kernel/linux.git/drivers/gpu, branch v4.9.33

drm/i915: Always recompute watermarks when distrust_bios_wm is set, v2.

2017-06-17T04:41:58+00:00

commit 4e3aed844547f63614363a386de126e6304e55fb upstream. On some systems there can be a race condition in which no crtc state is added to the first atomic commit. This results in all crtc's having a null DDB allocation, causing a FIFO underrun on any update until the first modeset. Changes since v1: - Do not take the connection_mutex, this is already done below. Reported-by: Maarten Lankhorst Inspired-by: Mahesh Kumar Signed-off-by: Maarten Lankhorst Fixes: 98d39494d375 ("drm/i915/gen9: Compute DDB allocation at atomic check time (v4)") Cc: Mahesh Kumar Cc: Matt Roper Link: http://patchwork.freedesktop.org/patch/msgid/20170531154236.27180-1-maarten.lankhorst@linux.intel.com Reviewed-by: Mahesh Kumar Reviewed-by: Matt Roper Signed-off-by: Greg Kroah-Hartman (cherry picked from commit 367d73d2806085bb507ab44c1f532640917fd5ca) Signed-off-by: Jani Nikula

drm/i915: Workaround VLV/CHV DSI scanline counter hardware fail

2017-06-17T04:41:58+00:00

commit 8f4d38099b3098eae75f7755e1801931f8141350 upstream. The scanline counter is bonkers on VLV/CHV DSI. The scanline counter increment is not lined up with the start of vblank like it is on every other platform and output type. This causes problems for both the vblank timestamping and atomic update vblank evasion. On my FFRD8 machine at least, the scanline counter increment happens about 1/3 of a scanline ahead of the start of vblank (which is where all register latching happens still). That means we can't trust the scanline counter to tell us whether we're in vblank or not while we're on that particular line. In order to keep vblank timestamping in working condition when called from the vblank irq, we'll leave scanline_offset at one, which means that the entire line containing the start of vblank is considered to be inside the vblank. For the vblank evasion we'll need to consider that entire line to be bad, since we can't tell whether the registers already got latched or not. And we can't actually use the start of vblank interrupt to get us past that line as the interrupt would fire too soon, and then we'd up waiting for the next start of vblank instead. One way around that would using the frame start interrupt instead since that wouldn't fire until the next scanline, but that would require some bigger changes in the interrupt code. So for simplicity we'll just poll until we get past the bad line. v2: Adjust the comments a bit Cc: Jonas Aaberg Tested-by: Jonas Aaberg Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99086 Signed-off-by: Ville Syrjälä Link: http://patchwork.freedesktop.org/patch/msgid/20161215174734.28779-1-ville.syrjala@linux.intel.com Tested-by: Mika Kahola Reviewed-by: Mika Kahola (cherry picked from commit ec1b4ee2834e66884e5b0d3d465f347ff212e372) Signed-off-by: Jani Nikula Signed-off-by: Greg Kroah-Hartman

drm/ast: Fixed system hanged if disable P2A

2017-06-17T04:41:54+00:00

[ Upstream commit 6c971c09f38704513c426ba6515f22fb3d6c87d5 ] The original ast driver will access some BMC configuration through P2A bridge that can be disabled since AST2300 and after. It will cause system hanged if P2A bridge is disabled. Here is the update to fix it. Signed-off-by: Y.C. Chen Signed-off-by: Dave Airlie Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman

drm/nouveau: Fix drm poll_helper handling

2017-06-17T04:41:54+00:00

[ Upstream commit 9a2eba337cacefc95b97c2726e3efdd435b3460e ] Commit cae9ff036eea effectively disabled the drm poll_helper by checking the wrong flag to see if the driver should enable the poll or not: mode_config.poll_enabled is only set to true by poll_init and it is not indicating if the poll is enabled or not. nouveau_display_create() will initialize the poll and going to disable it right away. After poll_init() the mode_config.poll_enabled will be true, but the poll itself is disabled. To avoid the race caused by calling the poll_enable() from different paths, this patch will enable the poll from one place, in the nouveau_display_hpd_work(). In case the pm_runtime is disabled we will enable the poll in nouveau_drm_load() once. Fixes: cae9ff036eea ("drm/nouveau: Don't enabling polling twice on runtime resume") Signed-off-by: Peter Ujfalusi Reviewed-by: Lyude Signed-off-by: Ben Skeggs Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman

drm/nouveau: Don't enabling polling twice on runtime resume

2017-06-17T04:41:54+00:00

[ Upstream commit cae9ff036eea577856d5b12860b4c79c5e71db4a ] As it turns out, on cards that actually have CRTCs on them we're already calling drm_kms_helper_poll_enable(drm_dev) from nouveau_display_resume() before we call it in nouveau_pmops_runtime_resume(). This leads us to accidentally trying to enable polling twice, which results in a potential deadlock between the RPM locks and drm_dev->mode_config.mutex if we end up trying to enable polling the second time while output_poll_execute is running and holding the mode_config lock. As such, make sure we only enable polling in nouveau_pmops_runtime_resume() if we need to. This fixes hangs observed on the ThinkPad W541 Signed-off-by: Lyude Cc: Hans de Goede Cc: Kilian Singer Cc: Lukas Wunner Cc: David Airlie Signed-off-by: Dave Airlie Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman

drm/nouveau: Handle fbcon suspend/resume in seperate worker

2017-06-17T04:41:54+00:00

[ Upstream commit 15266ae38fe09dae07bd8812cb7a7717b1e1d992 ] Resuming from RPM can happen while already holding dev->mode_config.mutex. This means we can't actually handle fbcon in any RPM resume workers, since restoring fbcon requires grabbing dev->mode_config.mutex again. So move the fbcon suspend/resume code into it's own worker, and rely on that instead to avoid deadlocking. This fixes more deadlocks for runtime suspending the GPU on the ThinkPad W541. Reproduction recipe: - Get a machine with both optimus and a nvidia card with connectors attached to it - Wait for the nvidia GPU to suspend - Attempt to manually reprobe any of the connectors on the nvidia GPU using sysfs - *deadlock* [airlied: use READ_ONCE to address Hans's comment] Signed-off-by: Lyude Cc: Hans de Goede Cc: Kilian Singer Cc: Lukas Wunner Cc: David Airlie Signed-off-by: Dave Airlie Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman

drm/nouveau: Rename acpi_work to hpd_work

2017-06-17T04:41:54+00:00

[ Upstream commit 81280d0e24e76c35f40f997af26c779bcb10b04d ] We need to call drm_helper_hpd_irq_event() on resume to properly detect monitor connection / disconnection on some laptops. For runtime-resume (which gets called on resume from normal suspend too) we must call drm_helper_hpd_irq_event() from a workqueue to avoid a deadlock. Rename acpi_work to hpd_work, and move it out of the #ifdef CONFIG_ACPI blocks to make it suitable for generic work. Signed-off-by: Hans de Goede Signed-off-by: Ben Skeggs Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman

drm/nouveau: Intercept ACPI_VIDEO_NOTIFY_PROBE

2017-06-17T04:41:54+00:00

[ Upstream commit 3a6536c51d5db3adf58dcd466a3aee6233b58544 ] Various notebooks with nvidia GPUs generate an ACPI_VIDEO_NOTIFY_PROBE acpi-video event when an external device gets plugged in (and again on modesets on that connector), the default behavior in the acpi-video driver for this is to send a KEY_SWITCHVIDEOMODE evdev event, which causes e.g. gnome-settings-daemon to ask us to rescan the connectors (good), but also causes g-s-d to switch to mirror mode on a newly plugged monitor rather then using the monitor to extend the desktop (bad) as KEY_SWITCHVIDEOMODE is supposed to switch between extend the desktop vs mirror mode. More troublesome are the repeated ACPI_VIDEO_NOTIFY_PROBE events on changing the mode on the connector, which cause g-s-d to switch between mirror/extend mode, which causes a new ACPI_VIDEO_NOTIFY_PROBE event and we end up with an endless loop. This commit fixes this by adding an acpi notifier block handler to nouveau_display.c to intercept ACPI_VIDEO_NOTIFY_PROBE and: 1) Wake-up runtime suspended GPUs and call drm_helper_hpd_irq_event() on them, this is necessary in some cases for the GPU to detect connector hotplug events while runtime suspended 2) Return NOTIFY_BAD to stop acpi-video from emitting a bogus KEY_SWITCHVIDEOMODE key-press event There already is another acpi notifier block handler registered in drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.c, but that is not suitable since that one gets unregistered on runtime suspend, and we also want to intercept ACPI_VIDEO_NOTIFY_PROBE when runtime suspended. Signed-off-by: Hans de Goede Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman

drm/i915: Check for NULL i915_vma in intel_unpin_fb_obj()

2017-06-17T04:41:53+00:00

[ Upstream commit 39cb2c9a316e77f6dfba96c543e55b6672d5a37e ] I've seen this trigger twice now, where the i915_gem_object_to_ggtt() call in intel_unpin_fb_obj() returns NULL, resulting in an oops immediately afterwards as the (inlined) call to i915_vma_unpin_fence() tries to dereference it. It seems to be some race condition where the object is going away at shutdown time, since both times happened when shutting down the X server. The call chains were different: - VT ioctl(KDSETMODE, KD_TEXT): intel_cleanup_plane_fb+0x5b/0xa0 [i915] drm_atomic_helper_cleanup_planes+0x6f/0x90 [drm_kms_helper] intel_atomic_commit_tail+0x749/0xfe0 [i915] intel_atomic_commit+0x3cb/0x4f0 [i915] drm_atomic_commit+0x4b/0x50 [drm] restore_fbdev_mode+0x14c/0x2a0 [drm_kms_helper] drm_fb_helper_restore_fbdev_mode_unlocked+0x34/0x80 [drm_kms_helper] drm_fb_helper_set_par+0x2d/0x60 [drm_kms_helper] intel_fbdev_set_par+0x18/0x70 [i915] fb_set_var+0x236/0x460 fbcon_blank+0x30f/0x350 do_unblank_screen+0xd2/0x1a0 vt_ioctl+0x507/0x12a0 tty_ioctl+0x355/0xc30 do_vfs_ioctl+0xa3/0x5e0 SyS_ioctl+0x79/0x90 entry_SYSCALL_64_fastpath+0x13/0x94 - i915 unpin_work workqueue: intel_unpin_work_fn+0x58/0x140 [i915] process_one_work+0x1f1/0x480 worker_thread+0x48/0x4d0 kthread+0x101/0x140 and this patch purely papers over the issue by adding a NULL pointer check and a WARN_ON_ONCE() to avoid the oops that would then generally make the machine unresponsive. Other callers of i915_gem_object_to_ggtt() seem to also check for the returned pointer being NULL and warn about it, so this clearly has happened before in other places. [ Reported it originally to the i915 developers on Jan 8, applying the ugly workaround on my own now after triggering the problem for the second time with no feedback. This is likely to be the same bug reported as https://bugs.freedesktop.org/show_bug.cgi?id=98829 https://bugs.freedesktop.org/show_bug.cgi?id=99134 which has a patch for the underlying problem, but it hasn't gotten to me, so I'm applying the workaround. ] Cc: Daniel Vetter Cc: Jani Nikula Cc: Ville Syrjälä Cc: Chris Wilson Cc: Maarten Lankhorst Cc: Tvrtko Ursulin Cc: Imre Deak Signed-off-by: Linus Torvalds Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman

drm: Don't race connector registration

2017-06-17T04:41:53+00:00

[ Upstream commit e6e7b48b295afa5a5ab440de0a94d9ad8b3ce2d0 ] I was under the misconception that the sysfs dev stuff can be fully set up, and then registered all in one step with device_add. That's true for properties and property groups, but not for parents and child devices. Those must be fully registered before you can register a child. Add a bit of tracking to make sure that asynchronous mst connector hotplugging gets this right. For consistency we rely upon the implicit barriers of the connector->mutex, which is taken anyway, to ensure that at least either the connector or device registration call will work out. Mildly tested since I can't reliably reproduce this on my mst box here. Reported-by: Dave Hansen Cc: Dave Hansen Acked-by: Chris Wilson Cc: Chris Wilson Signed-off-by: Daniel Vetter Link: http://patchwork.freedesktop.org/patch/msgid/1484237756-2720-1-git-send-email-daniel.vetter@ffwll.ch Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman