linux_dsm_epyc7002/drivers/base/firmware_class.c

1909 lines
45 KiB
C
Raw Normal View History

/*
* firmware_class.c - Multi purpose firmware loading support
*
* Copyright (c) 2003 Manuel Estrada Sainz
*
* Please see Documentation/firmware_class/ for more information.
*
*/
#include <linux/capability.h>
#include <linux/device.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/timer.h>
#include <linux/vmalloc.h>
#include <linux/interrupt.h>
#include <linux/bitops.h>
#include <linux/mutex.h>
#include <linux/workqueue.h>
#include <linux/highmem.h>
#include <linux/firmware.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 15:04:11 +07:00
#include <linux/slab.h>
#include <linux/sched.h>
#include <linux/file.h>
#include <linux/list.h>
#include <linux/fs.h>
#include <linux/async.h>
#include <linux/pm.h>
#include <linux/suspend.h>
#include <linux/syscore_ops.h>
#include <linux/reboot.h>
#include <linux/security.h>
#include <generated/utsrelease.h>
#include "base.h"
MODULE_AUTHOR("Manuel Estrada Sainz");
MODULE_DESCRIPTION("Multi purpose firmware loading support");
MODULE_LICENSE("GPL");
/* Builtin firmware support */
#ifdef CONFIG_FW_LOADER
extern struct builtin_fw __start_builtin_fw[];
extern struct builtin_fw __end_builtin_fw[];
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
static bool fw_get_builtin_firmware(struct firmware *fw, const char *name,
void *buf, size_t size)
{
struct builtin_fw *b_fw;
for (b_fw = __start_builtin_fw; b_fw != __end_builtin_fw; b_fw++) {
if (strcmp(name, b_fw->name) == 0) {
fw->size = b_fw->size;
fw->data = b_fw->data;
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
if (buf && fw->size <= size)
memcpy(buf, fw->data, fw->size);
return true;
}
}
return false;
}
static bool fw_is_builtin_firmware(const struct firmware *fw)
{
struct builtin_fw *b_fw;
for (b_fw = __start_builtin_fw; b_fw != __end_builtin_fw; b_fw++)
if (fw->data == b_fw->data)
return true;
return false;
}
#else /* Module case - no builtin firmware support */
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
static inline bool fw_get_builtin_firmware(struct firmware *fw,
const char *name, void *buf,
size_t size)
{
return false;
}
static inline bool fw_is_builtin_firmware(const struct firmware *fw)
{
return false;
}
#endif
enum fw_status {
FW_STATUS_UNKNOWN,
FW_STATUS_LOADING,
FW_STATUS_DONE,
FW_STATUS_ABORTED,
};
static int loading_timeout = 60; /* In seconds */
static inline long firmware_loading_timeout(void)
{
return loading_timeout > 0 ? loading_timeout * HZ : MAX_JIFFY_OFFSET;
}
/*
* Concurrent request_firmware() for the same firmware need to be
* serialized. struct fw_state is simple state machine which hold the
* state of the firmware loading.
*/
struct fw_state {
firmware: fix batched requests - wake all waiters The firmware cache mechanism serves two purposes, the secondary purpose is not well documented nor understood. This fixes a regression with the secondary purpose of the firmware cache mechanism: batched requests on successful lookups. Without this fix *any* time a batched request is triggered, secondary requests for which the batched request mechanism was designed for will seem to last forver and seem to never return. This issue is present for all kernel builds possible, and a hard reset is required. The firmware cache is used for: 1) Addressing races with file lookups during the suspend/resume cycle by keeping firmware in memory during the suspend/resume cycle 2) Batched requests for the same file rely only on work from the first file lookup, which keeps the firmware in memory until the last release_firmware() is called Batched requests *only* take effect if secondary requests come in prior to the first user calling release_firmware(). The devres name used for the internal firmware cache is used as a hint other pending requests are ongoing, the firmware buffer data is kept in memory until the last user of the buffer calls release_firmware(), therefore serializing requests and delaying the release until all requests are done. Batched requests wait for a wakup or signal so we can rely on the first file fetch to write to the pending secondary requests. Commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") ported the firmware API to use swait, and in doing so failed to convert complete_all() to swake_up_all() -- it used swake_up(), loosing the ability for *some* batched requests to take effect. We *could* fix this by just using swake_up_all() *but* swait is now known to be very special use case, so its best to just move away from it. So we just go back to using completions as before commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") given this was using complete_all(). Without this fix it has been reported plugging in two Intel 6260 Wifi cards on a system will end up enumerating the two devices only 50% of the time [0]. The ported swake_up() should have actually handled the case with two devices, however, *if more than two cards are used* the swake_up() would not have sufficed. This change is only part of the required fixes for batched requests. Another fix is provided in the next patch. This particular change should fix the cases where more than three requests with the same firmware name is used, otherwise batched requests will wait for MAX_SCHEDULE_TIMEOUT and just timeout eventually. Below is a summary of tests triggering batched requests on different kernel builds. Before this patch: ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=y Most common Linux distribution setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=n Only possible if CONFIG_DELL_RBU=n and CONFIG_LEDS_LP55XX_COMMON=n, rare. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y CONFIG_FW_LOADER_USER_HELPER=y Google Android setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ After this patch: ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=y Most common Linux distribution setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) FAIL OK request_firmware_nowait(uevent=false) FAIL OK ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=n Only possible if CONFIG_DELL_RBU=n and CONFIG_LEDS_LP55XX_COMMON=n, rare. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) FAIL OK request_firmware_nowait(uevent=false) FAIL OK ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y CONFIG_FW_LOADER_USER_HELPER=y Google Android setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() OK OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) OK OK request_firmware_nowait(uevent=false) OK OK ============================================================================ [0] https://bugzilla.kernel.org/show_bug.cgi?id=195477 CC: <stable@vger.kernel.org> [4.10+] Cc: Ming Lei <ming.lei@redhat.com> Fixes: 5b029624948d ("firmware: do not use fw_lock for fw_state protection") Reported-by: Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-07-21 03:13:09 +07:00
struct completion completion;
enum fw_status status;
};
static void fw_state_init(struct fw_state *fw_st)
{
firmware: fix batched requests - wake all waiters The firmware cache mechanism serves two purposes, the secondary purpose is not well documented nor understood. This fixes a regression with the secondary purpose of the firmware cache mechanism: batched requests on successful lookups. Without this fix *any* time a batched request is triggered, secondary requests for which the batched request mechanism was designed for will seem to last forver and seem to never return. This issue is present for all kernel builds possible, and a hard reset is required. The firmware cache is used for: 1) Addressing races with file lookups during the suspend/resume cycle by keeping firmware in memory during the suspend/resume cycle 2) Batched requests for the same file rely only on work from the first file lookup, which keeps the firmware in memory until the last release_firmware() is called Batched requests *only* take effect if secondary requests come in prior to the first user calling release_firmware(). The devres name used for the internal firmware cache is used as a hint other pending requests are ongoing, the firmware buffer data is kept in memory until the last user of the buffer calls release_firmware(), therefore serializing requests and delaying the release until all requests are done. Batched requests wait for a wakup or signal so we can rely on the first file fetch to write to the pending secondary requests. Commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") ported the firmware API to use swait, and in doing so failed to convert complete_all() to swake_up_all() -- it used swake_up(), loosing the ability for *some* batched requests to take effect. We *could* fix this by just using swake_up_all() *but* swait is now known to be very special use case, so its best to just move away from it. So we just go back to using completions as before commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") given this was using complete_all(). Without this fix it has been reported plugging in two Intel 6260 Wifi cards on a system will end up enumerating the two devices only 50% of the time [0]. The ported swake_up() should have actually handled the case with two devices, however, *if more than two cards are used* the swake_up() would not have sufficed. This change is only part of the required fixes for batched requests. Another fix is provided in the next patch. This particular change should fix the cases where more than three requests with the same firmware name is used, otherwise batched requests will wait for MAX_SCHEDULE_TIMEOUT and just timeout eventually. Below is a summary of tests triggering batched requests on different kernel builds. Before this patch: ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=y Most common Linux distribution setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=n Only possible if CONFIG_DELL_RBU=n and CONFIG_LEDS_LP55XX_COMMON=n, rare. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y CONFIG_FW_LOADER_USER_HELPER=y Google Android setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ After this patch: ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=y Most common Linux distribution setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) FAIL OK request_firmware_nowait(uevent=false) FAIL OK ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=n Only possible if CONFIG_DELL_RBU=n and CONFIG_LEDS_LP55XX_COMMON=n, rare. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) FAIL OK request_firmware_nowait(uevent=false) FAIL OK ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y CONFIG_FW_LOADER_USER_HELPER=y Google Android setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() OK OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) OK OK request_firmware_nowait(uevent=false) OK OK ============================================================================ [0] https://bugzilla.kernel.org/show_bug.cgi?id=195477 CC: <stable@vger.kernel.org> [4.10+] Cc: Ming Lei <ming.lei@redhat.com> Fixes: 5b029624948d ("firmware: do not use fw_lock for fw_state protection") Reported-by: Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-07-21 03:13:09 +07:00
init_completion(&fw_st->completion);
fw_st->status = FW_STATUS_UNKNOWN;
}
static inline bool __fw_state_is_done(enum fw_status status)
{
return status == FW_STATUS_DONE || status == FW_STATUS_ABORTED;
}
static int __fw_state_wait_common(struct fw_state *fw_st, long timeout)
{
long ret;
firmware: fix batched requests - wake all waiters The firmware cache mechanism serves two purposes, the secondary purpose is not well documented nor understood. This fixes a regression with the secondary purpose of the firmware cache mechanism: batched requests on successful lookups. Without this fix *any* time a batched request is triggered, secondary requests for which the batched request mechanism was designed for will seem to last forver and seem to never return. This issue is present for all kernel builds possible, and a hard reset is required. The firmware cache is used for: 1) Addressing races with file lookups during the suspend/resume cycle by keeping firmware in memory during the suspend/resume cycle 2) Batched requests for the same file rely only on work from the first file lookup, which keeps the firmware in memory until the last release_firmware() is called Batched requests *only* take effect if secondary requests come in prior to the first user calling release_firmware(). The devres name used for the internal firmware cache is used as a hint other pending requests are ongoing, the firmware buffer data is kept in memory until the last user of the buffer calls release_firmware(), therefore serializing requests and delaying the release until all requests are done. Batched requests wait for a wakup or signal so we can rely on the first file fetch to write to the pending secondary requests. Commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") ported the firmware API to use swait, and in doing so failed to convert complete_all() to swake_up_all() -- it used swake_up(), loosing the ability for *some* batched requests to take effect. We *could* fix this by just using swake_up_all() *but* swait is now known to be very special use case, so its best to just move away from it. So we just go back to using completions as before commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") given this was using complete_all(). Without this fix it has been reported plugging in two Intel 6260 Wifi cards on a system will end up enumerating the two devices only 50% of the time [0]. The ported swake_up() should have actually handled the case with two devices, however, *if more than two cards are used* the swake_up() would not have sufficed. This change is only part of the required fixes for batched requests. Another fix is provided in the next patch. This particular change should fix the cases where more than three requests with the same firmware name is used, otherwise batched requests will wait for MAX_SCHEDULE_TIMEOUT and just timeout eventually. Below is a summary of tests triggering batched requests on different kernel builds. Before this patch: ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=y Most common Linux distribution setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=n Only possible if CONFIG_DELL_RBU=n and CONFIG_LEDS_LP55XX_COMMON=n, rare. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y CONFIG_FW_LOADER_USER_HELPER=y Google Android setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ After this patch: ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=y Most common Linux distribution setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) FAIL OK request_firmware_nowait(uevent=false) FAIL OK ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=n Only possible if CONFIG_DELL_RBU=n and CONFIG_LEDS_LP55XX_COMMON=n, rare. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) FAIL OK request_firmware_nowait(uevent=false) FAIL OK ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y CONFIG_FW_LOADER_USER_HELPER=y Google Android setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() OK OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) OK OK request_firmware_nowait(uevent=false) OK OK ============================================================================ [0] https://bugzilla.kernel.org/show_bug.cgi?id=195477 CC: <stable@vger.kernel.org> [4.10+] Cc: Ming Lei <ming.lei@redhat.com> Fixes: 5b029624948d ("firmware: do not use fw_lock for fw_state protection") Reported-by: Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-07-21 03:13:09 +07:00
ret = wait_for_completion_interruptible_timeout(&fw_st->completion,
timeout);
if (ret != 0 && fw_st->status == FW_STATUS_ABORTED)
return -ENOENT;
if (!ret)
return -ETIMEDOUT;
return ret < 0 ? ret : 0;
}
static void __fw_state_set(struct fw_state *fw_st,
enum fw_status status)
{
WRITE_ONCE(fw_st->status, status);
if (status == FW_STATUS_DONE || status == FW_STATUS_ABORTED)
firmware: fix batched requests - wake all waiters The firmware cache mechanism serves two purposes, the secondary purpose is not well documented nor understood. This fixes a regression with the secondary purpose of the firmware cache mechanism: batched requests on successful lookups. Without this fix *any* time a batched request is triggered, secondary requests for which the batched request mechanism was designed for will seem to last forver and seem to never return. This issue is present for all kernel builds possible, and a hard reset is required. The firmware cache is used for: 1) Addressing races with file lookups during the suspend/resume cycle by keeping firmware in memory during the suspend/resume cycle 2) Batched requests for the same file rely only on work from the first file lookup, which keeps the firmware in memory until the last release_firmware() is called Batched requests *only* take effect if secondary requests come in prior to the first user calling release_firmware(). The devres name used for the internal firmware cache is used as a hint other pending requests are ongoing, the firmware buffer data is kept in memory until the last user of the buffer calls release_firmware(), therefore serializing requests and delaying the release until all requests are done. Batched requests wait for a wakup or signal so we can rely on the first file fetch to write to the pending secondary requests. Commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") ported the firmware API to use swait, and in doing so failed to convert complete_all() to swake_up_all() -- it used swake_up(), loosing the ability for *some* batched requests to take effect. We *could* fix this by just using swake_up_all() *but* swait is now known to be very special use case, so its best to just move away from it. So we just go back to using completions as before commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") given this was using complete_all(). Without this fix it has been reported plugging in two Intel 6260 Wifi cards on a system will end up enumerating the two devices only 50% of the time [0]. The ported swake_up() should have actually handled the case with two devices, however, *if more than two cards are used* the swake_up() would not have sufficed. This change is only part of the required fixes for batched requests. Another fix is provided in the next patch. This particular change should fix the cases where more than three requests with the same firmware name is used, otherwise batched requests will wait for MAX_SCHEDULE_TIMEOUT and just timeout eventually. Below is a summary of tests triggering batched requests on different kernel builds. Before this patch: ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=y Most common Linux distribution setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=n Only possible if CONFIG_DELL_RBU=n and CONFIG_LEDS_LP55XX_COMMON=n, rare. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y CONFIG_FW_LOADER_USER_HELPER=y Google Android setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL FAIL request_firmware_direct() FAIL FAIL request_firmware_nowait(uevent=true) FAIL FAIL request_firmware_nowait(uevent=false) FAIL FAIL ============================================================================ After this patch: ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=y Most common Linux distribution setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) FAIL OK request_firmware_nowait(uevent=false) FAIL OK ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=n CONFIG_FW_LOADER_USER_HELPER=n Only possible if CONFIG_DELL_RBU=n and CONFIG_LEDS_LP55XX_COMMON=n, rare. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() FAIL OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) FAIL OK request_firmware_nowait(uevent=false) FAIL OK ============================================================================ CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y CONFIG_FW_LOADER_USER_HELPER=y Google Android setup. API-type no-firmware-found firmware-found ---------------------------------------------------------------------- request_firmware() OK OK request_firmware_direct() FAIL OK request_firmware_nowait(uevent=true) OK OK request_firmware_nowait(uevent=false) OK OK ============================================================================ [0] https://bugzilla.kernel.org/show_bug.cgi?id=195477 CC: <stable@vger.kernel.org> [4.10+] Cc: Ming Lei <ming.lei@redhat.com> Fixes: 5b029624948d ("firmware: do not use fw_lock for fw_state protection") Reported-by: Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-07-21 03:13:09 +07:00
complete_all(&fw_st->completion);
}
#define fw_state_start(fw_st) \
__fw_state_set(fw_st, FW_STATUS_LOADING)
#define fw_state_done(fw_st) \
__fw_state_set(fw_st, FW_STATUS_DONE)
#define fw_state_wait(fw_st) \
__fw_state_wait_common(fw_st, MAX_SCHEDULE_TIMEOUT)
#ifndef CONFIG_FW_LOADER_USER_HELPER
#define fw_state_is_aborted(fw_st) false
#else /* CONFIG_FW_LOADER_USER_HELPER */
static int __fw_state_check(struct fw_state *fw_st, enum fw_status status)
{
return fw_st->status == status;
}
#define fw_state_aborted(fw_st) \
__fw_state_set(fw_st, FW_STATUS_ABORTED)
#define fw_state_is_done(fw_st) \
__fw_state_check(fw_st, FW_STATUS_DONE)
#define fw_state_is_loading(fw_st) \
__fw_state_check(fw_st, FW_STATUS_LOADING)
#define fw_state_is_aborted(fw_st) \
__fw_state_check(fw_st, FW_STATUS_ABORTED)
#define fw_state_wait_timeout(fw_st, timeout) \
__fw_state_wait_common(fw_st, timeout)
#endif /* CONFIG_FW_LOADER_USER_HELPER */
/* firmware behavior options */
#define FW_OPT_UEVENT (1U << 0)
#define FW_OPT_NOWAIT (1U << 1)
#ifdef CONFIG_FW_LOADER_USER_HELPER
firmware loader: allow disabling of udev as firmware loader [The patch was originally proposed by Tom Gundersen, and rewritten afterwards by me; most of changelogs below borrowed from Tom's original patch -- tiwai] Currently (at least) the dell-rbu driver selects FW_LOADER_USER_HELPER, which means that distros can't really stop loading firmware through udev without breaking other users (though some have). Ideally we would remove/disable the udev firmware helper in both the kernel and in udev, but if we were to disable it in udev and not the kernel, the result would be (seemingly) hung kernels as no one would be around to cancel firmware requests. This patch allows udev firmware loading to be disabled while still allowing non-udev firmware loading, as done by the dell-rbu driver, to continue working. This is achieved by only using the fallback mechanism when the uevent is suppressed. The patch renames the user-selectable Kconfig from FW_LOADER_USER_HELPER to FW_LOADER_USER_HELPER_FALLBACK, and the former is reverse-selected by the latter or the drivers that need userhelper like dell-rbu. Also, the "default y" is removed together with this change, since it's been deprecated in udev upstream, thus rather better to disable it nowadays. Tested with FW_LOADER_USER_HELPER=n LATTICE_ECP3_CONFIG=y DELL_RBU=y and udev without the firmware loading support, but I don't have the hardware to test the lattice/dell drivers, so additional testing would be appreciated. Reviewed-by: Tom Gundersen <teg@jklm.no> Cc: Ming Lei <ming.lei@canonical.com> Cc: Abhay Salunke <Abhay_Salunke@dell.com> Cc: Stefan Roese <sr@denx.de> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Kay Sievers <kay@vrfy.org> Tested-by: Balaji Singh <B_B_Singh@DELL.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-06-04 22:48:15 +07:00
#define FW_OPT_USERHELPER (1U << 2)
#else
firmware loader: allow disabling of udev as firmware loader [The patch was originally proposed by Tom Gundersen, and rewritten afterwards by me; most of changelogs below borrowed from Tom's original patch -- tiwai] Currently (at least) the dell-rbu driver selects FW_LOADER_USER_HELPER, which means that distros can't really stop loading firmware through udev without breaking other users (though some have). Ideally we would remove/disable the udev firmware helper in both the kernel and in udev, but if we were to disable it in udev and not the kernel, the result would be (seemingly) hung kernels as no one would be around to cancel firmware requests. This patch allows udev firmware loading to be disabled while still allowing non-udev firmware loading, as done by the dell-rbu driver, to continue working. This is achieved by only using the fallback mechanism when the uevent is suppressed. The patch renames the user-selectable Kconfig from FW_LOADER_USER_HELPER to FW_LOADER_USER_HELPER_FALLBACK, and the former is reverse-selected by the latter or the drivers that need userhelper like dell-rbu. Also, the "default y" is removed together with this change, since it's been deprecated in udev upstream, thus rather better to disable it nowadays. Tested with FW_LOADER_USER_HELPER=n LATTICE_ECP3_CONFIG=y DELL_RBU=y and udev without the firmware loading support, but I don't have the hardware to test the lattice/dell drivers, so additional testing would be appreciated. Reviewed-by: Tom Gundersen <teg@jklm.no> Cc: Ming Lei <ming.lei@canonical.com> Cc: Abhay Salunke <Abhay_Salunke@dell.com> Cc: Stefan Roese <sr@denx.de> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Kay Sievers <kay@vrfy.org> Tested-by: Balaji Singh <B_B_Singh@DELL.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-06-04 22:48:15 +07:00
#define FW_OPT_USERHELPER 0
#endif
#ifdef CONFIG_FW_LOADER_USER_HELPER_FALLBACK
#define FW_OPT_FALLBACK FW_OPT_USERHELPER
#else
#define FW_OPT_FALLBACK 0
#endif
#define FW_OPT_NO_WARN (1U << 3)
#define FW_OPT_NOCACHE (1U << 4)
struct firmware_cache {
/* firmware_buf instance will be added into the below list */
spinlock_t lock;
struct list_head head;
int state;
#ifdef CONFIG_PM_SLEEP
/*
* Names of firmware images which have been cached successfully
* will be added into the below list so that device uncache
* helper can trace which firmware images have been cached
* before.
*/
spinlock_t name_lock;
struct list_head fw_names;
struct delayed_work work;
struct notifier_block pm_notify;
#endif
};
struct firmware_buf {
struct kref ref;
struct list_head list;
struct firmware_cache *fwc;
struct fw_state fw_st;
void *data;
size_t size;
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
size_t allocated_size;
#ifdef CONFIG_FW_LOADER_USER_HELPER
bool is_paged_buf;
bool need_uevent;
struct page **pages;
int nr_pages;
int page_array_size;
struct list_head pending_list;
#endif
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
const char *fw_id;
};
struct fw_cache_entry {
struct list_head list;
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
const char *name;
};
struct fw_name_devm {
unsigned long magic;
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
const char *name;
};
#define to_fwbuf(d) container_of(d, struct firmware_buf, ref)
#define FW_LOADER_NO_CACHE 0
#define FW_LOADER_START_CACHE 1
static int fw_cache_piggyback_on_request(const char *name);
/* fw_lock could be moved to 'struct firmware_priv' but since it is just
* guarding for corner cases a global lock should be OK */
static DEFINE_MUTEX(fw_lock);
static bool __enable_firmware = false;
static void enable_firmware(void)
{
mutex_lock(&fw_lock);
__enable_firmware = true;
mutex_unlock(&fw_lock);
}
static void disable_firmware(void)
{
mutex_lock(&fw_lock);
__enable_firmware = false;
mutex_unlock(&fw_lock);
}
/*
* When disabled only the built-in firmware and the firmware cache will be
* used to look for firmware.
*/
static bool firmware_enabled(void)
{
bool enabled = false;
mutex_lock(&fw_lock);
if (__enable_firmware)
enabled = true;
mutex_unlock(&fw_lock);
return enabled;
}
static struct firmware_cache fw_cache;
static struct firmware_buf *__allocate_fw_buf(const char *fw_name,
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
struct firmware_cache *fwc,
void *dbuf, size_t size)
{
struct firmware_buf *buf;
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
buf = kzalloc(sizeof(*buf), GFP_ATOMIC);
if (!buf)
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
return NULL;
buf->fw_id = kstrdup_const(fw_name, GFP_ATOMIC);
if (!buf->fw_id) {
kfree(buf);
return NULL;
}
kref_init(&buf->ref);
buf->fwc = fwc;
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
buf->data = dbuf;
buf->allocated_size = size;
fw_state_init(&buf->fw_st);
#ifdef CONFIG_FW_LOADER_USER_HELPER
INIT_LIST_HEAD(&buf->pending_list);
#endif
pr_debug("%s: fw-%s buf=%p\n", __func__, fw_name, buf);
return buf;
}
static struct firmware_buf *__fw_lookup_buf(const char *fw_name)
{
struct firmware_buf *tmp;
struct firmware_cache *fwc = &fw_cache;
list_for_each_entry(tmp, &fwc->head, list)
if (!strcmp(tmp->fw_id, fw_name))
return tmp;
return NULL;
}
static int fw_lookup_and_allocate_buf(const char *fw_name,
struct firmware_cache *fwc,
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
struct firmware_buf **buf, void *dbuf,
size_t size)
{
struct firmware_buf *tmp;
spin_lock(&fwc->lock);
tmp = __fw_lookup_buf(fw_name);
if (tmp) {
kref_get(&tmp->ref);
spin_unlock(&fwc->lock);
*buf = tmp;
return 1;
}
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
tmp = __allocate_fw_buf(fw_name, fwc, dbuf, size);
if (tmp)
list_add(&tmp->list, &fwc->head);
spin_unlock(&fwc->lock);
*buf = tmp;
return tmp ? 0 : -ENOMEM;
}
static void __fw_free_buf(struct kref *ref)
__releases(&fwc->lock)
{
struct firmware_buf *buf = to_fwbuf(ref);
struct firmware_cache *fwc = buf->fwc;
pr_debug("%s: fw-%s buf=%p data=%p size=%u\n",
__func__, buf->fw_id, buf, buf->data,
(unsigned int)buf->size);
list_del(&buf->list);
spin_unlock(&fwc->lock);
#ifdef CONFIG_FW_LOADER_USER_HELPER
if (buf->is_paged_buf) {
int i;
vunmap(buf->data);
for (i = 0; i < buf->nr_pages; i++)
__free_page(buf->pages[i]);
vfree(buf->pages);
} else
#endif
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
if (!buf->allocated_size)
vfree(buf->data);
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
kfree_const(buf->fw_id);
kfree(buf);
}
static void fw_free_buf(struct firmware_buf *buf)
{
struct firmware_cache *fwc = buf->fwc;
spin_lock(&fwc->lock);
if (!kref_put(&buf->ref, __fw_free_buf))
spin_unlock(&fwc->lock);
}
/* direct firmware loading support */
static char fw_path_para[256];
static const char * const fw_path[] = {
fw_path_para,
"/lib/firmware/updates/" UTS_RELEASE,
"/lib/firmware/updates",
"/lib/firmware/" UTS_RELEASE,
"/lib/firmware"
};
/*
* Typical usage is that passing 'firmware_class.path=$CUSTOMIZED_PATH'
* from kernel command line because firmware_class is generally built in
* kernel instead of module.
*/
module_param_string(path, fw_path_para, sizeof(fw_path_para), 0644);
MODULE_PARM_DESC(path, "customized firmware image search path with a higher priority than default path");
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
static int
fw_get_filesystem_firmware(struct device *device, struct firmware_buf *buf)
{
loff_t size;
int i, len;
int rc = -ENOENT;
char *path;
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
enum kernel_read_file_id id = READING_FIRMWARE;
size_t msize = INT_MAX;
/* Already populated data member means we're loading into a buffer */
if (buf->data) {
id = READING_FIRMWARE_PREALLOC_BUFFER;
msize = buf->allocated_size;
}
path = __getname();
if (!path)
return -ENOMEM;
for (i = 0; i < ARRAY_SIZE(fw_path); i++) {
/* skip the unset customized path */
if (!fw_path[i][0])
continue;
len = snprintf(path, PATH_MAX, "%s/%s",
fw_path[i], buf->fw_id);
if (len >= PATH_MAX) {
rc = -ENAMETOOLONG;
break;
}
buf->size = 0;
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
rc = kernel_read_file_from_path(path, &buf->data, &size, msize,
id);
if (rc) {
if (rc == -ENOENT)
dev_dbg(device, "loading %s failed with error %d\n",
path, rc);
else
dev_warn(device, "loading %s failed with error %d\n",
path, rc);
continue;
}
dev_dbg(device, "direct-loading %s\n", buf->fw_id);
buf->size = size;
fw_state_done(&buf->fw_st);
break;
}
__putname(path);
return rc;
}
/* firmware holds the ownership of pages */
static void firmware_free_data(const struct firmware *fw)
{
/* Loaded directly? */
if (!fw->priv) {
vfree(fw->data);
return;
}
fw_free_buf(fw->priv);
}
/* store the pages buffer info firmware from buf */
static void fw_set_page_data(struct firmware_buf *buf, struct firmware *fw)
{
fw->priv = buf;
#ifdef CONFIG_FW_LOADER_USER_HELPER
fw->pages = buf->pages;
#endif
fw->size = buf->size;
fw->data = buf->data;
pr_debug("%s: fw-%s buf=%p data=%p size=%u\n",
__func__, buf->fw_id, buf, buf->data,
(unsigned int)buf->size);
}
#ifdef CONFIG_PM_SLEEP
static void fw_name_devm_release(struct device *dev, void *res)
{
struct fw_name_devm *fwn = res;
if (fwn->magic == (unsigned long)&fw_cache)
pr_debug("%s: fw_name-%s devm-%p released\n",
__func__, fwn->name, res);
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
kfree_const(fwn->name);
}
static int fw_devm_match(struct device *dev, void *res,
void *match_data)
{
struct fw_name_devm *fwn = res;
return (fwn->magic == (unsigned long)&fw_cache) &&
!strcmp(fwn->name, match_data);
}
static struct fw_name_devm *fw_find_devm_name(struct device *dev,
const char *name)
{
struct fw_name_devm *fwn;
fwn = devres_find(dev, fw_name_devm_release,
fw_devm_match, (void *)name);
return fwn;
}
/* add firmware name into devres list */
static int fw_add_devm_name(struct device *dev, const char *name)
{
struct fw_name_devm *fwn;
fwn = fw_find_devm_name(dev, name);
if (fwn)
return 1;
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
fwn = devres_alloc(fw_name_devm_release, sizeof(struct fw_name_devm),
GFP_KERNEL);
if (!fwn)
return -ENOMEM;
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
fwn->name = kstrdup_const(name, GFP_KERNEL);
if (!fwn->name) {
devres_free(fwn);
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
return -ENOMEM;
}
fwn->magic = (unsigned long)&fw_cache;
devres_add(dev, fwn);
return 0;
}
#else
static int fw_add_devm_name(struct device *dev, const char *name)
{
return 0;
}
#endif
static int assign_firmware_buf(struct firmware *fw, struct device *device,
unsigned int opt_flags)
{
struct firmware_buf *buf = fw->priv;
mutex_lock(&fw_lock);
if (!buf->size || fw_state_is_aborted(&buf->fw_st)) {
mutex_unlock(&fw_lock);
return -ENOENT;
}
/*
* add firmware name into devres list so that we can auto cache
* and uncache firmware for device.
*
* device may has been deleted already, but the problem
* should be fixed in devres or driver core.
*/
/* don't cache firmware handled without uevent */
if (device && (opt_flags & FW_OPT_UEVENT) &&
!(opt_flags & FW_OPT_NOCACHE))
fw_add_devm_name(device, buf->fw_id);
/*
* After caching firmware image is started, let it piggyback
* on request firmware.
*/
if (!(opt_flags & FW_OPT_NOCACHE) &&
buf->fwc->state == FW_LOADER_START_CACHE) {
if (fw_cache_piggyback_on_request(buf->fw_id))
kref_get(&buf->ref);
}
/* pass the pages buffer to driver at the last minute */
fw_set_page_data(buf, fw);
mutex_unlock(&fw_lock);
return 0;
}
/*
* user-mode helper code
*/
#ifdef CONFIG_FW_LOADER_USER_HELPER
struct firmware_priv {
bool nowait;
struct device dev;
struct firmware_buf *buf;
struct firmware *fw;
};
static struct firmware_priv *to_firmware_priv(struct device *dev)
{
return container_of(dev, struct firmware_priv, dev);
}
static void __fw_load_abort(struct firmware_buf *buf)
{
/*
* There is a small window in which user can write to 'loading'
* between loading done and disappearance of 'loading'
*/
if (fw_state_is_done(&buf->fw_st))
return;
list_del_init(&buf->pending_list);
fw_state_aborted(&buf->fw_st);
}
static void fw_load_abort(struct firmware_priv *fw_priv)
{
struct firmware_buf *buf = fw_priv->buf;
__fw_load_abort(buf);
}
static LIST_HEAD(pending_fw_head);
static void kill_pending_fw_fallback_reqs(bool only_kill_custom)
{
struct firmware_buf *buf;
struct firmware_buf *next;
mutex_lock(&fw_lock);
list_for_each_entry_safe(buf, next, &pending_fw_head, pending_list) {
if (!buf->need_uevent || !only_kill_custom)
__fw_load_abort(buf);
}
mutex_unlock(&fw_lock);
}
static ssize_t timeout_show(struct class *class, struct class_attribute *attr,
char *buf)
{
return sprintf(buf, "%d\n", loading_timeout);
}
/**
* firmware_timeout_store - set number of seconds to wait for firmware
* @class: device class pointer
* @attr: device attribute pointer
* @buf: buffer to scan for timeout value
* @count: number of bytes in @buf
*
* Sets the number of seconds to wait for the firmware. Once
* this expires an error will be returned to the driver and no
* firmware will be provided.
*
* Note: zero means 'wait forever'.
**/
static ssize_t timeout_store(struct class *class, struct class_attribute *attr,
const char *buf, size_t count)
{
loading_timeout = simple_strtol(buf, NULL, 10);
if (loading_timeout < 0)
loading_timeout = 0;
return count;
}
static CLASS_ATTR_RW(timeout);
static struct attribute *firmware_class_attrs[] = {
&class_attr_timeout.attr,
NULL,
};
ATTRIBUTE_GROUPS(firmware_class);
static void fw_dev_release(struct device *dev)
{
struct firmware_priv *fw_priv = to_firmware_priv(dev);
kfree(fw_priv);
}
static int do_firmware_uevent(struct firmware_priv *fw_priv, struct kobj_uevent_env *env)
{
if (add_uevent_var(env, "FIRMWARE=%s", fw_priv->buf->fw_id))
return -ENOMEM;
if (add_uevent_var(env, "TIMEOUT=%i", loading_timeout))
return -ENOMEM;
if (add_uevent_var(env, "ASYNC=%d", fw_priv->nowait))
return -ENOMEM;
return 0;
}
static int firmware_uevent(struct device *dev, struct kobj_uevent_env *env)
{
struct firmware_priv *fw_priv = to_firmware_priv(dev);
int err = 0;
mutex_lock(&fw_lock);
if (fw_priv->buf)
err = do_firmware_uevent(fw_priv, env);
mutex_unlock(&fw_lock);
return err;
}
static struct class firmware_class = {
.name = "firmware",
.class_groups = firmware_class_groups,
.dev_uevent = firmware_uevent,
.dev_release = fw_dev_release,
};
static ssize_t firmware_loading_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct firmware_priv *fw_priv = to_firmware_priv(dev);
int loading = 0;
mutex_lock(&fw_lock);
if (fw_priv->buf)
loading = fw_state_is_loading(&fw_priv->buf->fw_st);
mutex_unlock(&fw_lock);
return sprintf(buf, "%d\n", loading);
}
/* Some architectures don't have PAGE_KERNEL_RO */
#ifndef PAGE_KERNEL_RO
#define PAGE_KERNEL_RO PAGE_KERNEL
#endif
/* one pages buffer should be mapped/unmapped only once */
static int fw_map_pages_buf(struct firmware_buf *buf)
{
if (!buf->is_paged_buf)
return 0;
vunmap(buf->data);
buf->data = vmap(buf->pages, buf->nr_pages, 0, PAGE_KERNEL_RO);
if (!buf->data)
return -ENOMEM;
return 0;
}
/**
* firmware_loading_store - set value in the 'loading' control file
* @dev: device pointer
* @attr: device attribute pointer
* @buf: buffer to scan for loading control value
* @count: number of bytes in @buf
*
* The relevant values are:
*
* 1: Start a load, discarding any previous partial load.
* 0: Conclude the load and hand the data to the driver code.
* -1: Conclude the load with an error and discard any written data.
**/
static ssize_t firmware_loading_store(struct device *dev,
struct device_attribute *attr,
const char *buf, size_t count)
{
struct firmware_priv *fw_priv = to_firmware_priv(dev);
struct firmware_buf *fw_buf;
ssize_t written = count;
int loading = simple_strtol(buf, NULL, 10);
int i;
mutex_lock(&fw_lock);
fw_buf = fw_priv->buf;
firmware: fix NULL pointer dereference in __fw_load_abort() Since commit 5d47ec02c37ea6 ("firmware: Correct handling of fw_state_wait() return value") fw_load_abort() could be called twice and lead us to a kernel crash. This happens only when the firmware fallback mechanism (regular or custom) is used. The fallback mechanism exposes a sysfs interface for userspace to upload a file and notify the kernel when the file is loaded and ready, or to cancel an upload by echo'ing -1 into on the loading file: echo -n "-1" > /sys/$DEVPATH/loading This will call fw_load_abort(). Some distributions actually have a udev rule in place to *always* immediately cancel all firmware fallback mechanism requests (Debian), they have: $ cat /lib/udev/rules.d/50-firmware.rules # stub for immediately telling the kernel that userspace firmware loading # failed; necessary to avoid long timeouts with CONFIG_FW_LOADER_USER_HELPER=y SUBSYSTEM=="firmware", ACTION=="add", ATTR{loading}="-1 Distributions with this udev rule would run into this crash only if the fallback mechanism is used. Since most distributions disable by default using the fallback mechanism (CONFIG_FW_LOADER_USER_HELPER_FALLBACK), this would typicaly mean only 2 drivers which *require* the fallback mechanism could typically incur a crash: drivers/firmware/dell_rbu.c and the drivers/leds/leds-lp55xx-common.c driver. Distributions enabling CONFIG_FW_LOADER_USER_HELPER_FALLBACK by default are obviously more exposed to this crash. The crash happens because after commit 5b029624948d ("firmware: do not use fw_lock for fw_state protection") and subsequent fix commit 5d47ec02c37ea6 ("firmware: Correct handling of fw_state_wait() return value") a race can happen between this cancelation and the firmware fw_state_wait_timeout() being woken up after a state change with which fw_load_abort() as that calls swake_up(). Upon error fw_state_wait_timeout() will also again call fw_load_abort() and trigger a null reference. At first glance we could just fix this with a !buf check on fw_load_abort() before accessing buf->fw_st, however there is a logical issue in having a state machine used for the fallback mechanism and preventing access from it once we abort as its inside the buf (buf->fw_st). The firmware_class.c code is setting the buf to NULL to annotate an abort has occurred. Replace this mechanism by simply using the state check instead. All the other code in place already uses similar checks for aborting as well so no further changes are needed. An oops can be reproduced with the new fw_fallback.sh fallback mechanism cancellation test. Either cancelling the fallback mechanism or the custom fallback mechanism triggers a crash. mcgrof@piggy ~/linux-next/tools/testing/selftests/firmware (git::20170111-fw-fixes)$ sudo ./fw_fallback.sh ./fw_fallback.sh: timeout works ./fw_fallback.sh: firmware comparison works ./fw_fallback.sh: fallback mechanism works [ this then sits here when it is trying the cancellation test ] Kernel log: test_firmware: loading 'nope-test-firmware.bin' misc test_firmware: Direct firmware load for nope-test-firmware.bin failed with error -2 misc test_firmware: Falling back to user helper BUG: unable to handle kernel NULL pointer dereference at 0000000000000038 IP: _request_firmware+0xa27/0xad0 PGD 0 Oops: 0000 [#1] SMP Modules linked in: test_firmware(E) ... etc ... CPU: 1 PID: 1396 Comm: fw_fallback.sh Tainted: G W E 4.10.0-rc3-next-20170111+ #30 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.10.1-0-g8891697-prebuilt.qemu-project.org 04/01/2014 task: ffff9740b27f4340 task.stack: ffffbb15c0bc8000 RIP: 0010:_request_firmware+0xa27/0xad0 RSP: 0018:ffffbb15c0bcbd10 EFLAGS: 00010246 RAX: 00000000fffffffe RBX: ffff9740afe5aa80 RCX: 0000000000000000 RDX: ffff9740b27f4340 RSI: 0000000000000283 RDI: 0000000000000000 RBP: ffffbb15c0bcbd90 R08: ffffbb15c0bcbcd8 R09: 0000000000000000 R10: 0000000894a0d4b1 R11: 000000000000008c R12: ffffffffc0312480 R13: 0000000000000005 R14: ffff9740b1c32400 R15: 00000000000003e8 FS: 00007f8604422700(0000) GS:ffff9740bfc80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000038 CR3: 000000012164c000 CR4: 00000000000006e0 Call Trace: request_firmware+0x37/0x50 trigger_request_store+0x79/0xd0 [test_firmware] dev_attr_store+0x18/0x30 sysfs_kf_write+0x37/0x40 kernfs_fop_write+0x110/0x1a0 __vfs_write+0x37/0x160 ? _cond_resched+0x1a/0x50 vfs_write+0xb5/0x1a0 SyS_write+0x55/0xc0 ? trace_do_page_fault+0x37/0xd0 entry_SYSCALL_64_fastpath+0x1e/0xad RIP: 0033:0x7f8603f49620 RSP: 002b:00007fff6287b788 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 000055c307b110a0 RCX: 00007f8603f49620 RDX: 0000000000000016 RSI: 000055c3084d8a90 RDI: 0000000000000001 RBP: 0000000000000016 R08: 000000000000c0ff R09: 000055c3084d6336 R10: 000055c307b108b0 R11: 0000000000000246 R12: 000055c307b13c80 R13: 000055c3084d6320 R14: 0000000000000000 R15: 00007fff6287b950 Code: 9f 64 84 e8 9c 61 fe ff b8 f4 ff ff ff e9 6b f9 ff ff 48 c7 c7 40 6b 8d 84 89 45 a8 e8 43 84 18 00 49 8b be 00 03 00 00 8b 45 a8 <83> 7f 38 02 74 08 e8 6e ec ff ff 8b 45 a8 49 c7 86 00 03 00 00 RIP: _request_firmware+0xa27/0xad0 RSP: ffffbb15c0bcbd10 CR2: 0000000000000038 ---[ end trace 6d94ac339c133e6f ]--- Fixes: 5d47ec02c37e ("firmware: Correct handling of fw_state_wait() return value") Reported-and-Tested-by: Jakub Kicinski <jakub.kicinski@netronome.com> Reported-and-Tested-by: Patrick Bruenn <p.bruenn@beckhoff.com> Reported-by: Chris Wilson <chris@chris-wilson.co.uk> CC: <stable@vger.kernel.org> [3.10+] Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-01-26 01:31:52 +07:00
if (fw_state_is_aborted(&fw_buf->fw_st))
goto out;
switch (loading) {
case 1:
/* discarding any previous partial load */
if (!fw_state_is_done(&fw_buf->fw_st)) {
for (i = 0; i < fw_buf->nr_pages; i++)
__free_page(fw_buf->pages[i]);
vfree(fw_buf->pages);
fw_buf->pages = NULL;
fw_buf->page_array_size = 0;
fw_buf->nr_pages = 0;
fw_state_start(&fw_buf->fw_st);
}
break;
case 0:
if (fw_state_is_loading(&fw_buf->fw_st)) {
int rc;
/*
* Several loading requests may be pending on
* one same firmware buf, so let all requests
* see the mapped 'buf->data' once the loading
* is completed.
* */
rc = fw_map_pages_buf(fw_buf);
if (rc)
firmware: give a protection when map page failed so, we need give a protection and return a error value. [ 7341.474236] [drm:do_intel_finish_page_flip] *ERROR* invalid or inactive unpin_work! [ 7341.494464] atomisp-css2400b0_v21 0000:00:03.0: unhandled css stored event: 0x20 [ 7341.503627] vmap allocation for size 208896 failed: use vmalloc=<size> to increase size.<=================== map failed [ 7341.507135] [drm:do_intel_finish_page_flip] *ERROR* invalid or inactive unpin_work! [ 7341.503848] BUG: unable to handle kernel NULL pointer dereference at (null) [ 7341.520394] IP: [<c18f5c1b>] sst_load_all_modules_elf+0x1bb/0x850 [ 7341.527216] *pdpt = 0000000030dfe001 *pde = 0000000000000000 [ 7341.533640] Oops: 0000 [#1] PREEMPT SMP [ 7341.540360] [drm:do_intel_finish_page_flip] *ERROR* invalid or inactive unpin_work! [ 7341.538037] Modules linked in: atomisp_css2400b0_v21 lm3554 ov2722 imx1x5 atmel_mxt_ts vxd392 videobuf_vmalloc videobuf_core lm_dump(O) bcm_bt_lpm hdmi_audio bcm4334x(O) [ 7341.563531] CPU: 1 PID: 525 Comm: mediaserver Tainted: G W O 3.10.20-262518-ga83c053 #1 [ 7341.573253] task: f0994ec0 ti: f09f0000 task.ti: f09f0000 [ 7341.579284] EIP: 0060:[<c18f5c1b>] EFLAGS: 00010246 CPU: 1 [ 7341.585415] EIP is at sst_load_all_modules_elf+0x1bb/0x850 [ 7341.591541] EAX: 00000000 EBX: e3595ba0 ECX: 00000000 EDX: 00031c1c [ 7341.598541] ESI: e04a0000 EDI: 00000000 EBP: f09f1d80 ESP: f09f1cf4 [ 7341.605542] DS: 007b ES: 007b FS: 00d8 GS: 003b SS: 0068 [ 7341.611573] CR0: 80050033 CR2: 00000000 CR3: 30db4000 CR4: 001007f0 [ 7341.618573] DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000 [ 7341.625575] DR6: ffff0ff0 DR7: 00000400 [ 7341.629856] Stack: [ 7341.632097] f09f1d57 00000019 c1d656d7 c1d658d3 c1d56409 00000f28 c1d64af9 18000103 [ 7341.640766] 01000001 00080000 c1f910a0 f326f4c8 00000034 f326f520 00000002 e04a02bc [ 7341.649465] 00000001 f326e014 c1f910b0 e04a0000 c0080100 00031c1c e3595ba0 c0080100 [ 7341.658149] Call Trace: [ 7341.660888] [<c18f6308>] sst_post_download_byt+0x58/0xb0 [ 7341.666925] [<c18f4fbc>] sst_load_fw+0xdc/0x510 [ 7341.672086] [<c1a7b2c0>] ? __mutex_lock_slowpath+0x250/0x370 [ 7341.678507] [<c1a80e05>] ? sub_preempt_count+0x55/0xe0 [ 7341.684346] [<c18f1294>] sst_download_fw+0x14/0x60 [ 7341.689796] [<c1a7b403>] ? mutex_lock+0x23/0x30 [ 7341.694954] [<c18f191c>] intel_sst_check_device+0x6c/0x120 [ 7341.701181] [<c18f1d08>] sst_set_generic_params+0x1b8/0x4a0 [ 7341.707504] [<c1a80e05>] ? sub_preempt_count+0x55/0xe0 [ 7341.713341] [<c1a80e05>] ? sub_preempt_count+0x55/0xe0 [ 7341.719178] [<c1a7b2c0>] ? __mutex_lock_slowpath+0x250/0x370 [ 7341.725600] [<c1320d84>] ? __kmalloc_track_caller+0xe4/0x1d0 [ 7341.732022] [<c18e35f5>] sst_set_mixer_param+0x25/0x40 [ 7341.737859] [<c18e3853>] lpe_mixer_ihf_set+0xb3/0x160 [ 7341.743602] [<c1855b99>] snd_ctl_ioctl+0xa89/0xb40 [ 7341.749052] [<c1331e65>] ? path_openat+0xa5/0x3d0 [ 7341.754409] [<c1447857>] ? avc_has_perm_flags+0xc7/0x170 [ 7341.760441] [<c1855110>] ? snd_ctl_elem_add_user+0x540/0x540 [ 7341.766862] [<c1334047>] do_vfs_ioctl+0x77/0x5e0 [ 7341.772117] [<c144842a>] ? inode_has_perm.isra.42.constprop.79+0x3a/0x50 [ 7341.779705] [<c14490a0>] ? file_has_perm+0xa0/0xb0 [ 7341.785155] [<c14493b8>] ? selinux_file_ioctl+0x48/0xe0 [ 7341.791090] [<c1334628>] SyS_ioctl+0x78/0x90 [ 7341.795958] [<c1a7dde8>] syscall_call+0x7/0xb [ 7341.800925] [<c1a70000>] ? mm_fault_error+0x13c/0x198 Signed-off-by: zhang jun <jun.zhang@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-13 14:18:47 +07:00
dev_err(dev, "%s: map pages failed\n",
__func__);
else
rc = security_kernel_post_read_file(NULL,
fw_buf->data, fw_buf->size,
READING_FIRMWARE);
/*
* Same logic as fw_load_abort, only the DONE bit
* is ignored and we set ABORT only on failure.
*/
list_del_init(&fw_buf->pending_list);
if (rc) {
fw_state_aborted(&fw_buf->fw_st);
written = rc;
} else {
fw_state_done(&fw_buf->fw_st);
}
break;
}
/* fallthrough */
default:
dev_err(dev, "%s: unexpected value (%d)\n", __func__, loading);
/* fallthrough */
case -1:
fw_load_abort(fw_priv);
break;
}
out:
mutex_unlock(&fw_lock);
return written;
}
static DEVICE_ATTR(loading, 0644, firmware_loading_show, firmware_loading_store);
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
static void firmware_rw_buf(struct firmware_buf *buf, char *buffer,
loff_t offset, size_t count, bool read)
{
if (read)
memcpy(buffer, buf->data + offset, count);
else
memcpy(buf->data + offset, buffer, count);
}
firmware: consolidate kmap/read/write logic Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This design creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. This patch sets adds support to the request firmware API to load the firmware directly into a pre-allocated buffer, skipping the intermediate copying step and alleviating memory pressure during firmware loading. The drawback is that we can't use the firmware caching feature because the memory for the firmware cache is not managed by the firmware layer. This patch (of 3): We use similar structured code to read and write the kmapped firmware pages. The only difference is read copies from the kmap region and write copies to it. Consolidate this into one function to reduce duplication. Link: http://lkml.kernel.org/r/20160607164741.31849-2-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:22 +07:00
static void firmware_rw(struct firmware_buf *buf, char *buffer,
loff_t offset, size_t count, bool read)
{
while (count) {
void *page_data;
int page_nr = offset >> PAGE_SHIFT;
int page_ofs = offset & (PAGE_SIZE-1);
int page_cnt = min_t(size_t, PAGE_SIZE - page_ofs, count);
page_data = kmap(buf->pages[page_nr]);
if (read)
memcpy(buffer, page_data + page_ofs, page_cnt);
else
memcpy(page_data + page_ofs, buffer, page_cnt);
kunmap(buf->pages[page_nr]);
buffer += page_cnt;
offset += page_cnt;
count -= page_cnt;
}
}
static ssize_t firmware_data_read(struct file *filp, struct kobject *kobj,
struct bin_attribute *bin_attr,
char *buffer, loff_t offset, size_t count)
{
struct device *dev = kobj_to_dev(kobj);
struct firmware_priv *fw_priv = to_firmware_priv(dev);
struct firmware_buf *buf;
ssize_t ret_count;
mutex_lock(&fw_lock);
buf = fw_priv->buf;
if (!buf || fw_state_is_done(&buf->fw_st)) {
ret_count = -ENODEV;
goto out;
}
if (offset > buf->size) {
ret_count = 0;
goto out;
}
if (count > buf->size - offset)
count = buf->size - offset;
ret_count = count;
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
if (buf->data)
firmware_rw_buf(buf, buffer, offset, count, true);
else
firmware_rw(buf, buffer, offset, count, true);
out:
mutex_unlock(&fw_lock);
return ret_count;
}
static int fw_realloc_buffer(struct firmware_priv *fw_priv, int min_size)
{
struct firmware_buf *buf = fw_priv->buf;
int pages_needed = PAGE_ALIGN(min_size) >> PAGE_SHIFT;
/* If the array of pages is too small, grow it... */
if (buf->page_array_size < pages_needed) {
int new_array_size = max(pages_needed,
buf->page_array_size * 2);
struct page **new_pages;
new_pages = vmalloc(new_array_size * sizeof(void *));
if (!new_pages) {
fw_load_abort(fw_priv);
return -ENOMEM;
}
memcpy(new_pages, buf->pages,
buf->page_array_size * sizeof(void *));
memset(&new_pages[buf->page_array_size], 0, sizeof(void *) *
(new_array_size - buf->page_array_size));
vfree(buf->pages);
buf->pages = new_pages;
buf->page_array_size = new_array_size;
}
while (buf->nr_pages < pages_needed) {
buf->pages[buf->nr_pages] =
alloc_page(GFP_KERNEL | __GFP_HIGHMEM);
if (!buf->pages[buf->nr_pages]) {
fw_load_abort(fw_priv);
return -ENOMEM;
}
buf->nr_pages++;
}
return 0;
}
/**
* firmware_data_write - write method for firmware
* @filp: open sysfs file
* @kobj: kobject for the device
* @bin_attr: bin_attr structure
* @buffer: buffer being written
* @offset: buffer offset for write in total data store area
* @count: buffer size
*
* Data written to the 'data' attribute will be later handed to
* the driver as a firmware image.
**/
static ssize_t firmware_data_write(struct file *filp, struct kobject *kobj,
struct bin_attribute *bin_attr,
char *buffer, loff_t offset, size_t count)
{
struct device *dev = kobj_to_dev(kobj);
struct firmware_priv *fw_priv = to_firmware_priv(dev);
struct firmware_buf *buf;
ssize_t retval;
if (!capable(CAP_SYS_RAWIO))
return -EPERM;
mutex_lock(&fw_lock);
buf = fw_priv->buf;
if (!buf || fw_state_is_done(&buf->fw_st)) {
retval = -ENODEV;
goto out;
}
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
if (buf->data) {
if (offset + count > buf->allocated_size) {
retval = -ENOMEM;
goto out;
}
firmware_rw_buf(buf, buffer, offset, count, false);
retval = count;
} else {
retval = fw_realloc_buffer(fw_priv, offset + count);
if (retval)
goto out;
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
retval = count;
firmware_rw(buf, buffer, offset, count, false);
}
firmware: consolidate kmap/read/write logic Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This design creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. This patch sets adds support to the request firmware API to load the firmware directly into a pre-allocated buffer, skipping the intermediate copying step and alleviating memory pressure during firmware loading. The drawback is that we can't use the firmware caching feature because the memory for the firmware cache is not managed by the firmware layer. This patch (of 3): We use similar structured code to read and write the kmapped firmware pages. The only difference is read copies from the kmap region and write copies to it. Consolidate this into one function to reduce duplication. Link: http://lkml.kernel.org/r/20160607164741.31849-2-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:22 +07:00
buf->size = max_t(size_t, offset + count, buf->size);
out:
mutex_unlock(&fw_lock);
return retval;
}
static struct bin_attribute firmware_attr_data = {
.attr = { .name = "data", .mode = 0644 },
.size = 0,
.read = firmware_data_read,
.write = firmware_data_write,
};
static struct attribute *fw_dev_attrs[] = {
&dev_attr_loading.attr,
NULL
};
static struct bin_attribute *fw_dev_bin_attrs[] = {
&firmware_attr_data,
NULL
};
static const struct attribute_group fw_dev_attr_group = {
.attrs = fw_dev_attrs,
.bin_attrs = fw_dev_bin_attrs,
};
static const struct attribute_group *fw_dev_attr_groups[] = {
&fw_dev_attr_group,
NULL
};
static struct firmware_priv *
fw_create_instance(struct firmware *firmware, const char *fw_name,
struct device *device, unsigned int opt_flags)
{
struct firmware_priv *fw_priv;
struct device *f_dev;
fw_priv = kzalloc(sizeof(*fw_priv), GFP_KERNEL);
if (!fw_priv) {
fw_priv = ERR_PTR(-ENOMEM);
goto exit;
}
fw_priv->nowait = !!(opt_flags & FW_OPT_NOWAIT);
fw_priv->fw = firmware;
f_dev = &fw_priv->dev;
device_initialize(f_dev);
dev_set_name(f_dev, "%s", fw_name);
f_dev->parent = device;
f_dev->class = &firmware_class;
f_dev->groups = fw_dev_attr_groups;
exit:
return fw_priv;
}
/* load a firmware via user helper */
static int _request_firmware_load(struct firmware_priv *fw_priv,
unsigned int opt_flags, long timeout)
{
int retval = 0;
struct device *f_dev = &fw_priv->dev;
struct firmware_buf *buf = fw_priv->buf;
/* fall back on userspace loading */
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
if (!buf->data)
buf->is_paged_buf = true;
dev_set_uevent_suppress(f_dev, true);
retval = device_add(f_dev);
if (retval) {
dev_err(f_dev, "%s: device_register failed\n", __func__);
goto err_put_dev;
}
mutex_lock(&fw_lock);
list_add(&buf->pending_list, &pending_fw_head);
mutex_unlock(&fw_lock);
if (opt_flags & FW_OPT_UEVENT) {
buf->need_uevent = true;
dev_set_uevent_suppress(f_dev, false);
dev_dbg(f_dev, "firmware: requesting %s\n", buf->fw_id);
kobject_uevent(&fw_priv->dev.kobj, KOBJ_ADD);
} else {
timeout = MAX_JIFFY_OFFSET;
}
retval = fw_state_wait_timeout(&buf->fw_st, timeout);
if (retval < 0) {
mutex_lock(&fw_lock);
fw_load_abort(fw_priv);
mutex_unlock(&fw_lock);
}
if (fw_state_is_aborted(&buf->fw_st))
retval = -EAGAIN;
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
else if (buf->is_paged_buf && !buf->data)
firmware: give a protection when map page failed so, we need give a protection and return a error value. [ 7341.474236] [drm:do_intel_finish_page_flip] *ERROR* invalid or inactive unpin_work! [ 7341.494464] atomisp-css2400b0_v21 0000:00:03.0: unhandled css stored event: 0x20 [ 7341.503627] vmap allocation for size 208896 failed: use vmalloc=<size> to increase size.<=================== map failed [ 7341.507135] [drm:do_intel_finish_page_flip] *ERROR* invalid or inactive unpin_work! [ 7341.503848] BUG: unable to handle kernel NULL pointer dereference at (null) [ 7341.520394] IP: [<c18f5c1b>] sst_load_all_modules_elf+0x1bb/0x850 [ 7341.527216] *pdpt = 0000000030dfe001 *pde = 0000000000000000 [ 7341.533640] Oops: 0000 [#1] PREEMPT SMP [ 7341.540360] [drm:do_intel_finish_page_flip] *ERROR* invalid or inactive unpin_work! [ 7341.538037] Modules linked in: atomisp_css2400b0_v21 lm3554 ov2722 imx1x5 atmel_mxt_ts vxd392 videobuf_vmalloc videobuf_core lm_dump(O) bcm_bt_lpm hdmi_audio bcm4334x(O) [ 7341.563531] CPU: 1 PID: 525 Comm: mediaserver Tainted: G W O 3.10.20-262518-ga83c053 #1 [ 7341.573253] task: f0994ec0 ti: f09f0000 task.ti: f09f0000 [ 7341.579284] EIP: 0060:[<c18f5c1b>] EFLAGS: 00010246 CPU: 1 [ 7341.585415] EIP is at sst_load_all_modules_elf+0x1bb/0x850 [ 7341.591541] EAX: 00000000 EBX: e3595ba0 ECX: 00000000 EDX: 00031c1c [ 7341.598541] ESI: e04a0000 EDI: 00000000 EBP: f09f1d80 ESP: f09f1cf4 [ 7341.605542] DS: 007b ES: 007b FS: 00d8 GS: 003b SS: 0068 [ 7341.611573] CR0: 80050033 CR2: 00000000 CR3: 30db4000 CR4: 001007f0 [ 7341.618573] DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000 [ 7341.625575] DR6: ffff0ff0 DR7: 00000400 [ 7341.629856] Stack: [ 7341.632097] f09f1d57 00000019 c1d656d7 c1d658d3 c1d56409 00000f28 c1d64af9 18000103 [ 7341.640766] 01000001 00080000 c1f910a0 f326f4c8 00000034 f326f520 00000002 e04a02bc [ 7341.649465] 00000001 f326e014 c1f910b0 e04a0000 c0080100 00031c1c e3595ba0 c0080100 [ 7341.658149] Call Trace: [ 7341.660888] [<c18f6308>] sst_post_download_byt+0x58/0xb0 [ 7341.666925] [<c18f4fbc>] sst_load_fw+0xdc/0x510 [ 7341.672086] [<c1a7b2c0>] ? __mutex_lock_slowpath+0x250/0x370 [ 7341.678507] [<c1a80e05>] ? sub_preempt_count+0x55/0xe0 [ 7341.684346] [<c18f1294>] sst_download_fw+0x14/0x60 [ 7341.689796] [<c1a7b403>] ? mutex_lock+0x23/0x30 [ 7341.694954] [<c18f191c>] intel_sst_check_device+0x6c/0x120 [ 7341.701181] [<c18f1d08>] sst_set_generic_params+0x1b8/0x4a0 [ 7341.707504] [<c1a80e05>] ? sub_preempt_count+0x55/0xe0 [ 7341.713341] [<c1a80e05>] ? sub_preempt_count+0x55/0xe0 [ 7341.719178] [<c1a7b2c0>] ? __mutex_lock_slowpath+0x250/0x370 [ 7341.725600] [<c1320d84>] ? __kmalloc_track_caller+0xe4/0x1d0 [ 7341.732022] [<c18e35f5>] sst_set_mixer_param+0x25/0x40 [ 7341.737859] [<c18e3853>] lpe_mixer_ihf_set+0xb3/0x160 [ 7341.743602] [<c1855b99>] snd_ctl_ioctl+0xa89/0xb40 [ 7341.749052] [<c1331e65>] ? path_openat+0xa5/0x3d0 [ 7341.754409] [<c1447857>] ? avc_has_perm_flags+0xc7/0x170 [ 7341.760441] [<c1855110>] ? snd_ctl_elem_add_user+0x540/0x540 [ 7341.766862] [<c1334047>] do_vfs_ioctl+0x77/0x5e0 [ 7341.772117] [<c144842a>] ? inode_has_perm.isra.42.constprop.79+0x3a/0x50 [ 7341.779705] [<c14490a0>] ? file_has_perm+0xa0/0xb0 [ 7341.785155] [<c14493b8>] ? selinux_file_ioctl+0x48/0xe0 [ 7341.791090] [<c1334628>] SyS_ioctl+0x78/0x90 [ 7341.795958] [<c1a7dde8>] syscall_call+0x7/0xb [ 7341.800925] [<c1a70000>] ? mm_fault_error+0x13c/0x198 Signed-off-by: zhang jun <jun.zhang@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-13 14:18:47 +07:00
retval = -ENOMEM;
device_del(f_dev);
err_put_dev:
put_device(f_dev);
return retval;
}
static int fw_load_from_user_helper(struct firmware *firmware,
const char *name, struct device *device,
firmware: move umh try locks into the umh code This moves the usermode helper locks into only code paths that use the usermode helper API from the kernel. The usermode helper locks were originally added to prevent stalling suspend, later the firmware cache was added to help with this, and further later direct filesystem lookup was added by Linus to completely bypass udev due to the amount of issues the umh approach had. The usermode helper locks were kept even when the direct filesystem lookup mechanism is used though. A lot has changed since the original usermode helper locks were added but the recent commit which added the code for firmware_enabled() are intended to address any possible races cured only as collateral by using the locks as though side consequence of code evolution and this not being addressed any time sooner. With the firmware_enabled() code in place we are a bit more sure to move the usermode helper locks to UMH only code. There is a bit of history here so let's recap a bit of it to ensure nothing is lost and things are clear. The direct filesystem approach to loading firmware is rather new, it was added via commit abb139e75c2cdb ("firmware: teach the kernel to load firmware files directly from the filesystem") by Linus merged on the v3.7 release, to enable to bypass udev. usermodehelper_read_lock_wait() was added earlier via commit 9b78c1da60b3c ("firmware_class: Do not warn that system is not ready from async loads") merged on v3.4, after Rafael noted that the async firmware API call request_firmware_nowait() should not be penalized to fail if userspace is not available yet or frozen, it'd allow for a timeout grace period before giving up. The WARN_ON() was kept for the sync firmware API call though on request_firmware(). At this time there was no direct filesystem lookup for firmware though. The original usermode helper lock came from commit a144c6a6c924a ("PM: Print a warning if firmware is requested when tasks are frozen") merged on the v3.0 kernel by Rafael to print a warning back when firmware requests were used on resume(), thaw() or restore() callbacks and there was no direct fs lookups or the firmware cache. Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-02 15:31:07 +07:00
unsigned int opt_flags)
{
struct firmware_priv *fw_priv;
firmware: move umh try locks into the umh code This moves the usermode helper locks into only code paths that use the usermode helper API from the kernel. The usermode helper locks were originally added to prevent stalling suspend, later the firmware cache was added to help with this, and further later direct filesystem lookup was added by Linus to completely bypass udev due to the amount of issues the umh approach had. The usermode helper locks were kept even when the direct filesystem lookup mechanism is used though. A lot has changed since the original usermode helper locks were added but the recent commit which added the code for firmware_enabled() are intended to address any possible races cured only as collateral by using the locks as though side consequence of code evolution and this not being addressed any time sooner. With the firmware_enabled() code in place we are a bit more sure to move the usermode helper locks to UMH only code. There is a bit of history here so let's recap a bit of it to ensure nothing is lost and things are clear. The direct filesystem approach to loading firmware is rather new, it was added via commit abb139e75c2cdb ("firmware: teach the kernel to load firmware files directly from the filesystem") by Linus merged on the v3.7 release, to enable to bypass udev. usermodehelper_read_lock_wait() was added earlier via commit 9b78c1da60b3c ("firmware_class: Do not warn that system is not ready from async loads") merged on v3.4, after Rafael noted that the async firmware API call request_firmware_nowait() should not be penalized to fail if userspace is not available yet or frozen, it'd allow for a timeout grace period before giving up. The WARN_ON() was kept for the sync firmware API call though on request_firmware(). At this time there was no direct filesystem lookup for firmware though. The original usermode helper lock came from commit a144c6a6c924a ("PM: Print a warning if firmware is requested when tasks are frozen") merged on the v3.0 kernel by Rafael to print a warning back when firmware requests were used on resume(), thaw() or restore() callbacks and there was no direct fs lookups or the firmware cache. Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-02 15:31:07 +07:00
long timeout;
int ret;
timeout = firmware_loading_timeout();
if (opt_flags & FW_OPT_NOWAIT) {
timeout = usermodehelper_read_lock_wait(timeout);
if (!timeout) {
dev_dbg(device, "firmware: %s loading timed out\n",
name);
return -EBUSY;
}
} else {
ret = usermodehelper_read_trylock();
if (WARN_ON(ret)) {
dev_err(device, "firmware: %s will not be loaded\n",
name);
return ret;
}
}
fw_priv = fw_create_instance(firmware, name, device, opt_flags);
firmware: move umh try locks into the umh code This moves the usermode helper locks into only code paths that use the usermode helper API from the kernel. The usermode helper locks were originally added to prevent stalling suspend, later the firmware cache was added to help with this, and further later direct filesystem lookup was added by Linus to completely bypass udev due to the amount of issues the umh approach had. The usermode helper locks were kept even when the direct filesystem lookup mechanism is used though. A lot has changed since the original usermode helper locks were added but the recent commit which added the code for firmware_enabled() are intended to address any possible races cured only as collateral by using the locks as though side consequence of code evolution and this not being addressed any time sooner. With the firmware_enabled() code in place we are a bit more sure to move the usermode helper locks to UMH only code. There is a bit of history here so let's recap a bit of it to ensure nothing is lost and things are clear. The direct filesystem approach to loading firmware is rather new, it was added via commit abb139e75c2cdb ("firmware: teach the kernel to load firmware files directly from the filesystem") by Linus merged on the v3.7 release, to enable to bypass udev. usermodehelper_read_lock_wait() was added earlier via commit 9b78c1da60b3c ("firmware_class: Do not warn that system is not ready from async loads") merged on v3.4, after Rafael noted that the async firmware API call request_firmware_nowait() should not be penalized to fail if userspace is not available yet or frozen, it'd allow for a timeout grace period before giving up. The WARN_ON() was kept for the sync firmware API call though on request_firmware(). At this time there was no direct filesystem lookup for firmware though. The original usermode helper lock came from commit a144c6a6c924a ("PM: Print a warning if firmware is requested when tasks are frozen") merged on the v3.0 kernel by Rafael to print a warning back when firmware requests were used on resume(), thaw() or restore() callbacks and there was no direct fs lookups or the firmware cache. Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-02 15:31:07 +07:00
if (IS_ERR(fw_priv)) {
ret = PTR_ERR(fw_priv);
goto out_unlock;
}
fw_priv->buf = firmware->priv;
firmware: move umh try locks into the umh code This moves the usermode helper locks into only code paths that use the usermode helper API from the kernel. The usermode helper locks were originally added to prevent stalling suspend, later the firmware cache was added to help with this, and further later direct filesystem lookup was added by Linus to completely bypass udev due to the amount of issues the umh approach had. The usermode helper locks were kept even when the direct filesystem lookup mechanism is used though. A lot has changed since the original usermode helper locks were added but the recent commit which added the code for firmware_enabled() are intended to address any possible races cured only as collateral by using the locks as though side consequence of code evolution and this not being addressed any time sooner. With the firmware_enabled() code in place we are a bit more sure to move the usermode helper locks to UMH only code. There is a bit of history here so let's recap a bit of it to ensure nothing is lost and things are clear. The direct filesystem approach to loading firmware is rather new, it was added via commit abb139e75c2cdb ("firmware: teach the kernel to load firmware files directly from the filesystem") by Linus merged on the v3.7 release, to enable to bypass udev. usermodehelper_read_lock_wait() was added earlier via commit 9b78c1da60b3c ("firmware_class: Do not warn that system is not ready from async loads") merged on v3.4, after Rafael noted that the async firmware API call request_firmware_nowait() should not be penalized to fail if userspace is not available yet or frozen, it'd allow for a timeout grace period before giving up. The WARN_ON() was kept for the sync firmware API call though on request_firmware(). At this time there was no direct filesystem lookup for firmware though. The original usermode helper lock came from commit a144c6a6c924a ("PM: Print a warning if firmware is requested when tasks are frozen") merged on the v3.0 kernel by Rafael to print a warning back when firmware requests were used on resume(), thaw() or restore() callbacks and there was no direct fs lookups or the firmware cache. Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-02 15:31:07 +07:00
ret = _request_firmware_load(fw_priv, opt_flags, timeout);
if (!ret)
ret = assign_firmware_buf(firmware, device, opt_flags);
out_unlock:
usermodehelper_read_unlock();
return ret;
}
#else /* CONFIG_FW_LOADER_USER_HELPER */
static inline int
fw_load_from_user_helper(struct firmware *firmware, const char *name,
firmware: move umh try locks into the umh code This moves the usermode helper locks into only code paths that use the usermode helper API from the kernel. The usermode helper locks were originally added to prevent stalling suspend, later the firmware cache was added to help with this, and further later direct filesystem lookup was added by Linus to completely bypass udev due to the amount of issues the umh approach had. The usermode helper locks were kept even when the direct filesystem lookup mechanism is used though. A lot has changed since the original usermode helper locks were added but the recent commit which added the code for firmware_enabled() are intended to address any possible races cured only as collateral by using the locks as though side consequence of code evolution and this not being addressed any time sooner. With the firmware_enabled() code in place we are a bit more sure to move the usermode helper locks to UMH only code. There is a bit of history here so let's recap a bit of it to ensure nothing is lost and things are clear. The direct filesystem approach to loading firmware is rather new, it was added via commit abb139e75c2cdb ("firmware: teach the kernel to load firmware files directly from the filesystem") by Linus merged on the v3.7 release, to enable to bypass udev. usermodehelper_read_lock_wait() was added earlier via commit 9b78c1da60b3c ("firmware_class: Do not warn that system is not ready from async loads") merged on v3.4, after Rafael noted that the async firmware API call request_firmware_nowait() should not be penalized to fail if userspace is not available yet or frozen, it'd allow for a timeout grace period before giving up. The WARN_ON() was kept for the sync firmware API call though on request_firmware(). At this time there was no direct filesystem lookup for firmware though. The original usermode helper lock came from commit a144c6a6c924a ("PM: Print a warning if firmware is requested when tasks are frozen") merged on the v3.0 kernel by Rafael to print a warning back when firmware requests were used on resume(), thaw() or restore() callbacks and there was no direct fs lookups or the firmware cache. Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-02 15:31:07 +07:00
struct device *device, unsigned int opt_flags)
{
return -ENOENT;
}
static inline void kill_pending_fw_fallback_reqs(bool only_kill_custom) { }
#endif /* CONFIG_FW_LOADER_USER_HELPER */
/* prepare firmware and firmware_buf structs;
* return 0 if a firmware is already assigned, 1 if need to load one,
* or a negative error code
*/
static int
_request_firmware_prepare(struct firmware **firmware_p, const char *name,
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
struct device *device, void *dbuf, size_t size)
{
struct firmware *firmware;
struct firmware_buf *buf;
int ret;
*firmware_p = firmware = kzalloc(sizeof(*firmware), GFP_KERNEL);
if (!firmware) {
dev_err(device, "%s: kmalloc(struct firmware) failed\n",
__func__);
return -ENOMEM;
}
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
if (fw_get_builtin_firmware(firmware, name, dbuf, size)) {
firmware: simplify dev_*() print messages for generic helpers Simplify a few of the *generic* shared dev_warn() and dev_dbg() print messages for three reasons: 0) Historically firmware_class code was added to help get device driver firmware binaries but these days request_firmware*() helpers are being repurposed for general *system data* needed by the kernel. 1) This will also help generalize shared code as much as possible later in the future in consideration for a new extensible firmware API which will enable to separate usermode helper code out as much as possible. 2) Kees Cook pointed out the the prints already have the device associated as dev_*() helpers are used, that should help identify the user and case in which the helpers are used. That should provide enough context and simplifies the messages further. v4: generalize debug/warn messages even further as suggested by Kees Cook. Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: David Howells <dhowells@redhat.com> Cc: Kees Cook <keescook@chromium.org> Cc: Casey Schaufler <casey@schaufler-ca.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Takashi Iwai <tiwai@suse.de> Cc: Vojtěch Pavlík <vojtech@suse.cz> Cc: Kyle McMartin <kyle@kernel.org> Cc: Matthew Garrett <mjg59@srcf.ucam.org> Cc: linux-kernel@vger.kernel.org Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Mimi Zohar <zohar@linux.vnet.ibm.com> Acked-by: Kees Cook <keescook@chromium.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-04-30 06:30:43 +07:00
dev_dbg(device, "using built-in %s\n", name);
return 0; /* assigned */
}
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
ret = fw_lookup_and_allocate_buf(name, &fw_cache, &buf, dbuf, size);
/*
* bind with 'buf' now to avoid warning in failure path
* of requesting firmware.
*/
firmware->priv = buf;
if (ret > 0) {
ret = fw_state_wait(&buf->fw_st);
if (!ret) {
fw_set_page_data(buf, firmware);
return 0; /* assigned */
}
}
if (ret < 0)
return ret;
return 1; /* need to load */
}
/* called from request_firmware() and request_firmware_work_func() */
static int
_request_firmware(const struct firmware **firmware_p, const char *name,
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
struct device *device, void *buf, size_t size,
unsigned int opt_flags)
{
firmware: actually return NULL on failed request_firmware_nowait() The kerneldoc for request_firmware_nowait() says that it may call the provided cont() callback with @fw == NULL, if the firmware request fails. However, this is not the case when called with an empty string (""). This case is short-circuited by the 'name[0] == '\0'' check introduced in commit 471b095dfe0d ("firmware_class: make sure fw requests contain a name"), so _request_firmware() never gets to set the fw to NULL. Noticed while using the new 'trigger_async_request' testing hook: # printf '\x00' > /sys/devices/virtual/misc/test_firmware/trigger_async_request [10553.726178] test_firmware: loading '' [10553.729859] test_firmware: loaded: 995209091 # printf '\x00' > /sys/devices/virtual/misc/test_firmware/trigger_async_request [10733.676184] test_firmware: loading '' [10733.679855] Unable to handle kernel NULL pointer dereference at virtual address 00000004 [10733.687951] pgd = ec188000 [10733.690655] [00000004] *pgd=00000000 [10733.694240] Internal error: Oops: 5 [#1] SMP ARM [10733.698847] Modules linked in: btmrvl_sdio btmrvl bluetooth sbs_battery nf_conntrack_ipv6 nf_defrag_ipv6 ip6table_filter ip6_tables asix usbnet mwifiex_sdio mwifiex cfg80211 jitterentropy_rng drbg joydev snd_seq_midi snd_seq_midi_event snd_rawmidi snd_seq snd_seq_device ppp_async ppp_generic slhc tun [10733.725670] CPU: 0 PID: 6600 Comm: bash Not tainted 4.4.0-rc4-00351-g63d0877 #178 [10733.733137] Hardware name: Rockchip (Device Tree) [10733.737831] task: ed24f6c0 ti: ee322000 task.ti: ee322000 [10733.743222] PC is at do_raw_spin_lock+0x18/0x1a0 [10733.747831] LR is at _raw_spin_lock+0x18/0x1c [10733.752180] pc : [<c00653a0>] lr : [<c054c204>] psr: a00d0013 [10733.752180] sp : ee323df8 ip : ee323e20 fp : ee323e1c [10733.763634] r10: 00000051 r9 : b6f18000 r8 : ee323f80 [10733.768847] r7 : c089cebc r6 : 00000001 r5 : 00000000 r4 : ec0e6000 [10733.775360] r3 : dead4ead r2 : c06bd140 r1 : eef913b4 r0 : 00000000 [10733.781874] Flags: NzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none [10733.788995] Control: 10c5387d Table: 2c18806a DAC: 00000051 [10733.794728] Process bash (pid: 6600, stack limit = 0xee322218) [10733.800549] Stack: (0xee323df8 to 0xee324000) [10733.804896] 3de0: ec0e6000 00000000 [10733.813059] 3e00: 00000001 c089cebc ee323f80 b6f18000 ee323e2c ee323e20 c054c204 c0065394 [10733.821221] 3e20: ee323e44 ee323e30 c02fec60 c054c1f8 ec0e7ec0 ec3fcfc0 ee323e5c ee323e48 [10733.829384] 3e40: c02fed08 c02fec48 c07dbf74 eeb05a00 ee323e8c ee323e60 c0253828 c02fecac [10733.837547] 3e60: 00000001 c0116950 ee323eac ee323e78 00000001 ec3fce00 ed2d9700 ed2d970c [10733.845710] 3e80: ee323e9c ee323e90 c02e873c c02537d4 ee323eac ee323ea0 c017bd40 c02e8720 [10733.853873] 3ea0: ee323ee4 ee323eb0 c017b250 c017bd00 00000000 00000000 f3e47a54 ec128b00 [10733.862035] 3ec0: c017b10c ee323f80 00000001 c000f504 ee322000 00000000 ee323f4c ee323ee8 [10733.870197] 3ee0: c011b71c c017b118 ee323fb0 c011bc90 becfa8d9 00000001 ec128b00 00000001 [10733.878359] 3f00: b6f18000 ee323f80 ee323f4c ee323f18 c011bc90 c0063950 ee323f3c ee323f28 [10733.886522] 3f20: c0063950 c0549138 00000001 ec128b00 00000001 ec128b00 b6f18000 ee323f80 [10733.894684] 3f40: ee323f7c ee323f50 c011bed8 c011b6ec c0135fb8 c0135f24 ec128b00 ec128b00 [10733.902847] 3f60: 00000001 b6f18000 c000f504 ee322000 ee323fa4 ee323f80 c011c664 c011be24 [10733.911009] 3f80: 00000000 00000000 00000001 b6f18000 b6e79be0 00000004 00000000 ee323fa8 [10733.919172] 3fa0: c000f340 c011c618 00000001 b6f18000 00000001 b6f18000 00000001 00000000 [10733.927334] 3fc0: 00000001 b6f18000 b6e79be0 00000004 00000001 00000001 8068a3f1 b6e79c84 [10733.935496] 3fe0: 00000000 becfa7dc b6de194d b6e20246 400d0030 00000001 7a4536e8 49bda390 [10733.943664] [<c00653a0>] (do_raw_spin_lock) from [<c054c204>] (_raw_spin_lock+0x18/0x1c) [10733.951743] [<c054c204>] (_raw_spin_lock) from [<c02fec60>] (fw_free_buf+0x24/0x64) [10733.959388] [<c02fec60>] (fw_free_buf) from [<c02fed08>] (release_firmware+0x68/0x74) [10733.967207] [<c02fed08>] (release_firmware) from [<c0253828>] (trigger_async_request_store+0x60/0x124) [10733.976501] [<c0253828>] (trigger_async_request_store) from [<c02e873c>] (dev_attr_store+0x28/0x34) [10733.985533] [<c02e873c>] (dev_attr_store) from [<c017bd40>] (sysfs_kf_write+0x4c/0x58) [10733.993437] [<c017bd40>] (sysfs_kf_write) from [<c017b250>] (kernfs_fop_write+0x144/0x1a8) [10734.001689] [<c017b250>] (kernfs_fop_write) from [<c011b71c>] (__vfs_write+0x3c/0xe4) After this patch: # printf '\x00' > /sys/devices/virtual/misc/test_firmware/trigger_async_request [ 32.126322] test_firmware: loading '' [ 32.129995] test_firmware: failed to async load firmware -bash: printf: write error: No such device Fixes: 471b095dfe0d ("firmware_class: make sure fw requests contain a name") Signed-off-by: Brian Norris <computersforpeace@gmail.com> Acked-by: Ming Lei <ming.lei@canonical.com> Acked-by: Kees Cook <keescook@chromium.org> Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
2015-12-10 05:50:28 +07:00
struct firmware *fw = NULL;
int ret;
if (!firmware_p)
return -EINVAL;
firmware: actually return NULL on failed request_firmware_nowait() The kerneldoc for request_firmware_nowait() says that it may call the provided cont() callback with @fw == NULL, if the firmware request fails. However, this is not the case when called with an empty string (""). This case is short-circuited by the 'name[0] == '\0'' check introduced in commit 471b095dfe0d ("firmware_class: make sure fw requests contain a name"), so _request_firmware() never gets to set the fw to NULL. Noticed while using the new 'trigger_async_request' testing hook: # printf '\x00' > /sys/devices/virtual/misc/test_firmware/trigger_async_request [10553.726178] test_firmware: loading '' [10553.729859] test_firmware: loaded: 995209091 # printf '\x00' > /sys/devices/virtual/misc/test_firmware/trigger_async_request [10733.676184] test_firmware: loading '' [10733.679855] Unable to handle kernel NULL pointer dereference at virtual address 00000004 [10733.687951] pgd = ec188000 [10733.690655] [00000004] *pgd=00000000 [10733.694240] Internal error: Oops: 5 [#1] SMP ARM [10733.698847] Modules linked in: btmrvl_sdio btmrvl bluetooth sbs_battery nf_conntrack_ipv6 nf_defrag_ipv6 ip6table_filter ip6_tables asix usbnet mwifiex_sdio mwifiex cfg80211 jitterentropy_rng drbg joydev snd_seq_midi snd_seq_midi_event snd_rawmidi snd_seq snd_seq_device ppp_async ppp_generic slhc tun [10733.725670] CPU: 0 PID: 6600 Comm: bash Not tainted 4.4.0-rc4-00351-g63d0877 #178 [10733.733137] Hardware name: Rockchip (Device Tree) [10733.737831] task: ed24f6c0 ti: ee322000 task.ti: ee322000 [10733.743222] PC is at do_raw_spin_lock+0x18/0x1a0 [10733.747831] LR is at _raw_spin_lock+0x18/0x1c [10733.752180] pc : [<c00653a0>] lr : [<c054c204>] psr: a00d0013 [10733.752180] sp : ee323df8 ip : ee323e20 fp : ee323e1c [10733.763634] r10: 00000051 r9 : b6f18000 r8 : ee323f80 [10733.768847] r7 : c089cebc r6 : 00000001 r5 : 00000000 r4 : ec0e6000 [10733.775360] r3 : dead4ead r2 : c06bd140 r1 : eef913b4 r0 : 00000000 [10733.781874] Flags: NzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none [10733.788995] Control: 10c5387d Table: 2c18806a DAC: 00000051 [10733.794728] Process bash (pid: 6600, stack limit = 0xee322218) [10733.800549] Stack: (0xee323df8 to 0xee324000) [10733.804896] 3de0: ec0e6000 00000000 [10733.813059] 3e00: 00000001 c089cebc ee323f80 b6f18000 ee323e2c ee323e20 c054c204 c0065394 [10733.821221] 3e20: ee323e44 ee323e30 c02fec60 c054c1f8 ec0e7ec0 ec3fcfc0 ee323e5c ee323e48 [10733.829384] 3e40: c02fed08 c02fec48 c07dbf74 eeb05a00 ee323e8c ee323e60 c0253828 c02fecac [10733.837547] 3e60: 00000001 c0116950 ee323eac ee323e78 00000001 ec3fce00 ed2d9700 ed2d970c [10733.845710] 3e80: ee323e9c ee323e90 c02e873c c02537d4 ee323eac ee323ea0 c017bd40 c02e8720 [10733.853873] 3ea0: ee323ee4 ee323eb0 c017b250 c017bd00 00000000 00000000 f3e47a54 ec128b00 [10733.862035] 3ec0: c017b10c ee323f80 00000001 c000f504 ee322000 00000000 ee323f4c ee323ee8 [10733.870197] 3ee0: c011b71c c017b118 ee323fb0 c011bc90 becfa8d9 00000001 ec128b00 00000001 [10733.878359] 3f00: b6f18000 ee323f80 ee323f4c ee323f18 c011bc90 c0063950 ee323f3c ee323f28 [10733.886522] 3f20: c0063950 c0549138 00000001 ec128b00 00000001 ec128b00 b6f18000 ee323f80 [10733.894684] 3f40: ee323f7c ee323f50 c011bed8 c011b6ec c0135fb8 c0135f24 ec128b00 ec128b00 [10733.902847] 3f60: 00000001 b6f18000 c000f504 ee322000 ee323fa4 ee323f80 c011c664 c011be24 [10733.911009] 3f80: 00000000 00000000 00000001 b6f18000 b6e79be0 00000004 00000000 ee323fa8 [10733.919172] 3fa0: c000f340 c011c618 00000001 b6f18000 00000001 b6f18000 00000001 00000000 [10733.927334] 3fc0: 00000001 b6f18000 b6e79be0 00000004 00000001 00000001 8068a3f1 b6e79c84 [10733.935496] 3fe0: 00000000 becfa7dc b6de194d b6e20246 400d0030 00000001 7a4536e8 49bda390 [10733.943664] [<c00653a0>] (do_raw_spin_lock) from [<c054c204>] (_raw_spin_lock+0x18/0x1c) [10733.951743] [<c054c204>] (_raw_spin_lock) from [<c02fec60>] (fw_free_buf+0x24/0x64) [10733.959388] [<c02fec60>] (fw_free_buf) from [<c02fed08>] (release_firmware+0x68/0x74) [10733.967207] [<c02fed08>] (release_firmware) from [<c0253828>] (trigger_async_request_store+0x60/0x124) [10733.976501] [<c0253828>] (trigger_async_request_store) from [<c02e873c>] (dev_attr_store+0x28/0x34) [10733.985533] [<c02e873c>] (dev_attr_store) from [<c017bd40>] (sysfs_kf_write+0x4c/0x58) [10733.993437] [<c017bd40>] (sysfs_kf_write) from [<c017b250>] (kernfs_fop_write+0x144/0x1a8) [10734.001689] [<c017b250>] (kernfs_fop_write) from [<c011b71c>] (__vfs_write+0x3c/0xe4) After this patch: # printf '\x00' > /sys/devices/virtual/misc/test_firmware/trigger_async_request [ 32.126322] test_firmware: loading '' [ 32.129995] test_firmware: failed to async load firmware -bash: printf: write error: No such device Fixes: 471b095dfe0d ("firmware_class: make sure fw requests contain a name") Signed-off-by: Brian Norris <computersforpeace@gmail.com> Acked-by: Ming Lei <ming.lei@canonical.com> Acked-by: Kees Cook <keescook@chromium.org> Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
2015-12-10 05:50:28 +07:00
if (!name || name[0] == '\0') {
ret = -EINVAL;
goto out;
}
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
ret = _request_firmware_prepare(&fw, name, device, buf, size);
if (ret <= 0) /* error or already assigned */
goto out;
if (!firmware_enabled()) {
WARN(1, "firmware request while host is not available\n");
ret = -EHOSTDOWN;
goto out;
}
ret = fw_get_filesystem_firmware(device, fw->priv);
if (ret) {
if (!(opt_flags & FW_OPT_NO_WARN))
dev_warn(device,
"Direct firmware load for %s failed with error %d\n",
name, ret);
if (opt_flags & FW_OPT_USERHELPER) {
dev_warn(device, "Falling back to user helper\n");
ret = fw_load_from_user_helper(fw, name, device,
firmware: move umh try locks into the umh code This moves the usermode helper locks into only code paths that use the usermode helper API from the kernel. The usermode helper locks were originally added to prevent stalling suspend, later the firmware cache was added to help with this, and further later direct filesystem lookup was added by Linus to completely bypass udev due to the amount of issues the umh approach had. The usermode helper locks were kept even when the direct filesystem lookup mechanism is used though. A lot has changed since the original usermode helper locks were added but the recent commit which added the code for firmware_enabled() are intended to address any possible races cured only as collateral by using the locks as though side consequence of code evolution and this not being addressed any time sooner. With the firmware_enabled() code in place we are a bit more sure to move the usermode helper locks to UMH only code. There is a bit of history here so let's recap a bit of it to ensure nothing is lost and things are clear. The direct filesystem approach to loading firmware is rather new, it was added via commit abb139e75c2cdb ("firmware: teach the kernel to load firmware files directly from the filesystem") by Linus merged on the v3.7 release, to enable to bypass udev. usermodehelper_read_lock_wait() was added earlier via commit 9b78c1da60b3c ("firmware_class: Do not warn that system is not ready from async loads") merged on v3.4, after Rafael noted that the async firmware API call request_firmware_nowait() should not be penalized to fail if userspace is not available yet or frozen, it'd allow for a timeout grace period before giving up. The WARN_ON() was kept for the sync firmware API call though on request_firmware(). At this time there was no direct filesystem lookup for firmware though. The original usermode helper lock came from commit a144c6a6c924a ("PM: Print a warning if firmware is requested when tasks are frozen") merged on the v3.0 kernel by Rafael to print a warning back when firmware requests were used on resume(), thaw() or restore() callbacks and there was no direct fs lookups or the firmware cache. Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-02 15:31:07 +07:00
opt_flags);
}
firmware: move umh try locks into the umh code This moves the usermode helper locks into only code paths that use the usermode helper API from the kernel. The usermode helper locks were originally added to prevent stalling suspend, later the firmware cache was added to help with this, and further later direct filesystem lookup was added by Linus to completely bypass udev due to the amount of issues the umh approach had. The usermode helper locks were kept even when the direct filesystem lookup mechanism is used though. A lot has changed since the original usermode helper locks were added but the recent commit which added the code for firmware_enabled() are intended to address any possible races cured only as collateral by using the locks as though side consequence of code evolution and this not being addressed any time sooner. With the firmware_enabled() code in place we are a bit more sure to move the usermode helper locks to UMH only code. There is a bit of history here so let's recap a bit of it to ensure nothing is lost and things are clear. The direct filesystem approach to loading firmware is rather new, it was added via commit abb139e75c2cdb ("firmware: teach the kernel to load firmware files directly from the filesystem") by Linus merged on the v3.7 release, to enable to bypass udev. usermodehelper_read_lock_wait() was added earlier via commit 9b78c1da60b3c ("firmware_class: Do not warn that system is not ready from async loads") merged on v3.4, after Rafael noted that the async firmware API call request_firmware_nowait() should not be penalized to fail if userspace is not available yet or frozen, it'd allow for a timeout grace period before giving up. The WARN_ON() was kept for the sync firmware API call though on request_firmware(). At this time there was no direct filesystem lookup for firmware though. The original usermode helper lock came from commit a144c6a6c924a ("PM: Print a warning if firmware is requested when tasks are frozen") merged on the v3.0 kernel by Rafael to print a warning back when firmware requests were used on resume(), thaw() or restore() callbacks and there was no direct fs lookups or the firmware cache. Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-05-02 15:31:07 +07:00
} else
ret = assign_firmware_buf(fw, device, opt_flags);
out:
if (ret < 0) {
release_firmware(fw);
fw = NULL;
}
*firmware_p = fw;
return ret;
}
/**
* request_firmware: - send firmware request and wait for it
* @firmware_p: pointer to firmware image
* @name: name of firmware file
* @device: device for which firmware is being loaded
*
* @firmware_p will be used to return a firmware image by the name
* of @name for device @device.
*
* Should be called from user context where sleeping is allowed.
*
* @name will be used as $FIRMWARE in the uevent environment and
* should be distinctive enough not to be confused with any other
* firmware image for this or any other device.
*
* Caller must hold the reference count of @device.
*
* The function can be called safely inside device's suspend and
* resume callback.
**/
int
request_firmware(const struct firmware **firmware_p, const char *name,
struct device *device)
{
int ret;
/* Need to pin this module until return */
__module_get(THIS_MODULE);
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
ret = _request_firmware(firmware_p, name, device, NULL, 0,
FW_OPT_UEVENT | FW_OPT_FALLBACK);
module_put(THIS_MODULE);
return ret;
}
EXPORT_SYMBOL(request_firmware);
/**
* request_firmware_direct: - load firmware directly without usermode helper
* @firmware_p: pointer to firmware image
* @name: name of firmware file
* @device: device for which firmware is being loaded
*
* This function works pretty much like request_firmware(), but this doesn't
* fall back to usermode helper even if the firmware couldn't be loaded
* directly from fs. Hence it's useful for loading optional firmwares, which
* aren't always present, without extra long timeouts of udev.
**/
int request_firmware_direct(const struct firmware **firmware_p,
const char *name, struct device *device)
{
int ret;
__module_get(THIS_MODULE);
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
ret = _request_firmware(firmware_p, name, device, NULL, 0,
FW_OPT_UEVENT | FW_OPT_NO_WARN);
module_put(THIS_MODULE);
return ret;
}
EXPORT_SYMBOL_GPL(request_firmware_direct);
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
/**
* request_firmware_into_buf - load firmware into a previously allocated buffer
* @firmware_p: pointer to firmware image
* @name: name of firmware file
* @device: device for which firmware is being loaded and DMA region allocated
* @buf: address of buffer to load firmware into
* @size: size of buffer
*
* This function works pretty much like request_firmware(), but it doesn't
* allocate a buffer to hold the firmware data. Instead, the firmware
* is loaded directly into the buffer pointed to by @buf and the @firmware_p
* data member is pointed at @buf.
*
* This function doesn't cache firmware either.
*/
int
request_firmware_into_buf(const struct firmware **firmware_p, const char *name,
struct device *device, void *buf, size_t size)
{
int ret;
__module_get(THIS_MODULE);
ret = _request_firmware(firmware_p, name, device, buf, size,
FW_OPT_UEVENT | FW_OPT_FALLBACK |
FW_OPT_NOCACHE);
module_put(THIS_MODULE);
return ret;
}
EXPORT_SYMBOL(request_firmware_into_buf);
/**
* release_firmware: - release the resource associated with a firmware image
* @fw: firmware resource to release
**/
void release_firmware(const struct firmware *fw)
{
if (fw) {
if (!fw_is_builtin_firmware(fw))
firmware_free_data(fw);
kfree(fw);
}
}
EXPORT_SYMBOL(release_firmware);
/* Async support */
struct firmware_work {
struct work_struct work;
struct module *module;
const char *name;
struct device *device;
void *context;
void (*cont)(const struct firmware *fw, void *context);
unsigned int opt_flags;
};
static void request_firmware_work_func(struct work_struct *work)
{
struct firmware_work *fw_work;
const struct firmware *fw;
fw_work = container_of(work, struct firmware_work, work);
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
_request_firmware(&fw, fw_work->name, fw_work->device, NULL, 0,
fw_work->opt_flags);
fw_work->cont(fw, fw_work->context);
put_device(fw_work->device); /* taken in request_firmware_nowait() */
module_put(fw_work->module);
firmware: fix possible use after free on name on asynchronous request Asynchronous firmware loading copies the pointer to the name passed as an argument only to be scheduled later and used. This behaviour works well for synchronous calling but in asynchronous mode there's a chance the caller could immediately free the passed string after making the asynchronous call. This could trigger a use after free having the kernel look on disk for arbitrary file names. In order to force-test the issue you can use a test-driver designed to illustrate this issue on github [0], use the next-20150505-fix-use-after-free branch. With this patch applied you get: [ 283.512445] firmware name: test_module_stuff.bin [ 287.514020] firmware name: test_module_stuff.bin [ 287.532489] firmware found Without this patch applied you can end up with something such as: [ 135.624216] firmware name: \xffffff80BJ [ 135.624249] platform fake-dev.0: Direct firmware load for \xffffff80Bi failed with error -2 [ 135.624252] No firmware found [ 135.624252] firmware found Unfortunatley in the worst and most common case however you can typically crash your system with a page fault by trying to free something which you cannot, and/or a NULL pointer dereference [1]. The fix and issue using schedule_work() for asynchronous runs is generalized in the following SmPL grammar patch, when applied to next-20150505 only the firmware_class code is affected. This grammar patch can and should further be generalized to vet for for other kernel asynchronous mechanisms. @ calls_schedule_work @ type T; T *priv_work; identifier func, work_func; identifier work; identifier priv_name, name; expression gfp; @@ func(..., const char *name, ...) { ... priv_work = kzalloc(sizeof(T), gfp); ... - priv_work->priv_name = name; + priv_work->priv_name = kstrdup_const(name, gfp); ... (... when any if (...) { ... + kfree_const(priv_work->priv_name); kfree(priv_work); ... } ) ... when any INIT_WORK(&priv_work->work, work_func); ... schedule_work(&priv_work->work); ... } @ the_work_func depends on calls_schedule_work @ type calls_schedule_work.T; T *priv_work; identifier calls_schedule_work.work_func; identifier calls_schedule_work.priv_name; identifier calls_schedule_work.work; identifier some_work; @@ work_func(...) { ... priv_work = container_of(some_work, T, work); ... + kfree_const(priv_work->priv_name); kfree(priv_work); ... } [0] https://github.com/mcgrof/fake-firmware-test.git [1] The following kernel ring buffer splat: firmware name: test_module_stuff.bin firmware name: firmware found general protection fault: 0000 [#1] SMP Modules linked in: test(O) <...etc-it-does-not-matter> drm sr_mod cdrom xhci_pci xhci_hcd rtsx_pci mfd_core video button sg CPU: 3 PID: 87 Comm: kworker/3:2 Tainted: G O 4.0.0-00010-g22b5bb0-dirty #176 Hardware name: LENOVO 20AW000LUS/20AW000LUS, BIOS GLET43WW (1.18 ) 12/04/2013 Workqueue: events request_firmware_work_func task: ffff8800c7f8e290 ti: ffff8800c7f94000 task.ti: ffff8800c7f94000 RIP: 0010:[<ffffffff814a586c>] [<ffffffff814a586c>] fw_free_buf+0xc/0x40 RSP: 0000:ffff8800c7f97d78 EFLAGS: 00010286 RAX: ffffffff81ae3700 RBX: ffffffff816d1181 RCX: 0000000000000006 RDX: 0001ee850ff68500 RSI: 0000000000000246 RDI: c35d5f415e415d41 RBP: ffff8800c7f97d88 R08: 000000000000000a R09: 0000000000000000 R10: 0000000000000358 R11: ffff8800c7f97a7e R12: ffff8800c7ec1e80 R13: ffff88021e2d4cc0 R14: ffff88021e2dff00 R15: 00000000000000c0 FS: 0000000000000000(0000) GS:ffff88021e2c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000034b8cd8 CR3: 000000021073c000 CR4: 00000000001407e0 Stack: ffffffff816d1181 ffff8800c7ec1e80 ffff8800c7f97da8 ffffffff814a58f8 000000000000000a ffffffff816d1181 ffff8800c7f97dc8 ffffffffa047002c ffff88021e2dff00 ffff8802116ac1c0 ffff8800c7f97df8 ffffffff814a65fe Call Trace: [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a58f8>] release_firmware+0x58/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffffa047002c>] test_mod_cb+0x2c/0x43 [test] [<ffffffff814a65fe>] request_firmware_work_func+0x5e/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff8108d23a>] process_one_work+0x14a/0x3f0 [<ffffffff8108d911>] worker_thread+0x121/0x460 [<ffffffff8108d7f0>] ? rescuer_thread+0x310/0x310 [<ffffffff810928f9>] kthread+0xc9/0xe0 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 [<ffffffff816d52d8>] ret_from_fork+0x58/0x90 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 Code: c7 c6 dd ad a3 81 48 c7 c7 20 97 ce 81 31 c0 e8 0b b2 ed ff e9 78 ff ff ff 66 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 54 53 <4c> 8b 67 38 48 89 fb 4c 89 e7 e8 85 f7 22 00 f0 83 2b 01 74 0f RIP [<ffffffff814a586c>] fw_free_buf+0xc/0x40 RSP <ffff8800c7f97d78> ---[ end trace 4e62c56a58d0eac1 ]--- BUG: unable to handle kernel paging request at ffffffffffffffd8 IP: [<ffffffff81093ee0>] kthread_data+0x10/0x20 PGD 1c13067 PUD 1c15067 PMD 0 Oops: 0000 [#2] SMP Modules linked in: test(O) <...etc-it-does-not-matter> drm sr_mod cdrom xhci_pci xhci_hcd rtsx_pci mfd_core video button sg CPU: 3 PID: 87 Comm: kworker/3:2 Tainted: G D O 4.0.0-00010-g22b5bb0-dirty #176 Hardware name: LENOVO 20AW000LUS/20AW000LUS, BIOS GLET43WW (1.18 ) 12/04/2013 task: ffff8800c7f8e290 ti: ffff8800c7f94000 task.ti: ffff8800c7f94000 RIP: 0010:[<ffffffff81092ee0>] [<ffffffff81092ee0>] kthread_data+0x10/0x20 RSP: 0018:ffff8800c7f97b18 EFLAGS: 00010096 RAX: 0000000000000000 RBX: 0000000000000003 RCX: 000000000000000d RDX: 0000000000000003 RSI: 0000000000000003 RDI: ffff8800c7f8e290 RBP: ffff8800c7f97b18 R08: 000000000000bc00 R09: 0000000000007e76 R10: 0000000000000001 R11: 000000000000002f R12: ffff8800c7f8e290 R13: 00000000000154c0 R14: 0000000000000003 R15: 0000000000000000 FS: 0000000000000000(0000) GS:ffff88021e2c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000028 CR3: 0000000210675000 CR4: 00000000001407e0 Stack: ffff8800c7f97b38 ffffffff8108dcd5 ffff8800c7f97b38 ffff88021e2d54c0 ffff8800c7f97b88 ffffffff816d1500 ffff880213d42368 ffff8800c7f8e290 ffff8800c7f97b88 ffff8800c7f97fd8 ffff8800c7f8e710 0000000000000246 Call Trace: [<ffffffff8108dcd5>] wq_worker_sleeping+0x15/0xa0 [<ffffffff816d1500>] __schedule+0x6e0/0x940 [<ffffffff816d1797>] schedule+0x37/0x90 [<ffffffff810779bc>] do_exit+0x6bc/0xb40 [<ffffffff8101898f>] oops_end+0x9f/0xe0 [<ffffffff81018efb>] die+0x4b/0x70 [<ffffffff81015622>] do_general_protection+0xe2/0x170 [<ffffffff816d74e8>] general_protection+0x28/0x30 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a586c>] ? fw_free_buf+0xc/0x40 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a58f8>] release_firmware+0x58/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffffa047002c>] test_mod_cb+0x2c/0x43 [test] [<ffffffff814a65fe>] request_firmware_work_func+0x5e/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff8108d23a>] process_one_work+0x14a/0x3f0 [<ffffffff8108d911>] worker_thread+0x121/0x460 [<ffffffff8108d7f0>] ? rescuer_thread+0x310/0x310 [<ffffffff810928f9>] kthread+0xc9/0xe0 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 [<ffffffff816d52d8>] ret_from_fork+0x58/0x90 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 Code: 00 48 89 e5 5d 48 8b 40 c8 48 c1 e8 02 83 e0 01 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 8b 87 30 05 00 00 55 48 89 e5 <48> 8b 40 d8 5d c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 RIP [<ffffffff81092ee0>] kthread_data+0x10/0x20 RSP <ffff8800c7f97b18> CR2: ffffffffffffffd8 ---[ end trace 4e62c56a58d0eac2 ]--- Fixing recursive fault but reboot is needed! Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Generated-by: Coccinelle SmPL Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:42 +07:00
kfree_const(fw_work->name);
kfree(fw_work);
}
/**
* request_firmware_nowait - asynchronous version of request_firmware
* @module: module requesting the firmware
* @uevent: sends uevent to copy the firmware image if this flag
* is non-zero else the firmware copy must be done manually.
* @name: name of firmware file
* @device: device for which firmware is being loaded
* @gfp: allocation flags
* @context: will be passed over to @cont, and
* @fw may be %NULL if firmware request fails.
* @cont: function will be called asynchronously when the firmware
* request is over.
*
* Caller must hold the reference count of @device.
*
* Asynchronous variant of request_firmware() for user contexts:
* - sleep for as small periods as possible since it may
* increase kernel boot time of built-in device drivers
* requesting firmware in their ->probe() methods, if
* @gfp is GFP_KERNEL.
*
* - can't sleep at all if @gfp is GFP_ATOMIC.
**/
int
request_firmware_nowait(
struct module *module, bool uevent,
const char *name, struct device *device, gfp_t gfp, void *context,
void (*cont)(const struct firmware *fw, void *context))
{
struct firmware_work *fw_work;
fw_work = kzalloc(sizeof(struct firmware_work), gfp);
if (!fw_work)
return -ENOMEM;
fw_work->module = module;
firmware: fix possible use after free on name on asynchronous request Asynchronous firmware loading copies the pointer to the name passed as an argument only to be scheduled later and used. This behaviour works well for synchronous calling but in asynchronous mode there's a chance the caller could immediately free the passed string after making the asynchronous call. This could trigger a use after free having the kernel look on disk for arbitrary file names. In order to force-test the issue you can use a test-driver designed to illustrate this issue on github [0], use the next-20150505-fix-use-after-free branch. With this patch applied you get: [ 283.512445] firmware name: test_module_stuff.bin [ 287.514020] firmware name: test_module_stuff.bin [ 287.532489] firmware found Without this patch applied you can end up with something such as: [ 135.624216] firmware name: \xffffff80BJ [ 135.624249] platform fake-dev.0: Direct firmware load for \xffffff80Bi failed with error -2 [ 135.624252] No firmware found [ 135.624252] firmware found Unfortunatley in the worst and most common case however you can typically crash your system with a page fault by trying to free something which you cannot, and/or a NULL pointer dereference [1]. The fix and issue using schedule_work() for asynchronous runs is generalized in the following SmPL grammar patch, when applied to next-20150505 only the firmware_class code is affected. This grammar patch can and should further be generalized to vet for for other kernel asynchronous mechanisms. @ calls_schedule_work @ type T; T *priv_work; identifier func, work_func; identifier work; identifier priv_name, name; expression gfp; @@ func(..., const char *name, ...) { ... priv_work = kzalloc(sizeof(T), gfp); ... - priv_work->priv_name = name; + priv_work->priv_name = kstrdup_const(name, gfp); ... (... when any if (...) { ... + kfree_const(priv_work->priv_name); kfree(priv_work); ... } ) ... when any INIT_WORK(&priv_work->work, work_func); ... schedule_work(&priv_work->work); ... } @ the_work_func depends on calls_schedule_work @ type calls_schedule_work.T; T *priv_work; identifier calls_schedule_work.work_func; identifier calls_schedule_work.priv_name; identifier calls_schedule_work.work; identifier some_work; @@ work_func(...) { ... priv_work = container_of(some_work, T, work); ... + kfree_const(priv_work->priv_name); kfree(priv_work); ... } [0] https://github.com/mcgrof/fake-firmware-test.git [1] The following kernel ring buffer splat: firmware name: test_module_stuff.bin firmware name: firmware found general protection fault: 0000 [#1] SMP Modules linked in: test(O) <...etc-it-does-not-matter> drm sr_mod cdrom xhci_pci xhci_hcd rtsx_pci mfd_core video button sg CPU: 3 PID: 87 Comm: kworker/3:2 Tainted: G O 4.0.0-00010-g22b5bb0-dirty #176 Hardware name: LENOVO 20AW000LUS/20AW000LUS, BIOS GLET43WW (1.18 ) 12/04/2013 Workqueue: events request_firmware_work_func task: ffff8800c7f8e290 ti: ffff8800c7f94000 task.ti: ffff8800c7f94000 RIP: 0010:[<ffffffff814a586c>] [<ffffffff814a586c>] fw_free_buf+0xc/0x40 RSP: 0000:ffff8800c7f97d78 EFLAGS: 00010286 RAX: ffffffff81ae3700 RBX: ffffffff816d1181 RCX: 0000000000000006 RDX: 0001ee850ff68500 RSI: 0000000000000246 RDI: c35d5f415e415d41 RBP: ffff8800c7f97d88 R08: 000000000000000a R09: 0000000000000000 R10: 0000000000000358 R11: ffff8800c7f97a7e R12: ffff8800c7ec1e80 R13: ffff88021e2d4cc0 R14: ffff88021e2dff00 R15: 00000000000000c0 FS: 0000000000000000(0000) GS:ffff88021e2c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000034b8cd8 CR3: 000000021073c000 CR4: 00000000001407e0 Stack: ffffffff816d1181 ffff8800c7ec1e80 ffff8800c7f97da8 ffffffff814a58f8 000000000000000a ffffffff816d1181 ffff8800c7f97dc8 ffffffffa047002c ffff88021e2dff00 ffff8802116ac1c0 ffff8800c7f97df8 ffffffff814a65fe Call Trace: [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a58f8>] release_firmware+0x58/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffffa047002c>] test_mod_cb+0x2c/0x43 [test] [<ffffffff814a65fe>] request_firmware_work_func+0x5e/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff8108d23a>] process_one_work+0x14a/0x3f0 [<ffffffff8108d911>] worker_thread+0x121/0x460 [<ffffffff8108d7f0>] ? rescuer_thread+0x310/0x310 [<ffffffff810928f9>] kthread+0xc9/0xe0 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 [<ffffffff816d52d8>] ret_from_fork+0x58/0x90 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 Code: c7 c6 dd ad a3 81 48 c7 c7 20 97 ce 81 31 c0 e8 0b b2 ed ff e9 78 ff ff ff 66 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 54 53 <4c> 8b 67 38 48 89 fb 4c 89 e7 e8 85 f7 22 00 f0 83 2b 01 74 0f RIP [<ffffffff814a586c>] fw_free_buf+0xc/0x40 RSP <ffff8800c7f97d78> ---[ end trace 4e62c56a58d0eac1 ]--- BUG: unable to handle kernel paging request at ffffffffffffffd8 IP: [<ffffffff81093ee0>] kthread_data+0x10/0x20 PGD 1c13067 PUD 1c15067 PMD 0 Oops: 0000 [#2] SMP Modules linked in: test(O) <...etc-it-does-not-matter> drm sr_mod cdrom xhci_pci xhci_hcd rtsx_pci mfd_core video button sg CPU: 3 PID: 87 Comm: kworker/3:2 Tainted: G D O 4.0.0-00010-g22b5bb0-dirty #176 Hardware name: LENOVO 20AW000LUS/20AW000LUS, BIOS GLET43WW (1.18 ) 12/04/2013 task: ffff8800c7f8e290 ti: ffff8800c7f94000 task.ti: ffff8800c7f94000 RIP: 0010:[<ffffffff81092ee0>] [<ffffffff81092ee0>] kthread_data+0x10/0x20 RSP: 0018:ffff8800c7f97b18 EFLAGS: 00010096 RAX: 0000000000000000 RBX: 0000000000000003 RCX: 000000000000000d RDX: 0000000000000003 RSI: 0000000000000003 RDI: ffff8800c7f8e290 RBP: ffff8800c7f97b18 R08: 000000000000bc00 R09: 0000000000007e76 R10: 0000000000000001 R11: 000000000000002f R12: ffff8800c7f8e290 R13: 00000000000154c0 R14: 0000000000000003 R15: 0000000000000000 FS: 0000000000000000(0000) GS:ffff88021e2c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000028 CR3: 0000000210675000 CR4: 00000000001407e0 Stack: ffff8800c7f97b38 ffffffff8108dcd5 ffff8800c7f97b38 ffff88021e2d54c0 ffff8800c7f97b88 ffffffff816d1500 ffff880213d42368 ffff8800c7f8e290 ffff8800c7f97b88 ffff8800c7f97fd8 ffff8800c7f8e710 0000000000000246 Call Trace: [<ffffffff8108dcd5>] wq_worker_sleeping+0x15/0xa0 [<ffffffff816d1500>] __schedule+0x6e0/0x940 [<ffffffff816d1797>] schedule+0x37/0x90 [<ffffffff810779bc>] do_exit+0x6bc/0xb40 [<ffffffff8101898f>] oops_end+0x9f/0xe0 [<ffffffff81018efb>] die+0x4b/0x70 [<ffffffff81015622>] do_general_protection+0xe2/0x170 [<ffffffff816d74e8>] general_protection+0x28/0x30 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a586c>] ? fw_free_buf+0xc/0x40 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a58f8>] release_firmware+0x58/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffffa047002c>] test_mod_cb+0x2c/0x43 [test] [<ffffffff814a65fe>] request_firmware_work_func+0x5e/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff8108d23a>] process_one_work+0x14a/0x3f0 [<ffffffff8108d911>] worker_thread+0x121/0x460 [<ffffffff8108d7f0>] ? rescuer_thread+0x310/0x310 [<ffffffff810928f9>] kthread+0xc9/0xe0 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 [<ffffffff816d52d8>] ret_from_fork+0x58/0x90 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 Code: 00 48 89 e5 5d 48 8b 40 c8 48 c1 e8 02 83 e0 01 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 8b 87 30 05 00 00 55 48 89 e5 <48> 8b 40 d8 5d c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 RIP [<ffffffff81092ee0>] kthread_data+0x10/0x20 RSP <ffff8800c7f97b18> CR2: ffffffffffffffd8 ---[ end trace 4e62c56a58d0eac2 ]--- Fixing recursive fault but reboot is needed! Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Generated-by: Coccinelle SmPL Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:42 +07:00
fw_work->name = kstrdup_const(name, gfp);
if (!fw_work->name) {
kfree(fw_work);
firmware: fix possible use after free on name on asynchronous request Asynchronous firmware loading copies the pointer to the name passed as an argument only to be scheduled later and used. This behaviour works well for synchronous calling but in asynchronous mode there's a chance the caller could immediately free the passed string after making the asynchronous call. This could trigger a use after free having the kernel look on disk for arbitrary file names. In order to force-test the issue you can use a test-driver designed to illustrate this issue on github [0], use the next-20150505-fix-use-after-free branch. With this patch applied you get: [ 283.512445] firmware name: test_module_stuff.bin [ 287.514020] firmware name: test_module_stuff.bin [ 287.532489] firmware found Without this patch applied you can end up with something such as: [ 135.624216] firmware name: \xffffff80BJ [ 135.624249] platform fake-dev.0: Direct firmware load for \xffffff80Bi failed with error -2 [ 135.624252] No firmware found [ 135.624252] firmware found Unfortunatley in the worst and most common case however you can typically crash your system with a page fault by trying to free something which you cannot, and/or a NULL pointer dereference [1]. The fix and issue using schedule_work() for asynchronous runs is generalized in the following SmPL grammar patch, when applied to next-20150505 only the firmware_class code is affected. This grammar patch can and should further be generalized to vet for for other kernel asynchronous mechanisms. @ calls_schedule_work @ type T; T *priv_work; identifier func, work_func; identifier work; identifier priv_name, name; expression gfp; @@ func(..., const char *name, ...) { ... priv_work = kzalloc(sizeof(T), gfp); ... - priv_work->priv_name = name; + priv_work->priv_name = kstrdup_const(name, gfp); ... (... when any if (...) { ... + kfree_const(priv_work->priv_name); kfree(priv_work); ... } ) ... when any INIT_WORK(&priv_work->work, work_func); ... schedule_work(&priv_work->work); ... } @ the_work_func depends on calls_schedule_work @ type calls_schedule_work.T; T *priv_work; identifier calls_schedule_work.work_func; identifier calls_schedule_work.priv_name; identifier calls_schedule_work.work; identifier some_work; @@ work_func(...) { ... priv_work = container_of(some_work, T, work); ... + kfree_const(priv_work->priv_name); kfree(priv_work); ... } [0] https://github.com/mcgrof/fake-firmware-test.git [1] The following kernel ring buffer splat: firmware name: test_module_stuff.bin firmware name: firmware found general protection fault: 0000 [#1] SMP Modules linked in: test(O) <...etc-it-does-not-matter> drm sr_mod cdrom xhci_pci xhci_hcd rtsx_pci mfd_core video button sg CPU: 3 PID: 87 Comm: kworker/3:2 Tainted: G O 4.0.0-00010-g22b5bb0-dirty #176 Hardware name: LENOVO 20AW000LUS/20AW000LUS, BIOS GLET43WW (1.18 ) 12/04/2013 Workqueue: events request_firmware_work_func task: ffff8800c7f8e290 ti: ffff8800c7f94000 task.ti: ffff8800c7f94000 RIP: 0010:[<ffffffff814a586c>] [<ffffffff814a586c>] fw_free_buf+0xc/0x40 RSP: 0000:ffff8800c7f97d78 EFLAGS: 00010286 RAX: ffffffff81ae3700 RBX: ffffffff816d1181 RCX: 0000000000000006 RDX: 0001ee850ff68500 RSI: 0000000000000246 RDI: c35d5f415e415d41 RBP: ffff8800c7f97d88 R08: 000000000000000a R09: 0000000000000000 R10: 0000000000000358 R11: ffff8800c7f97a7e R12: ffff8800c7ec1e80 R13: ffff88021e2d4cc0 R14: ffff88021e2dff00 R15: 00000000000000c0 FS: 0000000000000000(0000) GS:ffff88021e2c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000034b8cd8 CR3: 000000021073c000 CR4: 00000000001407e0 Stack: ffffffff816d1181 ffff8800c7ec1e80 ffff8800c7f97da8 ffffffff814a58f8 000000000000000a ffffffff816d1181 ffff8800c7f97dc8 ffffffffa047002c ffff88021e2dff00 ffff8802116ac1c0 ffff8800c7f97df8 ffffffff814a65fe Call Trace: [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a58f8>] release_firmware+0x58/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffffa047002c>] test_mod_cb+0x2c/0x43 [test] [<ffffffff814a65fe>] request_firmware_work_func+0x5e/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff8108d23a>] process_one_work+0x14a/0x3f0 [<ffffffff8108d911>] worker_thread+0x121/0x460 [<ffffffff8108d7f0>] ? rescuer_thread+0x310/0x310 [<ffffffff810928f9>] kthread+0xc9/0xe0 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 [<ffffffff816d52d8>] ret_from_fork+0x58/0x90 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 Code: c7 c6 dd ad a3 81 48 c7 c7 20 97 ce 81 31 c0 e8 0b b2 ed ff e9 78 ff ff ff 66 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 54 53 <4c> 8b 67 38 48 89 fb 4c 89 e7 e8 85 f7 22 00 f0 83 2b 01 74 0f RIP [<ffffffff814a586c>] fw_free_buf+0xc/0x40 RSP <ffff8800c7f97d78> ---[ end trace 4e62c56a58d0eac1 ]--- BUG: unable to handle kernel paging request at ffffffffffffffd8 IP: [<ffffffff81093ee0>] kthread_data+0x10/0x20 PGD 1c13067 PUD 1c15067 PMD 0 Oops: 0000 [#2] SMP Modules linked in: test(O) <...etc-it-does-not-matter> drm sr_mod cdrom xhci_pci xhci_hcd rtsx_pci mfd_core video button sg CPU: 3 PID: 87 Comm: kworker/3:2 Tainted: G D O 4.0.0-00010-g22b5bb0-dirty #176 Hardware name: LENOVO 20AW000LUS/20AW000LUS, BIOS GLET43WW (1.18 ) 12/04/2013 task: ffff8800c7f8e290 ti: ffff8800c7f94000 task.ti: ffff8800c7f94000 RIP: 0010:[<ffffffff81092ee0>] [<ffffffff81092ee0>] kthread_data+0x10/0x20 RSP: 0018:ffff8800c7f97b18 EFLAGS: 00010096 RAX: 0000000000000000 RBX: 0000000000000003 RCX: 000000000000000d RDX: 0000000000000003 RSI: 0000000000000003 RDI: ffff8800c7f8e290 RBP: ffff8800c7f97b18 R08: 000000000000bc00 R09: 0000000000007e76 R10: 0000000000000001 R11: 000000000000002f R12: ffff8800c7f8e290 R13: 00000000000154c0 R14: 0000000000000003 R15: 0000000000000000 FS: 0000000000000000(0000) GS:ffff88021e2c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000028 CR3: 0000000210675000 CR4: 00000000001407e0 Stack: ffff8800c7f97b38 ffffffff8108dcd5 ffff8800c7f97b38 ffff88021e2d54c0 ffff8800c7f97b88 ffffffff816d1500 ffff880213d42368 ffff8800c7f8e290 ffff8800c7f97b88 ffff8800c7f97fd8 ffff8800c7f8e710 0000000000000246 Call Trace: [<ffffffff8108dcd5>] wq_worker_sleeping+0x15/0xa0 [<ffffffff816d1500>] __schedule+0x6e0/0x940 [<ffffffff816d1797>] schedule+0x37/0x90 [<ffffffff810779bc>] do_exit+0x6bc/0xb40 [<ffffffff8101898f>] oops_end+0x9f/0xe0 [<ffffffff81018efb>] die+0x4b/0x70 [<ffffffff81015622>] do_general_protection+0xe2/0x170 [<ffffffff816d74e8>] general_protection+0x28/0x30 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a586c>] ? fw_free_buf+0xc/0x40 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a58f8>] release_firmware+0x58/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffffa047002c>] test_mod_cb+0x2c/0x43 [test] [<ffffffff814a65fe>] request_firmware_work_func+0x5e/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff8108d23a>] process_one_work+0x14a/0x3f0 [<ffffffff8108d911>] worker_thread+0x121/0x460 [<ffffffff8108d7f0>] ? rescuer_thread+0x310/0x310 [<ffffffff810928f9>] kthread+0xc9/0xe0 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 [<ffffffff816d52d8>] ret_from_fork+0x58/0x90 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 Code: 00 48 89 e5 5d 48 8b 40 c8 48 c1 e8 02 83 e0 01 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 8b 87 30 05 00 00 55 48 89 e5 <48> 8b 40 d8 5d c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 RIP [<ffffffff81092ee0>] kthread_data+0x10/0x20 RSP <ffff8800c7f97b18> CR2: ffffffffffffffd8 ---[ end trace 4e62c56a58d0eac2 ]--- Fixing recursive fault but reboot is needed! Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Generated-by: Coccinelle SmPL Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:42 +07:00
return -ENOMEM;
}
fw_work->device = device;
fw_work->context = context;
fw_work->cont = cont;
fw_work->opt_flags = FW_OPT_NOWAIT | FW_OPT_FALLBACK |
firmware loader: allow disabling of udev as firmware loader [The patch was originally proposed by Tom Gundersen, and rewritten afterwards by me; most of changelogs below borrowed from Tom's original patch -- tiwai] Currently (at least) the dell-rbu driver selects FW_LOADER_USER_HELPER, which means that distros can't really stop loading firmware through udev without breaking other users (though some have). Ideally we would remove/disable the udev firmware helper in both the kernel and in udev, but if we were to disable it in udev and not the kernel, the result would be (seemingly) hung kernels as no one would be around to cancel firmware requests. This patch allows udev firmware loading to be disabled while still allowing non-udev firmware loading, as done by the dell-rbu driver, to continue working. This is achieved by only using the fallback mechanism when the uevent is suppressed. The patch renames the user-selectable Kconfig from FW_LOADER_USER_HELPER to FW_LOADER_USER_HELPER_FALLBACK, and the former is reverse-selected by the latter or the drivers that need userhelper like dell-rbu. Also, the "default y" is removed together with this change, since it's been deprecated in udev upstream, thus rather better to disable it nowadays. Tested with FW_LOADER_USER_HELPER=n LATTICE_ECP3_CONFIG=y DELL_RBU=y and udev without the firmware loading support, but I don't have the hardware to test the lattice/dell drivers, so additional testing would be appreciated. Reviewed-by: Tom Gundersen <teg@jklm.no> Cc: Ming Lei <ming.lei@canonical.com> Cc: Abhay Salunke <Abhay_Salunke@dell.com> Cc: Stefan Roese <sr@denx.de> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Kay Sievers <kay@vrfy.org> Tested-by: Balaji Singh <B_B_Singh@DELL.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-06-04 22:48:15 +07:00
(uevent ? FW_OPT_UEVENT : FW_OPT_USERHELPER);
if (!try_module_get(module)) {
firmware: fix possible use after free on name on asynchronous request Asynchronous firmware loading copies the pointer to the name passed as an argument only to be scheduled later and used. This behaviour works well for synchronous calling but in asynchronous mode there's a chance the caller could immediately free the passed string after making the asynchronous call. This could trigger a use after free having the kernel look on disk for arbitrary file names. In order to force-test the issue you can use a test-driver designed to illustrate this issue on github [0], use the next-20150505-fix-use-after-free branch. With this patch applied you get: [ 283.512445] firmware name: test_module_stuff.bin [ 287.514020] firmware name: test_module_stuff.bin [ 287.532489] firmware found Without this patch applied you can end up with something such as: [ 135.624216] firmware name: \xffffff80BJ [ 135.624249] platform fake-dev.0: Direct firmware load for \xffffff80Bi failed with error -2 [ 135.624252] No firmware found [ 135.624252] firmware found Unfortunatley in the worst and most common case however you can typically crash your system with a page fault by trying to free something which you cannot, and/or a NULL pointer dereference [1]. The fix and issue using schedule_work() for asynchronous runs is generalized in the following SmPL grammar patch, when applied to next-20150505 only the firmware_class code is affected. This grammar patch can and should further be generalized to vet for for other kernel asynchronous mechanisms. @ calls_schedule_work @ type T; T *priv_work; identifier func, work_func; identifier work; identifier priv_name, name; expression gfp; @@ func(..., const char *name, ...) { ... priv_work = kzalloc(sizeof(T), gfp); ... - priv_work->priv_name = name; + priv_work->priv_name = kstrdup_const(name, gfp); ... (... when any if (...) { ... + kfree_const(priv_work->priv_name); kfree(priv_work); ... } ) ... when any INIT_WORK(&priv_work->work, work_func); ... schedule_work(&priv_work->work); ... } @ the_work_func depends on calls_schedule_work @ type calls_schedule_work.T; T *priv_work; identifier calls_schedule_work.work_func; identifier calls_schedule_work.priv_name; identifier calls_schedule_work.work; identifier some_work; @@ work_func(...) { ... priv_work = container_of(some_work, T, work); ... + kfree_const(priv_work->priv_name); kfree(priv_work); ... } [0] https://github.com/mcgrof/fake-firmware-test.git [1] The following kernel ring buffer splat: firmware name: test_module_stuff.bin firmware name: firmware found general protection fault: 0000 [#1] SMP Modules linked in: test(O) <...etc-it-does-not-matter> drm sr_mod cdrom xhci_pci xhci_hcd rtsx_pci mfd_core video button sg CPU: 3 PID: 87 Comm: kworker/3:2 Tainted: G O 4.0.0-00010-g22b5bb0-dirty #176 Hardware name: LENOVO 20AW000LUS/20AW000LUS, BIOS GLET43WW (1.18 ) 12/04/2013 Workqueue: events request_firmware_work_func task: ffff8800c7f8e290 ti: ffff8800c7f94000 task.ti: ffff8800c7f94000 RIP: 0010:[<ffffffff814a586c>] [<ffffffff814a586c>] fw_free_buf+0xc/0x40 RSP: 0000:ffff8800c7f97d78 EFLAGS: 00010286 RAX: ffffffff81ae3700 RBX: ffffffff816d1181 RCX: 0000000000000006 RDX: 0001ee850ff68500 RSI: 0000000000000246 RDI: c35d5f415e415d41 RBP: ffff8800c7f97d88 R08: 000000000000000a R09: 0000000000000000 R10: 0000000000000358 R11: ffff8800c7f97a7e R12: ffff8800c7ec1e80 R13: ffff88021e2d4cc0 R14: ffff88021e2dff00 R15: 00000000000000c0 FS: 0000000000000000(0000) GS:ffff88021e2c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000034b8cd8 CR3: 000000021073c000 CR4: 00000000001407e0 Stack: ffffffff816d1181 ffff8800c7ec1e80 ffff8800c7f97da8 ffffffff814a58f8 000000000000000a ffffffff816d1181 ffff8800c7f97dc8 ffffffffa047002c ffff88021e2dff00 ffff8802116ac1c0 ffff8800c7f97df8 ffffffff814a65fe Call Trace: [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a58f8>] release_firmware+0x58/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffffa047002c>] test_mod_cb+0x2c/0x43 [test] [<ffffffff814a65fe>] request_firmware_work_func+0x5e/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff8108d23a>] process_one_work+0x14a/0x3f0 [<ffffffff8108d911>] worker_thread+0x121/0x460 [<ffffffff8108d7f0>] ? rescuer_thread+0x310/0x310 [<ffffffff810928f9>] kthread+0xc9/0xe0 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 [<ffffffff816d52d8>] ret_from_fork+0x58/0x90 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 Code: c7 c6 dd ad a3 81 48 c7 c7 20 97 ce 81 31 c0 e8 0b b2 ed ff e9 78 ff ff ff 66 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 54 53 <4c> 8b 67 38 48 89 fb 4c 89 e7 e8 85 f7 22 00 f0 83 2b 01 74 0f RIP [<ffffffff814a586c>] fw_free_buf+0xc/0x40 RSP <ffff8800c7f97d78> ---[ end trace 4e62c56a58d0eac1 ]--- BUG: unable to handle kernel paging request at ffffffffffffffd8 IP: [<ffffffff81093ee0>] kthread_data+0x10/0x20 PGD 1c13067 PUD 1c15067 PMD 0 Oops: 0000 [#2] SMP Modules linked in: test(O) <...etc-it-does-not-matter> drm sr_mod cdrom xhci_pci xhci_hcd rtsx_pci mfd_core video button sg CPU: 3 PID: 87 Comm: kworker/3:2 Tainted: G D O 4.0.0-00010-g22b5bb0-dirty #176 Hardware name: LENOVO 20AW000LUS/20AW000LUS, BIOS GLET43WW (1.18 ) 12/04/2013 task: ffff8800c7f8e290 ti: ffff8800c7f94000 task.ti: ffff8800c7f94000 RIP: 0010:[<ffffffff81092ee0>] [<ffffffff81092ee0>] kthread_data+0x10/0x20 RSP: 0018:ffff8800c7f97b18 EFLAGS: 00010096 RAX: 0000000000000000 RBX: 0000000000000003 RCX: 000000000000000d RDX: 0000000000000003 RSI: 0000000000000003 RDI: ffff8800c7f8e290 RBP: ffff8800c7f97b18 R08: 000000000000bc00 R09: 0000000000007e76 R10: 0000000000000001 R11: 000000000000002f R12: ffff8800c7f8e290 R13: 00000000000154c0 R14: 0000000000000003 R15: 0000000000000000 FS: 0000000000000000(0000) GS:ffff88021e2c0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000028 CR3: 0000000210675000 CR4: 00000000001407e0 Stack: ffff8800c7f97b38 ffffffff8108dcd5 ffff8800c7f97b38 ffff88021e2d54c0 ffff8800c7f97b88 ffffffff816d1500 ffff880213d42368 ffff8800c7f8e290 ffff8800c7f97b88 ffff8800c7f97fd8 ffff8800c7f8e710 0000000000000246 Call Trace: [<ffffffff8108dcd5>] wq_worker_sleeping+0x15/0xa0 [<ffffffff816d1500>] __schedule+0x6e0/0x940 [<ffffffff816d1797>] schedule+0x37/0x90 [<ffffffff810779bc>] do_exit+0x6bc/0xb40 [<ffffffff8101898f>] oops_end+0x9f/0xe0 [<ffffffff81018efb>] die+0x4b/0x70 [<ffffffff81015622>] do_general_protection+0xe2/0x170 [<ffffffff816d74e8>] general_protection+0x28/0x30 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a586c>] ? fw_free_buf+0xc/0x40 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff814a58f8>] release_firmware+0x58/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffffa047002c>] test_mod_cb+0x2c/0x43 [test] [<ffffffff814a65fe>] request_firmware_work_func+0x5e/0x80 [<ffffffff816d1181>] ? __schedule+0x361/0x940 [<ffffffff8108d23a>] process_one_work+0x14a/0x3f0 [<ffffffff8108d911>] worker_thread+0x121/0x460 [<ffffffff8108d7f0>] ? rescuer_thread+0x310/0x310 [<ffffffff810928f9>] kthread+0xc9/0xe0 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 [<ffffffff816d52d8>] ret_from_fork+0x58/0x90 [<ffffffff81092830>] ? kthread_create_on_node+0x180/0x180 Code: 00 48 89 e5 5d 48 8b 40 c8 48 c1 e8 02 83 e0 01 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 8b 87 30 05 00 00 55 48 89 e5 <48> 8b 40 d8 5d c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 RIP [<ffffffff81092ee0>] kthread_data+0x10/0x20 RSP <ffff8800c7f97b18> CR2: ffffffffffffffd8 ---[ end trace 4e62c56a58d0eac2 ]--- Fixing recursive fault but reboot is needed! Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Generated-by: Coccinelle SmPL Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:42 +07:00
kfree_const(fw_work->name);
kfree(fw_work);
return -EFAULT;
}
get_device(fw_work->device);
INIT_WORK(&fw_work->work, request_firmware_work_func);
schedule_work(&fw_work->work);
return 0;
}
EXPORT_SYMBOL(request_firmware_nowait);
#ifdef CONFIG_PM_SLEEP
static ASYNC_DOMAIN_EXCLUSIVE(fw_cache_domain);
/**
* cache_firmware - cache one firmware image in kernel memory space
* @fw_name: the firmware image name
*
* Cache firmware in kernel memory so that drivers can use it when
* system isn't ready for them to request firmware image from userspace.
* Once it returns successfully, driver can use request_firmware or its
* nowait version to get the cached firmware without any interacting
* with userspace
*
* Return 0 if the firmware image has been cached successfully
* Return !0 otherwise
*
*/
static int cache_firmware(const char *fw_name)
{
int ret;
const struct firmware *fw;
pr_debug("%s: %s\n", __func__, fw_name);
ret = request_firmware(&fw, fw_name, NULL);
if (!ret)
kfree(fw);
pr_debug("%s: %s ret=%d\n", __func__, fw_name, ret);
return ret;
}
static struct firmware_buf *fw_lookup_buf(const char *fw_name)
{
struct firmware_buf *tmp;
struct firmware_cache *fwc = &fw_cache;
spin_lock(&fwc->lock);
tmp = __fw_lookup_buf(fw_name);
spin_unlock(&fwc->lock);
return tmp;
}
/**
* uncache_firmware - remove one cached firmware image
* @fw_name: the firmware image name
*
* Uncache one firmware image which has been cached successfully
* before.
*
* Return 0 if the firmware cache has been removed successfully
* Return !0 otherwise
*
*/
static int uncache_firmware(const char *fw_name)
{
struct firmware_buf *buf;
struct firmware fw;
pr_debug("%s: %s\n", __func__, fw_name);
firmware: support loading into a pre-allocated buffer Some systems are memory constrained but they need to load very large firmwares. The firmware subsystem allows drivers to request this firmware be loaded from the filesystem, but this requires that the entire firmware be loaded into kernel memory first before it's provided to the driver. This can lead to a situation where we map the firmware twice, once to load the firmware into kernel memory and once to copy the firmware into the final resting place. This creates needless memory pressure and delays loading because we have to copy from kernel memory to somewhere else. Let's add a request_firmware_into_buf() API that allows drivers to request firmware be loaded directly into a pre-allocated buffer. This skips the intermediate step of allocating a buffer in kernel memory to hold the firmware image while it's read from the filesystem. It also requires that drivers know how much memory they'll require before requesting the firmware and negates any benefits of firmware caching because the firmware layer doesn't manage the buffer lifetime. For a 16MB buffer, about half the time is spent performing a memcpy from the buffer to the final resting place. I see loading times go from 0.081171 seconds to 0.047696 seconds after applying this patch. Plus the vmalloc pressure is reduced. This is based on a patch from Vikram Mulukutla on codeaurora.org: https://www.codeaurora.org/cgit/quic/la/kernel/msm-3.18/commit/drivers/base/firmware_class.c?h=rel/msm-3.18&id=0a328c5f6cd999f5c591f172216835636f39bcb5 Link: http://lkml.kernel.org/r/20160607164741.31849-4-stephen.boyd@linaro.org Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: Vikram Mulukutla <markivx@codeaurora.org> Cc: Mark Brown <broonie@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-08-03 04:04:28 +07:00
if (fw_get_builtin_firmware(&fw, fw_name, NULL, 0))
return 0;
buf = fw_lookup_buf(fw_name);
if (buf) {
fw_free_buf(buf);
return 0;
}
return -EINVAL;
}
static struct fw_cache_entry *alloc_fw_cache_entry(const char *name)
{
struct fw_cache_entry *fce;
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
fce = kzalloc(sizeof(*fce), GFP_ATOMIC);
if (!fce)
goto exit;
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
fce->name = kstrdup_const(name, GFP_ATOMIC);
if (!fce->name) {
kfree(fce);
fce = NULL;
goto exit;
}
exit:
return fce;
}
static int __fw_entry_found(const char *name)
{
struct firmware_cache *fwc = &fw_cache;
struct fw_cache_entry *fce;
list_for_each_entry(fce, &fwc->fw_names, list) {
if (!strcmp(fce->name, name))
return 1;
}
return 0;
}
static int fw_cache_piggyback_on_request(const char *name)
{
struct firmware_cache *fwc = &fw_cache;
struct fw_cache_entry *fce;
int ret = 0;
spin_lock(&fwc->name_lock);
if (__fw_entry_found(name))
goto found;
fce = alloc_fw_cache_entry(name);
if (fce) {
ret = 1;
list_add(&fce->list, &fwc->fw_names);
pr_debug("%s: fw: %s\n", __func__, name);
}
found:
spin_unlock(&fwc->name_lock);
return ret;
}
static void free_fw_cache_entry(struct fw_cache_entry *fce)
{
firmware: use const for remaining firmware names We currently use flexible arrays with a char at the end for the remaining internal firmware name uses. There are two limitations with the way we use this. Since we're using a flexible array for a string on the struct if we wanted to use two strings it means we'd have a disjoint means of handling the strings, one using the flexible array, and another a char * pointer. We're also currently not using 'const' for the string. We wish to later extend some firmware data structures with other string/char pointers, but we also want to be very pedantic about const usage. Since we're going to change things to use 'const' we might as well also address unified way to use multiple strings on the structs. Replace the flexible array practice for strings with kstrdup_const() and kfree_const(), this will avoid allocations when the vmlinux .rodata is used, and just allocate a new proper string for us when needed. This also means we can simplify the struct allocations by removing the string length from the allocation size computation, which would otherwise get even more complicated when supporting multiple strings. Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: David Howells <dhowells@redhat.com> Cc: Ming Lei <ming.lei@canonical.com> Cc: Seth Forshee <seth.forshee@canonical.com> Cc: Kyle McMartin <kyle@kernel.org> Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2015-05-13 04:49:43 +07:00
kfree_const(fce->name);
kfree(fce);
}
static void __async_dev_cache_fw_image(void *fw_entry,
async_cookie_t cookie)
{
struct fw_cache_entry *fce = fw_entry;
struct firmware_cache *fwc = &fw_cache;
int ret;
ret = cache_firmware(fce->name);
if (ret) {
spin_lock(&fwc->name_lock);
list_del(&fce->list);
spin_unlock(&fwc->name_lock);
free_fw_cache_entry(fce);
}
}
/* called with dev->devres_lock held */
static void dev_create_fw_entry(struct device *dev, void *res,
void *data)
{
struct fw_name_devm *fwn = res;
const char *fw_name = fwn->name;
struct list_head *head = data;
struct fw_cache_entry *fce;
fce = alloc_fw_cache_entry(fw_name);
if (fce)
list_add(&fce->list, head);
}
static int devm_name_match(struct device *dev, void *res,
void *match_data)
{
struct fw_name_devm *fwn = res;
return (fwn->magic == (unsigned long)match_data);
}
static void dev_cache_fw_image(struct device *dev, void *data)
{
LIST_HEAD(todo);
struct fw_cache_entry *fce;
struct fw_cache_entry *fce_next;
struct firmware_cache *fwc = &fw_cache;
devres_for_each_res(dev, fw_name_devm_release,
devm_name_match, &fw_cache,
dev_create_fw_entry, &todo);
list_for_each_entry_safe(fce, fce_next, &todo, list) {
list_del(&fce->list);
spin_lock(&fwc->name_lock);
/* only one cache entry for one firmware */
if (!__fw_entry_found(fce->name)) {
list_add(&fce->list, &fwc->fw_names);
} else {
free_fw_cache_entry(fce);
fce = NULL;
}
spin_unlock(&fwc->name_lock);
if (fce)
async_schedule_domain(__async_dev_cache_fw_image,
(void *)fce,
&fw_cache_domain);
}
}
static void __device_uncache_fw_images(void)
{
struct firmware_cache *fwc = &fw_cache;
struct fw_cache_entry *fce;
spin_lock(&fwc->name_lock);
while (!list_empty(&fwc->fw_names)) {
fce = list_entry(fwc->fw_names.next,
struct fw_cache_entry, list);
list_del(&fce->list);
spin_unlock(&fwc->name_lock);
uncache_firmware(fce->name);
free_fw_cache_entry(fce);
spin_lock(&fwc->name_lock);
}
spin_unlock(&fwc->name_lock);
}
/**
* device_cache_fw_images - cache devices' firmware
*
* If one device called request_firmware or its nowait version
* successfully before, the firmware names are recored into the
* device's devres link list, so device_cache_fw_images can call
* cache_firmware() to cache these firmwares for the device,
* then the device driver can load its firmwares easily at
* time when system is not ready to complete loading firmware.
*/
static void device_cache_fw_images(void)
{
struct firmware_cache *fwc = &fw_cache;
int old_timeout;
DEFINE_WAIT(wait);
pr_debug("%s\n", __func__);
/* cancel uncache work */
cancel_delayed_work_sync(&fwc->work);
/*
* use small loading timeout for caching devices' firmware
* because all these firmware images have been loaded
* successfully at lease once, also system is ready for
* completing firmware loading now. The maximum size of
* firmware in current distributions is about 2M bytes,
* so 10 secs should be enough.
*/
old_timeout = loading_timeout;
loading_timeout = 10;
mutex_lock(&fw_lock);
fwc->state = FW_LOADER_START_CACHE;
dpm_for_each_dev(NULL, dev_cache_fw_image);
mutex_unlock(&fw_lock);
/* wait for completion of caching firmware for all devices */
async_synchronize_full_domain(&fw_cache_domain);
loading_timeout = old_timeout;
}
/**
* device_uncache_fw_images - uncache devices' firmware
*
* uncache all firmwares which have been cached successfully
* by device_uncache_fw_images earlier
*/
static void device_uncache_fw_images(void)
{
pr_debug("%s\n", __func__);
__device_uncache_fw_images();
}
static void device_uncache_fw_images_work(struct work_struct *work)
{
device_uncache_fw_images();
}
/**
* device_uncache_fw_images_delay - uncache devices firmwares
* @delay: number of milliseconds to delay uncache device firmwares
*
* uncache all devices's firmwares which has been cached successfully
* by device_cache_fw_images after @delay milliseconds.
*/
static void device_uncache_fw_images_delay(unsigned long delay)
{
queue_delayed_work(system_power_efficient_wq, &fw_cache.work,
msecs_to_jiffies(delay));
}
/**
* fw_pm_notify - notifier for suspend/resume
* @notify_block: unused
* @mode: mode we are switching to
* @unused: unused
*
* Used to modify the firmware_class state as we move in between states.
* The firmware_class implements a firmware cache to enable device driver
* to fetch firmware upon resume before the root filesystem is ready. We
* disable API calls which do not use the built-in firmware or the firmware
* cache when we know these calls will not work.
*
* The inner logic behind all this is a bit complex so it is worth summarizing
* the kernel's own suspend/resume process with context and focus on how this
* can impact the firmware API.
*
* First a review on how we go to suspend::
*
* pm_suspend() --> enter_state() -->
* sys_sync()
* suspend_prepare() -->
* __pm_notifier_call_chain(PM_SUSPEND_PREPARE, ...);
* suspend_freeze_processes() -->
* freeze_processes() -->
* __usermodehelper_set_disable_depth(UMH_DISABLED);
* freeze all tasks ...
* freeze_kernel_threads()
* suspend_devices_and_enter() -->
* dpm_suspend_start() -->
* dpm_prepare()
* dpm_suspend()
* suspend_enter() -->
* platform_suspend_prepare()
* dpm_suspend_late()
* freeze_enter()
* syscore_suspend()
*
* When we resume we bail out of a loop from suspend_devices_and_enter() and
* unwind back out to the caller enter_state() where we were before as follows::
*
* enter_state() -->
* suspend_devices_and_enter() --> (bail from loop)
* dpm_resume_end() -->
* dpm_resume()
* dpm_complete()
* suspend_finish() -->
* suspend_thaw_processes() -->
* thaw_processes() -->
* __usermodehelper_set_disable_depth(UMH_FREEZING);
* thaw_workqueues();
* thaw all processes ...
* usermodehelper_enable();
* pm_notifier_call_chain(PM_POST_SUSPEND);
*
* fw_pm_notify() works through pm_notifier_call_chain().
*/
static int fw_pm_notify(struct notifier_block *notify_block,
unsigned long mode, void *unused)
{
switch (mode) {
case PM_HIBERNATION_PREPARE:
case PM_SUSPEND_PREPARE:
case PM_RESTORE_PREPARE:
/*
* kill pending fallback requests with a custom fallback
* to avoid stalling suspend.
*/
kill_pending_fw_fallback_reqs(true);
device_cache_fw_images();
disable_firmware();
break;
case PM_POST_SUSPEND:
case PM_POST_HIBERNATION:
case PM_POST_RESTORE:
/*
* In case that system sleep failed and syscore_suspend is
* not called.
*/
mutex_lock(&fw_lock);
fw_cache.state = FW_LOADER_NO_CACHE;
mutex_unlock(&fw_lock);
enable_firmware();
device_uncache_fw_images_delay(10 * MSEC_PER_SEC);
break;
}
return 0;
}
/* stop caching firmware once syscore_suspend is reached */
static int fw_suspend(void)
{
fw_cache.state = FW_LOADER_NO_CACHE;
return 0;
}
static struct syscore_ops fw_syscore_ops = {
.suspend = fw_suspend,
};
#else
static int fw_cache_piggyback_on_request(const char *name)
{
return 0;
}
#endif
static void __init fw_cache_init(void)
{
spin_lock_init(&fw_cache.lock);
INIT_LIST_HEAD(&fw_cache.head);
fw_cache.state = FW_LOADER_NO_CACHE;
#ifdef CONFIG_PM_SLEEP
spin_lock_init(&fw_cache.name_lock);
INIT_LIST_HEAD(&fw_cache.fw_names);
INIT_DELAYED_WORK(&fw_cache.work,
device_uncache_fw_images_work);
fw_cache.pm_notify.notifier_call = fw_pm_notify;
register_pm_notifier(&fw_cache.pm_notify);
register_syscore_ops(&fw_syscore_ops);
#endif
}
static int fw_shutdown_notify(struct notifier_block *unused1,
unsigned long unused2, void *unused3)
{
disable_firmware();
/*
* Kill all pending fallback requests to avoid both stalling shutdown,
* and avoid a deadlock with the usermode_lock.
*/
kill_pending_fw_fallback_reqs(false);
return NOTIFY_DONE;
}
static struct notifier_block fw_shutdown_nb = {
.notifier_call = fw_shutdown_notify,
};
static int __init firmware_class_init(void)
{
enable_firmware();
fw_cache_init();
register_reboot_notifier(&fw_shutdown_nb);
#ifdef CONFIG_FW_LOADER_USER_HELPER
return class_register(&firmware_class);
#else
return 0;
#endif
}
static void __exit firmware_class_exit(void)
{
disable_firmware();
#ifdef CONFIG_PM_SLEEP
unregister_syscore_ops(&fw_syscore_ops);
unregister_pm_notifier(&fw_cache.pm_notify);
#endif
unregister_reboot_notifier(&fw_shutdown_nb);
#ifdef CONFIG_FW_LOADER_USER_HELPER
class_unregister(&firmware_class);
#endif
}
fs_initcall(firmware_class_init);
module_exit(firmware_class_exit);