
struct ata_force_param param;
};
static struct ata_force_ent *ata_force_tbl;
static int ata_force_tbl_size;
static char ata_force_param_buf[PAGE_SIZE] __initdata;
/* param_buf is thrown away after initialization, disallow read */
module_param_string(force, ata_force_param_buf, sizeof(ata_force_param_buf), 0);
MODULE_PARM_DESC(force, "Force ATA configurations including cable type, link speed and transfer mode (see Documentation/kernel-parameters.txt for details)");
static int atapi_enabled = 1;
module_param(atapi_enabled, int, 0444);
MODULE_PARM_DESC(atapi_enabled, "Enable discovery of ATAPI devices (0=off, 1=on [default])");
static int atapi_dmadir = 0;
module_param(atapi_dmadir, int, 0444);
MODULE_PARM_DESC(atapi_dmadir, "Enable ATAPI DMADIR bridge support (0=off [default], 1=on)");
int atapi_passthru16 = 1;
module_param(atapi_passthru16, int, 0444);
MODULE_PARM_DESC(atapi_passthru16, "Enable ATA_16 passthru for ATAPI devices (0=off, 1=on [default])");
int libata_fua = 0;
module_param_named(fua, libata_fua, int, 0444);
MODULE_PARM_DESC(fua, "FUA support (0=off [default], 1=on)");
static int ata_ignore_hpa;
module_param_named(ignore_hpa, ata_ignore_hpa, int, 0644);
MODULE_PARM_DESC(ignore_hpa, "Ignore HPA limit (0=keep BIOS limits, 1=ignore limits, using full disk)");
static int libata_dma_mask = ATA_DMA_MASK_ATA|ATA_DMA_MASK_ATAPI|ATA_DMA_MASK_CFA;
module_param_named(dma, libata_dma_mask, int, 0444);
MODULE_PARM_DESC(dma, "DMA enable/disable (0x1==ATA, 0x2==ATAPI, 0x4==CF)");
static int ata_probe_timeout;
module_param(ata_probe_timeout, int, 0444);
MODULE_PARM_DESC(ata_probe_timeout, "Set ATA probing timeout (seconds)");
int libata_noacpi = 0;
module_param_named(noacpi, libata_noacpi, int, 0444);
MODULE_PARM_DESC(noacpi, "Disable the use of ACPI in probe/suspend/resume (0=off [default], 1=on)");
int libata_allow_tpm = 0;
module_param_named(allow_tpm, libata_allow_tpm, int, 0444);
MODULE_PARM_DESC(allow_tpm, "Permit the use of TPM commands (0=off [default], 1=on)");
static int atapi_an;
module_param(atapi_an, int, 0444);
MODULE_PARM_DESC(atapi_an, "Enable ATAPI AN media presence notification (0=0ff [default], 1=on)");
MODULE_AUTHOR("Jeff Garzik");
MODULE_DESCRIPTION("Library module for ATA devices");
MODULE_LICENSE("GPL");
MODULE_VERSION(DRV_VERSION);
2005-06-03 06:17:13 +08:00
static bool ata_sstatus_online(u32 sstatus)
{
return (sstatus & 0xf) == 0x3;
}
/**
* ata_link_next - link iteration helper
* @link: the previous link, NULL to start
* @ap: ATA port containing links to iterate
* @mode: iteration mode, one of ATA_LITER_*
*
* LOCKING:
* Host lock or EH context.
*
* RETURNS:
* Pointer to the next link.
*/
struct ata_link *ata_link_next(struct ata_link *link, struct ata_port *ap,
enum ata_link_iter_mode mode)
{
BUG_ON(mode != ATA_LITER_EDGE &&
mode != ATA_LITER_PMP_FIRST && mode != ATA_LITER_HOST_FIRST);
/* NULL link indicates start of iteration */
if (!link)
switch (mode) {
case ATA_LITER_EDGE:
case ATA_LITER_PMP_FIRST:
if (sata_pmp_attached(ap))
return ap->pmp_link;
/* fall through */
case ATA_LITER_HOST_FIRST:
return &ap->link;
}
/* we just iterated over the host link, what's next? */
if (link == &ap->link)
switch (mode) {
case ATA_LITER_HOST_FIRST:
if (sata_pmp_attached(ap))
return ap->pmp_link;
/* fall through */
case ATA_LITER_PMP_FIRST:
if (unlikely(ap->slave_link))
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
return ap->slave_link;
/* fall through */
case ATA_LITER_EDGE:
return NULL;
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
}
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
/* slave_link excludes PMP */
if (unlikely(link == ap->slave_link))
return NULL;
/* we were over a PMP link */
if (++link < ap->pmp_link + ap->nr_pmp_links)
return link;
if (mode == ATA_LITER_PMP_FIRST)
return &ap->link;
return NULL;
}
/**
* ata_dev_next - device iteration helper
* @dev: the previous device, NULL to start
* @link: ATA link containing devices to iterate
* @mode: iteration mode, one of ATA_DITER_*
*
* LOCKING:
* Host lock or EH context.
*
* RETURNS:
* Pointer to the next device.
*/
struct ata_device *ata_dev_next(struct ata_device *dev, struct ata_link *link,
enum ata_dev_iter_mode mode)
{
BUG_ON(mode != ATA_DITER_ENABLED && mode != ATA_DITER_ENABLED_REVERSE &&
mode != ATA_DITER_ALL && mode != ATA_DITER_ALL_REVERSE);
/* NULL dev indicates start of iteration */
if (!dev)
switch (mode) {
case ATA_DITER_ENABLED:
case ATA_DITER_ALL:
dev = link->device;
goto check;
case ATA_DITER_ENABLED_REVERSE:
case ATA_DITER_ALL_REVERSE:
dev = link->device + ata_link_max_devices(link) - 1;
goto check;
}
next:
/* move to the next one */
switch (mode) {
case ATA_DITER_ENABLED:
case ATA_DITER_ALL:
if (++dev < link->device + ata_link_max_devices(link))
goto check;
return NULL;
case ATA_DITER_ENABLED_REVERSE:
case ATA_DITER_ALL_REVERSE:
if (--dev >= link->device)
goto check;
return NULL;
}
check:
if ((mode == ATA_DITER_ENABLED || mode == ATA_DITER_ENABLED_REVERSE) &&
!ata_dev_enabled(dev))
goto next;
return dev;
}
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
/**
* ata_dev_phys_link - find physical link for a device
* @dev: ATA device to look up physical link for
*
* Look up physical link which @dev is attached to. Note that
* this is different from @dev->link only when @dev is on slave
* link. For all other cases, it's the same as @dev->link.
*
* LOCKING:
* Don't care.
*
* RETURNS:
* Pointer to the found physical link.
*/
struct ata_link *ata_dev_phys_link(struct ata_device *dev)
{
struct ata_port *ap = dev->link->ap;
if (!ap->slave_link)
return dev->link;
if (!dev->devno)
return &ap->link;
return ap->slave_link;
}
/**
* ata_force_cbl - force cable type according to libata.force
* @ap: ATA port of interest
*
* Force cable type according to libata.force and whine about it.
* The last entry which has matching port number is used, so it
* can be specified as part of device force parameters. For
* example, both "a:40c,1.00:udma4" and "1.00:40c,udma4" have the
* same effect.
*
* LOCKING:
* EH context.
*/
void ata_force_cbl(struct ata_port *ap)
{
int i;
for (i = ata_force_tbl_size - 1; i >= 0; i--) {
const struct ata_force_ent *fe = &ata_force_tbl[i];
if (fe->port != -1 && fe->port != ap->print_id)
continue;
if (fe->param.cbl == ATA_CBL_NONE)
continue;
ap->cbl = fe->param.cbl;
ata_port_notice(ap, "FORCE: cable set to %s\n", fe->param.name);
return;
}
}
/**
* ata_force_link_limits - force link limits according to libata.force
* @link: ATA link of interest
*
* Force link flags and SATA spd limit according to libata.force
* and whine about it. When only the port part is specified
* (e.g. 1:), the limit applies to all links connected to both
* the host link and all fan-out ports connected via PMP. If the
* device part is specified as 0 (e.g. 1.00:), it specifies the
* first fan-out link not the host link. Device number 15 always
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
* points to the host link whether PMP is attached or not. If the
* controller has slave link, device number 16 points to it.
*
* LOCKING:
* EH context.
*/
static void ata_force_link_limits(struct ata_link *link)
{
bool did_spd = false;
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
int linkno = link->pmp;
int i;
if (ata_is_host_link(link))
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
linkno += 15;
for (i = ata_force_tbl_size - 1; i >= 0; i--) {
const struct ata_force_ent *fe = &ata_force_tbl[i];
if (fe->port != -1 && fe->port != link->ap->print_id)
continue;
if (fe->device != -1 && fe->device != linkno)
continue;
/* only honor the first spd limit */
if (!did_spd && fe->param.spd_limit) {
link->hw_sata_spd_limit = (1 << fe->param.spd_limit) - 1;
ata_link_notice(link, "FORCE: PHY spd limit set to %s\n",
fe->param.name);
did_spd = true;
}
/* let lflags stack */
if (fe->param.lflags) {
link->flags |= fe->param.lflags;
ata_link_notice(link,
"FORCE: link flag 0x%x forced -> 0x%x\n",
fe->param.lflags, link->flags);
}
}
}
/**
* ata_force_xfermask - force xfermask according to libata.force
* @dev: ATA device of interest
*
* Force xfer_mask according to libata.force and whine about it.
* For consistency with link selection, device number 15 selects
* the first device connected to the host link.
*
* LOCKING:
* EH context.
*/
static void ata_force_xfermask(struct ata_device *dev)
{
int devno = dev->link->pmp + dev->devno;
int alt_devno = devno;
int i;
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
/* allow n.15/16 for devices attached to host port */
if (ata_is_host_link(dev->link))
alt_devno += 15;
for (i = ata_force_tbl_size - 1; i >= 0; i--) {
const struct ata_force_ent *fe = &ata_force_tbl[i];
unsigned long pio_mask, mwdma_mask, udma_mask;
if (fe->port != -1 && fe->port != dev->link->ap->print_id)
continue;
if (fe->device != -1 && fe->device != devno &&
fe->device != alt_devno)
continue;
if (!fe->param.xfer_mask)
continue;
ata_unpack_xfermask(fe->param.xfer_mask,
&pio_mask, &mwdma_mask, &udma_mask);
if (udma_mask)
dev->udma_mask = udma_mask;
else if (mwdma_mask) {
dev->udma_mask = 0;
dev->mwdma_mask = mwdma_mask;
} else {
dev->udma_mask = 0;
dev->mwdma_mask = 0;
dev->pio_mask = pio_mask;
}
ata_dev_notice(dev, "FORCE: xfer_mask set to %s\n",
fe->param.name);
return;
}
}
/**
* ata_force_horkage - force horkage according to libata.force
* @dev: ATA device of interest
*
* Force horkage according to libata.force and whine about it.
* For consistency with link selection, device number 15 selects
* the first device connected to the host link.
*
* LOCKING:
* EH context.
*/
static void ata_force_horkage(struct ata_device *dev)
{
int devno = dev->link->pmp + dev->devno;
int alt_devno = devno;
int i;
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
/* allow n.15/16 for devices attached to host port */
if (ata_is_host_link(dev->link))
alt_devno += 15;
for (i = 0; i < ata_force_tbl_size; i++) {
const struct ata_force_ent *fe = &ata_force_tbl[i];
if (fe->port != -1 && fe->port != dev->link->ap->print_id)
continue;
if (fe->device != -1 && fe->device != devno &&
fe->device != alt_devno)
continue;
if (!(~dev->horkage & fe->param.horkage_on) &&
!(dev->horkage & fe->param.horkage_off))
continue;
dev->horkage |= fe->param.horkage_on;
dev->horkage &= ~fe->param.horkage_off;
ata_dev_notice(dev, "FORCE: horkage modified (%s)\n",
fe->param.name);
}
}
/**
* atapi_cmd_type - Determine ATAPI command type from SCSI opcode
* @opcode: SCSI opcode
*
* Determine ATAPI command type from @opcode.
*
* LOCKING:
* None.
*
* RETURNS:
* ATAPI_{READ|WRITE|READ_CD|PASS_THRU|MISC}
*/
int atapi_cmd_type(u8 opcode)
{
switch (opcode) {
case GPCMD_READ_10:
case GPCMD_READ_12:
return ATAPI_READ;
case GPCMD_WRITE_10:
case GPCMD_WRITE_12:
case GPCMD_WRITE_AND_VERIFY_10:
return ATAPI_WRITE;
case GPCMD_READ_CD:
case GPCMD_READ_CD_MSF:
return ATAPI_READ_CD;
case ATA_16:
case ATA_12:
if (atapi_passthru16)
return ATAPI_PASS_THRU;
/* fall thru */
default:
return ATAPI_MISC;
}
}
/**
* ata_tf_to_fis - Convert ATA taskfile to SATA FIS structure
* @tf: Taskfile to convert
* @pmp: Port multiplier port
* @is_cmd: This FIS is for command
* @fis: Buffer into which data will output
*
* Converts a standard ATA taskfile to a Serial ATA
* FIS structure (Register - Host to Device).
*
* LOCKING:
* Inherited from caller.
*/
void ata_tf_to_fis(const struct ata_taskfile *tf, u8 pmp, int is_cmd, u8 *fis)
{
fis[0] = 0x27; /* Register - Host to Device FIS */
fis[1] = pmp & 0xf; /* Port multiplier number*/
if (is_cmd)
fis[1] |= (1 << 7); /* bit 7 indicates Command FIS */
fis[2] = tf->command;
fis[3] = tf->feature;
fis[4] = tf->lbal;
fis[5] = tf->lbam;
fis[6] = tf->lbah;
fis[7] = tf->device;
fis[8] = tf->hob_lbal;
fis[9] = tf->hob_lbam;
fis[10] = tf->hob_lbah;
fis[11] = tf->hob_feature;
fis[12] = tf->nsect;
fis[13] = tf->hob_nsect;
fis[14] = 0;
fis[15] = tf->ctl;
fis[16] = tf->auxiliary & 0xff;
fis[17] = (tf->auxiliary >> 8) & 0xff;
fis[18] = (tf->auxiliary >> 16) & 0xff;
fis[19] = (tf->auxiliary >> 24) & 0xff;
}
/**
* ata_tf_from_fis - Convert SATA FIS to ATA taskfile
* @fis: Buffer from which data will be input
* @tf: Taskfile to output
*
* Converts a serial ATA FIS structure to a standard ATA taskfile.
*
* LOCKING:
* Inherited from caller.
*/
void ata_tf_from_fis(const u8 *fis, struct ata_taskfile *tf)
{
tf->command = fis[2]; /* status */
tf->feature = fis[3]; /* error */
tf->lbal = fis[4];
tf->lbam = fis[5];
tf->lbah = fis[6];
tf->device = fis[7];
tf->hob_lbal = fis[8];
tf->hob_lbam = fis[9];
tf->hob_lbah = fis[10];
tf->nsect = fis[12];
tf->hob_nsect = fis[13];
}
static const u8 ata_rw_cmds[] = {
/* pio multi */
ATA_CMD_READ_MULTI,
ATA_CMD_WRITE_MULTI,
ATA_CMD_READ_MULTI_EXT,
ATA_CMD_WRITE_MULTI_EXT,
0,
0,
0,
ATA_CMD_WRITE_MULTI_FUA_EXT,
/* pio */
ATA_CMD_PIO_READ,
ATA_CMD_PIO_WRITE,
ATA_CMD_PIO_READ_EXT,
ATA_CMD_PIO_WRITE_EXT,
0,
0,
0,
0,
/* dma */
ATA_CMD_READ,
ATA_CMD_WRITE,
ATA_CMD_READ_EXT,
ATA_CMD_WRITE_EXT,
0,
0,
0,
ATA_CMD_WRITE_FUA_EXT
};
/**
* ata_rwcmd_protocol - set taskfile r/w commands and protocol
* @tf: command to examine and configure
* @dev: device tf belongs to
*
* Examine the device configuration and tf->flags to calculate
* the proper read/write commands and protocol to use.
*
* LOCKING:
* caller.
*/
static int ata_rwcmd_protocol(struct ata_taskfile *tf, struct ata_device *dev)
{
u8 cmd;
int index, fua, lba48, write;
fua = (tf->flags & ATA_TFLAG_FUA) ? 4 : 0;
lba48 = (tf->flags & ATA_TFLAG_LBA48) ? 2 : 0;
write = (tf->flags & ATA_TFLAG_WRITE) ? 1 : 0;
if (dev->flags & ATA_DFLAG_PIO) {
tf->protocol = ATA_PROT_PIO;
index = dev->multi_count ? 0 : 8;
} else if (lba48 && (dev->link->ap->flags & ATA_FLAG_PIO_LBA48)) {
/* Unable to use DMA due to host limitation */
tf->protocol = ATA_PROT_PIO;
index = dev->multi_count ? 0 : 8;
} else {
tf->protocol = ATA_PROT_DMA;
index = 16;
}
cmd = ata_rw_cmds[index + fua + lba48 + write];
if (cmd) {
tf->command = cmd;
return 0;
}
return -1;
}
/**
* ata_tf_read_block - Read block address from ATA taskfile
* @tf: ATA taskfile of interest
* @dev: ATA device @tf belongs to
*
* LOCKING:
* None.
*
* Read block address from @tf. This function can handle all
* three address formats - LBA, LBA48 and CHS. tf->protocol and
* flags select the address format to use.
*
* RETURNS:
* Block address read from @tf.
*/
u64 ata_tf_read_block(struct ata_taskfile *tf, struct ata_device *dev)
{
u64 block = 0;
if (tf->flags & ATA_TFLAG_LBA) {
if (tf->flags & ATA_TFLAG_LBA48) {
block |= (u64)tf->hob_lbah << 40;
block |= (u64)tf->hob_lbam << 32;
block |= (u64)tf->hob_lbal << 24;
} else
block |= (tf->device & 0xf) << 24;
block |= tf->lbah << 16;
block |= tf->lbam << 8;
block |= tf->lbal;
} else {
u32 cyl, head, sect;
cyl = tf->lbam | (tf->lbah << 8);
head = tf->device & 0xf;
sect = tf->lbal;
if (!sect) {
ata_dev_warn(dev,
"device reported invalid CHS sector 0\n");
sect = 1; /* oh well */
}
block = (cyl * dev->heads + head) * dev->sectors + sect - 1;
}
return block;
}
/**
* ata_build_rw_tf - Build ATA taskfile for given read/write request
* @tf: Target ATA taskfile
* @dev: ATA device @tf belongs to
* @block: Block address
* @n_block: Number of blocks
* @tf_flags: RW/FUA etc...
* @tag: tag
*
* LOCKING:
* None.
*
* Build ATA taskfile @tf for read/write request described by
* @block, @n_block, @tf_flags and @tag on @dev.
*
* RETURNS:
*
* 0 on success, -ERANGE if the request is too large for @dev,
* -EINVAL if the request is invalid.
*/
int ata_build_rw_tf(struct ata_taskfile *tf, struct ata_device *dev,
u64 block, u32 n_block, unsigned int tf_flags,
unsigned int tag)
{
tf->flags |= ATA_TFLAG_ISADDR | ATA_TFLAG_DEVICE;
tf->flags |= tf_flags;
if (ata_ncq_enabled(dev) && likely(tag != ATA_TAG_INTERNAL)) {
/* yay, NCQ */
if (!lba_48_ok(block, n_block))
return -ERANGE;
tf->protocol = ATA_PROT_NCQ;
tf->flags |= ATA_TFLAG_LBA | ATA_TFLAG_LBA48;
if (tf->flags & ATA_TFLAG_WRITE)
tf->command = ATA_CMD_FPDMA_WRITE;
else
tf->command = ATA_CMD_FPDMA_READ;
tf->nsect = tag << 3;
tf->hob_feature = (n_block >> 8) & 0xff;
tf->feature = n_block & 0xff;
tf->hob_lbah = (block >> 40) & 0xff;
tf->hob_lbam = (block >> 32) & 0xff;
tf->hob_lbal = (block >> 24) & 0xff;
tf->lbah = (block >> 16) & 0xff;
tf->lbam = (block >> 8) & 0xff;
tf->lbal = block & 0xff;
tf->device = ATA_LBA;
if (tf->flags & ATA_TFLAG_FUA)
tf->device |= 1 << 7;
} else if (dev->flags & ATA_DFLAG_LBA) {
tf->flags |= ATA_TFLAG_LBA;
if (lba_28_ok(block, n_block)) {
/* use LBA28 */
tf->device |= (block >> 24) & 0xf;
} else if (lba_48_ok(block, n_block)) {
if (!(dev->flags & ATA_DFLAG_LBA48))
return -ERANGE;
/* use LBA48 */
tf->flags |= ATA_TFLAG_LBA48;
tf->hob_nsect = (n_block >> 8) & 0xff;
tf->hob_lbah = (block >> 40) & 0xff;
tf->hob_lbam = (block >> 32) & 0xff;
tf->hob_lbal = (block >> 24) & 0xff;
} else
/* request too large even for LBA48 */
return -ERANGE;
if (unlikely(ata_rwcmd_protocol(tf, dev) < 0))
return -EINVAL;
tf->nsect = n_block & 0xff;
tf->lbah = (block >> 16) & 0xff;
tf->lbam = (block >> 8) & 0xff;
tf->lbal = block & 0xff;
tf->device |= ATA_LBA;
} else {
/* CHS */
u32 sect, head, cyl, track;
/* The request -may- be too large for CHS addressing. */
if (!lba_28_ok(block, n_block))
return -ERANGE;
if (unlikely(ata_rwcmd_protocol(tf, dev) < 0))
return -EINVAL;
/* Convert LBA to CHS */
track = (u32)block / dev->sectors;
cyl = track / dev->heads;
head = track % dev->heads;
sect = (u32)block % dev->sectors + 1;
DPRINTK("block %u track %u cyl %u head %u sect %u\n",
(u32)block, track, cyl, head, sect);
/* Check whether the converted CHS can fit.
Cylinder: 0-65535
Head: 0-15
Sector: 1-255*/
if ((cyl >> 16) || (head >> 4) || (sect >> 8) || (!sect))
return -ERANGE;
tf->nsect = n_block & 0xff; /* Sector count 0 means 256 sectors */
tf->lbal = sect;
tf->lbam = cyl;
tf->lbah = cyl >> 8;
tf->device |= head;
}
return 0;
}
/**
* ata_pack_xfermask - Pack pio, mwdma and udma masks into xfer_mask
* @pio_mask: pio_mask
* @mwdma_mask: mwdma_mask
* @udma_mask: udma_mask
*
* Pack @pio_mask, @mwdma_mask and @udma_mask into a single
* unsigned int xfer_mask.
*
* LOCKING:
* None.
*
* RETURNS:
* Packed xfer_mask.
*/
unsigned long ata_pack_xfermask(unsigned long pio_mask,
unsigned long mwdma_mask,
unsigned long udma_mask)
{
return ((pio_mask << ATA_SHIFT_PIO) & ATA_MASK_PIO) |
((mwdma_mask << ATA_SHIFT_MWDMA) & ATA_MASK_MWDMA) |
((udma_mask << ATA_SHIFT_UDMA) & ATA_MASK_UDMA);
}
/**
* ata_unpack_xfermask - Unpack xfer_mask into pio, mwdma and udma masks
* @xfer_mask: xfer_mask to unpack
* @pio_mask: resulting pio_mask
* @mwdma_mask: resulting mwdma_mask
* @udma_mask: resulting udma_mask
*
* Unpack @xfer_mask into @pio_mask, @mwdma_mask and @udma_mask.
* Any NULL distination masks will be ignored.
*/
void ata_unpack_xfermask(unsigned long xfer_mask, unsigned long *pio_mask,
unsigned long *mwdma_mask, unsigned long *udma_mask)
{
if (pio_mask)
*pio_mask = (xfer_mask & ATA_MASK_PIO) >> ATA_SHIFT_PIO;
if (mwdma_mask)
*mwdma_mask = (xfer_mask & ATA_MASK_MWDMA) >> ATA_SHIFT_MWDMA;
if (udma_mask)
*udma_mask = (xfer_mask & ATA_MASK_UDMA) >> ATA_SHIFT_UDMA;
}
static const struct ata_xfer_ent {
int shift, bits;
u8 base;
} ata_xfer_tbl[] = {
{ ATA_SHIFT_PIO, ATA_NR_PIO_MODES, XFER_PIO_0 },
{ ATA_SHIFT_MWDMA, ATA_NR_MWDMA_MODES, XFER_MW_DMA_0 },
{ ATA_SHIFT_UDMA, ATA_NR_UDMA_MODES, XFER_UDMA_0 },
{ -1, },
};
/**
* ata_xfer_mask2mode - Find matching XFER_* for the given xfer_mask
* @xfer_mask: xfer_mask of interest
*
* Return matching XFER_* value for @xfer_mask. Only the highest
* bit of @xfer_mask is considered.
*
* LOCKING:
* None.
*
* RETURNS:
* Matching XFER_* value, 0xff if no match found.
*/
u8 ata_xfer_mask2mode(unsigned long xfer_mask)
{
int highbit = fls(xfer_mask) - 1;
const struct ata_xfer_ent *ent;
for (ent = ata_xfer_tbl; ent->shift >= 0; ent++)
if (highbit >= ent->shift && highbit < ent->shift + ent->bits)
return ent->base + highbit - ent->shift;
return 0xff;
}
/**
* ata_xfer_mode2mask - Find matching xfer_mask for XFER_*
* @xfer_mode: XFER_* of interest
*
* Return matching xfer_mask for @xfer_mode.
*
* LOCKING:
* None.
*
* RETURNS:
* Matching xfer_mask, 0 if no match found.
*/
unsigned long ata_xfer_mode2mask(u8 xfer_mode)
{
const struct ata_xfer_ent *ent;
for (ent = ata_xfer_tbl; ent->shift >= 0; ent++)
if (xfer_mode >= ent->base && xfer_mode < ent->base + ent->bits)
return ((2 << (ent->shift + xfer_mode - ent->base)) - 1)
& ~((1 << ent->shift) - 1);
return 0;
}
/**
* ata_xfer_mode2shift - Find matching xfer_shift for XFER_*
* @xfer_mode: XFER_* of interest
*
* Return matching xfer_shift for @xfer_mode.
*
* LOCKING:
* None.
*
* RETURNS:
* Matching xfer_shift, -1 if no match found.
*/
int ata_xfer_mode2shift(unsigned long xfer_mode)
{
const struct ata_xfer_ent *ent;
for (ent = ata_xfer_tbl; ent->shift >= 0; ent++)
if (xfer_mode >= ent->base && xfer_mode < ent->base + ent->bits)
return ent->shift;
return -1;
}
/**
* ata_mode_string - convert xfer_mask to string
* @xfer_mask: mask of bits supported; only highest bit counts.
*
* Determine string which represents the highest speed
* (highest bit in @modemask).
*
* LOCKING:
* None.
*
* RETURNS:
* Constant C string representing highest speed listed in
* @mode_mask, or the constant C string "<n/a>".
*/
const char *ata_mode_string(unsigned long xfer_mask)
{
static const char * const xfer_mode_str[] = {
"PIO0",
"PIO1",
"PIO2",
"PIO3",
"PIO4",
"PIO5",
"PIO6",
"MWDMA0",
"MWDMA1",
"MWDMA2",
"MWDMA3",
"MWDMA4",
"UDMA/16",
"UDMA/25",
"UDMA/33",
"UDMA/44",
"UDMA/66",
"UDMA/100",
"UDMA/133",
"UDMA7",
};
int highbit;
highbit = fls(xfer_mask) - 1;
if (highbit >= 0 && highbit < ARRAY_SIZE(xfer_mode_str))
return xfer_mode_str[highbit];
return "<n/a>";
}
const char *sata_spd_string(unsigned int spd)
{
static const char * const spd_str[] = {
"1.5 Gbps",
"3.0 Gbps",
"6.0 Gbps",
};
if (spd == 0 || (spd - 1) >= ARRAY_SIZE(spd_str))
return "<unknown>";
return spd_str[spd - 1];
}
/**
* ata_dev_classify - determine device type based on ATA-spec signature
* @tf: ATA taskfile register set for device to be identified
*
* Determine from taskfile register contents whether a device is
* ATA or ATAPI, as per "Signature and persistence" section
* of ATA/PI spec (volume 1, sect 5.14).
*
* LOCKING:
* None.
*
* RETURNS:
* Device type, %ATA_DEV_ATA, %ATA_DEV_ATAPI, %ATA_DEV_PMP,
* %ATA_DEV_ZAC, or %ATA_DEV_UNKNOWN the event of failure.
*/
unsigned int ata_dev_classify(const struct ata_taskfile *tf)
{
/* Apple's open source Darwin code hints that some devices only
* put a proper signature into the LBA mid/high registers,
* So, we only check those. It's sufficient for uniqueness.
*
* ATA/ATAPI-7 (d1532v1r1: Feb. 19, 2003) specified separate
* signatures for ATA and ATAPI devices attached on SerialATA,
* 0x3c/0xc3 and 0x69/0x96 respectively. However, SerialATA
* spec has never mentioned about using different signatures
* for ATA/ATAPI devices. Then, Serial ATA II: Port
* Multiplier specification began to use 0x69/0x96 to identify
* port multpliers and 0x3c/0xc3 to identify SEMB device.
* ATA/ATAPI-7 dropped descriptions about 0x3c/0xc3 and
* 0x69/0x96 shortly and described them as reserved for
* SerialATA.
*
* We follow the current spec and consider that 0x69/0x96
* identifies a port multiplier and 0x3c/0xc3 a SEMB device.
* Unfortunately, WDC WD1600JS-62MHB5 (a hard drive) reports
* SEMB signature. This is worked around in
* ata_dev_read_id().
*/
if ((tf->lbam == 0) && (tf->lbah == 0)) {
DPRINTK("found ATA device by sig\n");
return ATA_DEV_ATA;
}
if ((tf->lbam == 0x14) && (tf->lbah == 0xeb)) {
DPRINTK("found ATAPI device by sig\n");
return ATA_DEV_ATAPI;
}
if ((tf->lbam == 0x69) && (tf->lbah == 0x96)) {
DPRINTK("found PMP device by sig\n");
return ATA_DEV_PMP;
}
if ((tf->lbam == 0x3c) && (tf->lbah == 0xc3)) {
DPRINTK("found SEMB device by sig (could be ATA device)\n");
return ATA_DEV_SEMB;
}
if ((tf->lbam == 0xcd) && (tf->lbah == 0xab)) {
DPRINTK("found ZAC device by sig\n");
return ATA_DEV_ZAC;
}
DPRINTK("unknown device\n");
return ATA_DEV_UNKNOWN;
}
/**
* ata_id_string - Convert IDENTIFY DEVICE page into string
* @id: IDENTIFY DEVICE results we will examine
* @s: string into which data is output
* @ofs: offset into identify device page
* @len: length of string to return. must be an even number.
*
* The strings in the IDENTIFY DEVICE page are broken up into
* 16-bit chunks. Run through the string, and output each
* 8-bit chunk linearly, regardless of platform.
*
* LOCKING:
* caller.
*/
void ata_id_string(const u16 *id, unsigned char *s,
unsigned int ofs, unsigned int len)
{
unsigned int c;
BUG_ON(len & 1);
while (len > 0) {
c = id[ofs] >> 8;
*s = c;
s++;
c = id[ofs] & 0xff;
*s = c;
s++;
ofs++;
len -= 2;
}
}
/**
* ata_id_c_string - Convert IDENTIFY DEVICE page into C string
* @id: IDENTIFY DEVICE results we will examine
* @s: string into which data is output
* @ofs: offset into identify device page
* @len: length of string to return. must be an odd number.
*
* This function is identical to ata_id_string except that it
* trims trailing spaces and terminates the resulting string with
* null. @len must be actual maximum length (even number) + 1.
*
* LOCKING:
* caller.
*/
void ata_id_c_string(const u16 *id, unsigned char *s,
unsigned int ofs, unsigned int len)
{
unsigned char *p;
ata_id_string(id, s, ofs, len - 1);
p = s + strnlen(s, len - 1);
while (p > s && p[-1] == ' ')
p--;
*p = '\0';
}
2005-06-03 06:17:13 +08:00
static u64 ata_id_n_sectors(const u16 *id)
{
if (ata_id_has_lba(id)) {
if (ata_id_has_lba48(id))
return ata_id_u64(id, ATA_ID_LBA_CAPACITY_2);
else
return ata_id_u32(id, ATA_ID_LBA_CAPACITY);
} else {
if (ata_id_current_chs_valid(id))
return id[ATA_ID_CUR_CYLS] * id[ATA_ID_CUR_HEADS] *
id[ATA_ID_CUR_SECTORS];
else
return id[ATA_ID_CYLS] * id[ATA_ID_HEADS] *
id[ATA_ID_SECTORS];
}
}
u64 ata_tf_to_lba48(const struct ata_taskfile *tf)
{
u64 sectors = 0;
sectors |= ((u64)(tf->hob_lbah & 0xff)) << 40;
sectors |= ((u64)(tf->hob_lbam & 0xff)) << 32;
sectors |= ((u64)(tf->hob_lbal & 0xff)) << 24;
sectors |= (tf->lbah & 0xff) << 16;
sectors |= (tf->lbam & 0xff) << 8;
sectors |= (tf->lbal & 0xff);
return sectors;
}
u64 ata_tf_to_lba(const struct ata_taskfile *tf)
{
u64 sectors = 0;
sectors |= (tf->device & 0x0f) << 24;
sectors |= (tf->lbah & 0xff) << 16;
sectors |= (tf->lbam & 0xff) << 8;
sectors |= (tf->lbal & 0xff);
return sectors;
}
/**
* ata_read_native_max_address - Read native max address
* @dev: target device
* @max_sectors: out parameter for the result native max address
*
* Perform an LBA48 or LBA28 native size query upon the device in
* question.
*
* RETURNS:
* 0 on success, -EACCES if command is aborted by the drive.
* -EIO on other errors.
*/
static int ata_read_native_max_address(struct ata_device *dev, u64 *max_sectors)
{
unsigned int err_mask;
struct ata_taskfile tf;
int lba48 = ata_id_has_lba48(dev->id);
ata_tf_init(dev, &tf);
/* always clear all address registers */
tf.flags |= ATA_TFLAG_DEVICE | ATA_TFLAG_ISADDR;
if (lba48) {
tf.command = ATA_CMD_READ_NATIVE_MAX_EXT;
tf.flags |= ATA_TFLAG_LBA48;
} else
tf.command = ATA_CMD_READ_NATIVE_MAX;
tf.protocol |= ATA_PROT_NODATA;
tf.device |= ATA_LBA;
err_mask = ata_exec_internal(dev, &tf, NULL, DMA_NONE, NULL, 0, 0);
if (err_mask) {
ata_dev_warn(dev,
"failed to read native max address (err_mask=0x%x)\n",
err_mask);
if (err_mask == AC_ERR_DEV && (tf.feature & ATA_ABORTED))
return -EACCES;
return -EIO;
}
if (lba48)
*max_sectors = ata_tf_to_lba48(&tf) + 1;
else
*max_sectors = ata_tf_to_lba(&tf) + 1;
if (dev->horkage & ATA_HORKAGE_HPA_SIZE)
(*max_sectors)--;
return 0;
}
/**
* ata_set_max_sectors - Set max sectors
* @dev: target device
* @new_sectors: new max sectors value to set for the device
*
* Set max sectors of @dev to @new_sectors.
*
* RETURNS:
* 0 on success, -EACCES if command is aborted or denied (due to
* previous non-volatile SET_MAX) by the drive. -EIO on other
* errors.
*/
static int ata_set_max_sectors(struct ata_device *dev, u64 new_sectors)
{
unsigned int err_mask;
struct ata_taskfile tf;
int lba48 = ata_id_has_lba48(dev->id);
new_sectors--;
ata_tf_init(dev, &tf);
tf.flags |= ATA_TFLAG_DEVICE | ATA_TFLAG_ISADDR;
if (lba48) {
tf.command = ATA_CMD_SET_MAX_EXT;
tf.flags |= ATA_TFLAG_LBA48;
tf.hob_lbal = (new_sectors >> 24) & 0xff;
tf.hob_lbam = (new_sectors >> 32) & 0xff;
tf.hob_lbah = (new_sectors >> 40) & 0xff;
} else {
tf.command = ATA_CMD_SET_MAX;
tf.device |= (new_sectors >> 24) & 0xf;
}
tf.protocol |= ATA_PROT_NODATA;
tf.device |= ATA_LBA;
tf.lbal = (new_sectors >> 0) & 0xff;
tf.lbam = (new_sectors >> 8) & 0xff;
tf.lbah = (new_sectors >> 16) & 0xff;
err_mask = ata_exec_internal(dev, &tf, NULL, DMA_NONE, NULL, 0, 0);
if (err_mask) {
ata_dev_warn(dev,
"failed to set max address (err_mask=0x%x)\n",
err_mask);
if (err_mask == AC_ERR_DEV &&
(tf.feature & (ATA_ABORTED | ATA_IDNF)))
return -EACCES;
return -EIO;
}
return 0;
}
/**
* ata_hpa_resize - Resize a device with an HPA set
* @dev: Device to resize
*
* Read the size of an LBA28 or LBA48 disk with HPA features and resize
* it if required to the full size of the media. The caller must check
* the drive has the HPA feature set enabled.
*
* RETURNS:
* 0 on success, -errno on failure.
*/
static int ata_hpa_resize(struct ata_device *dev)
{
struct ata_eh_context *ehc = &dev->link->eh_context;
int print_info = ehc->i.flags & ATA_EHI_PRINTINFO;
bool unlock_hpa = ata_ignore_hpa || dev->flags & ATA_DFLAG_UNLOCK_HPA;
u64 sectors = ata_id_n_sectors(dev->id);
u64 native_sectors;
int rc;
/* do we need to do it? */
if ((dev->class != ATA_DEV_ATA && dev->class != ATA_DEV_ZAC) ||
!ata_id_has_lba(dev->id) || !ata_id_hpa_enabled(dev->id) ||
(dev->horkage & ATA_HORKAGE_BROKEN_HPA))
return 0;
/* read native max address */
rc = ata_read_native_max_address(dev, &native_sectors);
if (rc) {
/* If device aborted the command or HPA isn't going to
* be unlocked, skip HPA resizing.
*/
if (rc == -EACCES || !unlock_hpa) {
ata_dev_warn(dev,
"HPA support seems broken, skipping HPA handling\n");
dev->horkage |= ATA_HORKAGE_BROKEN_HPA;
/* we can continue if device aborted the command */
if (rc == -EACCES)
rc = 0;
}
return rc;
}
dev->n_native_sectors = native_sectors;
/* nothing to do? */
if (native_sectors <= sectors || !unlock_hpa) {
if (!print_info || native_sectors == sectors)
return 0;
if (native_sectors > sectors)
ata_dev_info(dev,
"HPA detected: current %llu, native %llu\n",
(unsigned long long)sectors,
(unsigned long long)native_sectors);
else if (native_sectors < sectors)
ata_dev_warn(dev,
"native sectors (%llu) is smaller than sectors (%llu)\n",
(unsigned long long)native_sectors,
(unsigned long long)sectors);
return 0;
}
/* let's unlock HPA */
rc = ata_set_max_sectors(dev, native_sectors);
if (rc == -EACCES) {
/* if device aborted the command, skip HPA resizing */
ata_dev_warn(dev,
"device aborted resize (%llu -> %llu), skipping HPA handling\n",
(unsigned long long)sectors,
(unsigned long long)native_sectors);
dev->horkage |= ATA_HORKAGE_BROKEN_HPA;
return 0;
} else if (rc)
return rc;
/* re-read IDENTIFY data */
rc = ata_dev_reread_id(dev, 0);
if (rc) {
ata_dev_err(dev,
"failed to re-read IDENTIFY data after HPA resizing\n");
return rc;
}
if (print_info) {
u64 new_sectors = ata_id_n_sectors(dev->id);
ata_dev_info(dev,
"HPA unlocked: %llu -> %llu, native %llu\n",
(unsigned long long)sectors,
(unsigned long long)new_sectors,
(unsigned long long)native_sectors);
}
return 0;
}
/**
* ata_dump_id - IDENTIFY DEVICE info debugging output
* @id: IDENTIFY DEVICE page to dump
*
* Dump selected 16-bit words from the given IDENTIFY DEVICE
* page.
*
* LOCKING:
* caller.
*/
static inline void ata_dump_id(const u16 *id)
{
DPRINTK("49==0x%04x "
"53==0x%04x "
"63==0x%04x "
"64==0x%04x "
"75==0x%04x \n",
id[49],
id[53],
id[63],
id[64],
id[75]);
DPRINTK("80==0x%04x "
"81==0x%04x "
"82==0x%04x "
"83==0x%04x "
"84==0x%04x \n",
id[80],
id[81],
id[82],
id[83],
id[84]);
DPRINTK("88==0x%04x "
"93==0x%04x\n",
id[88],
id[93]);
}
/**
* ata_id_xfermask - Compute xfermask from the given IDENTIFY data
* @id: IDENTIFY data to compute xfer mask from
*
* Compute the xfermask for this device. This is not as trivial
* as it seems if we must consider early devices correctly.
*
* FIXME: pre IDE drive timing (do we care ?).
*
* LOCKING:
* None.
*
* RETURNS:
* Computed xfermask
*/
unsigned long ata_id_xfermask(const u16 *id)
{
unsigned long pio_mask, mwdma_mask, udma_mask;
/* Usual case. Word 53 indicates word 64 is valid */
if (id[ATA_ID_FIELD_VALID] & (1 << 1)) {
pio_mask = id[ATA_ID_PIO_MODES] & 0x03;
pio_mask <<= 3;
pio_mask |= 0x7;
} else {
/* If word 64 isn't valid then Word 51 high byte holds
* the PIO timing number for the maximum. Turn it into
* a mask.
*/
u8 mode = (id[ATA_ID_OLD_PIO_MODES] >> 8) & 0xFF;
if (mode < 5) /* Valid PIO range */
pio_mask = (2 << mode) - 1;
else
pio_mask = 1;
/* But wait.. there's more. Design your standards by
* committee and you too can get a free iordy field to
* process. However its the speeds not the modes that
* are supported... Note drivers using the timing API
* will get this right anyway
*/
}
mwdma_mask = id[ATA_ID_MWDMA_MODES] & 0x07;
if (ata_id_is_cfa(id)) {
/*
* Process compact flash extended modes
*/
int pio = (id[ATA_ID_CFA_MODES] >> 0) & 0x7;
int dma = (id[ATA_ID_CFA_MODES] >> 3) & 0x7;
if (pio)
pio_mask |= (1 << 5);
if (pio > 1)
pio_mask |= (1 << 6);
if (dma)
mwdma_mask |= (1 << 3);
if (dma > 1)
mwdma_mask |= (1 << 4);
}
udma_mask = 0;
if (id[ATA_ID_FIELD_VALID] & (1 << 2))
udma_mask = id[ATA_ID_UDMA_MODES] & 0xff;
return ata_pack_xfermask(pio_mask, mwdma_mask, udma_mask);
}
static void ata_qc_complete_internal(struct ata_queued_cmd *qc)
{
struct completion *waiting = qc->private_data;
complete(waiting);
}
/**
* ata_exec_internal_sg - execute libata internal command
* @dev: Device to which the command is sent
* @tf: Taskfile registers for the command and the result
* @cdb: CDB for packet command
* @dma_dir: Data transfer direction of the command
* @sgl: sg list for the data buffer of the command
* @n_elem: Number of sg entries
* @timeout: Timeout in msecs (0 for default)
*
* Executes libata internal command with timeout. @tf contains
* command on entry and result on return. Timeout and error
* conditions are reported via return value. No recovery action
* is taken after a command times out. It's caller's duty to
* clean up after timeout.
*
* LOCKING:
* None. Should be called with kernel context, might sleep.
*
* RETURNS:
* Zero on success, AC_ERR_* mask on failure
*/
unsigned ata_exec_internal_sg(struct ata_device *dev,
struct ata_taskfile *tf, const u8 *cdb,
int dma_dir, struct scatterlist *sgl,
unsigned int n_elem, unsigned long timeout)
{
struct ata_link *link = dev->link;
struct ata_port *ap = link->ap;
u8 command = tf->command;
int auto_timeout = 0;
struct ata_queued_cmd *qc;
unsigned int tag, preempted_tag;
u32 preempted_sactive, preempted_qc_active;
int preempted_nr_active_links;
DECLARE_COMPLETION_ONSTACK(wait);
unsigned long flags;
unsigned int err_mask;
int rc;
spin_lock_irqsave(ap->lock, flags);
/* no internal command while frozen */
if (ap->pflags & ATA_PFLAG_FROZEN) {
spin_unlock_irqrestore(ap->lock, flags);
return AC_ERR_SYSTEM;
}
/* initialize internal qc */
/* XXX: Tag 0 is used for drivers with legacy EH as some
* drivers choke if any other tag is given. This breaks
* ata_tag_internal() test for those drivers. Don't use new
* EH stuff without converting to it.
*/
if (ap->ops->error_handler)
tag = ATA_TAG_INTERNAL;
else
tag = 0;
qc = __ata_qc_from_tag(ap, tag);
qc->tag = tag;
qc->scsicmd = NULL;
qc->ap = ap;
qc->dev = dev;
ata_qc_reinit(qc);
preempted_tag = link->active_tag;
preempted_sactive = link->sactive;
preempted_qc_active = ap->qc_active;
preempted_nr_active_links = ap->nr_active_links;
link->active_tag = ATA_TAG_POISON;
link->sactive = 0;
ap->qc_active = 0;
ap->nr_active_links = 0;
/* prepare & issue qc */
qc->tf = *tf;
if (cdb)
memcpy(qc->cdb, cdb, ATAPI_CDB_LEN);
libata: make ata_exec_internal_sg honor DMADIR libata honors DMADIR for regular commands, but not for internal commands used (among other) during device initialisation. This makes SATA-host-to-PATA-device bridges based on Silicon Image SiL3611 (such as "Abit Serillel 2") end up disabled when used with an ATAPI device after a few tries. Log output of the bridge being hot-plugged with an ATAPI drive: [ 9631.212901] ata1: exception Emask 0x10 SAct 0x0 SErr 0x40c0000 action 0xe frozen [ 9631.212913] ata1: irq_stat 0x00000040, connection status changed [ 9631.212923] ata1: SError: { CommWake 10B8B DevExch } [ 9631.212939] ata1: hard resetting link [ 9632.104962] ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300) [ 9632.106393] ata1.00: ATAPI: PIONEER DVD-RW DVR-115, 1.06, max UDMA/33 [ 9632.106407] ata1.00: applying bridge limits [ 9632.108151] ata1.00: configured for UDMA/33 [ 9637.105303] ata1.00: qc timeout (cmd 0xa0) [ 9637.105324] ata1.00: failed to clear UNIT ATTENTION (err_mask=0x5) [ 9637.105335] ata1: hard resetting link [ 9638.044599] ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300) [ 9638.047878] ata1.00: configured for UDMA/33 [ 9643.044933] ata1.00: qc timeout (cmd 0xa0) [ 9643.044953] ata1.00: failed to clear UNIT ATTENTION (err_mask=0x5) [ 9643.044963] ata1: limiting SATA link speed to 1.5 Gbps [ 9643.044971] ata1.00: limiting speed to UDMA/33:PIO3 [ 9643.044979] ata1: hard resetting link [ 9643.984225] ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310) [ 9643.987471] ata1.00: configured for UDMA/33 [ 9648.984591] ata1.00: qc timeout (cmd 0xa0) [ 9648.984612] ata1.00: failed to clear UNIT ATTENTION (err_mask=0x5) [ 9648.984619] ata1.00: disabled [ 9649.000593] ata1: hard resetting link [ 9649.939902] ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310) [ 9649.955864] ata1: EH complete With this patch, the drive enumerates correctly when libata is loaded with atapi_dmadir=1: [ 9891.810863] ata1: exception Emask 0x10 SAct 0x0 SErr 0x40c0000 action 0xe frozen [ 9891.810874] ata1: irq_stat 0x00000040, connection status changed [ 9891.810884] ata1: SError: { CommWake 10B8B DevExch } [ 9891.810900] ata1: hard resetting link [ 9892.762105] ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300) [ 9892.763544] ata1.00: ATAPI: PIONEER DVD-RW DVR-115, 1.06, max UDMA/33, DMADIR [ 9892.763558] ata1.00: applying bridge limits [ 9892.765393] ata1.00: configured for UDMA/33 [ 9892.786063] ata1: EH complete [ 9892.792062] scsi 0:0:0:0: CD-ROM PIONEER DVD-RW DVR-115 1.06 PQ: 0 ANSI: 5 [ 9892.798455] sr2: scsi3-mmc drive: 12x/12x writer dvd-ram cd/rw xa/form2 cdda tray [ 9892.798837] sr 0:0:0:0: Attached scsi CD-ROM sr2 [ 9892.799109] sr 0:0:0:0: Attached scsi generic sg6 type 5 Based on a patch by Csaba Halász <csaba.halasz@gmail.com> on linux-ide: http://marc.info/?l=linux-ide&m=136121147832295&w=2 tj: minor formatting changes. Signed-off-by: Vincent Pelletier <plr.vincent@gmail.com> Signed-off-by: Tejun Heo <tj@kernel.org> Cc: stable@vger.kernel.org
2013-05-19 00:44:04 +08:00
/* some SATA bridges need us to indicate data xfer direction */
if (tf->protocol == ATAPI_PROT_DMA && (dev->flags & ATA_DFLAG_DMADIR) &&
dma_dir == DMA_FROM_DEVICE)
qc->tf.feature |= ATAPI_DMADIR;
qc->flags |= ATA_QCFLAG_RESULT_TF;
qc->dma_dir = dma_dir;
if (dma_dir != DMA_NONE) {
unsigned int i, buflen = 0;
struct scatterlist *sg;
for_each_sg(sgl, sg, n_elem, i)
buflen += sg->length;
ata_sg_init(qc, sgl, n_elem);
qc->nbytes = buflen;
}
qc->private_data = &wait;
qc->complete_fn = ata_qc_complete_internal;
ata_qc_issue(qc);
spin_unlock_irqrestore(ap->lock, flags);
if (!timeout) {
if (ata_probe_timeout)
timeout = ata_probe_timeout * 1000;
else {
timeout = ata_internal_cmd_timeout(dev, command);
auto_timeout = 1;
}
}
libata: implement cross-port EH exclusion In libata, the non-EH code paths should always take and release ap->lock explicitly when accessing hardware or shared data structures. However, once EH is active, it's assumed that the port is owned by EH and EH methods don't explicitly take ap->lock unless race from irq handler or other code paths are expected. However, libata EH didn't guarantee exclusion among EHs for ports of the same host. IOW, multiple EHs may execute in parallel on multiple ports of the same controller. In many cases, especially in SATA, the ports are completely independent of each other and this doesn't cause problems; however, there are cases where different ports share the same resource, which lead to obscure timing related bugs such as the one fixed by commit 213373cf (ata_piix: fix locking around SIDPR access). This patch implements exclusion among EHs of the same host. When EH begins, it acquires per-host EH ownership by calling ata_eh_acquire(). When EH finishes, the ownership is released by calling ata_eh_release(). EH ownership is also released whenever the EH thread goes to sleep from ata_msleep() or explicitly and reacquired after waking up. This ensures that while EH is actively accessing the hardware, it has exclusive access to it while allowing EHs to interleave and progress in parallel as they hit waiting stages, which dominate the time spent in EH. This achieves cross-port EH exclusion without pervasive and fragile changes while still allowing parallel EH for the most part. This was first reported by yuanding02@gmail.com more than three years ago in the following bugzilla. :-) https://bugzilla.kernel.org/show_bug.cgi?id=8223 Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Reported-by: yuanding02@gmail.com Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-09-06 23:57:14 +08:00
if (ap->ops->error_handler)
ata_eh_release(ap);
rc = wait_for_completion_timeout(&wait, msecs_to_jiffies(timeout));
libata: implement cross-port EH exclusion In libata, the non-EH code paths should always take and release ap->lock explicitly when accessing hardware or shared data structures. However, once EH is active, it's assumed that the port is owned by EH and EH methods don't explicitly take ap->lock unless race from irq handler or other code paths are expected. However, libata EH didn't guarantee exclusion among EHs for ports of the same host. IOW, multiple EHs may execute in parallel on multiple ports of the same controller. In many cases, especially in SATA, the ports are completely independent of each other and this doesn't cause problems; however, there are cases where different ports share the same resource, which lead to obscure timing related bugs such as the one fixed by commit 213373cf (ata_piix: fix locking around SIDPR access). This patch implements exclusion among EHs of the same host. When EH begins, it acquires per-host EH ownership by calling ata_eh_acquire(). When EH finishes, the ownership is released by calling ata_eh_release(). EH ownership is also released whenever the EH thread goes to sleep from ata_msleep() or explicitly and reacquired after waking up. This ensures that while EH is actively accessing the hardware, it has exclusive access to it while allowing EHs to interleave and progress in parallel as they hit waiting stages, which dominate the time spent in EH. This achieves cross-port EH exclusion without pervasive and fragile changes while still allowing parallel EH for the most part. This was first reported by yuanding02@gmail.com more than three years ago in the following bugzilla. :-) https://bugzilla.kernel.org/show_bug.cgi?id=8223 Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Reported-by: yuanding02@gmail.com Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-09-06 23:57:14 +08:00
if (ap->ops->error_handler)
ata_eh_acquire(ap);
ata_sff_flush_pio_task(ap);
if (!rc) {
spin_lock_irqsave(ap->lock, flags);
/* We're racing with irq here. If we lose, the
* following test prevents us from completing the qc
* twice. If we win, the port is frozen and will be
* cleaned up by ->post_internal_cmd().
*/
if (qc->flags & ATA_QCFLAG_ACTIVE) {
qc->err_mask |= AC_ERR_TIMEOUT;
if (ap->ops->error_handler)
ata_port_freeze(ap);
else
ata_qc_complete(qc);
if (ata_msg_warn(ap))
ata_dev_warn(dev, "qc timeout (cmd 0x%x)\n",
command);
}
spin_unlock_irqrestore(ap->lock, flags);
}
/* do post_internal_cmd */
if (ap->ops->post_internal_cmd)
ap->ops->post_internal_cmd(qc);
/* perform minimal error analysis */
if (qc->flags & ATA_QCFLAG_FAILED) {
if (qc->result_tf.command & (ATA_ERR | ATA_DF))
qc->err_mask |= AC_ERR_DEV;
if (!qc->err_mask)
qc->err_mask |= AC_ERR_OTHER;
if (qc->err_mask & ~AC_ERR_OTHER)
qc->err_mask &= ~AC_ERR_OTHER;
}
/* finish up */
spin_lock_irqsave(ap->lock, flags);
*tf = qc->result_tf;
err_mask = qc->err_mask;
ata_qc_free(qc);
link->active_tag = preempted_tag;
link->sactive = preempted_sactive;
ap->qc_active = preempted_qc_active;
ap->nr_active_links = preempted_nr_active_links;
spin_unlock_irqrestore(ap->lock, flags);
if ((err_mask & AC_ERR_TIMEOUT) && auto_timeout)
ata_internal_cmd_timed_out(dev, command);
return err_mask;
}
/**
* ata_exec_internal - execute libata internal command
* @dev: Device to which the command is sent
* @tf: Taskfile registers for the command and the result
* @cdb: CDB for packet command
* @dma_dir: Data transfer direction of the command
* @buf: Data buffer of the command
* @buflen: Length of data buffer
* @timeout: Timeout in msecs (0 for default)
*
* Wrapper around ata_exec_internal_sg() which takes simple
* buffer instead of sg list.
*
* LOCKING:
* None. Should be called with kernel context, might sleep.
*
* RETURNS:
* Zero on success, AC_ERR_* mask on failure
*/
unsigned ata_exec_internal(struct ata_device *dev,
struct ata_taskfile *tf, const u8 *cdb,
int dma_dir, void *buf, unsigned int buflen,
unsigned long timeout)
{
struct scatterlist *psg = NULL, sg;
unsigned int n_elem = 0;
if (dma_dir != DMA_NONE) {
WARN_ON(!buf);
sg_init_one(&sg, buf, buflen);
psg = &sg;
n_elem++;
}
return ata_exec_internal_sg(dev, tf, cdb, dma_dir, psg, n_elem,
timeout);
}
/**
* ata_pio_need_iordy - check if iordy needed
* @adev: ATA device
*
* Check if the current speed of the device requires IORDY. Used
* by various controllers for chip configuration.
*/
unsigned int ata_pio_need_iordy(const struct ata_device *adev)
{
/* Don't set IORDY if we're preparing for reset. IORDY may
* lead to controller lock up on certain controllers if the
* port is not occupied. See bko#11703 for details.
*/
if (adev->link->ap->pflags & ATA_PFLAG_RESETTING)
return 0;
/* Controller doesn't support IORDY. Probably a pointless
* check as the caller should know this.
*/
if (adev->link->ap->flags & ATA_FLAG_NO_IORDY)
return 0;
/* CF spec. r4.1 Table 22 says no iordy on PIO5 and PIO6. */
if (ata_id_is_cfa(adev->id)
&& (adev->pio_mode == XFER_PIO_5 || adev->pio_mode == XFER_PIO_6))
return 0;
/* PIO3 and higher it is mandatory */
if (adev->pio_mode > XFER_PIO_2)
return 1;
/* We turn it on when possible */
if (ata_id_has_iordy(adev->id))
return 1;
return 0;
}
/**
* ata_pio_mask_no_iordy - Return the non IORDY mask
* @adev: ATA device
*
* Compute the highest mode possible if we are not using iordy. Return
* -1 if no iordy mode is available.
*/
static u32 ata_pio_mask_no_iordy(const struct ata_device *adev)
{
/* If we have no drive specific rule, then PIO 2 is non IORDY */
if (adev->id[ATA_ID_FIELD_VALID] & 2) { /* EIDE */
u16 pio = adev->id[ATA_ID_EIDE_PIO];
/* Is the speed faster than the drive allows non IORDY ? */
if (pio) {
/* This is cycle times not frequency - watch the logic! */
if (pio > 240) /* PIO2 is 240nS per cycle */
return 3 << ATA_SHIFT_PIO;
return 7 << ATA_SHIFT_PIO;
}
}
return 3 << ATA_SHIFT_PIO;
}
/**
* ata_do_dev_read_id - default ID read method
* @dev: device
* @tf: proposed taskfile
* @id: data buffer
*
* Issue the identify taskfile and hand back the buffer containing
* identify data. For some RAID controllers and for pre ATA devices
* this function is wrapped or replaced by the driver
*/
unsigned int ata_do_dev_read_id(struct ata_device *dev,
struct ata_taskfile *tf, u16 *id)
{
return ata_exec_internal(dev, tf, NULL, DMA_FROM_DEVICE,
id, sizeof(id[0]) * ATA_ID_WORDS, 0);
}
/**
* ata_dev_read_id - Read ID data from the specified device
* @dev: target device
* @p_class: pointer to class of the target device (may be changed)
* @flags: ATA_READID_* flags
* @id: buffer to read IDENTIFY data into
*
* Read ID data from the specified device. ATA_CMD_ID_ATA is
* performed on ATA devices and ATA_CMD_ID_ATAPI on ATAPI
* devices. This function also issues ATA_CMD_INIT_DEV_PARAMS
* for pre-ATA4 drives.
*
* FIXME: ATA_CMD_ID_ATA is optional for early drives and right
* now we abort if we hit that case.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, -errno otherwise.
*/
int ata_dev_read_id(struct ata_device *dev, unsigned int *p_class,
unsigned int flags, u16 *id)
{
struct ata_port *ap = dev->link->ap;
unsigned int class = *p_class;
struct ata_taskfile tf;
unsigned int err_mask = 0;
const char *reason;
bool is_semb = class == ATA_DEV_SEMB;
int may_fallback = 1, tried_spinup = 0;
int rc;
if (ata_msg_ctl(ap))
ata_dev_dbg(dev, "%s: ENTER\n", __func__);
retry:
ata_tf_init(dev, &tf);
switch (class) {
case ATA_DEV_SEMB:
class = ATA_DEV_ATA; /* some hard drives report SEMB sig */
case ATA_DEV_ATA:
case ATA_DEV_ZAC:
tf.command = ATA_CMD_ID_ATA;
break;
case ATA_DEV_ATAPI:
tf.command = ATA_CMD_ID_ATAPI;
break;
default:
rc = -ENODEV;
reason = "unsupported class";
goto err_out;
}
tf.protocol = ATA_PROT_PIO;
/* Some devices choke if TF registers contain garbage. Make
* sure those are properly initialized.
*/
tf.flags |= ATA_TFLAG_ISADDR | ATA_TFLAG_DEVICE;
/* Device presence detection is unreliable on some
* controllers. Always poll IDENTIFY if available.
*/
tf.flags |= ATA_TFLAG_POLLING;
if (ap->ops->read_id)
err_mask = ap->ops->read_id(dev, &tf, id);
else
err_mask = ata_do_dev_read_id(dev, &tf, id);
if (err_mask) {
if (err_mask & AC_ERR_NODEV_HINT) {
ata_dev_dbg(dev, "NODEV after polling detection\n");
return -ENOENT;
}
if (is_semb) {
ata_dev_info(dev,
"IDENTIFY failed on device w/ SEMB sig, disabled\n");
/* SEMB is not supported yet */
*p_class = ATA_DEV_SEMB_UNSUP;
return 0;
}
if ((err_mask == AC_ERR_DEV) && (tf.feature & ATA_ABORTED)) {
/* Device or controller might have reported
* the wrong device class. Give a shot at the
* other IDENTIFY if the current one is
* aborted by the device.
*/
if (may_fallback) {
may_fallback = 0;
if (class == ATA_DEV_ATA)
class = ATA_DEV_ATAPI;
else
class = ATA_DEV_ATA;
goto retry;
}
/* Control reaches here iff the device aborted
* both flavors of IDENTIFYs which happens
* sometimes with phantom devices.
*/
ata_dev_dbg(dev,
"both IDENTIFYs aborted, assuming NODEV\n");
return -ENOENT;
}
rc = -EIO;
reason = "I/O error";
goto err_out;
}
if (dev->horkage & ATA_HORKAGE_DUMP_ID) {
ata_dev_dbg(dev, "dumping IDENTIFY data, "
"class=%d may_fallback=%d tried_spinup=%d\n",
class, may_fallback, tried_spinup);
print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET,
16, 2, id, ATA_ID_WORDS * sizeof(*id), true);
}
/* Falling back doesn't make sense if ID data was read
* successfully at least once.
*/
may_fallback = 0;
swap_buf_le16(id, ATA_ID_WORDS);
/* sanity check */
rc = -EINVAL;
reason = "device reports invalid type";
if (class == ATA_DEV_ATA || class == ATA_DEV_ZAC) {
if (!ata_id_is_ata(id) && !ata_id_is_cfa(id))
goto err_out;
if (ap->host->flags & ATA_HOST_IGNORE_ATA &&
ata_id_is_ata(id)) {
ata_dev_dbg(dev,
"host indicates ignore ATA devices, ignored\n");
return -ENOENT;
}
} else {
if (ata_id_is_ata(id))
goto err_out;
}
if (!tried_spinup && (id[2] == 0x37c8 || id[2] == 0x738c)) {
tried_spinup = 1;
/*
* Drive powered-up in standby mode, and requires a specific
* SET_FEATURES spin-up subcommand before it will accept
* anything other than the original IDENTIFY command.
*/
err_mask = ata_dev_set_feature(dev, SETFEATURES_SPINUP, 0);
if (err_mask && id[2] != 0x738c) {
rc = -EIO;
reason = "SPINUP failed";
goto err_out;
}
/*
* If the drive initially returned incomplete IDENTIFY info,
* we now must reissue the IDENTIFY command.
*/
if (id[2] == 0x37c8)
goto retry;
}
if ((flags & ATA_READID_POSTRESET) &&
(class == ATA_DEV_ATA || class == ATA_DEV_ZAC)) {
/*
* The exact sequence expected by certain pre-ATA4 drives is:
* SRST RESET
* IDENTIFY (optional in early ATA)
* INITIALIZE DEVICE PARAMETERS (later IDE and ATA)
* anything else..
* Some drives were very specific about that exact sequence.
*
* Note that ATA4 says lba is mandatory so the second check
* should never trigger.
*/
if (ata_id_major_version(id) < 4 || !ata_id_has_lba(id)) {
err_mask = ata_dev_init_params(dev, id[3], id[6]);
if (err_mask) {
rc = -EIO;
reason = "INIT_DEV_PARAMS failed";
goto err_out;
}
/* current CHS translation info (id[53-58]) might be
* changed. reread the identify device info.
*/
flags &= ~ATA_READID_POSTRESET;
goto retry;
}
}
*p_class = class;
return 0;
err_out:
if (ata_msg_warn(ap))
ata_dev_warn(dev, "failed to IDENTIFY (%s, err_mask=0x%x)\n",
reason, err_mask);
return rc;
}
static int ata_do_link_spd_horkage(struct ata_device *dev)
{
struct ata_link *plink = ata_dev_phys_link(dev);
u32 target, target_limit;
if (!sata_scr_valid(plink))
return 0;
if (dev->horkage & ATA_HORKAGE_1_5_GBPS)
target = 1;
else
return 0;
target_limit = (1 << target) - 1;
/* if already on stricter limit, no need to push further */
if (plink->sata_spd_limit <= target_limit)
return 0;
plink->sata_spd_limit = target_limit;
/* Request another EH round by returning -EAGAIN if link is
* going faster than the target speed. Forward progress is
* guaranteed by setting sata_spd_limit to target_limit above.
*/
if (plink->sata_spd > target) {
ata_dev_info(dev, "applying link speed limit horkage to %s\n",
sata_spd_string(target));
return -EAGAIN;
}
return 0;
}
static inline u8 ata_dev_knobble(struct ata_device *dev)
{
struct ata_port *ap = dev->link->ap;
if (ata_dev_blacklisted(dev) & ATA_HORKAGE_BRIDGE_OK)
return 0;
return ((ap->cbl == ATA_CBL_SATA) && (!ata_id_is_sata(dev->id)));
}
static int ata_dev_config_ncq(struct ata_device *dev,
char *desc, size_t desc_sz)
{
struct ata_port *ap = dev->link->ap;
int hdepth = 0, ddepth = ata_id_queue_depth(dev->id);
unsigned int err_mask;
char *aa_desc = "";
if (!ata_id_has_ncq(dev->id)) {
desc[0] = '\0';
return 0;
}
if (dev->horkage & ATA_HORKAGE_NONCQ) {
snprintf(desc, desc_sz, "NCQ (not used)");
return 0;
}
if (ap->flags & ATA_FLAG_NCQ) {
hdepth = min(ap->scsi_host->can_queue, ATA_MAX_QUEUE - 1);
dev->flags |= ATA_DFLAG_NCQ;
}
if (!(dev->horkage & ATA_HORKAGE_BROKEN_FPDMA_AA) &&
(ap->flags & ATA_FLAG_FPDMA_AA) &&
ata_id_has_fpdma_aa(dev->id)) {
err_mask = ata_dev_set_feature(dev, SETFEATURES_SATA_ENABLE,
SATA_FPDMA_AA);
if (err_mask) {
ata_dev_err(dev,
"failed to enable AA (error_mask=0x%x)\n",
err_mask);
if (err_mask != AC_ERR_DEV) {
dev->horkage |= ATA_HORKAGE_BROKEN_FPDMA_AA;
return -EIO;
}
} else
aa_desc = ", AA";
}
if (hdepth >= ddepth)
snprintf(desc, desc_sz, "NCQ (depth %d)%s", ddepth, aa_desc);
else
snprintf(desc, desc_sz, "NCQ (depth %d/%d)%s", hdepth,
ddepth, aa_desc);
if ((ap->flags & ATA_FLAG_FPDMA_AUX) &&
ata_id_has_ncq_send_and_recv(dev->id)) {
err_mask = ata_read_log_page(dev, ATA_LOG_NCQ_SEND_RECV,
0, ap->sector_buf, 1);
if (err_mask) {
ata_dev_dbg(dev,
"failed to get NCQ Send/Recv Log Emask 0x%x\n",
err_mask);
} else {
u8 *cmds = dev->ncq_send_recv_cmds;
dev->flags |= ATA_DFLAG_NCQ_SEND_RECV;
memcpy(cmds, ap->sector_buf, ATA_LOG_NCQ_SEND_RECV_SIZE);
if (dev->horkage & ATA_HORKAGE_NO_NCQ_TRIM) {
ata_dev_dbg(dev, "disabling queued TRIM support\n");
cmds[ATA_LOG_NCQ_SEND_RECV_DSM_OFFSET] &=
~ATA_LOG_NCQ_SEND_RECV_DSM_TRIM;
}
}
}
return 0;
}
/**
* ata_dev_configure - Configure the specified ATA/ATAPI device
* @dev: Target device to configure
*
* Configure @dev according to @dev->id. Generic and low-level
* driver specific fixups are also applied.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, -errno otherwise
*/
int ata_dev_configure(struct ata_device *dev)
{
struct ata_port *ap = dev->link->ap;
struct ata_eh_context *ehc = &dev->link->eh_context;
int print_info = ehc->i.flags & ATA_EHI_PRINTINFO;
const u16 *id = dev->id;
unsigned long xfer_mask;
unsigned int err_mask;
char revbuf[7]; /* XYZ-99\0 */
char fwrevbuf[ATA_ID_FW_REV_LEN+1];
char modelbuf[ATA_ID_PROD_LEN+1];
int rc;
if (!ata_dev_enabled(dev) && ata_msg_info(ap)) {
ata_dev_info(dev, "%s: ENTER/EXIT -- nodev\n", __func__);
return 0;
}
if (ata_msg_probe(ap))
ata_dev_dbg(dev, "%s: ENTER\n", __func__);
/* set horkage */
dev->horkage |= ata_dev_blacklisted(dev);
ata_force_horkage(dev);
if (dev->horkage & ATA_HORKAGE_DISABLE) {
ata_dev_info(dev, "unsupported device, disabling\n");
ata_dev_disable(dev);
return 0;
}
if ((!atapi_enabled || (ap->flags & ATA_FLAG_NO_ATAPI)) &&
dev->class == ATA_DEV_ATAPI) {
ata_dev_warn(dev, "WARNING: ATAPI is %s, device ignored\n",
atapi_enabled ? "not supported with this driver"
: "disabled");
ata_dev_disable(dev);
return 0;
}
rc = ata_do_link_spd_horkage(dev);
if (rc)
return rc;
/* some WD SATA-1 drives have issues with LPM, turn on NOLPM for them */
if ((dev->horkage & ATA_HORKAGE_WD_BROKEN_LPM) &&
(id[ATA_ID_SATA_CAPABILITY] & 0xe) == 0x2)
dev->horkage |= ATA_HORKAGE_NOLPM;
if (dev->horkage & ATA_HORKAGE_NOLPM) {
ata_dev_warn(dev, "LPM support broken, forcing max_power\n");
dev->link->ap->target_lpm_policy = ATA_LPM_MAX_POWER;
}
/* let ACPI work its magic */
rc = ata_acpi_on_devcfg(dev);
if (rc)
return rc;
/* massage HPA, do it early as it might change IDENTIFY data */
rc = ata_hpa_resize(dev);
if (rc)
return rc;
/* print device capabilities */
if (ata_msg_probe(ap))
ata_dev_dbg(dev,
"%s: cfg 49:%04x 82:%04x 83:%04x 84:%04x "
"85:%04x 86:%04x 87:%04x 88:%04x\n",
__func__,
id[49], id[82], id[83], id[84],
id[85], id[86], id[87], id[88]);
/* initialize to-be-configured parameters */
dev->flags &= ~ATA_DFLAG_CFG_MASK;
dev->max_sectors = 0;
dev->cdb_len = 0;
dev->n_sectors = 0;
dev->cylinders = 0;
dev->heads = 0;
dev->sectors = 0;
dev->multi_count = 0;
/*
* common ATA, ATAPI feature tests
*/
/* find max transfer mode; for printk only */
xfer_mask = ata_id_xfermask(id);
if (ata_msg_probe(ap))
ata_dump_id(id);
/* SCSI only uses 4-char revisions, dump full 8 chars from ATA */
ata_id_c_string(dev->id, fwrevbuf, ATA_ID_FW_REV,
sizeof(fwrevbuf));
ata_id_c_string(dev->id, modelbuf, ATA_ID_PROD,
sizeof(modelbuf));
/* ATA-specific feature tests */
if (dev->class == ATA_DEV_ATA || dev->class == ATA_DEV_ZAC) {
if (ata_id_is_cfa(id)) {
/* CPRM may make this media unusable */
if (id[ATA_ID_CFA_KEY_MGMT] & 1)
ata_dev_warn(dev,
"supports DRM functions and may not be fully accessible\n");
snprintf(revbuf, 7, "CFA");
} else {
snprintf(revbuf, 7, "ATA-%d", ata_id_major_version(id));
/* Warn the user if the device has TPM extensions */
if (ata_id_has_tpm(id))
ata_dev_warn(dev,
"supports DRM functions and may not be fully accessible\n");
}
dev->n_sectors = ata_id_n_sectors(id);
/* get current R/W Multiple count setting */
if ((dev->id[47] >> 8) == 0x80 && (dev->id[59] & 0x100)) {
unsigned int max = dev->id[47] & 0xff;
unsigned int cnt = dev->id[59] & 0xff;
/* only recognize/allow powers of two here */
if (is_power_of_2(max) && is_power_of_2(cnt))
if (cnt <= max)
dev->multi_count = cnt;
}
if (ata_id_has_lba(id)) {
const char *lba_desc;
char ncq_desc[24];
lba_desc = "LBA";
dev->flags |= ATA_DFLAG_LBA;
if (ata_id_has_lba48(id)) {
dev->flags |= ATA_DFLAG_LBA48;
lba_desc = "LBA48";
if (dev->n_sectors >= (1UL << 28) &&
ata_id_has_flush_ext(id))
dev->flags |= ATA_DFLAG_FLUSH_EXT;
}
/* config NCQ */
rc = ata_dev_config_ncq(dev, ncq_desc, sizeof(ncq_desc));
if (rc)
return rc;
/* print device info to dmesg */
if (ata_msg_drv(ap) && print_info) {
ata_dev_info(dev, "%s: %s, %s, max %s\n",
revbuf, modelbuf, fwrevbuf,
ata_mode_string(xfer_mask));
ata_dev_info(dev,
"%llu sectors, multi %u: %s %s\n",
(unsigned long long)dev->n_sectors,
dev->multi_count, lba_desc, ncq_desc);
}
} else {
/* CHS */
/* Default translation */
dev->cylinders = id[1];
dev->heads = id[3];
dev->sectors = id[6];
if (ata_id_current_chs_valid(id)) {
/* Current CHS translation is valid. */
dev->cylinders = id[54];
dev->heads = id[55];
dev->sectors = id[56];
}
/* print device info to dmesg */
if (ata_msg_drv(ap) && print_info) {
ata_dev_info(dev, "%s: %s, %s, max %s\n",
revbuf, modelbuf, fwrevbuf,
ata_mode_string(xfer_mask));
ata_dev_info(dev,
"%llu sectors, multi %u, CHS %u/%u/%u\n",
(unsigned long long)dev->n_sectors,
dev->multi_count, dev->cylinders,
dev->heads, dev->sectors);
}
}
/* Check and mark DevSlp capability. Get DevSlp timing variables
* from SATA Settings page of Identify Device Data Log.
*/
if (ata_id_has_devslp(dev->id)) {
u8 *sata_setting = ap->sector_buf;
int i, j;
dev->flags |= ATA_DFLAG_DEVSLP;
err_mask = ata_read_log_page(dev,
ATA_LOG_SATA_ID_DEV_DATA,
ATA_LOG_SATA_SETTINGS,
sata_setting,
1);
if (err_mask)
ata_dev_dbg(dev,
"failed to get Identify Device Data, Emask 0x%x\n",
err_mask);
else
for (i = 0; i < ATA_LOG_DEVSLP_SIZE; i++) {
j = ATA_LOG_DEVSLP_OFFSET + i;
dev->devslp_timing[i] = sata_setting[j];
}
}
dev->cdb_len = 16;
}
/* ATAPI-specific feature tests */
else if (dev->class == ATA_DEV_ATAPI) {
const char *cdb_intr_string = "";
const char *atapi_an_string = "";
const char *dma_dir_string = "";
u32 sntf;
rc = atapi_cdb_len(id);
if ((rc < 12) || (rc > ATAPI_CDB_LEN)) {
if (ata_msg_warn(ap))
ata_dev_warn(dev, "unsupported CDB len\n");
rc = -EINVAL;
goto err_out_nosup;
}
dev->cdb_len = (unsigned int) rc;
/* Enable ATAPI AN if both the host and device have
* the support. If PMP is attached, SNTF is required
* to enable ATAPI AN to discern between PHY status
* changed notifications and ATAPI ANs.
*/
if (atapi_an &&
(ap->flags & ATA_FLAG_AN) && ata_id_has_atapi_AN(id) &&
(!sata_pmp_attached(ap) ||
sata_scr_read(&ap->link, SCR_NOTIFICATION, &sntf) == 0)) {
/* issue SET feature command to turn this on */
err_mask = ata_dev_set_feature(dev,
SETFEATURES_SATA_ENABLE, SATA_AN);
if (err_mask)
ata_dev_err(dev,
"failed to enable ATAPI AN (err_mask=0x%x)\n",
err_mask);
else {
dev->flags |= ATA_DFLAG_AN;
atapi_an_string = ", ATAPI AN";
}
}
if (ata_id_cdb_intr(dev->id)) {
dev->flags |= ATA_DFLAG_CDB_INTR;
cdb_intr_string = ", CDB intr";
}
if (atapi_dmadir || (dev->horkage & ATA_HORKAGE_ATAPI_DMADIR) || atapi_id_dmadir(dev->id)) {
dev->flags |= ATA_DFLAG_DMADIR;
dma_dir_string = ", DMADIR";
}
if (ata_id_has_da(dev->id)) {
dev->flags |= ATA_DFLAG_DA;
zpodd_init(dev);
}
/* print device info to dmesg */
if (ata_msg_drv(ap) && print_info)
ata_dev_info(dev,
"ATAPI: %s, %s, max %s%s%s%s\n",
modelbuf, fwrevbuf,
ata_mode_string(xfer_mask),
cdb_intr_string, atapi_an_string,
dma_dir_string);
}
/* determine max_sectors */
dev->max_sectors = ATA_MAX_SECTORS;
if (dev->flags & ATA_DFLAG_LBA48)
dev->max_sectors = ATA_MAX_SECTORS_LBA48;
/* Limit PATA drive on SATA cable bridge transfers to udma5,
200 sectors */
if (ata_dev_knobble(dev)) {
if (ata_msg_drv(ap) && print_info)
ata_dev_info(dev, "applying bridge limits\n");
dev->udma_mask &= ATA_UDMA5;
dev->max_sectors = ATA_MAX_SECTORS;
}
if ((dev->class == ATA_DEV_ATAPI) &&
(atapi_command_packet_set(id) == TYPE_TAPE)) {
dev->max_sectors = ATA_MAX_SECTORS_TAPE;
dev->horkage |= ATA_HORKAGE_STUCK_ERR;
}
if (dev->horkage & ATA_HORKAGE_MAX_SEC_128)
dev->max_sectors = min_t(unsigned int, ATA_MAX_SECTORS_128,
dev->max_sectors);
libata: Set max sector to 65535 for Slimtype DVD A DS8A8SH drive The Slimtype DVD A DS8A8SH drive locks up when max sector is smaller than 65535, and the blow backtrace is observed on locking up: INFO: task flush-8:32:1130 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. flush-8:32 D ffffffff8180cf60 0 1130 2 0x00000000 ffff880273aef618 0000000000000046 0000000000000005 ffff880273aee000 ffff880273aee000 ffff880273aeffd8 ffff880273aee010 ffff880273aee000 ffff880273aeffd8 ffff880273aee000 ffff88026e842ea0 ffff880274a10000 Call Trace: [<ffffffff8168fc2d>] schedule+0x5d/0x70 [<ffffffff8168fccc>] io_schedule+0x8c/0xd0 [<ffffffff81324461>] get_request+0x731/0x7d0 [<ffffffff8133dc60>] ? cfq_allow_merge+0x50/0x90 [<ffffffff81083aa0>] ? wake_up_bit+0x40/0x40 [<ffffffff81320443>] ? bio_attempt_back_merge+0x33/0x110 [<ffffffff813248ea>] blk_queue_bio+0x23a/0x3f0 [<ffffffff81322176>] generic_make_request+0xc6/0x120 [<ffffffff81322308>] submit_bio+0x138/0x160 [<ffffffff811d7596>] ? bio_alloc_bioset+0x96/0x120 [<ffffffff811d1f61>] submit_bh+0x1f1/0x220 [<ffffffff811d48b8>] __block_write_full_page+0x228/0x340 [<ffffffff811d3650>] ? attach_nobh_buffers+0xc0/0xc0 [<ffffffff811d8960>] ? I_BDEV+0x10/0x10 [<ffffffff811d8960>] ? I_BDEV+0x10/0x10 [<ffffffff811d4ab6>] block_write_full_page_endio+0xe6/0x100 [<ffffffff811d4ae5>] block_write_full_page+0x15/0x20 [<ffffffff811d9268>] blkdev_writepage+0x18/0x20 [<ffffffff81142527>] __writepage+0x17/0x40 [<ffffffff811438ba>] write_cache_pages+0x34a/0x4a0 [<ffffffff81142510>] ? set_page_dirty+0x70/0x70 [<ffffffff81143a61>] generic_writepages+0x51/0x80 [<ffffffff81143ab0>] do_writepages+0x20/0x50 [<ffffffff811c9ed6>] __writeback_single_inode+0xa6/0x2b0 [<ffffffff811ca861>] writeback_sb_inodes+0x311/0x4d0 [<ffffffff811caaa6>] __writeback_inodes_wb+0x86/0xd0 [<ffffffff811cad43>] wb_writeback+0x1a3/0x330 [<ffffffff816916cf>] ? _raw_spin_lock_irqsave+0x3f/0x50 [<ffffffff811b8362>] ? get_nr_inodes+0x52/0x70 [<ffffffff811cb0ac>] wb_do_writeback+0x1dc/0x260 [<ffffffff8168dd34>] ? schedule_timeout+0x204/0x240 [<ffffffff811cb232>] bdi_writeback_thread+0x102/0x2b0 [<ffffffff811cb130>] ? wb_do_writeback+0x260/0x260 [<ffffffff81083550>] kthread+0xc0/0xd0 [<ffffffff81083490>] ? kthread_worker_fn+0x1b0/0x1b0 [<ffffffff8169a3ec>] ret_from_fork+0x7c/0xb0 [<ffffffff81083490>] ? kthread_worker_fn+0x1b0/0x1b0 The above trace was triggered by "dd if=/dev/zero of=/dev/sr0 bs=2048 count=32768" It was previously working by accident, since another bug introduced by 4dce8ba94c7 (libata: Use 'bool' return value for ata_id_XXX) caused all drives to use maxsect=65535. Cc: stable@vger.kernel.org Signed-off-by: Shan Hai <shan.hai@windriver.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2013-03-18 10:30:44 +08:00
if (dev->horkage & ATA_HORKAGE_MAX_SEC_LBA48)
dev->max_sectors = ATA_MAX_SECTORS_LBA48;
if (ap->ops->dev_config)
ap->ops->dev_config(dev);
if (dev->horkage & ATA_HORKAGE_DIAGNOSTIC) {
/* Let the user know. We don't want to disallow opens for
rescue purposes, or in case the vendor is just a blithering
idiot. Do this after the dev_config call as some controllers
with buggy firmware may want to avoid reporting false device
bugs */
if (print_info) {
ata_dev_warn(dev,
"Drive reports diagnostics failure. This may indicate a drive\n");
ata_dev_warn(dev,
"fault or invalid emulation. Contact drive vendor for information.\n");
}
}
if ((dev->horkage & ATA_HORKAGE_FIRMWARE_WARN) && print_info) {
ata_dev_warn(dev, "WARNING: device requires firmware update to be fully functional\n");
ata_dev_warn(dev, " contact the vendor or visit http://ata.wiki.kernel.org\n");
}
return 0;
err_out_nosup:
if (ata_msg_probe(ap))
ata_dev_dbg(dev, "%s: EXIT, err\n", __func__);
return rc;
}
/**
* ata_cable_40wire - return 40 wire cable type
* @ap: port
*
* Helper method for drivers which want to hardwire 40 wire cable
* detection.
*/
int ata_cable_40wire(struct ata_port *ap)
{
return ATA_CBL_PATA40;
}
/**
* ata_cable_80wire - return 80 wire cable type
* @ap: port
*
* Helper method for drivers which want to hardwire 80 wire cable
* detection.
*/
int ata_cable_80wire(struct ata_port *ap)
{
return ATA_CBL_PATA80;
}
/**
* ata_cable_unknown - return unknown PATA cable.
* @ap: port
*
* Helper method for drivers which have no PATA cable detection.
*/
int ata_cable_unknown(struct ata_port *ap)
{
return ATA_CBL_PATA_UNK;
}
/**
* ata_cable_ignore - return ignored PATA cable.
* @ap: port
*
* Helper method for drivers which don't use cable type to limit
* transfer mode.
*/
int ata_cable_ignore(struct ata_port *ap)
{
return ATA_CBL_PATA_IGN;
}
/**
* ata_cable_sata - return SATA cable type
* @ap: port
*
* Helper method for drivers which have SATA cables
*/
int ata_cable_sata(struct ata_port *ap)
{
return ATA_CBL_SATA;
}
/**
* ata_bus_probe - Reset and probe ATA bus
* @ap: Bus to probe
*
2005-05-31 07:49:12 +08:00
* Master ATA bus probing function. Initiates a hardware-dependent
* bus reset, then attempts to identify any devices found on
* the bus.
*
* LOCKING:
2005-05-31 07:49:12 +08:00
* PCI/etc. bus probe sem.
*
* RETURNS:
* Zero on success, negative errno otherwise.
*/
int ata_bus_probe(struct ata_port *ap)
{
unsigned int classes[ATA_MAX_DEVICES];
int tries[ATA_MAX_DEVICES];
int rc;
struct ata_device *dev;
ata_for_each_dev(dev, &ap->link, ALL)
tries[dev->devno] = ATA_PROBE_MAX_TRIES;
retry:
ata_for_each_dev(dev, &ap->link, ALL) {
/* If we issue an SRST then an ATA drive (not ATAPI)
* may change configuration and be in PIO0 timing. If
* we do a hard reset (or are coming from power on)
* this is true for ATA or ATAPI. Until we've set a
* suitable controller mode we should not touch the
* bus as we may be talking too fast.
*/
dev->pio_mode = XFER_PIO_0;
dev->dma_mode = 0xff;
/* If the controller has a pio mode setup function
* then use it to set the chipset to rights. Don't
* touch the DMA setup as that will be dealt with when
* configuring devices.
*/
if (ap->ops->set_piomode)
ap->ops->set_piomode(ap, dev);
}
/* reset and determine device classes */
ap->ops->phy_reset(ap);
ata_for_each_dev(dev, &ap->link, ALL) {
if (dev->class != ATA_DEV_UNKNOWN)
classes[dev->devno] = dev->class;
else
classes[dev->devno] = ATA_DEV_NONE;
dev->class = ATA_DEV_UNKNOWN;
}
/* read IDENTIFY page and configure devices. We have to do the identify
specific sequence bass-ackwards so that PDIAG- is released by
the slave device */
ata_for_each_dev(dev, &ap->link, ALL_REVERSE) {
if (tries[dev->devno])
dev->class = classes[dev->devno];
if (!ata_dev_enabled(dev))
continue;
rc = ata_dev_read_id(dev, &dev->class, ATA_READID_POSTRESET,
dev->id);
if (rc)
goto fail;
}
/* Now ask for the cable type as PDIAG- should have been released */
if (ap->ops->cable_detect)
ap->cbl = ap->ops->cable_detect(ap);
/* We may have SATA bridge glue hiding here irrespective of
* the reported cable types and sensed types. When SATA
* drives indicate we have a bridge, we don't know which end
* of the link the bridge is which is a problem.
*/
ata_for_each_dev(dev, &ap->link, ENABLED)
if (ata_id_is_sata(dev->id))
ap->cbl = ATA_CBL_SATA;
/* After the identify sequence we can now set up the devices. We do
this in the normal order so that the user doesn't get confused */
ata_for_each_dev(dev, &ap->link, ENABLED) {
ap->link.eh_context.i.flags |= ATA_EHI_PRINTINFO;
rc = ata_dev_configure(dev);
ap->link.eh_context.i.flags &= ~ATA_EHI_PRINTINFO;
if (rc)
goto fail;
}
/* configure transfer mode */
rc = ata_set_mode(&ap->link, &dev);
if (rc)
goto fail;
ata_for_each_dev(dev, &ap->link, ENABLED)
return 0;
return -ENODEV;
fail:
tries[dev->devno]--;
switch (rc) {
case -EINVAL:
/* eeek, something went very wrong, give up */
tries[dev->devno] = 0;
break;
case -ENODEV:
/* give it just one more chance */
tries[dev->devno] = min(tries[dev->devno], 1);
case -EIO:
if (tries[dev->devno] == 1) {
/* This is the last chance, better to slow
* down than lose it.
*/
sata_down_spd_limit(&ap->link, 0);
ata_down_xfermask_limit(dev, ATA_DNXFER_PIO);
}
}
if (!tries[dev->devno])
ata_dev_disable(dev);
goto retry;
}
/**
* sata_print_link_status - Print SATA link status
* @link: SATA link to printk link status about
*
* This function prints link speed and status of a SATA link.
*
* LOCKING:
* None.
*/
static void sata_print_link_status(struct ata_link *link)
{
u32 sstatus, scontrol, tmp;
if (sata_scr_read(link, SCR_STATUS, &sstatus))
return;
sata_scr_read(link, SCR_CONTROL, &scontrol);
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
if (ata_phys_link_online(link)) {
tmp = (sstatus >> 4) & 0xf;
ata_link_info(link, "SATA link up %s (SStatus %X SControl %X)\n",
sata_spd_string(tmp), sstatus, scontrol);
} else {
ata_link_info(link, "SATA link down (SStatus %X SControl %X)\n",
sstatus, scontrol);
}
}
/**
* ata_dev_pair - return other device on cable
* @adev: device
*
* Obtain the other device on the same cable, or if none is
* present NULL is returned
*/
struct ata_device *ata_dev_pair(struct ata_device *adev)
{
struct ata_link *link = adev->link;
struct ata_device *pair = &link->device[1 - adev->devno];
if (!ata_dev_enabled(pair))
return NULL;
return pair;
}
/**
* sata_down_spd_limit - adjust SATA spd limit downward
* @link: Link to adjust SATA spd limit for
* @spd_limit: Additional limit
*
* Adjust SATA spd limit of @link downward. Note that this
* function only adjusts the limit. The change must be applied
* using sata_set_spd().
*
* If @spd_limit is non-zero, the speed is limited to equal to or
* lower than @spd_limit if such speed is supported. If
* @spd_limit is slower than any supported speed, only the lowest
* supported speed is allowed.
*
* LOCKING:
* Inherited from caller.
*
* RETURNS:
* 0 on success, negative errno on failure
*/
int sata_down_spd_limit(struct ata_link *link, u32 spd_limit)
{
u32 sstatus, spd, mask;
int rc, bit;
if (!sata_scr_valid(link))
return -EOPNOTSUPP;
/* If SCR can be read, use it to determine the current SPD.
* If not, use cached value in link->sata_spd.
*/
rc = sata_scr_read(link, SCR_STATUS, &sstatus);
if (rc == 0 && ata_sstatus_online(sstatus))
spd = (sstatus >> 4) & 0xf;
else
spd = link->sata_spd;
mask = link->sata_spd_limit;
if (mask <= 1)
return -EINVAL;
/* unconditionally mask off the highest bit */
bit = fls(mask) - 1;
mask &= ~(1 << bit);
/* Mask off all speeds higher than or equal to the current
* one. Force 1.5Gbps if current SPD is not available.
*/
if (spd > 1)
mask &= (1 << (spd - 1)) - 1;
else
mask &= 1;
/* were we already at the bottom? */
if (!mask)
return -EINVAL;
if (spd_limit) {
if (mask & ((1 << spd_limit) - 1))
mask &= (1 << spd_limit) - 1;
else {
bit = ffs(mask) - 1;
mask = 1 << bit;
}
}
link->sata_spd_limit = mask;
ata_link_warn(link, "limiting SATA link speed to %s\n",
sata_spd_string(fls(mask)));
return 0;
}
static int __sata_set_spd_needed(struct ata_link *link, u32 *scontrol)
{
struct ata_link *host_link = &link->ap->link;
u32 limit, target, spd;
limit = link->sata_spd_limit;
/* Don't configure downstream link faster than upstream link.
* It doesn't speed up anything and some PMPs choke on such
* configuration.
*/
if (!ata_is_host_link(link) && host_link->sata_spd)
limit &= (1 << host_link->sata_spd) - 1;
if (limit == UINT_MAX)
target = 0;
else
target = fls(limit);
spd = (*scontrol >> 4) & 0xf;
*scontrol = (*scontrol & ~0xf0) | ((target & 0xf) << 4);
return spd != target;
}
/**
* sata_set_spd_needed - is SATA spd configuration needed
* @link: Link in question
*
* Test whether the spd limit in SControl matches
* @link->sata_spd_limit. This function is used to determine
* whether hardreset is necessary to apply SATA spd
* configuration.
*
* LOCKING:
* Inherited from caller.
*
* RETURNS:
* 1 if SATA spd configuration is needed, 0 otherwise.
*/
static int sata_set_spd_needed(struct ata_link *link)
{
u32 scontrol;
if (sata_scr_read(link, SCR_CONTROL, &scontrol))
return 1;
return __sata_set_spd_needed(link, &scontrol);
}
/**
* sata_set_spd - set SATA spd according to spd limit
* @link: Link to set SATA spd for
*
* Set SATA spd of @link according to sata_spd_limit.
*
* LOCKING:
* Inherited from caller.
*
* RETURNS:
* 0 if spd doesn't need to be changed, 1 if spd has been
* changed. Negative errno if SCR registers are inaccessible.
*/
int sata_set_spd(struct ata_link *link)
{
u32 scontrol;
int rc;
if ((rc = sata_scr_read(link, SCR_CONTROL, &scontrol)))
return rc;
if (!__sata_set_spd_needed(link, &scontrol))
return 0;
if ((rc = sata_scr_write(link, SCR_CONTROL, scontrol)))
return rc;
return 1;
}
/*
* This mode timing computation functionality is ported over from
* drivers/ide/ide-timing.h and was originally written by Vojtech Pavlik
*/
/*
* PIO 0-4, MWDMA 0-2 and UDMA 0-6 timings (in nanoseconds).
* These were taken from ATA/ATAPI-6 standard, rev 0a, except
* for UDMA6, which is currently supported only by Maxtor drives.
*
* For PIO 5/6 MWDMA 3/4 see the CFA specification 3.0.
*/
static const struct ata_timing ata_timing[] = {
/* { XFER_PIO_SLOW, 120, 290, 240, 960, 290, 240, 0, 960, 0 }, */
{ XFER_PIO_0, 70, 290, 240, 600, 165, 150, 0, 600, 0 },
{ XFER_PIO_1, 50, 290, 93, 383, 125, 100, 0, 383, 0 },
{ XFER_PIO_2, 30, 290, 40, 330, 100, 90, 0, 240, 0 },
{ XFER_PIO_3, 30, 80, 70, 180, 80, 70, 0, 180, 0 },
{ XFER_PIO_4, 25, 70, 25, 120, 70, 25, 0, 120, 0 },
{ XFER_PIO_5, 15, 65, 25, 100, 65, 25, 0, 100, 0 },
{ XFER_PIO_6, 10, 55, 20, 80, 55, 20, 0, 80, 0 },
{ XFER_SW_DMA_0, 120, 0, 0, 0, 480, 480, 50, 960, 0 },
{ XFER_SW_DMA_1, 90, 0, 0, 0, 240, 240, 30, 480, 0 },
{ XFER_SW_DMA_2, 60, 0, 0, 0, 120, 120, 20, 240, 0 },
{ XFER_MW_DMA_0, 60, 0, 0, 0, 215, 215, 20, 480, 0 },
{ XFER_MW_DMA_1, 45, 0, 0, 0, 80, 50, 5, 150, 0 },
{ XFER_MW_DMA_2, 25, 0, 0, 0, 70, 25, 5, 120, 0 },
{ XFER_MW_DMA_3, 25, 0, 0, 0, 65, 25, 5, 100, 0 },
{ XFER_MW_DMA_4, 25, 0, 0, 0, 55, 20, 5, 80, 0 },
/* { XFER_UDMA_SLOW, 0, 0, 0, 0, 0, 0, 0, 0, 150 }, */
{ XFER_UDMA_0, 0, 0, 0, 0, 0, 0, 0, 0, 120 },
{ XFER_UDMA_1, 0, 0, 0, 0, 0, 0, 0, 0, 80 },
{ XFER_UDMA_2, 0, 0, 0, 0, 0, 0, 0, 0, 60 },
{ XFER_UDMA_3, 0, 0, 0, 0, 0, 0, 0, 0, 45 },
{ XFER_UDMA_4, 0, 0, 0, 0, 0, 0, 0, 0, 30 },
{ XFER_UDMA_5, 0, 0, 0, 0, 0, 0, 0, 0, 20 },
{ XFER_UDMA_6, 0, 0, 0, 0, 0, 0, 0, 0, 15 },
{ 0xFF }
};
#define ENOUGH(v, unit) (((v)-1)/(unit)+1)
#define EZ(v, unit) ((v)?ENOUGH(v, unit):0)
static void ata_timing_quantize(const struct ata_timing *t, struct ata_timing *q, int T, int UT)
{
q->setup = EZ(t->setup * 1000, T);
q->act8b = EZ(t->act8b * 1000, T);
q->rec8b = EZ(t->rec8b * 1000, T);
q->cyc8b = EZ(t->cyc8b * 1000, T);
q->active = EZ(t->active * 1000, T);
q->recover = EZ(t->recover * 1000, T);
q->dmack_hold = EZ(t->dmack_hold * 1000, T);
q->cycle = EZ(t->cycle * 1000, T);
q->udma = EZ(t->udma * 1000, UT);
}
void ata_timing_merge(const struct ata_timing *a, const struct ata_timing *b,
struct ata_timing *m, unsigned int what)
{
if (what & ATA_TIMING_SETUP ) m->setup = max(a->setup, b->setup);
if (what & ATA_TIMING_ACT8B ) m->act8b = max(a->act8b, b->act8b);
if (what & ATA_TIMING_REC8B ) m->rec8b = max(a->rec8b, b->rec8b);
if (what & ATA_TIMING_CYC8B ) m->cyc8b = max(a->cyc8b, b->cyc8b);
if (what & ATA_TIMING_ACTIVE ) m->active = max(a->active, b->active);
if (what & ATA_TIMING_RECOVER) m->recover = max(a->recover, b->recover);
if (what & ATA_TIMING_DMACK_HOLD) m->dmack_hold = max(a->dmack_hold, b->dmack_hold);
if (what & ATA_TIMING_CYCLE ) m->cycle = max(a->cycle, b->cycle);
if (what & ATA_TIMING_UDMA ) m->udma = max(a->udma, b->udma);
}
const struct ata_timing *ata_timing_find_mode(u8 xfer_mode)
{
const struct ata_timing *t = ata_timing;
while (xfer_mode > t->mode)
t++;
if (xfer_mode == t->mode)
return t;
WARN_ONCE(true, "%s: unable to find timing for xfer_mode 0x%x\n",
__func__, xfer_mode);
return NULL;
}
int ata_timing_compute(struct ata_device *adev, unsigned short speed,
struct ata_timing *t, int T, int UT)
{
const u16 *id = adev->id;
const struct ata_timing *s;
struct ata_timing p;
/*
* Find the mode.
*/
if (!(s = ata_timing_find_mode(speed)))
return -EINVAL;
memcpy(t, s, sizeof(*s));
/*
* If the drive is an EIDE drive, it can tell us it needs extended
* PIO/MW_DMA cycle timing.
*/
if (id[ATA_ID_FIELD_VALID] & 2) { /* EIDE drive */
memset(&p, 0, sizeof(p));
if (speed >= XFER_PIO_0 && speed < XFER_SW_DMA_0) {
if (speed <= XFER_PIO_2)
p.cycle = p.cyc8b = id[ATA_ID_EIDE_PIO];
else if ((speed <= XFER_PIO_4) ||
(speed == XFER_PIO_5 && !ata_id_is_cfa(id)))
p.cycle = p.cyc8b = id[ATA_ID_EIDE_PIO_IORDY];
} else if (speed >= XFER_MW_DMA_0 && speed <= XFER_MW_DMA_2)
p.cycle = id[ATA_ID_EIDE_DMA_MIN];
ata_timing_merge(&p, t, t, ATA_TIMING_CYCLE | ATA_TIMING_CYC8B);
}
/*
* Convert the timing to bus clock counts.
*/
ata_timing_quantize(t, t, T, UT);
/*
* Even in DMA/UDMA modes we still use PIO access for IDENTIFY,
* S.M.A.R.T * and some other commands. We have to ensure that the
* DMA cycle timing is slower/equal than the fastest PIO timing.
*/
if (speed > XFER_PIO_6) {
ata_timing_compute(adev, adev->pio_mode, &p, T, UT);
ata_timing_merge(&p, t, t, ATA_TIMING_ALL);
}
/*
* Lengthen active & recovery time so that cycle time is correct.
*/
if (t->act8b + t->rec8b < t->cyc8b) {
t->act8b += (t->cyc8b - (t->act8b + t->rec8b)) / 2;
t->rec8b = t->cyc8b - t->act8b;
}
if (t->active + t->recover < t->cycle) {
t->active += (t->cycle - (t->active + t->recover)) / 2;
t->recover = t->cycle - t->active;
}
/* In a few cases quantisation may produce enough errors to
leave t->cycle too low for the sum of active and recovery
if so we must correct this */
if (t->active + t->recover > t->cycle)
t->cycle = t->active + t->recover;
return 0;
}
/**
* ata_timing_cycle2mode - find xfer mode for the specified cycle duration
* @xfer_shift: ATA_SHIFT_* value for transfer type to examine.
* @cycle: cycle duration in ns
*
* Return matching xfer mode for @cycle. The returned mode is of
* the transfer type specified by @xfer_shift. If @cycle is too
* slow for @xfer_shift, 0xff is returned. If @cycle is faster
* than the fastest known mode, the fasted mode is returned.
*
* LOCKING:
* None.
*
* RETURNS:
* Matching xfer_mode, 0xff if no match found.
*/
u8 ata_timing_cycle2mode(unsigned int xfer_shift, int cycle)
{
u8 base_mode = 0xff, last_mode = 0xff;
const struct ata_xfer_ent *ent;
const struct ata_timing *t;
for (ent = ata_xfer_tbl; ent->shift >= 0; ent++)
if (ent->shift == xfer_shift)
base_mode = ent->base;
for (t = ata_timing_find_mode(base_mode);
t && ata_xfer_mode2shift(t->mode) == xfer_shift; t++) {
unsigned short this_cycle;
switch (xfer_shift) {
case ATA_SHIFT_PIO:
case ATA_SHIFT_MWDMA:
this_cycle = t->cycle;
break;
case ATA_SHIFT_UDMA:
this_cycle = t->udma;
break;
default:
return 0xff;
}
if (cycle > this_cycle)
break;
last_mode = t->mode;
}
return last_mode;
}
/**
* ata_down_xfermask_limit - adjust dev xfer masks downward
* @dev: Device to adjust xfer masks
* @sel: ATA_DNXFER_* selector
*
* Adjust xfer masks of @dev downward. Note that this function
* does not apply the change. Invoking ata_set_mode() afterwards
* will apply the limit.
*
* LOCKING:
* Inherited from caller.
*
* RETURNS:
* 0 on success, negative errno on failure
*/
int ata_down_xfermask_limit(struct ata_device *dev, unsigned int sel)
{
char buf[32];
unsigned long orig_mask, xfer_mask;
unsigned long pio_mask, mwdma_mask, udma_mask;
int quiet, highbit;
quiet = !!(sel & ATA_DNXFER_QUIET);
sel &= ~ATA_DNXFER_QUIET;
xfer_mask = orig_mask = ata_pack_xfermask(dev->pio_mask,
dev->mwdma_mask,
dev->udma_mask);
ata_unpack_xfermask(xfer_mask, &pio_mask, &mwdma_mask, &udma_mask);
switch (sel) {
case ATA_DNXFER_PIO:
highbit = fls(pio_mask) - 1;
pio_mask &= ~(1 << highbit);
break;
case ATA_DNXFER_DMA:
if (udma_mask) {
highbit = fls(udma_mask) - 1;
udma_mask &= ~(1 << highbit);
if (!udma_mask)
return -ENOENT;
} else if (mwdma_mask) {
highbit = fls(mwdma_mask) - 1;
mwdma_mask &= ~(1 << highbit);
if (!mwdma_mask)
return -ENOENT;
}
break;
case ATA_DNXFER_40C:
udma_mask &= ATA_UDMA_MASK_40C;
break;
case ATA_DNXFER_FORCE_PIO0:
pio_mask &= 1;
case ATA_DNXFER_FORCE_PIO:
mwdma_mask = 0;
udma_mask = 0;
break;
default:
BUG();
}
xfer_mask &= ata_pack_xfermask(pio_mask, mwdma_mask, udma_mask);
if (!(xfer_mask & ATA_MASK_PIO) || xfer_mask == orig_mask)
return -ENOENT;
if (!quiet) {
if (xfer_mask & (ATA_MASK_MWDMA | ATA_MASK_UDMA))
snprintf(buf, sizeof(buf), "%s:%s",
ata_mode_string(xfer_mask),
ata_mode_string(xfer_mask & ATA_MASK_PIO));
else
snprintf(buf, sizeof(buf), "%s",
ata_mode_string(xfer_mask));
ata_dev_warn(dev, "limiting speed to %s\n", buf);
}
ata_unpack_xfermask(xfer_mask, &dev->pio_mask, &dev->mwdma_mask,
&dev->udma_mask);
return 0;
}
static int ata_dev_set_mode(struct ata_device *dev)
{
struct ata_port *ap = dev->link->ap;
struct ata_eh_context *ehc = &dev->link->eh_context;
const bool nosetxfer = dev->horkage & ATA_HORKAGE_NOSETXFER;
const char *dev_err_whine = "";
int ign_dev_err = 0;
unsigned int err_mask = 0;
int rc;
dev->flags &= ~ATA_DFLAG_PIO;
if (dev->xfer_shift == ATA_SHIFT_PIO)
dev->flags |= ATA_DFLAG_PIO;
if (nosetxfer && ap->flags & ATA_FLAG_SATA && ata_id_is_sata(dev->id))
dev_err_whine = " (SET_XFERMODE skipped)";
else {
if (nosetxfer)
ata_dev_warn(dev,
"NOSETXFER but PATA detected - can't "
"skip SETXFER, might malfunction\n");
err_mask = ata_dev_set_xfermode(dev);
}
if (err_mask & ~AC_ERR_DEV)
goto fail;
/* revalidate */
ehc->i.flags |= ATA_EHI_POST_SETMODE;
rc = ata_dev_revalidate(dev, ATA_DEV_UNKNOWN, 0);
ehc->i.flags &= ~ATA_EHI_POST_SETMODE;
if (rc)
return rc;
if (dev->xfer_shift == ATA_SHIFT_PIO) {
/* Old CFA may refuse this command, which is just fine */
if (ata_id_is_cfa(dev->id))
ign_dev_err = 1;
/* Catch several broken garbage emulations plus some pre
ATA devices */
if (ata_id_major_version(dev->id) == 0 &&
dev->pio_mode <= XFER_PIO_2)
ign_dev_err = 1;
/* Some very old devices and some bad newer ones fail
any kind of SET_XFERMODE request but support PIO0-2
timings and no IORDY */
if (!ata_id_has_iordy(dev->id) && dev->pio_mode <= XFER_PIO_2)
ign_dev_err = 1;
}
/* Early MWDMA devices do DMA but don't allow DMA mode setting.
Don't fail an MWDMA0 set IFF the device indicates it is in MWDMA0 */
if (dev->xfer_shift == ATA_SHIFT_MWDMA &&
dev->dma_mode == XFER_MW_DMA_0 &&
(dev->id[63] >> 8) & 1)
ign_dev_err = 1;
/* if the device is actually configured correctly, ignore dev err */
if (dev->xfer_mode == ata_xfer_mask2mode(ata_id_xfermask(dev->id)))
ign_dev_err = 1;
if (err_mask & AC_ERR_DEV) {
if (!ign_dev_err)
goto fail;
else
dev_err_whine = " (device error ignored)";
}
DPRINTK("xfer_shift=%u, xfer_mode=0x%x\n",
dev->xfer_shift, (int)dev->xfer_mode);
ata_dev_info(dev, "configured for %s%s\n",
ata_mode_string(ata_xfer_mode2mask(dev->xfer_mode)),
dev_err_whine);
return 0;
fail:
ata_dev_err(dev, "failed to set xfermode (err_mask=0x%x)\n", err_mask);
return -EIO;
}
/**
* ata_do_set_mode - Program timings and issue SET FEATURES - XFER
* @link: link on which timings will be programmed
* @r_failed_dev: out parameter for failed device
*
* Standard implementation of the function used to tune and set
* ATA device disk transfer mode (PIO3, UDMA6, etc.). If
* ata_dev_set_mode() fails, pointer to the failing device is
* returned in @r_failed_dev.
*
* LOCKING:
2005-05-31 07:49:12 +08:00
* PCI/etc. bus probe sem.
*
* RETURNS:
* 0 on success, negative errno otherwise
*/
int ata_do_set_mode(struct ata_link *link, struct ata_device **r_failed_dev)
{
struct ata_port *ap = link->ap;
struct ata_device *dev;
int rc = 0, used_dma = 0, found = 0;
/* step 1: calculate xfer_mask */
ata_for_each_dev(dev, link, ENABLED) {
unsigned long pio_mask, dma_mask;
unsigned int mode_mask;
mode_mask = ATA_DMA_MASK_ATA;
if (dev->class == ATA_DEV_ATAPI)
mode_mask = ATA_DMA_MASK_ATAPI;
else if (ata_id_is_cfa(dev->id))
mode_mask = ATA_DMA_MASK_CFA;
ata_dev_xfermask(dev);
ata_force_xfermask(dev);
pio_mask = ata_pack_xfermask(dev->pio_mask, 0, 0);
if (libata_dma_mask & mode_mask)
dma_mask = ata_pack_xfermask(0, dev->mwdma_mask,
dev->udma_mask);
else
dma_mask = 0;
dev->pio_mode = ata_xfer_mask2mode(pio_mask);
dev->dma_mode = ata_xfer_mask2mode(dma_mask);
found = 1;
if (ata_dma_enabled(dev))
used_dma = 1;
}
if (!found)
goto out;
/* step 2: always set host PIO timings */
ata_for_each_dev(dev, link, ENABLED) {
if (dev->pio_mode == 0xff) {
ata_dev_warn(dev, "no PIO support\n");
rc = -EINVAL;
goto out;
}
dev->xfer_mode = dev->pio_mode;
dev->xfer_shift = ATA_SHIFT_PIO;
if (ap->ops->set_piomode)
ap->ops->set_piomode(ap, dev);
}
/* step 3: set host DMA timings */
ata_for_each_dev(dev, link, ENABLED) {
if (!ata_dma_enabled(dev))
continue;
dev->xfer_mode = dev->dma_mode;
dev->xfer_shift = ata_xfer_mode2shift(dev->dma_mode);
if (ap->ops->set_dmamode)
ap->ops->set_dmamode(ap, dev);
}
/* step 4: update devices' xfer mode */
ata_for_each_dev(dev, link, ENABLED) {
rc = ata_dev_set_mode(dev);
if (rc)
goto out;
}
/* Record simplex status. If we selected DMA then the other
* host channels are not permitted to do so.
*/
if (used_dma && (ap->host->flags & ATA_HOST_SIMPLEX))
ap->host->simplex_claimed = ap;
out:
if (rc)
*r_failed_dev = dev;
return rc;
}
/**
* ata_wait_ready - wait for link to become ready
* @link: link to be waited on
* @deadline: deadline jiffies for the operation
* @check_ready: callback to check link readiness
*
* Wait for @link to become ready. @check_ready should return
* positive number if @link is ready, 0 if it isn't, -ENODEV if
* link doesn't seem to be occupied, other errno for other error
* conditions.
*
* Transient -ENODEV conditions are allowed for
* ATA_TMOUT_FF_WAIT.
*
* LOCKING:
* EH context.
*
* RETURNS:
* 0 if @linke is ready before @deadline; otherwise, -errno.
*/
int ata_wait_ready(struct ata_link *link, unsigned long deadline,
int (*check_ready)(struct ata_link *link))
{
unsigned long start = jiffies;
unsigned long nodev_deadline;
int warned = 0;
/* choose which 0xff timeout to use, read comment in libata.h */
if (link->ap->host->flags & ATA_HOST_PARALLEL_SCAN)
nodev_deadline = ata_deadline(start, ATA_TMOUT_FF_WAIT_LONG);
else
nodev_deadline = ata_deadline(start, ATA_TMOUT_FF_WAIT);
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
/* Slave readiness can't be tested separately from master. On
* M/S emulation configuration, this function should be called
* only on the master and it will handle both master and slave.
*/
WARN_ON(link == link->ap->slave_link);
if (time_after(nodev_deadline, deadline))
nodev_deadline = deadline;
while (1) {
unsigned long now = jiffies;
int ready, tmp;
ready = tmp = check_ready(link);
if (ready > 0)
return 0;
/*
* -ENODEV could be transient. Ignore -ENODEV if link
* is online. Also, some SATA devices take a long
* time to clear 0xff after reset. Wait for
* ATA_TMOUT_FF_WAIT[_LONG] on -ENODEV if link isn't
* offline.
*
* Note that some PATA controllers (pata_ali) explode
* if status register is read more than once when
* there's no device attached.
*/
if (ready == -ENODEV) {
if (ata_link_online(link))
ready = 0;
else if ((link->ap->flags & ATA_FLAG_SATA) &&
!ata_link_offline(link) &&
time_before(now, nodev_deadline))
ready = 0;
}
if (ready)
return ready;
if (time_after(now, deadline))
return -EBUSY;
if (!warned && time_after(now, start + 5 * HZ) &&
(deadline - now > 3 * HZ)) {
ata_link_warn(link,
"link is slow to respond, please be patient "
"(ready=%d)\n", tmp);
warned = 1;
}
ata_msleep(link->ap, 50);
}
}
/**
* ata_wait_after_reset - wait for link to become ready after reset
* @link: link to be waited on
* @deadline: deadline jiffies for the operation
* @check_ready: callback to check link readiness
*
* Wait for @link to become ready after reset.
*
* LOCKING:
* EH context.
*
* RETURNS:
* 0 if @linke is ready before @deadline; otherwise, -errno.
*/
int ata_wait_after_reset(struct ata_link *link, unsigned long deadline,
int (*check_ready)(struct ata_link *link))
{
ata_msleep(link->ap, ATA_WAIT_AFTER_RESET);
return ata_wait_ready(link, deadline, check_ready);
}
/**
* sata_link_debounce - debounce SATA phy status
* @link: ATA link to debounce SATA phy status for
* @params: timing parameters { interval, duratinon, timeout } in msec
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
* @deadline: deadline jiffies for the operation
*
* Make sure SStatus of @link reaches stable state, determined by
* holding the same value where DET is not 1 for @duration polled
* every @interval, before @timeout. Timeout constraints the
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
* beginning of the stable state. Because DET gets stuck at 1 on
* some controllers after hot unplugging, this functions waits
* until timeout then returns 0 if DET is stable at 1.
*
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
* @timeout is further limited by @deadline. The sooner of the
* two is used.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, -errno on failure.
*/
int sata_link_debounce(struct ata_link *link, const unsigned long *params,
unsigned long deadline)
{
unsigned long interval = params[0];
unsigned long duration = params[1];
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
unsigned long last_jiffies, t;
u32 last, cur;
int rc;
t = ata_deadline(jiffies, params[2]);
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
if (time_before(t, deadline))
deadline = t;
if ((rc = sata_scr_read(link, SCR_STATUS, &cur)))
return rc;
cur &= 0xf;
last = cur;
last_jiffies = jiffies;
while (1) {
ata_msleep(link->ap, interval);
if ((rc = sata_scr_read(link, SCR_STATUS, &cur)))
return rc;
cur &= 0xf;
/* DET stable? */
if (cur == last) {
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
if (cur == 1 && time_before(jiffies, deadline))
continue;
if (time_after(jiffies,
ata_deadline(last_jiffies, duration)))
return 0;
continue;
}
/* unstable, start over */
last = cur;
last_jiffies = jiffies;
/* Check deadline. If debouncing failed, return
* -EPIPE to tell upper layer to lower link speed.
*/
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
if (time_after(jiffies, deadline))
return -EPIPE;
}
}
/**
* sata_link_resume - resume SATA link
* @link: ATA link to resume SATA
* @params: timing parameters { interval, duratinon, timeout } in msec
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
* @deadline: deadline jiffies for the operation
*
* Resume SATA phy @link and debounce it.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, -errno on failure.
*/
int sata_link_resume(struct ata_link *link, const unsigned long *params,
unsigned long deadline)
{
int tries = ATA_LINK_RESUME_TRIES;
u32 scontrol, serror;
int rc;
if ((rc = sata_scr_read(link, SCR_CONTROL, &scontrol)))
return rc;
/*
* Writes to SControl sometimes get ignored under certain
* controllers (ata_piix SIDPR). Make sure DET actually is
* cleared.
*/
do {
scontrol = (scontrol & 0x0f0) | 0x300;
if ((rc = sata_scr_write(link, SCR_CONTROL, scontrol)))
return rc;
/*
* Some PHYs react badly if SStatus is pounded
* immediately after resuming. Delay 200ms before
* debouncing.
*/
ata_msleep(link->ap, 200);
/* is SControl restored correctly? */
if ((rc = sata_scr_read(link, SCR_CONTROL, &scontrol)))
return rc;
} while ((scontrol & 0xf0f) != 0x300 && --tries);
if ((scontrol & 0xf0f) != 0x300) {
ata_link_warn(link, "failed to resume link (SControl %X)\n",
scontrol);
return 0;
}
if (tries < ATA_LINK_RESUME_TRIES)
ata_link_warn(link, "link resume succeeded after %d retries\n",
ATA_LINK_RESUME_TRIES - tries);
if ((rc = sata_link_debounce(link, params, deadline)))
return rc;
/* clear SError, some PHYs require this even for SRST to work */
if (!(rc = sata_scr_read(link, SCR_ERROR, &serror)))
rc = sata_scr_write(link, SCR_ERROR, serror);
return rc != -EINVAL ? rc : 0;
}
/**
* sata_link_scr_lpm - manipulate SControl IPM and SPM fields
* @link: ATA link to manipulate SControl for
* @policy: LPM policy to configure
* @spm_wakeup: initiate LPM transition to active state
*
* Manipulate the IPM field of the SControl register of @link
* according to @policy. If @policy is ATA_LPM_MAX_POWER and
* @spm_wakeup is %true, the SPM field is manipulated to wake up
* the link. This function also clears PHYRDY_CHG before
* returning.
*
* LOCKING:
* EH context.
*
* RETURNS:
* 0 on succes, -errno otherwise.
*/
int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy,
bool spm_wakeup)
{
struct ata_eh_context *ehc = &link->eh_context;
bool woken_up = false;
u32 scontrol;
int rc;
rc = sata_scr_read(link, SCR_CONTROL, &scontrol);
if (rc)
return rc;
switch (policy) {
case ATA_LPM_MAX_POWER:
/* disable all LPM transitions */
scontrol |= (0x7 << 8);
/* initiate transition to active state */
if (spm_wakeup) {
scontrol |= (0x4 << 12);
woken_up = true;
}
break;
case ATA_LPM_MED_POWER:
/* allow LPM to PARTIAL */
scontrol &= ~(0x1 << 8);
scontrol |= (0x6 << 8);
break;
case ATA_LPM_MIN_POWER:
if (ata_link_nr_enabled(link) > 0)
/* no restrictions on LPM transitions */
scontrol &= ~(0x7 << 8);
else {
/* empty port, power off */
scontrol &= ~0xf;
scontrol |= (0x1 << 2);
}
break;
default:
WARN_ON(1);
}
rc = sata_scr_write(link, SCR_CONTROL, scontrol);
if (rc)
return rc;
/* give the link time to transit out of LPM state */
if (woken_up)
msleep(10);
/* clear PHYRDY_CHG from SError */
ehc->i.serror &= ~SERR_PHYRDY_CHG;
return sata_scr_write(link, SCR_ERROR, SERR_PHYRDY_CHG);
}
[PATCH] libata-hp-prep: add prereset() method and implement ata_std_prereset() With hotplug, every reset might be a probing reset and thus something similar to probe_init() is needed. prereset() method is called before a series of resets to a port and is the counterpart of postreset(). prereset() can tell EH to use different type of reset or skip reset by modifying ehc->i.action. This patch also implements ata_std_prereset(). Most controllers should be able to use this function directly or with some wrapping. After hotplug, different controllers need different actions to resume the PHY and detect the newly attached device. Controllers can be categorized as follows. * Controllers which can wait for the first D2H FIS after hotplug. Note that if the waiting is implemented by polling TF status, there needs to be a way to set BSY on PHY status change. It can be implemented by hardware or with the help of the driver. * Controllers which can wait for the first D2H FIS after sending COMRESET. These controllers need to issue COMRESET to wait for the first FIS. Note that the received D2H FIS could be the first D2H FIS after POR (power-on-reset) or D2H FIS in response to the COMRESET. Some controllers use COMRESET as TF status synchronization point and clear TF automatically (sata_sil). * Controllers which cannot wait for the first D2H FIS reliably. Blindly issuing SRST to spinning-up device often results in command issue failure or timeout, causing extended delay. For these controllers, ata_std_prereset() explicitly waits ATA_SPINUP_WAIT (currently 8s) to give newly attached device time to spin up, then issues reset. Note that failing to getting ready in ATA_SPINUP_WAIT is not critical. libata will retry. So, the timeout needs to be long enough to spin up most devices. LLDDs can tell ata_std_prereset() which of above action is needed with ATA_FLAG_HRST_TO_RESUME and ATA_FLAG_SKIP_D2H_BSY flags. These flags are PHY-specific property and will be moved to ata_link later. While at it, this patch unifies function typedef's such that they all have named arguments. Signed-off-by: Tejun Heo <htejun@gmail.com>
2006-05-31 17:27:48 +08:00
/**
* ata_std_prereset - prepare for reset
* @link: ATA link to be reset
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
* @deadline: deadline jiffies for the operation
[PATCH] libata-hp-prep: add prereset() method and implement ata_std_prereset() With hotplug, every reset might be a probing reset and thus something similar to probe_init() is needed. prereset() method is called before a series of resets to a port and is the counterpart of postreset(). prereset() can tell EH to use different type of reset or skip reset by modifying ehc->i.action. This patch also implements ata_std_prereset(). Most controllers should be able to use this function directly or with some wrapping. After hotplug, different controllers need different actions to resume the PHY and detect the newly attached device. Controllers can be categorized as follows. * Controllers which can wait for the first D2H FIS after hotplug. Note that if the waiting is implemented by polling TF status, there needs to be a way to set BSY on PHY status change. It can be implemented by hardware or with the help of the driver. * Controllers which can wait for the first D2H FIS after sending COMRESET. These controllers need to issue COMRESET to wait for the first FIS. Note that the received D2H FIS could be the first D2H FIS after POR (power-on-reset) or D2H FIS in response to the COMRESET. Some controllers use COMRESET as TF status synchronization point and clear TF automatically (sata_sil). * Controllers which cannot wait for the first D2H FIS reliably. Blindly issuing SRST to spinning-up device often results in command issue failure or timeout, causing extended delay. For these controllers, ata_std_prereset() explicitly waits ATA_SPINUP_WAIT (currently 8s) to give newly attached device time to spin up, then issues reset. Note that failing to getting ready in ATA_SPINUP_WAIT is not critical. libata will retry. So, the timeout needs to be long enough to spin up most devices. LLDDs can tell ata_std_prereset() which of above action is needed with ATA_FLAG_HRST_TO_RESUME and ATA_FLAG_SKIP_D2H_BSY flags. These flags are PHY-specific property and will be moved to ata_link later. While at it, this patch unifies function typedef's such that they all have named arguments. Signed-off-by: Tejun Heo <htejun@gmail.com>
2006-05-31 17:27:48 +08:00
*
* @link is about to be reset. Initialize it. Failure from
* prereset makes libata abort whole reset sequence and give up
* that port, so prereset should be best-effort. It does its
* best to prepare for reset sequence but if things go wrong, it
* should just whine, not fail.
[PATCH] libata-hp-prep: add prereset() method and implement ata_std_prereset() With hotplug, every reset might be a probing reset and thus something similar to probe_init() is needed. prereset() method is called before a series of resets to a port and is the counterpart of postreset(). prereset() can tell EH to use different type of reset or skip reset by modifying ehc->i.action. This patch also implements ata_std_prereset(). Most controllers should be able to use this function directly or with some wrapping. After hotplug, different controllers need different actions to resume the PHY and detect the newly attached device. Controllers can be categorized as follows. * Controllers which can wait for the first D2H FIS after hotplug. Note that if the waiting is implemented by polling TF status, there needs to be a way to set BSY on PHY status change. It can be implemented by hardware or with the help of the driver. * Controllers which can wait for the first D2H FIS after sending COMRESET. These controllers need to issue COMRESET to wait for the first FIS. Note that the received D2H FIS could be the first D2H FIS after POR (power-on-reset) or D2H FIS in response to the COMRESET. Some controllers use COMRESET as TF status synchronization point and clear TF automatically (sata_sil). * Controllers which cannot wait for the first D2H FIS reliably. Blindly issuing SRST to spinning-up device often results in command issue failure or timeout, causing extended delay. For these controllers, ata_std_prereset() explicitly waits ATA_SPINUP_WAIT (currently 8s) to give newly attached device time to spin up, then issues reset. Note that failing to getting ready in ATA_SPINUP_WAIT is not critical. libata will retry. So, the timeout needs to be long enough to spin up most devices. LLDDs can tell ata_std_prereset() which of above action is needed with ATA_FLAG_HRST_TO_RESUME and ATA_FLAG_SKIP_D2H_BSY flags. These flags are PHY-specific property and will be moved to ata_link later. While at it, this patch unifies function typedef's such that they all have named arguments. Signed-off-by: Tejun Heo <htejun@gmail.com>
2006-05-31 17:27:48 +08:00
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, -errno otherwise.
*/
int ata_std_prereset(struct ata_link *link, unsigned long deadline)
[PATCH] libata-hp-prep: add prereset() method and implement ata_std_prereset() With hotplug, every reset might be a probing reset and thus something similar to probe_init() is needed. prereset() method is called before a series of resets to a port and is the counterpart of postreset(). prereset() can tell EH to use different type of reset or skip reset by modifying ehc->i.action. This patch also implements ata_std_prereset(). Most controllers should be able to use this function directly or with some wrapping. After hotplug, different controllers need different actions to resume the PHY and detect the newly attached device. Controllers can be categorized as follows. * Controllers which can wait for the first D2H FIS after hotplug. Note that if the waiting is implemented by polling TF status, there needs to be a way to set BSY on PHY status change. It can be implemented by hardware or with the help of the driver. * Controllers which can wait for the first D2H FIS after sending COMRESET. These controllers need to issue COMRESET to wait for the first FIS. Note that the received D2H FIS could be the first D2H FIS after POR (power-on-reset) or D2H FIS in response to the COMRESET. Some controllers use COMRESET as TF status synchronization point and clear TF automatically (sata_sil). * Controllers which cannot wait for the first D2H FIS reliably. Blindly issuing SRST to spinning-up device often results in command issue failure or timeout, causing extended delay. For these controllers, ata_std_prereset() explicitly waits ATA_SPINUP_WAIT (currently 8s) to give newly attached device time to spin up, then issues reset. Note that failing to getting ready in ATA_SPINUP_WAIT is not critical. libata will retry. So, the timeout needs to be long enough to spin up most devices. LLDDs can tell ata_std_prereset() which of above action is needed with ATA_FLAG_HRST_TO_RESUME and ATA_FLAG_SKIP_D2H_BSY flags. These flags are PHY-specific property and will be moved to ata_link later. While at it, this patch unifies function typedef's such that they all have named arguments. Signed-off-by: Tejun Heo <htejun@gmail.com>
2006-05-31 17:27:48 +08:00
{
struct ata_port *ap = link->ap;
struct ata_eh_context *ehc = &link->eh_context;
const unsigned long *timing = sata_ehc_deb_timing(ehc);
[PATCH] libata-hp-prep: add prereset() method and implement ata_std_prereset() With hotplug, every reset might be a probing reset and thus something similar to probe_init() is needed. prereset() method is called before a series of resets to a port and is the counterpart of postreset(). prereset() can tell EH to use different type of reset or skip reset by modifying ehc->i.action. This patch also implements ata_std_prereset(). Most controllers should be able to use this function directly or with some wrapping. After hotplug, different controllers need different actions to resume the PHY and detect the newly attached device. Controllers can be categorized as follows. * Controllers which can wait for the first D2H FIS after hotplug. Note that if the waiting is implemented by polling TF status, there needs to be a way to set BSY on PHY status change. It can be implemented by hardware or with the help of the driver. * Controllers which can wait for the first D2H FIS after sending COMRESET. These controllers need to issue COMRESET to wait for the first FIS. Note that the received D2H FIS could be the first D2H FIS after POR (power-on-reset) or D2H FIS in response to the COMRESET. Some controllers use COMRESET as TF status synchronization point and clear TF automatically (sata_sil). * Controllers which cannot wait for the first D2H FIS reliably. Blindly issuing SRST to spinning-up device often results in command issue failure or timeout, causing extended delay. For these controllers, ata_std_prereset() explicitly waits ATA_SPINUP_WAIT (currently 8s) to give newly attached device time to spin up, then issues reset. Note that failing to getting ready in ATA_SPINUP_WAIT is not critical. libata will retry. So, the timeout needs to be long enough to spin up most devices. LLDDs can tell ata_std_prereset() which of above action is needed with ATA_FLAG_HRST_TO_RESUME and ATA_FLAG_SKIP_D2H_BSY flags. These flags are PHY-specific property and will be moved to ata_link later. While at it, this patch unifies function typedef's such that they all have named arguments. Signed-off-by: Tejun Heo <htejun@gmail.com>
2006-05-31 17:27:48 +08:00
int rc;
/* if we're about to do hardreset, nothing more to do */
if (ehc->i.action & ATA_EH_HARDRESET)
return 0;
/* if SATA, resume link */
if (ap->flags & ATA_FLAG_SATA) {
rc = sata_link_resume(link, timing, deadline);
/* whine about phy resume failure but proceed */
if (rc && rc != -EOPNOTSUPP)
ata_link_warn(link,
"failed to resume link for reset (errno=%d)\n",
rc);
[PATCH] libata-hp-prep: add prereset() method and implement ata_std_prereset() With hotplug, every reset might be a probing reset and thus something similar to probe_init() is needed. prereset() method is called before a series of resets to a port and is the counterpart of postreset(). prereset() can tell EH to use different type of reset or skip reset by modifying ehc->i.action. This patch also implements ata_std_prereset(). Most controllers should be able to use this function directly or with some wrapping. After hotplug, different controllers need different actions to resume the PHY and detect the newly attached device. Controllers can be categorized as follows. * Controllers which can wait for the first D2H FIS after hotplug. Note that if the waiting is implemented by polling TF status, there needs to be a way to set BSY on PHY status change. It can be implemented by hardware or with the help of the driver. * Controllers which can wait for the first D2H FIS after sending COMRESET. These controllers need to issue COMRESET to wait for the first FIS. Note that the received D2H FIS could be the first D2H FIS after POR (power-on-reset) or D2H FIS in response to the COMRESET. Some controllers use COMRESET as TF status synchronization point and clear TF automatically (sata_sil). * Controllers which cannot wait for the first D2H FIS reliably. Blindly issuing SRST to spinning-up device often results in command issue failure or timeout, causing extended delay. For these controllers, ata_std_prereset() explicitly waits ATA_SPINUP_WAIT (currently 8s) to give newly attached device time to spin up, then issues reset. Note that failing to getting ready in ATA_SPINUP_WAIT is not critical. libata will retry. So, the timeout needs to be long enough to spin up most devices. LLDDs can tell ata_std_prereset() which of above action is needed with ATA_FLAG_HRST_TO_RESUME and ATA_FLAG_SKIP_D2H_BSY flags. These flags are PHY-specific property and will be moved to ata_link later. While at it, this patch unifies function typedef's such that they all have named arguments. Signed-off-by: Tejun Heo <htejun@gmail.com>
2006-05-31 17:27:48 +08:00
}
/* no point in trying softreset on offline link */
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
if (ata_phys_link_offline(link))
ehc->i.action &= ~ATA_EH_SOFTRESET;
[PATCH] libata-hp-prep: add prereset() method and implement ata_std_prereset() With hotplug, every reset might be a probing reset and thus something similar to probe_init() is needed. prereset() method is called before a series of resets to a port and is the counterpart of postreset(). prereset() can tell EH to use different type of reset or skip reset by modifying ehc->i.action. This patch also implements ata_std_prereset(). Most controllers should be able to use this function directly or with some wrapping. After hotplug, different controllers need different actions to resume the PHY and detect the newly attached device. Controllers can be categorized as follows. * Controllers which can wait for the first D2H FIS after hotplug. Note that if the waiting is implemented by polling TF status, there needs to be a way to set BSY on PHY status change. It can be implemented by hardware or with the help of the driver. * Controllers which can wait for the first D2H FIS after sending COMRESET. These controllers need to issue COMRESET to wait for the first FIS. Note that the received D2H FIS could be the first D2H FIS after POR (power-on-reset) or D2H FIS in response to the COMRESET. Some controllers use COMRESET as TF status synchronization point and clear TF automatically (sata_sil). * Controllers which cannot wait for the first D2H FIS reliably. Blindly issuing SRST to spinning-up device often results in command issue failure or timeout, causing extended delay. For these controllers, ata_std_prereset() explicitly waits ATA_SPINUP_WAIT (currently 8s) to give newly attached device time to spin up, then issues reset. Note that failing to getting ready in ATA_SPINUP_WAIT is not critical. libata will retry. So, the timeout needs to be long enough to spin up most devices. LLDDs can tell ata_std_prereset() which of above action is needed with ATA_FLAG_HRST_TO_RESUME and ATA_FLAG_SKIP_D2H_BSY flags. These flags are PHY-specific property and will be moved to ata_link later. While at it, this patch unifies function typedef's such that they all have named arguments. Signed-off-by: Tejun Heo <htejun@gmail.com>
2006-05-31 17:27:48 +08:00
return 0;
}
/**
* sata_link_hardreset - reset link via SATA phy reset
* @link: link to reset
* @timing: timing parameters { interval, duratinon, timeout } in msec
libata: add deadline support to prereset and reset methods Add @deadline to prereset and reset methods and make them honor it. ata_wait_ready() which directly takes @deadline is implemented to be used as the wait function. This patch is in preparation for EH timing improvements. * ata_wait_ready() never does busy sleep. It's only used from EH and no wait in EH is that urgent. This function also prints 'be patient' message automatically after 5 secs of waiting if more than 3 secs is remaining till deadline. * ata_bus_post_reset() now fails with error code if any of its wait fails. This is important because earlier reset tries will have shorter timeout than the spec requires. If a device fails to respond before the short timeout, reset should be retried with longer timeout rather than silently ignoring the device. There are three behavior differences. 1. Timeout is applied to both devices at once, not separately. This is more consistent with what the spec says. 2. When a device passes devchk but fails to become ready before deadline. Previouly, post_reset would just succeed and let device classification remove the device. New code fails the reset thus causing reset retry. After a few times, EH will give up disabling the port. 3. When slave device passes devchk but fails to become accessible (TF-wise) after reset. Original code disables dev1 after 30s timeout and continues as if the device doesn't exist, while the patched code fails reset. When this happens, new code fails reset on whole port rather than proceeding with only the primary device. If the failing device is suffering transient problems, new code retries reset which is a better behavior. If the failing device is actually broken, the net effect is identical to it, but not to the other device sharing the channel. In the previous code, reset would have succeeded after 30s thus detecting the working one. In the new code, reset fails and whole port gets disabled. IMO, it's a pathological case anyway (broken device sharing bus with working one) and doesn't really matter. * ata_bus_softreset() is changed to return error code from ata_bus_post_reset(). It used to return 0 unconditionally. * Spin up waiting is to be removed and not converted to honor deadline. * To be on the safe side, deadline is set to 40s for the time being. Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-02-02 15:50:52 +08:00
* @deadline: deadline jiffies for the operation
* @online: optional out parameter indicating link onlineness
* @check_ready: optional callback to check link readiness
*
* SATA phy-reset @link using DET bits of SControl register.
* After hardreset, link readiness is waited upon using
* ata_wait_ready() if @check_ready is specified. LLDs are
* allowed to not specify @check_ready and wait itself after this
* function returns. Device classification is LLD's
* responsibility.
*
* *@online is set to one iff reset succeeded and @link is online
* after reset.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, -errno otherwise.
*/
int sata_link_hardreset(struct ata_link *link, const unsigned long *timing,
unsigned long deadline,
bool *online, int (*check_ready)(struct ata_link *))
{
u32 scontrol;
int rc;
DPRINTK("ENTER\n");
if (online)
*online = false;
if (sata_set_spd_needed(link)) {
/* SATA spec says nothing about how to reconfigure
* spd. To be on the safe side, turn off phy during
* reconfiguration. This works for at least ICH7 AHCI
* and Sil3124.
*/
if ((rc = sata_scr_read(link, SCR_CONTROL, &scontrol)))
goto out;
scontrol = (scontrol & 0x0f0) | 0x304;
if ((rc = sata_scr_write(link, SCR_CONTROL, scontrol)))
goto out;
sata_set_spd(link);
}
/* issue phy wake/reset */
if ((rc = sata_scr_read(link, SCR_CONTROL, &scontrol)))
goto out;
scontrol = (scontrol & 0x0f0) | 0x301;
if ((rc = sata_scr_write_flush(link, SCR_CONTROL, scontrol)))
goto out;
/* Couldn't find anything in SATA I/II specs, but AHCI-1.1
* 10.4.2 says at least 1 ms.
*/
ata_msleep(link->ap, 1);
/* bring link back */
rc = sata_link_resume(link, timing, deadline);
if (rc)
goto out;
/* if link is offline nothing more to do */
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
if (ata_phys_link_offline(link))
goto out;
/* Link is online. From this point, -ENODEV too is an error. */
if (online)
*online = true;
if (sata_pmp_supported(link->ap) && ata_is_host_link(link)) {
/* If PMP is supported, we have to do follow-up SRST.
* Some PMPs don't send D2H Reg FIS after hardreset if
* the first port is empty. Wait only for
* ATA_TMOUT_PMP_SRST_WAIT.
*/
if (check_ready) {
unsigned long pmp_deadline;
pmp_deadline = ata_deadline(jiffies,
ATA_TMOUT_PMP_SRST_WAIT);
if (time_after(pmp_deadline, deadline))
pmp_deadline = deadline;
ata_wait_ready(link, pmp_deadline, check_ready);
}
rc = -EAGAIN;
goto out;
}
rc = 0;
if (check_ready)
rc = ata_wait_ready(link, deadline, check_ready);
out:
if (rc && rc != -EAGAIN) {
/* online is set iff link is online && reset succeeded */
if (online)
*online = false;
ata_link_err(link, "COMRESET failed (errno=%d)\n", rc);
}
DPRINTK("EXIT, rc=%d\n", rc);
return rc;
}
/**
* sata_std_hardreset - COMRESET w/o waiting or classification
* @link: link to reset
* @class: resulting class of attached device
* @deadline: deadline jiffies for the operation
*
* Standard SATA COMRESET w/o waiting or classification.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 if link offline, -EAGAIN if link online, -errno on errors.
*/
int sata_std_hardreset(struct ata_link *link, unsigned int *class,
unsigned long deadline)
{
const unsigned long *timing = sata_ehc_deb_timing(&link->eh_context);
bool online;
int rc;
/* do hardreset */
rc = sata_link_hardreset(link, timing, deadline, &online, NULL);
return online ? -EAGAIN : rc;
}
/**
* ata_std_postreset - standard postreset callback
* @link: the target ata_link
* @classes: classes of attached devices
*
* This function is invoked after a successful reset. Note that
* the device might have been reset more than once using
* different reset methods before postreset is invoked.
*
* LOCKING:
* Kernel thread context (may sleep)
*/
void ata_std_postreset(struct ata_link *link, unsigned int *classes)
{
u32 serror;
DPRINTK("ENTER\n");
/* reset complete, clear SError */
if (!sata_scr_read(link, SCR_ERROR, &serror))
sata_scr_write(link, SCR_ERROR, serror);
/* print link status */
sata_print_link_status(link);
DPRINTK("EXIT\n");
}
/**
* ata_dev_same_device - Determine whether new ID matches configured device
* @dev: device to compare against
* @new_class: class of the new device
* @new_id: IDENTIFY page of the new device
*
* Compare @new_class and @new_id against @dev and determine
* whether @dev is the device indicated by @new_class and
* @new_id.
*
* LOCKING:
* None.
*
* RETURNS:
* 1 if @dev matches @new_class and @new_id, 0 otherwise.
*/
static int ata_dev_same_device(struct ata_device *dev, unsigned int new_class,
const u16 *new_id)
{
const u16 *old_id = dev->id;
unsigned char model[2][ATA_ID_PROD_LEN + 1];
unsigned char serial[2][ATA_ID_SERNO_LEN + 1];
if (dev->class != new_class) {
ata_dev_info(dev, "class mismatch %d != %d\n",
dev->class, new_class);
return 0;
}
ata_id_c_string(old_id, model[0], ATA_ID_PROD, sizeof(model[0]));
ata_id_c_string(new_id, model[1], ATA_ID_PROD, sizeof(model[1]));
ata_id_c_string(old_id, serial[0], ATA_ID_SERNO, sizeof(serial[0]));
ata_id_c_string(new_id, serial[1], ATA_ID_SERNO, sizeof(serial[1]));
if (strcmp(model[0], model[1])) {
ata_dev_info(dev, "model number mismatch '%s' != '%s'\n",
model[0], model[1]);
return 0;
}
if (strcmp(serial[0], serial[1])) {
ata_dev_info(dev, "serial number mismatch '%s' != '%s'\n",
serial[0], serial[1]);
return 0;
}
return 1;
}
/**
* ata_dev_reread_id - Re-read IDENTIFY data
* @dev: target ATA device
* @readid_flags: read ID flags
*
* Re-read IDENTIFY page and make sure @dev is still attached to
* the port.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, negative errno otherwise
*/
int ata_dev_reread_id(struct ata_device *dev, unsigned int readid_flags)
{
unsigned int class = dev->class;
u16 *id = (void *)dev->link->ap->sector_buf;
int rc;
/* read ID data */
rc = ata_dev_read_id(dev, &class, readid_flags, id);
if (rc)
return rc;
/* is the device still there? */
if (!ata_dev_same_device(dev, class, id))
return -ENODEV;
memcpy(dev->id, id, sizeof(id[0]) * ATA_ID_WORDS);
return 0;
}
/**
* ata_dev_revalidate - Revalidate ATA device
* @dev: device to revalidate
* @new_class: new class code
* @readid_flags: read ID flags
*
* Re-read IDENTIFY page, make sure @dev is still attached to the
* port and reconfigure it according to the new IDENTIFY page.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, negative errno otherwise
*/
int ata_dev_revalidate(struct ata_device *dev, unsigned int new_class,
unsigned int readid_flags)
{
u64 n_sectors = dev->n_sectors;
u64 n_native_sectors = dev->n_native_sectors;
int rc;
if (!ata_dev_enabled(dev))
return -ENODEV;
/* fail early if !ATA && !ATAPI to avoid issuing [P]IDENTIFY to PMP */
if (ata_class_enabled(new_class) &&
libata: fix suspend/resume for ATA SEMB devices 79b42babbac2a5a522b8e269fb2811b6e1063030 fixed identifying ATA devices reporting 3c/c3 signature which belongs to SEMB devices now. However, suspending the machine with such device (WDC WD2500AAJS-6 01.0) fails with the following: hda: host max PIO4 wanted PIO255(auto-tune) selected PIO4 hda: UDMA/100 mode selected hdb: host max PIO4 wanted PIO255(auto-tune) selected PIO4 hdb: UDMA/66 mode selected sd 1:0:0:0: [sda] Starting disk ata5: SATA link down (SStatus 0 SControl 300) ata1: SATA link down (SStatus 0 SControl 300) ata3: SATA link down (SStatus 0 SControl 300) ata6: SATA link down (SStatus 0 SControl 300) ata2: softreset failed (device not ready) ata2: failed due to HW bug, retry pmp=0 ata4: softreset failed (device not ready) ata4: failed due to HW bug, retry pmp=0 ata4: SATA link up 3.0 Gbps (SStatus 123 SControl 300) ata2: SATA link up 3.0 Gbps (SStatus 123 SControl 300) ata2.00: class mismatch 1 != 7 ata2.00: revalidation failed (errno=-19) ata2: limiting SATA link speed to 1.5 Gbps ata4.00: configured for UDMA/133 ata2: softreset failed (device not ready) ata2: failed due to HW bug, retry pmp=0 ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 310) ata2.00: class mismatch 1 != 7 ata2.00: revalidation failed (errno=-19) ata2.00: disabled sd 1:0:0:0: rejecting I/O to offline device sd 1:0:0:0: [sda] START_STOP FAILED sd 1:0:0:0: [sda] Result: hostbyte=0x01 driverbyte=0x00 PM: Device 1:0:0:0 failed to thaw: error 65536 sd 3:0:0:0: [sdb] Starting disk due to a class mismatch in ata_dev_revalidate(). Fix it by adding the ATA_DEV_SEMB device class to the check. CC: Tejun Heo <htejun@gmail.com> Signed-off-by: Borislav Petkov <petkovbb@gmail.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2009-04-26 22:33:34 +08:00
new_class != ATA_DEV_ATA &&
new_class != ATA_DEV_ATAPI &&
new_class != ATA_DEV_ZAC &&
libata: fix suspend/resume for ATA SEMB devices 79b42babbac2a5a522b8e269fb2811b6e1063030 fixed identifying ATA devices reporting 3c/c3 signature which belongs to SEMB devices now. However, suspending the machine with such device (WDC WD2500AAJS-6 01.0) fails with the following: hda: host max PIO4 wanted PIO255(auto-tune) selected PIO4 hda: UDMA/100 mode selected hdb: host max PIO4 wanted PIO255(auto-tune) selected PIO4 hdb: UDMA/66 mode selected sd 1:0:0:0: [sda] Starting disk ata5: SATA link down (SStatus 0 SControl 300) ata1: SATA link down (SStatus 0 SControl 300) ata3: SATA link down (SStatus 0 SControl 300) ata6: SATA link down (SStatus 0 SControl 300) ata2: softreset failed (device not ready) ata2: failed due to HW bug, retry pmp=0 ata4: softreset failed (device not ready) ata4: failed due to HW bug, retry pmp=0 ata4: SATA link up 3.0 Gbps (SStatus 123 SControl 300) ata2: SATA link up 3.0 Gbps (SStatus 123 SControl 300) ata2.00: class mismatch 1 != 7 ata2.00: revalidation failed (errno=-19) ata2: limiting SATA link speed to 1.5 Gbps ata4.00: configured for UDMA/133 ata2: softreset failed (device not ready) ata2: failed due to HW bug, retry pmp=0 ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 310) ata2.00: class mismatch 1 != 7 ata2.00: revalidation failed (errno=-19) ata2.00: disabled sd 1:0:0:0: rejecting I/O to offline device sd 1:0:0:0: [sda] START_STOP FAILED sd 1:0:0:0: [sda] Result: hostbyte=0x01 driverbyte=0x00 PM: Device 1:0:0:0 failed to thaw: error 65536 sd 3:0:0:0: [sdb] Starting disk due to a class mismatch in ata_dev_revalidate(). Fix it by adding the ATA_DEV_SEMB device class to the check. CC: Tejun Heo <htejun@gmail.com> Signed-off-by: Borislav Petkov <petkovbb@gmail.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2009-04-26 22:33:34 +08:00
new_class != ATA_DEV_SEMB) {
ata_dev_info(dev, "class mismatch %u != %u\n",
dev->class, new_class);
rc = -ENODEV;
goto fail;
}
/* re-read ID */
rc = ata_dev_reread_id(dev, readid_flags);
if (rc)
goto fail;
/* configure device according to the new ID */
rc = ata_dev_configure(dev);
if (rc)
goto fail;
/* verify n_sectors hasn't changed */
if (dev->class != ATA_DEV_ATA || !n_sectors ||
dev->n_sectors == n_sectors)
return 0;
/* n_sectors has changed */
ata_dev_warn(dev, "n_sectors mismatch %llu != %llu\n",
(unsigned long long)n_sectors,
(unsigned long long)dev->n_sectors);
/*
* Something could have caused HPA to be unlocked
* involuntarily. If n_native_sectors hasn't changed and the
* new size matches it, keep the device.
*/
if (dev->n_native_sectors == n_native_sectors &&
dev->n_sectors > n_sectors && dev->n_sectors == n_native_sectors) {
ata_dev_warn(dev,
"new n_sectors matches native, probably "
"late HPA unlock, n_sectors updated\n");
/* use the larger n_sectors */
return 0;
}
/*
* Some BIOSes boot w/o HPA but resume w/ HPA locked. Try
* unlocking HPA in those cases.
*
* https://bugzilla.kernel.org/show_bug.cgi?id=15396
*/
if (dev->n_native_sectors == n_native_sectors &&
dev->n_sectors < n_sectors && n_sectors == n_native_sectors &&
!(dev->horkage & ATA_HORKAGE_BROKEN_HPA)) {
ata_dev_warn(dev,
"old n_sectors matches native, probably "
"late HPA lock, will try to unlock HPA\n");
/* try unlocking HPA */
dev->flags |= ATA_DFLAG_UNLOCK_HPA;
rc = -EIO;
} else
rc = -ENODEV;
/* restore original n_[native_]sectors and fail */
dev->n_native_sectors = n_native_sectors;
dev->n_sectors = n_sectors;
fail:
ata_dev_err(dev, "revalidation failed (errno=%d)\n", rc);
return rc;
}
struct ata_blacklist_entry {
const char *model_num;
const char *model_rev;
unsigned long horkage;
};
static const struct ata_blacklist_entry ata_device_blacklist [] = {
/* Devices with DMA related problems under Linux */
{ "WDC AC11000H", NULL, ATA_HORKAGE_NODMA },
{ "WDC AC22100H", NULL, ATA_HORKAGE_NODMA },
{ "WDC AC32500H", NULL, ATA_HORKAGE_NODMA },
{ "WDC AC33100H", NULL, ATA_HORKAGE_NODMA },
{ "WDC AC31600H", NULL, ATA_HORKAGE_NODMA },
{ "WDC AC32100H", "24.09P07", ATA_HORKAGE_NODMA },
{ "WDC AC23200L", "21.10N21", ATA_HORKAGE_NODMA },
{ "Compaq CRD-8241B", NULL, ATA_HORKAGE_NODMA },
{ "CRD-8400B", NULL, ATA_HORKAGE_NODMA },
{ "CRD-848[02]B", NULL, ATA_HORKAGE_NODMA },
{ "CRD-84", NULL, ATA_HORKAGE_NODMA },
{ "SanDisk SDP3B", NULL, ATA_HORKAGE_NODMA },
{ "SanDisk SDP3B-64", NULL, ATA_HORKAGE_NODMA },
{ "SANYO CD-ROM CRD", NULL, ATA_HORKAGE_NODMA },
{ "HITACHI CDR-8", NULL, ATA_HORKAGE_NODMA },
{ "HITACHI CDR-8[34]35",NULL, ATA_HORKAGE_NODMA },
{ "Toshiba CD-ROM XM-6202B", NULL, ATA_HORKAGE_NODMA },
{ "TOSHIBA CD-ROM XM-1702BC", NULL, ATA_HORKAGE_NODMA },
{ "CD-532E-A", NULL, ATA_HORKAGE_NODMA },
{ "E-IDE CD-ROM CR-840",NULL, ATA_HORKAGE_NODMA },
{ "CD-ROM Drive/F5A", NULL, ATA_HORKAGE_NODMA },
{ "WPI CDD-820", NULL, ATA_HORKAGE_NODMA },
{ "SAMSUNG CD-ROM SC-148C", NULL, ATA_HORKAGE_NODMA },
{ "SAMSUNG CD-ROM SC", NULL, ATA_HORKAGE_NODMA },
{ "ATAPI CD-ROM DRIVE 40X MAXIMUM",NULL,ATA_HORKAGE_NODMA },
{ "_NEC DV5800A", NULL, ATA_HORKAGE_NODMA },
{ "SAMSUNG CD-ROM SN-124", "N001", ATA_HORKAGE_NODMA },
{ "Seagate STT20000A", NULL, ATA_HORKAGE_NODMA },
{ " 2GB ATA Flash Disk", "ADMA428M", ATA_HORKAGE_NODMA },
/* Odd clown on sil3726/4726 PMPs */
{ "Config Disk", NULL, ATA_HORKAGE_DISABLE },
/* Weird ATAPI devices */
{ "TORiSAN DVD-ROM DRD-N216", NULL, ATA_HORKAGE_MAX_SEC_128 },
{ "QUANTUM DAT DAT72-000", NULL, ATA_HORKAGE_ATAPI_MOD16_DMA },
libata: Set max sector to 65535 for Slimtype DVD A DS8A8SH drive The Slimtype DVD A DS8A8SH drive locks up when max sector is smaller than 65535, and the blow backtrace is observed on locking up: INFO: task flush-8:32:1130 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. flush-8:32 D ffffffff8180cf60 0 1130 2 0x00000000 ffff880273aef618 0000000000000046 0000000000000005 ffff880273aee000 ffff880273aee000 ffff880273aeffd8 ffff880273aee010 ffff880273aee000 ffff880273aeffd8 ffff880273aee000 ffff88026e842ea0 ffff880274a10000 Call Trace: [<ffffffff8168fc2d>] schedule+0x5d/0x70 [<ffffffff8168fccc>] io_schedule+0x8c/0xd0 [<ffffffff81324461>] get_request+0x731/0x7d0 [<ffffffff8133dc60>] ? cfq_allow_merge+0x50/0x90 [<ffffffff81083aa0>] ? wake_up_bit+0x40/0x40 [<ffffffff81320443>] ? bio_attempt_back_merge+0x33/0x110 [<ffffffff813248ea>] blk_queue_bio+0x23a/0x3f0 [<ffffffff81322176>] generic_make_request+0xc6/0x120 [<ffffffff81322308>] submit_bio+0x138/0x160 [<ffffffff811d7596>] ? bio_alloc_bioset+0x96/0x120 [<ffffffff811d1f61>] submit_bh+0x1f1/0x220 [<ffffffff811d48b8>] __block_write_full_page+0x228/0x340 [<ffffffff811d3650>] ? attach_nobh_buffers+0xc0/0xc0 [<ffffffff811d8960>] ? I_BDEV+0x10/0x10 [<ffffffff811d8960>] ? I_BDEV+0x10/0x10 [<ffffffff811d4ab6>] block_write_full_page_endio+0xe6/0x100 [<ffffffff811d4ae5>] block_write_full_page+0x15/0x20 [<ffffffff811d9268>] blkdev_writepage+0x18/0x20 [<ffffffff81142527>] __writepage+0x17/0x40 [<ffffffff811438ba>] write_cache_pages+0x34a/0x4a0 [<ffffffff81142510>] ? set_page_dirty+0x70/0x70 [<ffffffff81143a61>] generic_writepages+0x51/0x80 [<ffffffff81143ab0>] do_writepages+0x20/0x50 [<ffffffff811c9ed6>] __writeback_single_inode+0xa6/0x2b0 [<ffffffff811ca861>] writeback_sb_inodes+0x311/0x4d0 [<ffffffff811caaa6>] __writeback_inodes_wb+0x86/0xd0 [<ffffffff811cad43>] wb_writeback+0x1a3/0x330 [<ffffffff816916cf>] ? _raw_spin_lock_irqsave+0x3f/0x50 [<ffffffff811b8362>] ? get_nr_inodes+0x52/0x70 [<ffffffff811cb0ac>] wb_do_writeback+0x1dc/0x260 [<ffffffff8168dd34>] ? schedule_timeout+0x204/0x240 [<ffffffff811cb232>] bdi_writeback_thread+0x102/0x2b0 [<ffffffff811cb130>] ? wb_do_writeback+0x260/0x260 [<ffffffff81083550>] kthread+0xc0/0xd0 [<ffffffff81083490>] ? kthread_worker_fn+0x1b0/0x1b0 [<ffffffff8169a3ec>] ret_from_fork+0x7c/0xb0 [<ffffffff81083490>] ? kthread_worker_fn+0x1b0/0x1b0 The above trace was triggered by "dd if=/dev/zero of=/dev/sr0 bs=2048 count=32768" It was previously working by accident, since another bug introduced by 4dce8ba94c7 (libata: Use 'bool' return value for ata_id_XXX) caused all drives to use maxsect=65535. Cc: stable@vger.kernel.org Signed-off-by: Shan Hai <shan.hai@windriver.com> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2013-03-18 10:30:44 +08:00
{ "Slimtype DVD A DS8A8SH", NULL, ATA_HORKAGE_MAX_SEC_LBA48 },
drivers/libata: Set max sector to 65535 for Slimtype DVD A DS8A9SH drive The "Slimtype DVD A DS8A9SH" drive locks up with following backtrace when the max sector is smaller than 65535 bytes, fix it by adding a quirk to set the max sector to 65535 bytes. INFO: task flush-11:0:663 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. flush-11:0 D 00000000ffff5ceb 0 663 2 0x00000000 ffff88026d3b1710 0000000000000046 0000000000000001 0000000000000000 ffff88026f2530c0 ffff88026d365860 ffff88026d3b16e0 ffffffff812ffd52 ffff88026d4fd3d0 0000000100000001 ffff88026d3b16f0 ffff88026d3b1fd8 Call Trace: [<ffffffff812ffd52>] ? cfq_may_queue+0x52/0xf0 [<ffffffff81604338>] schedule+0x18/0x30 [<ffffffff81604392>] io_schedule+0x42/0x60 [<ffffffff812f22bb>] get_request_wait+0xeb/0x1f0 [<ffffffff81065660>] ? autoremove_wake_function+0x0/0x40 [<ffffffff812eb382>] ? elv_merge+0x42/0x210 [<ffffffff812f26ae>] __make_request+0x8e/0x4e0 [<ffffffff812f068e>] generic_make_request+0x21e/0x5e0 [<ffffffff812f0aad>] submit_bio+0x5d/0xd0 [<ffffffff81141422>] submit_bh+0xf2/0x130 [<ffffffff8114474c>] __block_write_full_page+0x1dc/0x3a0 [<ffffffff81143f60>] ? end_buffer_async_write+0x0/0x120 [<ffffffff811474e0>] ? blkdev_get_block+0x0/0x70 [<ffffffff811474e0>] ? blkdev_get_block+0x0/0x70 [<ffffffff81143f60>] ? end_buffer_async_write+0x0/0x120 [<ffffffff811449ee>] block_write_full_page_endio+0xde/0x100 [<ffffffff81144a20>] block_write_full_page+0x10/0x20 [<ffffffff81148703>] blkdev_writepage+0x13/0x20 [<ffffffff810d7525>] __writepage+0x15/0x40 [<ffffffff810d7c0f>] write_cache_pages+0x1cf/0x3e0 [<ffffffff810d7510>] ? __writepage+0x0/0x40 [<ffffffff810d7e42>] generic_writepages+0x22/0x30 [<ffffffff810d7e6f>] do_writepages+0x1f/0x40 [<ffffffff8113ae67>] writeback_single_inode+0xe7/0x3b0 [<ffffffff8113b574>] writeback_sb_inodes+0x184/0x280 [<ffffffff8113bedb>] writeback_inodes_wb+0x6b/0x1a0 [<ffffffff8113c24b>] wb_writeback+0x23b/0x2a0 [<ffffffff8113c42d>] wb_do_writeback+0x17d/0x190 [<ffffffff8113c48b>] bdi_writeback_task+0x4b/0xe0 [<ffffffff810e82a0>] ? bdi_start_fn+0x0/0x100 [<ffffffff810e8321>] bdi_start_fn+0x81/0x100 [<ffffffff810e82a0>] ? bdi_start_fn+0x0/0x100 [<ffffffff8106522e>] kthread+0x8e/0xa0 [<ffffffff81039274>] ? finish_task_switch+0x54/0xc0 [<ffffffff81003334>] kernel_thread_helper+0x4/0x10 [<ffffffff810651a0>] ? kthread+0x0/0xa0 [<ffffffff81003330>] ? kernel_thread_helper+0x0/0x10 The above trace was triggered by "dd if=/dev/zero of=/dev/sr0 bs=2048 count=32768" Cc: stable@vger.kernel.org Signed-off-by: Shan Hai <shan.hai@windriver.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2013-10-28 16:08:01 +08:00
{ "Slimtype DVD A DS8A9SH", NULL, ATA_HORKAGE_MAX_SEC_LBA48 },
/* Devices we expect to fail diagnostics */
/* Devices where NCQ should be avoided */
/* NCQ is slow */
{ "WDC WD740ADFD-00", NULL, ATA_HORKAGE_NONCQ },
libata: kill spurious NCQ completion detection Spurious NCQ completion detection implemented in ahci was incorrect. On AHCI receving and processing FISes and raising interrupts are not interlocked and spurious interrupts are expected. For example, if an interrupt occurs while interrupt handler is running and the running interrupt handler handles the event the new IRQ indicated, after IRQ handler finishes, it will be executed again because IRQ pending bit is set by the new interrupt but there won't be anything to process. Please read the following message for more information. http://article.gmane.org/gmane.linux.ide/26012 This patch... * Removes all spurious IRQ whining from ahci. Spurious NCQ completion detection was completely wrong. Spurious D2H Register FIS taught us that some early drives send spurious D2H Register FIS with I bit set while NCQ commands are in progress but none of recent drives does that and even the ones which show such behavior can do NCQ fine. * Kills all NCQ blacklist entries which were added because of spurious NCQ completions. I tracked down each commit and verified all removed ones are actually added because of spurious completions. WD740ADFD-00NLR1 wasn't deleted but moved upward because the drive not only had spurious NCQ completions but also is slow on sequential data transfers if NCQ is enabled. Maxtor 7V300F0 was added by 0e3dbc01d53940fe10e5a5cfec15ede3e929c918 from Alan Cox. I can only find evidences that the drive only had troubles with spuruious completions by searching the mailing list. This entry needs to be verified and removed if it doesn't have other NCQ related problems. Signed-off-by: Tejun Heo <htejun@gmail.com> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-12-07 11:46:23 +08:00
{ "WDC WD740ADFD-00NLR1", NULL, ATA_HORKAGE_NONCQ, },
/* http://thread.gmane.org/gmane.linux.ide/14907 */
{ "FUJITSU MHT2060BH", NULL, ATA_HORKAGE_NONCQ },
/* NCQ is broken */
{ "Maxtor *", "BANC*", ATA_HORKAGE_NONCQ },
{ "Maxtor 7V300F0", "VA111630", ATA_HORKAGE_NONCQ },
{ "ST380817AS", "3.42", ATA_HORKAGE_NONCQ },
{ "ST3160023AS", "3.42", ATA_HORKAGE_NONCQ },
{ "OCZ CORE_SSD", "02.10104", ATA_HORKAGE_NONCQ },
/* Seagate NCQ + FLUSH CACHE firmware bug */
{ "ST31500341AS", "SD1[5-9]", ATA_HORKAGE_NONCQ |
ATA_HORKAGE_FIRMWARE_WARN },
{ "ST31000333AS", "SD1[5-9]", ATA_HORKAGE_NONCQ |
ATA_HORKAGE_FIRMWARE_WARN },
{ "ST3640[36]23AS", "SD1[5-9]", ATA_HORKAGE_NONCQ |
ATA_HORKAGE_FIRMWARE_WARN },
{ "ST3320[68]13AS", "SD1[5-9]", ATA_HORKAGE_NONCQ |
ATA_HORKAGE_FIRMWARE_WARN },
/* Seagate Momentus SpinPoint M8 seem to have FPMDA_AA issues */
{ "ST1000LM024 HN-M101MBB", "2AR10001", ATA_HORKAGE_BROKEN_FPDMA_AA },
{ "ST1000LM024 HN-M101MBB", "2BA30001", ATA_HORKAGE_BROKEN_FPDMA_AA },
/* Blacklist entries taken from Silicon Image 3124/3132
Windows driver .inf file - also several Linux problem reports */
{ "HTS541060G9SA00", "MB3OC60D", ATA_HORKAGE_NONCQ, },
{ "HTS541080G9SA00", "MB4OC60D", ATA_HORKAGE_NONCQ, },
{ "HTS541010G9SA00", "MBZOC60D", ATA_HORKAGE_NONCQ, },
/* https://bugzilla.kernel.org/show_bug.cgi?id=15573 */
{ "C300-CTFDDAC128MAG", "0001", ATA_HORKAGE_NONCQ, },
/* devices which puke on READ_NATIVE_MAX */
{ "HDS724040KLSA80", "KFAOA20N", ATA_HORKAGE_BROKEN_HPA, },
{ "WDC WD3200JD-00KLB0", "WD-WCAMR1130137", ATA_HORKAGE_BROKEN_HPA },
{ "WDC WD2500JD-00HBB0", "WD-WMAL71490727", ATA_HORKAGE_BROKEN_HPA },
{ "MAXTOR 6L080L4", "A93.0500", ATA_HORKAGE_BROKEN_HPA },
/* this one allows HPA unlocking but fails IOs on the area */
{ "OCZ-VERTEX", "1.30", ATA_HORKAGE_BROKEN_HPA },
/* Devices which report 1 sector over size HPA */
{ "ST340823A", NULL, ATA_HORKAGE_HPA_SIZE, },
{ "ST320413A", NULL, ATA_HORKAGE_HPA_SIZE, },
{ "ST310211A", NULL, ATA_HORKAGE_HPA_SIZE, },
/* Devices which get the IVB wrong */
{ "QUANTUM FIREBALLlct10 05", "A03.0900", ATA_HORKAGE_IVB, },
/* Maybe we should just blacklist TSSTcorp... */
{ "TSSTcorp CDDVDW SH-S202[HJN]", "SB0[01]", ATA_HORKAGE_IVB, },
/* Devices that do not need bridging limits applied */
{ "MTRON MSP-SATA*", NULL, ATA_HORKAGE_BRIDGE_OK, },
{ "BUFFALO HD-QSU2/R5", NULL, ATA_HORKAGE_BRIDGE_OK, },
/* Devices which aren't very happy with higher link speeds */
{ "WD My Book", NULL, ATA_HORKAGE_1_5_GBPS, },
{ "Seagate FreeAgent GoFlex", NULL, ATA_HORKAGE_1_5_GBPS, },
/*
* Devices which choke on SETXFER. Applies only if both the
* device and controller are SATA.
*/
{ "PIONEER DVD-RW DVRTD08", NULL, ATA_HORKAGE_NOSETXFER },
{ "PIONEER DVD-RW DVRTD08A", NULL, ATA_HORKAGE_NOSETXFER },
{ "PIONEER DVD-RW DVR-215", NULL, ATA_HORKAGE_NOSETXFER },
{ "PIONEER DVD-RW DVR-212D", NULL, ATA_HORKAGE_NOSETXFER },
{ "PIONEER DVD-RW DVR-216D", NULL, ATA_HORKAGE_NOSETXFER },
/* devices that don't properly handle queued TRIM commands */
{ "Micron_M500*", NULL, ATA_HORKAGE_NO_NCQ_TRIM |
ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "Crucial_CT*M500*", NULL, ATA_HORKAGE_NO_NCQ_TRIM |
ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "Micron_M5[15]0*", "MU01", ATA_HORKAGE_NO_NCQ_TRIM |
ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "Crucial_CT*M550*", "MU01", ATA_HORKAGE_NO_NCQ_TRIM |
ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "Crucial_CT*MX100*", "MU01", ATA_HORKAGE_NO_NCQ_TRIM |
libata: Whitelist SSDs that are known to properly return zeroes after TRIM As defined, the DRAT (Deterministic Read After Trim) and RZAT (Return Zero After Trim) flags in the ATA Command Set are unreliable in the sense that they only define what happens if the device successfully executed the DSM TRIM command. TRIM is only advisory, however, and the device is free to silently ignore all or parts of the request. In practice this renders the DRAT and RZAT flags completely useless and because the results are unpredictable we decided to disable discard in MD for 3.18 to avoid the risk of data corruption. Hardware vendors in the real world obviously need better guarantees than what the standards bodies provide. Unfortuntely those guarantees are encoded in product requirements documents rather than somewhere we can key off of them programatically. So we are compelled to disabling discard_zeroes_data for all devices unless we explicitly have data to support whitelisting them. This patch whitelists SSDs from a few of the main vendors. None of the whitelists are based on written guarantees. They are purely based on empirical evidence collected from internal and external users that have tested or qualified these drives in RAID deployments. The whitelist is only meant as a starting point and is by no means comprehensive: - All intel SSD models except for 510 - Micron M5?0/M600 - Samsung SSDs - Seagate SSDs Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Tejun Heo <tj@kernel.org>
2015-01-08 23:34:27 +08:00
ATA_HORKAGE_ZERO_AFTER_TRIM, },
/*
* As defined, the DRAT (Deterministic Read After Trim) and RZAT
* (Return Zero After Trim) flags in the ATA Command Set are
* unreliable in the sense that they only define what happens if
* the device successfully executed the DSM TRIM command. TRIM
* is only advisory, however, and the device is free to silently
* ignore all or parts of the request.
*
* Whitelist drives that are known to reliably return zeroes
* after TRIM.
*/
/*
* The intel 510 drive has buggy DRAT/RZAT. Explicitly exclude
* that model before whitelisting all other intel SSDs.
*/
{ "INTEL*SSDSC2MH*", NULL, 0, },
{ "Micron*", NULL, ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "Crucial*", NULL, ATA_HORKAGE_ZERO_AFTER_TRIM, },
libata: Whitelist SSDs that are known to properly return zeroes after TRIM As defined, the DRAT (Deterministic Read After Trim) and RZAT (Return Zero After Trim) flags in the ATA Command Set are unreliable in the sense that they only define what happens if the device successfully executed the DSM TRIM command. TRIM is only advisory, however, and the device is free to silently ignore all or parts of the request. In practice this renders the DRAT and RZAT flags completely useless and because the results are unpredictable we decided to disable discard in MD for 3.18 to avoid the risk of data corruption. Hardware vendors in the real world obviously need better guarantees than what the standards bodies provide. Unfortuntely those guarantees are encoded in product requirements documents rather than somewhere we can key off of them programatically. So we are compelled to disabling discard_zeroes_data for all devices unless we explicitly have data to support whitelisting them. This patch whitelists SSDs from a few of the main vendors. None of the whitelists are based on written guarantees. They are purely based on empirical evidence collected from internal and external users that have tested or qualified these drives in RAID deployments. The whitelist is only meant as a starting point and is by no means comprehensive: - All intel SSD models except for 510 - Micron M5?0/M600 - Samsung SSDs - Seagate SSDs Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Tejun Heo <tj@kernel.org>
2015-01-08 23:34:27 +08:00
{ "INTEL*SSD*", NULL, ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "SSD*INTEL*", NULL, ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "Samsung*SSD*", NULL, ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "SAMSUNG*SSD*", NULL, ATA_HORKAGE_ZERO_AFTER_TRIM, },
{ "ST[1248][0248]0[FH]*", NULL, ATA_HORKAGE_ZERO_AFTER_TRIM, },
/*
* Some WD SATA-I drives spin up and down erratically when the link
* is put into the slumber mode. We don't have full list of the
* affected devices. Disable LPM if the device matches one of the
* known prefixes and is SATA-1. As a side effect LPM partial is
* lost too.
*
* https://bugzilla.kernel.org/show_bug.cgi?id=57211
*/
{ "WDC WD800JD-*", NULL, ATA_HORKAGE_WD_BROKEN_LPM },
{ "WDC WD1200JD-*", NULL, ATA_HORKAGE_WD_BROKEN_LPM },
{ "WDC WD1600JD-*", NULL, ATA_HORKAGE_WD_BROKEN_LPM },
{ "WDC WD2000JD-*", NULL, ATA_HORKAGE_WD_BROKEN_LPM },
{ "WDC WD2500JD-*", NULL, ATA_HORKAGE_WD_BROKEN_LPM },
{ "WDC WD3000JD-*", NULL, ATA_HORKAGE_WD_BROKEN_LPM },
{ "WDC WD3200JD-*", NULL, ATA_HORKAGE_WD_BROKEN_LPM },
/* End Marker */
{ }
};
static unsigned long ata_dev_blacklisted(const struct ata_device *dev)
{
unsigned char model_num[ATA_ID_PROD_LEN + 1];
unsigned char model_rev[ATA_ID_FW_REV_LEN + 1];
const struct ata_blacklist_entry *ad = ata_device_blacklist;
ata_id_c_string(dev->id, model_num, ATA_ID_PROD, sizeof(model_num));
ata_id_c_string(dev->id, model_rev, ATA_ID_FW_REV, sizeof(model_rev));
while (ad->model_num) {
if (glob_match(ad->model_num, model_num)) {
if (ad->model_rev == NULL)
return ad->horkage;
if (glob_match(ad->model_rev, model_rev))
return ad->horkage;
}
ad++;
}
return 0;
}
static int ata_dma_blacklisted(const struct ata_device *dev)
{
/* We don't support polling DMA.
* DMA blacklist those ATAPI devices with CDB-intr (and use PIO)
* if the LLDD handles only interrupts in the HSM_ST_LAST state.
*/
if ((dev->link->ap->flags & ATA_FLAG_PIO_POLLING) &&
(dev->flags & ATA_DFLAG_CDB_INTR))
return 1;
return (dev->horkage & ATA_HORKAGE_NODMA) ? 1 : 0;
}
/**
* ata_is_40wire - check drive side detection
* @dev: device
*
* Perform drive side detection decoding, allowing for device vendors
* who can't follow the documentation.
*/
static int ata_is_40wire(struct ata_device *dev)
{
if (dev->horkage & ATA_HORKAGE_IVB)
return ata_drive_40wire_relaxed(dev->id);
return ata_drive_40wire(dev->id);
}
/**
* cable_is_40wire - 40/80/SATA decider
* @ap: port to consider
*
* This function encapsulates the policy for speed management
* in one place. At the moment we don't cache the result but
* there is a good case for setting ap->cbl to the result when
* we are called with unknown cables (and figuring out if it
* impacts hotplug at all).
*
* Return 1 if the cable appears to be 40 wire.
*/
static int cable_is_40wire(struct ata_port *ap)
{
struct ata_link *link;
struct ata_device *dev;
/* If the controller thinks we are 40 wire, we are. */
if (ap->cbl == ATA_CBL_PATA40)
return 1;
/* If the controller thinks we are 80 wire, we are. */
if (ap->cbl == ATA_CBL_PATA80 || ap->cbl == ATA_CBL_SATA)
return 0;
/* If the system is known to be 40 wire short cable (eg
* laptop), then we allow 80 wire modes even if the drive
* isn't sure.
*/
if (ap->cbl == ATA_CBL_PATA40_SHORT)
return 0;
/* If the controller doesn't know, we scan.
*
* Note: We look for all 40 wire detects at this point. Any
* 80 wire detect is taken to be 80 wire cable because
* - in many setups only the one drive (slave if present) will
* give a valid detect
* - if you have a non detect capable drive you don't want it
* to colour the choice
*/
ata_for_each_link(link, ap, EDGE) {
ata_for_each_dev(dev, link, ENABLED) {
if (!ata_is_40wire(dev))
return 0;
}
}
return 1;
}
/**
* ata_dev_xfermask - Compute supported xfermask of the given device
* @dev: Device to compute xfermask for
*
* Compute supported xfermask of @dev and store it in
* dev->*_mask. This function is responsible for applying all
* known limits including host controller limits, device
* blacklist, etc...
*
* LOCKING:
* None.
*/
static void ata_dev_xfermask(struct ata_device *dev)
{
struct ata_link *link = dev->link;
struct ata_port *ap = link->ap;
struct ata_host *host = ap->host;
unsigned long xfer_mask;
/* controller modes available */
xfer_mask = ata_pack_xfermask(ap->pio_mask,
ap->mwdma_mask, ap->udma_mask);
/* drive modes available */
xfer_mask &= ata_pack_xfermask(dev->pio_mask,
dev->mwdma_mask, dev->udma_mask);
xfer_mask &= ata_id_xfermask(dev->id);
/*
* CFA Advanced TrueIDE timings are not allowed on a shared
* cable
*/
if (ata_dev_pair(dev)) {
/* No PIO5 or PIO6 */
xfer_mask &= ~(0x03 << (ATA_SHIFT_PIO + 5));
/* No MWDMA3 or MWDMA 4 */
xfer_mask &= ~(0x03 << (ATA_SHIFT_MWDMA + 3));
}
if (ata_dma_blacklisted(dev)) {
xfer_mask &= ~(ATA_MASK_MWDMA | ATA_MASK_UDMA);
ata_dev_warn(dev,
"device is on DMA blacklist, disabling DMA\n");
}
Fix simplex adapters with libata Recently I got my hands on nVidia's MCP61 PM-AM board, and it contains IDE chip configured by BIOS with only primary channel enabled. This confuses code which probes for device DMA capabilities - it gets 0x60 (happy duplex device) from primary channel BMDMA, but 0xFF (nobody here) from secondary channel BMDMA. Due to this code then believes that chip is simplex. I do not address this problem in my patch, as I'm not sure how to handle this. Probably ata_pci_init_one should have bitmap of enabled/possible interfaces instead of their count, but it looks like quite intrusive change, and maybe we do not care - for device with only one channel simplex and regular DMA engines are same. But making device simplex pointed out that support for DMA on simplex devices is currently broken - ata_dev_xfermask tests whether device is simplex and if it is whether DMA engine was assigned to this port. If not then it strips out DMA bits from device. Problem is that code which assigns DMA engine to port in ata_set_mode first detect device mode and assigns DMA engine to channel only if some DMA capable device was found. And as xfermask stripped out DMA bits, host->simplex_claimed is always NULL with current implementation. By allowing DMA either if simplex_claimed is NULL or if it points to current port DMA can be finally used - it gets assigned to first port which contains any DMA capable device. Before: pata_amd 0000:00:06.0: version 0.2.8 PCI: Setting latency timer of device 0000:00:06.0 to 64 ata5: PATA max UDMA/133 cmd 0x000101f0 ctl 0x000103f6 bmdma 0x0001f000 irq 14 ata6: PATA max UDMA/133 cmd 0x00010170 ctl 0x00010376 bmdma 0x0001f008 irq 15 scsi4 : pata_amd ata5.00: ATAPI, max UDMA/66 ata5.00: simplex DMA is claimed by other device, disabling DMA ata5.00: configured for PIO4 scsi5 : pata_amd ata6: port disabled. ignoring. ata6: reset failed, giving up scsi 4:0:0:0: CD-ROM ATAPI DVD W DH16W1P LG12 PQ: 0 ANSI: 5 After: pata_amd 0000:00:06.0: version 0.2.8 PCI: Setting latency timer of device 0000:00:06.0 to 64 ata5: PATA max UDMA/133 cmd 0x000101f0 ctl 0x000103f6 bmdma 0x0001f000 irq 14 ata6: PATA max UDMA/133 cmd 0x00010170 ctl 0x00010376 bmdma 0x0001f008 irq 15 scsi4 : pata_amd ata5.00: ATAPI, max UDMA/66 ata5.00: configured for UDMA/33 scsi5 : pata_amd ata6: port disabled. ignoring. ata6: reset failed, giving up scsi 4:0:0:0: CD-ROM ATAPI DVD W DH16W1P LG12 PQ: 0 ANSI: 5 Signed-off-by: Petr Vandrovec <petr@vandrovec.name> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-03-08 17:12:12 +08:00
if ((host->flags & ATA_HOST_SIMPLEX) &&
host->simplex_claimed && host->simplex_claimed != ap) {
xfer_mask &= ~(ATA_MASK_MWDMA | ATA_MASK_UDMA);
ata_dev_warn(dev,
"simplex DMA is claimed by other device, disabling DMA\n");
}
if (ap->flags & ATA_FLAG_NO_IORDY)
xfer_mask &= ata_pio_mask_no_iordy(dev);
if (ap->ops->mode_filter)
xfer_mask = ap->ops->mode_filter(dev, xfer_mask);
/* Apply cable rule here. Don't apply it early because when
* we handle hot plug the cable type can itself change.
* Check this last so that we know if the transfer rate was
* solely limited by the cable.
* Unknown or 80 wire cables reported host side are checked
* drive side as well. Cases where we know a 40wire cable
* is used safely for 80 are not checked here.
*/
if (xfer_mask & (0xF8 << ATA_SHIFT_UDMA))
/* UDMA/44 or higher would be available */
if (cable_is_40wire(ap)) {
ata_dev_warn(dev,
"limited to UDMA/33 due to 40-wire cable\n");
xfer_mask &= ~(0xF8 << ATA_SHIFT_UDMA);
}
ata_unpack_xfermask(xfer_mask, &dev->pio_mask,
&dev->mwdma_mask, &dev->udma_mask);
}
/**
* ata_dev_set_xfermode - Issue SET FEATURES - XFER MODE command
* @dev: Device to which command will be sent
*
* Issue SET FEATURES - XFER MODE command to device @dev
* on port @ap.
*
* LOCKING:
2005-05-31 07:49:12 +08:00
* PCI/etc. bus probe sem.
*
* RETURNS:
* 0 on success, AC_ERR_* mask otherwise.
*/
static unsigned int ata_dev_set_xfermode(struct ata_device *dev)
{
struct ata_taskfile tf;
unsigned int err_mask;
/* set up set-features taskfile */
DPRINTK("set features - xfer mode\n");
/* Some controllers and ATAPI devices show flaky interrupt
* behavior after setting xfer mode. Use polling instead.
*/
ata_tf_init(dev, &tf);
tf.command = ATA_CMD_SET_FEATURES;
tf.feature = SETFEATURES_XFER;
tf.flags |= ATA_TFLAG_ISADDR | ATA_TFLAG_DEVICE | ATA_TFLAG_POLLING;
tf.protocol = ATA_PROT_NODATA;
/* If we are using IORDY we must send the mode setting command */
if (ata_pio_need_iordy(dev))
tf.nsect = dev->xfer_mode;
/* If the device has IORDY and the controller does not - turn it off */
else if (ata_id_has_iordy(dev->id))
tf.nsect = 0x01;
else /* In the ancient relic department - skip all of this */
return 0;
err_mask = ata_exec_internal(dev, &tf, NULL, DMA_NONE, NULL, 0, 0);
DPRINTK("EXIT, err_mask=%x\n", err_mask);
return err_mask;
}
/**
* ata_dev_set_feature - Issue SET FEATURES - SATA FEATURES
* @dev: Device to which command will be sent
* @enable: Whether to enable or disable the feature
* @feature: The sector count represents the feature to set
*
* Issue SET FEATURES - SATA FEATURES command to device @dev
* on port @ap with sector count
*
* LOCKING:
* PCI/etc. bus probe sem.
*
* RETURNS:
* 0 on success, AC_ERR_* mask otherwise.
*/
unsigned int ata_dev_set_feature(struct ata_device *dev, u8 enable, u8 feature)
{
struct ata_taskfile tf;
unsigned int err_mask;
/* set up set-features taskfile */
DPRINTK("set features - SATA features\n");
ata_tf_init(dev, &tf);
tf.command = ATA_CMD_SET_FEATURES;
tf.feature = enable;
tf.flags |= ATA_TFLAG_ISADDR | ATA_TFLAG_DEVICE;
tf.protocol = ATA_PROT_NODATA;
tf.nsect = feature;
err_mask = ata_exec_internal(dev, &tf, NULL, DMA_NONE, NULL, 0, 0);
DPRINTK("EXIT, err_mask=%x\n", err_mask);
return err_mask;
}
EXPORT_SYMBOL_GPL(ata_dev_set_feature);
/**
* ata_dev_init_params - Issue INIT DEV PARAMS command
* @dev: Device to which command will be sent
* @heads: Number of heads (taskfile parameter)
* @sectors: Number of sectors (taskfile parameter)
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* 0 on success, AC_ERR_* mask otherwise.
*/
static unsigned int ata_dev_init_params(struct ata_device *dev,
u16 heads, u16 sectors)
{
struct ata_taskfile tf;
unsigned int err_mask;
/* Number of sectors per track 1-255. Number of heads 1-16 */
if (sectors < 1 || sectors > 255 || heads < 1 || heads > 16)
return AC_ERR_INVALID;
/* set up init dev params taskfile */
DPRINTK("init dev params \n");
ata_tf_init(dev, &tf);
tf.command = ATA_CMD_INIT_DEV_PARAMS;
tf.flags |= ATA_TFLAG_ISADDR | ATA_TFLAG_DEVICE;
tf.protocol = ATA_PROT_NODATA;
tf.nsect = sectors;
tf.device |= (heads - 1) & 0x0f; /* max head = num. of heads - 1 */
err_mask = ata_exec_internal(dev, &tf, NULL, DMA_NONE, NULL, 0, 0);
/* A clean abort indicates an original or just out of spec drive
and we should continue as we issue the setup based on the
drive reported working geometry */
if (err_mask == AC_ERR_DEV && (tf.feature & ATA_ABORTED))
err_mask = 0;
DPRINTK("EXIT, err_mask=%x\n", err_mask);
return err_mask;
}
/**
2005-05-31 07:49:12 +08:00
* ata_sg_clean - Unmap DMA memory associated with command
* @qc: Command containing DMA memory to be released
*
* Unmap all mapped DMA memory associated with this command.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*/
void ata_sg_clean(struct ata_queued_cmd *qc)
{
struct ata_port *ap = qc->ap;
struct scatterlist *sg = qc->sg;
int dir = qc->dma_dir;
WARN_ON_ONCE(sg == NULL);
libata: eliminate the home grown dma padding in favour of that provided by the block layer ATA requires that all DMA transfers begin and end on word boundaries. Because of this, a large amount of machinery grew up in ide to adjust scatterlists on this basis. However, as of 2.5, the block layer has a dma_alignment variable which ensures both the beginning and length of a DMA transfer are aligned on the dma_alignment boundary. Although the block layer does adjust the beginning of the transfer to ensure this happens, it doesn't actually adjust the length, it merely makes sure that space is allocated for transfers beyond the declared length. The upshot of this is that scatterlists may be padded to any size between the actual length and the length adjusted to the dma_alignment safely knowing that memory is allocated in this region. Right at the moment, SCSI takes the default dma_aligment which is on a 512 byte boundary. Note that this aligment only applies to transfers coming in from user space. However, since all kernel allocations are automatically aligned on a minimum of 32 byte boundaries, it is safe to adjust them in this manner as well. tj: * Adjusting sg after padding is done in block layer. Make libata set queue alignment correctly for ATAPI devices and drop broken sg mangling from ata_sg_setup(). * Use request->raw_data_len for ATAPI transfer chunk size. * Killed qc->raw_nbytes. * Separated out killing qc->n_iter. Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-02-19 18:36:56 +08:00
VPRINTK("unmapping %u sg elements\n", qc->n_elem);
libata: eliminate the home grown dma padding in favour of that provided by the block layer ATA requires that all DMA transfers begin and end on word boundaries. Because of this, a large amount of machinery grew up in ide to adjust scatterlists on this basis. However, as of 2.5, the block layer has a dma_alignment variable which ensures both the beginning and length of a DMA transfer are aligned on the dma_alignment boundary. Although the block layer does adjust the beginning of the transfer to ensure this happens, it doesn't actually adjust the length, it merely makes sure that space is allocated for transfers beyond the declared length. The upshot of this is that scatterlists may be padded to any size between the actual length and the length adjusted to the dma_alignment safely knowing that memory is allocated in this region. Right at the moment, SCSI takes the default dma_aligment which is on a 512 byte boundary. Note that this aligment only applies to transfers coming in from user space. However, since all kernel allocations are automatically aligned on a minimum of 32 byte boundaries, it is safe to adjust them in this manner as well. tj: * Adjusting sg after padding is done in block layer. Make libata set queue alignment correctly for ATAPI devices and drop broken sg mangling from ata_sg_setup(). * Use request->raw_data_len for ATAPI transfer chunk size. * Killed qc->raw_nbytes. * Separated out killing qc->n_iter. Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Tejun Heo <htejun@gmail.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2008-02-19 18:36:56 +08:00
if (qc->n_elem)
dma_unmap_sg(ap->dev, sg, qc->orig_n_elem, dir);
qc->flags &= ~ATA_QCFLAG_DMAMAP;
qc->sg = NULL;
}
/**
* atapi_check_dma - Check whether ATAPI DMA can be supported
* @qc: Metadata associated with taskfile to check
*
* Allow low-level driver to filter ATA PACKET commands, returning
* a status indicating whether or not it is OK to use DMA for the
* supplied PACKET command.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*
* RETURNS: 0 when ATAPI DMA can be used
* nonzero otherwise
*/
int atapi_check_dma(struct ata_queued_cmd *qc)
{
struct ata_port *ap = qc->ap;
/* Don't allow DMA if it isn't multiple of 16 bytes. Quite a
* few ATAPI devices choke on such DMA requests.
*/
if (!(qc->dev->horkage & ATA_HORKAGE_ATAPI_MOD16_DMA) &&
unlikely(qc->nbytes & 15))
return 1;
if (ap->ops->check_atapi_dma)
return ap->ops->check_atapi_dma(qc);
return 0;
}
/**
* ata_std_qc_defer - Check whether a qc needs to be deferred
* @qc: ATA command in question
*
* Non-NCQ commands cannot run with any other command, NCQ or
* not. As upper layer only knows the queue depth, we are
* responsible for maintaining exclusion. This function checks
* whether a new command @qc can be issued.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*
* RETURNS:
* ATA_DEFER_* if deferring is needed, 0 otherwise.
*/
int ata_std_qc_defer(struct ata_queued_cmd *qc)
{
struct ata_link *link = qc->dev->link;
if (qc->tf.protocol == ATA_PROT_NCQ) {
if (!ata_tag_valid(link->active_tag))
return 0;
} else {
if (!ata_tag_valid(link->active_tag) && !link->sactive)
return 0;
}
return ATA_DEFER_LINK;
}
void ata_noop_qc_prep(struct ata_queued_cmd *qc) { }
/**
* ata_sg_init - Associate command with scatter-gather table.
* @qc: Command to be associated
* @sg: Scatter-gather table.
* @n_elem: Number of elements in s/g table.
*
* Initialize the data-related elements of queued_cmd @qc
* to point to a scatter-gather table @sg, containing @n_elem
* elements.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*/
void ata_sg_init(struct ata_queued_cmd *qc, struct scatterlist *sg,
unsigned int n_elem)
{
qc->sg = sg;
qc->n_elem = n_elem;
qc->cursg = qc->sg;
}
/**
* ata_sg_setup - DMA-map the scatter-gather table associated with a command.
* @qc: Command with scatter-gather table to be mapped.
*
* DMA-map the scatter-gather table associated with queued_cmd @qc.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*
* RETURNS:
* Zero on success, negative on error.
*
*/
static int ata_sg_setup(struct ata_queued_cmd *qc)
{
struct ata_port *ap = qc->ap;
unsigned int n_elem;
VPRINTK("ENTER, ata%u\n", ap->print_id);
n_elem = dma_map_sg(ap->dev, qc->sg, qc->n_elem, qc->dma_dir);
if (n_elem < 1)
return -1;
DPRINTK("%d sg elements mapped\n", n_elem);
qc->orig_n_elem = qc->n_elem;
qc->n_elem = n_elem;
qc->flags |= ATA_QCFLAG_DMAMAP;
return 0;
}
/**
* swap_buf_le16 - swap halves of 16-bit words in place
* @buf: Buffer to swap
* @buf_words: Number of 16-bit words in buffer.
*
* Swap halves of 16-bit words if needed to convert from
* little-endian byte order to native cpu byte order, or
* vice-versa.
*
* LOCKING:
* Inherited from caller.
*/
void swap_buf_le16(u16 *buf, unsigned int buf_words)
{
#ifdef __BIG_ENDIAN
unsigned int i;
for (i = 0; i < buf_words; i++)
buf[i] = le16_to_cpu(buf[i]);
#endif /* __BIG_ENDIAN */
}
/**
* ata_qc_new_init - Request an available ATA command, and initialize it
* @dev: Device from whom we request an available command structure
*
* LOCKING:
* None.
*/
struct ata_queued_cmd *ata_qc_new_init(struct ata_device *dev, int tag)
{
struct ata_port *ap = dev->link->ap;
struct ata_queued_cmd *qc;
/* no command while frozen */
if (unlikely(ap->pflags & ATA_PFLAG_FROZEN))
return NULL;
/* libsas case */
ata: Add a new flag to destinguish sas controller SAS controller has its own tag allocation, which doesn't directly match to ATA tag, so SAS and SATA have different code path for ata tags. Originally we use port->scsi_host (98bd4be1) to destinguish SAS controller, but libsas set ->scsi_host too, so we can't use it for the destinguish, we add a new flag for this purpose. Without this patch, the following oops can happen because scsi-mq uses a host-wide tag map shared among all devices with some integer tag values >= ATA_MAX_QUEUE. These unexpectedly high tag values cause __ata_qc_from_tag() to return NULL, which is then dereferenced in ata_qc_new_init(). BUG: unable to handle kernel NULL pointer dereference at 0000000000000058 IP: [<ffffffff804fd46e>] ata_qc_new_init+0x3e/0x120 PGD 32adf0067 PUD 32adf1067 PMD 0 Oops: 0002 [#1] SMP DEBUG_PAGEALLOC Modules linked in: iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi igb i2c_algo_bit ptp pps_core pm80xx libsas scsi_transport_sas sg coretemp eeprom w83795 i2c_i801 CPU: 4 PID: 1450 Comm: cydiskbench Not tainted 4.0.0-rc3 #1 Hardware name: Supermicro X8DTH-i/6/iF/6F/X8DTH, BIOS 2.1b 05/04/12 task: ffff8800ba86d500 ti: ffff88032a064000 task.ti: ffff88032a064000 RIP: 0010:[<ffffffff804fd46e>] [<ffffffff804fd46e>] ata_qc_new_init+0x3e/0x120 RSP: 0018:ffff88032a067858 EFLAGS: 00010046 RAX: 0000000000000000 RBX: ffff8800ba0d2230 RCX: 000000000000002a RDX: ffffffff80505ae0 RSI: 0000000000000020 RDI: ffff8800ba0d2230 RBP: ffff88032a067868 R08: 0000000000000201 R09: 0000000000000001 R10: 0000000000000000 R11: 0000000000000000 R12: ffff8800ba0d0000 R13: ffff8800ba0d2230 R14: ffffffff80505ae0 R15: ffff8800ba0d0000 FS: 0000000041223950(0063) GS:ffff88033e480000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 0000000000000058 CR3: 000000032a0a3000 CR4: 00000000000006e0 Stack: ffff880329eee758 ffff880329eee758 ffff88032a0678a8 ffffffff80502dad ffff8800ba167978 ffff880329eee758 ffff88032bf9c520 ffff8800ba167978 ffff88032bf9c520 ffff88032bf9a290 ffff88032a0678b8 ffffffff80506909 Call Trace: [<ffffffff80502dad>] ata_scsi_translate+0x3d/0x1b0 [<ffffffff80506909>] ata_sas_queuecmd+0x149/0x2a0 [<ffffffffa0046650>] sas_queuecommand+0xa0/0x1f0 [libsas] [<ffffffff804ea544>] scsi_dispatch_cmd+0xd4/0x1a0 [<ffffffff804eb50f>] scsi_queue_rq+0x66f/0x7f0 [<ffffffff803e5098>] __blk_mq_run_hw_queue+0x208/0x3f0 [<ffffffff803e54b8>] blk_mq_run_hw_queue+0x88/0xc0 [<ffffffff803e5c74>] blk_mq_insert_request+0xc4/0x130 [<ffffffff803e0b63>] blk_execute_rq_nowait+0x73/0x160 [<ffffffffa0023fca>] sg_common_write+0x3da/0x720 [sg] [<ffffffffa0025100>] sg_new_write+0x250/0x360 [sg] [<ffffffffa0025feb>] sg_write+0x13b/0x450 [sg] [<ffffffff8032ec91>] vfs_write+0xd1/0x1b0 [<ffffffff8032ee54>] SyS_write+0x54/0xc0 [<ffffffff80689932>] system_call_fastpath+0x12/0x17 tj: updated description. Fixes: 12cb5ce101ab ("libata: use blk taging") Reported-and-tested-by: Tony Battersby <tonyb@cybernetics.com> Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2015-03-13 01:32:18 +08:00
if (ap->flags & ATA_FLAG_SAS_HOST) {
tag = ata_sas_allocate_tag(ap);
if (tag < 0)
return NULL;
libata/ahci: accommodate tag ordered controllers The AHCI spec allows implementations to issue commands in tag order rather than FIFO order: 5.3.2.12 P:SelectCmd HBA sets pSlotLoc = (pSlotLoc + 1) mod (CAP.NCS + 1) or HBA selects the command to issue that has had the PxCI bit set to '1' longer than any other command pending to be issued. The result is that commands posted sequentially (time-wise) may play out of sequence when issued by hardware. This behavior has likely been hidden by drives that arrange for commands to complete in issue order. However, it appears recent drives (two from different vendors that we have found so far) inflict out-of-order completions as a matter of course. So, we need to take care to maintain ordered submission, otherwise we risk triggering a drive to fall out of sequential-io automation and back to random-io processing, which incurs large latency and degrades throughput. This issue was found in simple benchmarks where QD=2 seq-write performance was 30-50% *greater* than QD=32 seq-write performance. Tagging for -stable and making the change globally since it has a low risk-to-reward ratio. Also, word is that recent versions of an unnamed OS also does it this way now. So, drives in the field are already experienced with this tag ordering scheme. Cc: <stable@vger.kernel.org> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Ed Ciechanowski <ed.ciechanowski@intel.com> Reviewed-by: Matthew Wilcox <matthew.r.wilcox@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-04-18 02:48:21 +08:00
}
qc = __ata_qc_from_tag(ap, tag);
qc->tag = tag;
qc->scsicmd = NULL;
qc->ap = ap;
qc->dev = dev;
ata_qc_reinit(qc);
return qc;
}
/**
* ata_qc_free - free unused ata_queued_cmd
* @qc: Command to complete
*
* Designed to free unused ata_queued_cmd object
* in case something prevents using it.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*/
void ata_qc_free(struct ata_queued_cmd *qc)
{
struct ata_port *ap;
unsigned int tag;
WARN_ON_ONCE(qc == NULL); /* ata_qc_from_tag _might_ return NULL */
ap = qc->ap;
qc->flags = 0;
tag = qc->tag;
if (likely(ata_tag_valid(tag))) {
qc->tag = ATA_TAG_POISON;
ata: Add a new flag to destinguish sas controller SAS controller has its own tag allocation, which doesn't directly match to ATA tag, so SAS and SATA have different code path for ata tags. Originally we use port->scsi_host (98bd4be1) to destinguish SAS controller, but libsas set ->scsi_host too, so we can't use it for the destinguish, we add a new flag for this purpose. Without this patch, the following oops can happen because scsi-mq uses a host-wide tag map shared among all devices with some integer tag values >= ATA_MAX_QUEUE. These unexpectedly high tag values cause __ata_qc_from_tag() to return NULL, which is then dereferenced in ata_qc_new_init(). BUG: unable to handle kernel NULL pointer dereference at 0000000000000058 IP: [<ffffffff804fd46e>] ata_qc_new_init+0x3e/0x120 PGD 32adf0067 PUD 32adf1067 PMD 0 Oops: 0002 [#1] SMP DEBUG_PAGEALLOC Modules linked in: iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi igb i2c_algo_bit ptp pps_core pm80xx libsas scsi_transport_sas sg coretemp eeprom w83795 i2c_i801 CPU: 4 PID: 1450 Comm: cydiskbench Not tainted 4.0.0-rc3 #1 Hardware name: Supermicro X8DTH-i/6/iF/6F/X8DTH, BIOS 2.1b 05/04/12 task: ffff8800ba86d500 ti: ffff88032a064000 task.ti: ffff88032a064000 RIP: 0010:[<ffffffff804fd46e>] [<ffffffff804fd46e>] ata_qc_new_init+0x3e/0x120 RSP: 0018:ffff88032a067858 EFLAGS: 00010046 RAX: 0000000000000000 RBX: ffff8800ba0d2230 RCX: 000000000000002a RDX: ffffffff80505ae0 RSI: 0000000000000020 RDI: ffff8800ba0d2230 RBP: ffff88032a067868 R08: 0000000000000201 R09: 0000000000000001 R10: 0000000000000000 R11: 0000000000000000 R12: ffff8800ba0d0000 R13: ffff8800ba0d2230 R14: ffffffff80505ae0 R15: ffff8800ba0d0000 FS: 0000000041223950(0063) GS:ffff88033e480000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 0000000000000058 CR3: 000000032a0a3000 CR4: 00000000000006e0 Stack: ffff880329eee758 ffff880329eee758 ffff88032a0678a8 ffffffff80502dad ffff8800ba167978 ffff880329eee758 ffff88032bf9c520 ffff8800ba167978 ffff88032bf9c520 ffff88032bf9a290 ffff88032a0678b8 ffffffff80506909 Call Trace: [<ffffffff80502dad>] ata_scsi_translate+0x3d/0x1b0 [<ffffffff80506909>] ata_sas_queuecmd+0x149/0x2a0 [<ffffffffa0046650>] sas_queuecommand+0xa0/0x1f0 [libsas] [<ffffffff804ea544>] scsi_dispatch_cmd+0xd4/0x1a0 [<ffffffff804eb50f>] scsi_queue_rq+0x66f/0x7f0 [<ffffffff803e5098>] __blk_mq_run_hw_queue+0x208/0x3f0 [<ffffffff803e54b8>] blk_mq_run_hw_queue+0x88/0xc0 [<ffffffff803e5c74>] blk_mq_insert_request+0xc4/0x130 [<ffffffff803e0b63>] blk_execute_rq_nowait+0x73/0x160 [<ffffffffa0023fca>] sg_common_write+0x3da/0x720 [sg] [<ffffffffa0025100>] sg_new_write+0x250/0x360 [sg] [<ffffffffa0025feb>] sg_write+0x13b/0x450 [sg] [<ffffffff8032ec91>] vfs_write+0xd1/0x1b0 [<ffffffff8032ee54>] SyS_write+0x54/0xc0 [<ffffffff80689932>] system_call_fastpath+0x12/0x17 tj: updated description. Fixes: 12cb5ce101ab ("libata: use blk taging") Reported-and-tested-by: Tony Battersby <tonyb@cybernetics.com> Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2015-03-13 01:32:18 +08:00
if (ap->flags & ATA_FLAG_SAS_HOST)
ata_sas_free_tag(tag, ap);
}
}
void __ata_qc_complete(struct ata_queued_cmd *qc)
{
struct ata_port *ap;
struct ata_link *link;
WARN_ON_ONCE(qc == NULL); /* ata_qc_from_tag _might_ return NULL */
WARN_ON_ONCE(!(qc->flags & ATA_QCFLAG_ACTIVE));
ap = qc->ap;
link = qc->dev->link;
if (likely(qc->flags & ATA_QCFLAG_DMAMAP))
ata_sg_clean(qc);
/* command should be marked inactive atomically with qc completion */
if (qc->tf.protocol == ATA_PROT_NCQ) {
link->sactive &= ~(1 << qc->tag);
if (!link->sactive)
ap->nr_active_links--;
} else {
link->active_tag = ATA_TAG_POISON;
ap->nr_active_links--;
}
/* clear exclusive status */
if (unlikely(qc->flags & ATA_QCFLAG_CLEAR_EXCL &&
ap->excl_link == link))
ap->excl_link = NULL;
/* atapi: mark qc as inactive to prevent the interrupt handler
* from completing the command twice later, before the error handler
* is called. (when rc != 0 and atapi request sense is needed)
*/
qc->flags &= ~ATA_QCFLAG_ACTIVE;
ap->qc_active &= ~(1 << qc->tag);
/* call completion callback */
qc->complete_fn(qc);
}
static void fill_result_tf(struct ata_queued_cmd *qc)
{
struct ata_port *ap = qc->ap;
qc->result_tf.flags = qc->tf.flags;
ap->ops->qc_fill_rtf(qc);
}
static void ata_verify_xfer(struct ata_queued_cmd *qc)
{
struct ata_device *dev = qc->dev;
if (ata_is_nodata(qc->tf.protocol))
return;
if ((dev->mwdma_mask || dev->udma_mask) && ata_is_pio(qc->tf.protocol))
return;
dev->flags &= ~ATA_DFLAG_DUBIOUS_XFER;
}
/**
* ata_qc_complete - Complete an active ATA command
* @qc: Command to complete
*
libata: always use ata_qc_complete_multiple() for NCQ command completions Currently, sata_fsl, mv and nv call ata_qc_complete() multiple times from their interrupt handlers to indicate completion of NCQ commands. This limits the visibility the libata core layer has into how commands are being executed and completed, which is necessary to support IRQ expecting in generic way. libata already has an interface to complete multiple commands at once - ata_qc_complete_multiple() which ahci and sata_sil24 already use. This patch updates the three drivers to use ata_qc_complete_multiple() too and updates comments on ata_qc_complete[_multiple]() regarding their usages with NCQ completions. This change not only provides better visibility into command execution to the core layer but also simplifies low level drivers. * sata_fsl: It already builds done_mask. Conversion is straight forward. * sata_mv: mv_process_crpb_response() no longer checks for illegal completions, it just returns whether the tag is completed or not. mv_process_crpb_entries() builds done_mask from it and passes it to ata_qc_complete_multiple() which will check for illegal completions. * sata_nv adma: Similar to sata_mv. nv_adma_check_cpb() now just returns the tag status and nv_adma_interrupt() builds done_mask from it and passes it to ata_qc_complete_multiple(). * sata_nv swncq: It already builds done_mask. Drop unnecessary illegal transition checks and call ata_qc_complete_multiple(). In the long run, it might be a good idea to make ata_qc_complete() whine if called when multiple NCQ commands are in flight. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Ashish Kalra <ashish.kalra@freescale.com> Cc: Saeed Bishara <saeed@marvell.com> Cc: Mark Lord <liml@rtr.ca> Cc: Robert Hancock <hancockr@shaw.ca> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-06-25 21:03:34 +08:00
* Indicate to the mid and upper layers that an ATA command has
* completed, with either an ok or not-ok status.
*
* Refrain from calling this function multiple times when
* successfully completing multiple NCQ commands.
* ata_qc_complete_multiple() should be used instead, which will
* properly update IRQ expect state.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*/
void ata_qc_complete(struct ata_queued_cmd *qc)
{
struct ata_port *ap = qc->ap;
/* XXX: New EH and old EH use different mechanisms to
* synchronize EH with regular execution path.
*
* In new EH, a failed qc is marked with ATA_QCFLAG_FAILED.
* Normal execution path is responsible for not accessing a
* failed qc. libata core enforces the rule by returning NULL
* from ata_qc_from_tag() for failed qcs.
*
* Old EH depends on ata_qc_complete() nullifying completion
* requests if ATA_QCFLAG_EH_SCHEDULED is set. Old EH does
* not synchronize with interrupt handler. Only PIO task is
* taken care of.
*/
if (ap->ops->error_handler) {
struct ata_device *dev = qc->dev;
struct ata_eh_info *ehi = &dev->link->eh_info;
if (unlikely(qc->err_mask))
qc->flags |= ATA_QCFLAG_FAILED;
/*
* Finish internal commands without any further processing
* and always with the result TF filled.
*/
if (unlikely(ata_tag_internal(qc->tag))) {
fill_result_tf(qc);
__ata_qc_complete(qc);
return;
}
/*
* Non-internal qc has failed. Fill the result TF and
* summon EH.
*/
if (unlikely(qc->flags & ATA_QCFLAG_FAILED)) {
fill_result_tf(qc);
ata_qc_schedule_eh(qc);
return;
}
WARN_ON_ONCE(ap->pflags & ATA_PFLAG_FROZEN);
/* read result TF if requested */
if (qc->flags & ATA_QCFLAG_RESULT_TF)
fill_result_tf(qc);
/* Some commands need post-processing after successful
* completion.
*/
switch (qc->tf.command) {
case ATA_CMD_SET_FEATURES:
if (qc->tf.feature != SETFEATURES_WC_ON &&
qc->tf.feature != SETFEATURES_WC_OFF)
break;
/* fall through */
case ATA_CMD_INIT_DEV_PARAMS: /* CHS translation changed */
case ATA_CMD_SET_MULTI: /* multi_count changed */
/* revalidate device */
ehi->dev_action[dev->devno] |= ATA_EH_REVALIDATE;
ata_port_schedule_eh(ap);
break;
case ATA_CMD_SLEEP:
dev->flags |= ATA_DFLAG_SLEEPING;
break;
}
if (unlikely(dev->flags & ATA_DFLAG_DUBIOUS_XFER))
ata_verify_xfer(qc);
__ata_qc_complete(qc);
} else {
if (qc->flags & ATA_QCFLAG_EH_SCHEDULED)
return;
/* read result TF if failed or requested */
if (qc->err_mask || qc->flags & ATA_QCFLAG_RESULT_TF)
fill_result_tf(qc);
__ata_qc_complete(qc);
}
}
/**
* ata_qc_complete_multiple - Complete multiple qcs successfully
* @ap: port in question
* @qc_active: new qc_active mask
*
* Complete in-flight commands. This functions is meant to be
* called from low-level driver's interrupt routine to complete
* requests normally. ap->qc_active and @qc_active is compared
* and commands are completed accordingly.
*
libata: always use ata_qc_complete_multiple() for NCQ command completions Currently, sata_fsl, mv and nv call ata_qc_complete() multiple times from their interrupt handlers to indicate completion of NCQ commands. This limits the visibility the libata core layer has into how commands are being executed and completed, which is necessary to support IRQ expecting in generic way. libata already has an interface to complete multiple commands at once - ata_qc_complete_multiple() which ahci and sata_sil24 already use. This patch updates the three drivers to use ata_qc_complete_multiple() too and updates comments on ata_qc_complete[_multiple]() regarding their usages with NCQ completions. This change not only provides better visibility into command execution to the core layer but also simplifies low level drivers. * sata_fsl: It already builds done_mask. Conversion is straight forward. * sata_mv: mv_process_crpb_response() no longer checks for illegal completions, it just returns whether the tag is completed or not. mv_process_crpb_entries() builds done_mask from it and passes it to ata_qc_complete_multiple() which will check for illegal completions. * sata_nv adma: Similar to sata_mv. nv_adma_check_cpb() now just returns the tag status and nv_adma_interrupt() builds done_mask from it and passes it to ata_qc_complete_multiple(). * sata_nv swncq: It already builds done_mask. Drop unnecessary illegal transition checks and call ata_qc_complete_multiple(). In the long run, it might be a good idea to make ata_qc_complete() whine if called when multiple NCQ commands are in flight. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Ashish Kalra <ashish.kalra@freescale.com> Cc: Saeed Bishara <saeed@marvell.com> Cc: Mark Lord <liml@rtr.ca> Cc: Robert Hancock <hancockr@shaw.ca> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-06-25 21:03:34 +08:00
* Always use this function when completing multiple NCQ commands
* from IRQ handlers instead of calling ata_qc_complete()
* multiple times to keep IRQ expect status properly in sync.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*
* RETURNS:
* Number of completed commands on success, -errno otherwise.
*/
int ata_qc_complete_multiple(struct ata_port *ap, u32 qc_active)
{
int nr_done = 0;
u32 done_mask;
done_mask = ap->qc_active ^ qc_active;
if (unlikely(done_mask & qc_active)) {
ata_port_err(ap, "illegal qc_active transition (%08x->%08x)\n",
ap->qc_active, qc_active);
return -EINVAL;
}
while (done_mask) {
struct ata_queued_cmd *qc;
unsigned int tag = __ffs(done_mask);
qc = ata_qc_from_tag(ap, tag);
if (qc) {
ata_qc_complete(qc);
nr_done++;
}
done_mask &= ~(1 << tag);
}
return nr_done;
}
/**
* ata_qc_issue - issue taskfile to device
* @qc: command to issue to device
*
* Prepare an ATA command to submission to device.
* This includes mapping the data into a DMA-able
* area, filling in the S/G table, and finally
* writing the taskfile to hardware, starting the command.
*
* LOCKING:
* spin_lock_irqsave(host lock)
*/
void ata_qc_issue(struct ata_queued_cmd *qc)
{
struct ata_port *ap = qc->ap;
struct ata_link *link = qc->dev->link;
u8 prot = qc->tf.protocol;
/* Make sure only one non-NCQ command is outstanding. The
* check is skipped for old EH because it reuses active qc to
* request ATAPI sense.
*/
WARN_ON_ONCE(ap->ops->error_handler && ata_tag_valid(link->active_tag));
if (ata_is_ncq(prot)) {
WARN_ON_ONCE(link->sactive & (1 << qc->tag));
if (!link->sactive)
ap->nr_active_links++;
link->sactive |= 1 << qc->tag;
} else {
WARN_ON_ONCE(link->sactive);
ap->nr_active_links++;
link->active_tag = qc->tag;
}
qc->flags |= ATA_QCFLAG_ACTIVE;
ap->qc_active |= 1 << qc->tag;
/*
* We guarantee to LLDs that they will have at least one
* non-zero sg if the command is a data command.
*/
if (WARN_ON_ONCE(ata_is_data(prot) &&
(!qc->sg || !qc->n_elem || !qc->nbytes)))
goto sys_err;
if (ata_is_dma(prot) || (ata_is_pio(prot) &&
(ap->flags & ATA_FLAG_PIO_DMA)))
if (ata_sg_setup(qc))
goto sys_err;
libata: prefer hardreset When both soft and hard resets are available, libata preferred softreset till now. The logic behind it was to be softer to devices; however, this doesn't really help much. Rationales for the change: * BIOS may freeze lock certain things during boot and softreset can't unlock those. This by itself is okay but during operation PHY event or other error conditions can trigger hardreset and the device may end up with different configuration. For example, after a hardreset, previously unlockable HPA can be unlocked resulting in different device size and thus revalidation failure. Similar condition can occur during or after resume. * Certain ATAPI devices require hardreset to recover after certain error conditions. On PATA, this is done by issuing the DEVICE RESET command. On SATA, COMRESET has equivalent effect. The problem is that DEVICE RESET needs its own execution protocol. For SFF controllers with bare TF access, it can be easily implemented but more advanced controllers (e.g. ahci and sata_sil24) require specialized implementations. Simply using hardreset solves the problem nicely. * COMRESET initialization sequence is the norm in SATA land and many SATA devices don't work properly if only SRST is used. For example, some PMPs behave this way and libata works around by always issuing hardreset if the host supports PMP. Like the above example, libata has developed a number of mechanisms aiming to promote softreset to hardreset if softreset is not going to work. This approach is time consuming and error prone. Also, note that, dependingon how you read the specs, it could be argued that PMP fan-out ports require COMRESET to start operation. In fact, all the PMPs on the market except one don't work properly if COMRESET is not issued to fan-out ports after PMP reset. * COMRESET is an integral part of SATA connection and any working device should be able to handle COMRESET properly. After all, it's the way to signal hardreset during reboot. This is the most used and recommended (at least by the ahci spec) method of resetting devices. So, this patch makes libata prefer hardreset over softreset by making the following changes. * Rename ATA_EH_RESET_MASK to ATA_EH_RESET and use it whereever ATA_EH_{SOFT|HARD}RESET used to be used. ATA_EH_{SOFT|HARD}RESET is now only used to tell prereset whether soft or hard reset will be issued. * Strip out now unneeded promote-to-hardreset logics from ata_eh_reset(), ata_std_prereset(), sata_pmp_std_prereset() and other places. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-01-23 23:05:14 +08:00
/* if device is sleeping, schedule reset and abort the link */
if (unlikely(qc->dev->flags & ATA_DFLAG_SLEEPING)) {
libata: prefer hardreset When both soft and hard resets are available, libata preferred softreset till now. The logic behind it was to be softer to devices; however, this doesn't really help much. Rationales for the change: * BIOS may freeze lock certain things during boot and softreset can't unlock those. This by itself is okay but during operation PHY event or other error conditions can trigger hardreset and the device may end up with different configuration. For example, after a hardreset, previously unlockable HPA can be unlocked resulting in different device size and thus revalidation failure. Similar condition can occur during or after resume. * Certain ATAPI devices require hardreset to recover after certain error conditions. On PATA, this is done by issuing the DEVICE RESET command. On SATA, COMRESET has equivalent effect. The problem is that DEVICE RESET needs its own execution protocol. For SFF controllers with bare TF access, it can be easily implemented but more advanced controllers (e.g. ahci and sata_sil24) require specialized implementations. Simply using hardreset solves the problem nicely. * COMRESET initialization sequence is the norm in SATA land and many SATA devices don't work properly if only SRST is used. For example, some PMPs behave this way and libata works around by always issuing hardreset if the host supports PMP. Like the above example, libata has developed a number of mechanisms aiming to promote softreset to hardreset if softreset is not going to work. This approach is time consuming and error prone. Also, note that, dependingon how you read the specs, it could be argued that PMP fan-out ports require COMRESET to start operation. In fact, all the PMPs on the market except one don't work properly if COMRESET is not issued to fan-out ports after PMP reset. * COMRESET is an integral part of SATA connection and any working device should be able to handle COMRESET properly. After all, it's the way to signal hardreset during reboot. This is the most used and recommended (at least by the ahci spec) method of resetting devices. So, this patch makes libata prefer hardreset over softreset by making the following changes. * Rename ATA_EH_RESET_MASK to ATA_EH_RESET and use it whereever ATA_EH_{SOFT|HARD}RESET used to be used. ATA_EH_{SOFT|HARD}RESET is now only used to tell prereset whether soft or hard reset will be issued. * Strip out now unneeded promote-to-hardreset logics from ata_eh_reset(), ata_std_prereset(), sata_pmp_std_prereset() and other places. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-01-23 23:05:14 +08:00
link->eh_info.action |= ATA_EH_RESET;
ata_ehi_push_desc(&link->eh_info, "waking up from sleep");
ata_link_abort(link);
return;
}
ap->ops->qc_prep(qc);
qc->err_mask |= ap->ops->qc_issue(qc);
if (unlikely(qc->err_mask))
goto err;
return;
sys_err:
qc->err_mask |= AC_ERR_SYSTEM;
err:
ata_qc_complete(qc);
}
/**
* sata_scr_valid - test whether SCRs are accessible
* @link: ATA link to test SCR accessibility for
*
* Test whether SCRs are accessible for @link.
*
* LOCKING:
* None.
*
* RETURNS:
* 1 if SCRs are accessible, 0 otherwise.
*/
int sata_scr_valid(struct ata_link *link)
{
struct ata_port *ap = link->ap;
return (ap->flags & ATA_FLAG_SATA) && ap->ops->scr_read;
}
/**
* sata_scr_read - read SCR register of the specified port
* @link: ATA link to read SCR for
* @reg: SCR to read
* @val: Place to store read value
*
* Read SCR register @reg of @link into *@val. This function is
* guaranteed to succeed if @link is ap->link, the cable type of
* the port is SATA and the port implements ->scr_read.
*
* LOCKING:
* None if @link is ap->link. Kernel thread context otherwise.
*
* RETURNS:
* 0 on success, negative errno on failure.
*/
int sata_scr_read(struct ata_link *link, int reg, u32 *val)
{
if (ata_is_host_link(link)) {
if (sata_scr_valid(link))
return link->ap->ops->scr_read(link, reg, val);
return -EOPNOTSUPP;
}
return sata_pmp_scr_read(link, reg, val);
}
/**
* sata_scr_write - write SCR register of the specified port
* @link: ATA link to write SCR for
* @reg: SCR to write
* @val: value to write
*
* Write @val to SCR register @reg of @link. This function is
* guaranteed to succeed if @link is ap->link, the cable type of
* the port is SATA and the port implements ->scr_read.
*
* LOCKING:
* None if @link is ap->link. Kernel thread context otherwise.
*
* RETURNS:
* 0 on success, negative errno on failure.
*/
int sata_scr_write(struct ata_link *link, int reg, u32 val)
{
if (ata_is_host_link(link)) {
if (sata_scr_valid(link))
return link->ap->ops->scr_write(link, reg, val);
return -EOPNOTSUPP;
}
return sata_pmp_scr_write(link, reg, val);
}
/**
* sata_scr_write_flush - write SCR register of the specified port and flush
* @link: ATA link to write SCR for
* @reg: SCR to write
* @val: value to write
*
* This function is identical to sata_scr_write() except that this
* function performs flush after writing to the register.
*
* LOCKING:
* None if @link is ap->link. Kernel thread context otherwise.
*
* RETURNS:
* 0 on success, negative errno on failure.
*/
int sata_scr_write_flush(struct ata_link *link, int reg, u32 val)
{
if (ata_is_host_link(link)) {
int rc;
if (sata_scr_valid(link)) {
rc = link->ap->ops->scr_write(link, reg, val);
if (rc == 0)
rc = link->ap->ops->scr_read(link, reg, &val);
return rc;
}
return -EOPNOTSUPP;
}
return sata_pmp_scr_write(link, reg, val);
}
/**
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
* ata_phys_link_online - test whether the given link is online
* @link: ATA link to test
*
* Test whether @link is online. Note that this function returns
* 0 if online status of @link cannot be obtained, so
* ata_link_online(link) != !ata_link_offline(link).
*
* LOCKING:
* None.
*
* RETURNS:
* True if the port online status is available and online.
*/
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
bool ata_phys_link_online(struct ata_link *link)
{
u32 sstatus;
if (sata_scr_read(link, SCR_STATUS, &sstatus) == 0 &&
ata_sstatus_online(sstatus))
return true;
return false;
}
/**
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
* ata_phys_link_offline - test whether the given link is offline
* @link: ATA link to test
*
* Test whether @link is offline. Note that this function
* returns 0 if offline status of @link cannot be obtained, so
* ata_link_online(link) != !ata_link_offline(link).
*
* LOCKING:
* None.
*
* RETURNS:
* True if the port offline status is available and offline.
*/
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
bool ata_phys_link_offline(struct ata_link *link)
{
u32 sstatus;
if (sata_scr_read(link, SCR_STATUS, &sstatus) == 0 &&
!ata_sstatus_online(sstatus))
return true;
return false;
}
2005-06-03 06:17:13 +08:00
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
/**
* ata_link_online - test whether the given link is online
* @link: ATA link to test
*
* Test whether @link is online. This is identical to
* ata_phys_link_online() when there's no slave link. When
* there's a slave link, this function should only be called on
* the master link and will return true if any of M/S links is
* online.
*
* LOCKING:
* None.
*
* RETURNS:
* True if the port online status is available and online.
*/
bool ata_link_online(struct ata_link *link)
{
struct ata_link *slave = link->ap->slave_link;
WARN_ON(link == slave); /* shouldn't be called on slave link */
return ata_phys_link_online(link) ||
(slave && ata_phys_link_online(slave));
}
/**
* ata_link_offline - test whether the given link is offline
* @link: ATA link to test
*
* Test whether @link is offline. This is identical to
* ata_phys_link_offline() when there's no slave link. When
* there's a slave link, this function should only be called on
* the master link and will return true if both M/S links are
* offline.
*
* LOCKING:
* None.
*
* RETURNS:
* True if the port offline status is available and offline.
*/
bool ata_link_offline(struct ata_link *link)
{
struct ata_link *slave = link->ap->slave_link;
WARN_ON(link == slave); /* shouldn't be called on slave link */
return ata_phys_link_offline(link) &&
(!slave || ata_phys_link_offline(slave));
}
#ifdef CONFIG_PM
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static void ata_port_request_pm(struct ata_port *ap, pm_message_t mesg,
unsigned int action, unsigned int ehi_flags,
bool async)
{
struct ata_link *link;
unsigned long flags;
/* Previous resume operation might still be in
* progress. Wait for PM_PENDING to clear.
*/
if (ap->pflags & ATA_PFLAG_PM_PENDING) {
ata_port_wait_eh(ap);
WARN_ON(ap->pflags & ATA_PFLAG_PM_PENDING);
}
/* request PM ops to EH */
spin_lock_irqsave(ap->lock, flags);
ap->pm_mesg = mesg;
ap->pflags |= ATA_PFLAG_PM_PENDING;
ata_for_each_link(link, ap, HOST_FIRST) {
link->eh_info.action |= action;
link->eh_info.flags |= ehi_flags;
}
ata_port_schedule_eh(ap);
spin_unlock_irqrestore(ap->lock, flags);
if (!async) {
ata_port_wait_eh(ap);
WARN_ON(ap->pflags & ATA_PFLAG_PM_PENDING);
}
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
/*
* On some hardware, device fails to respond after spun down for suspend. As
* the device won't be used before being resumed, we don't need to touch the
* device. Ask EH to skip the usual stuff and proceed directly to suspend.
*
* http://thread.gmane.org/gmane.linux.ide/46764
*/
static const unsigned int ata_port_suspend_ehi = ATA_EHI_QUIET
| ATA_EHI_NO_AUTOPSY
| ATA_EHI_NO_RECOVERY;
static void ata_port_suspend(struct ata_port *ap, pm_message_t mesg)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_request_pm(ap, mesg, 0, ata_port_suspend_ehi, false);
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static void ata_port_suspend_async(struct ata_port *ap, pm_message_t mesg)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_request_pm(ap, mesg, 0, ata_port_suspend_ehi, true);
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static int ata_port_pm_suspend(struct device *dev)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
struct ata_port *ap = to_ata_port(dev);
if (pm_runtime_suspended(dev))
return 0;
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_suspend(ap, PMSG_SUSPEND);
return 0;
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static int ata_port_pm_freeze(struct device *dev)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
struct ata_port *ap = to_ata_port(dev);
if (pm_runtime_suspended(dev))
return 0;
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_suspend(ap, PMSG_FREEZE);
return 0;
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static int ata_port_pm_poweroff(struct device *dev)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_suspend(to_ata_port(dev), PMSG_HIBERNATE);
return 0;
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static const unsigned int ata_port_resume_ehi = ATA_EHI_NO_AUTOPSY
| ATA_EHI_QUIET;
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static void ata_port_resume(struct ata_port *ap, pm_message_t mesg)
{
ata_port_request_pm(ap, mesg, ATA_EH_RESET, ata_port_resume_ehi, false);
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static void ata_port_resume_async(struct ata_port *ap, pm_message_t mesg)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_request_pm(ap, mesg, ATA_EH_RESET, ata_port_resume_ehi, true);
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
static int ata_port_pm_resume(struct device *dev)
{
ata_port_resume_async(to_ata_port(dev), PMSG_RESUME);
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
pm_runtime_disable(dev);
pm_runtime_set_active(dev);
pm_runtime_enable(dev);
return 0;
}
/*
* For ODDs, the upper layer will poll for media change every few seconds,
* which will make it enter and leave suspend state every few seconds. And
* as each suspend will cause a hard/soft reset, the gain of runtime suspend
* is very little and the ODD may malfunction after constantly being reset.
* So the idle callback here will not proceed to suspend if a non-ZPODD capable
* ODD is attached to the port.
*/
static int ata_port_runtime_idle(struct device *dev)
{
struct ata_port *ap = to_ata_port(dev);
struct ata_link *link;
struct ata_device *adev;
ata_for_each_link(link, ap, HOST_FIRST) {
ata_for_each_dev(adev, link, ENABLED)
if (adev->class == ATA_DEV_ATAPI &&
!zpodd_dev_enabled(adev))
return -EBUSY;
}
return 0;
}
static int ata_port_runtime_suspend(struct device *dev)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_suspend(to_ata_port(dev), PMSG_AUTO_SUSPEND);
return 0;
}
static int ata_port_runtime_resume(struct device *dev)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_resume(to_ata_port(dev), PMSG_AUTO_RESUME);
return 0;
}
static const struct dev_pm_ops ata_port_pm_ops = {
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
.suspend = ata_port_pm_suspend,
.resume = ata_port_pm_resume,
.freeze = ata_port_pm_freeze,
.thaw = ata_port_pm_resume,
.poweroff = ata_port_pm_poweroff,
.restore = ata_port_pm_resume,
.runtime_suspend = ata_port_runtime_suspend,
.runtime_resume = ata_port_runtime_resume,
.runtime_idle = ata_port_runtime_idle,
};
/* sas ports don't participate in pm runtime management of ata_ports,
* and need to resume ata devices at the domain level, not the per-port
* level. sas suspend/resume is async to allow parallel port recovery
* since sas has multiple ata_port instances per Scsi_Host.
*/
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
void ata_sas_port_suspend(struct ata_port *ap)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_suspend_async(ap, PMSG_SUSPEND);
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
EXPORT_SYMBOL_GPL(ata_sas_port_suspend);
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
void ata_sas_port_resume(struct ata_port *ap)
{
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
ata_port_resume_async(ap, PMSG_RESUME);
}
libata, libsas: kill pm_result and related cleanup Tejun says: "At least for libata, worrying about suspend/resume failures don't make whole lot of sense. If suspend failed, just proceed with suspend. If the device can't be woken up afterwards, that's that. There isn't anything we could have done differently anyway. The same for resume, if spinup fails, the device is dud and the following commands will invoke EH actions and will eventually fail. Again, there really isn't any *choice* to make. Just making sure the errors are handled gracefully (ie. don't crash) and the following commands are handled correctly should be enough." The only libata user that actually cares about the result from a suspend operation is libsas. However, it only cares about whether queuing a new operation collides with an in-flight one. All libsas does with the error is retry, but we can just let libata wait for the previous operation before continuing. Other cleanups include: 1/ Unifying all ata port pm operations on an ata_port_pm_ prefix 2/ Marking all ata port pm helper routines as returning void, only ata_port_pm_ entry points need to fake a 0 return value. 3/ Killing ata_port_{suspend|resume}_common() in favor of calling ata_port_request_pm() directly 4/ Killing the wrappers that just do a to_ata_port() conversion 5/ Clearly marking the entry points that do async operations with an _async suffix. Reference: http://marc.info/?l=linux-scsi&m=138995409532286&w=2 Cc: Phillip Susi <psusi@ubuntu.com> Cc: Alan Stern <stern@rowland.harvard.edu> Suggested-by: Tejun Heo <tj@kernel.org> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2014-03-15 04:52:48 +08:00
EXPORT_SYMBOL_GPL(ata_sas_port_resume);
/**
* ata_host_suspend - suspend host
* @host: host to suspend
* @mesg: PM message
*
* Suspend @host. Actual operation is performed by port suspend.
*/
int ata_host_suspend(struct ata_host *host, pm_message_t mesg)
{
host->dev->power.power_state = mesg;
return 0;
}
/**
* ata_host_resume - resume host
* @host: host to resume
*
* Resume @host. Actual operation is performed by port resume.
*/
void ata_host_resume(struct ata_host *host)
{
host->dev->power.power_state = PMSG_ON;
}
#endif
struct device_type ata_port_type = {
.name = "ata_port",
#ifdef CONFIG_PM
.pm = &ata_port_pm_ops,
#endif
};
/**
* ata_dev_init - Initialize an ata_device structure
* @dev: Device structure to initialize
*
* Initialize @dev in preparation for probing.
*
* LOCKING:
* Inherited from caller.
*/
void ata_dev_init(struct ata_device *dev)
{
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
struct ata_link *link = ata_dev_phys_link(dev);
struct ata_port *ap = link->ap;
unsigned long flags;
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
/* SATA spd limit is bound to the attached device, reset together */
link->sata_spd_limit = link->hw_sata_spd_limit;
link->sata_spd = 0;
/* High bits of dev->flags are used to record warm plug
* requests which occur asynchronously. Synchronize using
* host lock.
*/
spin_lock_irqsave(ap->lock, flags);
dev->flags &= ~ATA_DFLAG_INIT_MASK;
dev->horkage = 0;
spin_unlock_irqrestore(ap->lock, flags);
memset((void *)dev + ATA_DEVICE_CLEAR_BEGIN, 0,
ATA_DEVICE_CLEAR_END - ATA_DEVICE_CLEAR_BEGIN);
dev->pio_mask = UINT_MAX;
dev->mwdma_mask = UINT_MAX;
dev->udma_mask = UINT_MAX;
}
/**
* ata_link_init - Initialize an ata_link structure
* @ap: ATA port link is attached to
* @link: Link structure to initialize
* @pmp: Port multiplier port number
*
* Initialize @link.
*
* LOCKING:
* Kernel thread context (may sleep)
*/
void ata_link_init(struct ata_port *ap, struct ata_link *link, int pmp)
{
int i;
/* clear everything except for devices */
memset((void *)link + ATA_LINK_CLEAR_BEGIN, 0,
ATA_LINK_CLEAR_END - ATA_LINK_CLEAR_BEGIN);
link->ap = ap;
link->pmp = pmp;
link->active_tag = ATA_TAG_POISON;
link->hw_sata_spd_limit = UINT_MAX;
/* can't use iterator, ap isn't initialized yet */
for (i = 0; i < ATA_MAX_DEVICES; i++) {
struct ata_device *dev = &link->device[i];
dev->link = link;
dev->devno = dev - link->device;
#ifdef CONFIG_ATA_ACPI
dev->gtf_filter = ata_acpi_gtf_filter;
#endif
ata_dev_init(dev);
}
}
/**
* sata_link_init_spd - Initialize link->sata_spd_limit
* @link: Link to configure sata_spd_limit for
*
* Initialize @link->[hw_]sata_spd_limit to the currently
* configured value.
*
* LOCKING:
* Kernel thread context (may sleep).
*
* RETURNS:
* 0 on success, -errno on failure.
*/
int sata_link_init_spd(struct ata_link *link)
{
u8 spd;
int rc;
rc = sata_scr_read(link, SCR_CONTROL, &link->saved_scontrol);
if (rc)
return rc;
spd = (link->saved_scontrol >> 4) & 0xf;
if (spd)
link->hw_sata_spd_limit &= (1 << spd) - 1;
ata_force_link_limits(link);
link->sata_spd_limit = link->hw_sata_spd_limit;
return 0;
}
/**
* ata_port_alloc - allocate and initialize basic ATA port resources
* @host: ATA host this allocated port belongs to
*
* Allocate and initialize basic ATA port resources.
*
* RETURNS:
* Allocate ATA port on success, NULL on failure.
2005-05-31 07:49:12 +08:00
*
* LOCKING:
* Inherited from calling layer (may sleep).
*/
struct ata_port *ata_port_alloc(struct ata_host *host)
{
struct ata_port *ap;
DPRINTK("ENTER\n");
ap = kzalloc(sizeof(*ap), GFP_KERNEL);
if (!ap)
return NULL;
ap->pflags |= ATA_PFLAG_INITIALIZING | ATA_PFLAG_FROZEN;
ap->lock = &host->lock;
ap->print_id = -1;
ap->local_port_no = -1;
ap->host = host;
ap->dev = host->dev;
#if defined(ATA_VERBOSE_DEBUG)
/* turn on all debugging levels */
ap->msg_enable = 0x00FF;
#elif defined(ATA_DEBUG)
ap->msg_enable = ATA_MSG_DRV | ATA_MSG_INFO | ATA_MSG_CTL | ATA_MSG_WARN | ATA_MSG_ERR;
#else
ap->msg_enable = ATA_MSG_DRV | ATA_MSG_ERR | ATA_MSG_WARN;
#endif
mutex_init(&ap->scsi_scan_mutex);
2006-11-22 22:55:48 +08:00
INIT_DELAYED_WORK(&ap->hotplug_task, ata_scsi_hotplug);
INIT_WORK(&ap->scsi_rescan_task, ata_scsi_dev_rescan);
INIT_LIST_HEAD(&ap->eh_done_q);
init_waitqueue_head(&ap->eh_wait_q);
init_completion(&ap->park_req_pending);
init_timer_deferrable(&ap->fastdrain_timer);
ap->fastdrain_timer.function = ata_eh_fastdrain_timerfn;
ap->fastdrain_timer.data = (unsigned long)ap;
ap->cbl = ATA_CBL_NONE;
ata_link_init(ap, &ap->link, 0);
#ifdef ATA_IRQ_TRAP
ap->stats.unhandled_irq = 1;
ap->stats.idle_irq = 1;
#endif
ata_sff_port_init(ap);
return ap;
}
static void ata_host_release(struct device *gendev, void *res)
{
struct ata_host *host = dev_get_drvdata(gendev);
int i;
for (i = 0; i < host->n_ports; i++) {
struct ata_port *ap = host->ports[i];
if (!ap)
continue;
if (ap->scsi_host)
scsi_host_put(ap->scsi_host);
kfree(ap->pmp_link);
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
kfree(ap->slave_link);
kfree(ap);
host->ports[i] = NULL;
}
dev_set_drvdata(gendev, NULL);
}
/**
* ata_host_alloc - allocate and init basic ATA host resources
* @dev: generic device this host is associated with
* @max_ports: maximum number of ATA ports associated with this host
*
* Allocate and initialize basic ATA host resources. LLD calls
* this function to allocate a host, initializes it fully and
* attaches it using ata_host_register().
*
* @max_ports ports are allocated and host->n_ports is
* initialized to @max_ports. The caller is allowed to decrease
* host->n_ports before calling ata_host_register(). The unused
* ports will be automatically freed on registration.
*
* RETURNS:
* Allocate ATA host on success, NULL on failure.
*
* LOCKING:
* Inherited from calling layer (may sleep).
*/
struct ata_host *ata_host_alloc(struct device *dev, int max_ports)
{
struct ata_host *host;
size_t sz;
int i;
DPRINTK("ENTER\n");
if (!devres_open_group(dev, NULL, GFP_KERNEL))
return NULL;
/* alloc a container for our list of ATA ports (buses) */
sz = sizeof(struct ata_host) + (max_ports + 1) * sizeof(void *);
/* alloc a container for our list of ATA ports (buses) */
host = devres_alloc(ata_host_release, sz, GFP_KERNEL);
if (!host)
goto err_out;
devres_add(dev, host);
dev_set_drvdata(dev, host);
spin_lock_init(&host->lock);
libata: implement cross-port EH exclusion In libata, the non-EH code paths should always take and release ap->lock explicitly when accessing hardware or shared data structures. However, once EH is active, it's assumed that the port is owned by EH and EH methods don't explicitly take ap->lock unless race from irq handler or other code paths are expected. However, libata EH didn't guarantee exclusion among EHs for ports of the same host. IOW, multiple EHs may execute in parallel on multiple ports of the same controller. In many cases, especially in SATA, the ports are completely independent of each other and this doesn't cause problems; however, there are cases where different ports share the same resource, which lead to obscure timing related bugs such as the one fixed by commit 213373cf (ata_piix: fix locking around SIDPR access). This patch implements exclusion among EHs of the same host. When EH begins, it acquires per-host EH ownership by calling ata_eh_acquire(). When EH finishes, the ownership is released by calling ata_eh_release(). EH ownership is also released whenever the EH thread goes to sleep from ata_msleep() or explicitly and reacquired after waking up. This ensures that while EH is actively accessing the hardware, it has exclusive access to it while allowing EHs to interleave and progress in parallel as they hit waiting stages, which dominate the time spent in EH. This achieves cross-port EH exclusion without pervasive and fragile changes while still allowing parallel EH for the most part. This was first reported by yuanding02@gmail.com more than three years ago in the following bugzilla. :-) https://bugzilla.kernel.org/show_bug.cgi?id=8223 Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Reported-by: yuanding02@gmail.com Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-09-06 23:57:14 +08:00
mutex_init(&host->eh_mutex);
host->dev = dev;
host->n_ports = max_ports;
/* allocate ports bound to this host */
for (i = 0; i < max_ports; i++) {
struct ata_port *ap;
ap = ata_port_alloc(host);
if (!ap)
goto err_out;
ap->port_no = i;
host->ports[i] = ap;
}
devres_remove_group(dev, NULL);
return host;
err_out:
devres_release_group(dev, NULL);
return NULL;
}
/**
* ata_host_alloc_pinfo - alloc host and init with port_info array
* @dev: generic device this host is associated with
* @ppi: array of ATA port_info to initialize host with
* @n_ports: number of ATA ports attached to this host
*
* Allocate ATA host and initialize with info from @ppi. If NULL
* terminated, @ppi may contain fewer entries than @n_ports. The
* last entry will be used for the remaining ports.
*
* RETURNS:
* Allocate ATA host on success, NULL on failure.
*
* LOCKING:
* Inherited from calling layer (may sleep).
*/
struct ata_host *ata_host_alloc_pinfo(struct device *dev,
const struct ata_port_info * const * ppi,
int n_ports)
{
const struct ata_port_info *pi;
struct ata_host *host;
int i, j;
host = ata_host_alloc(dev, n_ports);
if (!host)
return NULL;
for (i = 0, j = 0, pi = NULL; i < host->n_ports; i++) {
struct ata_port *ap = host->ports[i];
if (ppi[j])
pi = ppi[j++];
ap->pio_mask = pi->pio_mask;
ap->mwdma_mask = pi->mwdma_mask;
ap->udma_mask = pi->udma_mask;
ap->flags |= pi->flags;
ap->link.flags |= pi->link_flags;
ap->ops = pi->port_ops;
if (!host->ops && (pi->port_ops != &ata_dummy_port_ops))
host->ops = pi->port_ops;
}
return host;
}
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
/**
* ata_slave_link_init - initialize slave link
* @ap: port to initialize slave link for
*
* Create and initialize slave link for @ap. This enables slave
* link handling on the port.
*
* In libata, a port contains links and a link contains devices.
* There is single host link but if a PMP is attached to it,
* there can be multiple fan-out links. On SATA, there's usually
* a single device connected to a link but PATA and SATA
* controllers emulating TF based interface can have two - master
* and slave.
*
* However, there are a few controllers which don't fit into this
* abstraction too well - SATA controllers which emulate TF
* interface with both master and slave devices but also have
* separate SCR register sets for each device. These controllers
* need separate links for physical link handling
* (e.g. onlineness, link speed) but should be treated like a
* traditional M/S controller for everything else (e.g. command
* issue, softreset).
*
* slave_link is libata's way of handling this class of
* controllers without impacting core layer too much. For
* anything other than physical link handling, the default host
* link is used for both master and slave. For physical link
* handling, separate @ap->slave_link is used. All dirty details
* are implemented inside libata core layer. From LLD's POV, the
* only difference is that prereset, hardreset and postreset are
* called once more for the slave link, so the reset sequence
* looks like the following.
*
* prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) ->
* softreset(M) -> postreset(M) -> postreset(S)
*
* Note that softreset is called only for the master. Softreset
* resets both M/S by definition, so SRST on master should handle
* both (the standard method will work just fine).
*
* LOCKING:
* Should be called before host is registered.
*
* RETURNS:
* 0 on success, -errno on failure.
*/
int ata_slave_link_init(struct ata_port *ap)
{
struct ata_link *link;
WARN_ON(ap->slave_link);
WARN_ON(ap->flags & ATA_FLAG_PMP);
link = kzalloc(sizeof(*link), GFP_KERNEL);
if (!link)
return -ENOMEM;
ata_link_init(ap, link, 1);
ap->slave_link = link;
return 0;
}
static void ata_host_stop(struct device *gendev, void *res)
{
struct ata_host *host = dev_get_drvdata(gendev);
int i;
WARN_ON(!(host->flags & ATA_HOST_STARTED));
for (i = 0; i < host->n_ports; i++) {
struct ata_port *ap = host->ports[i];
if (ap->ops->port_stop)
ap->ops->port_stop(ap);
}
if (host->ops->host_stop)
host->ops->host_stop(host);
}
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
/**
* ata_finalize_port_ops - finalize ata_port_operations
* @ops: ata_port_operations to finalize
*
* An ata_port_operations can inherit from another ops and that
* ops can again inherit from another. This can go on as many
* times as necessary as long as there is no loop in the
* inheritance chain.
*
* Ops tables are finalized when the host is started. NULL or
* unspecified entries are inherited from the closet ancestor
* which has the method and the entry is populated with it.
* After finalization, the ops table directly points to all the
* methods and ->inherits is no longer necessary and cleared.
*
* Using ATA_OP_NULL, inheriting ops can force a method to NULL.
*
* LOCKING:
* None.
*/
static void ata_finalize_port_ops(struct ata_port_operations *ops)
{
static DEFINE_SPINLOCK(lock);
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
const struct ata_port_operations *cur;
void **begin = (void **)ops;
void **end = (void **)&ops->inherits;
void **pp;
if (!ops || !ops->inherits)
return;
spin_lock(&lock);
for (cur = ops->inherits; cur; cur = cur->inherits) {
void **inherit = (void **)cur;
for (pp = begin; pp < end; pp++, inherit++)
if (!*pp)
*pp = *inherit;
}
for (pp = begin; pp < end; pp++)
if (IS_ERR(*pp))
*pp = NULL;
ops->inherits = NULL;
spin_unlock(&lock);
}
/**
* ata_host_start - start and freeze ports of an ATA host
* @host: ATA host to start ports for
*
* Start and then freeze ports of @host. Started status is
* recorded in host->flags, so this function can be called
* multiple times. Ports are guaranteed to get started only
* once. If host->ops isn't initialized yet, its set to the
* first non-dummy port ops.
*
* LOCKING:
* Inherited from calling layer (may sleep).
*
* RETURNS:
* 0 if all ports are started successfully, -errno otherwise.
*/
int ata_host_start(struct ata_host *host)
{
int have_stop = 0;
void *start_dr = NULL;
int i, rc;
if (host->flags & ATA_HOST_STARTED)
return 0;
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
ata_finalize_port_ops(host->ops);
for (i = 0; i < host->n_ports; i++) {
struct ata_port *ap = host->ports[i];
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
ata_finalize_port_ops(ap->ops);
if (!host->ops && !ata_port_is_dummy(ap))
host->ops = ap->ops;
if (ap->ops->port_stop)
have_stop = 1;
}
if (host->ops->host_stop)
have_stop = 1;
if (have_stop) {
start_dr = devres_alloc(ata_host_stop, 0, GFP_KERNEL);
if (!start_dr)
return -ENOMEM;
}
for (i = 0; i < host->n_ports; i++) {
struct ata_port *ap = host->ports[i];
if (ap->ops->port_start) {
rc = ap->ops->port_start(ap);
if (rc) {
if (rc != -ENODEV)
dev_err(host->dev,
"failed to start port %d (errno=%d)\n",
i, rc);
goto err_out;
}
}
ata_eh_freeze_port(ap);
}
if (start_dr)
devres_add(host->dev, start_dr);
host->flags |= ATA_HOST_STARTED;
return 0;
err_out:
while (--i >= 0) {
struct ata_port *ap = host->ports[i];
if (ap->ops->port_stop)
ap->ops->port_stop(ap);
}
devres_free(start_dr);
return rc;
}
/**
* ata_sas_host_init - Initialize a host struct for sas (ipr, libsas)
* @host: host to initialize
* @dev: device host is attached to
* @ops: port_ops
*
*/
void ata_host_init(struct ata_host *host, struct device *dev,
struct ata_port_operations *ops)
{
spin_lock_init(&host->lock);
libata: implement cross-port EH exclusion In libata, the non-EH code paths should always take and release ap->lock explicitly when accessing hardware or shared data structures. However, once EH is active, it's assumed that the port is owned by EH and EH methods don't explicitly take ap->lock unless race from irq handler or other code paths are expected. However, libata EH didn't guarantee exclusion among EHs for ports of the same host. IOW, multiple EHs may execute in parallel on multiple ports of the same controller. In many cases, especially in SATA, the ports are completely independent of each other and this doesn't cause problems; however, there are cases where different ports share the same resource, which lead to obscure timing related bugs such as the one fixed by commit 213373cf (ata_piix: fix locking around SIDPR access). This patch implements exclusion among EHs of the same host. When EH begins, it acquires per-host EH ownership by calling ata_eh_acquire(). When EH finishes, the ownership is released by calling ata_eh_release(). EH ownership is also released whenever the EH thread goes to sleep from ata_msleep() or explicitly and reacquired after waking up. This ensures that while EH is actively accessing the hardware, it has exclusive access to it while allowing EHs to interleave and progress in parallel as they hit waiting stages, which dominate the time spent in EH. This achieves cross-port EH exclusion without pervasive and fragile changes while still allowing parallel EH for the most part. This was first reported by yuanding02@gmail.com more than three years ago in the following bugzilla. :-) https://bugzilla.kernel.org/show_bug.cgi?id=8223 Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Reported-by: yuanding02@gmail.com Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-09-06 23:57:14 +08:00
mutex_init(&host->eh_mutex);
libata: introduce ata_host->n_tags to avoid oops on SAS controllers 1871ee134b73 ("libata: support the ata host which implements a queue depth less than 32") directly used ata_port->scsi_host->can_queue from ata_qc_new() to determine the number of tags supported by the host; unfortunately, SAS controllers doing SATA don't initialize ->scsi_host leading to the following oops. BUG: unable to handle kernel NULL pointer dereference at 0000000000000058 IP: [<ffffffff814e0618>] ata_qc_new_init+0x188/0x1b0 PGD 0 Oops: 0002 [#1] SMP Modules linked in: isci libsas scsi_transport_sas mgag200 drm_kms_helper ttm CPU: 1 PID: 518 Comm: udevd Not tainted 3.16.0-rc6+ #62 Hardware name: Intel Corporation S2600CO/S2600CO, BIOS SE5C600.86B.02.02.0002.122320131210 12/23/2013 task: ffff880c1a00b280 ti: ffff88061a000000 task.ti: ffff88061a000000 RIP: 0010:[<ffffffff814e0618>] [<ffffffff814e0618>] ata_qc_new_init+0x188/0x1b0 RSP: 0018:ffff88061a003ae8 EFLAGS: 00010012 RAX: 0000000000000001 RBX: ffff88000241ca80 RCX: 00000000000000fa RDX: 0000000000000020 RSI: 0000000000000020 RDI: ffff8806194aa298 RBP: ffff88061a003ae8 R08: ffff8806194a8000 R09: 0000000000000000 R10: 0000000000000000 R11: ffff88000241ca80 R12: ffff88061ad58200 R13: ffff8806194aa298 R14: ffffffff814e67a0 R15: ffff8806194a8000 FS: 00007f3ad7fe3840(0000) GS:ffff880627620000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000058 CR3: 000000061a118000 CR4: 00000000001407e0 Stack: ffff88061a003b20 ffffffff814e96e1 ffff88000241ca80 ffff88061ad58200 ffff8800b6bf6000 ffff880c1c988000 ffff880619903850 ffff88061a003b68 ffffffffa0056ce1 ffff88061a003b48 0000000013d6e6f8 ffff88000241ca80 Call Trace: [<ffffffff814e96e1>] ata_sas_queuecmd+0xa1/0x430 [<ffffffffa0056ce1>] sas_queuecommand+0x191/0x220 [libsas] [<ffffffff8149afee>] scsi_dispatch_cmd+0x10e/0x300 [<ffffffff814a3bc5>] scsi_request_fn+0x2f5/0x550 [<ffffffff81317613>] __blk_run_queue+0x33/0x40 [<ffffffff8131781a>] queue_unplugged+0x2a/0x90 [<ffffffff8131ceb4>] blk_flush_plug_list+0x1b4/0x210 [<ffffffff8131d274>] blk_finish_plug+0x14/0x50 [<ffffffff8117eaa8>] __do_page_cache_readahead+0x198/0x1f0 [<ffffffff8117ee21>] force_page_cache_readahead+0x31/0x50 [<ffffffff8117ee7e>] page_cache_sync_readahead+0x3e/0x50 [<ffffffff81172ac6>] generic_file_read_iter+0x496/0x5a0 [<ffffffff81219897>] blkdev_read_iter+0x37/0x40 [<ffffffff811e307e>] new_sync_read+0x7e/0xb0 [<ffffffff811e3734>] vfs_read+0x94/0x170 [<ffffffff811e43c6>] SyS_read+0x46/0xb0 [<ffffffff811e33d1>] ? SyS_lseek+0x91/0xb0 [<ffffffff8171ee29>] system_call_fastpath+0x16/0x1b Code: 00 00 00 88 50 29 83 7f 08 01 19 d2 83 e2 f0 83 ea 50 88 50 34 c6 81 1d 02 00 00 40 c6 81 17 02 00 00 00 5d c3 66 0f 1f 44 00 00 <89> 14 25 58 00 00 00 Fix it by introducing ata_host->n_tags which is initialized to ATA_MAX_QUEUE - 1 in ata_host_init() for SAS controllers and set to scsi_host_template->can_queue in ata_host_register() for !SAS ones. As SAS hosts are never registered, this will give them the same ATA_MAX_QUEUE - 1 as before. Note that we can't use scsi_host->can_queue directly for SAS hosts anyway as they can go higher than the libata maximum. Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Mike Qiu <qiudayu@linux.vnet.ibm.com> Reported-by: Jesse Brandeburg <jesse.brandeburg@gmail.com> Reported-by: Peter Hurley <peter@hurleysoftware.com> Reported-by: Peter Zijlstra <peterz@infradead.org> Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru> Fixes: 1871ee134b73 ("libata: support the ata host which implements a queue depth less than 32") Cc: Kevin Hao <haokexin@gmail.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: stable@vger.kernel.org
2014-07-23 21:05:27 +08:00
host->n_tags = ATA_MAX_QUEUE - 1;
host->dev = dev;
host->ops = ops;
}
void __ata_port_probe(struct ata_port *ap)
{
struct ata_eh_info *ehi = &ap->link.eh_info;
unsigned long flags;
/* kick EH for boot probing */
spin_lock_irqsave(ap->lock, flags);
ehi->probe_mask |= ATA_ALL_DEVICES;
ehi->action |= ATA_EH_RESET;
ehi->flags |= ATA_EHI_NO_AUTOPSY | ATA_EHI_QUIET;
ap->pflags &= ~ATA_PFLAG_INITIALIZING;
ap->pflags |= ATA_PFLAG_LOADING;
ata_port_schedule_eh(ap);
spin_unlock_irqrestore(ap->lock, flags);
}
int ata_port_probe(struct ata_port *ap)
{
int rc = 0;
if (ap->ops->error_handler) {
__ata_port_probe(ap);
ata_port_wait_eh(ap);
} else {
DPRINTK("ata%u: bus probe begin\n", ap->print_id);
rc = ata_bus_probe(ap);
DPRINTK("ata%u: bus probe end\n", ap->print_id);
}
return rc;
}
static void async_port_probe(void *data, async_cookie_t cookie)
{
struct ata_port *ap = data;
/*
* If we're not allowed to scan this host in parallel,
* we need to wait until all previous scans have completed
* before going further.
* Jeff Garzik says this is only within a controller, so we
* don't need to wait for port 0, only for later ports.
*/
if (!(ap->host->flags & ATA_HOST_PARALLEL_SCAN) && ap->port_no != 0)
async_synchronize_cookie(cookie);
(void)ata_port_probe(ap);
/* in order to keep device order, we need to synchronize at this point */
async_synchronize_cookie(cookie);
ata_scsi_scan_host(ap, 1);
}
/**
* ata_host_register - register initialized ATA host
* @host: ATA host to register
* @sht: template for SCSI host
*
* Register initialized ATA host. @host is allocated using
* ata_host_alloc() and fully initialized by LLD. This function
* starts ports, registers @host with ATA and SCSI layers and
* probe registered devices.
*
* LOCKING:
* Inherited from calling layer (may sleep).
*
* RETURNS:
* 0 on success, -errno otherwise.
*/
int ata_host_register(struct ata_host *host, struct scsi_host_template *sht)
{
int i, rc;
libata: introduce ata_host->n_tags to avoid oops on SAS controllers 1871ee134b73 ("libata: support the ata host which implements a queue depth less than 32") directly used ata_port->scsi_host->can_queue from ata_qc_new() to determine the number of tags supported by the host; unfortunately, SAS controllers doing SATA don't initialize ->scsi_host leading to the following oops. BUG: unable to handle kernel NULL pointer dereference at 0000000000000058 IP: [<ffffffff814e0618>] ata_qc_new_init+0x188/0x1b0 PGD 0 Oops: 0002 [#1] SMP Modules linked in: isci libsas scsi_transport_sas mgag200 drm_kms_helper ttm CPU: 1 PID: 518 Comm: udevd Not tainted 3.16.0-rc6+ #62 Hardware name: Intel Corporation S2600CO/S2600CO, BIOS SE5C600.86B.02.02.0002.122320131210 12/23/2013 task: ffff880c1a00b280 ti: ffff88061a000000 task.ti: ffff88061a000000 RIP: 0010:[<ffffffff814e0618>] [<ffffffff814e0618>] ata_qc_new_init+0x188/0x1b0 RSP: 0018:ffff88061a003ae8 EFLAGS: 00010012 RAX: 0000000000000001 RBX: ffff88000241ca80 RCX: 00000000000000fa RDX: 0000000000000020 RSI: 0000000000000020 RDI: ffff8806194aa298 RBP: ffff88061a003ae8 R08: ffff8806194a8000 R09: 0000000000000000 R10: 0000000000000000 R11: ffff88000241ca80 R12: ffff88061ad58200 R13: ffff8806194aa298 R14: ffffffff814e67a0 R15: ffff8806194a8000 FS: 00007f3ad7fe3840(0000) GS:ffff880627620000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000058 CR3: 000000061a118000 CR4: 00000000001407e0 Stack: ffff88061a003b20 ffffffff814e96e1 ffff88000241ca80 ffff88061ad58200 ffff8800b6bf6000 ffff880c1c988000 ffff880619903850 ffff88061a003b68 ffffffffa0056ce1 ffff88061a003b48 0000000013d6e6f8 ffff88000241ca80 Call Trace: [<ffffffff814e96e1>] ata_sas_queuecmd+0xa1/0x430 [<ffffffffa0056ce1>] sas_queuecommand+0x191/0x220 [libsas] [<ffffffff8149afee>] scsi_dispatch_cmd+0x10e/0x300 [<ffffffff814a3bc5>] scsi_request_fn+0x2f5/0x550 [<ffffffff81317613>] __blk_run_queue+0x33/0x40 [<ffffffff8131781a>] queue_unplugged+0x2a/0x90 [<ffffffff8131ceb4>] blk_flush_plug_list+0x1b4/0x210 [<ffffffff8131d274>] blk_finish_plug+0x14/0x50 [<ffffffff8117eaa8>] __do_page_cache_readahead+0x198/0x1f0 [<ffffffff8117ee21>] force_page_cache_readahead+0x31/0x50 [<ffffffff8117ee7e>] page_cache_sync_readahead+0x3e/0x50 [<ffffffff81172ac6>] generic_file_read_iter+0x496/0x5a0 [<ffffffff81219897>] blkdev_read_iter+0x37/0x40 [<ffffffff811e307e>] new_sync_read+0x7e/0xb0 [<ffffffff811e3734>] vfs_read+0x94/0x170 [<ffffffff811e43c6>] SyS_read+0x46/0xb0 [<ffffffff811e33d1>] ? SyS_lseek+0x91/0xb0 [<ffffffff8171ee29>] system_call_fastpath+0x16/0x1b Code: 00 00 00 88 50 29 83 7f 08 01 19 d2 83 e2 f0 83 ea 50 88 50 34 c6 81 1d 02 00 00 40 c6 81 17 02 00 00 00 5d c3 66 0f 1f 44 00 00 <89> 14 25 58 00 00 00 Fix it by introducing ata_host->n_tags which is initialized to ATA_MAX_QUEUE - 1 in ata_host_init() for SAS controllers and set to scsi_host_template->can_queue in ata_host_register() for !SAS ones. As SAS hosts are never registered, this will give them the same ATA_MAX_QUEUE - 1 as before. Note that we can't use scsi_host->can_queue directly for SAS hosts anyway as they can go higher than the libata maximum. Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Mike Qiu <qiudayu@linux.vnet.ibm.com> Reported-by: Jesse Brandeburg <jesse.brandeburg@gmail.com> Reported-by: Peter Hurley <peter@hurleysoftware.com> Reported-by: Peter Zijlstra <peterz@infradead.org> Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru> Fixes: 1871ee134b73 ("libata: support the ata host which implements a queue depth less than 32") Cc: Kevin Hao <haokexin@gmail.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: stable@vger.kernel.org
2014-07-23 21:05:27 +08:00
host->n_tags = clamp(sht->can_queue, 1, ATA_MAX_QUEUE - 1);
/* host must have been started */
if (!(host->flags & ATA_HOST_STARTED)) {
dev_err(host->dev, "BUG: trying to register unstarted host\n");
WARN_ON(1);
return -EINVAL;
}
/* Blow away unused ports. This happens when LLD can't
* determine the exact number of ports to allocate at
* allocation time.
*/
for (i = host->n_ports; host->ports[i]; i++)
kfree(host->ports[i]);
/* give ports names and add SCSI hosts */
for (i = 0; i < host->n_ports; i++) {
host->ports[i]->print_id = atomic_inc_return(&ata_print_id);
host->ports[i]->local_port_no = i + 1;
}
/* Create associated sysfs transport objects */
for (i = 0; i < host->n_ports; i++) {
rc = ata_tport_add(host->dev,host->ports[i]);
if (rc) {
goto err_tadd;
}
}
rc = ata_scsi_add_hosts(host, sht);
if (rc)
goto err_tadd;
/* set cable, sata_spd_limit and report */
for (i = 0; i < host->n_ports; i++) {
struct ata_port *ap = host->ports[i];
unsigned long xfer_mask;
/* set SATA cable type if still unset */
if (ap->cbl == ATA_CBL_NONE && (ap->flags & ATA_FLAG_SATA))
ap->cbl = ATA_CBL_SATA;
/* init sata_spd_limit to the current value */
sata_link_init_spd(&ap->link);
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
if (ap->slave_link)
sata_link_init_spd(ap->slave_link);
/* print per-port info to dmesg */
xfer_mask = ata_pack_xfermask(ap->pio_mask, ap->mwdma_mask,
ap->udma_mask);
if (!ata_port_is_dummy(ap)) {
ata_port_info(ap, "%cATA max %s %s\n",
(ap->flags & ATA_FLAG_SATA) ? 'S' : 'P',
ata_mode_string(xfer_mask),
ap->link.eh_info.desc);
ata_ehi_clear_desc(&ap->link.eh_info);
} else
ata_port_info(ap, "DUMMY\n");
}
/* perform each probe asynchronously */
for (i = 0; i < host->n_ports; i++) {
struct ata_port *ap = host->ports[i];
async_schedule(async_port_probe, ap);
}
return 0;
err_tadd:
while (--i >= 0) {
ata_tport_delete(host->ports[i]);
}
return rc;
}
/**
* ata_host_activate - start host, request IRQ and register it
* @host: target ATA host
* @irq: IRQ to request
* @irq_handler: irq_handler used when requesting IRQ
* @irq_flags: irq_flags used when requesting IRQ
* @sht: scsi_host_template to use when registering the host
*
* After allocating an ATA host and initializing it, most libata
* LLDs perform three steps to activate the host - start host,
* request IRQ and register it. This helper takes necessasry
* arguments and performs the three steps in one go.
*
* An invalid IRQ skips the IRQ registration and expects the host to
* have set polling mode on the port. In this case, @irq_handler
* should be NULL.
*
* LOCKING:
* Inherited from calling layer (may sleep).
*
* RETURNS:
* 0 on success, -errno otherwise.
*/
int ata_host_activate(struct ata_host *host, int irq,
irq_handler_t irq_handler, unsigned long irq_flags,
struct scsi_host_template *sht)
{
int i, rc;
rc = ata_host_start(host);
if (rc)
return rc;
/* Special case for polling mode */
if (!irq) {
WARN_ON(irq_handler);
return ata_host_register(host, sht);
}
rc = devm_request_irq(host->dev, irq, irq_handler, irq_flags,
dev_name(host->dev), host);
if (rc)
return rc;
for (i = 0; i < host->n_ports; i++)
ata_port_desc(host->ports[i], "irq %d", irq);
rc = ata_host_register(host, sht);
/* if failed, just free the IRQ and leave ports alone */
if (rc)
devm_free_irq(host->dev, irq, host);
return rc;
}
/**
* ata_port_detach - Detach ATA port in prepration of device removal
* @ap: ATA port to be detached
*
* Detach all ATA devices and the associated SCSI devices of @ap;
* then, remove the associated SCSI host. @ap is guaranteed to
* be quiescent on return from this function.
*
* LOCKING:
* Kernel thread context (may sleep).
*/
static void ata_port_detach(struct ata_port *ap)
{
unsigned long flags;
struct ata_link *link;
struct ata_device *dev;
if (!ap->ops->error_handler)
goto skip_eh;
/* tell EH we're leaving & flush EH */
spin_lock_irqsave(ap->lock, flags);
ap->pflags |= ATA_PFLAG_UNLOADING;
ata_port_schedule_eh(ap);
spin_unlock_irqrestore(ap->lock, flags);
/* wait till EH commits suicide */
ata_port_wait_eh(ap);
/* it better be dead now */
WARN_ON(!(ap->pflags & ATA_PFLAG_UNLOADED));
cancel_delayed_work_sync(&ap->hotplug_task);
skip_eh:
/* clean up zpodd on port removal */
ata_for_each_link(link, ap, HOST_FIRST) {
ata_for_each_dev(dev, link, ALL) {
if (zpodd_dev_enabled(dev))
zpodd_exit(dev);
}
}
if (ap->pmp_link) {
int i;
for (i = 0; i < SATA_PMP_MAX_PORTS; i++)
ata_tlink_delete(&ap->pmp_link[i]);
}
/* remove the associated SCSI host */
scsi_remove_host(ap->scsi_host);
ATA: Fix port removal ordering After commit bcdde7e221a8 (sysfs: make __sysfs_remove_dir() recursive) Mika Westerberg sees traces analogous to the one below in Thunderbolt hot-remove testing: WARNING: CPU: 0 PID: 4 at fs/sysfs/group.c:214 sysfs_remove_group+0xc6/0xd0() sysfs group ffffffff81c6f1e0 not found for kobject 'host7' Modules linked in: CPU: 0 PID: 4 Comm: kworker/0:0 Not tainted 3.12.0+ #13 Hardware name: /D33217CK, BIOS GKPPT10H.86A.0042.2013.0422.1439 04/22/2013 Workqueue: kacpi_hotplug acpi_hotplug_work_fn 0000000000000009 ffff8801002459b0 ffffffff817daab1 ffff8801002459f8 ffff8801002459e8 ffffffff810436b8 0000000000000000 ffffffff81c6f1e0 ffff88006d440358 ffff88006d440188 ffff88006e8b4c28 ffff880100245a48 Call Trace: [<ffffffff817daab1>] dump_stack+0x45/0x56 [<ffffffff810436b8>] warn_slowpath_common+0x78/0xa0 [<ffffffff81043727>] warn_slowpath_fmt+0x47/0x50 [<ffffffff811ad319>] ? sysfs_get_dirent_ns+0x49/0x70 [<ffffffff811ae526>] sysfs_remove_group+0xc6/0xd0 [<ffffffff81432f7e>] dpm_sysfs_remove+0x3e/0x50 [<ffffffff8142a0d0>] device_del+0x40/0x1b0 [<ffffffff8142a24d>] device_unregister+0xd/0x20 [<ffffffff8144131a>] scsi_remove_host+0xba/0x110 [<ffffffff8145f526>] ata_host_detach+0xc6/0x100 [<ffffffff8145f578>] ata_pci_remove_one+0x18/0x20 [<ffffffff812e8f48>] pci_device_remove+0x28/0x60 [<ffffffff8142d854>] __device_release_driver+0x64/0xd0 [<ffffffff8142d8de>] device_release_driver+0x1e/0x30 [<ffffffff8142d257>] bus_remove_device+0xf7/0x140 [<ffffffff8142a1b1>] device_del+0x121/0x1b0 [<ffffffff812e43d4>] pci_stop_bus_device+0x94/0xa0 [<ffffffff812e437b>] pci_stop_bus_device+0x3b/0xa0 [<ffffffff812e437b>] pci_stop_bus_device+0x3b/0xa0 [<ffffffff812e44dd>] pci_stop_and_remove_bus_device+0xd/0x20 [<ffffffff812fc743>] trim_stale_devices+0x73/0xe0 [<ffffffff812fc78b>] trim_stale_devices+0xbb/0xe0 [<ffffffff812fc78b>] trim_stale_devices+0xbb/0xe0 [<ffffffff812fcb6e>] acpiphp_check_bridge+0x7e/0xd0 [<ffffffff812fd90d>] hotplug_event+0xcd/0x160 [<ffffffff812fd9c5>] hotplug_event_work+0x25/0x60 [<ffffffff81316749>] acpi_hotplug_work_fn+0x17/0x22 [<ffffffff8105cf3a>] process_one_work+0x17a/0x430 [<ffffffff8105db29>] worker_thread+0x119/0x390 [<ffffffff8105da10>] ? manage_workers.isra.25+0x2a0/0x2a0 [<ffffffff81063a5d>] kthread+0xcd/0xf0 [<ffffffff81063990>] ? kthread_create_on_node+0x180/0x180 [<ffffffff817eb33c>] ret_from_fork+0x7c/0xb0 [<ffffffff81063990>] ? kthread_create_on_node+0x180/0x180 The source of this problem is that SCSI hosts are removed from ATA ports after calling ata_tport_delete() which removes the port's sysfs directory, among other things. Now, after commit bcdde7e221a8, the sysfs directory is removed along with all of its subdirectories that include the SCSI host's sysfs directory and its subdirectories at this point. Consequently, when device_del() is finally called for any child device of the SCSI host and tries to remove its "power" group (which is already gone then), it triggers the above warning. To make the warnings go away, change the removal ordering in ata_port_detach() so that the SCSI host is removed from the port before ata_tport_delete() is called. References: https://bugzilla.kernel.org/show_bug.cgi?id=65281 Reported-and-tested-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Jingoo Han <jg1.han@samsung.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2013-11-25 20:19:01 +08:00
ata_tport_delete(ap);
}
/**
* ata_host_detach - Detach all ports of an ATA host
* @host: Host to detach
*
* Detach all ports of @host.
*
* LOCKING:
* Kernel thread context (may sleep).
*/
void ata_host_detach(struct ata_host *host)
{
int i;
for (i = 0; i < host->n_ports; i++)
ata_port_detach(host->ports[i]);
/* the host is dead now, dissociate ACPI */
ata_acpi_dissociate(host);
}
#ifdef CONFIG_PCI
/**
* ata_pci_remove_one - PCI layer callback for device removal
* @pdev: PCI device that was removed
*
* PCI layer indicates to libata via this hook that hot-unplug or
* module unload event has occurred. Detach all ports. Resource
* release is handled via devres.
*
* LOCKING:
* Inherited from PCI layer (may sleep).
*/
void ata_pci_remove_one(struct pci_dev *pdev)
{
struct ata_host *host = pci_get_drvdata(pdev);
ata_host_detach(host);
}
/* move to PCI subsystem */
int pci_test_config_bits(struct pci_dev *pdev, const struct pci_bits *bits)
{
unsigned long tmp = 0;
switch (bits->width) {
case 1: {
u8 tmp8 = 0;
pci_read_config_byte(pdev, bits->reg, &tmp8);
tmp = tmp8;
break;
}
case 2: {
u16 tmp16 = 0;
pci_read_config_word(pdev, bits->reg, &tmp16);
tmp = tmp16;
break;
}
case 4: {
u32 tmp32 = 0;
pci_read_config_dword(pdev, bits->reg, &tmp32);
tmp = tmp32;
break;
}
default:
return -EINVAL;
}
tmp &= bits->mask;
return (tmp == bits->val) ? 1 : 0;
}
#ifdef CONFIG_PM
void ata_pci_device_do_suspend(struct pci_dev *pdev, pm_message_t mesg)
{
pci_save_state(pdev);
pci_disable_device(pdev);
if (mesg.event & PM_EVENT_SLEEP)
pci_set_power_state(pdev, PCI_D3hot);
}
int ata_pci_device_do_resume(struct pci_dev *pdev)
{
int rc;
pci_set_power_state(pdev, PCI_D0);
pci_restore_state(pdev);
rc = pcim_enable_device(pdev);
if (rc) {
dev_err(&pdev->dev,
"failed to enable device after resume (%d)\n", rc);
return rc;
}
pci_set_master(pdev);
return 0;
}
int ata_pci_device_suspend(struct pci_dev *pdev, pm_message_t mesg)
{
struct ata_host *host = pci_get_drvdata(pdev);
int rc = 0;
rc = ata_host_suspend(host, mesg);
if (rc)
return rc;
ata_pci_device_do_suspend(pdev, mesg);
return 0;
}
int ata_pci_device_resume(struct pci_dev *pdev)
{
struct ata_host *host = pci_get_drvdata(pdev);
int rc;
rc = ata_pci_device_do_resume(pdev);
if (rc == 0)
ata_host_resume(host);
return rc;
}
#endif /* CONFIG_PM */
#endif /* CONFIG_PCI */
/**
* ata_platform_remove_one - Platform layer callback for device removal
* @pdev: Platform device that was removed
*
* Platform layer indicates to libata via this hook that hot-unplug or
* module unload event has occurred. Detach all ports. Resource
* release is handled via devres.
*
* LOCKING:
* Inherited from platform layer (may sleep).
*/
int ata_platform_remove_one(struct platform_device *pdev)
{
struct ata_host *host = platform_get_drvdata(pdev);
ata_host_detach(host);
return 0;
}
static int __init ata_parse_force_one(char **cur,
struct ata_force_ent *force_ent,
const char **reason)
{
/* FIXME: Currently, there's no way to tag init const data and
* using __initdata causes build failure on some versions of
* gcc. Once __initdataconst is implemented, add const to the
* following structure.
*/
static struct ata_force_param force_tbl[] __initdata = {
{ "40c", .cbl = ATA_CBL_PATA40 },
{ "80c", .cbl = ATA_CBL_PATA80 },
{ "short40c", .cbl = ATA_CBL_PATA40_SHORT },
{ "unk", .cbl = ATA_CBL_PATA_UNK },
{ "ign", .cbl = ATA_CBL_PATA_IGN },
{ "sata", .cbl = ATA_CBL_SATA },
{ "1.5Gbps", .spd_limit = 1 },
{ "3.0Gbps", .spd_limit = 2 },
{ "noncq", .horkage_on = ATA_HORKAGE_NONCQ },
{ "ncq", .horkage_off = ATA_HORKAGE_NONCQ },
{ "dump_id", .horkage_on = ATA_HORKAGE_DUMP_ID },
{ "pio0", .xfer_mask = 1 << (ATA_SHIFT_PIO + 0) },
{ "pio1", .xfer_mask = 1 << (ATA_SHIFT_PIO + 1) },
{ "pio2", .xfer_mask = 1 << (ATA_SHIFT_PIO + 2) },
{ "pio3", .xfer_mask = 1 << (ATA_SHIFT_PIO + 3) },
{ "pio4", .xfer_mask = 1 << (ATA_SHIFT_PIO + 4) },
{ "pio5", .xfer_mask = 1 << (ATA_SHIFT_PIO + 5) },
{ "pio6", .xfer_mask = 1 << (ATA_SHIFT_PIO + 6) },
{ "mwdma0", .xfer_mask = 1 << (ATA_SHIFT_MWDMA + 0) },
{ "mwdma1", .xfer_mask = 1 << (ATA_SHIFT_MWDMA + 1) },
{ "mwdma2", .xfer_mask = 1 << (ATA_SHIFT_MWDMA + 2) },
{ "mwdma3", .xfer_mask = 1 << (ATA_SHIFT_MWDMA + 3) },
{ "mwdma4", .xfer_mask = 1 << (ATA_SHIFT_MWDMA + 4) },
{ "udma0", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 0) },
{ "udma16", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 0) },
{ "udma/16", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 0) },
{ "udma1", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 1) },
{ "udma25", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 1) },
{ "udma/25", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 1) },
{ "udma2", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 2) },
{ "udma33", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 2) },
{ "udma/33", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 2) },
{ "udma3", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 3) },
{ "udma44", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 3) },
{ "udma/44", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 3) },
{ "udma4", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 4) },
{ "udma66", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 4) },
{ "udma/66", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 4) },
{ "udma5", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 5) },
{ "udma100", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 5) },
{ "udma/100", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 5) },
{ "udma6", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 6) },
{ "udma133", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 6) },
{ "udma/133", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 6) },
{ "udma7", .xfer_mask = 1 << (ATA_SHIFT_UDMA + 7) },
{ "nohrst", .lflags = ATA_LFLAG_NO_HRST },
{ "nosrst", .lflags = ATA_LFLAG_NO_SRST },
{ "norst", .lflags = ATA_LFLAG_NO_HRST | ATA_LFLAG_NO_SRST },
{ "rstonce", .lflags = ATA_LFLAG_RST_ONCE },
{ "atapi_dmadir", .horkage_on = ATA_HORKAGE_ATAPI_DMADIR },
{ "disable", .horkage_on = ATA_HORKAGE_DISABLE },
};
char *start = *cur, *p = *cur;
char *id, *val, *endp;
const struct ata_force_param *match_fp = NULL;
int nr_matches = 0, i;
/* find where this param ends and update *cur */
while (*p != '\0' && *p != ',')
p++;
if (*p == '\0')
*cur = p;
else
*cur = p + 1;
*p = '\0';
/* parse */
p = strchr(start, ':');
if (!p) {
val = strstrip(start);
goto parse_val;
}
*p = '\0';
id = strstrip(start);
val = strstrip(p + 1);
/* parse id */
p = strchr(id, '.');
if (p) {
*p++ = '\0';
force_ent->device = simple_strtoul(p, &endp, 10);
if (p == endp || *endp != '\0') {
*reason = "invalid device";
return -EINVAL;
}
}
force_ent->port = simple_strtoul(id, &endp, 10);
if (p == endp || *endp != '\0') {
*reason = "invalid port/link";
return -EINVAL;
}
parse_val:
/* parse val, allow shortcuts so that both 1.5 and 1.5Gbps work */
for (i = 0; i < ARRAY_SIZE(force_tbl); i++) {
const struct ata_force_param *fp = &force_tbl[i];
if (strncasecmp(val, fp->name, strlen(val)))
continue;
nr_matches++;
match_fp = fp;
if (strcasecmp(val, fp->name) == 0) {
nr_matches = 1;
break;
}
}
if (!nr_matches) {
*reason = "unknown value";
return -EINVAL;
}
if (nr_matches > 1) {
*reason = "ambigious value";
return -EINVAL;
}
force_ent->param = *match_fp;
return 0;
}
static void __init ata_parse_force_param(void)
{
int idx = 0, size = 1;
int last_port = -1, last_device = -1;
char *p, *cur, *next;
/* calculate maximum number of params and allocate force_tbl */
for (p = ata_force_param_buf; *p; p++)
if (*p == ',')
size++;
ata_force_tbl = kzalloc(sizeof(ata_force_tbl[0]) * size, GFP_KERNEL);
if (!ata_force_tbl) {
printk(KERN_WARNING "ata: failed to extend force table, "
"libata.force ignored\n");
return;
}
/* parse and populate the table */
for (cur = ata_force_param_buf; *cur != '\0'; cur = next) {
const char *reason = "";
struct ata_force_ent te = { .port = -1, .device = -1 };
next = cur;
if (ata_parse_force_one(&next, &te, &reason)) {
printk(KERN_WARNING "ata: failed to parse force "
"parameter \"%s\" (%s)\n",
cur, reason);
continue;
}
if (te.port == -1) {
te.port = last_port;
te.device = last_device;
}
ata_force_tbl[idx++] = te;
last_port = te.port;
last_device = te.device;
}
ata_force_tbl_size = idx;
}
static int __init ata_init(void)
{
int rc;
ata_parse_force_param();
rc = ata_sff_init();
if (rc) {
kfree(ata_force_tbl);
return rc;
}
libata_transport_init();
ata_scsi_transport_template = ata_attach_transport();
if (!ata_scsi_transport_template) {
ata_sff_exit();
rc = -ENOMEM;
goto err_out;
}
printk(KERN_DEBUG "libata version " DRV_VERSION " loaded.\n");
return 0;
err_out:
return rc;
}
static void __exit ata_exit(void)
{
ata_release_transport(ata_scsi_transport_template);
libata_transport_exit();
ata_sff_exit();
kfree(ata_force_tbl);
}
subsys_initcall(ata_init);
module_exit(ata_exit);
static DEFINE_RATELIMIT_STATE(ratelimit, HZ / 5, 1);
int ata_ratelimit(void)
{
return __ratelimit(&ratelimit);
}
libata: implement cross-port EH exclusion In libata, the non-EH code paths should always take and release ap->lock explicitly when accessing hardware or shared data structures. However, once EH is active, it's assumed that the port is owned by EH and EH methods don't explicitly take ap->lock unless race from irq handler or other code paths are expected. However, libata EH didn't guarantee exclusion among EHs for ports of the same host. IOW, multiple EHs may execute in parallel on multiple ports of the same controller. In many cases, especially in SATA, the ports are completely independent of each other and this doesn't cause problems; however, there are cases where different ports share the same resource, which lead to obscure timing related bugs such as the one fixed by commit 213373cf (ata_piix: fix locking around SIDPR access). This patch implements exclusion among EHs of the same host. When EH begins, it acquires per-host EH ownership by calling ata_eh_acquire(). When EH finishes, the ownership is released by calling ata_eh_release(). EH ownership is also released whenever the EH thread goes to sleep from ata_msleep() or explicitly and reacquired after waking up. This ensures that while EH is actively accessing the hardware, it has exclusive access to it while allowing EHs to interleave and progress in parallel as they hit waiting stages, which dominate the time spent in EH. This achieves cross-port EH exclusion without pervasive and fragile changes while still allowing parallel EH for the most part. This was first reported by yuanding02@gmail.com more than three years ago in the following bugzilla. :-) https://bugzilla.kernel.org/show_bug.cgi?id=8223 Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Reported-by: yuanding02@gmail.com Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-09-06 23:57:14 +08:00
/**
* ata_msleep - ATA EH owner aware msleep
* @ap: ATA port to attribute the sleep to
* @msecs: duration to sleep in milliseconds
*
* Sleeps @msecs. If the current task is owner of @ap's EH, the
* ownership is released before going to sleep and reacquired
* after the sleep is complete. IOW, other ports sharing the
* @ap->host will be allowed to own the EH while this task is
* sleeping.
*
* LOCKING:
* Might sleep.
*/
void ata_msleep(struct ata_port *ap, unsigned int msecs)
{
libata: implement cross-port EH exclusion In libata, the non-EH code paths should always take and release ap->lock explicitly when accessing hardware or shared data structures. However, once EH is active, it's assumed that the port is owned by EH and EH methods don't explicitly take ap->lock unless race from irq handler or other code paths are expected. However, libata EH didn't guarantee exclusion among EHs for ports of the same host. IOW, multiple EHs may execute in parallel on multiple ports of the same controller. In many cases, especially in SATA, the ports are completely independent of each other and this doesn't cause problems; however, there are cases where different ports share the same resource, which lead to obscure timing related bugs such as the one fixed by commit 213373cf (ata_piix: fix locking around SIDPR access). This patch implements exclusion among EHs of the same host. When EH begins, it acquires per-host EH ownership by calling ata_eh_acquire(). When EH finishes, the ownership is released by calling ata_eh_release(). EH ownership is also released whenever the EH thread goes to sleep from ata_msleep() or explicitly and reacquired after waking up. This ensures that while EH is actively accessing the hardware, it has exclusive access to it while allowing EHs to interleave and progress in parallel as they hit waiting stages, which dominate the time spent in EH. This achieves cross-port EH exclusion without pervasive and fragile changes while still allowing parallel EH for the most part. This was first reported by yuanding02@gmail.com more than three years ago in the following bugzilla. :-) https://bugzilla.kernel.org/show_bug.cgi?id=8223 Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Reported-by: yuanding02@gmail.com Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-09-06 23:57:14 +08:00
bool owns_eh = ap && ap->host->eh_owner == current;
if (owns_eh)
ata_eh_release(ap);
msleep(msecs);
libata: implement cross-port EH exclusion In libata, the non-EH code paths should always take and release ap->lock explicitly when accessing hardware or shared data structures. However, once EH is active, it's assumed that the port is owned by EH and EH methods don't explicitly take ap->lock unless race from irq handler or other code paths are expected. However, libata EH didn't guarantee exclusion among EHs for ports of the same host. IOW, multiple EHs may execute in parallel on multiple ports of the same controller. In many cases, especially in SATA, the ports are completely independent of each other and this doesn't cause problems; however, there are cases where different ports share the same resource, which lead to obscure timing related bugs such as the one fixed by commit 213373cf (ata_piix: fix locking around SIDPR access). This patch implements exclusion among EHs of the same host. When EH begins, it acquires per-host EH ownership by calling ata_eh_acquire(). When EH finishes, the ownership is released by calling ata_eh_release(). EH ownership is also released whenever the EH thread goes to sleep from ata_msleep() or explicitly and reacquired after waking up. This ensures that while EH is actively accessing the hardware, it has exclusive access to it while allowing EHs to interleave and progress in parallel as they hit waiting stages, which dominate the time spent in EH. This achieves cross-port EH exclusion without pervasive and fragile changes while still allowing parallel EH for the most part. This was first reported by yuanding02@gmail.com more than three years ago in the following bugzilla. :-) https://bugzilla.kernel.org/show_bug.cgi?id=8223 Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Reported-by: yuanding02@gmail.com Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2010-09-06 23:57:14 +08:00
if (owns_eh)
ata_eh_acquire(ap);
}
/**
* ata_wait_register - wait until register value changes
* @ap: ATA port to wait register for, can be NULL
* @reg: IO-mapped register
* @mask: Mask to apply to read register value
* @val: Wait condition
* @interval: polling interval in milliseconds
* @timeout: timeout in milliseconds
*
* Waiting for some bits of register to change is a common
* operation for ATA controllers. This function reads 32bit LE
* IO-mapped register @reg and tests for the following condition.
*
* (*@reg & mask) != val
*
* If the condition is met, it returns; otherwise, the process is
* repeated after @interval_msec until timeout.
*
* LOCKING:
* Kernel thread context (may sleep)
*
* RETURNS:
* The final register value.
*/
u32 ata_wait_register(struct ata_port *ap, void __iomem *reg, u32 mask, u32 val,
unsigned long interval, unsigned long timeout)
{
unsigned long deadline;
u32 tmp;
tmp = ioread32(reg);
/* Calculate timeout _after_ the first read to make sure
* preceding writes reach the controller before starting to
* eat away the timeout.
*/
deadline = ata_deadline(jiffies, timeout);
while ((tmp & mask) == val && time_before(jiffies, deadline)) {
ata_msleep(ap, interval);
tmp = ioread32(reg);
}
return tmp;
}
/*
* Dummy port_ops
*/
static unsigned int ata_dummy_qc_issue(struct ata_queued_cmd *qc)
{
return AC_ERR_SYSTEM;
}
static void ata_dummy_error_handler(struct ata_port *ap)
{
/* truly dummy */
}
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
struct ata_port_operations ata_dummy_port_ops = {
.qc_prep = ata_noop_qc_prep,
.qc_issue = ata_dummy_qc_issue,
.error_handler = ata_dummy_error_handler,
.sched_eh = ata_std_sched_eh,
.end_eh = ata_std_end_eh,
};
const struct ata_port_info ata_dummy_port_info = {
.port_ops = &ata_dummy_port_ops,
};
/*
* Utility print functions
*/
void ata_port_printk(const struct ata_port *ap, const char *level,
const char *fmt, ...)
{
struct va_format vaf;
va_list args;
va_start(args, fmt);
vaf.fmt = fmt;
vaf.va = &args;
printk("%sata%u: %pV", level, ap->print_id, &vaf);
va_end(args);
}
EXPORT_SYMBOL(ata_port_printk);
void ata_link_printk(const struct ata_link *link, const char *level,
const char *fmt, ...)
{
struct va_format vaf;
va_list args;
va_start(args, fmt);
vaf.fmt = fmt;
vaf.va = &args;
if (sata_pmp_attached(link->ap) || link->ap->slave_link)
printk("%sata%u.%02u: %pV",
level, link->ap->print_id, link->pmp, &vaf);
else
printk("%sata%u: %pV",
level, link->ap->print_id, &vaf);
va_end(args);
}
EXPORT_SYMBOL(ata_link_printk);
void ata_dev_printk(const struct ata_device *dev, const char *level,
const char *fmt, ...)
{
struct va_format vaf;
va_list args;
va_start(args, fmt);
vaf.fmt = fmt;
vaf.va = &args;
printk("%sata%u.%02u: %pV",
level, dev->link->ap->print_id, dev->link->pmp + dev->devno,
&vaf);
va_end(args);
}
EXPORT_SYMBOL(ata_dev_printk);
void ata_print_version(const struct device *dev, const char *version)
{
dev_printk(KERN_DEBUG, dev, "version %s\n", version);
}
EXPORT_SYMBOL(ata_print_version);
/*
* libata is essentially a library of internal helper functions for
* low-level ATA host controller drivers. As such, the API/ABI is
* likely to change as new drivers are added and updated.
* Do not depend on ABI/API stability.
*/
EXPORT_SYMBOL_GPL(sata_deb_timing_normal);
EXPORT_SYMBOL_GPL(sata_deb_timing_hotplug);
EXPORT_SYMBOL_GPL(sata_deb_timing_long);
libata: implement and use ops inheritance libata lets low level drivers build ata_port_operations table and register it with libata core layer. This allows low level drivers high level of flexibility but also burdens them with lots of boilerplate entries. This becomes worse for drivers which support related similar controllers which differ slightly. They share most of the operations except for a few. However, the driver still needs to list all operations for each variant. This results in large number of duplicate entries, which is not only inefficient but also error-prone as it becomes very difficult to tell what the actual differences are. This duplicate boilerplates all over the low level drivers also make updating the core layer exteremely difficult and error-prone. When compounded with multi-branched development model, it ends up accumulating inconsistencies over time. Some of those inconsistencies cause immediate problems and fixed. Others just remain there dormant making maintenance increasingly difficult. To rectify the problem, this patch implements ata_port_operations inheritance. To allow LLDs to easily re-use their own ops tables overriding only specific methods, this patch implements poor man's class inheritance. An ops table has ->inherits field which can be set to any ops table as long as it doesn't create a loop. When the host is started, the inheritance chain is followed and any operation which isn't specified is taken from the nearest ancestor which has it specified. This operation is called finalization and done only once per an ops table and the LLD doesn't have to do anything special about it other than making the ops table non-const such that libata can update it. libata provides four base ops tables lower drivers can inherit from - base, sata, pmp, sff and bmdma. To avoid overriding these ops accidentaly, these ops are declared const and LLDs should always inherit these instead of using them directly. After finalization, all the ops table are identical before and after the patch except for setting .irq_handler to ata_interrupt in drivers which didn't use to. The .irq_handler doesn't have any actual effect and the field will soon be removed by later patch. * sata_sx4 is still using old style EH and currently doesn't take advantage of ops inheritance. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:49 +08:00
EXPORT_SYMBOL_GPL(ata_base_port_ops);
EXPORT_SYMBOL_GPL(sata_port_ops);
EXPORT_SYMBOL_GPL(ata_dummy_port_ops);
EXPORT_SYMBOL_GPL(ata_dummy_port_info);
EXPORT_SYMBOL_GPL(ata_link_next);
EXPORT_SYMBOL_GPL(ata_dev_next);
EXPORT_SYMBOL_GPL(ata_std_bios_param);
EXPORT_SYMBOL_GPL(ata_scsi_unlock_native_capacity);
EXPORT_SYMBOL_GPL(ata_host_init);
EXPORT_SYMBOL_GPL(ata_host_alloc);
EXPORT_SYMBOL_GPL(ata_host_alloc_pinfo);
libata: implement slave_link Explanation taken from the comment of ata_slave_link_init(). In libata, a port contains links and a link contains devices. There is single host link but if a PMP is attached to it, there can be multiple fan-out links. On SATA, there's usually a single device connected to a link but PATA and SATA controllers emulating TF based interface can have two - master and slave. However, there are a few controllers which don't fit into this abstraction too well - SATA controllers which emulate TF interface with both master and slave devices but also have separate SCR register sets for each device. These controllers need separate links for physical link handling (e.g. onlineness, link speed) but should be treated like a traditional M/S controller for everything else (e.g. command issue, softreset). slave_link is libata's way of handling this class of controllers without impacting core layer too much. For anything other than physical link handling, the default host link is used for both master and slave. For physical link handling, separate @ap->slave_link is used. All dirty details are implemented inside libata core layer. From LLD's POV, the only difference is that prereset, hardreset and postreset are called once more for the slave link, so the reset sequence looks like the following. prereset(M) -> prereset(S) -> hardreset(M) -> hardreset(S) -> softreset(M) -> postreset(M) -> postreset(S) Note that softreset is called only for the master. Softreset resets both M/S by definition, so SRST on master should handle both (the standard method will work just fine). As slave_link excludes PMP support and only code paths which deal with the attributes of physical link are affected, all the changes are localized to libata.h, libata-core.c and libata-eh.c. * ata_is_host_link() updated so that slave_link is considered as host link too. * iterator extended to iterate over the slave_link when using the underbarred version. * force param handling updated such that devno 16 is mapped to the slave link/device. * ata_link_on/offline() updated to return the combined result from master and slave link. ata_phys_link_on/offline() are the direct versions. * EH autopsy and report are performed separately for master slave links. Reset is udpated to implement the above described reset sequence. Except for reset update, most changes are minor, many of them just modifying dev->link to ata_dev_phys_link(dev) or using phys online test instead. After this update, LLDs can take full advantage of per-dev SCR registers by simply turning on slave link. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
2008-07-31 16:02:43 +08:00
EXPORT_SYMBOL_GPL(ata_slave_link_init);
EXPORT_SYMBOL_GPL(ata_host_start);
EXPORT_SYMBOL_GPL(ata_host_register);
EXPORT_SYMBOL_GPL(ata_host_activate);
EXPORT_SYMBOL_GPL(ata_host_detach);
EXPORT_SYMBOL_GPL(ata_sg_init);
EXPORT_SYMBOL_GPL(ata_qc_complete);
EXPORT_SYMBOL_GPL(ata_qc_complete_multiple);
EXPORT_SYMBOL_GPL(atapi_cmd_type);
EXPORT_SYMBOL_GPL(ata_tf_to_fis);
EXPORT_SYMBOL_GPL(ata_tf_from_fis);
EXPORT_SYMBOL_GPL(ata_pack_xfermask);
EXPORT_SYMBOL_GPL(ata_unpack_xfermask);
EXPORT_SYMBOL_GPL(ata_xfer_mask2mode);
EXPORT_SYMBOL_GPL(ata_xfer_mode2mask);
EXPORT_SYMBOL_GPL(ata_xfer_mode2shift);
EXPORT_SYMBOL_GPL(ata_mode_string);
EXPORT_SYMBOL_GPL(ata_id_xfermask);
EXPORT_SYMBOL_GPL(ata_do_set_mode);
EXPORT_SYMBOL_GPL(ata_std_qc_defer);
EXPORT_SYMBOL_GPL(ata_noop_qc_prep);
EXPORT_SYMBOL_GPL(ata_dev_disable);
EXPORT_SYMBOL_GPL(sata_set_spd);
EXPORT_SYMBOL_GPL(ata_wait_after_reset);
EXPORT_SYMBOL_GPL(sata_link_debounce);
EXPORT_SYMBOL_GPL(sata_link_resume);
EXPORT_SYMBOL_GPL(sata_link_scr_lpm);
EXPORT_SYMBOL_GPL(ata_std_prereset);
EXPORT_SYMBOL_GPL(sata_link_hardreset);
EXPORT_SYMBOL_GPL(sata_std_hardreset);
EXPORT_SYMBOL_GPL(ata_std_postreset);
EXPORT_SYMBOL_GPL(ata_dev_classify);
EXPORT_SYMBOL_GPL(ata_dev_pair);
EXPORT_SYMBOL_GPL(ata_ratelimit);
EXPORT_SYMBOL_GPL(ata_msleep);
EXPORT_SYMBOL_GPL(ata_wait_register);
EXPORT_SYMBOL_GPL(ata_scsi_queuecmd);
EXPORT_SYMBOL_GPL(ata_scsi_slave_config);
EXPORT_SYMBOL_GPL(ata_scsi_slave_destroy);
EXPORT_SYMBOL_GPL(ata_scsi_change_queue_depth);
EXPORT_SYMBOL_GPL(__ata_change_queue_depth);
EXPORT_SYMBOL_GPL(sata_scr_valid);
EXPORT_SYMBOL_GPL(sata_scr_read);
EXPORT_SYMBOL_GPL(sata_scr_write);
EXPORT_SYMBOL_GPL(sata_scr_write_flush);
EXPORT_SYMBOL_GPL(ata_link_online);
EXPORT_SYMBOL_GPL(ata_link_offline);
#ifdef CONFIG_PM
EXPORT_SYMBOL_GPL(ata_host_suspend);
EXPORT_SYMBOL_GPL(ata_host_resume);
#endif /* CONFIG_PM */
EXPORT_SYMBOL_GPL(ata_id_string);
EXPORT_SYMBOL_GPL(ata_id_c_string);
EXPORT_SYMBOL_GPL(ata_do_dev_read_id);
EXPORT_SYMBOL_GPL(ata_scsi_simulate);
EXPORT_SYMBOL_GPL(ata_pio_need_iordy);
EXPORT_SYMBOL_GPL(ata_timing_find_mode);
EXPORT_SYMBOL_GPL(ata_timing_compute);
EXPORT_SYMBOL_GPL(ata_timing_merge);
EXPORT_SYMBOL_GPL(ata_timing_cycle2mode);
#ifdef CONFIG_PCI
EXPORT_SYMBOL_GPL(pci_test_config_bits);
EXPORT_SYMBOL_GPL(ata_pci_remove_one);
#ifdef CONFIG_PM
EXPORT_SYMBOL_GPL(ata_pci_device_do_suspend);
EXPORT_SYMBOL_GPL(ata_pci_device_do_resume);
EXPORT_SYMBOL_GPL(ata_pci_device_suspend);
EXPORT_SYMBOL_GPL(ata_pci_device_resume);
#endif /* CONFIG_PM */
#endif /* CONFIG_PCI */
EXPORT_SYMBOL_GPL(ata_platform_remove_one);
EXPORT_SYMBOL_GPL(__ata_ehi_push_desc);
EXPORT_SYMBOL_GPL(ata_ehi_push_desc);
EXPORT_SYMBOL_GPL(ata_ehi_clear_desc);
EXPORT_SYMBOL_GPL(ata_port_desc);
#ifdef CONFIG_PCI
EXPORT_SYMBOL_GPL(ata_port_pbar_desc);
#endif /* CONFIG_PCI */
EXPORT_SYMBOL_GPL(ata_port_schedule_eh);
EXPORT_SYMBOL_GPL(ata_link_abort);
EXPORT_SYMBOL_GPL(ata_port_abort);
EXPORT_SYMBOL_GPL(ata_port_freeze);
EXPORT_SYMBOL_GPL(sata_async_notification);
EXPORT_SYMBOL_GPL(ata_eh_freeze_port);
EXPORT_SYMBOL_GPL(ata_eh_thaw_port);
EXPORT_SYMBOL_GPL(ata_eh_qc_complete);
EXPORT_SYMBOL_GPL(ata_eh_qc_retry);
EXPORT_SYMBOL_GPL(ata_eh_analyze_ncq_error);
EXPORT_SYMBOL_GPL(ata_do_eh);
libata: make reset related methods proper port operations Currently reset methods are not specified directly in the ata_port_operations table. If a LLD wants to use custom reset methods, it should construct and use a error_handler which uses those reset methods. It's done this way for two reasons. First, the ops table already contained too many methods and adding four more of them would noticeably increase the amount of necessary boilerplate code all over low level drivers. Second, as ->error_handler uses those reset methods, it can get confusing. ie. By overriding ->error_handler, those reset ops can be made useless making layering a bit hazy. Now that ops table uses inheritance, the first problem doesn't exist anymore. The second isn't completely solved but is relieved by providing default values - most drivers can just override what it has implemented and don't have to concern itself about higher level callbacks. In fact, there currently is no driver which actually modifies error handling behavior. Drivers which override ->error_handler just wraps the standard error handler only to prepare the controller for EH. I don't think making ops layering strict has any noticeable benefit. This patch makes ->prereset, ->softreset, ->hardreset, ->postreset and their PMP counterparts propoer ops. Default ops are provided in the base ops tables and drivers are converted to override individual reset methods instead of creating custom error_handler. * ata_std_error_handler() doesn't use sata_std_hardreset() if SCRs aren't accessible. sata_promise doesn't need to use separate error_handlers for PATA and SATA anymore. * softreset is broken for sata_inic162x and sata_sx4. As libata now always prefers hardreset, this doesn't really matter but the ops are forced to NULL using ATA_OP_NULL for documentation purpose. * pata_hpt374 needs to use different prereset for the first and second PCI functions. This used to be done by branching from hpt374_error_handler(). The proper way to do this is to use separate ops and port_info tables for each function. Converted. Signed-off-by: Tejun Heo <htejun@gmail.com>
2008-03-25 11:22:50 +08:00
EXPORT_SYMBOL_GPL(ata_std_error_handler);
EXPORT_SYMBOL_GPL(ata_cable_40wire);
EXPORT_SYMBOL_GPL(ata_cable_80wire);
EXPORT_SYMBOL_GPL(ata_cable_unknown);
EXPORT_SYMBOL_GPL(ata_cable_ignore);
EXPORT_SYMBOL_GPL(ata_cable_sata);