linux/drivers/usb/gadget/function/f_mass_storage.c

3469 lines
93 KiB
C
Raw Normal View History

// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause)
/*
* f_mass_storage.c -- Mass Storage USB Composite Function
*
* Copyright (C) 2003-2008 Alan Stern
* Copyright (C) 2009 Samsung Electronics
* Author: Michal Nazarewicz <mina86@mina86.com>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions, and the following disclaimer,
* without modification.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The names of the above-listed copyright holders may not be used
* to endorse or promote products derived from this software without
* specific prior written permission.
*
* ALTERNATIVELY, this software may be distributed under the terms of the
* GNU General Public License ("GPL") as published by the Free Software
* Foundation, either version 2 of that License or (at your option) any
* later version.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
* IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
* THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/*
* The Mass Storage Function acts as a USB Mass Storage device,
* appearing to the host as a disk drive or as a CD-ROM drive. In
* addition to providing an example of a genuinely useful composite
* function for a USB device, it also illustrates a technique of
* double-buffering for increased throughput.
*
* For more information about MSF and in particular its module
* parameters and sysfs interface read the
* <Documentation/usb/mass-storage.txt> file.
*/
/*
* MSF is configured by specifying a fsg_config structure. It has the
* following fields:
*
* nluns Number of LUNs function have (anywhere from 1
* to FSG_MAX_LUNS).
* luns An array of LUN configuration values. This
* should be filled for each LUN that
* function will include (ie. for "nluns"
* LUNs). Each element of the array has
* the following fields:
* ->filename The path to the backing file for the LUN.
* Required if LUN is not marked as
* removable.
* ->ro Flag specifying access to the LUN shall be
* read-only. This is implied if CD-ROM
* emulation is enabled as well as when
* it was impossible to open "filename"
* in R/W mode.
* ->removable Flag specifying that LUN shall be indicated as
* being removable.
* ->cdrom Flag specifying that LUN shall be reported as
* being a CD-ROM.
* ->nofua Flag specifying that FUA flag in SCSI WRITE(10,12)
* commands for this LUN shall be ignored.
*
* vendor_name
* product_name
* release Information used as a reply to INQUIRY
* request. To use default set to NULL,
* NULL, 0xffff respectively. The first
* field should be 8 and the second 16
* characters or less.
*
* can_stall Set to permit function to halt bulk endpoints.
* Disabled on some USB devices known not
* to work correctly. You should set it
* to true.
*
* If "removable" is not set for a LUN then a backing file must be
* specified. If it is set, then NULL filename means the LUN's medium
* is not loaded (an empty string as "filename" in the fsg_config
* structure causes error). The CD-ROM emulation includes a single
* data track and no audio tracks; hence there need be only one
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
* backing file per LUN.
*
* This function is heavily based on "File-backed Storage Gadget" by
* Alan Stern which in turn is heavily based on "Gadget Zero" by David
* Brownell. The driver's SCSI command interface was based on the
* "Information technology - Small Computer System Interface - 2"
* document from X3T9.2 Project 375D, Revision 10L, 7-SEP-93,
* available at <http://www.t10.org/ftp/t10/drafts/s2/s2-r10l.pdf>.
* The single exception is opcode 0x23 (READ FORMAT CAPACITIES), which
* was based on the "Universal Serial Bus Mass Storage Class UFI
* Command Specification" document, Revision 1.0, December 14, 1998,
* available at
* <http://www.usb.org/developers/devclass_docs/usbmass-ufi10.pdf>.
*/
/*
* Driver Design
*
* The MSF is fairly straightforward. There is a main kernel
* thread that handles most of the work. Interrupt routines field
* callbacks from the controller driver: bulk- and interrupt-request
* completion notifications, endpoint-0 events, and disconnect events.
* Completion events are passed to the main thread by wakeup calls. Many
* ep0 requests are handled at interrupt time, but SetInterface,
* SetConfiguration, and device reset requests are forwarded to the
* thread in the form of "exceptions" using SIGUSR1 signals (since they
* should interrupt any ongoing file I/O operations).
*
* The thread's main routine implements the standard command/data/status
* parts of a SCSI interaction. It and its subroutines are full of tests
* for pending signals/exceptions -- all this polling is necessary since
* the kernel has no setjmp/longjmp equivalents. (Maybe this is an
* indication that the driver really wants to be running in userspace.)
* An important point is that so long as the thread is alive it keeps an
* open reference to the backing file. This will prevent unmounting
* the backing file's underlying filesystem and could cause problems
* during system shutdown, for example. To prevent such problems, the
* thread catches INT, TERM, and KILL signals and converts them into
* an EXIT exception.
*
* In normal operation the main thread is started during the gadget's
* fsg_bind() callback and stopped during fsg_unbind(). But it can
* also exit when it receives a signal, and there's no point leaving
* the gadget running when the thread is dead. As of this moment, MSF
* provides no way to deregister the gadget when thread dies -- maybe
* a callback functions is needed.
*
* To provide maximum throughput, the driver uses a circular pipeline of
* buffer heads (struct fsg_buffhd). In principle the pipeline can be
* arbitrarily long; in practice the benefits don't justify having more
* than 2 stages (i.e., double buffering). But it helps to think of the
* pipeline as being a long one. Each buffer head contains a bulk-in and
* a bulk-out request pointer (since the buffer can be used for both
* output and input -- directions always are given from the host's
* point of view) as well as a pointer to the buffer and various state
* variables.
*
* Use of the pipeline follows a simple protocol. There is a variable
* (fsg->next_buffhd_to_fill) that points to the next buffer head to use.
* At any time that buffer head may still be in use from an earlier
* request, so each buffer head has a state variable indicating whether
* it is EMPTY, FULL, or BUSY. Typical use involves waiting for the
* buffer head to be EMPTY, filling the buffer either by file I/O or by
* USB I/O (during which the buffer head is BUSY), and marking the buffer
* head FULL when the I/O is complete. Then the buffer will be emptied
* (again possibly by USB I/O, during which it is marked BUSY) and
* finally marked EMPTY again (possibly by a completion routine).
*
* A module parameter tells the driver to avoid stalling the bulk
* endpoints wherever the transport specification allows. This is
* necessary for some UDCs like the SuperH, which cannot reliably clear a
* halt on a bulk endpoint. However, under certain circumstances the
* Bulk-only specification requires a stall. In such cases the driver
* will halt the endpoint and set a flag indicating that it should clear
* the halt in software during the next device reset. Hopefully this
* will permit everything to work correctly. Furthermore, although the
* specification allows the bulk-out endpoint to halt when the host sends
* too much data, implementing this would cause an unavoidable race.
* The driver will always use the "no-stall" approach for OUT transfers.
*
* One subtle point concerns sending status-stage responses for ep0
* requests. Some of these requests, such as device reset, can involve
* interrupting an ongoing file I/O operation, which might take an
* arbitrarily long time. During that delay the host might give up on
* the original ep0 request and issue a new one. When that happens the
* driver should not notify the host about completion of the original
* request, as the host will no longer be waiting for it. So the driver
* assigns to each ep0 request a unique tag, and it keeps track of the
* tag value of the request associated with a long-running exception
* (device-reset, interface-change, or configuration-change). When the
* exception handler is finished, the status-stage response is submitted
* only if the current ep0 request tag is equal to the exception request
* tag. Thus only the most recently received ep0 request will get a
* status-stage response.
*
* Warning: This driver source file is too long. It ought to be split up
* into a header file plus about 3 separate .c files, to handle the details
* of the Gadget, USB Mass Storage, and SCSI protocols.
*/
/* #define VERBOSE_DEBUG */
/* #define DUMP_MSGS */
#include <linux/blkdev.h>
#include <linux/completion.h>
#include <linux/dcache.h>
#include <linux/delay.h>
#include <linux/device.h>
#include <linux/fcntl.h>
#include <linux/file.h>
#include <linux/fs.h>
#include <linux/kthread.h>
#include <linux/sched/signal.h>
#include <linux/limits.h>
#include <linux/rwsem.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/string.h>
#include <linux/freezer.h>
#include <linux/module.h>
#include <linux/uaccess.h>
#include <linux/usb/ch9.h>
#include <linux/usb/gadget.h>
USB, Mass Storage, composite, gadget: Fix build failure and memset of a struct Trying to compile drivers/usb/gadget/f_mass_storage.o currently fails and spews a ton of warnings : CC drivers/usb/gadget/f_mass_storage.o drivers/usb/gadget/f_mass_storage.c:436:22: error: field ‘function’ has incomplete type drivers/usb/gadget/f_mass_storage.c: In function ‘fsg_from_func’: drivers/usb/gadget/f_mass_storage.c:466:9: warning: type defaults to ‘int’ in declaration of ‘__mptr’ drivers/usb/gadget/f_mass_storage.c:466:9: warning: initialization from incompatible pointer type drivers/usb/gadget/f_mass_storage.c: At top level: drivers/usb/gadget/f_mass_storage.c:2743:15: warning: ‘struct usb_composite_dev’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c:2743:15: warning: its scope is only this definition or declaration, which is probably not what you want drivers/usb/gadget/f_mass_storage.c: In function ‘fsg_common_init’: drivers/usb/gadget/f_mass_storage.c:2745:34: error: dereferencing pointer to incomplete type drivers/usb/gadget/f_mass_storage.c:2775:23: error: dereferencing pointer to incomplete type drivers/usb/gadget/f_mass_storage.c:2779:3: error: implicit declaration of function ‘usb_string_id’ drivers/usb/gadget/f_mass_storage.c: At top level: drivers/usb/gadget/f_mass_storage.c:2984:60: warning: ‘struct usb_configuration’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c:3003:57: warning: ‘struct usb_configuration’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c: In function ‘fsg_bind’: drivers/usb/gadget/f_mass_storage.c:3006:31: error: dereferencing pointer to incomplete type drivers/usb/gadget/f_mass_storage.c:3013:2: error: implicit declaration of function ‘usb_interface_id’ drivers/usb/gadget/f_mass_storage.c:3033:3: error: dereferencing pointer to incomplete type drivers/usb/gadget/f_mass_storage.c:3034:6: error: dereferencing pointer to incomplete type drivers/usb/gadget/f_mass_storage.c:3043:4: error: dereferencing pointer to incomplete type drivers/usb/gadget/f_mass_storage.c:3044:7: error: dereferencing pointer to incomplete type drivers/usb/gadget/f_mass_storage.c:3045:26: error: dereferencing pointer to incomplete type drivers/usb/gadget/f_mass_storage.c: At top level: drivers/usb/gadget/f_mass_storage.c:3067:14: warning: ‘struct usb_configuration’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c:3067:14: warning: ‘struct usb_composite_dev’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c: In function ‘fsg_bind_config’: drivers/usb/gadget/f_mass_storage.c:3093:2: error: implicit declaration of function ‘usb_add_function’ drivers/usb/gadget/f_mass_storage.c: At top level: drivers/usb/gadget/f_mass_storage.c:3103:9: warning: ‘struct usb_configuration’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c:3103:9: warning: ‘struct usb_composite_dev’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c: In function ‘fsg_add’: drivers/usb/gadget/f_mass_storage.c:3105:2: warning: passing argument 1 of ‘fsg_bind_config’ from incompatible pointer type drivers/usb/gadget/f_mass_storage.c:3065:12: note: expected ‘struct usb_composite_dev *’ but argument is of type ‘struct usb_composite_dev *’ drivers/usb/gadget/f_mass_storage.c:3105:2: warning: passing argument 2 of ‘fsg_bind_config’ from incompatible pointer type drivers/usb/gadget/f_mass_storage.c:3065:12: note: expected ‘struct usb_configuration *’ but argument is of type ‘struct usb_configuration *’ drivers/usb/gadget/f_mass_storage.c: At top level: drivers/usb/gadget/f_mass_storage.c:3190:23: warning: ‘struct usb_composite_dev’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c:3195:23: warning: ‘struct usb_composite_dev’ declared inside parameter list drivers/usb/gadget/f_mass_storage.c:3193:1: error: conflicting types for ‘fsg_common_from_params’ drivers/usb/gadget/f_mass_storage.c:3188:1: note: previous declaration of ‘fsg_common_from_params’ was here drivers/usb/gadget/f_mass_storage.c: In function ‘fsg_common_from_params’: drivers/usb/gadget/f_mass_storage.c:3199:2: warning: passing argument 2 of ‘fsg_common_init’ from incompatible pointer type drivers/usb/gadget/f_mass_storage.c:2741:27: note: expected ‘struct usb_composite_dev *’ but argument is of type ‘struct usb_composite_dev *’ make[1]: *** [drivers/usb/gadget/f_mass_storage.o] Error 1 make: *** [drivers/usb/gadget/f_mass_storage.o] Error 2 This is due to the missing include of linux/usb/composite.h - this patch adds the missing include. In addition there's also a problem in fsg_common_init() where we memset 'common', but we use the size of a pointer to 'struct fsg_common' as the size argument to memset(), not the actual size of the struct. This patch fixes the sizeof so we zero the entire struct as intended. Signed-off-by: Jesper Juhl <jj@chaosbits.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-01-29 09:26:51 +08:00
#include <linux/usb/composite.h>
#include <linux/nospec.h>
#include "configfs.h"
/*------------------------------------------------------------------------*/
#define FSG_DRIVER_DESC "Mass Storage Function"
#define FSG_DRIVER_VERSION "2009/09/11"
static const char fsg_string_interface[] = "Mass Storage";
#include "storage_common.h"
#include "f_mass_storage.h"
/* Static strings, in UTF-8 (for simplicity we use only ASCII characters) */
static struct usb_string fsg_strings[] = {
{FSG_STRING_INTERFACE, fsg_string_interface},
{}
};
static struct usb_gadget_strings fsg_stringtab = {
.language = 0x0409, /* en-us */
.strings = fsg_strings,
};
static struct usb_gadget_strings *fsg_strings_array[] = {
&fsg_stringtab,
NULL,
};
/*-------------------------------------------------------------------------*/
struct fsg_dev;
struct fsg_common;
/* Data shared by all the FSG instances. */
struct fsg_common {
struct usb_gadget *gadget;
struct usb_composite_dev *cdev;
struct fsg_dev *fsg, *new_fsg;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
wait_queue_head_t io_wait;
wait_queue_head_t fsg_wait;
/* filesem protects: backing files in use */
struct rw_semaphore filesem;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* lock protects: state and thread_task */
spinlock_t lock;
struct usb_ep *ep0; /* Copy of gadget->ep0 */
struct usb_request *ep0req; /* Copy of cdev->req */
unsigned int ep0_req_tag;
struct fsg_buffhd *next_buffhd_to_fill;
struct fsg_buffhd *next_buffhd_to_drain;
usb: gadget: storage: make FSG_NUM_BUFFERS variable size FSG_NUM_BUFFERS is set to 2 as default. Usually 2 buffers are enough to establish a good buffering pipeline. The number may be increased in order to compensate a for bursty VFS behaviour. Here follows a description of system that may require more than 2 buffers. * CPU ondemand governor active * latency cost for wake up and/or frequency change * DMA for IO Use case description. * Data transfer from MMC via VFS to USB. * DMA shuffles data from MMC and to USB. * The CPU wakes up every now and then to pass data in and out from VFS, which cause the bursty VFS behaviour. Test set up * Running dd on the host reading from the mass storage device * cmdline: dd if=/dev/sdb of=/dev/null bs=4k count=$((256*100)) * Caches are dropped on the host and on the device before each run Measurements on a Snowball board with ondemand_governor active. FSG_NUM_BUFFERS 2 104857600 bytes (105 MB) copied, 5.62173 s, 18.7 MB/s 104857600 bytes (105 MB) copied, 5.61811 s, 18.7 MB/s 104857600 bytes (105 MB) copied, 5.57817 s, 18.8 MB/s FSG_NUM_BUFFERS 4 104857600 bytes (105 MB) copied, 5.26839 s, 19.9 MB/s 104857600 bytes (105 MB) copied, 5.2691 s, 19.9 MB/s 104857600 bytes (105 MB) copied, 5.2711 s, 19.9 MB/s There may not be one optimal number for all boards. This is why the number is added to Kconfig. If selecting USB_GADGET_DEBUG_FILES this value may be set by a module parameter as well. Signed-off-by: Per Forlin <per.forlin@linaro.org> Acked-by: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-20 03:21:27 +08:00
struct fsg_buffhd *buffhds;
unsigned int fsg_num_buffers;
int cmnd_size;
u8 cmnd[MAX_COMMAND_SIZE];
unsigned int lun;
struct fsg_lun *luns[FSG_MAX_LUNS];
struct fsg_lun *curlun;
unsigned int bulk_out_maxpacket;
enum fsg_state state; /* For exception handling */
unsigned int exception_req_tag;
enum data_direction data_dir;
u32 data_size;
u32 data_size_from_cmnd;
u32 tag;
u32 residue;
u32 usb_amount_left;
unsigned int can_stall:1;
unsigned int free_storage_on_release:1;
unsigned int phase_error:1;
unsigned int short_packet_received:1;
unsigned int bad_lun_okay:1;
unsigned int running:1;
unsigned int sysfs:1;
struct completion thread_notifier;
struct task_struct *thread_task;
/* Gadget's private data. */
void *private_data;
char inquiry_string[INQUIRY_STRING_LEN];
};
struct fsg_dev {
struct usb_function function;
struct usb_gadget *gadget; /* Copy of cdev->gadget */
struct fsg_common *common;
u16 interface_number;
unsigned int bulk_in_enabled:1;
unsigned int bulk_out_enabled:1;
unsigned long atomic_bitflags;
#define IGNORE_BULK_OUT 0
struct usb_ep *bulk_in;
struct usb_ep *bulk_out;
};
static inline int __fsg_is_set(struct fsg_common *common,
const char *func, unsigned line)
{
if (common->fsg)
return 1;
ERROR(common, "common->fsg is NULL in %s at %u\n", func, line);
WARN_ON(1);
return 0;
}
#define fsg_is_set(common) likely(__fsg_is_set(common, __func__, __LINE__))
static inline struct fsg_dev *fsg_from_func(struct usb_function *f)
{
return container_of(f, struct fsg_dev, function);
}
typedef void (*fsg_routine_t)(struct fsg_dev *);
static int exception_in_progress(struct fsg_common *common)
{
return common->state > FSG_STATE_NORMAL;
}
/* Make bulk-out requests be divisible by the maxpacket size */
static void set_bulk_out_req_length(struct fsg_common *common,
struct fsg_buffhd *bh, unsigned int length)
{
unsigned int rem;
bh->bulk_out_intended_length = length;
rem = length % common->bulk_out_maxpacket;
if (rem > 0)
length += common->bulk_out_maxpacket - rem;
bh->outreq->length = length;
}
/*-------------------------------------------------------------------------*/
static int fsg_set_halt(struct fsg_dev *fsg, struct usb_ep *ep)
{
const char *name;
if (ep == fsg->bulk_in)
name = "bulk-in";
else if (ep == fsg->bulk_out)
name = "bulk-out";
else
name = ep->name;
DBG(fsg, "%s set halt\n", name);
return usb_ep_set_halt(ep);
}
/*-------------------------------------------------------------------------*/
/* These routines may be called in process context or in_irq */
static void raise_exception(struct fsg_common *common, enum fsg_state new_state)
{
unsigned long flags;
/*
* Do nothing if a higher-priority exception is already in progress.
* If a lower-or-equal priority exception is in progress, preempt it
* and notify the main thread by sending it a signal.
*/
spin_lock_irqsave(&common->lock, flags);
if (common->state <= new_state) {
common->exception_req_tag = common->ep0_req_tag;
common->state = new_state;
if (common->thread_task)
send_sig_info(SIGUSR1, SEND_SIG_PRIV,
common->thread_task);
}
spin_unlock_irqrestore(&common->lock, flags);
}
/*-------------------------------------------------------------------------*/
static int ep0_queue(struct fsg_common *common)
{
int rc;
rc = usb_ep_queue(common->ep0, common->ep0req, GFP_ATOMIC);
common->ep0->driver_data = common;
if (rc != 0 && rc != -ESHUTDOWN) {
/* We can't do much more than wait for a reset */
WARNING(common, "error in submission: %s --> %d\n",
common->ep0->name, rc);
}
return rc;
}
/*-------------------------------------------------------------------------*/
/* Completion handlers. These always run in_irq. */
static void bulk_in_complete(struct usb_ep *ep, struct usb_request *req)
{
struct fsg_common *common = ep->driver_data;
struct fsg_buffhd *bh = req->context;
if (req->status || req->actual != req->length)
DBG(common, "%s --> %d, %u/%u\n", __func__,
req->status, req->actual, req->length);
if (req->status == -ECONNRESET) /* Request was cancelled */
usb_ep_fifo_flush(ep);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Synchronize with the smp_load_acquire() in sleep_thread() */
smp_store_release(&bh->state, BUF_STATE_EMPTY);
wake_up(&common->io_wait);
}
static void bulk_out_complete(struct usb_ep *ep, struct usb_request *req)
{
struct fsg_common *common = ep->driver_data;
struct fsg_buffhd *bh = req->context;
dump_msg(common, "bulk-out", req->buf, req->actual);
if (req->status || req->actual != bh->bulk_out_intended_length)
DBG(common, "%s --> %d, %u/%u\n", __func__,
req->status, req->actual, bh->bulk_out_intended_length);
if (req->status == -ECONNRESET) /* Request was cancelled */
usb_ep_fifo_flush(ep);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Synchronize with the smp_load_acquire() in sleep_thread() */
smp_store_release(&bh->state, BUF_STATE_FULL);
wake_up(&common->io_wait);
}
static int _fsg_common_get_max_lun(struct fsg_common *common)
{
int i = ARRAY_SIZE(common->luns) - 1;
while (i >= 0 && !common->luns[i])
--i;
return i;
}
static int fsg_setup(struct usb_function *f,
const struct usb_ctrlrequest *ctrl)
{
struct fsg_dev *fsg = fsg_from_func(f);
struct usb_request *req = fsg->common->ep0req;
u16 w_index = le16_to_cpu(ctrl->wIndex);
u16 w_value = le16_to_cpu(ctrl->wValue);
u16 w_length = le16_to_cpu(ctrl->wLength);
if (!fsg_is_set(fsg->common))
return -EOPNOTSUPP;
++fsg->common->ep0_req_tag; /* Record arrival of a new request */
req->context = NULL;
req->length = 0;
dump_msg(fsg, "ep0-setup", (u8 *) ctrl, sizeof(*ctrl));
switch (ctrl->bRequest) {
case US_BULK_RESET_REQUEST:
if (ctrl->bRequestType !=
(USB_DIR_OUT | USB_TYPE_CLASS | USB_RECIP_INTERFACE))
break;
if (w_index != fsg->interface_number || w_value != 0 ||
w_length != 0)
return -EDOM;
/*
* Raise an exception to stop the current operation
* and reinitialize our state.
*/
DBG(fsg, "bulk reset request\n");
raise_exception(fsg->common, FSG_STATE_PROTOCOL_RESET);
return USB_GADGET_DELAYED_STATUS;
case US_BULK_GET_MAX_LUN:
if (ctrl->bRequestType !=
(USB_DIR_IN | USB_TYPE_CLASS | USB_RECIP_INTERFACE))
break;
if (w_index != fsg->interface_number || w_value != 0 ||
w_length != 1)
return -EDOM;
VDBG(fsg, "get max LUN\n");
*(u8 *)req->buf = _fsg_common_get_max_lun(fsg->common);
/* Respond with data/status */
req->length = min((u16)1, w_length);
return ep0_queue(fsg->common);
}
VDBG(fsg,
"unknown class-specific control req %02x.%02x v%04x i%04x l%u\n",
ctrl->bRequestType, ctrl->bRequest,
le16_to_cpu(ctrl->wValue), w_index, w_length);
return -EOPNOTSUPP;
}
/*-------------------------------------------------------------------------*/
/* All the following routines run in process context */
/* Use this for bulk or interrupt transfers, not ep0 */
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
static int start_transfer(struct fsg_dev *fsg, struct usb_ep *ep,
struct usb_request *req)
{
int rc;
if (ep == fsg->bulk_in)
dump_msg(fsg, "bulk-in", req->buf, req->length);
rc = usb_ep_queue(ep, req, GFP_KERNEL);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
if (rc) {
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* We can't do much more than wait for a reset */
req->status = rc;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/*
* Note: currently the net2280 driver fails zero-length
* submissions if DMA is enabled.
*/
if (rc != -ESHUTDOWN &&
!(rc == -EOPNOTSUPP && req->length == 0))
WARNING(fsg, "error in submission: %s --> %d\n",
ep->name, rc);
}
return rc;
}
static bool start_in_transfer(struct fsg_common *common, struct fsg_buffhd *bh)
{
if (!fsg_is_set(common))
return false;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
bh->state = BUF_STATE_SENDING;
if (start_transfer(common->fsg, common->fsg->bulk_in, bh->inreq))
bh->state = BUF_STATE_EMPTY;
return true;
}
static bool start_out_transfer(struct fsg_common *common, struct fsg_buffhd *bh)
{
if (!fsg_is_set(common))
return false;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
bh->state = BUF_STATE_RECEIVING;
if (start_transfer(common->fsg, common->fsg->bulk_out, bh->outreq))
bh->state = BUF_STATE_FULL;
return true;
}
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
static int sleep_thread(struct fsg_common *common, bool can_freeze,
struct fsg_buffhd *bh)
{
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
int rc;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Wait until a signal arrives or bh is no longer busy */
if (can_freeze)
/*
* synchronize with the smp_store_release(&bh->state) in
* bulk_in_complete() or bulk_out_complete()
*/
rc = wait_event_freezable(common->io_wait,
bh && smp_load_acquire(&bh->state) >=
BUF_STATE_EMPTY);
else
rc = wait_event_interruptible(common->io_wait,
bh && smp_load_acquire(&bh->state) >=
BUF_STATE_EMPTY);
return rc ? -EINTR : 0;
}
/*-------------------------------------------------------------------------*/
static int do_read(struct fsg_common *common)
{
struct fsg_lun *curlun = common->curlun;
u32 lba;
struct fsg_buffhd *bh;
int rc;
u32 amount_left;
loff_t file_offset, file_offset_tmp;
unsigned int amount;
ssize_t nread;
/*
* Get the starting Logical Block Address and check that it's
* not too big.
*/
if (common->cmnd[0] == READ_6)
lba = get_unaligned_be24(&common->cmnd[1]);
else {
lba = get_unaligned_be32(&common->cmnd[2]);
/*
* We allow DPO (Disable Page Out = don't save data in the
* cache) and FUA (Force Unit Access = don't read from the
* cache), but we don't implement them.
*/
if ((common->cmnd[1] & ~0x18) != 0) {
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
}
if (lba >= curlun->num_sectors) {
curlun->sense_data = SS_LOGICAL_BLOCK_ADDRESS_OUT_OF_RANGE;
return -EINVAL;
}
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
file_offset = ((loff_t) lba) << curlun->blkbits;
/* Carry out the file reads */
amount_left = common->data_size_from_cmnd;
if (unlikely(amount_left == 0))
return -EIO; /* No default reply */
for (;;) {
/*
* Figure out how much we need to read:
* Try to read the remaining amount.
* But don't read more than the buffer size.
* And don't try to read past the end of the file.
*/
amount = min(amount_left, FSG_BUFLEN);
amount = min((loff_t)amount,
curlun->file_length - file_offset);
/* Wait for the next buffer to become available */
bh = common->next_buffhd_to_fill;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
rc = sleep_thread(common, false, bh);
if (rc)
return rc;
/*
* If we were asked to read past the end of file,
* end with an empty buffer.
*/
if (amount == 0) {
curlun->sense_data =
SS_LOGICAL_BLOCK_ADDRESS_OUT_OF_RANGE;
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
curlun->sense_data_info =
file_offset >> curlun->blkbits;
curlun->info_valid = 1;
bh->inreq->length = 0;
bh->state = BUF_STATE_FULL;
break;
}
/* Perform the read */
file_offset_tmp = file_offset;
nread = kernel_read(curlun->filp, bh->buf, amount,
&file_offset_tmp);
VLDBG(curlun, "file read %u @ %llu -> %d\n", amount,
(unsigned long long)file_offset, (int)nread);
if (signal_pending(current))
return -EINTR;
if (nread < 0) {
LDBG(curlun, "error in file read: %d\n", (int)nread);
nread = 0;
} else if (nread < amount) {
LDBG(curlun, "partial file read: %d/%u\n",
(int)nread, amount);
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
nread = round_down(nread, curlun->blksize);
}
file_offset += nread;
amount_left -= nread;
common->residue -= nread;
/*
* Except at the end of the transfer, nread will be
* equal to the buffer size, which is divisible by the
* bulk-in maxpacket size.
*/
bh->inreq->length = nread;
bh->state = BUF_STATE_FULL;
/* If an error occurred, report it and its position */
if (nread < amount) {
curlun->sense_data = SS_UNRECOVERED_READ_ERROR;
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
curlun->sense_data_info =
file_offset >> curlun->blkbits;
curlun->info_valid = 1;
break;
}
if (amount_left == 0)
break; /* No more left to read */
/* Send this buffer and go read some more */
bh->inreq->zero = 0;
if (!start_in_transfer(common, bh))
/* Don't know what to do if common->fsg is NULL */
return -EIO;
common->next_buffhd_to_fill = bh->next;
}
return -EIO; /* No default reply */
}
/*-------------------------------------------------------------------------*/
static int do_write(struct fsg_common *common)
{
struct fsg_lun *curlun = common->curlun;
u32 lba;
struct fsg_buffhd *bh;
int get_some_more;
u32 amount_left_to_req, amount_left_to_write;
loff_t usb_offset, file_offset, file_offset_tmp;
unsigned int amount;
ssize_t nwritten;
int rc;
if (curlun->ro) {
curlun->sense_data = SS_WRITE_PROTECTED;
return -EINVAL;
}
spin_lock(&curlun->filp->f_lock);
curlun->filp->f_flags &= ~O_SYNC; /* Default is not to wait */
spin_unlock(&curlun->filp->f_lock);
/*
* Get the starting Logical Block Address and check that it's
* not too big
*/
if (common->cmnd[0] == WRITE_6)
lba = get_unaligned_be24(&common->cmnd[1]);
else {
lba = get_unaligned_be32(&common->cmnd[2]);
/*
* We allow DPO (Disable Page Out = don't save data in the
* cache) and FUA (Force Unit Access = write directly to the
* medium). We don't implement DPO; we implement FUA by
* performing synchronous output.
*/
if (common->cmnd[1] & ~0x18) {
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
if (!curlun->nofua && (common->cmnd[1] & 0x08)) { /* FUA */
spin_lock(&curlun->filp->f_lock);
curlun->filp->f_flags |= O_SYNC;
spin_unlock(&curlun->filp->f_lock);
}
}
if (lba >= curlun->num_sectors) {
curlun->sense_data = SS_LOGICAL_BLOCK_ADDRESS_OUT_OF_RANGE;
return -EINVAL;
}
/* Carry out the file writes */
get_some_more = 1;
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
file_offset = usb_offset = ((loff_t) lba) << curlun->blkbits;
amount_left_to_req = common->data_size_from_cmnd;
amount_left_to_write = common->data_size_from_cmnd;
while (amount_left_to_write > 0) {
/* Queue a request for more data from the host */
bh = common->next_buffhd_to_fill;
if (bh->state == BUF_STATE_EMPTY && get_some_more) {
/*
* Figure out how much we want to get:
* Try to get the remaining amount,
* but not more than the buffer size.
*/
amount = min(amount_left_to_req, FSG_BUFLEN);
/* Beyond the end of the backing file? */
if (usb_offset >= curlun->file_length) {
get_some_more = 0;
curlun->sense_data =
SS_LOGICAL_BLOCK_ADDRESS_OUT_OF_RANGE;
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
curlun->sense_data_info =
usb_offset >> curlun->blkbits;
curlun->info_valid = 1;
continue;
}
/* Get the next buffer */
usb_offset += amount;
common->usb_amount_left -= amount;
amount_left_to_req -= amount;
if (amount_left_to_req == 0)
get_some_more = 0;
/*
* Except at the end of the transfer, amount will be
* equal to the buffer size, which is divisible by
* the bulk-out maxpacket size.
*/
set_bulk_out_req_length(common, bh, amount);
if (!start_out_transfer(common, bh))
/* Dunno what to do if common->fsg is NULL */
return -EIO;
common->next_buffhd_to_fill = bh->next;
continue;
}
/* Write the received data to the backing file */
bh = common->next_buffhd_to_drain;
if (bh->state == BUF_STATE_EMPTY && !get_some_more)
break; /* We stopped early */
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Wait for the data to be received */
rc = sleep_thread(common, false, bh);
if (rc)
return rc;
common->next_buffhd_to_drain = bh->next;
bh->state = BUF_STATE_EMPTY;
/* Did something go wrong with the transfer? */
if (bh->outreq->status != 0) {
curlun->sense_data = SS_COMMUNICATION_FAILURE;
curlun->sense_data_info =
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
file_offset >> curlun->blkbits;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
curlun->info_valid = 1;
break;
}
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
amount = bh->outreq->actual;
if (curlun->file_length - file_offset < amount) {
LERROR(curlun, "write %u @ %llu beyond end %llu\n",
amount, (unsigned long long)file_offset,
(unsigned long long)curlun->file_length);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
amount = curlun->file_length - file_offset;
}
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/*
* Don't accept excess data. The spec doesn't say
* what to do in this case. We'll ignore the error.
*/
amount = min(amount, bh->bulk_out_intended_length);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Don't write a partial block */
amount = round_down(amount, curlun->blksize);
if (amount == 0)
goto empty_write;
/* Perform the write */
file_offset_tmp = file_offset;
nwritten = kernel_write(curlun->filp, bh->buf, amount,
&file_offset_tmp);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
VLDBG(curlun, "file write %u @ %llu -> %d\n", amount,
(unsigned long long)file_offset, (int)nwritten);
if (signal_pending(current))
return -EINTR; /* Interrupted! */
if (nwritten < 0) {
LDBG(curlun, "error in file write: %d\n",
(int) nwritten);
nwritten = 0;
} else if (nwritten < amount) {
LDBG(curlun, "partial file write: %d/%u\n",
(int) nwritten, amount);
nwritten = round_down(nwritten, curlun->blksize);
}
file_offset += nwritten;
amount_left_to_write -= nwritten;
common->residue -= nwritten;
/* If an error occurred, report it and its position */
if (nwritten < amount) {
curlun->sense_data = SS_WRITE_ERROR;
curlun->sense_data_info =
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
file_offset >> curlun->blkbits;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
curlun->info_valid = 1;
break;
}
empty_write:
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Did the host decide to stop early? */
if (bh->outreq->actual < bh->bulk_out_intended_length) {
common->short_packet_received = 1;
break;
}
}
return -EIO; /* No default reply */
}
/*-------------------------------------------------------------------------*/
static int do_synchronize_cache(struct fsg_common *common)
{
struct fsg_lun *curlun = common->curlun;
int rc;
/* We ignore the requested LBA and write out all file's
* dirty data buffers. */
rc = fsg_lun_fsync_sub(curlun);
if (rc)
curlun->sense_data = SS_WRITE_ERROR;
return 0;
}
/*-------------------------------------------------------------------------*/
static void invalidate_sub(struct fsg_lun *curlun)
{
struct file *filp = curlun->filp;
struct inode *inode = file_inode(filp);
unsigned long rc;
rc = invalidate_mapping_pages(inode->i_mapping, 0, -1);
VLDBG(curlun, "invalidate_mapping_pages -> %ld\n", rc);
}
static int do_verify(struct fsg_common *common)
{
struct fsg_lun *curlun = common->curlun;
u32 lba;
u32 verification_length;
struct fsg_buffhd *bh = common->next_buffhd_to_fill;
loff_t file_offset, file_offset_tmp;
u32 amount_left;
unsigned int amount;
ssize_t nread;
/*
* Get the starting Logical Block Address and check that it's
* not too big.
*/
lba = get_unaligned_be32(&common->cmnd[2]);
if (lba >= curlun->num_sectors) {
curlun->sense_data = SS_LOGICAL_BLOCK_ADDRESS_OUT_OF_RANGE;
return -EINVAL;
}
/*
* We allow DPO (Disable Page Out = don't save data in the
* cache) but we don't implement it.
*/
if (common->cmnd[1] & ~0x10) {
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
verification_length = get_unaligned_be16(&common->cmnd[7]);
if (unlikely(verification_length == 0))
return -EIO; /* No default reply */
/* Prepare to carry out the file verify */
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
amount_left = verification_length << curlun->blkbits;
file_offset = ((loff_t) lba) << curlun->blkbits;
/* Write out all the dirty buffers before invalidating them */
fsg_lun_fsync_sub(curlun);
if (signal_pending(current))
return -EINTR;
invalidate_sub(curlun);
if (signal_pending(current))
return -EINTR;
/* Just try to read the requested blocks */
while (amount_left > 0) {
/*
* Figure out how much we need to read:
* Try to read the remaining amount, but not more than
* the buffer size.
* And don't try to read past the end of the file.
*/
amount = min(amount_left, FSG_BUFLEN);
amount = min((loff_t)amount,
curlun->file_length - file_offset);
if (amount == 0) {
curlun->sense_data =
SS_LOGICAL_BLOCK_ADDRESS_OUT_OF_RANGE;
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
curlun->sense_data_info =
file_offset >> curlun->blkbits;
curlun->info_valid = 1;
break;
}
/* Perform the read */
file_offset_tmp = file_offset;
nread = kernel_read(curlun->filp, bh->buf, amount,
&file_offset_tmp);
VLDBG(curlun, "file read %u @ %llu -> %d\n", amount,
(unsigned long long) file_offset,
(int) nread);
if (signal_pending(current))
return -EINTR;
if (nread < 0) {
LDBG(curlun, "error in file verify: %d\n", (int)nread);
nread = 0;
} else if (nread < amount) {
LDBG(curlun, "partial file verify: %d/%u\n",
(int)nread, amount);
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
nread = round_down(nread, curlun->blksize);
}
if (nread == 0) {
curlun->sense_data = SS_UNRECOVERED_READ_ERROR;
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
curlun->sense_data_info =
file_offset >> curlun->blkbits;
curlun->info_valid = 1;
break;
}
file_offset += nread;
amount_left -= nread;
}
return 0;
}
/*-------------------------------------------------------------------------*/
static int do_inquiry(struct fsg_common *common, struct fsg_buffhd *bh)
{
struct fsg_lun *curlun = common->curlun;
u8 *buf = (u8 *) bh->buf;
if (!curlun) { /* Unsupported LUNs are okay */
common->bad_lun_okay = 1;
memset(buf, 0, 36);
buf[0] = TYPE_NO_LUN; /* Unsupported, no device-type */
buf[4] = 31; /* Additional length */
return 36;
}
buf[0] = curlun->cdrom ? TYPE_ROM : TYPE_DISK;
buf[1] = curlun->removable ? 0x80 : 0;
buf[2] = 2; /* ANSI SCSI level 2 */
buf[3] = 2; /* SCSI-2 INQUIRY data format */
buf[4] = 31; /* Additional length */
buf[5] = 0; /* No special options */
buf[6] = 0;
buf[7] = 0;
if (curlun->inquiry_string[0])
memcpy(buf + 8, curlun->inquiry_string,
sizeof(curlun->inquiry_string));
else
memcpy(buf + 8, common->inquiry_string,
sizeof(common->inquiry_string));
return 36;
}
static int do_request_sense(struct fsg_common *common, struct fsg_buffhd *bh)
{
struct fsg_lun *curlun = common->curlun;
u8 *buf = (u8 *) bh->buf;
u32 sd, sdinfo;
int valid;
/*
* From the SCSI-2 spec., section 7.9 (Unit attention condition):
*
* If a REQUEST SENSE command is received from an initiator
* with a pending unit attention condition (before the target
* generates the contingent allegiance condition), then the
* target shall either:
* a) report any pending sense data and preserve the unit
* attention condition on the logical unit, or,
* b) report the unit attention condition, may discard any
* pending sense data, and clear the unit attention
* condition on the logical unit for that initiator.
*
* FSG normally uses option a); enable this code to use option b).
*/
#if 0
if (curlun && curlun->unit_attention_data != SS_NO_SENSE) {
curlun->sense_data = curlun->unit_attention_data;
curlun->unit_attention_data = SS_NO_SENSE;
}
#endif
if (!curlun) { /* Unsupported LUNs are okay */
common->bad_lun_okay = 1;
sd = SS_LOGICAL_UNIT_NOT_SUPPORTED;
sdinfo = 0;
valid = 0;
} else {
sd = curlun->sense_data;
sdinfo = curlun->sense_data_info;
valid = curlun->info_valid << 7;
curlun->sense_data = SS_NO_SENSE;
curlun->sense_data_info = 0;
curlun->info_valid = 0;
}
memset(buf, 0, 18);
buf[0] = valid | 0x70; /* Valid, current error */
buf[2] = SK(sd);
put_unaligned_be32(sdinfo, &buf[3]); /* Sense information */
buf[7] = 18 - 8; /* Additional sense length */
buf[12] = ASC(sd);
buf[13] = ASCQ(sd);
return 18;
}
static int do_read_capacity(struct fsg_common *common, struct fsg_buffhd *bh)
{
struct fsg_lun *curlun = common->curlun;
u32 lba = get_unaligned_be32(&common->cmnd[2]);
int pmi = common->cmnd[8];
u8 *buf = (u8 *)bh->buf;
/* Check the PMI and LBA fields */
if (pmi > 1 || (pmi == 0 && lba != 0)) {
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
put_unaligned_be32(curlun->num_sectors - 1, &buf[0]);
/* Max logical block */
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
put_unaligned_be32(curlun->blksize, &buf[4]);/* Block length */
return 8;
}
static int do_read_header(struct fsg_common *common, struct fsg_buffhd *bh)
{
struct fsg_lun *curlun = common->curlun;
int msf = common->cmnd[1] & 0x02;
u32 lba = get_unaligned_be32(&common->cmnd[2]);
u8 *buf = (u8 *)bh->buf;
if (common->cmnd[1] & ~0x02) { /* Mask away MSF */
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
if (lba >= curlun->num_sectors) {
curlun->sense_data = SS_LOGICAL_BLOCK_ADDRESS_OUT_OF_RANGE;
return -EINVAL;
}
memset(buf, 0, 8);
buf[0] = 0x01; /* 2048 bytes of user data, rest is EC */
store_cdrom_address(&buf[4], msf, lba);
return 8;
}
static int do_read_toc(struct fsg_common *common, struct fsg_buffhd *bh)
{
struct fsg_lun *curlun = common->curlun;
int msf = common->cmnd[1] & 0x02;
int start_track = common->cmnd[6];
u8 *buf = (u8 *)bh->buf;
if ((common->cmnd[1] & ~0x02) != 0 || /* Mask away MSF */
start_track > 1) {
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
memset(buf, 0, 20);
buf[1] = (20-2); /* TOC data length */
buf[2] = 1; /* First track number */
buf[3] = 1; /* Last track number */
buf[5] = 0x16; /* Data track, copying allowed */
buf[6] = 0x01; /* Only track is number 1 */
store_cdrom_address(&buf[8], msf, 0);
buf[13] = 0x16; /* Lead-out track is data */
buf[14] = 0xAA; /* Lead-out track number */
store_cdrom_address(&buf[16], msf, curlun->num_sectors);
return 20;
}
static int do_mode_sense(struct fsg_common *common, struct fsg_buffhd *bh)
{
struct fsg_lun *curlun = common->curlun;
int mscmnd = common->cmnd[0];
u8 *buf = (u8 *) bh->buf;
u8 *buf0 = buf;
int pc, page_code;
int changeable_values, all_pages;
int valid_page = 0;
int len, limit;
if ((common->cmnd[1] & ~0x08) != 0) { /* Mask away DBD */
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
pc = common->cmnd[2] >> 6;
page_code = common->cmnd[2] & 0x3f;
if (pc == 3) {
curlun->sense_data = SS_SAVING_PARAMETERS_NOT_SUPPORTED;
return -EINVAL;
}
changeable_values = (pc == 1);
all_pages = (page_code == 0x3f);
/*
* Write the mode parameter header. Fixed values are: default
* medium type, no cache control (DPOFUA), and no block descriptors.
* The only variable value is the WriteProtect bit. We will fill in
* the mode data length later.
*/
memset(buf, 0, 8);
if (mscmnd == MODE_SENSE) {
buf[2] = (curlun->ro ? 0x80 : 0x00); /* WP, DPOFUA */
buf += 4;
limit = 255;
} else { /* MODE_SENSE_10 */
buf[3] = (curlun->ro ? 0x80 : 0x00); /* WP, DPOFUA */
buf += 8;
limit = 65535; /* Should really be FSG_BUFLEN */
}
/* No block descriptors */
/*
* The mode pages, in numerical order. The only page we support
* is the Caching page.
*/
if (page_code == 0x08 || all_pages) {
valid_page = 1;
buf[0] = 0x08; /* Page code */
buf[1] = 10; /* Page length */
memset(buf+2, 0, 10); /* None of the fields are changeable */
if (!changeable_values) {
buf[2] = 0x04; /* Write cache enable, */
/* Read cache not disabled */
/* No cache retention priorities */
put_unaligned_be16(0xffff, &buf[4]);
/* Don't disable prefetch */
/* Minimum prefetch = 0 */
put_unaligned_be16(0xffff, &buf[8]);
/* Maximum prefetch */
put_unaligned_be16(0xffff, &buf[10]);
/* Maximum prefetch ceiling */
}
buf += 12;
}
/*
* Check that a valid page was requested and the mode data length
* isn't too long.
*/
len = buf - buf0;
if (!valid_page || len > limit) {
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
/* Store the mode data length */
if (mscmnd == MODE_SENSE)
buf0[0] = len - 1;
else
put_unaligned_be16(len - 2, buf0);
return len;
}
static int do_start_stop(struct fsg_common *common)
{
struct fsg_lun *curlun = common->curlun;
int loej, start;
if (!curlun) {
return -EINVAL;
} else if (!curlun->removable) {
curlun->sense_data = SS_INVALID_COMMAND;
return -EINVAL;
} else if ((common->cmnd[1] & ~0x01) != 0 || /* Mask away Immed */
(common->cmnd[4] & ~0x03) != 0) { /* Mask LoEj, Start */
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
loej = common->cmnd[4] & 0x02;
start = common->cmnd[4] & 0x01;
/*
* Our emulation doesn't support mounting; the medium is
* available for use as soon as it is loaded.
*/
if (start) {
if (!fsg_lun_is_open(curlun)) {
curlun->sense_data = SS_MEDIUM_NOT_PRESENT;
return -EINVAL;
}
return 0;
}
/* Are we allowed to unload the media? */
if (curlun->prevent_medium_removal) {
LDBG(curlun, "unload attempt prevented\n");
curlun->sense_data = SS_MEDIUM_REMOVAL_PREVENTED;
return -EINVAL;
}
if (!loej)
return 0;
up_read(&common->filesem);
down_write(&common->filesem);
fsg_lun_close(curlun);
up_write(&common->filesem);
down_read(&common->filesem);
return 0;
}
static int do_prevent_allow(struct fsg_common *common)
{
struct fsg_lun *curlun = common->curlun;
int prevent;
if (!common->curlun) {
return -EINVAL;
} else if (!common->curlun->removable) {
common->curlun->sense_data = SS_INVALID_COMMAND;
return -EINVAL;
}
prevent = common->cmnd[4] & 0x01;
if ((common->cmnd[4] & ~0x01) != 0) { /* Mask away Prevent */
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
if (curlun->prevent_medium_removal && !prevent)
fsg_lun_fsync_sub(curlun);
curlun->prevent_medium_removal = prevent;
return 0;
}
static int do_read_format_capacities(struct fsg_common *common,
struct fsg_buffhd *bh)
{
struct fsg_lun *curlun = common->curlun;
u8 *buf = (u8 *) bh->buf;
buf[0] = buf[1] = buf[2] = 0;
buf[3] = 8; /* Only the Current/Maximum Capacity Descriptor */
buf += 4;
put_unaligned_be32(curlun->num_sectors, &buf[0]);
/* Number of blocks */
usb: gadget: storage: adapt logic block size to bound block devices Now the mass storage driver has fixed logic block size of 512 bytes. The mass storage gadget read/write bound devices only through VFS, so the bottom level devices actually are just RAW devices to the driver and connected PC. As a RAW, hosts can always format, read and write it right in 512 bytes logic block and don't care about the actual logic block size of devices bound to the gadget. But if we want to share the bound block device partition between target board and PC, in case the logic block size of the bound block device is 4KB, we execute the following steps: 1. connect a board with mass storage gadget to PC(the board has set one partition of on-board block device as file name of the mass storage) 2. PC format the mass storage to VFAT by default logic block size and read/write it 3. disconnect boards from PC 4. target board mount the partition as VFAT Step 4 will fail since kernel on target thinks the logic block size of the bound partition as 4KB. A typical error is "FAT: logical sector size too small for device (logical sector size = 512)" If we execute opposite steps: 1. format the partition to VFAT on target board and read/write this partition 2. connect the board to Windows PC as usb mass storage gadget, windows will think the disk is not formatted So the conclusion is that only as a gadget, the mass storage driver has no any problem. But being shared VFAT or other filesystem on PC and target board, it will fail. This patch adapts logic block size to bound block devices and fix the issue. Cc: Michal Nazarewicz <mina86@mina86.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Peiyu Li <peiyu.li@csr.com> Signed-off-by: Xianglong Du <xianglong.du@csr.com> Signed-off-by: Huayi Li <huayi.li@csr.com> Signed-off-by: Barry Song <Baohua.Song@csr.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2011-08-18 13:52:59 +08:00
put_unaligned_be32(curlun->blksize, &buf[4]);/* Block length */
buf[4] = 0x02; /* Current capacity */
return 12;
}
static int do_mode_select(struct fsg_common *common, struct fsg_buffhd *bh)
{
struct fsg_lun *curlun = common->curlun;
/* We don't support MODE SELECT */
if (curlun)
curlun->sense_data = SS_INVALID_COMMAND;
return -EINVAL;
}
/*-------------------------------------------------------------------------*/
static int halt_bulk_in_endpoint(struct fsg_dev *fsg)
{
int rc;
rc = fsg_set_halt(fsg, fsg->bulk_in);
if (rc == -EAGAIN)
VDBG(fsg, "delayed bulk-in endpoint halt\n");
while (rc != 0) {
if (rc != -EAGAIN) {
WARNING(fsg, "usb_ep_set_halt -> %d\n", rc);
rc = 0;
break;
}
/* Wait for a short time and then try again */
if (msleep_interruptible(100) != 0)
return -EINTR;
rc = usb_ep_set_halt(fsg->bulk_in);
}
return rc;
}
static int wedge_bulk_in_endpoint(struct fsg_dev *fsg)
{
int rc;
DBG(fsg, "bulk-in set wedge\n");
rc = usb_ep_set_wedge(fsg->bulk_in);
if (rc == -EAGAIN)
VDBG(fsg, "delayed bulk-in endpoint wedge\n");
while (rc != 0) {
if (rc != -EAGAIN) {
WARNING(fsg, "usb_ep_set_wedge -> %d\n", rc);
rc = 0;
break;
}
/* Wait for a short time and then try again */
if (msleep_interruptible(100) != 0)
return -EINTR;
rc = usb_ep_set_wedge(fsg->bulk_in);
}
return rc;
}
static int throw_away_data(struct fsg_common *common)
{
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
struct fsg_buffhd *bh, *bh2;
u32 amount;
int rc;
for (bh = common->next_buffhd_to_drain;
bh->state != BUF_STATE_EMPTY || common->usb_amount_left > 0;
bh = common->next_buffhd_to_drain) {
/* Try to submit another request if we need one */
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
bh2 = common->next_buffhd_to_fill;
if (bh2->state == BUF_STATE_EMPTY &&
common->usb_amount_left > 0) {
amount = min(common->usb_amount_left, FSG_BUFLEN);
/*
* Except at the end of the transfer, amount will be
* equal to the buffer size, which is divisible by
* the bulk-out maxpacket size.
*/
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
set_bulk_out_req_length(common, bh2, amount);
if (!start_out_transfer(common, bh2))
/* Dunno what to do if common->fsg is NULL */
return -EIO;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
common->next_buffhd_to_fill = bh2->next;
common->usb_amount_left -= amount;
continue;
}
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Wait for the data to be received */
rc = sleep_thread(common, false, bh);
if (rc)
return rc;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Throw away the data in a filled buffer */
bh->state = BUF_STATE_EMPTY;
common->next_buffhd_to_drain = bh->next;
/* A short packet or an error ends everything */
if (bh->outreq->actual < bh->bulk_out_intended_length ||
bh->outreq->status != 0) {
raise_exception(common, FSG_STATE_ABORT_BULK_OUT);
return -EINTR;
}
}
return 0;
}
static int finish_reply(struct fsg_common *common)
{
struct fsg_buffhd *bh = common->next_buffhd_to_fill;
int rc = 0;
switch (common->data_dir) {
case DATA_DIR_NONE:
break; /* Nothing to send */
/*
* If we don't know whether the host wants to read or write,
* this must be CB or CBI with an unknown command. We mustn't
* try to send or receive any data. So stall both bulk pipes
* if we can and wait for a reset.
*/
case DATA_DIR_UNKNOWN:
if (!common->can_stall) {
/* Nothing */
} else if (fsg_is_set(common)) {
fsg_set_halt(common->fsg, common->fsg->bulk_out);
rc = halt_bulk_in_endpoint(common->fsg);
} else {
/* Don't know what to do if common->fsg is NULL */
rc = -EIO;
}
break;
/* All but the last buffer of data must have already been sent */
case DATA_DIR_TO_HOST:
if (common->data_size == 0) {
/* Nothing to send */
/* Don't know what to do if common->fsg is NULL */
} else if (!fsg_is_set(common)) {
rc = -EIO;
/* If there's no residue, simply send the last buffer */
} else if (common->residue == 0) {
bh->inreq->zero = 0;
if (!start_in_transfer(common, bh))
return -EIO;
common->next_buffhd_to_fill = bh->next;
/*
* For Bulk-only, mark the end of the data with a short
* packet. If we are allowed to stall, halt the bulk-in
* endpoint. (Note: This violates the Bulk-Only Transport
* specification, which requires us to pad the data if we
* don't halt the endpoint. Presumably nobody will mind.)
*/
} else {
bh->inreq->zero = 1;
if (!start_in_transfer(common, bh))
rc = -EIO;
common->next_buffhd_to_fill = bh->next;
if (common->can_stall)
rc = halt_bulk_in_endpoint(common->fsg);
}
break;
/*
* We have processed all we want from the data the host has sent.
* There may still be outstanding bulk-out requests.
*/
case DATA_DIR_FROM_HOST:
if (common->residue == 0) {
/* Nothing to receive */
/* Did the host stop sending unexpectedly early? */
} else if (common->short_packet_received) {
raise_exception(common, FSG_STATE_ABORT_BULK_OUT);
rc = -EINTR;
/*
* We haven't processed all the incoming data. Even though
* we may be allowed to stall, doing so would cause a race.
* The controller may already have ACK'ed all the remaining
* bulk-out packets, in which case the host wouldn't see a
* STALL. Not realizing the endpoint was halted, it wouldn't
* clear the halt -- leading to problems later on.
*/
#if 0
} else if (common->can_stall) {
if (fsg_is_set(common))
fsg_set_halt(common->fsg,
common->fsg->bulk_out);
raise_exception(common, FSG_STATE_ABORT_BULK_OUT);
rc = -EINTR;
#endif
/*
* We can't stall. Read in the excess data and throw it
* all away.
*/
} else {
rc = throw_away_data(common);
}
break;
}
return rc;
}
static void send_status(struct fsg_common *common)
{
struct fsg_lun *curlun = common->curlun;
struct fsg_buffhd *bh;
struct bulk_cs_wrap *csw;
int rc;
u8 status = US_BULK_STAT_OK;
u32 sd, sdinfo = 0;
/* Wait for the next buffer to become available */
bh = common->next_buffhd_to_fill;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
rc = sleep_thread(common, false, bh);
if (rc)
return;
if (curlun) {
sd = curlun->sense_data;
sdinfo = curlun->sense_data_info;
} else if (common->bad_lun_okay)
sd = SS_NO_SENSE;
else
sd = SS_LOGICAL_UNIT_NOT_SUPPORTED;
if (common->phase_error) {
DBG(common, "sending phase-error status\n");
status = US_BULK_STAT_PHASE;
sd = SS_INVALID_COMMAND;
} else if (sd != SS_NO_SENSE) {
DBG(common, "sending command-failure status\n");
status = US_BULK_STAT_FAIL;
VDBG(common, " sense data: SK x%02x, ASC x%02x, ASCQ x%02x;"
" info x%x\n",
SK(sd), ASC(sd), ASCQ(sd), sdinfo);
}
/* Store and send the Bulk-only CSW */
csw = (void *)bh->buf;
csw->Signature = cpu_to_le32(US_BULK_CS_SIGN);
csw->Tag = common->tag;
csw->Residue = cpu_to_le32(common->residue);
csw->Status = status;
bh->inreq->length = US_BULK_CS_WRAP_LEN;
bh->inreq->zero = 0;
if (!start_in_transfer(common, bh))
/* Don't know what to do if common->fsg is NULL */
return;
common->next_buffhd_to_fill = bh->next;
return;
}
/*-------------------------------------------------------------------------*/
/*
* Check whether the command is properly formed and whether its data size
* and direction agree with the values we already have.
*/
static int check_command(struct fsg_common *common, int cmnd_size,
enum data_direction data_dir, unsigned int mask,
int needs_medium, const char *name)
{
int i;
unsigned int lun = common->cmnd[1] >> 5;
static const char dirletter[4] = {'u', 'o', 'i', 'n'};
char hdlen[20];
struct fsg_lun *curlun;
hdlen[0] = 0;
if (common->data_dir != DATA_DIR_UNKNOWN)
sprintf(hdlen, ", H%c=%u", dirletter[(int) common->data_dir],
common->data_size);
VDBG(common, "SCSI command: %s; Dc=%d, D%c=%u; Hc=%d%s\n",
name, cmnd_size, dirletter[(int) data_dir],
common->data_size_from_cmnd, common->cmnd_size, hdlen);
/*
* We can't reply at all until we know the correct data direction
* and size.
*/
if (common->data_size_from_cmnd == 0)
data_dir = DATA_DIR_NONE;
if (common->data_size < common->data_size_from_cmnd) {
/*
* Host data size < Device data size is a phase error.
* Carry out the command, but only transfer as much as
* we are allowed.
*/
common->data_size_from_cmnd = common->data_size;
common->phase_error = 1;
}
common->residue = common->data_size;
common->usb_amount_left = common->data_size;
/* Conflicting data directions is a phase error */
if (common->data_dir != data_dir && common->data_size_from_cmnd > 0) {
common->phase_error = 1;
return -EINVAL;
}
/* Verify the length of the command itself */
if (cmnd_size != common->cmnd_size) {
/*
* Special case workaround: There are plenty of buggy SCSI
* implementations. Many have issues with cbw->Length
* field passing a wrong command size. For those cases we
* always try to work around the problem by using the length
* sent by the host side provided it is at least as large
* as the correct command length.
* Examples of such cases would be MS-Windows, which issues
* REQUEST SENSE with cbw->Length == 12 where it should
* be 6, and xbox360 issuing INQUIRY, TEST UNIT READY and
* REQUEST SENSE with cbw->Length == 10 where it should
* be 6 as well.
*/
if (cmnd_size <= common->cmnd_size) {
DBG(common, "%s is buggy! Expected length %d "
"but we got %d\n", name,
cmnd_size, common->cmnd_size);
cmnd_size = common->cmnd_size;
} else {
common->phase_error = 1;
return -EINVAL;
}
}
/* Check that the LUN values are consistent */
if (common->lun != lun)
DBG(common, "using LUN %u from CBW, not LUN %u from CDB\n",
common->lun, lun);
/* Check the LUN */
curlun = common->curlun;
if (curlun) {
if (common->cmnd[0] != REQUEST_SENSE) {
curlun->sense_data = SS_NO_SENSE;
curlun->sense_data_info = 0;
curlun->info_valid = 0;
}
} else {
common->bad_lun_okay = 0;
/*
* INQUIRY and REQUEST SENSE commands are explicitly allowed
* to use unsupported LUNs; all others may not.
*/
if (common->cmnd[0] != INQUIRY &&
common->cmnd[0] != REQUEST_SENSE) {
DBG(common, "unsupported LUN %u\n", common->lun);
return -EINVAL;
}
}
/*
* If a unit attention condition exists, only INQUIRY and
* REQUEST SENSE commands are allowed; anything else must fail.
*/
if (curlun && curlun->unit_attention_data != SS_NO_SENSE &&
common->cmnd[0] != INQUIRY &&
common->cmnd[0] != REQUEST_SENSE) {
curlun->sense_data = curlun->unit_attention_data;
curlun->unit_attention_data = SS_NO_SENSE;
return -EINVAL;
}
/* Check that only command bytes listed in the mask are non-zero */
common->cmnd[1] &= 0x1f; /* Mask away the LUN */
for (i = 1; i < cmnd_size; ++i) {
if (common->cmnd[i] && !(mask & (1 << i))) {
if (curlun)
curlun->sense_data = SS_INVALID_FIELD_IN_CDB;
return -EINVAL;
}
}
/* If the medium isn't mounted and the command needs to access
* it, return an error. */
if (curlun && !fsg_lun_is_open(curlun) && needs_medium) {
curlun->sense_data = SS_MEDIUM_NOT_PRESENT;
return -EINVAL;
}
return 0;
}
/* wrapper of check_command for data size in blocks handling */
static int check_command_size_in_blocks(struct fsg_common *common,
int cmnd_size, enum data_direction data_dir,
unsigned int mask, int needs_medium, const char *name)
{
if (common->curlun)
common->data_size_from_cmnd <<= common->curlun->blkbits;
return check_command(common, cmnd_size, data_dir,
mask, needs_medium, name);
}
static int do_scsi_command(struct fsg_common *common)
{
struct fsg_buffhd *bh;
int rc;
int reply = -EINVAL;
int i;
static char unknown[16];
dump_cdb(common);
/* Wait for the next buffer to become available for data or status */
bh = common->next_buffhd_to_fill;
common->next_buffhd_to_drain = bh;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
rc = sleep_thread(common, false, bh);
if (rc)
return rc;
common->phase_error = 0;
common->short_packet_received = 0;
down_read(&common->filesem); /* We're using the backing file */
switch (common->cmnd[0]) {
case INQUIRY:
common->data_size_from_cmnd = common->cmnd[4];
reply = check_command(common, 6, DATA_DIR_TO_HOST,
(1<<4), 0,
"INQUIRY");
if (reply == 0)
reply = do_inquiry(common, bh);
break;
case MODE_SELECT:
common->data_size_from_cmnd = common->cmnd[4];
reply = check_command(common, 6, DATA_DIR_FROM_HOST,
(1<<1) | (1<<4), 0,
"MODE SELECT(6)");
if (reply == 0)
reply = do_mode_select(common, bh);
break;
case MODE_SELECT_10:
common->data_size_from_cmnd =
get_unaligned_be16(&common->cmnd[7]);
reply = check_command(common, 10, DATA_DIR_FROM_HOST,
(1<<1) | (3<<7), 0,
"MODE SELECT(10)");
if (reply == 0)
reply = do_mode_select(common, bh);
break;
case MODE_SENSE:
common->data_size_from_cmnd = common->cmnd[4];
reply = check_command(common, 6, DATA_DIR_TO_HOST,
(1<<1) | (1<<2) | (1<<4), 0,
"MODE SENSE(6)");
if (reply == 0)
reply = do_mode_sense(common, bh);
break;
case MODE_SENSE_10:
common->data_size_from_cmnd =
get_unaligned_be16(&common->cmnd[7]);
reply = check_command(common, 10, DATA_DIR_TO_HOST,
(1<<1) | (1<<2) | (3<<7), 0,
"MODE SENSE(10)");
if (reply == 0)
reply = do_mode_sense(common, bh);
break;
case ALLOW_MEDIUM_REMOVAL:
common->data_size_from_cmnd = 0;
reply = check_command(common, 6, DATA_DIR_NONE,
(1<<4), 0,
"PREVENT-ALLOW MEDIUM REMOVAL");
if (reply == 0)
reply = do_prevent_allow(common);
break;
case READ_6:
i = common->cmnd[4];
common->data_size_from_cmnd = (i == 0) ? 256 : i;
reply = check_command_size_in_blocks(common, 6,
DATA_DIR_TO_HOST,
(7<<1) | (1<<4), 1,
"READ(6)");
if (reply == 0)
reply = do_read(common);
break;
case READ_10:
common->data_size_from_cmnd =
get_unaligned_be16(&common->cmnd[7]);
reply = check_command_size_in_blocks(common, 10,
DATA_DIR_TO_HOST,
(1<<1) | (0xf<<2) | (3<<7), 1,
"READ(10)");
if (reply == 0)
reply = do_read(common);
break;
case READ_12:
common->data_size_from_cmnd =
get_unaligned_be32(&common->cmnd[6]);
reply = check_command_size_in_blocks(common, 12,
DATA_DIR_TO_HOST,
(1<<1) | (0xf<<2) | (0xf<<6), 1,
"READ(12)");
if (reply == 0)
reply = do_read(common);
break;
case READ_CAPACITY:
common->data_size_from_cmnd = 8;
reply = check_command(common, 10, DATA_DIR_TO_HOST,
(0xf<<2) | (1<<8), 1,
"READ CAPACITY");
if (reply == 0)
reply = do_read_capacity(common, bh);
break;
case READ_HEADER:
if (!common->curlun || !common->curlun->cdrom)
goto unknown_cmnd;
common->data_size_from_cmnd =
get_unaligned_be16(&common->cmnd[7]);
reply = check_command(common, 10, DATA_DIR_TO_HOST,
(3<<7) | (0x1f<<1), 1,
"READ HEADER");
if (reply == 0)
reply = do_read_header(common, bh);
break;
case READ_TOC:
if (!common->curlun || !common->curlun->cdrom)
goto unknown_cmnd;
common->data_size_from_cmnd =
get_unaligned_be16(&common->cmnd[7]);
reply = check_command(common, 10, DATA_DIR_TO_HOST,
(7<<6) | (1<<1), 1,
"READ TOC");
if (reply == 0)
reply = do_read_toc(common, bh);
break;
case READ_FORMAT_CAPACITIES:
common->data_size_from_cmnd =
get_unaligned_be16(&common->cmnd[7]);
reply = check_command(common, 10, DATA_DIR_TO_HOST,
(3<<7), 1,
"READ FORMAT CAPACITIES");
if (reply == 0)
reply = do_read_format_capacities(common, bh);
break;
case REQUEST_SENSE:
common->data_size_from_cmnd = common->cmnd[4];
reply = check_command(common, 6, DATA_DIR_TO_HOST,
(1<<4), 0,
"REQUEST SENSE");
if (reply == 0)
reply = do_request_sense(common, bh);
break;
case START_STOP:
common->data_size_from_cmnd = 0;
reply = check_command(common, 6, DATA_DIR_NONE,
(1<<1) | (1<<4), 0,
"START-STOP UNIT");
if (reply == 0)
reply = do_start_stop(common);
break;
case SYNCHRONIZE_CACHE:
common->data_size_from_cmnd = 0;
reply = check_command(common, 10, DATA_DIR_NONE,
(0xf<<2) | (3<<7), 1,
"SYNCHRONIZE CACHE");
if (reply == 0)
reply = do_synchronize_cache(common);
break;
case TEST_UNIT_READY:
common->data_size_from_cmnd = 0;
reply = check_command(common, 6, DATA_DIR_NONE,
0, 1,
"TEST UNIT READY");
break;
/*
* Although optional, this command is used by MS-Windows. We
* support a minimal version: BytChk must be 0.
*/
case VERIFY:
common->data_size_from_cmnd = 0;
reply = check_command(common, 10, DATA_DIR_NONE,
(1<<1) | (0xf<<2) | (3<<7), 1,
"VERIFY");
if (reply == 0)
reply = do_verify(common);
break;
case WRITE_6:
i = common->cmnd[4];
common->data_size_from_cmnd = (i == 0) ? 256 : i;
reply = check_command_size_in_blocks(common, 6,
DATA_DIR_FROM_HOST,
(7<<1) | (1<<4), 1,
"WRITE(6)");
if (reply == 0)
reply = do_write(common);
break;
case WRITE_10:
common->data_size_from_cmnd =
get_unaligned_be16(&common->cmnd[7]);
reply = check_command_size_in_blocks(common, 10,
DATA_DIR_FROM_HOST,
(1<<1) | (0xf<<2) | (3<<7), 1,
"WRITE(10)");
if (reply == 0)
reply = do_write(common);
break;
case WRITE_12:
common->data_size_from_cmnd =
get_unaligned_be32(&common->cmnd[6]);
reply = check_command_size_in_blocks(common, 12,
DATA_DIR_FROM_HOST,
(1<<1) | (0xf<<2) | (0xf<<6), 1,
"WRITE(12)");
if (reply == 0)
reply = do_write(common);
break;
/*
* Some mandatory commands that we recognize but don't implement.
* They don't mean much in this setting. It's left as an exercise
* for anyone interested to implement RESERVE and RELEASE in terms
* of Posix locks.
*/
case FORMAT_UNIT:
case RELEASE:
case RESERVE:
case SEND_DIAGNOSTIC:
/* Fall through */
default:
unknown_cmnd:
common->data_size_from_cmnd = 0;
sprintf(unknown, "Unknown x%02x", common->cmnd[0]);
reply = check_command(common, common->cmnd_size,
DATA_DIR_UNKNOWN, ~0, 0, unknown);
if (reply == 0) {
common->curlun->sense_data = SS_INVALID_COMMAND;
reply = -EINVAL;
}
break;
}
up_read(&common->filesem);
if (reply == -EINTR || signal_pending(current))
return -EINTR;
/* Set up the single reply buffer for finish_reply() */
if (reply == -EINVAL)
reply = 0; /* Error reply length */
if (reply >= 0 && common->data_dir == DATA_DIR_TO_HOST) {
reply = min((u32)reply, common->data_size_from_cmnd);
bh->inreq->length = reply;
bh->state = BUF_STATE_FULL;
common->residue -= reply;
} /* Otherwise it's already set */
return 0;
}
/*-------------------------------------------------------------------------*/
static int received_cbw(struct fsg_dev *fsg, struct fsg_buffhd *bh)
{
struct usb_request *req = bh->outreq;
struct bulk_cb_wrap *cbw = req->buf;
struct fsg_common *common = fsg->common;
/* Was this a real packet? Should it be ignored? */
if (req->status || test_bit(IGNORE_BULK_OUT, &fsg->atomic_bitflags))
return -EINVAL;
/* Is the CBW valid? */
if (req->actual != US_BULK_CB_WRAP_LEN ||
cbw->Signature != cpu_to_le32(
US_BULK_CB_SIGN)) {
DBG(fsg, "invalid CBW: len %u sig 0x%x\n",
req->actual,
le32_to_cpu(cbw->Signature));
/*
* The Bulk-only spec says we MUST stall the IN endpoint
* (6.6.1), so it's unavoidable. It also says we must
* retain this state until the next reset, but there's
* no way to tell the controller driver it should ignore
* Clear-Feature(HALT) requests.
*
* We aren't required to halt the OUT endpoint; instead
* we can simply accept and discard any data received
* until the next reset.
*/
wedge_bulk_in_endpoint(fsg);
set_bit(IGNORE_BULK_OUT, &fsg->atomic_bitflags);
return -EINVAL;
}
/* Is the CBW meaningful? */
if (cbw->Lun >= ARRAY_SIZE(common->luns) ||
cbw->Flags & ~US_BULK_FLAG_IN || cbw->Length <= 0 ||
cbw->Length > MAX_COMMAND_SIZE) {
DBG(fsg, "non-meaningful CBW: lun = %u, flags = 0x%x, "
"cmdlen %u\n",
cbw->Lun, cbw->Flags, cbw->Length);
/*
* We can do anything we want here, so let's stall the
* bulk pipes if we are allowed to.
*/
if (common->can_stall) {
fsg_set_halt(fsg, fsg->bulk_out);
halt_bulk_in_endpoint(fsg);
}
return -EINVAL;
}
/* Save the command for later */
common->cmnd_size = cbw->Length;
memcpy(common->cmnd, cbw->CDB, common->cmnd_size);
if (cbw->Flags & US_BULK_FLAG_IN)
common->data_dir = DATA_DIR_TO_HOST;
else
common->data_dir = DATA_DIR_FROM_HOST;
common->data_size = le32_to_cpu(cbw->DataTransferLength);
if (common->data_size == 0)
common->data_dir = DATA_DIR_NONE;
common->lun = cbw->Lun;
if (common->lun < ARRAY_SIZE(common->luns))
usb: gadget: f_mass_storage: add a level of indirection for luns storage This is needed to prepare for configfs integration. So far the luns have been allocated during gadget's initialization, based on the nluns module parameter's value; the exact number is known when the gadget is initialized and that number of luns is allocated in one go; they all will be used. When configfs is in place, the luns will be created one-by-one by the user. Once the user is satisfied with the number of luns, they activate the gadget. The number of luns must be <= FSG_MAX_LUN (currently 8), but other than that it is not known up front and the user need not use contiguous numbering (apart from the default lun #0). On the other hand, the function code uses lun numbers to identify them and the number needs to be used as an index into an array. Given the above, an array needs to be allocated, but it might happen that 7 out of its 8 elements will not be used. On my machine sizeof(struct fsg_lun) == 462, so > 3k of memory is allocated but not used in the worst case. By adding another level of indirection (allocating an array of pointers to struct fsg_lun and then allocating individual luns instead of an array of struct fsg_luns) at most 7 pointers are wasted, which is much less. This patch also changes some for/while loops to cope with the fact that in the luns array some entries are potentially empty. Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@samsung.com> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2013-09-26 20:38:18 +08:00
common->curlun = common->luns[common->lun];
else
common->curlun = NULL;
common->tag = cbw->Tag;
return 0;
}
static int get_next_command(struct fsg_common *common)
{
struct fsg_buffhd *bh;
int rc = 0;
/* Wait for the next buffer to become available */
bh = common->next_buffhd_to_fill;
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
rc = sleep_thread(common, true, bh);
if (rc)
return rc;
/* Queue a request to read a Bulk-only CBW */
set_bulk_out_req_length(common, bh, US_BULK_CB_WRAP_LEN);
if (!start_out_transfer(common, bh))
/* Don't know what to do if common->fsg is NULL */
return -EIO;
/*
* We will drain the buffer in software, which means we
* can reuse it for the next filling. No need to advance
* next_buffhd_to_fill.
*/
/* Wait for the CBW to arrive */
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
rc = sleep_thread(common, true, bh);
if (rc)
return rc;
rc = fsg_is_set(common) ? received_cbw(common->fsg, bh) : -EIO;
bh->state = BUF_STATE_EMPTY;
return rc;
}
/*-------------------------------------------------------------------------*/
static int alloc_request(struct fsg_common *common, struct usb_ep *ep,
struct usb_request **preq)
{
*preq = usb_ep_alloc_request(ep, GFP_ATOMIC);
if (*preq)
return 0;
ERROR(common, "can't allocate request for %s\n", ep->name);
return -ENOMEM;
}
/* Reset interface setting and re-init endpoint state (toggle etc). */
static int do_set_interface(struct fsg_common *common, struct fsg_dev *new_fsg)
{
struct fsg_dev *fsg;
int i, rc = 0;
if (common->running)
DBG(common, "reset interface\n");
reset:
/* Deallocate the requests */
if (common->fsg) {
fsg = common->fsg;
for (i = 0; i < common->fsg_num_buffers; ++i) {
struct fsg_buffhd *bh = &common->buffhds[i];
if (bh->inreq) {
usb_ep_free_request(fsg->bulk_in, bh->inreq);
bh->inreq = NULL;
}
if (bh->outreq) {
usb_ep_free_request(fsg->bulk_out, bh->outreq);
bh->outreq = NULL;
}
}
/* Disable the endpoints */
if (fsg->bulk_in_enabled) {
usb_ep_disable(fsg->bulk_in);
fsg->bulk_in_enabled = 0;
}
if (fsg->bulk_out_enabled) {
usb_ep_disable(fsg->bulk_out);
fsg->bulk_out_enabled = 0;
}
common->fsg = NULL;
wake_up(&common->fsg_wait);
}
common->running = 0;
if (!new_fsg || rc)
return rc;
common->fsg = new_fsg;
fsg = common->fsg;
/* Enable the endpoints */
rc = config_ep_by_speed(common->gadget, &(fsg->function), fsg->bulk_in);
if (rc)
goto reset;
rc = usb_ep_enable(fsg->bulk_in);
if (rc)
goto reset;
fsg->bulk_in->driver_data = common;
fsg->bulk_in_enabled = 1;
rc = config_ep_by_speed(common->gadget, &(fsg->function),
fsg->bulk_out);
if (rc)
goto reset;
rc = usb_ep_enable(fsg->bulk_out);
if (rc)
goto reset;
fsg->bulk_out->driver_data = common;
fsg->bulk_out_enabled = 1;
USB: use usb_endpoint_maxp() instead of le16_to_cpu() Now ${LINUX}/drivers/usb/* can use usb_endpoint_maxp(desc) to get maximum packet size instead of le16_to_cpu(desc->wMaxPacketSize). This patch fix it up Cc: Armin Fuerst <fuerst@in.tum.de> Cc: Pavel Machek <pavel@ucw.cz> Cc: Johannes Erdfelt <johannes@erdfelt.com> Cc: Vojtech Pavlik <vojtech@suse.cz> Cc: Oliver Neukum <oliver@neukum.name> Cc: David Kubicek <dave@awk.cz> Cc: Johan Hovold <jhovold@gmail.com> Cc: Brad Hards <bhards@bigpond.net.au> Acked-by: Felipe Balbi <balbi@ti.com> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Thomas Dahlmann <dahlmann.thomas@arcor.de> Cc: David Brownell <david-b@pacbell.net> Cc: David Lopo <dlopo@chipidea.mips.com> Cc: Alan Stern <stern@rowland.harvard.edu> Cc: Michal Nazarewicz <m.nazarewicz@samsung.com> Cc: Xie Xiaobo <X.Xie@freescale.com> Cc: Li Yang <leoli@freescale.com> Cc: Jiang Bo <tanya.jiang@freescale.com> Cc: Yuan-hsin Chen <yhchen@faraday-tech.com> Cc: Darius Augulis <augulis.darius@gmail.com> Cc: Xiaochen Shen <xiaochen.shen@intel.com> Cc: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com> Cc: OKI SEMICONDUCTOR, <toshiharu-linux@dsn.okisemi.com> Cc: Robert Jarzmik <robert.jarzmik@free.fr> Cc: Ben Dooks <ben@simtec.co.uk> Cc: Thomas Abraham <thomas.ab@samsung.com> Cc: Herbert Pötzl <herbert@13thfloor.at> Cc: Arnaud Patard <arnaud.patard@rtp-net.org> Cc: Roman Weissgaerber <weissg@vienna.at> Acked-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Cc: Tony Olech <tony.olech@elandigitalsystems.com> Cc: Florian Floe Echtler <echtler@fs.tum.de> Cc: Christian Lucht <lucht@codemercs.com> Cc: Juergen Stuber <starblue@sourceforge.net> Cc: Georges Toth <g.toth@e-biz.lu> Cc: Bill Ryder <bryder@sgi.com> Cc: Kuba Ober <kuba@mareimbrium.org> Cc: Inaky Perez-Gonzalez <inaky.perez-gonzalez@intel.com> Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-08-23 18:12:03 +08:00
common->bulk_out_maxpacket = usb_endpoint_maxp(fsg->bulk_out->desc);
clear_bit(IGNORE_BULK_OUT, &fsg->atomic_bitflags);
/* Allocate the requests */
for (i = 0; i < common->fsg_num_buffers; ++i) {
struct fsg_buffhd *bh = &common->buffhds[i];
rc = alloc_request(common, fsg->bulk_in, &bh->inreq);
if (rc)
goto reset;
rc = alloc_request(common, fsg->bulk_out, &bh->outreq);
if (rc)
goto reset;
bh->inreq->buf = bh->outreq->buf = bh->buf;
bh->inreq->context = bh->outreq->context = bh;
bh->inreq->complete = bulk_in_complete;
bh->outreq->complete = bulk_out_complete;
}
common->running = 1;
for (i = 0; i < ARRAY_SIZE(common->luns); ++i)
usb: gadget: f_mass_storage: add a level of indirection for luns storage This is needed to prepare for configfs integration. So far the luns have been allocated during gadget's initialization, based on the nluns module parameter's value; the exact number is known when the gadget is initialized and that number of luns is allocated in one go; they all will be used. When configfs is in place, the luns will be created one-by-one by the user. Once the user is satisfied with the number of luns, they activate the gadget. The number of luns must be <= FSG_MAX_LUN (currently 8), but other than that it is not known up front and the user need not use contiguous numbering (apart from the default lun #0). On the other hand, the function code uses lun numbers to identify them and the number needs to be used as an index into an array. Given the above, an array needs to be allocated, but it might happen that 7 out of its 8 elements will not be used. On my machine sizeof(struct fsg_lun) == 462, so > 3k of memory is allocated but not used in the worst case. By adding another level of indirection (allocating an array of pointers to struct fsg_lun and then allocating individual luns instead of an array of struct fsg_luns) at most 7 pointers are wasted, which is much less. This patch also changes some for/while loops to cope with the fact that in the luns array some entries are potentially empty. Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@samsung.com> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2013-09-26 20:38:18 +08:00
if (common->luns[i])
common->luns[i]->unit_attention_data =
SS_RESET_OCCURRED;
return rc;
}
/****************************** ALT CONFIGS ******************************/
static int fsg_set_alt(struct usb_function *f, unsigned intf, unsigned alt)
{
struct fsg_dev *fsg = fsg_from_func(f);
fsg->common->new_fsg = fsg;
raise_exception(fsg->common, FSG_STATE_CONFIG_CHANGE);
return USB_GADGET_DELAYED_STATUS;
}
static void fsg_disable(struct usb_function *f)
{
struct fsg_dev *fsg = fsg_from_func(f);
fsg->common->new_fsg = NULL;
raise_exception(fsg->common, FSG_STATE_CONFIG_CHANGE);
}
/*-------------------------------------------------------------------------*/
static void handle_exception(struct fsg_common *common)
{
int i;
struct fsg_buffhd *bh;
enum fsg_state old_state;
struct fsg_lun *curlun;
unsigned int exception_req_tag;
/*
* Clear the existing signals. Anything but SIGUSR1 is converted
* into a high-priority EXIT exception.
*/
for (;;) {
int sig = kernel_dequeue_signal();
if (!sig)
break;
if (sig != SIGUSR1) {
spin_lock_irq(&common->lock);
if (common->state < FSG_STATE_EXIT)
DBG(common, "Main thread exiting on signal\n");
common->state = FSG_STATE_EXIT;
spin_unlock_irq(&common->lock);
}
}
/* Cancel all the pending transfers */
if (likely(common->fsg)) {
for (i = 0; i < common->fsg_num_buffers; ++i) {
bh = &common->buffhds[i];
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
if (bh->state == BUF_STATE_SENDING)
usb_ep_dequeue(common->fsg->bulk_in, bh->inreq);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
if (bh->state == BUF_STATE_RECEIVING)
usb_ep_dequeue(common->fsg->bulk_out,
bh->outreq);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
/* Wait for a transfer to become idle */
if (sleep_thread(common, false, bh))
return;
}
/* Clear out the controller's fifos */
if (common->fsg->bulk_in_enabled)
usb_ep_fifo_flush(common->fsg->bulk_in);
if (common->fsg->bulk_out_enabled)
usb_ep_fifo_flush(common->fsg->bulk_out);
}
/*
* Reset the I/O buffer states and pointers, the SCSI
* state, and the exception. Then invoke the handler.
*/
spin_lock_irq(&common->lock);
for (i = 0; i < common->fsg_num_buffers; ++i) {
bh = &common->buffhds[i];
bh->state = BUF_STATE_EMPTY;
}
common->next_buffhd_to_fill = &common->buffhds[0];
common->next_buffhd_to_drain = &common->buffhds[0];
exception_req_tag = common->exception_req_tag;
old_state = common->state;
common->state = FSG_STATE_NORMAL;
if (old_state != FSG_STATE_ABORT_BULK_OUT) {
for (i = 0; i < ARRAY_SIZE(common->luns); ++i) {
usb: gadget: f_mass_storage: add a level of indirection for luns storage This is needed to prepare for configfs integration. So far the luns have been allocated during gadget's initialization, based on the nluns module parameter's value; the exact number is known when the gadget is initialized and that number of luns is allocated in one go; they all will be used. When configfs is in place, the luns will be created one-by-one by the user. Once the user is satisfied with the number of luns, they activate the gadget. The number of luns must be <= FSG_MAX_LUN (currently 8), but other than that it is not known up front and the user need not use contiguous numbering (apart from the default lun #0). On the other hand, the function code uses lun numbers to identify them and the number needs to be used as an index into an array. Given the above, an array needs to be allocated, but it might happen that 7 out of its 8 elements will not be used. On my machine sizeof(struct fsg_lun) == 462, so > 3k of memory is allocated but not used in the worst case. By adding another level of indirection (allocating an array of pointers to struct fsg_lun and then allocating individual luns instead of an array of struct fsg_luns) at most 7 pointers are wasted, which is much less. This patch also changes some for/while loops to cope with the fact that in the luns array some entries are potentially empty. Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@samsung.com> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2013-09-26 20:38:18 +08:00
curlun = common->luns[i];
if (!curlun)
continue;
curlun->prevent_medium_removal = 0;
curlun->sense_data = SS_NO_SENSE;
curlun->unit_attention_data = SS_NO_SENSE;
curlun->sense_data_info = 0;
curlun->info_valid = 0;
}
}
spin_unlock_irq(&common->lock);
/* Carry out any extra actions required for the exception */
switch (old_state) {
case FSG_STATE_NORMAL:
break;
case FSG_STATE_ABORT_BULK_OUT:
send_status(common);
break;
case FSG_STATE_PROTOCOL_RESET:
/*
* In case we were forced against our will to halt a
* bulk endpoint, clear the halt now. (The SuperH UDC
* requires this.)
*/
if (!fsg_is_set(common))
break;
if (test_and_clear_bit(IGNORE_BULK_OUT,
&common->fsg->atomic_bitflags))
usb_ep_clear_halt(common->fsg->bulk_in);
if (common->ep0_req_tag == exception_req_tag)
ep0_queue(common); /* Complete the status stage */
/*
* Technically this should go here, but it would only be
* a waste of time. Ditto for the INTERFACE_CHANGE and
* CONFIG_CHANGE cases.
*/
/* for (i = 0; i < common->ARRAY_SIZE(common->luns); ++i) */
usb: gadget: f_mass_storage: add a level of indirection for luns storage This is needed to prepare for configfs integration. So far the luns have been allocated during gadget's initialization, based on the nluns module parameter's value; the exact number is known when the gadget is initialized and that number of luns is allocated in one go; they all will be used. When configfs is in place, the luns will be created one-by-one by the user. Once the user is satisfied with the number of luns, they activate the gadget. The number of luns must be <= FSG_MAX_LUN (currently 8), but other than that it is not known up front and the user need not use contiguous numbering (apart from the default lun #0). On the other hand, the function code uses lun numbers to identify them and the number needs to be used as an index into an array. Given the above, an array needs to be allocated, but it might happen that 7 out of its 8 elements will not be used. On my machine sizeof(struct fsg_lun) == 462, so > 3k of memory is allocated but not used in the worst case. By adding another level of indirection (allocating an array of pointers to struct fsg_lun and then allocating individual luns instead of an array of struct fsg_luns) at most 7 pointers are wasted, which is much less. This patch also changes some for/while loops to cope with the fact that in the luns array some entries are potentially empty. Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@samsung.com> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2013-09-26 20:38:18 +08:00
/* if (common->luns[i]) */
/* common->luns[i]->unit_attention_data = */
/* SS_RESET_OCCURRED; */
break;
case FSG_STATE_CONFIG_CHANGE:
do_set_interface(common, common->new_fsg);
if (common->new_fsg)
usb_composite_setup_continue(common->cdev);
break;
case FSG_STATE_EXIT:
do_set_interface(common, NULL); /* Free resources */
spin_lock_irq(&common->lock);
common->state = FSG_STATE_TERMINATED; /* Stop the thread */
spin_unlock_irq(&common->lock);
break;
case FSG_STATE_TERMINATED:
break;
}
}
/*-------------------------------------------------------------------------*/
static int fsg_main_thread(void *common_)
{
struct fsg_common *common = common_;
USB: g_mass_storage: Fix deadlock when driver is unbound As a holdover from the old g_file_storage gadget, the g_mass_storage legacy gadget driver attempts to unregister itself when its main operating thread terminates (if it hasn't been unregistered already). This is not strictly necessary; it was never more than an attempt to have the gadget fail cleanly if something went wrong and the main thread was killed. However, now that the UDC core manages gadget drivers independently of UDC drivers, this scheme doesn't work any more. A simple test: modprobe dummy-hcd modprobe g-mass-storage file=... rmmod dummy-hcd ends up in a deadlock with the following backtrace: sysrq: SysRq : Show Blocked State task PC stack pid father file-storage D 0 1130 2 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_preempt_disabled+0xd/0xf __mutex_lock.isra.1+0x129/0x224 ? _raw_spin_unlock_irqrestore+0x12/0x14 __mutex_lock_slowpath+0x12/0x14 mutex_lock+0x28/0x2b usb_gadget_unregister_driver+0x29/0x9b [udc_core] usb_composite_unregister+0x10/0x12 [libcomposite] msg_cleanup+0x1d/0x20 [g_mass_storage] msg_thread_exits+0xd/0xdd7 [g_mass_storage] fsg_main_thread+0x1395/0x13d6 [usb_f_mass_storage] ? __schedule+0x573/0x58c kthread+0xd9/0xdb ? do_set_interface+0x25c/0x25c [usb_f_mass_storage] ? init_completion+0x1e/0x1e ret_from_fork+0x19/0x24 rmmod D 0 1155 683 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_timeout+0x26/0xbc ? __schedule+0x573/0x58c do_wait_for_common+0xb3/0x128 ? usleep_range+0x81/0x81 ? wake_up_q+0x3f/0x3f wait_for_common+0x2e/0x45 wait_for_completion+0x17/0x19 fsg_common_put+0x34/0x81 [usb_f_mass_storage] fsg_free_inst+0x13/0x1e [usb_f_mass_storage] usb_put_function_instance+0x1a/0x25 [libcomposite] msg_unbind+0x2a/0x42 [g_mass_storage] __composite_unbind+0x4a/0x6f [libcomposite] composite_unbind+0x12/0x14 [libcomposite] usb_gadget_remove_driver+0x4f/0x77 [udc_core] usb_del_gadget_udc+0x52/0xcc [udc_core] dummy_udc_remove+0x27/0x2c [dummy_hcd] platform_drv_remove+0x1d/0x31 device_release_driver_internal+0xe9/0x16d device_release_driver+0x11/0x13 bus_remove_device+0xd2/0xe2 device_del+0x19f/0x221 ? selinux_capable+0x22/0x27 platform_device_del+0x21/0x63 platform_device_unregister+0x10/0x1a cleanup+0x20/0x817 [dummy_hcd] SyS_delete_module+0x10c/0x197 ? ____fput+0xd/0xf ? task_work_run+0x55/0x62 ? prepare_exit_to_usermode+0x65/0x75 do_fast_syscall_32+0x86/0xc3 entry_SYSENTER_32+0x4e/0x7c What happens is that removing the dummy-hcd driver causes the UDC core to unbind the gadget driver, which it does while holding the udc_lock mutex. The unbind routine in g_mass_storage tells the main thread to exit and waits for it to terminate. But as mentioned above, when the main thread exits it tries to unregister the mass-storage function driver. Via the composite framework this ends up calling usb_gadget_unregister_driver(), which tries to acquire the udc_lock mutex. The result is deadlock. The simplest way to fix the problem is not to be so clever: The main thread doesn't have to unregister the function driver. The side effects won't be so terrible; if the gadget is still attached to a USB host when the main thread is killed, it will appear to the host as though the gadget's firmware has crashed -- a reasonably accurate interpretation, and an all-too-common occurrence for USB mass-storage devices. In fact, the code to unregister the driver when the main thread exits is specific to g-mass-storage; it is not used when f-mass-storage is included as a function in a larger composite device. Therefore the entire mechanism responsible for this (the fsg_operations structure with its ->thread_exits method, the fsg_common_set_ops() routine, and the msg_thread_exits() callback routine) can all be eliminated. Even the msg_registered bitflag can be removed, because now the driver is unregistered in only one place rather than in two places. Signed-off-by: Alan Stern <stern@rowland.harvard.edu> CC: <stable@vger.kernel.org> Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com> Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-09-22 01:22:00 +08:00
int i;
/*
* Allow the thread to be killed by a signal, but set the signal mask
* to block everything but INT, TERM, KILL, and USR1.
*/
allow_signal(SIGINT);
allow_signal(SIGTERM);
allow_signal(SIGKILL);
allow_signal(SIGUSR1);
/* Allow the thread to be frozen */
set_freezable();
/* The main loop */
while (common->state != FSG_STATE_TERMINATED) {
if (exception_in_progress(common) || signal_pending(current)) {
handle_exception(common);
continue;
}
if (!common->running) {
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
sleep_thread(common, true, NULL);
continue;
}
if (get_next_command(common) || exception_in_progress(common))
continue;
if (do_scsi_command(common) || exception_in_progress(common))
continue;
if (finish_reply(common) || exception_in_progress(common))
continue;
send_status(common);
}
spin_lock_irq(&common->lock);
common->thread_task = NULL;
spin_unlock_irq(&common->lock);
USB: g_mass_storage: Fix deadlock when driver is unbound As a holdover from the old g_file_storage gadget, the g_mass_storage legacy gadget driver attempts to unregister itself when its main operating thread terminates (if it hasn't been unregistered already). This is not strictly necessary; it was never more than an attempt to have the gadget fail cleanly if something went wrong and the main thread was killed. However, now that the UDC core manages gadget drivers independently of UDC drivers, this scheme doesn't work any more. A simple test: modprobe dummy-hcd modprobe g-mass-storage file=... rmmod dummy-hcd ends up in a deadlock with the following backtrace: sysrq: SysRq : Show Blocked State task PC stack pid father file-storage D 0 1130 2 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_preempt_disabled+0xd/0xf __mutex_lock.isra.1+0x129/0x224 ? _raw_spin_unlock_irqrestore+0x12/0x14 __mutex_lock_slowpath+0x12/0x14 mutex_lock+0x28/0x2b usb_gadget_unregister_driver+0x29/0x9b [udc_core] usb_composite_unregister+0x10/0x12 [libcomposite] msg_cleanup+0x1d/0x20 [g_mass_storage] msg_thread_exits+0xd/0xdd7 [g_mass_storage] fsg_main_thread+0x1395/0x13d6 [usb_f_mass_storage] ? __schedule+0x573/0x58c kthread+0xd9/0xdb ? do_set_interface+0x25c/0x25c [usb_f_mass_storage] ? init_completion+0x1e/0x1e ret_from_fork+0x19/0x24 rmmod D 0 1155 683 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_timeout+0x26/0xbc ? __schedule+0x573/0x58c do_wait_for_common+0xb3/0x128 ? usleep_range+0x81/0x81 ? wake_up_q+0x3f/0x3f wait_for_common+0x2e/0x45 wait_for_completion+0x17/0x19 fsg_common_put+0x34/0x81 [usb_f_mass_storage] fsg_free_inst+0x13/0x1e [usb_f_mass_storage] usb_put_function_instance+0x1a/0x25 [libcomposite] msg_unbind+0x2a/0x42 [g_mass_storage] __composite_unbind+0x4a/0x6f [libcomposite] composite_unbind+0x12/0x14 [libcomposite] usb_gadget_remove_driver+0x4f/0x77 [udc_core] usb_del_gadget_udc+0x52/0xcc [udc_core] dummy_udc_remove+0x27/0x2c [dummy_hcd] platform_drv_remove+0x1d/0x31 device_release_driver_internal+0xe9/0x16d device_release_driver+0x11/0x13 bus_remove_device+0xd2/0xe2 device_del+0x19f/0x221 ? selinux_capable+0x22/0x27 platform_device_del+0x21/0x63 platform_device_unregister+0x10/0x1a cleanup+0x20/0x817 [dummy_hcd] SyS_delete_module+0x10c/0x197 ? ____fput+0xd/0xf ? task_work_run+0x55/0x62 ? prepare_exit_to_usermode+0x65/0x75 do_fast_syscall_32+0x86/0xc3 entry_SYSENTER_32+0x4e/0x7c What happens is that removing the dummy-hcd driver causes the UDC core to unbind the gadget driver, which it does while holding the udc_lock mutex. The unbind routine in g_mass_storage tells the main thread to exit and waits for it to terminate. But as mentioned above, when the main thread exits it tries to unregister the mass-storage function driver. Via the composite framework this ends up calling usb_gadget_unregister_driver(), which tries to acquire the udc_lock mutex. The result is deadlock. The simplest way to fix the problem is not to be so clever: The main thread doesn't have to unregister the function driver. The side effects won't be so terrible; if the gadget is still attached to a USB host when the main thread is killed, it will appear to the host as though the gadget's firmware has crashed -- a reasonably accurate interpretation, and an all-too-common occurrence for USB mass-storage devices. In fact, the code to unregister the driver when the main thread exits is specific to g-mass-storage; it is not used when f-mass-storage is included as a function in a larger composite device. Therefore the entire mechanism responsible for this (the fsg_operations structure with its ->thread_exits method, the fsg_common_set_ops() routine, and the msg_thread_exits() callback routine) can all be eliminated. Even the msg_registered bitflag can be removed, because now the driver is unregistered in only one place rather than in two places. Signed-off-by: Alan Stern <stern@rowland.harvard.edu> CC: <stable@vger.kernel.org> Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com> Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-09-22 01:22:00 +08:00
/* Eject media from all LUNs */
USB: g_mass_storage: Fix deadlock when driver is unbound As a holdover from the old g_file_storage gadget, the g_mass_storage legacy gadget driver attempts to unregister itself when its main operating thread terminates (if it hasn't been unregistered already). This is not strictly necessary; it was never more than an attempt to have the gadget fail cleanly if something went wrong and the main thread was killed. However, now that the UDC core manages gadget drivers independently of UDC drivers, this scheme doesn't work any more. A simple test: modprobe dummy-hcd modprobe g-mass-storage file=... rmmod dummy-hcd ends up in a deadlock with the following backtrace: sysrq: SysRq : Show Blocked State task PC stack pid father file-storage D 0 1130 2 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_preempt_disabled+0xd/0xf __mutex_lock.isra.1+0x129/0x224 ? _raw_spin_unlock_irqrestore+0x12/0x14 __mutex_lock_slowpath+0x12/0x14 mutex_lock+0x28/0x2b usb_gadget_unregister_driver+0x29/0x9b [udc_core] usb_composite_unregister+0x10/0x12 [libcomposite] msg_cleanup+0x1d/0x20 [g_mass_storage] msg_thread_exits+0xd/0xdd7 [g_mass_storage] fsg_main_thread+0x1395/0x13d6 [usb_f_mass_storage] ? __schedule+0x573/0x58c kthread+0xd9/0xdb ? do_set_interface+0x25c/0x25c [usb_f_mass_storage] ? init_completion+0x1e/0x1e ret_from_fork+0x19/0x24 rmmod D 0 1155 683 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_timeout+0x26/0xbc ? __schedule+0x573/0x58c do_wait_for_common+0xb3/0x128 ? usleep_range+0x81/0x81 ? wake_up_q+0x3f/0x3f wait_for_common+0x2e/0x45 wait_for_completion+0x17/0x19 fsg_common_put+0x34/0x81 [usb_f_mass_storage] fsg_free_inst+0x13/0x1e [usb_f_mass_storage] usb_put_function_instance+0x1a/0x25 [libcomposite] msg_unbind+0x2a/0x42 [g_mass_storage] __composite_unbind+0x4a/0x6f [libcomposite] composite_unbind+0x12/0x14 [libcomposite] usb_gadget_remove_driver+0x4f/0x77 [udc_core] usb_del_gadget_udc+0x52/0xcc [udc_core] dummy_udc_remove+0x27/0x2c [dummy_hcd] platform_drv_remove+0x1d/0x31 device_release_driver_internal+0xe9/0x16d device_release_driver+0x11/0x13 bus_remove_device+0xd2/0xe2 device_del+0x19f/0x221 ? selinux_capable+0x22/0x27 platform_device_del+0x21/0x63 platform_device_unregister+0x10/0x1a cleanup+0x20/0x817 [dummy_hcd] SyS_delete_module+0x10c/0x197 ? ____fput+0xd/0xf ? task_work_run+0x55/0x62 ? prepare_exit_to_usermode+0x65/0x75 do_fast_syscall_32+0x86/0xc3 entry_SYSENTER_32+0x4e/0x7c What happens is that removing the dummy-hcd driver causes the UDC core to unbind the gadget driver, which it does while holding the udc_lock mutex. The unbind routine in g_mass_storage tells the main thread to exit and waits for it to terminate. But as mentioned above, when the main thread exits it tries to unregister the mass-storage function driver. Via the composite framework this ends up calling usb_gadget_unregister_driver(), which tries to acquire the udc_lock mutex. The result is deadlock. The simplest way to fix the problem is not to be so clever: The main thread doesn't have to unregister the function driver. The side effects won't be so terrible; if the gadget is still attached to a USB host when the main thread is killed, it will appear to the host as though the gadget's firmware has crashed -- a reasonably accurate interpretation, and an all-too-common occurrence for USB mass-storage devices. In fact, the code to unregister the driver when the main thread exits is specific to g-mass-storage; it is not used when f-mass-storage is included as a function in a larger composite device. Therefore the entire mechanism responsible for this (the fsg_operations structure with its ->thread_exits method, the fsg_common_set_ops() routine, and the msg_thread_exits() callback routine) can all be eliminated. Even the msg_registered bitflag can be removed, because now the driver is unregistered in only one place rather than in two places. Signed-off-by: Alan Stern <stern@rowland.harvard.edu> CC: <stable@vger.kernel.org> Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com> Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-09-22 01:22:00 +08:00
down_write(&common->filesem);
for (i = 0; i < ARRAY_SIZE(common->luns); i++) {
struct fsg_lun *curlun = common->luns[i];
USB: g_mass_storage: Fix deadlock when driver is unbound As a holdover from the old g_file_storage gadget, the g_mass_storage legacy gadget driver attempts to unregister itself when its main operating thread terminates (if it hasn't been unregistered already). This is not strictly necessary; it was never more than an attempt to have the gadget fail cleanly if something went wrong and the main thread was killed. However, now that the UDC core manages gadget drivers independently of UDC drivers, this scheme doesn't work any more. A simple test: modprobe dummy-hcd modprobe g-mass-storage file=... rmmod dummy-hcd ends up in a deadlock with the following backtrace: sysrq: SysRq : Show Blocked State task PC stack pid father file-storage D 0 1130 2 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_preempt_disabled+0xd/0xf __mutex_lock.isra.1+0x129/0x224 ? _raw_spin_unlock_irqrestore+0x12/0x14 __mutex_lock_slowpath+0x12/0x14 mutex_lock+0x28/0x2b usb_gadget_unregister_driver+0x29/0x9b [udc_core] usb_composite_unregister+0x10/0x12 [libcomposite] msg_cleanup+0x1d/0x20 [g_mass_storage] msg_thread_exits+0xd/0xdd7 [g_mass_storage] fsg_main_thread+0x1395/0x13d6 [usb_f_mass_storage] ? __schedule+0x573/0x58c kthread+0xd9/0xdb ? do_set_interface+0x25c/0x25c [usb_f_mass_storage] ? init_completion+0x1e/0x1e ret_from_fork+0x19/0x24 rmmod D 0 1155 683 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_timeout+0x26/0xbc ? __schedule+0x573/0x58c do_wait_for_common+0xb3/0x128 ? usleep_range+0x81/0x81 ? wake_up_q+0x3f/0x3f wait_for_common+0x2e/0x45 wait_for_completion+0x17/0x19 fsg_common_put+0x34/0x81 [usb_f_mass_storage] fsg_free_inst+0x13/0x1e [usb_f_mass_storage] usb_put_function_instance+0x1a/0x25 [libcomposite] msg_unbind+0x2a/0x42 [g_mass_storage] __composite_unbind+0x4a/0x6f [libcomposite] composite_unbind+0x12/0x14 [libcomposite] usb_gadget_remove_driver+0x4f/0x77 [udc_core] usb_del_gadget_udc+0x52/0xcc [udc_core] dummy_udc_remove+0x27/0x2c [dummy_hcd] platform_drv_remove+0x1d/0x31 device_release_driver_internal+0xe9/0x16d device_release_driver+0x11/0x13 bus_remove_device+0xd2/0xe2 device_del+0x19f/0x221 ? selinux_capable+0x22/0x27 platform_device_del+0x21/0x63 platform_device_unregister+0x10/0x1a cleanup+0x20/0x817 [dummy_hcd] SyS_delete_module+0x10c/0x197 ? ____fput+0xd/0xf ? task_work_run+0x55/0x62 ? prepare_exit_to_usermode+0x65/0x75 do_fast_syscall_32+0x86/0xc3 entry_SYSENTER_32+0x4e/0x7c What happens is that removing the dummy-hcd driver causes the UDC core to unbind the gadget driver, which it does while holding the udc_lock mutex. The unbind routine in g_mass_storage tells the main thread to exit and waits for it to terminate. But as mentioned above, when the main thread exits it tries to unregister the mass-storage function driver. Via the composite framework this ends up calling usb_gadget_unregister_driver(), which tries to acquire the udc_lock mutex. The result is deadlock. The simplest way to fix the problem is not to be so clever: The main thread doesn't have to unregister the function driver. The side effects won't be so terrible; if the gadget is still attached to a USB host when the main thread is killed, it will appear to the host as though the gadget's firmware has crashed -- a reasonably accurate interpretation, and an all-too-common occurrence for USB mass-storage devices. In fact, the code to unregister the driver when the main thread exits is specific to g-mass-storage; it is not used when f-mass-storage is included as a function in a larger composite device. Therefore the entire mechanism responsible for this (the fsg_operations structure with its ->thread_exits method, the fsg_common_set_ops() routine, and the msg_thread_exits() callback routine) can all be eliminated. Even the msg_registered bitflag can be removed, because now the driver is unregistered in only one place rather than in two places. Signed-off-by: Alan Stern <stern@rowland.harvard.edu> CC: <stable@vger.kernel.org> Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com> Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-09-22 01:22:00 +08:00
if (curlun && fsg_lun_is_open(curlun))
fsg_lun_close(curlun);
}
USB: g_mass_storage: Fix deadlock when driver is unbound As a holdover from the old g_file_storage gadget, the g_mass_storage legacy gadget driver attempts to unregister itself when its main operating thread terminates (if it hasn't been unregistered already). This is not strictly necessary; it was never more than an attempt to have the gadget fail cleanly if something went wrong and the main thread was killed. However, now that the UDC core manages gadget drivers independently of UDC drivers, this scheme doesn't work any more. A simple test: modprobe dummy-hcd modprobe g-mass-storage file=... rmmod dummy-hcd ends up in a deadlock with the following backtrace: sysrq: SysRq : Show Blocked State task PC stack pid father file-storage D 0 1130 2 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_preempt_disabled+0xd/0xf __mutex_lock.isra.1+0x129/0x224 ? _raw_spin_unlock_irqrestore+0x12/0x14 __mutex_lock_slowpath+0x12/0x14 mutex_lock+0x28/0x2b usb_gadget_unregister_driver+0x29/0x9b [udc_core] usb_composite_unregister+0x10/0x12 [libcomposite] msg_cleanup+0x1d/0x20 [g_mass_storage] msg_thread_exits+0xd/0xdd7 [g_mass_storage] fsg_main_thread+0x1395/0x13d6 [usb_f_mass_storage] ? __schedule+0x573/0x58c kthread+0xd9/0xdb ? do_set_interface+0x25c/0x25c [usb_f_mass_storage] ? init_completion+0x1e/0x1e ret_from_fork+0x19/0x24 rmmod D 0 1155 683 0x00000000 Call Trace: __schedule+0x53e/0x58c schedule+0x6e/0x77 schedule_timeout+0x26/0xbc ? __schedule+0x573/0x58c do_wait_for_common+0xb3/0x128 ? usleep_range+0x81/0x81 ? wake_up_q+0x3f/0x3f wait_for_common+0x2e/0x45 wait_for_completion+0x17/0x19 fsg_common_put+0x34/0x81 [usb_f_mass_storage] fsg_free_inst+0x13/0x1e [usb_f_mass_storage] usb_put_function_instance+0x1a/0x25 [libcomposite] msg_unbind+0x2a/0x42 [g_mass_storage] __composite_unbind+0x4a/0x6f [libcomposite] composite_unbind+0x12/0x14 [libcomposite] usb_gadget_remove_driver+0x4f/0x77 [udc_core] usb_del_gadget_udc+0x52/0xcc [udc_core] dummy_udc_remove+0x27/0x2c [dummy_hcd] platform_drv_remove+0x1d/0x31 device_release_driver_internal+0xe9/0x16d device_release_driver+0x11/0x13 bus_remove_device+0xd2/0xe2 device_del+0x19f/0x221 ? selinux_capable+0x22/0x27 platform_device_del+0x21/0x63 platform_device_unregister+0x10/0x1a cleanup+0x20/0x817 [dummy_hcd] SyS_delete_module+0x10c/0x197 ? ____fput+0xd/0xf ? task_work_run+0x55/0x62 ? prepare_exit_to_usermode+0x65/0x75 do_fast_syscall_32+0x86/0xc3 entry_SYSENTER_32+0x4e/0x7c What happens is that removing the dummy-hcd driver causes the UDC core to unbind the gadget driver, which it does while holding the udc_lock mutex. The unbind routine in g_mass_storage tells the main thread to exit and waits for it to terminate. But as mentioned above, when the main thread exits it tries to unregister the mass-storage function driver. Via the composite framework this ends up calling usb_gadget_unregister_driver(), which tries to acquire the udc_lock mutex. The result is deadlock. The simplest way to fix the problem is not to be so clever: The main thread doesn't have to unregister the function driver. The side effects won't be so terrible; if the gadget is still attached to a USB host when the main thread is killed, it will appear to the host as though the gadget's firmware has crashed -- a reasonably accurate interpretation, and an all-too-common occurrence for USB mass-storage devices. In fact, the code to unregister the driver when the main thread exits is specific to g-mass-storage; it is not used when f-mass-storage is included as a function in a larger composite device. Therefore the entire mechanism responsible for this (the fsg_operations structure with its ->thread_exits method, the fsg_common_set_ops() routine, and the msg_thread_exits() callback routine) can all be eliminated. Even the msg_registered bitflag can be removed, because now the driver is unregistered in only one place rather than in two places. Signed-off-by: Alan Stern <stern@rowland.harvard.edu> CC: <stable@vger.kernel.org> Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com> Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-09-22 01:22:00 +08:00
up_write(&common->filesem);
/* Let fsg_unbind() know the thread has exited */
complete_and_exit(&common->thread_notifier, 0);
}
/*************************** DEVICE ATTRIBUTES ***************************/
static ssize_t ro_show(struct device *dev, struct device_attribute *attr, char *buf)
{
struct fsg_lun *curlun = fsg_lun_from_dev(dev);
return fsg_show_ro(curlun, buf);
}
static ssize_t nofua_show(struct device *dev, struct device_attribute *attr,
char *buf)
{
struct fsg_lun *curlun = fsg_lun_from_dev(dev);
return fsg_show_nofua(curlun, buf);
}
static ssize_t file_show(struct device *dev, struct device_attribute *attr,
char *buf)
{
struct fsg_lun *curlun = fsg_lun_from_dev(dev);
struct rw_semaphore *filesem = dev_get_drvdata(dev);
return fsg_show_file(curlun, filesem, buf);
}
static ssize_t ro_store(struct device *dev, struct device_attribute *attr,
const char *buf, size_t count)
{
struct fsg_lun *curlun = fsg_lun_from_dev(dev);
struct rw_semaphore *filesem = dev_get_drvdata(dev);
return fsg_store_ro(curlun, filesem, buf, count);
}
static ssize_t nofua_store(struct device *dev, struct device_attribute *attr,
const char *buf, size_t count)
{
struct fsg_lun *curlun = fsg_lun_from_dev(dev);
return fsg_store_nofua(curlun, buf, count);
}
static ssize_t file_store(struct device *dev, struct device_attribute *attr,
const char *buf, size_t count)
{
struct fsg_lun *curlun = fsg_lun_from_dev(dev);
struct rw_semaphore *filesem = dev_get_drvdata(dev);
return fsg_store_file(curlun, filesem, buf, count);
}
static DEVICE_ATTR_RW(nofua);
/* mode wil be set in fsg_lun_attr_is_visible() */
static DEVICE_ATTR(ro, 0, ro_show, ro_store);
static DEVICE_ATTR(file, 0, file_show, file_store);
/****************************** FSG COMMON ******************************/
static void fsg_lun_release(struct device *dev)
{
/* Nothing needs to be done */
}
static struct fsg_common *fsg_common_setup(struct fsg_common *common)
{
if (!common) {
common = kzalloc(sizeof(*common), GFP_KERNEL);
if (!common)
return ERR_PTR(-ENOMEM);
common->free_storage_on_release = 1;
} else {
common->free_storage_on_release = 0;
}
init_rwsem(&common->filesem);
spin_lock_init(&common->lock);
init_completion(&common->thread_notifier);
USB: f_mass_storage: improve memory barriers and synchronization This patch reworks the way f_mass_storage.c handles memory barriers and synchronization: The driver now uses a wait_queue instead of doing its own task-state manipulations (even though only one task will ever use the wait_queue). The thread_wakeup_needed variable is removed. It was only a source of trouble; although it was what the driver tested to see whether it should wake up, what we really wanted to see was whether a USB transfer had completed. All the explicit memory barriers scattered throughout the driver are replaced by a few calls to smp_load_acquire() and smp_store_release(). The inreq_busy and outreq_busy fields are removed. In their place, the driver keeps track of the current I/O direction by splitting BUF_STATE_BUSY into two states: BUF_STATE_SENDING and BUF_STATE_RECEIVING. The buffer states are no longer protected by a lock. Mutual exclusion isn't needed; the state is changed only by the driver's main thread when it owns the buffer, and only by the request completion routine when the gadget core owns the buffer. The do_write() and throw_away_data() routines were reorganized to make efficient use of the new sleeping mechanism. This resulted in the removal of one indentation level in those routines, making the patch appear to be more more complicated than it really is. In a few places, the driver allowed itself to be frozen although it really shouldn't have (in the middle of executing a SCSI command). Those places have been fixed. The logic in the exception handler for aborting transfers and waiting for them to stop has been simplified. Tested-by: Thinh Nguyen <thinhn@synopsys.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
2017-04-14 04:37:01 +08:00
init_waitqueue_head(&common->io_wait);
init_waitqueue_head(&common->fsg_wait);
common->state = FSG_STATE_TERMINATED;
memset(common->luns, 0, sizeof(common->luns));
return common;
}
void fsg_common_set_sysfs(struct fsg_common *common, bool sysfs)
{
common->sysfs = sysfs;
}
EXPORT_SYMBOL_GPL(fsg_common_set_sysfs);
static void _fsg_common_free_buffers(struct fsg_buffhd *buffhds, unsigned n)
{
if (buffhds) {
struct fsg_buffhd *bh = buffhds;
while (n--) {
kfree(bh->buf);
++bh;
}
kfree(buffhds);
}
}
int fsg_common_set_num_buffers(struct fsg_common *common, unsigned int n)
{
struct fsg_buffhd *bh, *buffhds;
int i;
buffhds = kcalloc(n, sizeof(*buffhds), GFP_KERNEL);
if (!buffhds)
return -ENOMEM;
/* Data buffers cyclic list */
bh = buffhds;
i = n;
goto buffhds_first_it;
do {
bh->next = bh + 1;
++bh;
buffhds_first_it:
bh->buf = kmalloc(FSG_BUFLEN, GFP_KERNEL);
if (unlikely(!bh->buf))
goto error_release;
} while (--i);
bh->next = buffhds;
_fsg_common_free_buffers(common->buffhds, common->fsg_num_buffers);
common->fsg_num_buffers = n;
common->buffhds = buffhds;
return 0;
error_release:
/*
* "buf"s pointed to by heads after n - i are NULL
* so releasing them won't hurt
*/
_fsg_common_free_buffers(buffhds, n);
return -ENOMEM;
}
EXPORT_SYMBOL_GPL(fsg_common_set_num_buffers);
usb: gadget: mass_storage: Fix freeing luns sysfs implementation Use device_is_registered() instad of sysfs flag to determine if we should free sysfs representation of particular LUN. sysfs flag in fsg common determines if luns attributes should be exposed using sysfs. This flag is used when creating and freeing luns. Unfortunately there is no guarantee that this flag will not be changed between creation and removal of particular LUN. Especially because of lun.0 which is created during allocating instance of function. This may lead to resource leak or NULL pointer dereference: [ 62.539925] Unable to handle kernel NULL pointer dereference at virtual address 00000044 [ 62.548014] pgd = ec994000 [ 62.550679] [00000044] *pgd=6d7be831, *pte=00000000, *ppte=00000000 [ 62.556933] Internal error: Oops: 17 [#1] PREEMPT SMP ARM [ 62.562310] Modules linked in: g_mass_storage(+) [ 62.566916] CPU: 2 PID: 613 Comm: insmod Not tainted 4.2.0-rc4-00077-ge29ee91-dirty #125 [ 62.574984] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree) [ 62.581061] task: eca56e80 ti: eca76000 task.ti: eca76000 [ 62.586450] PC is at kernfs_find_ns+0x8/0xe8 [ 62.590698] LR is at kernfs_find_and_get_ns+0x30/0x48 [ 62.595732] pc : [<c01277c0>] lr : [<c0127b88>] psr: 40010053 [ 62.595732] sp : eca77c40 ip : eca77c38 fp : 000008c1 [ 62.607187] r10: 00000001 r9 : c0082f38 r8 : ed41ce40 [ 62.612395] r7 : c05c1484 r6 : 00000000 r5 : 00000000 r4 : c0814488 [ 62.618904] r3 : 00000000 r2 : 00000000 r1 : c05c1484 r0 : 00000000 [ 62.625417] Flags: nZcv IRQs on FIQs off Mode SVC_32 ISA ARM Segment user [ 62.632620] Control: 10c5387d Table: 6c99404a DAC: 00000015 [ 62.638348] Process insmod (pid: 613, stack limit = 0xeca76210) [ 62.644251] Stack: (0xeca77c40 to 0xeca78000) [ 62.648594] 7c40: c0814488 00000000 00000000 c05c1484 ed41ce40 c0127b88 00000000 c0824888 [ 62.656753] 7c60: ed41d038 ed41d030 ed41d000 c012af4c 00000000 c0824858 ed41d038 c02e3314 [ 62.664912] 7c80: ed41d030 00000000 ed41ce04 c02d9e8c c070eda8 eca77cb4 000008c1 c058317c [ 62.673071] 7ca0: 000008c1 ed41d030 ed41ce00 ed41ce04 ed41d000 c02da044 ed41cf48 c0375870 [ 62.681230] 7cc0: ed9d3c04 ed9d3c00 ed52df80 bf000940 fffffff0 c03758f4 c03758c0 00000000 [ 62.689389] 7ce0: bf000564 c03614e0 ed9d3c04 bf000194 c0082f38 00000001 00000000 c0000100 [ 62.697548] 7d00: c0814488 c0814488 c086b1dc c05893a8 00000000 ed7e8320 00000000 c0128b88 [ 62.705707] 7d20: ed8a6b40 00000000 00000000 ed410500 ed8a6b40 c0594818 ed7e8320 00000000 [ 62.713867] 7d40: 00000000 c0129f20 00000000 c082c444 ed8a6b40 c012a684 00001000 00000000 [ 62.722026] 7d60: c0594818 c082c444 00000000 00000000 ed52df80 ed52df80 00000000 00000000 [ 62.730185] 7d80: 00000000 00000000 00000001 00000002 ed8e9b70 ed52df80 bf0006d0 00000000 [ 62.738345] 7da0: ed8e9b70 ed410500 ed618340 c036129c ed8c1c00 bf0006d0 c080b158 ed8c1c00 [ 62.746504] 7dc0: bf0006d0 c080b158 ed8c1c08 ed410500 c0082f38 ed618340 000008c1 c03640ac [ 62.754663] 7de0: 00000000 bf0006d0 c082c8dc c080b158 c080b158 c03642d4 00000000 bf003000 [ 62.762822] 7e00: 00000000 c0009784 00000000 00000001 00000000 c05849b0 00000002 ee7ab780 [ 62.770981] 7e20: 00000002 ed4105c0 0000c53e 000000d0 c0808600 eca77e5c 00000004 00000000 [ 62.779140] 7e40: bf000000 c0095680 c08075a0 ee001f00 ed4105c0 c00cadc0 ed52df80 bf000780 [ 62.787300] 7e60: ed4105c0 bf000780 00000001 bf0007c8 c0082f38 ed618340 000008c1 c0083e24 [ 62.795459] 7e80: 00000001 bf000780 00000001 eca77f58 00000001 bf000780 00000001 c00857f4 [ 62.803618] 7ea0: bf00078c 00007fff 00000000 c00835b4 eca77f58 00000000 c0082fac eca77f58 [ 62.811777] 7ec0: f05038c0 0003b008 bf000904 00000000 00000000 bf00078c 6e72656b 00006c65 [ 62.819936] 7ee0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.828095] 7f00: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.836255] 7f20: 00000000 00000000 00000000 00000000 00000000 00000000 00000003 0003b008 [ 62.844414] 7f40: 0000017b c000f5c8 eca76000 00000000 0003b008 c0085df8 f04ef000 0001b8a9 [ 62.852573] 7f60: f0503258 f05030c2 f0509fe8 00000968 00000dc8 00000000 00000000 00000000 [ 62.860732] 7f80: 00000029 0000002a 00000011 00000000 0000000a 00000000 33f6eb00 0003b008 [ 62.868892] 7fa0: bef01cac c000f400 33f6eb00 0003b008 00000003 0003b008 00000000 00000003 [ 62.877051] 7fc0: 33f6eb00 0003b008 bef01cac 0000017b 00000000 0003b008 0000000b 0003b008 [ 62.885210] 7fe0: bef01ae0 bef01ad0 0001dc23 b6e8c162 800b0070 00000003 c0c0c0c0 c0c0c0c0 [ 62.893380] [<c01277c0>] (kernfs_find_ns) from [<c0824888>] (pm_qos_latency_tolerance_attr_group+0x0/0x10) [ 62.903005] Code: e28dd00c e8bd80f0 e92d41f0 e2923000 (e1d0e4b4) [ 62.909115] ---[ end trace 02fb4373ef095c7b ]--- Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2015-07-31 19:37:45 +08:00
void fsg_common_remove_lun(struct fsg_lun *lun)
{
usb: gadget: mass_storage: Fix freeing luns sysfs implementation Use device_is_registered() instad of sysfs flag to determine if we should free sysfs representation of particular LUN. sysfs flag in fsg common determines if luns attributes should be exposed using sysfs. This flag is used when creating and freeing luns. Unfortunately there is no guarantee that this flag will not be changed between creation and removal of particular LUN. Especially because of lun.0 which is created during allocating instance of function. This may lead to resource leak or NULL pointer dereference: [ 62.539925] Unable to handle kernel NULL pointer dereference at virtual address 00000044 [ 62.548014] pgd = ec994000 [ 62.550679] [00000044] *pgd=6d7be831, *pte=00000000, *ppte=00000000 [ 62.556933] Internal error: Oops: 17 [#1] PREEMPT SMP ARM [ 62.562310] Modules linked in: g_mass_storage(+) [ 62.566916] CPU: 2 PID: 613 Comm: insmod Not tainted 4.2.0-rc4-00077-ge29ee91-dirty #125 [ 62.574984] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree) [ 62.581061] task: eca56e80 ti: eca76000 task.ti: eca76000 [ 62.586450] PC is at kernfs_find_ns+0x8/0xe8 [ 62.590698] LR is at kernfs_find_and_get_ns+0x30/0x48 [ 62.595732] pc : [<c01277c0>] lr : [<c0127b88>] psr: 40010053 [ 62.595732] sp : eca77c40 ip : eca77c38 fp : 000008c1 [ 62.607187] r10: 00000001 r9 : c0082f38 r8 : ed41ce40 [ 62.612395] r7 : c05c1484 r6 : 00000000 r5 : 00000000 r4 : c0814488 [ 62.618904] r3 : 00000000 r2 : 00000000 r1 : c05c1484 r0 : 00000000 [ 62.625417] Flags: nZcv IRQs on FIQs off Mode SVC_32 ISA ARM Segment user [ 62.632620] Control: 10c5387d Table: 6c99404a DAC: 00000015 [ 62.638348] Process insmod (pid: 613, stack limit = 0xeca76210) [ 62.644251] Stack: (0xeca77c40 to 0xeca78000) [ 62.648594] 7c40: c0814488 00000000 00000000 c05c1484 ed41ce40 c0127b88 00000000 c0824888 [ 62.656753] 7c60: ed41d038 ed41d030 ed41d000 c012af4c 00000000 c0824858 ed41d038 c02e3314 [ 62.664912] 7c80: ed41d030 00000000 ed41ce04 c02d9e8c c070eda8 eca77cb4 000008c1 c058317c [ 62.673071] 7ca0: 000008c1 ed41d030 ed41ce00 ed41ce04 ed41d000 c02da044 ed41cf48 c0375870 [ 62.681230] 7cc0: ed9d3c04 ed9d3c00 ed52df80 bf000940 fffffff0 c03758f4 c03758c0 00000000 [ 62.689389] 7ce0: bf000564 c03614e0 ed9d3c04 bf000194 c0082f38 00000001 00000000 c0000100 [ 62.697548] 7d00: c0814488 c0814488 c086b1dc c05893a8 00000000 ed7e8320 00000000 c0128b88 [ 62.705707] 7d20: ed8a6b40 00000000 00000000 ed410500 ed8a6b40 c0594818 ed7e8320 00000000 [ 62.713867] 7d40: 00000000 c0129f20 00000000 c082c444 ed8a6b40 c012a684 00001000 00000000 [ 62.722026] 7d60: c0594818 c082c444 00000000 00000000 ed52df80 ed52df80 00000000 00000000 [ 62.730185] 7d80: 00000000 00000000 00000001 00000002 ed8e9b70 ed52df80 bf0006d0 00000000 [ 62.738345] 7da0: ed8e9b70 ed410500 ed618340 c036129c ed8c1c00 bf0006d0 c080b158 ed8c1c00 [ 62.746504] 7dc0: bf0006d0 c080b158 ed8c1c08 ed410500 c0082f38 ed618340 000008c1 c03640ac [ 62.754663] 7de0: 00000000 bf0006d0 c082c8dc c080b158 c080b158 c03642d4 00000000 bf003000 [ 62.762822] 7e00: 00000000 c0009784 00000000 00000001 00000000 c05849b0 00000002 ee7ab780 [ 62.770981] 7e20: 00000002 ed4105c0 0000c53e 000000d0 c0808600 eca77e5c 00000004 00000000 [ 62.779140] 7e40: bf000000 c0095680 c08075a0 ee001f00 ed4105c0 c00cadc0 ed52df80 bf000780 [ 62.787300] 7e60: ed4105c0 bf000780 00000001 bf0007c8 c0082f38 ed618340 000008c1 c0083e24 [ 62.795459] 7e80: 00000001 bf000780 00000001 eca77f58 00000001 bf000780 00000001 c00857f4 [ 62.803618] 7ea0: bf00078c 00007fff 00000000 c00835b4 eca77f58 00000000 c0082fac eca77f58 [ 62.811777] 7ec0: f05038c0 0003b008 bf000904 00000000 00000000 bf00078c 6e72656b 00006c65 [ 62.819936] 7ee0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.828095] 7f00: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.836255] 7f20: 00000000 00000000 00000000 00000000 00000000 00000000 00000003 0003b008 [ 62.844414] 7f40: 0000017b c000f5c8 eca76000 00000000 0003b008 c0085df8 f04ef000 0001b8a9 [ 62.852573] 7f60: f0503258 f05030c2 f0509fe8 00000968 00000dc8 00000000 00000000 00000000 [ 62.860732] 7f80: 00000029 0000002a 00000011 00000000 0000000a 00000000 33f6eb00 0003b008 [ 62.868892] 7fa0: bef01cac c000f400 33f6eb00 0003b008 00000003 0003b008 00000000 00000003 [ 62.877051] 7fc0: 33f6eb00 0003b008 bef01cac 0000017b 00000000 0003b008 0000000b 0003b008 [ 62.885210] 7fe0: bef01ae0 bef01ad0 0001dc23 b6e8c162 800b0070 00000003 c0c0c0c0 c0c0c0c0 [ 62.893380] [<c01277c0>] (kernfs_find_ns) from [<c0824888>] (pm_qos_latency_tolerance_attr_group+0x0/0x10) [ 62.903005] Code: e28dd00c e8bd80f0 e92d41f0 e2923000 (e1d0e4b4) [ 62.909115] ---[ end trace 02fb4373ef095c7b ]--- Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2015-07-31 19:37:45 +08:00
if (device_is_registered(&lun->dev))
device_unregister(&lun->dev);
fsg_lun_close(lun);
kfree(lun);
}
EXPORT_SYMBOL_GPL(fsg_common_remove_lun);
static void _fsg_common_remove_luns(struct fsg_common *common, int n)
{
int i;
for (i = 0; i < n; ++i)
if (common->luns[i]) {
usb: gadget: mass_storage: Fix freeing luns sysfs implementation Use device_is_registered() instad of sysfs flag to determine if we should free sysfs representation of particular LUN. sysfs flag in fsg common determines if luns attributes should be exposed using sysfs. This flag is used when creating and freeing luns. Unfortunately there is no guarantee that this flag will not be changed between creation and removal of particular LUN. Especially because of lun.0 which is created during allocating instance of function. This may lead to resource leak or NULL pointer dereference: [ 62.539925] Unable to handle kernel NULL pointer dereference at virtual address 00000044 [ 62.548014] pgd = ec994000 [ 62.550679] [00000044] *pgd=6d7be831, *pte=00000000, *ppte=00000000 [ 62.556933] Internal error: Oops: 17 [#1] PREEMPT SMP ARM [ 62.562310] Modules linked in: g_mass_storage(+) [ 62.566916] CPU: 2 PID: 613 Comm: insmod Not tainted 4.2.0-rc4-00077-ge29ee91-dirty #125 [ 62.574984] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree) [ 62.581061] task: eca56e80 ti: eca76000 task.ti: eca76000 [ 62.586450] PC is at kernfs_find_ns+0x8/0xe8 [ 62.590698] LR is at kernfs_find_and_get_ns+0x30/0x48 [ 62.595732] pc : [<c01277c0>] lr : [<c0127b88>] psr: 40010053 [ 62.595732] sp : eca77c40 ip : eca77c38 fp : 000008c1 [ 62.607187] r10: 00000001 r9 : c0082f38 r8 : ed41ce40 [ 62.612395] r7 : c05c1484 r6 : 00000000 r5 : 00000000 r4 : c0814488 [ 62.618904] r3 : 00000000 r2 : 00000000 r1 : c05c1484 r0 : 00000000 [ 62.625417] Flags: nZcv IRQs on FIQs off Mode SVC_32 ISA ARM Segment user [ 62.632620] Control: 10c5387d Table: 6c99404a DAC: 00000015 [ 62.638348] Process insmod (pid: 613, stack limit = 0xeca76210) [ 62.644251] Stack: (0xeca77c40 to 0xeca78000) [ 62.648594] 7c40: c0814488 00000000 00000000 c05c1484 ed41ce40 c0127b88 00000000 c0824888 [ 62.656753] 7c60: ed41d038 ed41d030 ed41d000 c012af4c 00000000 c0824858 ed41d038 c02e3314 [ 62.664912] 7c80: ed41d030 00000000 ed41ce04 c02d9e8c c070eda8 eca77cb4 000008c1 c058317c [ 62.673071] 7ca0: 000008c1 ed41d030 ed41ce00 ed41ce04 ed41d000 c02da044 ed41cf48 c0375870 [ 62.681230] 7cc0: ed9d3c04 ed9d3c00 ed52df80 bf000940 fffffff0 c03758f4 c03758c0 00000000 [ 62.689389] 7ce0: bf000564 c03614e0 ed9d3c04 bf000194 c0082f38 00000001 00000000 c0000100 [ 62.697548] 7d00: c0814488 c0814488 c086b1dc c05893a8 00000000 ed7e8320 00000000 c0128b88 [ 62.705707] 7d20: ed8a6b40 00000000 00000000 ed410500 ed8a6b40 c0594818 ed7e8320 00000000 [ 62.713867] 7d40: 00000000 c0129f20 00000000 c082c444 ed8a6b40 c012a684 00001000 00000000 [ 62.722026] 7d60: c0594818 c082c444 00000000 00000000 ed52df80 ed52df80 00000000 00000000 [ 62.730185] 7d80: 00000000 00000000 00000001 00000002 ed8e9b70 ed52df80 bf0006d0 00000000 [ 62.738345] 7da0: ed8e9b70 ed410500 ed618340 c036129c ed8c1c00 bf0006d0 c080b158 ed8c1c00 [ 62.746504] 7dc0: bf0006d0 c080b158 ed8c1c08 ed410500 c0082f38 ed618340 000008c1 c03640ac [ 62.754663] 7de0: 00000000 bf0006d0 c082c8dc c080b158 c080b158 c03642d4 00000000 bf003000 [ 62.762822] 7e00: 00000000 c0009784 00000000 00000001 00000000 c05849b0 00000002 ee7ab780 [ 62.770981] 7e20: 00000002 ed4105c0 0000c53e 000000d0 c0808600 eca77e5c 00000004 00000000 [ 62.779140] 7e40: bf000000 c0095680 c08075a0 ee001f00 ed4105c0 c00cadc0 ed52df80 bf000780 [ 62.787300] 7e60: ed4105c0 bf000780 00000001 bf0007c8 c0082f38 ed618340 000008c1 c0083e24 [ 62.795459] 7e80: 00000001 bf000780 00000001 eca77f58 00000001 bf000780 00000001 c00857f4 [ 62.803618] 7ea0: bf00078c 00007fff 00000000 c00835b4 eca77f58 00000000 c0082fac eca77f58 [ 62.811777] 7ec0: f05038c0 0003b008 bf000904 00000000 00000000 bf00078c 6e72656b 00006c65 [ 62.819936] 7ee0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.828095] 7f00: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.836255] 7f20: 00000000 00000000 00000000 00000000 00000000 00000000 00000003 0003b008 [ 62.844414] 7f40: 0000017b c000f5c8 eca76000 00000000 0003b008 c0085df8 f04ef000 0001b8a9 [ 62.852573] 7f60: f0503258 f05030c2 f0509fe8 00000968 00000dc8 00000000 00000000 00000000 [ 62.860732] 7f80: 00000029 0000002a 00000011 00000000 0000000a 00000000 33f6eb00 0003b008 [ 62.868892] 7fa0: bef01cac c000f400 33f6eb00 0003b008 00000003 0003b008 00000000 00000003 [ 62.877051] 7fc0: 33f6eb00 0003b008 bef01cac 0000017b 00000000 0003b008 0000000b 0003b008 [ 62.885210] 7fe0: bef01ae0 bef01ad0 0001dc23 b6e8c162 800b0070 00000003 c0c0c0c0 c0c0c0c0 [ 62.893380] [<c01277c0>] (kernfs_find_ns) from [<c0824888>] (pm_qos_latency_tolerance_attr_group+0x0/0x10) [ 62.903005] Code: e28dd00c e8bd80f0 e92d41f0 e2923000 (e1d0e4b4) [ 62.909115] ---[ end trace 02fb4373ef095c7b ]--- Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2015-07-31 19:37:45 +08:00
fsg_common_remove_lun(common->luns[i]);
common->luns[i] = NULL;
}
}
void fsg_common_remove_luns(struct fsg_common *common)
{
_fsg_common_remove_luns(common, ARRAY_SIZE(common->luns));
}
EXPORT_SYMBOL_GPL(fsg_common_remove_luns);
void fsg_common_free_buffers(struct fsg_common *common)
{
_fsg_common_free_buffers(common->buffhds, common->fsg_num_buffers);
common->buffhds = NULL;
}
EXPORT_SYMBOL_GPL(fsg_common_free_buffers);
int fsg_common_set_cdev(struct fsg_common *common,
struct usb_composite_dev *cdev, bool can_stall)
{
struct usb_string *us;
common->gadget = cdev->gadget;
common->ep0 = cdev->gadget->ep0;
common->ep0req = cdev->req;
common->cdev = cdev;
us = usb_gstrings_attach(cdev, fsg_strings_array,
ARRAY_SIZE(fsg_strings));
if (IS_ERR(us))
return PTR_ERR(us);
fsg_intf_desc.iInterface = us[FSG_STRING_INTERFACE].id;
/*
* Some peripheral controllers are known not to be able to
* halt bulk endpoints correctly. If one of them is present,
* disable stalls.
*/
common->can_stall = can_stall &&
gadget_is_stall_supported(common->gadget);
return 0;
}
EXPORT_SYMBOL_GPL(fsg_common_set_cdev);
static struct attribute *fsg_lun_dev_attrs[] = {
&dev_attr_ro.attr,
&dev_attr_file.attr,
&dev_attr_nofua.attr,
NULL
};
static umode_t fsg_lun_dev_is_visible(struct kobject *kobj,
struct attribute *attr, int idx)
{
struct device *dev = kobj_to_dev(kobj);
struct fsg_lun *lun = fsg_lun_from_dev(dev);
if (attr == &dev_attr_ro.attr)
return lun->cdrom ? S_IRUGO : (S_IWUSR | S_IRUGO);
if (attr == &dev_attr_file.attr)
return lun->removable ? (S_IWUSR | S_IRUGO) : S_IRUGO;
return attr->mode;
}
static const struct attribute_group fsg_lun_dev_group = {
.attrs = fsg_lun_dev_attrs,
.is_visible = fsg_lun_dev_is_visible,
};
static const struct attribute_group *fsg_lun_dev_groups[] = {
&fsg_lun_dev_group,
NULL
};
int fsg_common_create_lun(struct fsg_common *common, struct fsg_lun_config *cfg,
unsigned int id, const char *name,
const char **name_pfx)
{
struct fsg_lun *lun;
char *pathbuf, *p;
int rc = -ENOMEM;
if (id >= ARRAY_SIZE(common->luns))
return -ENODEV;
if (common->luns[id])
return -EBUSY;
if (!cfg->filename && !cfg->removable) {
pr_err("no file given for LUN%d\n", id);
return -EINVAL;
}
lun = kzalloc(sizeof(*lun), GFP_KERNEL);
if (!lun)
return -ENOMEM;
lun->name_pfx = name_pfx;
lun->cdrom = !!cfg->cdrom;
lun->ro = cfg->cdrom || cfg->ro;
lun->initially_ro = lun->ro;
lun->removable = !!cfg->removable;
if (!common->sysfs) {
/* we DON'T own the name!*/
lun->name = name;
} else {
lun->dev.release = fsg_lun_release;
lun->dev.parent = &common->gadget->dev;
lun->dev.groups = fsg_lun_dev_groups;
dev_set_drvdata(&lun->dev, &common->filesem);
dev_set_name(&lun->dev, "%s", name);
lun->name = dev_name(&lun->dev);
rc = device_register(&lun->dev);
if (rc) {
pr_info("failed to register LUN%d: %d\n", id, rc);
put_device(&lun->dev);
goto error_sysfs;
}
}
common->luns[id] = lun;
if (cfg->filename) {
rc = fsg_lun_open(lun, cfg->filename);
if (rc)
goto error_lun;
}
pathbuf = kmalloc(PATH_MAX, GFP_KERNEL);
p = "(no medium)";
if (fsg_lun_is_open(lun)) {
p = "(error)";
if (pathbuf) {
p = file_path(lun->filp, pathbuf, PATH_MAX);
if (IS_ERR(p))
p = "(error)";
}
}
pr_info("LUN: %s%s%sfile: %s\n",
lun->removable ? "removable " : "",
lun->ro ? "read only " : "",
lun->cdrom ? "CD-ROM " : "",
p);
kfree(pathbuf);
return 0;
error_lun:
usb: gadget: mass_storage: Fix freeing luns sysfs implementation Use device_is_registered() instad of sysfs flag to determine if we should free sysfs representation of particular LUN. sysfs flag in fsg common determines if luns attributes should be exposed using sysfs. This flag is used when creating and freeing luns. Unfortunately there is no guarantee that this flag will not be changed between creation and removal of particular LUN. Especially because of lun.0 which is created during allocating instance of function. This may lead to resource leak or NULL pointer dereference: [ 62.539925] Unable to handle kernel NULL pointer dereference at virtual address 00000044 [ 62.548014] pgd = ec994000 [ 62.550679] [00000044] *pgd=6d7be831, *pte=00000000, *ppte=00000000 [ 62.556933] Internal error: Oops: 17 [#1] PREEMPT SMP ARM [ 62.562310] Modules linked in: g_mass_storage(+) [ 62.566916] CPU: 2 PID: 613 Comm: insmod Not tainted 4.2.0-rc4-00077-ge29ee91-dirty #125 [ 62.574984] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree) [ 62.581061] task: eca56e80 ti: eca76000 task.ti: eca76000 [ 62.586450] PC is at kernfs_find_ns+0x8/0xe8 [ 62.590698] LR is at kernfs_find_and_get_ns+0x30/0x48 [ 62.595732] pc : [<c01277c0>] lr : [<c0127b88>] psr: 40010053 [ 62.595732] sp : eca77c40 ip : eca77c38 fp : 000008c1 [ 62.607187] r10: 00000001 r9 : c0082f38 r8 : ed41ce40 [ 62.612395] r7 : c05c1484 r6 : 00000000 r5 : 00000000 r4 : c0814488 [ 62.618904] r3 : 00000000 r2 : 00000000 r1 : c05c1484 r0 : 00000000 [ 62.625417] Flags: nZcv IRQs on FIQs off Mode SVC_32 ISA ARM Segment user [ 62.632620] Control: 10c5387d Table: 6c99404a DAC: 00000015 [ 62.638348] Process insmod (pid: 613, stack limit = 0xeca76210) [ 62.644251] Stack: (0xeca77c40 to 0xeca78000) [ 62.648594] 7c40: c0814488 00000000 00000000 c05c1484 ed41ce40 c0127b88 00000000 c0824888 [ 62.656753] 7c60: ed41d038 ed41d030 ed41d000 c012af4c 00000000 c0824858 ed41d038 c02e3314 [ 62.664912] 7c80: ed41d030 00000000 ed41ce04 c02d9e8c c070eda8 eca77cb4 000008c1 c058317c [ 62.673071] 7ca0: 000008c1 ed41d030 ed41ce00 ed41ce04 ed41d000 c02da044 ed41cf48 c0375870 [ 62.681230] 7cc0: ed9d3c04 ed9d3c00 ed52df80 bf000940 fffffff0 c03758f4 c03758c0 00000000 [ 62.689389] 7ce0: bf000564 c03614e0 ed9d3c04 bf000194 c0082f38 00000001 00000000 c0000100 [ 62.697548] 7d00: c0814488 c0814488 c086b1dc c05893a8 00000000 ed7e8320 00000000 c0128b88 [ 62.705707] 7d20: ed8a6b40 00000000 00000000 ed410500 ed8a6b40 c0594818 ed7e8320 00000000 [ 62.713867] 7d40: 00000000 c0129f20 00000000 c082c444 ed8a6b40 c012a684 00001000 00000000 [ 62.722026] 7d60: c0594818 c082c444 00000000 00000000 ed52df80 ed52df80 00000000 00000000 [ 62.730185] 7d80: 00000000 00000000 00000001 00000002 ed8e9b70 ed52df80 bf0006d0 00000000 [ 62.738345] 7da0: ed8e9b70 ed410500 ed618340 c036129c ed8c1c00 bf0006d0 c080b158 ed8c1c00 [ 62.746504] 7dc0: bf0006d0 c080b158 ed8c1c08 ed410500 c0082f38 ed618340 000008c1 c03640ac [ 62.754663] 7de0: 00000000 bf0006d0 c082c8dc c080b158 c080b158 c03642d4 00000000 bf003000 [ 62.762822] 7e00: 00000000 c0009784 00000000 00000001 00000000 c05849b0 00000002 ee7ab780 [ 62.770981] 7e20: 00000002 ed4105c0 0000c53e 000000d0 c0808600 eca77e5c 00000004 00000000 [ 62.779140] 7e40: bf000000 c0095680 c08075a0 ee001f00 ed4105c0 c00cadc0 ed52df80 bf000780 [ 62.787300] 7e60: ed4105c0 bf000780 00000001 bf0007c8 c0082f38 ed618340 000008c1 c0083e24 [ 62.795459] 7e80: 00000001 bf000780 00000001 eca77f58 00000001 bf000780 00000001 c00857f4 [ 62.803618] 7ea0: bf00078c 00007fff 00000000 c00835b4 eca77f58 00000000 c0082fac eca77f58 [ 62.811777] 7ec0: f05038c0 0003b008 bf000904 00000000 00000000 bf00078c 6e72656b 00006c65 [ 62.819936] 7ee0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.828095] 7f00: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.836255] 7f20: 00000000 00000000 00000000 00000000 00000000 00000000 00000003 0003b008 [ 62.844414] 7f40: 0000017b c000f5c8 eca76000 00000000 0003b008 c0085df8 f04ef000 0001b8a9 [ 62.852573] 7f60: f0503258 f05030c2 f0509fe8 00000968 00000dc8 00000000 00000000 00000000 [ 62.860732] 7f80: 00000029 0000002a 00000011 00000000 0000000a 00000000 33f6eb00 0003b008 [ 62.868892] 7fa0: bef01cac c000f400 33f6eb00 0003b008 00000003 0003b008 00000000 00000003 [ 62.877051] 7fc0: 33f6eb00 0003b008 bef01cac 0000017b 00000000 0003b008 0000000b 0003b008 [ 62.885210] 7fe0: bef01ae0 bef01ad0 0001dc23 b6e8c162 800b0070 00000003 c0c0c0c0 c0c0c0c0 [ 62.893380] [<c01277c0>] (kernfs_find_ns) from [<c0824888>] (pm_qos_latency_tolerance_attr_group+0x0/0x10) [ 62.903005] Code: e28dd00c e8bd80f0 e92d41f0 e2923000 (e1d0e4b4) [ 62.909115] ---[ end trace 02fb4373ef095c7b ]--- Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2015-07-31 19:37:45 +08:00
if (device_is_registered(&lun->dev))
device_unregister(&lun->dev);
fsg_lun_close(lun);
common->luns[id] = NULL;
error_sysfs:
kfree(lun);
return rc;
}
EXPORT_SYMBOL_GPL(fsg_common_create_lun);
int fsg_common_create_luns(struct fsg_common *common, struct fsg_config *cfg)
{
char buf[8]; /* enough for 100000000 different numbers, decimal */
int i, rc;
fsg_common_remove_luns(common);
for (i = 0; i < cfg->nluns; ++i) {
snprintf(buf, sizeof(buf), "lun%d", i);
rc = fsg_common_create_lun(common, &cfg->luns[i], i, buf, NULL);
if (rc)
goto fail;
}
pr_info("Number of LUNs=%d\n", cfg->nluns);
return 0;
fail:
_fsg_common_remove_luns(common, i);
return rc;
}
EXPORT_SYMBOL_GPL(fsg_common_create_luns);
void fsg_common_set_inquiry_string(struct fsg_common *common, const char *vn,
const char *pn)
{
int i;
/* Prepare inquiryString */
i = get_default_bcdDevice();
snprintf(common->inquiry_string, sizeof(common->inquiry_string),
"%-8s%-16s%04x", vn ?: "Linux",
/* Assume product name dependent on the first LUN */
pn ?: ((*common->luns)->cdrom
? "File-CD Gadget"
: "File-Stor Gadget"),
i);
}
EXPORT_SYMBOL_GPL(fsg_common_set_inquiry_string);
static void fsg_common_release(struct fsg_common *common)
{
int i;
/* If the thread isn't already dead, tell it to exit now */
if (common->state != FSG_STATE_TERMINATED) {
raise_exception(common, FSG_STATE_EXIT);
wait_for_completion(&common->thread_notifier);
}
for (i = 0; i < ARRAY_SIZE(common->luns); ++i) {
struct fsg_lun *lun = common->luns[i];
if (!lun)
continue;
fsg_lun_close(lun);
usb: gadget: mass_storage: Fix freeing luns sysfs implementation Use device_is_registered() instad of sysfs flag to determine if we should free sysfs representation of particular LUN. sysfs flag in fsg common determines if luns attributes should be exposed using sysfs. This flag is used when creating and freeing luns. Unfortunately there is no guarantee that this flag will not be changed between creation and removal of particular LUN. Especially because of lun.0 which is created during allocating instance of function. This may lead to resource leak or NULL pointer dereference: [ 62.539925] Unable to handle kernel NULL pointer dereference at virtual address 00000044 [ 62.548014] pgd = ec994000 [ 62.550679] [00000044] *pgd=6d7be831, *pte=00000000, *ppte=00000000 [ 62.556933] Internal error: Oops: 17 [#1] PREEMPT SMP ARM [ 62.562310] Modules linked in: g_mass_storage(+) [ 62.566916] CPU: 2 PID: 613 Comm: insmod Not tainted 4.2.0-rc4-00077-ge29ee91-dirty #125 [ 62.574984] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree) [ 62.581061] task: eca56e80 ti: eca76000 task.ti: eca76000 [ 62.586450] PC is at kernfs_find_ns+0x8/0xe8 [ 62.590698] LR is at kernfs_find_and_get_ns+0x30/0x48 [ 62.595732] pc : [<c01277c0>] lr : [<c0127b88>] psr: 40010053 [ 62.595732] sp : eca77c40 ip : eca77c38 fp : 000008c1 [ 62.607187] r10: 00000001 r9 : c0082f38 r8 : ed41ce40 [ 62.612395] r7 : c05c1484 r6 : 00000000 r5 : 00000000 r4 : c0814488 [ 62.618904] r3 : 00000000 r2 : 00000000 r1 : c05c1484 r0 : 00000000 [ 62.625417] Flags: nZcv IRQs on FIQs off Mode SVC_32 ISA ARM Segment user [ 62.632620] Control: 10c5387d Table: 6c99404a DAC: 00000015 [ 62.638348] Process insmod (pid: 613, stack limit = 0xeca76210) [ 62.644251] Stack: (0xeca77c40 to 0xeca78000) [ 62.648594] 7c40: c0814488 00000000 00000000 c05c1484 ed41ce40 c0127b88 00000000 c0824888 [ 62.656753] 7c60: ed41d038 ed41d030 ed41d000 c012af4c 00000000 c0824858 ed41d038 c02e3314 [ 62.664912] 7c80: ed41d030 00000000 ed41ce04 c02d9e8c c070eda8 eca77cb4 000008c1 c058317c [ 62.673071] 7ca0: 000008c1 ed41d030 ed41ce00 ed41ce04 ed41d000 c02da044 ed41cf48 c0375870 [ 62.681230] 7cc0: ed9d3c04 ed9d3c00 ed52df80 bf000940 fffffff0 c03758f4 c03758c0 00000000 [ 62.689389] 7ce0: bf000564 c03614e0 ed9d3c04 bf000194 c0082f38 00000001 00000000 c0000100 [ 62.697548] 7d00: c0814488 c0814488 c086b1dc c05893a8 00000000 ed7e8320 00000000 c0128b88 [ 62.705707] 7d20: ed8a6b40 00000000 00000000 ed410500 ed8a6b40 c0594818 ed7e8320 00000000 [ 62.713867] 7d40: 00000000 c0129f20 00000000 c082c444 ed8a6b40 c012a684 00001000 00000000 [ 62.722026] 7d60: c0594818 c082c444 00000000 00000000 ed52df80 ed52df80 00000000 00000000 [ 62.730185] 7d80: 00000000 00000000 00000001 00000002 ed8e9b70 ed52df80 bf0006d0 00000000 [ 62.738345] 7da0: ed8e9b70 ed410500 ed618340 c036129c ed8c1c00 bf0006d0 c080b158 ed8c1c00 [ 62.746504] 7dc0: bf0006d0 c080b158 ed8c1c08 ed410500 c0082f38 ed618340 000008c1 c03640ac [ 62.754663] 7de0: 00000000 bf0006d0 c082c8dc c080b158 c080b158 c03642d4 00000000 bf003000 [ 62.762822] 7e00: 00000000 c0009784 00000000 00000001 00000000 c05849b0 00000002 ee7ab780 [ 62.770981] 7e20: 00000002 ed4105c0 0000c53e 000000d0 c0808600 eca77e5c 00000004 00000000 [ 62.779140] 7e40: bf000000 c0095680 c08075a0 ee001f00 ed4105c0 c00cadc0 ed52df80 bf000780 [ 62.787300] 7e60: ed4105c0 bf000780 00000001 bf0007c8 c0082f38 ed618340 000008c1 c0083e24 [ 62.795459] 7e80: 00000001 bf000780 00000001 eca77f58 00000001 bf000780 00000001 c00857f4 [ 62.803618] 7ea0: bf00078c 00007fff 00000000 c00835b4 eca77f58 00000000 c0082fac eca77f58 [ 62.811777] 7ec0: f05038c0 0003b008 bf000904 00000000 00000000 bf00078c 6e72656b 00006c65 [ 62.819936] 7ee0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.828095] 7f00: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.836255] 7f20: 00000000 00000000 00000000 00000000 00000000 00000000 00000003 0003b008 [ 62.844414] 7f40: 0000017b c000f5c8 eca76000 00000000 0003b008 c0085df8 f04ef000 0001b8a9 [ 62.852573] 7f60: f0503258 f05030c2 f0509fe8 00000968 00000dc8 00000000 00000000 00000000 [ 62.860732] 7f80: 00000029 0000002a 00000011 00000000 0000000a 00000000 33f6eb00 0003b008 [ 62.868892] 7fa0: bef01cac c000f400 33f6eb00 0003b008 00000003 0003b008 00000000 00000003 [ 62.877051] 7fc0: 33f6eb00 0003b008 bef01cac 0000017b 00000000 0003b008 0000000b 0003b008 [ 62.885210] 7fe0: bef01ae0 bef01ad0 0001dc23 b6e8c162 800b0070 00000003 c0c0c0c0 c0c0c0c0 [ 62.893380] [<c01277c0>] (kernfs_find_ns) from [<c0824888>] (pm_qos_latency_tolerance_attr_group+0x0/0x10) [ 62.903005] Code: e28dd00c e8bd80f0 e92d41f0 e2923000 (e1d0e4b4) [ 62.909115] ---[ end trace 02fb4373ef095c7b ]--- Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2015-07-31 19:37:45 +08:00
if (device_is_registered(&lun->dev))
device_unregister(&lun->dev);
kfree(lun);
}
_fsg_common_free_buffers(common->buffhds, common->fsg_num_buffers);
if (common->free_storage_on_release)
kfree(common);
}
/*-------------------------------------------------------------------------*/
static int fsg_bind(struct usb_configuration *c, struct usb_function *f)
{
struct fsg_dev *fsg = fsg_from_func(f);
struct fsg_common *common = fsg->common;
struct usb_gadget *gadget = c->cdev->gadget;
int i;
struct usb_ep *ep;
unsigned max_burst;
int ret;
struct fsg_opts *opts;
/* Don't allow to bind if we don't have at least one LUN */
ret = _fsg_common_get_max_lun(common);
if (ret < 0) {
pr_err("There should be at least one LUN.\n");
return -EINVAL;
}
opts = fsg_opts_from_func_inst(f->fi);
if (!opts->no_configfs) {
ret = fsg_common_set_cdev(fsg->common, c->cdev,
fsg->common->can_stall);
if (ret)
return ret;
fsg_common_set_inquiry_string(fsg->common, NULL, NULL);
}
if (!common->thread_task) {
common->state = FSG_STATE_NORMAL;
common->thread_task =
kthread_create(fsg_main_thread, common, "file-storage");
if (IS_ERR(common->thread_task)) {
ret = PTR_ERR(common->thread_task);
common->thread_task = NULL;
common->state = FSG_STATE_TERMINATED;
return ret;
}
DBG(common, "I/O thread pid: %d\n",
task_pid_nr(common->thread_task));
wake_up_process(common->thread_task);
}
fsg->gadget = gadget;
/* New interface */
i = usb_interface_id(c, f);
if (i < 0)
goto fail;
fsg_intf_desc.bInterfaceNumber = i;
fsg->interface_number = i;
/* Find all the endpoints we will use */
ep = usb_ep_autoconfig(gadget, &fsg_fs_bulk_in_desc);
if (!ep)
goto autoconf_fail;
fsg->bulk_in = ep;
ep = usb_ep_autoconfig(gadget, &fsg_fs_bulk_out_desc);
if (!ep)
goto autoconf_fail;
fsg->bulk_out = ep;
/* Assume endpoint addresses are the same for both speeds */
fsg_hs_bulk_in_desc.bEndpointAddress =
fsg_fs_bulk_in_desc.bEndpointAddress;
fsg_hs_bulk_out_desc.bEndpointAddress =
fsg_fs_bulk_out_desc.bEndpointAddress;
/* Calculate bMaxBurst, we know packet size is 1024 */
max_burst = min_t(unsigned, FSG_BUFLEN / 1024, 15);
fsg_ss_bulk_in_desc.bEndpointAddress =
fsg_fs_bulk_in_desc.bEndpointAddress;
fsg_ss_bulk_in_comp_desc.bMaxBurst = max_burst;
fsg_ss_bulk_out_desc.bEndpointAddress =
fsg_fs_bulk_out_desc.bEndpointAddress;
fsg_ss_bulk_out_comp_desc.bMaxBurst = max_burst;
ret = usb_assign_descriptors(f, fsg_fs_function, fsg_hs_function,
fsg_ss_function, fsg_ss_function);
if (ret)
goto autoconf_fail;
return 0;
autoconf_fail:
ERROR(fsg, "unable to autoconfigure all endpoints\n");
i = -ENOTSUPP;
fail:
/* terminate the thread */
if (fsg->common->state != FSG_STATE_TERMINATED) {
raise_exception(fsg->common, FSG_STATE_EXIT);
wait_for_completion(&fsg->common->thread_notifier);
}
return i;
}
/****************************** ALLOCATE FUNCTION *************************/
static void fsg_unbind(struct usb_configuration *c, struct usb_function *f)
{
struct fsg_dev *fsg = fsg_from_func(f);
struct fsg_common *common = fsg->common;
DBG(fsg, "unbind\n");
if (fsg->common->fsg == fsg) {
fsg->common->new_fsg = NULL;
raise_exception(fsg->common, FSG_STATE_CONFIG_CHANGE);
/* FIXME: make interruptible or killable somehow? */
wait_event(common->fsg_wait, common->fsg != fsg);
}
usb_free_all_descriptors(&fsg->function);
}
static inline struct fsg_lun_opts *to_fsg_lun_opts(struct config_item *item)
{
return container_of(to_config_group(item), struct fsg_lun_opts, group);
}
static inline struct fsg_opts *to_fsg_opts(struct config_item *item)
{
return container_of(to_config_group(item), struct fsg_opts,
func_inst.group);
}
static void fsg_lun_attr_release(struct config_item *item)
{
struct fsg_lun_opts *lun_opts;
lun_opts = to_fsg_lun_opts(item);
kfree(lun_opts);
}
static struct configfs_item_operations fsg_lun_item_ops = {
.release = fsg_lun_attr_release,
};
static ssize_t fsg_lun_opts_file_show(struct config_item *item, char *page)
{
struct fsg_lun_opts *opts = to_fsg_lun_opts(item);
struct fsg_opts *fsg_opts = to_fsg_opts(opts->group.cg_item.ci_parent);
return fsg_show_file(opts->lun, &fsg_opts->common->filesem, page);
}
static ssize_t fsg_lun_opts_file_store(struct config_item *item,
const char *page, size_t len)
{
struct fsg_lun_opts *opts = to_fsg_lun_opts(item);
struct fsg_opts *fsg_opts = to_fsg_opts(opts->group.cg_item.ci_parent);
return fsg_store_file(opts->lun, &fsg_opts->common->filesem, page, len);
}
CONFIGFS_ATTR(fsg_lun_opts_, file);
static ssize_t fsg_lun_opts_ro_show(struct config_item *item, char *page)
{
return fsg_show_ro(to_fsg_lun_opts(item)->lun, page);
}
static ssize_t fsg_lun_opts_ro_store(struct config_item *item,
const char *page, size_t len)
{
struct fsg_lun_opts *opts = to_fsg_lun_opts(item);
struct fsg_opts *fsg_opts = to_fsg_opts(opts->group.cg_item.ci_parent);
return fsg_store_ro(opts->lun, &fsg_opts->common->filesem, page, len);
}
CONFIGFS_ATTR(fsg_lun_opts_, ro);
static ssize_t fsg_lun_opts_removable_show(struct config_item *item,
char *page)
{
return fsg_show_removable(to_fsg_lun_opts(item)->lun, page);
}
static ssize_t fsg_lun_opts_removable_store(struct config_item *item,
const char *page, size_t len)
{
return fsg_store_removable(to_fsg_lun_opts(item)->lun, page, len);
}
CONFIGFS_ATTR(fsg_lun_opts_, removable);
static ssize_t fsg_lun_opts_cdrom_show(struct config_item *item, char *page)
{
return fsg_show_cdrom(to_fsg_lun_opts(item)->lun, page);
}
static ssize_t fsg_lun_opts_cdrom_store(struct config_item *item,
const char *page, size_t len)
{
struct fsg_lun_opts *opts = to_fsg_lun_opts(item);
struct fsg_opts *fsg_opts = to_fsg_opts(opts->group.cg_item.ci_parent);
return fsg_store_cdrom(opts->lun, &fsg_opts->common->filesem, page,
len);
}
CONFIGFS_ATTR(fsg_lun_opts_, cdrom);
static ssize_t fsg_lun_opts_nofua_show(struct config_item *item, char *page)
{
return fsg_show_nofua(to_fsg_lun_opts(item)->lun, page);
}
static ssize_t fsg_lun_opts_nofua_store(struct config_item *item,
const char *page, size_t len)
{
return fsg_store_nofua(to_fsg_lun_opts(item)->lun, page, len);
}
CONFIGFS_ATTR(fsg_lun_opts_, nofua);
static ssize_t fsg_lun_opts_inquiry_string_show(struct config_item *item,
char *page)
{
return fsg_show_inquiry_string(to_fsg_lun_opts(item)->lun, page);
}
static ssize_t fsg_lun_opts_inquiry_string_store(struct config_item *item,
const char *page, size_t len)
{
return fsg_store_inquiry_string(to_fsg_lun_opts(item)->lun, page, len);
}
CONFIGFS_ATTR(fsg_lun_opts_, inquiry_string);
static struct configfs_attribute *fsg_lun_attrs[] = {
&fsg_lun_opts_attr_file,
&fsg_lun_opts_attr_ro,
&fsg_lun_opts_attr_removable,
&fsg_lun_opts_attr_cdrom,
&fsg_lun_opts_attr_nofua,
&fsg_lun_opts_attr_inquiry_string,
NULL,
};
static const struct config_item_type fsg_lun_type = {
.ct_item_ops = &fsg_lun_item_ops,
.ct_attrs = fsg_lun_attrs,
.ct_owner = THIS_MODULE,
};
static struct config_group *fsg_lun_make(struct config_group *group,
const char *name)
{
struct fsg_lun_opts *opts;
struct fsg_opts *fsg_opts;
struct fsg_lun_config config;
char *num_str;
u8 num;
int ret;
num_str = strchr(name, '.');
if (!num_str) {
pr_err("Unable to locate . in LUN.NUMBER\n");
return ERR_PTR(-EINVAL);
}
num_str++;
ret = kstrtou8(num_str, 0, &num);
if (ret)
return ERR_PTR(ret);
fsg_opts = to_fsg_opts(&group->cg_item);
if (num >= FSG_MAX_LUNS)
return ERR_PTR(-ERANGE);
num = array_index_nospec(num, FSG_MAX_LUNS);
mutex_lock(&fsg_opts->lock);
if (fsg_opts->refcnt || fsg_opts->common->luns[num]) {
ret = -EBUSY;
goto out;
}
opts = kzalloc(sizeof(*opts), GFP_KERNEL);
if (!opts) {
ret = -ENOMEM;
goto out;
}
memset(&config, 0, sizeof(config));
config.removable = true;
ret = fsg_common_create_lun(fsg_opts->common, &config, num, name,
(const char **)&group->cg_item.ci_name);
if (ret) {
kfree(opts);
goto out;
}
opts->lun = fsg_opts->common->luns[num];
opts->lun_id = num;
mutex_unlock(&fsg_opts->lock);
config_group_init_type_name(&opts->group, name, &fsg_lun_type);
return &opts->group;
out:
mutex_unlock(&fsg_opts->lock);
return ERR_PTR(ret);
}
static void fsg_lun_drop(struct config_group *group, struct config_item *item)
{
struct fsg_lun_opts *lun_opts;
struct fsg_opts *fsg_opts;
lun_opts = to_fsg_lun_opts(item);
fsg_opts = to_fsg_opts(&group->cg_item);
mutex_lock(&fsg_opts->lock);
if (fsg_opts->refcnt) {
struct config_item *gadget;
gadget = group->cg_item.ci_parent->ci_parent;
unregister_gadget_item(gadget);
}
usb: gadget: mass_storage: Fix freeing luns sysfs implementation Use device_is_registered() instad of sysfs flag to determine if we should free sysfs representation of particular LUN. sysfs flag in fsg common determines if luns attributes should be exposed using sysfs. This flag is used when creating and freeing luns. Unfortunately there is no guarantee that this flag will not be changed between creation and removal of particular LUN. Especially because of lun.0 which is created during allocating instance of function. This may lead to resource leak or NULL pointer dereference: [ 62.539925] Unable to handle kernel NULL pointer dereference at virtual address 00000044 [ 62.548014] pgd = ec994000 [ 62.550679] [00000044] *pgd=6d7be831, *pte=00000000, *ppte=00000000 [ 62.556933] Internal error: Oops: 17 [#1] PREEMPT SMP ARM [ 62.562310] Modules linked in: g_mass_storage(+) [ 62.566916] CPU: 2 PID: 613 Comm: insmod Not tainted 4.2.0-rc4-00077-ge29ee91-dirty #125 [ 62.574984] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree) [ 62.581061] task: eca56e80 ti: eca76000 task.ti: eca76000 [ 62.586450] PC is at kernfs_find_ns+0x8/0xe8 [ 62.590698] LR is at kernfs_find_and_get_ns+0x30/0x48 [ 62.595732] pc : [<c01277c0>] lr : [<c0127b88>] psr: 40010053 [ 62.595732] sp : eca77c40 ip : eca77c38 fp : 000008c1 [ 62.607187] r10: 00000001 r9 : c0082f38 r8 : ed41ce40 [ 62.612395] r7 : c05c1484 r6 : 00000000 r5 : 00000000 r4 : c0814488 [ 62.618904] r3 : 00000000 r2 : 00000000 r1 : c05c1484 r0 : 00000000 [ 62.625417] Flags: nZcv IRQs on FIQs off Mode SVC_32 ISA ARM Segment user [ 62.632620] Control: 10c5387d Table: 6c99404a DAC: 00000015 [ 62.638348] Process insmod (pid: 613, stack limit = 0xeca76210) [ 62.644251] Stack: (0xeca77c40 to 0xeca78000) [ 62.648594] 7c40: c0814488 00000000 00000000 c05c1484 ed41ce40 c0127b88 00000000 c0824888 [ 62.656753] 7c60: ed41d038 ed41d030 ed41d000 c012af4c 00000000 c0824858 ed41d038 c02e3314 [ 62.664912] 7c80: ed41d030 00000000 ed41ce04 c02d9e8c c070eda8 eca77cb4 000008c1 c058317c [ 62.673071] 7ca0: 000008c1 ed41d030 ed41ce00 ed41ce04 ed41d000 c02da044 ed41cf48 c0375870 [ 62.681230] 7cc0: ed9d3c04 ed9d3c00 ed52df80 bf000940 fffffff0 c03758f4 c03758c0 00000000 [ 62.689389] 7ce0: bf000564 c03614e0 ed9d3c04 bf000194 c0082f38 00000001 00000000 c0000100 [ 62.697548] 7d00: c0814488 c0814488 c086b1dc c05893a8 00000000 ed7e8320 00000000 c0128b88 [ 62.705707] 7d20: ed8a6b40 00000000 00000000 ed410500 ed8a6b40 c0594818 ed7e8320 00000000 [ 62.713867] 7d40: 00000000 c0129f20 00000000 c082c444 ed8a6b40 c012a684 00001000 00000000 [ 62.722026] 7d60: c0594818 c082c444 00000000 00000000 ed52df80 ed52df80 00000000 00000000 [ 62.730185] 7d80: 00000000 00000000 00000001 00000002 ed8e9b70 ed52df80 bf0006d0 00000000 [ 62.738345] 7da0: ed8e9b70 ed410500 ed618340 c036129c ed8c1c00 bf0006d0 c080b158 ed8c1c00 [ 62.746504] 7dc0: bf0006d0 c080b158 ed8c1c08 ed410500 c0082f38 ed618340 000008c1 c03640ac [ 62.754663] 7de0: 00000000 bf0006d0 c082c8dc c080b158 c080b158 c03642d4 00000000 bf003000 [ 62.762822] 7e00: 00000000 c0009784 00000000 00000001 00000000 c05849b0 00000002 ee7ab780 [ 62.770981] 7e20: 00000002 ed4105c0 0000c53e 000000d0 c0808600 eca77e5c 00000004 00000000 [ 62.779140] 7e40: bf000000 c0095680 c08075a0 ee001f00 ed4105c0 c00cadc0 ed52df80 bf000780 [ 62.787300] 7e60: ed4105c0 bf000780 00000001 bf0007c8 c0082f38 ed618340 000008c1 c0083e24 [ 62.795459] 7e80: 00000001 bf000780 00000001 eca77f58 00000001 bf000780 00000001 c00857f4 [ 62.803618] 7ea0: bf00078c 00007fff 00000000 c00835b4 eca77f58 00000000 c0082fac eca77f58 [ 62.811777] 7ec0: f05038c0 0003b008 bf000904 00000000 00000000 bf00078c 6e72656b 00006c65 [ 62.819936] 7ee0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.828095] 7f00: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 62.836255] 7f20: 00000000 00000000 00000000 00000000 00000000 00000000 00000003 0003b008 [ 62.844414] 7f40: 0000017b c000f5c8 eca76000 00000000 0003b008 c0085df8 f04ef000 0001b8a9 [ 62.852573] 7f60: f0503258 f05030c2 f0509fe8 00000968 00000dc8 00000000 00000000 00000000 [ 62.860732] 7f80: 00000029 0000002a 00000011 00000000 0000000a 00000000 33f6eb00 0003b008 [ 62.868892] 7fa0: bef01cac c000f400 33f6eb00 0003b008 00000003 0003b008 00000000 00000003 [ 62.877051] 7fc0: 33f6eb00 0003b008 bef01cac 0000017b 00000000 0003b008 0000000b 0003b008 [ 62.885210] 7fe0: bef01ae0 bef01ad0 0001dc23 b6e8c162 800b0070 00000003 c0c0c0c0 c0c0c0c0 [ 62.893380] [<c01277c0>] (kernfs_find_ns) from [<c0824888>] (pm_qos_latency_tolerance_attr_group+0x0/0x10) [ 62.903005] Code: e28dd00c e8bd80f0 e92d41f0 e2923000 (e1d0e4b4) [ 62.909115] ---[ end trace 02fb4373ef095c7b ]--- Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Krzysztof Opasiak <k.opasiak@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com>
2015-07-31 19:37:45 +08:00
fsg_common_remove_lun(lun_opts->lun);
fsg_opts->common->luns[lun_opts->lun_id] = NULL;
lun_opts->lun_id = 0;
mutex_unlock(&fsg_opts->lock);
config_item_put(item);
}
static void fsg_attr_release(struct config_item *item)
{
struct fsg_opts *opts = to_fsg_opts(item);
usb_put_function_instance(&opts->func_inst);
}
static struct configfs_item_operations fsg_item_ops = {
.release = fsg_attr_release,
};
static ssize_t fsg_opts_stall_show(struct config_item *item, char *page)
{
struct fsg_opts *opts = to_fsg_opts(item);
int result;
mutex_lock(&opts->lock);
result = sprintf(page, "%d", opts->common->can_stall);
mutex_unlock(&opts->lock);
return result;
}
static ssize_t fsg_opts_stall_store(struct config_item *item, const char *page,
size_t len)
{
struct fsg_opts *opts = to_fsg_opts(item);
int ret;
bool stall;
mutex_lock(&opts->lock);
if (opts->refcnt) {
mutex_unlock(&opts->lock);
return -EBUSY;
}
ret = strtobool(page, &stall);
if (!ret) {
opts->common->can_stall = stall;
ret = len;
}
mutex_unlock(&opts->lock);
return ret;
}
CONFIGFS_ATTR(fsg_opts_, stall);
#ifdef CONFIG_USB_GADGET_DEBUG_FILES
static ssize_t fsg_opts_num_buffers_show(struct config_item *item, char *page)
{
struct fsg_opts *opts = to_fsg_opts(item);
int result;
mutex_lock(&opts->lock);
result = sprintf(page, "%d", opts->common->fsg_num_buffers);
mutex_unlock(&opts->lock);
return result;
}
static ssize_t fsg_opts_num_buffers_store(struct config_item *item,
const char *page, size_t len)
{
struct fsg_opts *opts = to_fsg_opts(item);
int ret;
u8 num;
mutex_lock(&opts->lock);
if (opts->refcnt) {
ret = -EBUSY;
goto end;
}
ret = kstrtou8(page, 0, &num);
if (ret)
goto end;
ret = fsg_common_set_num_buffers(opts->common, num);
if (ret)
goto end;
ret = len;
end:
mutex_unlock(&opts->lock);
return ret;
}
CONFIGFS_ATTR(fsg_opts_, num_buffers);
#endif
static struct configfs_attribute *fsg_attrs[] = {
&fsg_opts_attr_stall,
#ifdef CONFIG_USB_GADGET_DEBUG_FILES
&fsg_opts_attr_num_buffers,
#endif
NULL,
};
static struct configfs_group_operations fsg_group_ops = {
.make_group = fsg_lun_make,
.drop_item = fsg_lun_drop,
};
static const struct config_item_type fsg_func_type = {
.ct_item_ops = &fsg_item_ops,
.ct_group_ops = &fsg_group_ops,
.ct_attrs = fsg_attrs,
.ct_owner = THIS_MODULE,
};
static void fsg_free_inst(struct usb_function_instance *fi)
{
struct fsg_opts *opts;
opts = fsg_opts_from_func_inst(fi);
fsg_common_release(opts->common);
kfree(opts);
}
static struct usb_function_instance *fsg_alloc_inst(void)
{
struct fsg_opts *opts;
struct fsg_lun_config config;
int rc;
opts = kzalloc(sizeof(*opts), GFP_KERNEL);
if (!opts)
return ERR_PTR(-ENOMEM);
mutex_init(&opts->lock);
opts->func_inst.free_func_inst = fsg_free_inst;
opts->common = fsg_common_setup(opts->common);
if (IS_ERR(opts->common)) {
rc = PTR_ERR(opts->common);
goto release_opts;
}
rc = fsg_common_set_num_buffers(opts->common,
CONFIG_USB_GADGET_STORAGE_NUM_BUFFERS);
if (rc)
goto release_common;
pr_info(FSG_DRIVER_DESC ", version: " FSG_DRIVER_VERSION "\n");
memset(&config, 0, sizeof(config));
config.removable = true;
rc = fsg_common_create_lun(opts->common, &config, 0, "lun.0",
(const char **)&opts->func_inst.group.cg_item.ci_name);
if (rc)
goto release_buffers;
opts->lun0.lun = opts->common->luns[0];
opts->lun0.lun_id = 0;
config_group_init_type_name(&opts->func_inst.group, "", &fsg_func_type);
config_group_init_type_name(&opts->lun0.group, "lun.0", &fsg_lun_type);
configfs_add_default_group(&opts->lun0.group, &opts->func_inst.group);
return &opts->func_inst;
release_buffers:
fsg_common_free_buffers(opts->common);
release_common:
kfree(opts->common);
release_opts:
kfree(opts);
return ERR_PTR(rc);
}
static void fsg_free(struct usb_function *f)
{
struct fsg_dev *fsg;
struct fsg_opts *opts;
fsg = container_of(f, struct fsg_dev, function);
opts = container_of(f->fi, struct fsg_opts, func_inst);
mutex_lock(&opts->lock);
opts->refcnt--;
mutex_unlock(&opts->lock);
kfree(fsg);
}
static struct usb_function *fsg_alloc(struct usb_function_instance *fi)
{
struct fsg_opts *opts = fsg_opts_from_func_inst(fi);
struct fsg_common *common = opts->common;
struct fsg_dev *fsg;
fsg = kzalloc(sizeof(*fsg), GFP_KERNEL);
if (unlikely(!fsg))
return ERR_PTR(-ENOMEM);
mutex_lock(&opts->lock);
opts->refcnt++;
mutex_unlock(&opts->lock);
fsg->function.name = FSG_DRIVER_DESC;
fsg->function.bind = fsg_bind;
fsg->function.unbind = fsg_unbind;
fsg->function.setup = fsg_setup;
fsg->function.set_alt = fsg_set_alt;
fsg->function.disable = fsg_disable;
fsg->function.free_func = fsg_free;
fsg->common = common;
return &fsg->function;
}
DECLARE_USB_FUNCTION_INIT(mass_storage, fsg_alloc_inst, fsg_alloc);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Michal Nazarewicz");
/************************* Module parameters *************************/
void fsg_config_from_params(struct fsg_config *cfg,
const struct fsg_module_parameters *params,
unsigned int fsg_num_buffers)
{
struct fsg_lun_config *lun;
unsigned i;
/* Configure LUNs */
cfg->nluns =
min(params->luns ?: (params->file_count ?: 1u),
(unsigned)FSG_MAX_LUNS);
for (i = 0, lun = cfg->luns; i < cfg->nluns; ++i, ++lun) {
lun->ro = !!params->ro[i];
lun->cdrom = !!params->cdrom[i];
lun->removable = !!params->removable[i];
lun->filename =
params->file_count > i && params->file[i][0]
? params->file[i]
: NULL;
}
/* Let MSF use defaults */
cfg->vendor_name = NULL;
cfg->product_name = NULL;
cfg->ops = NULL;
cfg->private_data = NULL;
/* Finalise */
cfg->can_stall = params->stall;
cfg->fsg_num_buffers = fsg_num_buffers;
}
EXPORT_SYMBOL_GPL(fsg_config_from_params);