NO_EVICT bos that are not idle when all references are dropped are put on
the delayed destroy list. However, since they are not on LRU lists, they
are not available to shrinkers at that point, and buffers on the delayed
destroy list are not checked very often for idle.
So when these buffers are put on the delayed destroy list, clear the
NO_EVICT flag and put them on the right LRU list. This way they are
immediately available for eviction or shrinkers and will not cause false
OOMS.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Make use of the FAULT_FLAG_ALLOW_RETRY flag to allow dropping the
mmap_sem while waiting for bo idle.
FAULT_FLAG_ALLOW_RETRY appears to be primarily designed for disk waits
but should work just as fine for GPU waits..
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
So GNOME userspace has an issue with when it rescans for modes on hotplug
events, if the monitor has no EDID it assumes that nothing has changed on
EDID as with real hw we'd never have new modes without a new EDID, and they
kind off rely on the behaviour now, however with virtual GPUs we would
like to rescan the modes and get a new preferred mode on hotplug events
to handle dynamic guest resizing (where you resize the host window and the
guest resizes with it).
This is a simple property we can make userspace watch for to trigger new
behaviour based on it, and can be used to replaced EDID hacks in virtual
drivers.
Acked-by: Marc-André Lureau <marcandre.lureau@gmail.com> (on irc)
Signed-off-by: Dave Airlie <airlied@redhat.com>
Instead of freeing minors in drm_dev_unregister(), we only unplug them and
delay the free to drm_dev_free(). Note that if drm_dev_register() has
never been called, minors are NULL and this has no effect.
This change is needed to allow early device unregistration. If we want to
call drm_dev_unregister() on live devices, we need to guarantee that
minors are still valid (but unplugged). This way, any open file can still
access file_priv->minor->dev to get the DRM device. However, the minor is
unplugged so no new users can occur.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Don't delay minor removal to drm_put_minor(). Otherwise, user-space can
still open the minor and cause the kernel to oops. Instead, remove the
minor during unplug so any new open() will fail to access this minor.
Note that open() and drm_unplug_minor() are both protected by the global
DRM mutex so we're fine.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
There is no reason to delay debugfs-cleanup to drm_put_minor(). We should
forbid any access to debugfs files once the device is dead. Chances they
oops once a card was unplugged are very high, anyway.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
drm_get_minor() is only used in one file. Make it static and add a
kernel-doc comment which documents the current semantics.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Allow passing NULL as minor to simplify DRM destruction paths. Also remove
the double-pointer reset as it is no longer needed. drm_put_minor() is
only called when the underlying object is destroyed. Hence, resetting
minors to NULL is not necessary.
As drm_put_minor() is no longer used by other DRM files, we can make it
static, too.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This protects drm_unplug_minor() against repeated calls so we can use it
in drm_put_minor(). This allows us to further simplify it in follow-ups as
we no longer do minor-destruction in both functions but only in
drm_unplug_minor().
Also add kernel-doc comments about what these calls do.
[airlied: fixup for changes to kdev stuff]
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Bit-copying restoration of CRTC structure in failure-recovery
path of drm_crtc_helper_set_config function evokes a
subtle and rare, but very dangerous, corruption of
CRTC mutex structure.
Namely, if drm_crtc_helper_set_config takes the path under
'fail:' label *and* some other process has attempted to
grab the crtc mutex (and got blocked), restoring the CRTC
structure by bit-copying it will overwrite the CRTC mutex
state and the waiters list pointer within the mutex structure.
Consequently the blocked process will never be scheduled.
This patch fixes the issue by eliminating the bit-copy
restoration. The elimination is possible because previous
patches have cleaned up the resoration path so that only
the fields touched by the drm_crtc_helper_set_config function
are saved and restored if necessary.
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
There is no need to set crtc->enabled field in
drm_crtc_helper_set_config. This is already done (and
properly restored in case of failure) in
drm_crtc_helper_set_mode that is called by
drm_crtc_helper_set_config. Doing it at only one
place makes restoration in case of failure easier.
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
There is no need to save or restore hwmode field, because by
the time this function sets this field, it cannot fail any more.
However, we should save old enabled field because if
the function fails, we want to return with unchanged CRTC.
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Old framebuffer is stored in save_set.fb and it is
the same value that is later stored in old_fb.
This makes old_fb redundant so we can replace
it with save_set.fb.
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Replace the sparse array of booleans with a bitfield.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
When a second process opens the device and master transferrence is
complete, we walk the list of open devices and remove their
authentication. This also revokes our root privilege. Instead of simply
dropping the authentication, this patch reverts the authenticated state
back to its original value.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This causes problems with never going busy due to ptherm polling,
and after talking to Ben I can't see it being required.
Signed-off-by: Dave Airlie <airlied@redhat.com>
drm_connector_sysfs_add() explicitly checks if connector->kdev
is already populated and returns success. So it clearly now allows
being called multiple times. Remove some stale comments to the contrary.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Don't block forever if there is nothing to wait for.
Signed-off-by: Christian König <christian.koenig@amd.com>
Tested-by: Rafa? Mi?ecki <zajec5@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
qxl devices can have a 64bit surface bar, which is quite handy if
you need a bit more surface memory. So try to use it if it is
present. Note that this bar might be mapped above 4g.
QEMU command line to check that out:
qemu-system-x86_64 -m 4g \
-vga qxl -global qxl-vga.vram64_size_mb=512 \
$otheroptions
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Explicitly set 1024x768 as default mode, so the display doesn't come up
with the largest supported mode.
While being at it drop first three drm_add_modes_noedid calls. As
drm_add_modes_noedid fills the mode list with modes from the database
*up to* the specified size it is pretty pointless to call it multiple
times with different sizes.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
New helper function to set the preferred video mode. Can be called
after drm_add_modes_noedid if you don't want the largest supported
video mode be used by default.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
drm_fb_get_bpp_depth() likes to complain about unsupported pixel formats
but doesn't bother telling us what the format was. Also format_check()
just returns an error when it encouters an invalid format, leaving the
user scratching his head trying to figure out why addfb failed. Make
life a bit easier by using drm_get_format_name() in both places.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
I got very confused when I tried to compare the EST modes with the spec.
Bring over a comment from xf86EdidModes.c that actually describes some
of history where these things came from.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Also check the est3 modes whose presence is indicated by bit 0.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The correct refresh rate for this mode is 75, not 85.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Let's be a bit more consistent with our error values.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Let's be a bit more consistent with our error values.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Let's be a bit more consistent with our error values.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Let's be a bit more consistent with our error values.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Let's be a bit more consistent with our error values.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Return -ENOENT for framebuffers like we do for other mode objects that
can't be found.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We tend to return -EINVAL for everything. Let's try to help poor
userland developers a bit by at least returning -ENONET for missing
objects.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Most architectures define virt_to_page() as a macro that casts its
argument such that an argument of type unsigned long will be accepted
without complaint. However, the proper type is void *, and passing
unsigned long results in a warning on MIPS.
Compile-tested only.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
By definition, the page offset will not affect the result.
Compile-tested only.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
When there are unconsumed pending events, the events are
destroyed by calling destroy callback, but the events list
are remained, because there is no list_del().
It is possible that the page flip request is handled after
drm_events_release() is called and before drm_fb_release().
In this case a drm_pending_event is remained not freed.
So exynos driver checks again to remove it in its post
close routine. But the file_priv->event_list contains
undeleted ones, this can make oops for accessing invalid
memory.
Signed-off-by: YoungJun Cho <yj44.cho@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Those structures are not used anywhere.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Those functions are just reading data from those pointers.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This removes an open coded simple_open() function and replaces file
operations references to the function with simple_open() instead.
Signed-off-by: Duan Jiong <duanj.fnst@cn.fujitsu.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The driver registers a backlight device and thus requires
BACKLIGHT_CLASS_DEVICE to be selected to avoid compilation breakages.
Cc: stable@vger.kernel.org
Reported-by: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Move the ktime_get() clock readouts and potential preempt_disable()
calls from drm core into kms driver to make it compatible with the
api changes in the drm core.
The intel-kms driver needs to take the uncore.lock inside
i915_get_crtc_scanoutpos() and intel_pipe_in_vblank().
This is incompatible with the preempt_disable() on a
PREEMPT_RT patched kernel, as regular spin locks must not
be taken within a preempt_disable'd section. Lock contention
on the uncore.lock also introduced too much uncertainty in vblank
timestamps.
Push the ktime_get() timestamping for scanoutpos queries and
potential preempt_disable_rt() into i915_get_crtc_scanoutpos(),
so these problems can be avoided:
1. First lock the uncore.lock (might sleep on a PREEMPT_RT kernel).
2. preempt_disable_rt() (will be added by the rt-linux folks).
3. ktime_get() a timestamp before scanout pos query.
4. Do all mmio reads as fast as possible without grabbing any new locks!
5. ktime_get() a post-query timestamp.
6. preempt_enable_rt()
7. Unlock the uncore.lock.
This reduces timestamp uncertainty on a low-end HP Atom Mini netbook
with Intel GMA-950 nicely:
Before: 3-8 usecs with spikes > 20 usecs, triggering query retries.
After : Typically 1 usec (98% of all samples), occassionally 2 usecs
(2% of all samples), with maximum of 3 usecs (a handful).
v2: Fix formatting of new multi-line code comments.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Move the ktime_get() clock readouts and potential preempt_disable()
calls from drm core into kms driver to make it compatible with the
api changes in the drm core.
This should not introduce any change in functionality or behaviour
in radeon-kms, just a reshuffling of code.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
A change in locking of some kms drivers (currently intel-kms) make
the old approach too inaccurate and also incompatible with the
PREEMPT_RT realtime kernel patchset.
The driver->get_scanout_position() method of intel-kms now needs
to aquire a spinlock, which clashes badly with the former
preempt_disable() calls in the drm, and it also introduces larger
delays and timing uncertainty on a contended lock than acceptable.
This patch changes the prototype of driver->get_scanout_position()
to require/allow kms drivers to perform the ktime_get() system time
queries which go along with actual scanout position readout in a way
that provides maximum precision and to return those timestamps to
the drm. kms drivers implementations of get_scanout_position() are
asked to implement timestamping and scanoutpos readout in a way
that is as precise as possible and compatible with preempt_disable()
on a PREMPT_RT kernel. A driver should follow this pattern in
get_scanout_position() for precision and compatibility:
spin_lock...(...);
preempt_disable_rt(); // On a PREEMPT_RT kernel, otherwise omit.
if (stime) *stime = ktime_get();
... Minimum amount of MMIO register reads to get scanout position ...
... no taking of locks allowed here! ...
if (etime) *etime = ktime_get();
preempt_enable_rt(); // On PREEMPT_RT kernel, otherwise omit.
spin_unlock...(...);
v2: Fix formatting of new multi-line code comments.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Preemption handling will get pushed into the kms
drivers in followup patches, to make timestamping
more robust and PREEMPT_RT friendly.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Initial pull request for radeon drm-next 3.13. Highlights:
- Enable DPM on a number of asics by default
- Enable audio by default
- Dynamically power down dGPUs on PowerXpress systems
- Lots of bug fixes
* 'drm-next-3.13' of git://people.freedesktop.org/~agd5f/linux: (36 commits)
drm/radeon: don't share PPLLs on DCE4.1
drm/radeon/dpm: fix typo in setting smc flag
drm/radeon: fixup locking inversion between, mmap_sem and reservations
drm/radeon: clear the page directory using the DMA
drm/radeon: initially clear page tables
drm/radeon: drop CP page table updates & cleanup v2
drm/radeon: add vm_set_page tracepoint
drm/radeon: rework and fix reset detection v2
drm/radeon: don't use PACKET2 on CIK
drm/radeon: fix UVD destroy IB size
drm/radeon: activate UVD clocks before sending the destroy msg
drm/radeon/si: fix define for MC_SEQ_TRAIN_WAKEUP_CNTL
drm/radeon: fix endian handling in rlc buffer setup
drm/radeon/dpm: retain user selected performance level across state changes
drm/radeon: disable force performance state when thermal state is active
drm/radeon: enable DPM by default on r7xx asics
drm/radeon: enable DPM by default on evergreen asics
drm/radeon: enable DPM by default on BTC asics
drm/radeon: enable DPM by default on SI asics
drm/radeon: enable DPM by default on SUMO/PALM APUs
...
The biggest part of the changes is the decoupling of the host1x and DRM
drivers followed by the move of Tegra DRM back to drivers/gpu/drm/tegra
from whence it came. There is a lot of cleanup as well, and the drivers
can now be properly unloaded and reloaded.
HDMI support for the Tegra114 SoC was contributed by Mikko Perttunen.
gr2d support was extended to Tegra114 and the gr3d driver that has been
in the works for quite some time finally made it in. All pieces to run
an OpenGL driver on top of an upstream kernel are now available.
Support for syncpoint bases was added by Arto Merilainen. This is useful
for synchronizing between command streams from different engines such as
gr2d and gr3d.
Erik Faye-Lund and Wei Yongjun contributed various small fixes. Thanks!
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)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=qc03
-----END PGP SIGNATURE-----
Merge tag 'drm/for-3.13-rc1' of git://anongit.freedesktop.org/tegra/linux into drm-next
drm/tegra: Changes for v3.13-rc1
The biggest part of the changes is the decoupling of the host1x and DRM
drivers followed by the move of Tegra DRM back to drivers/gpu/drm/tegra
from whence it came. There is a lot of cleanup as well, and the drivers
can now be properly unloaded and reloaded.
HDMI support for the Tegra114 SoC was contributed by Mikko Perttunen.
gr2d support was extended to Tegra114 and the gr3d driver that has been
in the works for quite some time finally made it in. All pieces to run
an OpenGL driver on top of an upstream kernel are now available.
Support for syncpoint bases was added by Arto Merilainen. This is useful
for synchronizing between command streams from different engines such as
gr2d and gr3d.
Erik Faye-Lund and Wei Yongjun contributed various small fixes. Thanks!
* tag 'drm/for-3.13-rc1' of git://anongit.freedesktop.org/tegra/linux: (45 commits)
drm/tegra: Reserve syncpoint base for gr3d
drm/tegra: Reserve base for gr2d
drm/tegra: Deliver syncpoint base to user space
gpu: host1x: Add syncpoint base support
gpu: host1x: Add 'flags' field to syncpt request
drm/tegra: Disable clock on probe failure
gpu: host1x: Disable clock on probe failure
drm/tegra: Support bottom-up buffer objects
drm/tegra: Add support for tiled buffer objects
drm/tegra: Add 3D support
drm/tegra: Introduce tegra_drm_submit()
drm/tegra: Use symbolic names for gr2d registers
drm/tegra: Start connectors with correct DPMS mode
drm/tegra: hdmi: Enable VDD earlier for hotplug/DDC
drm/tegra: hdmi: Fix build warnings
drm/tegra: hdmi: Detect DVI-only displays
drm/tegra: Add Tegra114 HDMI support
drm/tegra: hdmi: Parameterize based on compatible property
drm/tegra: hdmi: Rename tegra{2,3} to tegra{20,30}
gpu: host1x: Add support for Tegra114
...
