kernel.git/drivers/base/firmware_class.c, branch linux-4.12.y

firmware: avoid invalid fallback aborts by using killable wait

2017-08-16T20:46:57Z

commit 260d9f2fc5655a2552701cdd0b909c4466732f19 upstream. Commit 0cb64249ca500 ("firmware_loader: abort request if wait_for_completion is interrupted") added via 4.0 added support to abort the fallback mechanism when a signal was detected and wait_for_completion_interruptible() returned -ERESTARTSYS -- for instance when a user hits CTRL-C. The abort was overly *too* effective. When a child process terminates (successful or not) the signal SIGCHLD can be sent to the parent process which ran the child in the background and later triggered a sync request for firmware through a sysfs interface which relies on the fallback mechanism. This signal in turn can be recieved by the interruptible wait we constructed on firmware_class and detects it as an abort *before* userspace could get a chance to write the firmware. Upon failure -EAGAIN is returned, so userspace is also kept in the dark about exactly what happened. We can reproduce the issue with the fw_fallback.sh selftest: Before this patch: $ sudo tools/testing/selftests/firmware/fw_fallback.sh ... tools/testing/selftests/firmware/fw_fallback.sh: error - sync firmware request cancelled due to SIGCHLD After this patch: $ sudo tools/testing/selftests/firmware/fw_fallback.sh ... tools/testing/selftests/firmware/fw_fallback.sh: SIGCHLD on sync ignored as expected Fix this by making the wait killable -- only killable by SIGKILL (kill -9). We loose the ability to allow userspace to cancel a write with CTRL-C (SIGINT), however its been decided the compromise to require SIGKILL is worth the gains. Chances of this issue occuring are low due to the number of drivers upstream exclusively relying on the fallback mechanism for firmware (2 drivers), however this is observed in the field with custom drivers with sysfs triggers to load firmware. Only distributions relying on the fallback mechanism are impacted as well. An example reported issue was on Android, as follows: 1) Android init (pid=1) fork()s (say pid=42) [this child process is totally unrelated to firmware loading, it could be sleep 2; for all we care ] 2) Android init (pid=1) does a write() on a (driver custom) sysfs file which ends up calling request_firmware() kernel side 3) The firmware loading fallback mechanism is used, the request is sent to userspace and pid 1 waits in the kernel on wait_* 4) before firmware loading completes pid 42 dies (for any reason, even normal termination) 5) Kernel delivers SIGCHLD to pid=1 to tell it a child has died, which causes -ERESTARTSYS to be returned from wait_* 6) The kernel's wait aborts and return -EAGAIN for the request_firmware() caller. Fixes: 0cb64249ca500 ("firmware_loader: abort request if wait_for_completion is interrupted") Suggested-by: "Eric W. Biederman" Suggested-by: Dmitry Torokhov Tested-by: Martin Fuzzey Reported-by: Martin Fuzzey Signed-off-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: fix batched requests - send wake up on failure on direct lookups

2017-08-16T20:46:57Z

commit 90d41e74a9c36a84e2efbd2a5b8d79299feee6fa upstream. Fix batched requests from waiting forever on failure. The firmware API batched requests feature has been broken since the API call request_firmware_direct() was introduced on commit bba3a87e982ad ("firmware: Introduce request_firmware_direct()"), added on v3.14 *iff* the firmware being requested was not present in *certain kernel builds* [0]. When no firmware is found the worker which goes on to finish never informs waiters queued up of this, so any batched request will stall in what seems to be forever (MAX_SCHEDULE_TIMEOUT). Sadly, a reboot will also stall, as the reboot notifier was only designed to kill custom fallback workers. The issue seems to the user as a type of soft lockup, what *actually* happens underneath the hood is a wait call which never completes as we failed to issue a completion on error. For device drivers with optional firmware schemes (ie, Intel iwlwifi, or Netronome -- even though it uses request_firmware() and not request_firmware_direct()), this could mean that when you boot a system with multiple cards the firmware will seem to never load on the system, or that the card is just not responsive even the driver initialization. Due to differences in scheduling possible this should not always trigger -- one would need to to ensure that multiple requests are in place at the right time for this to work, also release_firmware() must not be called prior to any other incoming request. The complexity may not be worth supporting batched requests in the future given the wait mechanism is only used also for the fallback mechanism. We'll keep it for now and just fix it. Its reported that at least with the Intel WiFi cards on one system this issue was creeping up 50% of the boots [0]. Before this commit batched requests testing revealed: ============================================================================ 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 ============================================================================ Ater this commit batched testing results: ============================================================================ 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() OK OK request_firmware_direct() OK OK request_firmware_nowait(uevent=true) OK OK request_firmware_nowait(uevent=false) OK 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() OK OK request_firmware_direct() OK OK request_firmware_nowait(uevent=true) OK OK request_firmware_nowait(uevent=false) OK 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() OK 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 Fixes: bba3a87e982ad ("firmware: Introduce request_firmware_direct()" Reported-by: Nicolas Reported-by: John Ewalt Reported-by: Jakub Kicinski Signed-off-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: fix batched requests - wake all waiters

2017-08-16T20:46:57Z

commit e44565f62a72064e686f7a852137595ec94d78f2 upstream. 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: Ming Lei Fixes: 5b029624948d ("firmware: do not use fw_lock for fw_state protection") Reported-by: Jakub Kicinski Signed-off-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: fix NULL pointer dereference in __fw_load_abort()

2017-01-27T08:19:48Z

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 Reported-and-Tested-by: Patrick Bruenn Reported-by: Chris Wilson CC: [3.10+] Signed-off-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: Correct handling of fw_state_wait() return value

2016-12-08T20:05:54Z

When request_firmware() finds an already open firmware object it will wait for that object to become fully loaded and then check the status. As __fw_state_wait_common() succeeds the timeout value returned will be truncated in _request_firmware_prepare() and interpreted as -EPERM. Prior to "firmware: do not use fw_lock for fw_state protection" the code did test if we where in the "done" state before sleeping, causing this particular code path to succeed, in some cases. As the callers are interested in the result of the wait and not the remaining timeout the return value of __fw_state_wait_common() is changed to signal "done" or "error", which simplifies the logic in _request_firmware_load() as well. Fixes: 5b029624948d ("firmware: do not use fw_lock for fw_state protection") Signed-off-by: Bjorn Andersson Reviewed-by: Daniel Wagner Acked-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: remove warning at documentation generation time

2016-12-01T20:41:13Z

This patch removes following error at for `make htmldocs`. No functional change. ./drivers/base/firmware_class.c:1348: WARNING: Bullet list ends without a blank line; unexpected unindent. Signed-off-by: Silvio Fricke Reviewed-by: Mauro Carvalho Chehab Acked-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: move fw_state_is_done() into UHM section

2016-11-29T20:28:54Z

fw_state_is_done() is only used for UHM so moved into that section. Cc: Ming Lei Signed-off-by: Daniel Wagner Acked-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: do not use fw_lock for fw_state protection

2016-11-29T20:28:54Z

fw_lock is to use to protect 'corner cases' inside firmware_class. It is not exactly clear what those corner cases are nor what it exactly protects. fw_state can be used without needing the fw_lock to protect its state transition and wake ups. fw_state is holds the state in status and the completion is used to wake up all waiters (in this case that is the user land helper so only one). This operation has to be 'atomic' to avoid races. We can do this by using swait which takes care we don't miss any wake up. We use also swait instead of wait because don't need all the additional features wait provides. Note there some more cleanups possible after with this change. For example for !CONFIG_FW_LOADER_USER_HELPER we don't check for the state anymore. Let's to this in the next patch instead mingling to many changes into this one. And yes you get a gcc warning "‘__fw_state_check’ defined but not used [-Wunused-function] code." for the time beeing. Cc: Ming Lei Signed-off-by: Daniel Wagner Acked-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: drop bit ops in favor of simple state machine

2016-11-29T20:28:54Z

We track the state of the firmware loading with bit ops. Since the state machine has only a few states and they are all mutual exclusive there are only a few simple state transition we can model this simplify. UNKNOWN -> LOADING -> DONE | ABORTED Because we don't use any bit ops on fw_state::status anymore we are able to change the data type to enum fw_status and update the function arguments accordingly. READ_ONCE() and WRITE_ONCE() are propably not needed because there are a lot of load and stores around fw_st->status. But let's make it explicit and not be sorry later. Cc: Ming Lei Signed-off-by: Daniel Wagner Acked-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman

firmware: refactor loading status

2016-11-29T20:28:54Z

The firmware loader tracks the current state of the loading process via unsigned long status and a completion in struct firmware_buf. Instead of open code tracking the state, introduce data structure which encapsulate the state tracking and synchronization. While at it also separate UHM states from direct loading states, e.g. the loading_timeout is only defined when CONFIG_FW_LOADER_USER_HELPER. Cc: Ming Lei Signed-off-by: Daniel Wagner Acked-by: Luis R. Rodriguez Signed-off-by: Greg Kroah-Hartman