History tells us that if we cannot reset the GPU now, we never will. This
then impacts everything that is run subsequently. On failing the reset,
we mark the driver as wedged, trying to prevent further execution on the
GPU, forcing userspace to fallback to using the CPU to update its
framebuffers and let the user know what happened.
We also want to go one step further and add a taint to the kernel so that
any subsequent faults can be traced back to this failure. This is
useful for CI, where if the GPU/driver fails we want to reboot and
restart testing rather than continue on into oblivion. For everyone
else, the warning taint is a testament to the system unreliability.
TAINT_WARN is used anytime a WARN() is emitted, which is suitable for
our purposes here as well; the driver/system may behave unexpectedly
after the failure.
v2: Also taint if the recovery fails (again history shows us that is
typically fatal).
v3: Use TAINT_WARN
References: https://bugs.freedesktop.org/show_bug.cgi?id=103514
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Tomi Sarvela <tomi.p.sarvela@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171205172757.32609-1-chris@chris-wilson.co.uk
From the shrinker paths, we want to relinquish the GPU and GGTT access to
the object, releasing the backing storage back to the system for
swapout. As a part of that process we would unpin the pages, marking
them for access by the CPU (for the swapout/swapin). However, if that
process was interrupted after unbind the vma, we missed a flush of the
inflight GGTT writes before we made that GTT space available again for
reuse, with the prospect that we would redirect them to another page.
The bug dates back to the introduction of multiple GGTT vma, but the
code itself dates to commit 02bef8f98d ("drm/i915: Unbind closed vma
for i915_gem_object_unbind()").
Fixes: 02bef8f98d ("drm/i915: Unbind closed vma for i915_gem_object_unbind()")
Fixes: c5ad54cf7d ("drm/i915: Use partial view in mmap fault handler")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: stable@vger.kernel.org
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171204132513.7303-1-chris@chris-wilson.co.uk
When we detect consecutive allocation of pages swap them to avoid
accidentally freeing them as huge page.
v2: use swap
v3: check if it's really the first allocated page
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Roger He <Hongbo.He@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Previously I was under the impression that the scanline counter
reads 0 when the pipe is off. Turns out that's not correct, and
instead the scanline counter simply stops when the pipe stops, and
it retains it's last value until the pipe starts up again, at which
point the scanline counter jumps to vblank start.
These jumps can cause the timestamp to jump backwards by one frame.
Since we use the timestamps to guesstimage also the frame counter
value on gen2, that would cause the frame counter to also jump
backwards, which leads to a massice difference from the previous value.
The end result is that flips/vblank events don't appear to complete as
they're stuck waiting for the frame counter to catch up to that massive
difference.
Fix the problem properly by actually making sure the scanline counter
has started to move before we assume that it's safe to enable vblank
processing.
v2: Less pointless duplication in the code (Chris)
Cc: stable@vger.kernel.org
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Fixes: b7792d8b54 ("drm/i915: Wait for pipe to start before sampling vblank timestamps on gen2")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171129153732.3612-1-ville.syrjala@linux.intel.com
(cherry picked from commit 8fedd64dab)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Till now, DE2 driver supported only UI planes. Before we add support for
VI planes, lets split out UI layer specific code from common parts. This
commit does the following:
- renames sun8i_layer.c to sun8i_ui_layer.c
- moves UI channel specific code to sun8i_ui_layer.c
- moves common code from sun8i_layer.c to sun8i_mixer.c
- renames function and structure names so it is apparent where they
belong to
No functional change is made.
Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171201060550.10392-19-jernej.skrabec@siol.net
Till now, plane selection was hardcoded to first overlay in first UI
channel and layer parameter is unused.
Rename and add parameters to layer functions so they would represent HW
more accurately and start using then.
It turns out that overlays don't fit well in current DRM design, because
they can't be blended together or scaled independetly when they are set
to same channel. Because of that, always use only first overlay in each
channel.
Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171201060550.10392-11-jernej.skrabec@siol.net
Current code sets alpha mode to global alpha mode and global alpha
value to 0xff which is totaly opaque. That is not needed for two
reasons:
- only one plane is active and thus it can be blended only with
background, which is black,
- it will hinder proper blending when more than one plane is supported
Default mode (0) considers pixel alpha value or 0xff if pixel has
no alpha information. Global alpha value is ignored in this case.
Because of that, just remove the code.
Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171201060550.10392-4-jernej.skrabec@siol.net
The ring_id maximum boundary is being compared using the > operator
instead of >=, leading to an off-by-one error and an out of bounds
write into array vgpu->hws_pga[]. Fix this by simply using the
correct comparison operator. Also re-work another comparison that
uses the comparison > I915_NUM_ENGINES - 1 to use the >= idiom using
to keep this consistent in this code.
Detected by CoverityScan, CID#1462404 ("Out-of-bounds write")
Fixes: a2ae95af96 ("drm/i915/gvt: update CSB and CSB write pointer in virtual HWSP")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
The switch statement is missing breaks for the cases of
GVT_FAILSAFE_INSUFFICIENT_RESOURCE and GVT_FAILSAFE_GUEST_ERR. Add them
in.
Detected by CoverityScan, CID#1462416 ("Missing break in switch")
Fixes: e011c6ce2b ("drm/i915/gvt: Add VM healthy check for workload_thread")
Fixes: a33fc7a048 ("drm/i915/gvt: enter failsafe mode when guest requires more resources")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Ideally we could use the VBT for this, that would be simple, in
intel_dsi_init() check dev_priv->vbt.dsi.config->rotation, set
connector->display_info.panel_orientation accordingly and call
drm_connector_init_panel_orientation_property(), done.
Unfortunately vbt.dsi.config->rotation is always 0 even on tablets
with an upside down LCD and where the GOP is properly rotating the
EFI fb in hardware.
So instead we end up reading the rotation from the primary plane.
This commit only implements the panel orientation property for DSI
panels on BYT / CHT hardware, as all known non normal oriented panels
sofar are only found on this hardware.
Changes in v2:
-Read back the rotation applied by the GOP from the primary plane
instead of relying on dev_priv->vbt.dsi.config->rotation, because it
seems that the VBT rotation filed is always 0 even on devices where the
GOP does apply a rotation
Changes in v3:
-Rewrite the code to read back the orientation from the primary
plane to contain all of this in intel_dsi.c instead of poking a bunch
of holes between all the different layers
Changes in v6:
-Move hardware readout to intel_dsi_init()
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171125193553.23986-6-hdegoede@redhat.com
Apply the "panel orientation" drm connector prop to the primary plane so
that fbcon and fbdev using userspace programs display the right way up.
Changes in v3:
-Use a rotation member in struct drm_fb_helper_crtc and set that from
drm_setup_crtcs instead of looping over all crtc's to find the right one
later
-Since we now no longer look at rotation quirks directly in the fbcon
code, set fb_info.fbcon_rotate_hint when the panel is not mounted upright
and we cannot use hardware rotation
Changes in v4:
-Make drm_fb_helper_init() init drm_fb_helper_crtc.rotation to
DRM_MODE_ROTATE_0 for all crtcs, so that we do not end up setting the
plane_state's rotation to an invalid value for disabled crtcs
(caught by Fi.CI)
Changes in v5:
-Only use hardware (crtc primary plane) rotation for DRM_ROTATE_180,
90 / 270 degree rotation requires special handling which we lack atm
-Add a TODO comment for 90 / 270 degree hardware rotation
-Add some comments to better document the default case when mapping
sw_rotations to fbcon_rotate_hints
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=94894
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171125193553.23986-5-hdegoede@redhat.com
On some devices the LCD panel is mounted in the casing in such a way that
the up/top side of the panel does not match with the top side of the
device (e.g. it is mounted upside-down).
This commit adds the necessary infra for lcd-panel drm_connector-s to
have a "panel orientation" property to communicate how the panel is
orientated vs the casing.
Userspace can use this property to check for non-normal orientation and
then adjust the displayed image accordingly by rotating it to compensate.
Changes in v2:
-Store panel_orientation in drm_display_info, so that drm_fb_helper.c can
access it easily
-Have a single drm_connector_init_panel_orientation_property rather then
create and attach functions. The caller is expected to set
drm_display_info.panel_orientation before calling this, then this will
check for platform specific quirks overriding the panel_orientation and if
the panel_orientation is set after this then it will attach the property.
Changes in v6:
-Use an enum (with kerneldoc) rather then #defines for
DRM_MODE_PANEL_ORIENTATION_*
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171125193553.23986-4-hdegoede@redhat.com
Some x86 clamshell design devices use portrait tablet screens and a display
engine which cannot rotate in hardware, so the firmware just leaves things
as is and we cannot figure out that the display is oriented non upright
from the hardware.
So at least on x86, we need a quirk table for this. This commit adds a DMI
based quirk table which is initially populated with 5 such devices: Asus
T100HA, GPD Pocket, GPD win, I.T.Works TW891 and the VIOS LTH17.
This quirk table will be used by the drm code to let userspace know that
the display is not mounted upright inside the devices case through a new
panel orientation drm-connector property, as well as to tell fbcon to
rotate the console so that it shows the right way up.
Changes in v5:
-Add a kernel-doc comment documenting drm_get_panel_orientation_quirk()
-Remove board_* matches from the dmi-matches for the VIOS LTH17 laptop,
keeping only the (identical) sys_vendor and product_name matches.
This is necessary because an older version of the bios has
board_vendor set to VOIS instead of VIOS
Changes in v6:
-Add reference to added kernel-docs in Documentation/gpu/drm-kms-helpers.rst
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171125193553.23986-3-hdegoede@redhat.com
if app close CTX right after IB submit, gpu recover
will fail to find out the entity behind this guilty
job thus lead to no job skipping for this guilty job.
to fix this corner case just move the increasement of
job->karma out of the entity iteration.
v2:
only do karma increasment if bad->s_priority != KERNEL
because we always consider KERNEL job be correct and always
want to recover an unfinished kernel job (sometimes kernel
job is interrupted by VF FLR or other GPU hang event)
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-By: Xiangliang Yu <Xiangliang.Yu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
