linux/drivers/nvdimm/nd.h

681 lines
18 KiB
C
Raw Normal View History

/* SPDX-License-Identifier: GPL-2.0-only */
/*
* Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
*/
#ifndef __ND_H__
#define __ND_H__
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
#include <linux/libnvdimm.h>
#include <linux/badblocks.h>
#include <linux/blkdev.h>
#include <linux/device.h>
#include <linux/mutex.h>
#include <linux/ndctl.h>
#include <linux/types.h>
#include <linux/nd.h>
#include "label.h"
enum {
nd_btt: atomic sector updates BTT stands for Block Translation Table, and is a way to provide power fail sector atomicity semantics for block devices that have the ability to perform byte granularity IO. It relies on the capability of libnvdimm namespace devices to do byte aligned IO. The BTT works as a stacked blocked device, and reserves a chunk of space from the backing device for its accounting metadata. It is a bio-based driver because all IO is done synchronously, and there is no queuing or asynchronous completions at either the device or the driver level. The BTT uses 'lanes' to index into various 'on-disk' data structures, and lanes also act as a synchronization mechanism in case there are more CPUs than available lanes. We did a comparison between two lane lock strategies - first where we kept an atomic counter around that tracked which was the last lane that was used, and 'our' lane was determined by atomically incrementing that. That way, for the nr_cpus > nr_lanes case, theoretically, no CPU would be blocked waiting for a lane. The other strategy was to use the cpu number we're scheduled on to and hash it to a lane number. Theoretically, this could block an IO that could've otherwise run using a different, free lane. But some fio workloads showed that the direct cpu -> lane hash performed faster than tracking 'last lane' - my reasoning is the cache thrash caused by moving the atomic variable made that approach slower than simply waiting out the in-progress IO. This supports the conclusion that the driver can be a very simple bio-based one that does synchronous IOs instead of queuing. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Boaz Harrosh <boaz@plexistor.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jens Axboe <axboe@fb.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Neil Brown <neilb@suse.de> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Greg KH <gregkh@linuxfoundation.org> [jmoyer: fix nmi watchdog timeout in btt_map_init] [jmoyer: move btt initialization to module load path] [jmoyer: fix memory leak in the btt initialization path] [jmoyer: Don't overwrite corrupted arenas] Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 16:20:32 +08:00
/*
* Limits the maximum number of block apertures a dimm can
* support and is an input to the geometry/on-disk-format of a
* BTT instance
*/
ND_MAX_LANES = 256,
INT_LBASIZE_ALIGNMENT = 64,
NVDIMM_IO_ATOMIC = 1,
};
struct nvdimm_drvdata {
struct device *dev;
int nslabel_size;
struct nd_cmd_get_config_size nsarea;
void *data;
bool cxl;
int ns_current, ns_next;
struct resource dpa;
struct kref kref;
};
static inline const u8 *nsl_ref_name(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return nd_label->cxl.name;
return nd_label->efi.name;
}
static inline u8 *nsl_get_name(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label, u8 *name)
{
if (ndd->cxl)
return memcpy(name, nd_label->cxl.name, NSLABEL_NAME_LEN);
return memcpy(name, nd_label->efi.name, NSLABEL_NAME_LEN);
}
static inline u8 *nsl_set_name(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label, u8 *name)
{
if (!name)
return NULL;
if (ndd->cxl)
return memcpy(nd_label->cxl.name, name, NSLABEL_NAME_LEN);
return memcpy(nd_label->efi.name, name, NSLABEL_NAME_LEN);
}
static inline u32 nsl_get_slot(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return __le32_to_cpu(nd_label->cxl.slot);
return __le32_to_cpu(nd_label->efi.slot);
}
static inline void nsl_set_slot(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label, u32 slot)
{
if (ndd->cxl)
nd_label->cxl.slot = __cpu_to_le32(slot);
else
nd_label->efi.slot = __cpu_to_le32(slot);
}
static inline u64 nsl_get_checksum(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return __le64_to_cpu(nd_label->cxl.checksum);
return __le64_to_cpu(nd_label->efi.checksum);
}
static inline void nsl_set_checksum(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
u64 checksum)
{
if (ndd->cxl)
nd_label->cxl.checksum = __cpu_to_le64(checksum);
else
nd_label->efi.checksum = __cpu_to_le64(checksum);
}
static inline u32 nsl_get_flags(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return __le32_to_cpu(nd_label->cxl.flags);
return __le32_to_cpu(nd_label->efi.flags);
}
static inline void nsl_set_flags(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label, u32 flags)
{
if (ndd->cxl)
nd_label->cxl.flags = __cpu_to_le32(flags);
else
nd_label->efi.flags = __cpu_to_le32(flags);
}
static inline u64 nsl_get_dpa(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return __le64_to_cpu(nd_label->cxl.dpa);
return __le64_to_cpu(nd_label->efi.dpa);
}
static inline void nsl_set_dpa(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label, u64 dpa)
{
if (ndd->cxl)
nd_label->cxl.dpa = __cpu_to_le64(dpa);
else
nd_label->efi.dpa = __cpu_to_le64(dpa);
}
static inline u64 nsl_get_rawsize(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return __le64_to_cpu(nd_label->cxl.rawsize);
return __le64_to_cpu(nd_label->efi.rawsize);
}
static inline void nsl_set_rawsize(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
u64 rawsize)
{
if (ndd->cxl)
nd_label->cxl.rawsize = __cpu_to_le64(rawsize);
else
nd_label->efi.rawsize = __cpu_to_le64(rawsize);
}
static inline u64 nsl_get_isetcookie(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
/* WARN future refactor attempts that break this assumption */
if (dev_WARN_ONCE(ndd->dev, ndd->cxl,
"CXL labels do not use the isetcookie concept\n"))
return 0;
return __le64_to_cpu(nd_label->efi.isetcookie);
}
static inline void nsl_set_isetcookie(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
u64 isetcookie)
{
if (!ndd->cxl)
nd_label->efi.isetcookie = __cpu_to_le64(isetcookie);
}
static inline bool nsl_validate_isetcookie(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
u64 cookie)
{
/*
* Let the EFI and CXL validation comingle, where fields that
* don't matter to CXL always validate.
*/
if (ndd->cxl)
return true;
return cookie == __le64_to_cpu(nd_label->efi.isetcookie);
}
static inline u16 nsl_get_position(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return __le16_to_cpu(nd_label->cxl.position);
return __le16_to_cpu(nd_label->efi.position);
}
static inline void nsl_set_position(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
u16 position)
{
if (ndd->cxl)
nd_label->cxl.position = __cpu_to_le16(position);
else
nd_label->efi.position = __cpu_to_le16(position);
}
static inline u16 nsl_get_nlabel(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return 0;
return __le16_to_cpu(nd_label->efi.nlabel);
}
static inline void nsl_set_nlabel(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
u16 nlabel)
{
if (!ndd->cxl)
nd_label->efi.nlabel = __cpu_to_le16(nlabel);
}
static inline u16 nsl_get_nrange(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return __le16_to_cpu(nd_label->cxl.nrange);
return 1;
}
static inline void nsl_set_nrange(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
u16 nrange)
{
if (ndd->cxl)
nd_label->cxl.nrange = __cpu_to_le16(nrange);
}
static inline u64 nsl_get_lbasize(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
/*
* Yes, for some reason the EFI labels convey a massive 64-bit
* lbasize, that got fixed for CXL.
*/
if (ndd->cxl)
return __le16_to_cpu(nd_label->cxl.lbasize);
return __le64_to_cpu(nd_label->efi.lbasize);
}
static inline void nsl_set_lbasize(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
u64 lbasize)
{
if (ndd->cxl)
nd_label->cxl.lbasize = __cpu_to_le16(lbasize);
else
nd_label->efi.lbasize = __cpu_to_le64(lbasize);
}
static inline const uuid_t *nsl_get_uuid(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
uuid_t *uuid)
{
if (ndd->cxl)
import_uuid(uuid, nd_label->cxl.uuid);
else
import_uuid(uuid, nd_label->efi.uuid);
return uuid;
}
static inline const uuid_t *nsl_set_uuid(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
const uuid_t *uuid)
{
if (ndd->cxl)
export_uuid(nd_label->cxl.uuid, uuid);
else
export_uuid(nd_label->efi.uuid, uuid);
return uuid;
}
static inline bool nsl_uuid_equal(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label,
const uuid_t *uuid)
{
uuid_t tmp;
if (ndd->cxl)
import_uuid(&tmp, nd_label->cxl.uuid);
else
import_uuid(&tmp, nd_label->efi.uuid);
return uuid_equal(&tmp, uuid);
}
static inline const u8 *nsl_uuid_raw(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return nd_label->cxl.uuid;
return nd_label->efi.uuid;
}
bool nsl_validate_type_guid(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label, guid_t *guid);
enum nvdimm_claim_class nsl_get_claim_class(struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label);
struct nd_region_data {
int ns_count;
int ns_active;
unsigned int hints_shift;
void __iomem *flush_wpq[];
};
static inline void __iomem *ndrd_get_flush_wpq(struct nd_region_data *ndrd,
int dimm, int hint)
{
unsigned int num = 1 << ndrd->hints_shift;
unsigned int mask = num - 1;
return ndrd->flush_wpq[dimm * num + (hint & mask)];
}
static inline void ndrd_set_flush_wpq(struct nd_region_data *ndrd, int dimm,
int hint, void __iomem *flush)
{
unsigned int num = 1 << ndrd->hints_shift;
unsigned int mask = num - 1;
ndrd->flush_wpq[dimm * num + (hint & mask)] = flush;
}
static inline struct nd_namespace_index *to_namespace_index(
struct nvdimm_drvdata *ndd, int i)
{
if (i < 0)
return NULL;
return ndd->data + sizeof_namespace_index(ndd) * i;
}
static inline struct nd_namespace_index *to_current_namespace_index(
struct nvdimm_drvdata *ndd)
{
return to_namespace_index(ndd, ndd->ns_current);
}
static inline struct nd_namespace_index *to_next_namespace_index(
struct nvdimm_drvdata *ndd)
{
return to_namespace_index(ndd, ndd->ns_next);
}
unsigned sizeof_namespace_label(struct nvdimm_drvdata *ndd);
#define efi_namespace_label_has(ndd, field) \
(!ndd->cxl && offsetof(struct nvdimm_efi_label, field) \
< sizeof_namespace_label(ndd))
#define nd_dbg_dpa(r, d, res, fmt, arg...) \
dev_dbg((r) ? &(r)->dev : (d)->dev, "%s: %.13s: %#llx @ %#llx " fmt, \
(r) ? dev_name((d)->dev) : "", res ? res->name : "null", \
(unsigned long long) (res ? resource_size(res) : 0), \
(unsigned long long) (res ? res->start : 0), ##arg)
#define for_each_dpa_resource(ndd, res) \
for (res = (ndd)->dpa.child; res; res = res->sibling)
#define for_each_dpa_resource_safe(ndd, res, next) \
for (res = (ndd)->dpa.child, next = res ? res->sibling : NULL; \
res; res = next, next = next ? next->sibling : NULL)
nd_btt: atomic sector updates BTT stands for Block Translation Table, and is a way to provide power fail sector atomicity semantics for block devices that have the ability to perform byte granularity IO. It relies on the capability of libnvdimm namespace devices to do byte aligned IO. The BTT works as a stacked blocked device, and reserves a chunk of space from the backing device for its accounting metadata. It is a bio-based driver because all IO is done synchronously, and there is no queuing or asynchronous completions at either the device or the driver level. The BTT uses 'lanes' to index into various 'on-disk' data structures, and lanes also act as a synchronization mechanism in case there are more CPUs than available lanes. We did a comparison between two lane lock strategies - first where we kept an atomic counter around that tracked which was the last lane that was used, and 'our' lane was determined by atomically incrementing that. That way, for the nr_cpus > nr_lanes case, theoretically, no CPU would be blocked waiting for a lane. The other strategy was to use the cpu number we're scheduled on to and hash it to a lane number. Theoretically, this could block an IO that could've otherwise run using a different, free lane. But some fio workloads showed that the direct cpu -> lane hash performed faster than tracking 'last lane' - my reasoning is the cache thrash caused by moving the atomic variable made that approach slower than simply waiting out the in-progress IO. This supports the conclusion that the driver can be a very simple bio-based one that does synchronous IOs instead of queuing. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Boaz Harrosh <boaz@plexistor.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jens Axboe <axboe@fb.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Neil Brown <neilb@suse.de> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Greg KH <gregkh@linuxfoundation.org> [jmoyer: fix nmi watchdog timeout in btt_map_init] [jmoyer: move btt initialization to module load path] [jmoyer: fix memory leak in the btt initialization path] [jmoyer: Don't overwrite corrupted arenas] Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 16:20:32 +08:00
struct nd_percpu_lane {
int count;
spinlock_t lock;
};
libnvdimm/namespace: Fix label tracking error Users have reported intermittent occurrences of DIMM initialization failures due to duplicate allocations of address capacity detected in the labels, or errors of the form below, both have the same root cause. nd namespace1.4: failed to track label: 0 WARNING: CPU: 17 PID: 1381 at drivers/nvdimm/label.c:863 RIP: 0010:__pmem_label_update+0x56c/0x590 [libnvdimm] Call Trace: ? nd_pmem_namespace_label_update+0xd6/0x160 [libnvdimm] nd_pmem_namespace_label_update+0xd6/0x160 [libnvdimm] uuid_store+0x17e/0x190 [libnvdimm] kernfs_fop_write+0xf0/0x1a0 vfs_write+0xb7/0x1b0 ksys_write+0x57/0xd0 do_syscall_64+0x60/0x210 Unfortunately those reports were typically with a busy parallel namespace creation / destruction loop making it difficult to see the components of the bug. However, Jane provided a simple reproducer using the work-in-progress sub-section implementation. When ndctl is reconfiguring a namespace it may take an existing defunct / disabled namespace and reconfigure it with a new uuid and other parameters. Critically namespace_update_uuid() takes existing address resources and renames them for the new namespace to use / reconfigure as it sees fit. The bug is that this rename only happens in the resource tracking tree. Existing labels with the old uuid are not reaped leading to a scenario where multiple active labels reference the same span of address range. Teach namespace_update_uuid() to flag any references to the old uuid for reaping at the next label update attempt. Cc: <stable@vger.kernel.org> Fixes: bf9bccc14c05 ("libnvdimm: pmem label sets and namespace instantiation") Link: https://github.com/pmem/ndctl/issues/91 Reported-by: Jane Chu <jane.chu@oracle.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reported-by: Erwin Tsaur <erwin.tsaur@oracle.com> Cc: Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2019-05-01 12:51:21 +08:00
enum nd_label_flags {
ND_LABEL_REAP,
};
struct nd_label_ent {
struct list_head list;
libnvdimm/namespace: Fix label tracking error Users have reported intermittent occurrences of DIMM initialization failures due to duplicate allocations of address capacity detected in the labels, or errors of the form below, both have the same root cause. nd namespace1.4: failed to track label: 0 WARNING: CPU: 17 PID: 1381 at drivers/nvdimm/label.c:863 RIP: 0010:__pmem_label_update+0x56c/0x590 [libnvdimm] Call Trace: ? nd_pmem_namespace_label_update+0xd6/0x160 [libnvdimm] nd_pmem_namespace_label_update+0xd6/0x160 [libnvdimm] uuid_store+0x17e/0x190 [libnvdimm] kernfs_fop_write+0xf0/0x1a0 vfs_write+0xb7/0x1b0 ksys_write+0x57/0xd0 do_syscall_64+0x60/0x210 Unfortunately those reports were typically with a busy parallel namespace creation / destruction loop making it difficult to see the components of the bug. However, Jane provided a simple reproducer using the work-in-progress sub-section implementation. When ndctl is reconfiguring a namespace it may take an existing defunct / disabled namespace and reconfigure it with a new uuid and other parameters. Critically namespace_update_uuid() takes existing address resources and renames them for the new namespace to use / reconfigure as it sees fit. The bug is that this rename only happens in the resource tracking tree. Existing labels with the old uuid are not reaped leading to a scenario where multiple active labels reference the same span of address range. Teach namespace_update_uuid() to flag any references to the old uuid for reaping at the next label update attempt. Cc: <stable@vger.kernel.org> Fixes: bf9bccc14c05 ("libnvdimm: pmem label sets and namespace instantiation") Link: https://github.com/pmem/ndctl/issues/91 Reported-by: Jane Chu <jane.chu@oracle.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reported-by: Erwin Tsaur <erwin.tsaur@oracle.com> Cc: Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2019-05-01 12:51:21 +08:00
unsigned long flags;
struct nd_namespace_label *label;
};
enum nd_mapping_lock_class {
ND_MAPPING_CLASS0,
ND_MAPPING_UUID_SCAN,
};
struct nd_mapping {
struct nvdimm *nvdimm;
u64 start;
u64 size;
int position;
struct list_head labels;
struct mutex lock;
/*
* @ndd is for private use at region enable / disable time for
* get_ndd() + put_ndd(), all other nd_mapping to ndd
* conversions use to_ndd() which respects enabled state of the
* nvdimm.
*/
struct nvdimm_drvdata *ndd;
};
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
struct nd_region {
struct device dev;
struct ida ns_ida;
struct ida btt_ida;
struct ida pfn_ida;
struct ida dax_ida;
unsigned long flags;
struct device *ns_seed;
struct device *btt_seed;
struct device *pfn_seed;
struct device *dax_seed;
libnvdimm/region: Introduce an 'align' attribute The align attribute applies an alignment constraint for namespace creation in a region. Whereas the 'align' attribute of a namespace applied alignment padding via an info block, the 'align' attribute applies alignment constraints to the free space allocation. The default for 'align' is the maximum known memremap_compat_align() across all archs (16MiB from PowerPC at time of writing) multiplied by the number of interleave ways if there is blk-aliasing. The minimum is PAGE_SIZE and allows for the creation of cross-arch incompatible namespaces, just as previous kernels allowed, but the expectation is cross-arch and mode-independent compatibility by default. The regression risk with this change is limited to cases that were dependent on the ability to create unaligned namespaces, *and* for some reason are unable to opt-out of aligned namespaces by writing to 'regionX/align'. If such a scenario arises the default can be flipped from opt-out to opt-in of compat-aligned namespace creation, but that is a last resort. The kernel will otherwise continue to support existing defined misaligned namespaces. Unfortunately this change needs to touch several parts of the implementation at once: - region/available_size: expand busy extents to current align - region/max_available_extent: expand busy extents to current align - namespace/size: trim free space to current align ...to keep the free space accounting conforming to the dynamic align setting. Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reported-by: Jeff Moyer <jmoyer@redhat.com> Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Jeff Moyer <jmoyer@redhat.com> Link: https://lore.kernel.org/r/158041478371.3889308.14542630147672668068.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2020-01-31 04:06:23 +08:00
unsigned long align;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
u16 ndr_mappings;
u64 ndr_size;
u64 ndr_start;
acpi/nfit, device-dax: Identify differentiated memory with a unique numa-node Persistent memory, as described by the ACPI NFIT (NVDIMM Firmware Interface Table), is the first known instance of a memory range described by a unique "target" proximity domain. Where "initiator" and "target" proximity domains is an approach that the ACPI HMAT (Heterogeneous Memory Attributes Table) uses to described the unique performance properties of a memory range relative to a given initiator (e.g. CPU or DMA device). Currently the numa-node for a /dev/pmemX block-device or /dev/daxX.Y char-device follows the traditional notion of 'numa-node' where the attribute conveys the closest online numa-node. That numa-node attribute is useful for cpu-binding and memory-binding processes *near* the device. However, when the memory range backing a 'pmem', or 'dax' device is onlined (memory hot-add) the memory-only-numa-node representing that address needs to be differentiated from the set of online nodes. In other words, the numa-node association of the device depends on whether you can bind processes *near* the cpu-numa-node in the offline device-case, or bind process *on* the memory-range directly after the backing address range is onlined. Allow for the case that platform firmware describes persistent memory with a unique proximity domain, i.e. when it is distinct from the proximity of DRAM and CPUs that are on the same socket. Plumb the Linux numa-node translation of that proximity through the libnvdimm region device to namespaces that are in device-dax mode. With this in place the proposed kmem driver [1] can optionally discover a unique numa-node number for the address range as it transitions the memory from an offline state managed by a device-driver to an online memory range managed by the core-mm. [1]: https://lore.kernel.org/lkml/20181022201317.8558C1D8@viggo.jf.intel.com Reported-by: Fan Du <fan.du@intel.com> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: "Oliver O'Halloran" <oohall@gmail.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Jérôme Glisse <jglisse@redhat.com> Reviewed-by: Yang Shi <yang.shi@linux.alibaba.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2018-11-10 04:43:07 +08:00
int id, num_lanes, ro, numa_node, target_node;
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
void *provider_data;
struct kernfs_node *bb_state;
struct badblocks bb;
2015-05-02 01:11:27 +08:00
struct nd_interleave_set *nd_set;
nd_btt: atomic sector updates BTT stands for Block Translation Table, and is a way to provide power fail sector atomicity semantics for block devices that have the ability to perform byte granularity IO. It relies on the capability of libnvdimm namespace devices to do byte aligned IO. The BTT works as a stacked blocked device, and reserves a chunk of space from the backing device for its accounting metadata. It is a bio-based driver because all IO is done synchronously, and there is no queuing or asynchronous completions at either the device or the driver level. The BTT uses 'lanes' to index into various 'on-disk' data structures, and lanes also act as a synchronization mechanism in case there are more CPUs than available lanes. We did a comparison between two lane lock strategies - first where we kept an atomic counter around that tracked which was the last lane that was used, and 'our' lane was determined by atomically incrementing that. That way, for the nr_cpus > nr_lanes case, theoretically, no CPU would be blocked waiting for a lane. The other strategy was to use the cpu number we're scheduled on to and hash it to a lane number. Theoretically, this could block an IO that could've otherwise run using a different, free lane. But some fio workloads showed that the direct cpu -> lane hash performed faster than tracking 'last lane' - my reasoning is the cache thrash caused by moving the atomic variable made that approach slower than simply waiting out the in-progress IO. This supports the conclusion that the driver can be a very simple bio-based one that does synchronous IOs instead of queuing. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Boaz Harrosh <boaz@plexistor.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jens Axboe <axboe@fb.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Neil Brown <neilb@suse.de> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Greg KH <gregkh@linuxfoundation.org> [jmoyer: fix nmi watchdog timeout in btt_map_init] [jmoyer: move btt initialization to module load path] [jmoyer: fix memory leak in the btt initialization path] [jmoyer: Don't overwrite corrupted arenas] Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 16:20:32 +08:00
struct nd_percpu_lane __percpu *lane;
int (*flush)(struct nd_region *nd_region, struct bio *bio);
struct nd_mapping mapping[];
libnvdimm, nfit: regions (block-data-window, persistent memory, volatile memory) A "region" device represents the maximum capacity of a BLK range (mmio block-data-window(s)), or a PMEM range (DAX-capable persistent memory or volatile memory), without regard for aliasing. Aliasing, in the dimm-local address space (DPA), is resolved by metadata on a dimm to designate which exclusive interface will access the aliased DPA ranges. Support for the per-dimm metadata/label arrvies is in a subsequent patch. The name format of "region" devices is "regionN" where, like dimms, N is a global ida index assigned at discovery time. This id is not reliable across reboots nor in the presence of hotplug. Look to attributes of the region or static id-data of the sub-namespace to generate a persistent name. However, if the platform configuration does not change it is reasonable to expect the same region id to be assigned at the next boot. "region"s have 2 generic attributes "size", and "mapping"s where: - size: the BLK accessible capacity or the span of the system physical address range in the case of PMEM. - mappingN: a tuple describing a dimm's contribution to the region's capacity in the format (<nmemX>,<dpa>,<size>). For a PMEM-region there will be at least one mapping per dimm in the interleave set. For a BLK-region there is only "mapping0" listing the starting DPA of the BLK-region and the available DPA capacity of that space (matches "size" above). The max number of mappings per "region" is hard coded per the constraints of sysfs attribute groups. That said the number of mappings per region should never exceed the maximum number of possible dimms in the system. If the current number turns out to not be enough then the "mappings" attribute clarifies how many there are supposed to be. "32 should be enough for anybody...". Cc: Neil Brown <neilb@suse.de> Cc: <linux-acpi@vger.kernel.org> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Robert Moore <robert.moore@intel.com> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Christoph Hellwig <hch@lst.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-10 08:13:14 +08:00
};
static inline bool nsl_validate_nlabel(struct nd_region *nd_region,
struct nvdimm_drvdata *ndd,
struct nd_namespace_label *nd_label)
{
if (ndd->cxl)
return true;
return nsl_get_nlabel(ndd, nd_label) == nd_region->ndr_mappings;
}
/*
* Lookup next in the repeating sequence of 01, 10, and 11.
*/
static inline unsigned nd_inc_seq(unsigned seq)
{
static const unsigned next[] = { 0, 2, 3, 1 };
return next[seq & 3];
}
nd_btt: atomic sector updates BTT stands for Block Translation Table, and is a way to provide power fail sector atomicity semantics for block devices that have the ability to perform byte granularity IO. It relies on the capability of libnvdimm namespace devices to do byte aligned IO. The BTT works as a stacked blocked device, and reserves a chunk of space from the backing device for its accounting metadata. It is a bio-based driver because all IO is done synchronously, and there is no queuing or asynchronous completions at either the device or the driver level. The BTT uses 'lanes' to index into various 'on-disk' data structures, and lanes also act as a synchronization mechanism in case there are more CPUs than available lanes. We did a comparison between two lane lock strategies - first where we kept an atomic counter around that tracked which was the last lane that was used, and 'our' lane was determined by atomically incrementing that. That way, for the nr_cpus > nr_lanes case, theoretically, no CPU would be blocked waiting for a lane. The other strategy was to use the cpu number we're scheduled on to and hash it to a lane number. Theoretically, this could block an IO that could've otherwise run using a different, free lane. But some fio workloads showed that the direct cpu -> lane hash performed faster than tracking 'last lane' - my reasoning is the cache thrash caused by moving the atomic variable made that approach slower than simply waiting out the in-progress IO. This supports the conclusion that the driver can be a very simple bio-based one that does synchronous IOs instead of queuing. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Boaz Harrosh <boaz@plexistor.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jens Axboe <axboe@fb.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Neil Brown <neilb@suse.de> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Greg KH <gregkh@linuxfoundation.org> [jmoyer: fix nmi watchdog timeout in btt_map_init] [jmoyer: move btt initialization to module load path] [jmoyer: fix memory leak in the btt initialization path] [jmoyer: Don't overwrite corrupted arenas] Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 16:20:32 +08:00
struct btt;
struct nd_btt {
struct device dev;
struct nd_namespace_common *ndns;
nd_btt: atomic sector updates BTT stands for Block Translation Table, and is a way to provide power fail sector atomicity semantics for block devices that have the ability to perform byte granularity IO. It relies on the capability of libnvdimm namespace devices to do byte aligned IO. The BTT works as a stacked blocked device, and reserves a chunk of space from the backing device for its accounting metadata. It is a bio-based driver because all IO is done synchronously, and there is no queuing or asynchronous completions at either the device or the driver level. The BTT uses 'lanes' to index into various 'on-disk' data structures, and lanes also act as a synchronization mechanism in case there are more CPUs than available lanes. We did a comparison between two lane lock strategies - first where we kept an atomic counter around that tracked which was the last lane that was used, and 'our' lane was determined by atomically incrementing that. That way, for the nr_cpus > nr_lanes case, theoretically, no CPU would be blocked waiting for a lane. The other strategy was to use the cpu number we're scheduled on to and hash it to a lane number. Theoretically, this could block an IO that could've otherwise run using a different, free lane. But some fio workloads showed that the direct cpu -> lane hash performed faster than tracking 'last lane' - my reasoning is the cache thrash caused by moving the atomic variable made that approach slower than simply waiting out the in-progress IO. This supports the conclusion that the driver can be a very simple bio-based one that does synchronous IOs instead of queuing. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Boaz Harrosh <boaz@plexistor.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jens Axboe <axboe@fb.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Neil Brown <neilb@suse.de> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Greg KH <gregkh@linuxfoundation.org> [jmoyer: fix nmi watchdog timeout in btt_map_init] [jmoyer: move btt initialization to module load path] [jmoyer: fix memory leak in the btt initialization path] [jmoyer: Don't overwrite corrupted arenas] Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 16:20:32 +08:00
struct btt *btt;
unsigned long lbasize;
u64 size;
uuid_t *uuid;
int id;
int initial_offset;
u16 version_major;
u16 version_minor;
};
enum nd_pfn_mode {
PFN_MODE_NONE,
PFN_MODE_RAM,
PFN_MODE_PMEM,
};
struct nd_pfn {
int id;
uuid_t *uuid;
struct device dev;
unsigned long align;
unsigned long npfns;
enum nd_pfn_mode mode;
struct nd_pfn_sb *pfn_sb;
struct nd_namespace_common *ndns;
};
struct nd_dax {
struct nd_pfn nd_pfn;
};
static inline u32 nd_info_block_reserve(void)
{
return ALIGN(SZ_8K, PAGE_SIZE);
}
enum nd_async_mode {
ND_SYNC,
ND_ASYNC,
};
int nd_integrity_init(struct gendisk *disk, unsigned long meta_size);
void wait_nvdimm_bus_probe_idle(struct device *dev);
void nd_device_register(struct device *dev);
void nd_device_unregister(struct device *dev, enum nd_async_mode mode);
void nd_device_notify(struct device *dev, enum nvdimm_event event);
int nd_uuid_store(struct device *dev, uuid_t **uuid_out, const char *buf,
size_t len);
ssize_t nd_size_select_show(unsigned long current_size,
const unsigned long *supported, char *buf);
ssize_t nd_size_select_store(struct device *dev, const char *buf,
unsigned long *current_size, const unsigned long *supported);
int __init nvdimm_init(void);
int __init nd_region_init(void);
int __init nd_label_init(void);
void nvdimm_exit(void);
void nd_region_exit(void);
struct nvdimm;
extern const struct attribute_group nd_device_attribute_group;
extern const struct attribute_group nd_numa_attribute_group;
extern const struct attribute_group *nvdimm_bus_attribute_groups[];
struct nvdimm_drvdata *to_ndd(struct nd_mapping *nd_mapping);
int nvdimm_check_config_data(struct device *dev);
int nvdimm_init_nsarea(struct nvdimm_drvdata *ndd);
int nvdimm_init_config_data(struct nvdimm_drvdata *ndd);
int nvdimm_get_config_data(struct nvdimm_drvdata *ndd, void *buf,
size_t offset, size_t len);
int nvdimm_set_config_data(struct nvdimm_drvdata *ndd, size_t offset,
void *buf, size_t len);
long nvdimm_clear_poison(struct device *dev, phys_addr_t phys,
unsigned int len);
void nvdimm_set_labeling(struct device *dev);
void nvdimm_set_locked(struct device *dev);
void nvdimm_clear_locked(struct device *dev);
int nvdimm_security_setup_events(struct device *dev);
#if IS_ENABLED(CONFIG_NVDIMM_KEYS)
int nvdimm_security_unlock(struct device *dev);
#else
static inline int nvdimm_security_unlock(struct device *dev)
{
return 0;
}
#endif
struct nd_btt *to_nd_btt(struct device *dev);
struct nd_gen_sb {
char reserved[SZ_4K - 8];
__le64 checksum;
};
u64 nd_sb_checksum(struct nd_gen_sb *sb);
#if IS_ENABLED(CONFIG_BTT)
int nd_btt_probe(struct device *dev, struct nd_namespace_common *ndns);
bool is_nd_btt(struct device *dev);
struct device *nd_btt_create(struct nd_region *nd_region);
#else
static inline int nd_btt_probe(struct device *dev,
struct nd_namespace_common *ndns)
{
return -ENODEV;
}
static inline bool is_nd_btt(struct device *dev)
{
return false;
}
static inline struct device *nd_btt_create(struct nd_region *nd_region)
{
return NULL;
}
#endif
struct nd_pfn *to_nd_pfn(struct device *dev);
#if IS_ENABLED(CONFIG_NVDIMM_PFN)
libnvdimm/dax: Pick the right alignment default when creating dax devices Allow arch to provide the supported alignments and use hugepage alignment only if we support hugepage. Right now we depend on compile time configs whereas this patch switch this to runtime discovery. Architectures like ppc64 can have THP enabled in code, but then can have hugepage size disabled by the hypervisor. This allows us to create dax devices with PAGE_SIZE alignment in this case. Existing dax namespace with alignment larger than PAGE_SIZE will fail to initialize in this specific case. We still allow fsdax namespace initialization. With respect to identifying whether to enable hugepage fault for a dax device, if THP is enabled during compile, we default to taking hugepage fault and in dax fault handler if we find the fault size > alignment we retry with PAGE_SIZE fault size. This also addresses the below failure scenario on ppc64 ndctl create-namespace --mode=devdax | grep align "align":16777216, "align":16777216 cat /sys/devices/ndbus0/region0/dax0.0/supported_alignments 65536 16777216 daxio.static-debug -z -o /dev/dax0.0 Bus error (core dumped) $ dmesg | tail lpar: Failed hash pte insert with error -4 hash-mmu: mm: Hashing failure ! EA=0x7fff17000000 access=0x8000000000000006 current=daxio hash-mmu: trap=0x300 vsid=0x22cb7a3 ssize=1 base psize=2 psize 10 pte=0xc000000501002b86 daxio[3860]: bus error (7) at 7fff17000000 nip 7fff973c007c lr 7fff973bff34 code 2 in libpmem.so.1.0.0[7fff973b0000+20000] daxio[3860]: code: 792945e4 7d494b78 e95f0098 7d494b78 f93f00a0 4800012c e93f0088 f93f0120 daxio[3860]: code: e93f00a0 f93f0128 e93f0120 e95f0128 <f9490000> e93f0088 39290008 f93f0110 The failure was due to guest kernel using wrong page size. The namespaces created with 16M alignment will appear as below on a config with 16M page size disabled. $ ndctl list -Ni [ { "dev":"namespace0.1", "mode":"fsdax", "map":"dev", "size":5351931904, "uuid":"fc6e9667-461a-4718-82b4-69b24570bddb", "align":16777216, "blockdev":"pmem0.1", "supported_alignments":[ 65536 ] }, { "dev":"namespace0.0", "mode":"fsdax", <==== devdax 16M alignment marked disabled. "map":"mem", "size":5368709120, "uuid":"a4bdf81a-f2ee-4bc6-91db-7b87eddd0484", "state":"disabled" } ] Cc: linux-mm@kvack.org Cc: "Kirill A. Shutemov" <kirill@shutemov.name> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Link: https://lore.kernel.org/r/20190905154603.10349-8-aneesh.kumar@linux.ibm.com Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2019-09-05 23:46:03 +08:00
#define MAX_NVDIMM_ALIGN 4
int nd_pfn_probe(struct device *dev, struct nd_namespace_common *ndns);
bool is_nd_pfn(struct device *dev);
struct device *nd_pfn_create(struct nd_region *nd_region);
struct device *nd_pfn_devinit(struct nd_pfn *nd_pfn,
struct nd_namespace_common *ndns);
int nd_pfn_validate(struct nd_pfn *nd_pfn, const char *sig);
extern const struct attribute_group *nd_pfn_attribute_groups[];
#else
static inline int nd_pfn_probe(struct device *dev,
struct nd_namespace_common *ndns)
{
return -ENODEV;
}
static inline bool is_nd_pfn(struct device *dev)
{
return false;
}
static inline struct device *nd_pfn_create(struct nd_region *nd_region)
{
return NULL;
}
static inline int nd_pfn_validate(struct nd_pfn *nd_pfn, const char *sig)
{
return -ENODEV;
}
#endif
struct nd_dax *to_nd_dax(struct device *dev);
#if IS_ENABLED(CONFIG_NVDIMM_DAX)
int nd_dax_probe(struct device *dev, struct nd_namespace_common *ndns);
bool is_nd_dax(struct device *dev);
struct device *nd_dax_create(struct nd_region *nd_region);
#else
static inline int nd_dax_probe(struct device *dev,
struct nd_namespace_common *ndns)
{
return -ENODEV;
}
static inline bool is_nd_dax(struct device *dev)
{
return false;
}
static inline struct device *nd_dax_create(struct nd_region *nd_region)
{
return NULL;
}
#endif
int nd_region_to_nstype(struct nd_region *nd_region);
int nd_region_register_namespaces(struct nd_region *nd_region, int *err);
u64 nd_region_interleave_set_cookie(struct nd_region *nd_region,
struct nd_namespace_index *nsindex);
2017-03-01 10:32:48 +08:00
u64 nd_region_interleave_set_altcookie(struct nd_region *nd_region);
void nvdimm_bus_lock(struct device *dev);
void nvdimm_bus_unlock(struct device *dev);
bool is_nvdimm_bus_locked(struct device *dev);
void nvdimm_check_and_set_ro(struct gendisk *disk);
void nvdimm_drvdata_release(struct kref *kref);
void put_ndd(struct nvdimm_drvdata *ndd);
int nd_label_reserve_dpa(struct nvdimm_drvdata *ndd);
void nvdimm_free_dpa(struct nvdimm_drvdata *ndd, struct resource *res);
struct resource *nvdimm_allocate_dpa(struct nvdimm_drvdata *ndd,
struct nd_label_id *label_id, resource_size_t start,
resource_size_t n);
resource_size_t nvdimm_namespace_capacity(struct nd_namespace_common *ndns);
bool nvdimm_namespace_locked(struct nd_namespace_common *ndns);
struct nd_namespace_common *nvdimm_namespace_common_probe(struct device *dev);
nd_btt: atomic sector updates BTT stands for Block Translation Table, and is a way to provide power fail sector atomicity semantics for block devices that have the ability to perform byte granularity IO. It relies on the capability of libnvdimm namespace devices to do byte aligned IO. The BTT works as a stacked blocked device, and reserves a chunk of space from the backing device for its accounting metadata. It is a bio-based driver because all IO is done synchronously, and there is no queuing or asynchronous completions at either the device or the driver level. The BTT uses 'lanes' to index into various 'on-disk' data structures, and lanes also act as a synchronization mechanism in case there are more CPUs than available lanes. We did a comparison between two lane lock strategies - first where we kept an atomic counter around that tracked which was the last lane that was used, and 'our' lane was determined by atomically incrementing that. That way, for the nr_cpus > nr_lanes case, theoretically, no CPU would be blocked waiting for a lane. The other strategy was to use the cpu number we're scheduled on to and hash it to a lane number. Theoretically, this could block an IO that could've otherwise run using a different, free lane. But some fio workloads showed that the direct cpu -> lane hash performed faster than tracking 'last lane' - my reasoning is the cache thrash caused by moving the atomic variable made that approach slower than simply waiting out the in-progress IO. This supports the conclusion that the driver can be a very simple bio-based one that does synchronous IOs instead of queuing. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Boaz Harrosh <boaz@plexistor.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jens Axboe <axboe@fb.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Neil Brown <neilb@suse.de> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Greg KH <gregkh@linuxfoundation.org> [jmoyer: fix nmi watchdog timeout in btt_map_init] [jmoyer: move btt initialization to module load path] [jmoyer: fix memory leak in the btt initialization path] [jmoyer: Don't overwrite corrupted arenas] Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 16:20:32 +08:00
int nvdimm_namespace_attach_btt(struct nd_namespace_common *ndns);
int nvdimm_namespace_detach_btt(struct nd_btt *nd_btt);
nd_btt: atomic sector updates BTT stands for Block Translation Table, and is a way to provide power fail sector atomicity semantics for block devices that have the ability to perform byte granularity IO. It relies on the capability of libnvdimm namespace devices to do byte aligned IO. The BTT works as a stacked blocked device, and reserves a chunk of space from the backing device for its accounting metadata. It is a bio-based driver because all IO is done synchronously, and there is no queuing or asynchronous completions at either the device or the driver level. The BTT uses 'lanes' to index into various 'on-disk' data structures, and lanes also act as a synchronization mechanism in case there are more CPUs than available lanes. We did a comparison between two lane lock strategies - first where we kept an atomic counter around that tracked which was the last lane that was used, and 'our' lane was determined by atomically incrementing that. That way, for the nr_cpus > nr_lanes case, theoretically, no CPU would be blocked waiting for a lane. The other strategy was to use the cpu number we're scheduled on to and hash it to a lane number. Theoretically, this could block an IO that could've otherwise run using a different, free lane. But some fio workloads showed that the direct cpu -> lane hash performed faster than tracking 'last lane' - my reasoning is the cache thrash caused by moving the atomic variable made that approach slower than simply waiting out the in-progress IO. This supports the conclusion that the driver can be a very simple bio-based one that does synchronous IOs instead of queuing. Cc: Andy Lutomirski <luto@amacapital.net> Cc: Boaz Harrosh <boaz@plexistor.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jens Axboe <axboe@fb.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Neil Brown <neilb@suse.de> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Greg KH <gregkh@linuxfoundation.org> [jmoyer: fix nmi watchdog timeout in btt_map_init] [jmoyer: move btt initialization to module load path] [jmoyer: fix memory leak in the btt initialization path] [jmoyer: Don't overwrite corrupted arenas] Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 16:20:32 +08:00
const char *nvdimm_namespace_disk_name(struct nd_namespace_common *ndns,
char *name);
unsigned int pmem_sector_size(struct nd_namespace_common *ndns);
mm/memremap_pages: convert to 'struct range' The 'struct resource' in 'struct dev_pagemap' is only used for holding resource span information. The other fields, 'name', 'flags', 'desc', 'parent', 'sibling', and 'child' are all unused wasted space. This is in preparation for introducing a multi-range extension of devm_memremap_pages(). The bulk of this change is unwinding all the places internal to libnvdimm that used 'struct resource' unnecessarily, and replacing instances of 'struct dev_pagemap'.res with 'struct dev_pagemap'.range. P2PDMA had a minor usage of the resource flags field, but only to report failures with "%pR". That is replaced with an open coded print of the range. [dan.carpenter@oracle.com: mm/hmm/test: use after free in dmirror_allocate_chunk()] Link: https://lkml.kernel.org/r/20200926121402.GA7467@kadam Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> [xen] Cc: Paul Mackerras <paulus@ozlabs.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Vishal Verma <vishal.l.verma@intel.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Ben Skeggs <bskeggs@redhat.com> Cc: David Airlie <airlied@linux.ie> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Ira Weiny <ira.weiny@intel.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Cc: Juergen Gross <jgross@suse.com> Cc: Stefano Stabellini <sstabellini@kernel.org> Cc: "Jérôme Glisse" <jglisse@redhat.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Ard Biesheuvel <ardb@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brice Goglin <Brice.Goglin@inria.fr> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: David Hildenbrand <david@redhat.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Hulk Robot <hulkci@huawei.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jason Gunthorpe <jgg@mellanox.com> Cc: Jason Yan <yanaijie@huawei.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Jia He <justin.he@arm.com> Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com> Cc: kernel test robot <lkp@intel.com> Cc: Mike Rapoport <rppt@linux.ibm.com> Cc: Pavel Tatashin <pasha.tatashin@soleen.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tom Lendacky <thomas.lendacky@amd.com> Cc: Wei Yang <richard.weiyang@linux.alibaba.com> Cc: Will Deacon <will@kernel.org> Link: https://lkml.kernel.org/r/159643103173.4062302.768998885691711532.stgit@dwillia2-desk3.amr.corp.intel.com Link: https://lkml.kernel.org/r/160106115761.30709.13539840236873663620.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-14 07:50:29 +08:00
struct range;
void nvdimm_badblocks_populate(struct nd_region *nd_region,
mm/memremap_pages: convert to 'struct range' The 'struct resource' in 'struct dev_pagemap' is only used for holding resource span information. The other fields, 'name', 'flags', 'desc', 'parent', 'sibling', and 'child' are all unused wasted space. This is in preparation for introducing a multi-range extension of devm_memremap_pages(). The bulk of this change is unwinding all the places internal to libnvdimm that used 'struct resource' unnecessarily, and replacing instances of 'struct dev_pagemap'.res with 'struct dev_pagemap'.range. P2PDMA had a minor usage of the resource flags field, but only to report failures with "%pR". That is replaced with an open coded print of the range. [dan.carpenter@oracle.com: mm/hmm/test: use after free in dmirror_allocate_chunk()] Link: https://lkml.kernel.org/r/20200926121402.GA7467@kadam Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> [xen] Cc: Paul Mackerras <paulus@ozlabs.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Vishal Verma <vishal.l.verma@intel.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Ben Skeggs <bskeggs@redhat.com> Cc: David Airlie <airlied@linux.ie> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Ira Weiny <ira.weiny@intel.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Cc: Juergen Gross <jgross@suse.com> Cc: Stefano Stabellini <sstabellini@kernel.org> Cc: "Jérôme Glisse" <jglisse@redhat.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Ard Biesheuvel <ardb@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brice Goglin <Brice.Goglin@inria.fr> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: David Hildenbrand <david@redhat.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Hulk Robot <hulkci@huawei.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jason Gunthorpe <jgg@mellanox.com> Cc: Jason Yan <yanaijie@huawei.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Jia He <justin.he@arm.com> Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com> Cc: kernel test robot <lkp@intel.com> Cc: Mike Rapoport <rppt@linux.ibm.com> Cc: Pavel Tatashin <pasha.tatashin@soleen.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tom Lendacky <thomas.lendacky@amd.com> Cc: Wei Yang <richard.weiyang@linux.alibaba.com> Cc: Will Deacon <will@kernel.org> Link: https://lkml.kernel.org/r/159643103173.4062302.768998885691711532.stgit@dwillia2-desk3.amr.corp.intel.com Link: https://lkml.kernel.org/r/160106115761.30709.13539840236873663620.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-14 07:50:29 +08:00
struct badblocks *bb, const struct range *range);
int devm_namespace_enable(struct device *dev, struct nd_namespace_common *ndns,
resource_size_t size);
void devm_namespace_disable(struct device *dev,
struct nd_namespace_common *ndns);
#if IS_ENABLED(CONFIG_ND_CLAIM)
/* max struct page size independent of kernel config */
#define MAX_STRUCT_PAGE_SIZE 64
int nvdimm_setup_pfn(struct nd_pfn *nd_pfn, struct dev_pagemap *pgmap);
#else
static inline int nvdimm_setup_pfn(struct nd_pfn *nd_pfn,
struct dev_pagemap *pgmap)
{
return -ENXIO;
}
#endif
int nd_region_activate(struct nd_region *nd_region);
static inline bool is_bad_pmem(struct badblocks *bb, sector_t sector,
unsigned int len)
{
if (bb->count) {
sector_t first_bad;
int num_bad;
return !!badblocks_check(bb, sector, len / 512, &first_bad,
&num_bad);
}
return false;
}
const uuid_t *nd_dev_to_uuid(struct device *dev);
bool pmem_should_map_pages(struct device *dev);
#endif /* __ND_H__ */