Re: [PATCH v7 3/5] drm/nouveau: Fix deadlock with fb_helper with async RPM requests

From: Daniel Vetter
Date: Tue Aug 14 2018 - 09:27:36 EST

On Mon, Aug 13, 2018 at 03:16:43PM -0400, Lyude Paul wrote:
> I'm sure I don't need to tell you that fb_helper's locking is a mess.

Uh, loaded statement :-) Especially since we've just reworked all the
fbdev locking to untangle it from drm_modeset locking. What's still a mess
about it? It's documented, and very clearly wrapped around all the
drm_modeset locks. If there's gaps, please bring them up.

> That being said; fb_helper's locking mess can seriously complicate the
> runtime suspend/resume operations of drivers because it can invoke
> atomic commits and connector probing from anywhere that calls
> drm_fb_helper_hotplug_event(). Since most drivers use
> drm_fb_helper_output_poll_changed() as their output_poll_changed
> handler, this can happen in every single context that can fire off a
> hotplug event. An example:
>
> [ 246.669625] INFO: task kworker/4:0:37 blocked for more than 120 seconds.
> [ 246.673398] Not tainted 4.18.0-rc5Lyude-Test+ #2
> [ 246.675271] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> [ 246.676527] kworker/4:0 D 0 37 2 0x80000000
> [ 246.677580] Workqueue: events output_poll_execute [drm_kms_helper]
> [ 246.678704] Call Trace:
> [ 246.679753] __schedule+0x322/0xaf0
> [ 246.680916] schedule+0x33/0x90
> [ 246.681924] schedule_preempt_disabled+0x15/0x20
> [ 246.683023] __mutex_lock+0x569/0x9a0
> [ 246.684035] ? kobject_uevent_env+0x117/0x7b0
> [ 246.685132] ? drm_fb_helper_hotplug_event.part.28+0x20/0xb0 [drm_kms_helper]
> [ 246.686179] mutex_lock_nested+0x1b/0x20
> [ 246.687278] ? mutex_lock_nested+0x1b/0x20
> [ 246.688307] drm_fb_helper_hotplug_event.part.28+0x20/0xb0 [drm_kms_helper]
> [ 246.689420] drm_fb_helper_output_poll_changed+0x23/0x30 [drm_kms_helper]
> [ 246.690462] drm_kms_helper_hotplug_event+0x2a/0x30 [drm_kms_helper]
> [ 246.691570] output_poll_execute+0x198/0x1c0 [drm_kms_helper]
> [ 246.692611] process_one_work+0x231/0x620
> [ 246.693725] worker_thread+0x214/0x3a0
> [ 246.694756] kthread+0x12b/0x150
> [ 246.695856] ? wq_pool_ids_show+0x140/0x140
> [ 246.696888] ? kthread_create_worker_on_cpu+0x70/0x70
> [ 246.697998] ret_from_fork+0x3a/0x50
> [ 246.699034] INFO: task kworker/0:1:60 blocked for more than 120 seconds.
> [ 246.700153] Not tainted 4.18.0-rc5Lyude-Test+ #2
> [ 246.701182] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> [ 246.702278] kworker/0:1 D 0 60 2 0x80000000
> [ 246.703293] Workqueue: pm pm_runtime_work
> [ 246.704393] Call Trace:
> [ 246.705403] __schedule+0x322/0xaf0
> [ 246.706439] ? wait_for_completion+0x104/0x190
> [ 246.707393] schedule+0x33/0x90
> [ 246.708375] schedule_timeout+0x3a5/0x590
> [ 246.709289] ? mark_held_locks+0x58/0x80
> [ 246.710208] ? _raw_spin_unlock_irq+0x2c/0x40
> [ 246.711222] ? wait_for_completion+0x104/0x190
> [ 246.712134] ? trace_hardirqs_on_caller+0xf4/0x190
> [ 246.713094] ? wait_for_completion+0x104/0x190
> [ 246.713964] wait_for_completion+0x12c/0x190
> [ 246.714895] ? wake_up_q+0x80/0x80
> [ 246.715727] ? get_work_pool+0x90/0x90
> [ 246.716649] flush_work+0x1c9/0x280
> [ 246.717483] ? flush_workqueue_prep_pwqs+0x1b0/0x1b0
> [ 246.718442] __cancel_work_timer+0x146/0x1d0
> [ 246.719247] cancel_delayed_work_sync+0x13/0x20
> [ 246.720043] drm_kms_helper_poll_disable+0x1f/0x30 [drm_kms_helper]
> [ 246.721123] nouveau_pmops_runtime_suspend+0x3d/0xb0 [nouveau]
> [ 246.721897] pci_pm_runtime_suspend+0x6b/0x190
> [ 246.722825] ? pci_has_legacy_pm_support+0x70/0x70
> [ 246.723737] __rpm_callback+0x7a/0x1d0
> [ 246.724721] ? pci_has_legacy_pm_support+0x70/0x70
> [ 246.725607] rpm_callback+0x24/0x80
> [ 246.726553] ? pci_has_legacy_pm_support+0x70/0x70
> [ 246.727376] rpm_suspend+0x142/0x6b0
> [ 246.728185] pm_runtime_work+0x97/0xc0
> [ 246.728938] process_one_work+0x231/0x620
> [ 246.729796] worker_thread+0x44/0x3a0
> [ 246.730614] kthread+0x12b/0x150
> [ 246.731395] ? wq_pool_ids_show+0x140/0x140
> [ 246.732202] ? kthread_create_worker_on_cpu+0x70/0x70
> [ 246.732878] ret_from_fork+0x3a/0x50
> [ 246.733768] INFO: task kworker/4:2:422 blocked for more than 120 seconds.
> [ 246.734587] Not tainted 4.18.0-rc5Lyude-Test+ #2
> [ 246.735393] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> [ 246.736113] kworker/4:2 D 0 422 2 0x80000080
> [ 246.736789] Workqueue: events_long drm_dp_mst_link_probe_work [drm_kms_helper]
> [ 246.737665] Call Trace:
> [ 246.738490] __schedule+0x322/0xaf0
> [ 246.739250] schedule+0x33/0x90
> [ 246.739908] rpm_resume+0x19c/0x850
> [ 246.740750] ? finish_wait+0x90/0x90
> [ 246.741541] __pm_runtime_resume+0x4e/0x90
> [ 246.742370] nv50_disp_atomic_commit+0x31/0x210 [nouveau]
> [ 246.743124] drm_atomic_commit+0x4a/0x50 [drm]
> [ 246.743775] restore_fbdev_mode_atomic+0x1c8/0x240 [drm_kms_helper]
> [ 246.744603] restore_fbdev_mode+0x31/0x140 [drm_kms_helper]
> [ 246.745373] drm_fb_helper_restore_fbdev_mode_unlocked+0x54/0xb0 [drm_kms_helper]
> [ 246.746220] drm_fb_helper_set_par+0x2d/0x50 [drm_kms_helper]
> [ 246.746884] drm_fb_helper_hotplug_event.part.28+0x96/0xb0 [drm_kms_helper]
> [ 246.747675] drm_fb_helper_output_poll_changed+0x23/0x30 [drm_kms_helper]
> [ 246.748544] drm_kms_helper_hotplug_event+0x2a/0x30 [drm_kms_helper]
> [ 246.749439] nv50_mstm_hotplug+0x15/0x20 [nouveau]
> [ 246.750111] drm_dp_send_link_address+0x177/0x1c0 [drm_kms_helper]
> [ 246.750764] drm_dp_check_and_send_link_address+0xa8/0xd0 [drm_kms_helper]
> [ 246.751602] drm_dp_mst_link_probe_work+0x51/0x90 [drm_kms_helper]
> [ 246.752314] process_one_work+0x231/0x620
> [ 246.752979] worker_thread+0x44/0x3a0
> [ 246.753838] kthread+0x12b/0x150
> [ 246.754619] ? wq_pool_ids_show+0x140/0x140
> [ 246.755386] ? kthread_create_worker_on_cpu+0x70/0x70
> [ 246.756162] ret_from_fork+0x3a/0x50
> [ 246.756847]
> Showing all locks held in the system:
> [ 246.758261] 3 locks held by kworker/4:0/37:
> [ 246.759016] #0: 00000000f8df4d2d ((wq_completion)"events"){+.+.}, at: process_one_work+0x1b3/0x620
> [ 246.759856] #1: 00000000e6065461 ((work_completion)(&(&dev->mode_config.output_poll_work)->work)){+.+.}, at: process_one_work+0x1b3/0x620
> [ 246.760670] #2: 00000000cb66735f (&helper->lock){+.+.}, at: drm_fb_helper_hotplug_event.part.28+0x20/0xb0 [drm_kms_helper]
> [ 246.761516] 2 locks held by kworker/0:1/60:
> [ 246.762274] #0: 00000000fff6be0f ((wq_completion)"pm"){+.+.}, at: process_one_work+0x1b3/0x620
> [ 246.762982] #1: 000000005ab44fb4 ((work_completion)(&dev->power.work)){+.+.}, at: process_one_work+0x1b3/0x620
> [ 246.763890] 1 lock held by khungtaskd/64:
> [ 246.764664] #0: 000000008cb8b5c3 (rcu_read_lock){....}, at: debug_show_all_locks+0x23/0x185
> [ 246.765588] 5 locks held by kworker/4:2/422:
> [ 246.766440] #0: 00000000232f0959 ((wq_completion)"events_long"){+.+.}, at: process_one_work+0x1b3/0x620
> [ 246.767390] #1: 00000000bb59b134 ((work_completion)(&mgr->work)){+.+.}, at: process_one_work+0x1b3/0x620
> [ 246.768154] #2: 00000000cb66735f (&helper->lock){+.+.}, at: drm_fb_helper_restore_fbdev_mode_unlocked+0x4c/0xb0 [drm_kms_helper]
> [ 246.768966] #3: 000000004c8f0b6b (crtc_ww_class_acquire){+.+.}, at: restore_fbdev_mode_atomic+0x4b/0x240 [drm_kms_helper]
> [ 246.769921] #4: 000000004c34a296 (crtc_ww_class_mutex){+.+.}, at: drm_modeset_backoff+0x8a/0x1b0 [drm]
> [ 246.770839] 1 lock held by dmesg/1038:
> [ 246.771739] 2 locks held by zsh/1172:
> [ 246.772650] #0: 00000000836d0438 (&tty->ldisc_sem){++++}, at: ldsem_down_read+0x37/0x40
> [ 246.773680] #1: 000000001f4f4d48 (&ldata->atomic_read_lock){+.+.}, at: n_tty_read+0xc1/0x870
>
> [ 246.775522] =============================================
>
> After trying dozens of different solutions, I found one very simple one
> that should also have the benefit of preventing us from having to fight
> locking for the rest of our lives. So, we work around these deadlocks by
> deferring all fbcon hotplug events that happen after the runtime suspend
> process starts until after the device is resumed again.
>
> Changes since v6:
> - Remove unused nouveau_fbcon_hotplugged_in_suspend() - Ilia
>
> Changes since v5:
> - Come up with the (hopefully final) solution for solving this dumb
> problem, one that is a lot less likely to cause issues with locking in
> the future. This should work around all deadlock conditions with fbcon
> brought up thus far.
>
> Changes since v4:
> - Add nouveau_fbcon_hotplugged_in_suspend() to workaround deadlock
> condition that Lukas described
> - Just move all of this out of drm_fb_helper. It seems that other DRM
> drivers have already figured out other workarounds for this. If other
> drivers do end up needing this in the future, we can just move this
> back into drm_fb_helper again.
>
> Changes since v3:
> - Actually check if fb_helper is NULL in both new helpers
> - Actually check drm_fbdev_emulation in both new helpers
> - Don't fire off a fb_helper hotplug unconditionally; only do it if
> the following conditions are true (as otherwise, calling this in the
> wrong spot will cause Bad Things to happen):
> - fb_helper hotplug handling was actually inhibited previously
> - fb_helper actually has a delayed hotplug pending
> - fb_helper is actually bound
> - fb_helper is actually initialized
> - Add __must_check to drm_fb_helper_suspend_hotplug(). There's no
> situation where a driver would actually want to use this without
> checking the return value, so enforce that
> - Rewrite and clarify the documentation for both helpers.
> - Make sure to return true in the drm_fb_helper_suspend_hotplug() stub
> that's provided in drm_fb_helper.h when CONFIG_DRM_FBDEV_EMULATION
> isn't enabled
> - Actually grab the toplevel fb_helper lock in
> drm_fb_helper_resume_hotplug(), since it's possible other activity
> (such as a hotplug) could be going on at the same time the driver
> calls drm_fb_helper_resume_hotplug(). We need this to check whether or
> not drm_fb_helper_hotplug_event() needs to be called anyway
>
> Signed-off-by: Lyude Paul <lyude@xxxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> Cc: Lukas Wunner <lukas@xxxxxxxxx>
> Cc: Karol Herbst <karolherbst@xxxxxxxxx>
> ---
> drivers/gpu/drm/nouveau/dispnv50/disp.c | 2 +-
> drivers/gpu/drm/nouveau/nouveau_display.c | 2 +-
> drivers/gpu/drm/nouveau/nouveau_fbcon.c | 57 +++++++++++++++++++++++
> drivers/gpu/drm/nouveau/nouveau_fbcon.h | 5 ++
> 4 files changed, 64 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.c b/drivers/gpu/drm/nouveau/dispnv50/disp.c
> index 8b522a9b12f6..a0772389ed90 100644
> --- a/drivers/gpu/drm/nouveau/dispnv50/disp.c
> +++ b/drivers/gpu/drm/nouveau/dispnv50/disp.c
> @@ -2049,7 +2049,7 @@ nv50_disp_atomic_state_alloc(struct drm_device *dev)
> static const struct drm_mode_config_funcs
> nv50_disp_func = {
> .fb_create = nouveau_user_framebuffer_create,
> - .output_poll_changed = drm_fb_helper_output_poll_changed,
> + .output_poll_changed = nouveau_fbcon_output_poll_changed,
> .atomic_check = nv50_disp_atomic_check,
> .atomic_commit = nv50_disp_atomic_commit,
> .atomic_state_alloc = nv50_disp_atomic_state_alloc,
> diff --git a/drivers/gpu/drm/nouveau/nouveau_display.c b/drivers/gpu/drm/nouveau/nouveau_display.c
> index 1d36ab5d4796..4b873e668b26 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_display.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_display.c
> @@ -293,7 +293,7 @@ nouveau_user_framebuffer_create(struct drm_device *dev,
>
> static const struct drm_mode_config_funcs nouveau_mode_config_funcs = {
> .fb_create = nouveau_user_framebuffer_create,
> - .output_poll_changed = drm_fb_helper_output_poll_changed,
> + .output_poll_changed = nouveau_fbcon_output_poll_changed,
> };
>
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.c b/drivers/gpu/drm/nouveau/nouveau_fbcon.c
> index 85c1f10bc2b6..8cf966690963 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_fbcon.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.c
> @@ -466,6 +466,7 @@ nouveau_fbcon_set_suspend_work(struct work_struct *work)
> console_unlock();
>
> if (state == FBINFO_STATE_RUNNING) {
> + nouveau_fbcon_hotplug_resume(drm->fbcon);
> pm_runtime_mark_last_busy(drm->dev->dev);
> pm_runtime_put_sync(drm->dev->dev);
> }
> @@ -487,6 +488,61 @@ nouveau_fbcon_set_suspend(struct drm_device *dev, int state)
> schedule_work(&drm->fbcon_work);
> }
>
> +void
> +nouveau_fbcon_output_poll_changed(struct drm_device *dev)
> +{
> + struct nouveau_drm *drm = nouveau_drm(dev);
> + struct nouveau_fbdev *fbcon = drm->fbcon;
> + int ret;
> +
> + if (!fbcon)
> + return;
> +
> + mutex_lock(&fbcon->hotplug_lock);
> +
> + ret = pm_runtime_get(dev->dev);
> + if (ret == 1 || ret == -EACCES) {
> + drm_fb_helper_hotplug_event(&fbcon->helper);
> +
> + pm_runtime_mark_last_busy(dev->dev);
> + pm_runtime_put_autosuspend(dev->dev);
> + } else if (ret == 0) {
> + /* If the GPU was already in the process of suspending before
> + * this event happened, then we can't block here as we'll
> + * deadlock the runtime pmops since they wait for us to
> + * finish. So, just defer this event for when we runtime
> + * resume again. It will be handled by fbcon_work.
> + */

Hm, I thought pm_runtime_get() never fails, and you need one of the other
variants if you want special behaviour here? Like
pm_runtime_get_noresume().

Otherwise lgtm.
-Daniel

> + NV_DEBUG(drm, "fbcon HPD event deferred until runtime resume\n");
> + fbcon->hotplug_waiting = true;
> + pm_runtime_put_noidle(drm->dev->dev);
> + } else {
> + DRM_WARN("fbcon HPD event lost due to RPM failure: %d\n",
> + ret);
> + }
> +
> + mutex_unlock(&fbcon->hotplug_lock);
> +}
> +
> +void
> +nouveau_fbcon_hotplug_resume(struct nouveau_fbdev *fbcon)
> +{
> + struct nouveau_drm *drm;
> +
> + if (!fbcon)
> + return;
> + drm = nouveau_drm(fbcon->helper.dev);
> +
> + mutex_lock(&fbcon->hotplug_lock);
> + if (fbcon->hotplug_waiting) {
> + fbcon->hotplug_waiting = false;
> +
> + NV_DEBUG(drm, "Handling deferred fbcon HPD events\n");
> + drm_fb_helper_hotplug_event(&fbcon->helper);
> + }
> + mutex_unlock(&fbcon->hotplug_lock);
> +}
> +
> int
> nouveau_fbcon_init(struct drm_device *dev)
> {
> @@ -505,6 +561,7 @@ nouveau_fbcon_init(struct drm_device *dev)
>
> drm->fbcon = fbcon;
> INIT_WORK(&drm->fbcon_work, nouveau_fbcon_set_suspend_work);
> + mutex_init(&fbcon->hotplug_lock);
>
> drm_fb_helper_prepare(dev, &fbcon->helper, &nouveau_fbcon_helper_funcs);
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.h b/drivers/gpu/drm/nouveau/nouveau_fbcon.h
> index a6f192ea3fa6..db9d52047ef8 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_fbcon.h
> +++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.h
> @@ -41,6 +41,9 @@ struct nouveau_fbdev {
> struct nvif_object gdi;
> struct nvif_object blit;
> struct nvif_object twod;
> +
> + struct mutex hotplug_lock;
> + bool hotplug_waiting;
> };
>
> void nouveau_fbcon_restore(void);
> @@ -68,6 +71,8 @@ void nouveau_fbcon_set_suspend(struct drm_device *dev, int state);
> void nouveau_fbcon_accel_save_disable(struct drm_device *dev);
> void nouveau_fbcon_accel_restore(struct drm_device *dev);
>
> +void nouveau_fbcon_output_poll_changed(struct drm_device *dev);
> +void nouveau_fbcon_hotplug_resume(struct nouveau_fbdev *fbcon);
> extern int nouveau_nofbaccel;
>
> #endif /* __NV50_FBCON_H__ */
> --
> 2.17.1
>

--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch