| Message ID | 20220426221859.v2.1.I650b809482e1af8d0156ed88b5dc2677a0711d46@changeid (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | b72a4aff947ba807177bdabb43debaf2c66bee05 |
| Delegated to: | Kalle Valo |
| Headers | show |
| Series | [v2] ath10k: skip ath10k_halt during suspend for driver state RESTARTING | expand |
On Tue, Apr 26, 2022 at 3:20 PM Abhishek Kumar <kuabhs@chromium.org> wrote: > > Double free crash is observed when FW recovery(caused by wmi > timeout/crash) is followed by immediate suspend event. The FW recovery > is triggered by ath10k_core_restart() which calls driver clean up via > ath10k_halt(). When the suspend event occurs between the FW recovery, > the restart worker thread is put into frozen state until suspend completes. > The suspend event triggers ath10k_stop() which again triggers ath10k_halt() > The double invocation of ath10k_halt() causes ath10k_htt_rx_free() to be > called twice(Note: ath10k_htt_rx_alloc was not called by restart worker > thread because of its frozen state), causing the crash. ... > Tested-on: QCA6174 hw3.2 PCI WLAN.RM.4.4.1-00288-QCARMSWPZ-1 > Co-developed-by: Wen Gong <quic_wgong@quicinc.com> > Signed-off-by: Wen Gong <quic_wgong@quicinc.com> > Signed-off-by: Abhishek Kumar <kuabhs@chromium.org> > --- > > Changes in v2: > - Fixed typo, replaced ath11k by ath10k in the comments. > - Adjusted the position of my S-O-B tag. > - Added the Tested-on tag. You could have retained my: Reviewed-by: Brian Norris <briannorris@chromium.org> but no worries; it's just a few characters ;)
On Tue, Apr 26, 2022 at 3:34 PM Brian Norris <briannorris@chromium.org> wrote: > > On Tue, Apr 26, 2022 at 3:20 PM Abhishek Kumar <kuabhs@chromium.org> wrote: > > > > Double free crash is observed when FW recovery(caused by wmi > > timeout/crash) is followed by immediate suspend event. The FW recovery > > is triggered by ath10k_core_restart() which calls driver clean up via > > ath10k_halt(). When the suspend event occurs between the FW recovery, > > the restart worker thread is put into frozen state until suspend completes. > > The suspend event triggers ath10k_stop() which again triggers ath10k_halt() > > The double invocation of ath10k_halt() causes ath10k_htt_rx_free() to be > > called twice(Note: ath10k_htt_rx_alloc was not called by restart worker > > thread because of its frozen state), causing the crash. > ... > > Tested-on: QCA6174 hw3.2 PCI WLAN.RM.4.4.1-00288-QCARMSWPZ-1 > > Co-developed-by: Wen Gong <quic_wgong@quicinc.com> > > Signed-off-by: Wen Gong <quic_wgong@quicinc.com> > > Signed-off-by: Abhishek Kumar <kuabhs@chromium.org> > > --- > > > > Changes in v2: > > - Fixed typo, replaced ath11k by ath10k in the comments. > > - Adjusted the position of my S-O-B tag. > > - Added the Tested-on tag. > > You could have retained my: > > Reviewed-by: Brian Norris <briannorris@chromium.org> > > but no worries; it's just a few characters ;) Oh! sorry about that, I was under the impression that if the next iteration is posted, then I cannot just add the Reviewed-by tag provided in the previous iteration by myself.
Abhishek Kumar <kuabhs@chromium.org> writes: > Double free crash is observed when FW recovery(caused by wmi > timeout/crash) is followed by immediate suspend event. The FW recovery > is triggered by ath10k_core_restart() which calls driver clean up via > ath10k_halt(). When the suspend event occurs between the FW recovery, > the restart worker thread is put into frozen state until suspend completes. > The suspend event triggers ath10k_stop() which again triggers ath10k_halt() > The double invocation of ath10k_halt() causes ath10k_htt_rx_free() to be > called twice(Note: ath10k_htt_rx_alloc was not called by restart worker > thread because of its frozen state), causing the crash. > > To fix this, during the suspend flow, skip call to ath10k_halt() in > ath10k_stop() when the current driver state is ATH10K_STATE_RESTARTING. > Also, for driver state ATH10K_STATE_RESTARTING, call > ath10k_wait_for_suspend() in ath10k_stop(). This is because call to > ath10k_wait_for_suspend() is skipped later in > [ath10k_halt() > ath10k_core_stop()] for the driver state > ATH10K_STATE_RESTARTING. > > The frozen restart worker thread will be cancelled during resume when the > device comes out of suspend. > > Below is the crash stack for reference: > > [ 428.469167] ------------[ cut here ]------------ > [ 428.469180] kernel BUG at mm/slub.c:4150! > [ 428.469193] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI > [ 428.469219] Workqueue: events_unbound async_run_entry_fn > [ 428.469230] RIP: 0010:kfree+0x319/0x31b > [ 428.469241] RSP: 0018:ffffa1fac015fc30 EFLAGS: 00010246 > [ 428.469247] RAX: ffffedb10419d108 RBX: ffff8c05262b0000 > [ 428.469252] RDX: ffff8c04a8c07000 RSI: 0000000000000000 > [ 428.469256] RBP: ffffa1fac015fc78 R08: 0000000000000000 > [ 428.469276] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > [ 428.469285] Call Trace: > [ 428.469295] ? dma_free_attrs+0x5f/0x7d > [ 428.469320] ath10k_core_stop+0x5b/0x6f > [ 428.469336] ath10k_halt+0x126/0x177 > [ 428.469352] ath10k_stop+0x41/0x7e > [ 428.469387] drv_stop+0x88/0x10e > [ 428.469410] __ieee80211_suspend+0x297/0x411 > [ 428.469441] rdev_suspend+0x6e/0xd0 > [ 428.469462] wiphy_suspend+0xb1/0x105 > [ 428.469483] ? name_show+0x2d/0x2d > [ 428.469490] dpm_run_callback+0x8c/0x126 > [ 428.469511] ? name_show+0x2d/0x2d > [ 428.469517] __device_suspend+0x2e7/0x41b > [ 428.469523] async_suspend+0x1f/0x93 > [ 428.469529] async_run_entry_fn+0x3d/0xd1 > [ 428.469535] process_one_work+0x1b1/0x329 > [ 428.469541] worker_thread+0x213/0x372 > [ 428.469547] kthread+0x150/0x15f > [ 428.469552] ? pr_cont_work+0x58/0x58 > [ 428.469558] ? kthread_blkcg+0x31/0x31 > > Tested-on: QCA6174 hw3.2 PCI WLAN.RM.4.4.1-00288-QCARMSWPZ-1 > Co-developed-by: Wen Gong <quic_wgong@quicinc.com> > Signed-off-by: Wen Gong <quic_wgong@quicinc.com> > Signed-off-by: Abhishek Kumar <kuabhs@chromium.org> > --- > > Changes in v2: > - Fixed typo, replaced ath11k by ath10k in the comments. > - Adjusted the position of my S-O-B tag. > - Added the Tested-on tag. > > drivers/net/wireless/ath/ath10k/mac.c | 18 ++++++++++++++++-- > 1 file changed, 16 insertions(+), 2 deletions(-) > > diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c > index d804e19a742a..e9c1f11fef0a 100644 > --- a/drivers/net/wireless/ath/ath10k/mac.c > +++ b/drivers/net/wireless/ath/ath10k/mac.c > @@ -5345,8 +5345,22 @@ static void ath10k_stop(struct ieee80211_hw *hw) > > mutex_lock(&ar->conf_mutex); > if (ar->state != ATH10K_STATE_OFF) { > - if (!ar->hw_rfkill_on) > - ath10k_halt(ar); > + if (!ar->hw_rfkill_on) { > + /* If the current driver state is RESTARTING but not yet > + * fully RESTARTED because of incoming suspend event, > + * then ath10k_halt is already called via > + * ath10k_core_restart and should not be called here. > + */ > + if (ar->state != ATH10K_STATE_RESTARTING) > + ath10k_halt(ar); > + else > + /* Suspending here, because when in RESTARTING > + * state, ath10k_core_stop skips > + * ath10k_wait_for_suspend. > + */ > + ath10k_wait_for_suspend(ar, > + WMI_PDEV_SUSPEND_AND_DISABLE_INTR); > + } I'm nitpicking but I prefer to use parenthesis with function names, so I changed the comments. Also there was one ath10k-check warning: drivers/net/wireless/ath/ath10k/mac.c:5360: line length of 91 exceeds 90 columns In the pending branch I changed it to: if (ar->state != ATH10K_STATE_RESTARTING) { ath10k_halt(ar); } else { /* Suspending here, because when in RESTARTING * state, ath10k_core_stop() skips * ath10k_wait_for_suspend(). */ opt = WMI_PDEV_SUSPEND_AND_DISABLE_INTR; ath10k_wait_for_suspend(ar, opt); } Not really pretty but I prefer to keep ath10k warning free.
On Tue, Apr 26, 2022 at 10:20 PM Abhishek Kumar <kuabhs@chromium.org> wrote: > On Tue, Apr 26, 2022 at 3:34 PM Brian Norris <briannorris@chromium.org> wrote: > > You could have retained my: > > > > Reviewed-by: Brian Norris <briannorris@chromium.org> > > > > but no worries; it's just a few characters ;) > Oh! sorry about that, I was under the impression that if the next > iteration is posted, then I cannot just add the Reviewed-by tag > provided in the previous iteration by myself. You certainly *can* add it, but it's a judgment call on whether you should. In this case, you were only making cosmetic changes (commit message and comments, requested by reviewers) between versions, so IMO that's totally fine. If there are more substantial changes between versions, it may be better to skip it, and let the reviewer re-review. Sometimes a reviewer might even explicitly say, "with changes X and Y, consider this Reviewed-by: foo <foo@bar.tld>", which is a pretty strong indication you can add it. Anyway, it's probably better to omit a review line that should have been included, rather than include one that should not have been (e.g., because undesirable changes were made). So again, no worries from me! Brian
Abhishek Kumar <kuabhs@chromium.org> wrote: > Double free crash is observed when FW recovery(caused by wmi > timeout/crash) is followed by immediate suspend event. The FW recovery > is triggered by ath10k_core_restart() which calls driver clean up via > ath10k_halt(). When the suspend event occurs between the FW recovery, > the restart worker thread is put into frozen state until suspend completes. > The suspend event triggers ath10k_stop() which again triggers ath10k_halt() > The double invocation of ath10k_halt() causes ath10k_htt_rx_free() to be > called twice(Note: ath10k_htt_rx_alloc was not called by restart worker > thread because of its frozen state), causing the crash. > > To fix this, during the suspend flow, skip call to ath10k_halt() in > ath10k_stop() when the current driver state is ATH10K_STATE_RESTARTING. > Also, for driver state ATH10K_STATE_RESTARTING, call > ath10k_wait_for_suspend() in ath10k_stop(). This is because call to > ath10k_wait_for_suspend() is skipped later in > [ath10k_halt() > ath10k_core_stop()] for the driver state > ATH10K_STATE_RESTARTING. > > The frozen restart worker thread will be cancelled during resume when the > device comes out of suspend. > > Below is the crash stack for reference: > > [ 428.469167] ------------[ cut here ]------------ > [ 428.469180] kernel BUG at mm/slub.c:4150! > [ 428.469193] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI > [ 428.469219] Workqueue: events_unbound async_run_entry_fn > [ 428.469230] RIP: 0010:kfree+0x319/0x31b > [ 428.469241] RSP: 0018:ffffa1fac015fc30 EFLAGS: 00010246 > [ 428.469247] RAX: ffffedb10419d108 RBX: ffff8c05262b0000 > [ 428.469252] RDX: ffff8c04a8c07000 RSI: 0000000000000000 > [ 428.469256] RBP: ffffa1fac015fc78 R08: 0000000000000000 > [ 428.469276] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > [ 428.469285] Call Trace: > [ 428.469295] ? dma_free_attrs+0x5f/0x7d > [ 428.469320] ath10k_core_stop+0x5b/0x6f > [ 428.469336] ath10k_halt+0x126/0x177 > [ 428.469352] ath10k_stop+0x41/0x7e > [ 428.469387] drv_stop+0x88/0x10e > [ 428.469410] __ieee80211_suspend+0x297/0x411 > [ 428.469441] rdev_suspend+0x6e/0xd0 > [ 428.469462] wiphy_suspend+0xb1/0x105 > [ 428.469483] ? name_show+0x2d/0x2d > [ 428.469490] dpm_run_callback+0x8c/0x126 > [ 428.469511] ? name_show+0x2d/0x2d > [ 428.469517] __device_suspend+0x2e7/0x41b > [ 428.469523] async_suspend+0x1f/0x93 > [ 428.469529] async_run_entry_fn+0x3d/0xd1 > [ 428.469535] process_one_work+0x1b1/0x329 > [ 428.469541] worker_thread+0x213/0x372 > [ 428.469547] kthread+0x150/0x15f > [ 428.469552] ? pr_cont_work+0x58/0x58 > [ 428.469558] ? kthread_blkcg+0x31/0x31 > > Tested-on: QCA6174 hw3.2 PCI WLAN.RM.4.4.1-00288-QCARMSWPZ-1 > Co-developed-by: Wen Gong <quic_wgong@quicinc.com> > Signed-off-by: Wen Gong <quic_wgong@quicinc.com> > Signed-off-by: Abhishek Kumar <kuabhs@chromium.org> > Reviewed-by: Brian Norris <briannorris@chromium.org> > Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com> Patch applied to ath-next branch of ath.git, thanks. b72a4aff947b ath10k: skip ath10k_halt during suspend for driver state RESTARTING
diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c index d804e19a742a..e9c1f11fef0a 100644 --- a/drivers/net/wireless/ath/ath10k/mac.c +++ b/drivers/net/wireless/ath/ath10k/mac.c @@ -5345,8 +5345,22 @@ static void ath10k_stop(struct ieee80211_hw *hw) mutex_lock(&ar->conf_mutex); if (ar->state != ATH10K_STATE_OFF) { - if (!ar->hw_rfkill_on) - ath10k_halt(ar); + if (!ar->hw_rfkill_on) { + /* If the current driver state is RESTARTING but not yet + * fully RESTARTED because of incoming suspend event, + * then ath10k_halt is already called via + * ath10k_core_restart and should not be called here. + */ + if (ar->state != ATH10K_STATE_RESTARTING) + ath10k_halt(ar); + else + /* Suspending here, because when in RESTARTING + * state, ath10k_core_stop skips + * ath10k_wait_for_suspend. + */ + ath10k_wait_for_suspend(ar, + WMI_PDEV_SUSPEND_AND_DISABLE_INTR); + } ar->state = ATH10K_STATE_OFF; } mutex_unlock(&ar->conf_mutex);