Age | Commit message (Collapse) | Author | Files | Lines |
|
This just resets the dma_fence to NULL so it looks like it's never been
signaled. This will be useful once we add the new wait API for allowing
wait on "submit and signal" behavior.
v2:
- Take an array of sync objects (Dave Airlie)
v3:
- Throw -EINVAL if pad != 0
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Christian König <[email protected]> (v1)
Signed-off-by: Dave Airlie <[email protected]>
|
|
The wait ioctl has a bunch of code to read an syncobj handle array from
userspace and turn it into an array of syncobj pointers. We're about to
add two new IOCTLs which will need to work with arrays of syncobj
handles so let's make some helpers.
Signed-off-by: Jason Ekstrand <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
Vulkan VkFence semantics require that the application be able to perform
a CPU wait on work which may not yet have been submitted. This is
perfectly safe because the CPU wait has a timeout which will get
triggered eventually if no work is ever submitted. This behavior is
advantageous for multi-threaded workloads because, so long as all of the
threads agree on what fences to use up-front, you don't have the extra
cross-thread synchronization cost of thread A telling thread B that it
has submitted its dependent work and thread B is now free to wait.
Within a single process, this can be implemented in the userspace driver
by doing exactly the same kind of tracking the app would have to do
using posix condition variables or similar. However, in order for this
to work cross-process (as is required by VK_KHR_external_fence), we need
to handle this in the kernel.
This commit adds a WAIT_FOR_SUBMIT flag to DRM_IOCTL_SYNCOBJ_WAIT which
instructs the IOCTL to wait for the syncobj to have a non-null fence and
then wait on the fence. Combined with DRM_IOCTL_SYNCOBJ_RESET, you can
easily get the Vulkan behavior.
v2:
- Fix a bug in the invalid syncobj error path
- Unify the wait-all and wait-any cases
v3:
- Unify the timeout == 0 case a bit with the timeout > 0 case
- Use wait_event_interruptible_timeout
v4:
- Use proxy fence
v5:
- Revert to a combination of v2 and v3
- Don't use proxy fences
- Don't use wait_event_interruptible_timeout because it just adds an
extra layer of callbacks
Signed-off-by: Jason Ekstrand <[email protected]>
Cc: Dave Airlie <[email protected]>
Cc: Chris Wilson <[email protected]>
Cc: Christian König <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
This requests that the driver create the sync object such that it
already has a signaled dma_fence attached. Because we don't need
anything in particular (just something signaled), we use a dummy null
fence. This is useful for Vulkan which has a similar flag that can be
passed to vkCreateFence.
Signed-off-by: Jason Ekstrand <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
It is useful in certain circumstances to know when the fence is replaced
in a syncobj. Specifically, it may be useful to know when the fence
goes from NULL to something valid. This does make syncobj_replace_fence
a little more expensive because it has to take a lock but, in the common
case where there is no callback list, it spends a very short amount of
time inside the lock.
v2:
- Don't lock in drm_syncobj_fence_get. We only really need to lock
around fence_replace to make the callback work.
v3:
- Fix the cb_list comment to make kbuild happy
Signed-off-by: Jason Ekstrand <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
This interface will allow sync object to be used to back
Vulkan fences. This API is pretty much the vulkan fence waiting
API, and I've ported the code from amdgpu.
v2: accept relative timeout, pass remaining time back
to userspace.
v3: return to absolute timeouts.
v4: absolute zero = poll,
rewrite any/all code to have same operation for arrays
return -EINVAL for 0 fences.
v4.1: fixup fences allocation check, use u64_to_user_ptr
v5: move to sec/nsec, and use timespec64 for calcs.
v6: use -ETIME and drop the out status flag. (-ETIME
is suggested by ickle, I can feel a shed painting)
v7: talked to Daniel/Arnd, use ktime and ns everywhere.
v8: be more careful in the timeout calculations
use uint32_t for counter variables so we don't overflow
graciously handle -ENOINT being returned from dma_fence_wait_timeout
Signed-off-by: Dave Airlie <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Acked-by: Christian König <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
Signed-off-by: Jason Ekstrand <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
The atomic exchange operation in drm_syncobj_replace_fence is sufficient
for the case where it races with itself. However, if you have a race
between a replace_fence and dma_fence_get(syncobj->fence), you may end
up with the entire replace_fence happening between the point in time
where the one thread gets the syncobj->fence pointer and when it calls
dma_fence_get() on it. If this happens, then the reference may be
dropped before we get a chance to get a new one. The new helper uses
dma_fence_get_rcu_safe to get rid of the race.
This is also needed because it allows us to do a bit more than just get
a reference in drm_syncobj_fence_get should we wish to do so.
v2:
- RCU isn't that scary
- Call rcu_read_lock/unlock
- Don't rename fence to _fence
- Make the helper static inline
Signed-off-by: Jason Ekstrand <[email protected]>
Acked-by: Christian König <[email protected]> (v1)
Signed-off-by: Dave Airlie <[email protected]>
|
|
The function has far more in common with drm_syncobj_find than with
any in the get/put functions.
Signed-off-by: Jason Ekstrand <[email protected]>
Acked-by: Christian König <[email protected]> (v1)
Signed-off-by: Dave Airlie <[email protected]>
|
|
Currently the hikey dsi logic cannot generate accurate byte
clocks values for all pixel clock values. Thus if a mode clock
is selected that cannot match the calculated byte clock, the
device will boot with a blank screen.
This patch uses the new mode_valid callback (many thanks to
Jose Abreu for upstreaming it!) to ensure we don't select
modes we cannot generate.
Also, since the ade crtc code will adjust the mode in mode_set,
this patch also adds a mode_fixup callback which we use to make
sure we are validating the mode clock that will eventually be
used.
Cc: Daniel Vetter <[email protected]>
Cc: Jani Nikula <[email protected]>
Cc: Sean Paul <[email protected]>
Cc: David Airlie <[email protected]>
Cc: Rob Clark <[email protected]>
Cc: Xinliang Liu <[email protected]>
Cc: Xinliang Liu <[email protected]>
Cc: Rongrong Zou <[email protected]>
Cc: Xinwei Kong <[email protected]>
Cc: Chen Feng <[email protected]>
Cc: Jose Abreu <[email protected]>
Cc: Archit Taneja <[email protected]>
Cc: [email protected]
Reviewed-by: Sean Paul <[email protected]>
Signed-off-by: John Stultz <[email protected]>
Reviewed-by: Xinliang Liu <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
Minor version bump to indicate support for fence FD
Signed-off-by: Sinclair Yeh <[email protected]>
Reviewed-by: Deepak Singh Rawat <[email protected]>
Reviewed-by: Thomas Hellstrom <[email protected]>
|
|
Added code to link a fence to a out_fence_fd file descriptor and
thread out_fence_fd down to vmw_execbuf_copy_fence_user() so it can be
copied into the IOCTL reply and be passed back up the the user.
v2:
Make sure to sync and clean up in case of failure
Signed-off-by: Sinclair Yeh <[email protected]>
Reviewed-by: Deepak Singh Rawat <[email protected]>
Reviewed-by: Thomas Hellstrom <[email protected]>
|
|
This allows vmwgfx to wait on a fence created by another
device.
v2:
* Remove special handling for vmwgfx fence and just use dma_fence_wait()
* Use interruptible waits
* Added function documentation
Signed-off-by: Sinclair Yeh <[email protected]>
Reviewed-by: Deepak Singh Rawat <[email protected]>
Reviewed-by: Thomas Hellstrom <[email protected]>
|
|
Make the fields and flags available.
Signed-off-by: Sinclair Yeh <[email protected]>
Reviewed-by: Deepak Singh Rawat <[email protected]>
Reviewed-by: Thomas Hellstrom <[email protected]>
|
|
Sometimes it appears like the device modifies the command header offset
member. So explicitly clear it when restarting after an error.
Signed-off-by: Thomas Hellstrom <[email protected]>
Reviewed-by: Sinclair Yeh <[email protected]>
|
|
Can be used by user-space applications to test and verify the kernel
command buffer error recovery functionality.
Malicious user-space apps could potentially use this command to slow down
graphics processing somewhat, but they could also accomplish the same thing
using a random malformed command so this should be considered safe.
At least as safe as it gets.
Signed-off-by: Thomas Hellstrom <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
Reviewed-by: Sinclair Yeh <[email protected]>
|
|
Previously we skipped the command buffer and added an extra fence to
avoid hangs due to skipped fence commands.
Now we instead restart the command buffer after the failing command,
if there are any commands left.
In addition we print out some information about the failing command
and its location in the command buffer.
Testing Done: ran glxgears using mesa modified to send the NOP_ERROR
command before each 10th clear and verified that we detected the device
error properly and that there were no other device errors caused by
incorrectly ordered command buffers. Also ran the piglit "quick" test
suite which generates a couple of device errors and verified that
they were handled as intended.
Signed-off-by: Thomas Hellstrom <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
Reviewed-by: Sinclair Yeh <[email protected]>
|
|
This gets rid of the irq bottom half tasklets and instead performs the
work needed in process context. We also convert irq-disabling spinlocks to
ordinary spinlocks.
This should decrease system latency for other system components, like
sound for example but has the potential to increase latency for processes
that wait on the GPU.
Signed-off-by: Thomas Hellstrom <[email protected]>
Reviewed-by: Sinclair Yeh <[email protected]>
|
|
We're not allowed to change the upstream version of the drm_irq_install
function to be able to incorporate threaded irqs. So roll our own irq
install- and uninstall functions instead of relying on the drm core ones.
Signed-off-by: Thomas Hellstrom <[email protected]>
|
|
i2c/for-4.14
|
|
Conflicts:
arch/x86/kernel/head64.c
arch/x86/mm/mmap.c
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Signed-off-by: Ingo Molnar <[email protected]>
|
|
DRM core already checks the validity of the pixelformat.
Signed-off-by: Tobias Jakobi <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
A recent commit (272725c7db4da1fd3229d944fc76d2e98e3a144e) has removed
the use of 'bits_per_pixel' in DRM. However the corresponding Exynos
driver code still uses the ambiguous 'bpp', even though it is now
initialized from fb->cpp[0].
Consistenly use 'cpp' in FIMD, DECON7 and DECON5433 drivers.
Signed-off-by: Tobias Jakobi <[email protected]>
|
|
We always translate the dma address such that the offsets of
the source image are zero. Hence we can remove manipulation of
the MXR_GRAPHIC_SXY(win) register and just zero them once
in mixer_win_reset().
Signed-off-by: Tobias Jakobi <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
DRM core already checks in drm_atomic_plane_check() if the
pixelformat is valid. Hence we can collapse the default case
of the switch statement with the XRGB8888 case.
No functional change.
Signed-off-by: Tobias Jakobi <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
DRM core already checks in drm_atomic_plane_check() if the
pixelformat is valid. Hence we can drop the default case of
the switch statement and collapse most of the code.
Also rename the two booleans to reflect what true/false
actually means, and to avoid mixing CrCb/NV21 descriptions.
No functional change.
Signed-off-by: Tobias Jakobi <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
The video processor supports a tiled version of the NV12 format,
known as NV12MT in V4L2 terms. The support was removed in commit
083500baefd5f4c215a5a93aef2492c1aa775828 due to not being a real
pixel format, but rather NV12 with a special memory layout.
With the introduction of FB modifiers, we can now properly support
this format again.
Tested with a hacked up modetest from libdrm's test suite on
an ODROID-X2 (Exynos4412).
Signed-off-by: Tobias Jakobi <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
The current comment sounds like the division op is done to
compensate for some hardware erratum. But the chroma plane
having half the height of the luma plane is just the way
NV12/NV21 is defined, so clarify this behaviour.
Signed-off-by: Tobias Jakobi <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
Since panel's mode of work is propagated properly from panel to DECON,
there is no need to use redundant private device tree property.
The only issue with such approach is that check for required interrupts
should be postponed until panel communicate its requirements, ie to
mode validation phase - mode_valid callback.
Signed-off-by: Andrzej Hajda <[email protected]>
|
|
crtc::mode_valid callback is required to implement proper pipeline
validation for command/video modes. Since Exynos uses private
framework such callback should be added to it.
Signed-off-by: Andrzej Hajda <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
To allow runtime validation of mode of work irq request
code should be split into two separate phases:
- irq reqesting,
- irq checking.
Following patches will move 2nd phase to mode validation phase.
Signed-off-by: Andrzej Hajda <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
MIC driver should use info from CRTC to check mode of work instead of
illegally peeking into nodes of other devices.
Signed-off-by: Andrzej Hajda <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
mipi_dsi framework provides information about panel's mode of work.
This info should be propagated upstream to configure all elements of
the pipeline. As CRTC is the common denominator of the pipeline we can
put such info into its structures.
Signed-off-by: Andrzej Hajda <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
Description of drm_helper_hpd_irq_event clearly states that drivers
supporting hotplug events per connector should use different helper -
drm_kms_helper_hotplug_event. To achieve it following changes have
been performed:
- moved down all DSI ops - they require exynos_dsi_disable function
to be defined earlier,
- simplified exynos_dsi_detect - there is no real detection, it just
returns if panel is attached,
- DSI attach/detach callbacks attaches/detaches DRM panel and sets
connector status and other context fields accordingly, all this is
performed under mutex, as these callbacks are asynchronous.
Signed-off-by: Andrzej Hajda <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
All encoders share the same code to set encoders possible_crtcs field.
The patch creates helper to abstract out this code.
Signed-off-by: Andrzej Hajda <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
Linux core provide helpers for polling with timeout, lets use them.
Signed-off-by: Andrzej Hajda <[email protected]>
Signed-off-by: Inki Dae <[email protected]>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux into drm-next
sun4i DRM changes for 4.14, take 2
A single patch switching to a new OF helper.
* tag 'sunxi-drm-for-4.14-2' of https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux:
drm/sun4i: use of_graph_get_remote_endpoint()
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tomba/linux into drm-next
omapdrm fixes for 4.14
* fix compilation when compiling omapfb driver
* WA for OMAP3 endless sync lost issue
* WA for OMAP5 DSI PLL issue
* fix analog TV out modecheck
* tag 'omapdrm-4.14-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tomba/linux:
ARM: OMAP2+: fix missing variable declaration
drm/omap: work-around for omap3 display enable
drm/omap: fix i886 work-around
drm/omap: fix analog tv-out modecheck
|
|
git://people.freedesktop.org/~robclark/linux into drm-next
Updates for 4.14.. I have some further patches from Jordan to add
multiple priority levels and pre-emption, but those will probably be
for 4.15 to give me time for the mesa parts.
* tag 'drm-msm-next-2017-08-22' of git://people.freedesktop.org/~robclark/linux:
drm/msm/mdp5: mark runtime_pm functions as __maybe_unused
drm/msm: remove unused variable
drm/msm/mdp5: make helper function static
drm/msm: make msm_framebuffer_init() static
drm/msm: add helper to allocate stolen fb
drm/msm: don't track fbdev's gem object separately
drm/msm: add modeset module param
drm/msm/mdp5: add tracking for clk enable-count
drm/msm: remove unused define
drm/msm: Add a helper function for in-kernel buffer allocations
drm/msm: Attach the GPU MMU when it is created
drm/msm: Add A5XX hardware fault detection
drm/msm: Remove uneeded platform dev members
drm/msm/mdp5: Set up runtime PM for MDSS
drm/msm/mdp5: Write to SMP registers even if allocations don't change
drm/msm/mdp5: Don't use mode_set helper funcs for encoders and CRTCs
drm/msm/dsi: Implement RPM suspend/resume callbacks
drm/msm/dsi: Set up runtime PM for DSI
drm/msm/hdmi: Set up runtime PM for HDMI
drm/msm/mdp5: Use runtime PM get/put API instead of toggling clocks
|
|
git://anongit.freedesktop.org/git/drm-misc into drm-fixes
Core Changes:
- Release driver tracking before making the object available again (Chris)
Cc: Chris Wilson <[email protected]>
* tag 'drm-misc-fixes-2017-08-24' of git://anongit.freedesktop.org/git/drm-misc:
drm: Release driver tracking before making the object available again
|
|
git://anongit.freedesktop.org/git/drm-intel into drm-fixes
drm/i915 fixes for v4.13-rc7
* tag 'drm-intel-fixes-2017-08-24' of git://anongit.freedesktop.org/git/drm-intel:
drm/i915/gvt: Fix the kernel null pointer error
drm/i915: Clear lost context-switch interrupts across reset
drm/i915/bxt: use NULL for GPIO connection ID
drm/i915/cnl: Fix LSPCON support.
drm/i915/vbt: ignore extraneous child devices for a port
drm/i915: Initialize 'data' in intel_dsi_dcs_backlight.c
|
|
Remove a redundant identical return statement, it has no use.
Detected by CoverityScan, CID#1454586 ("Structurally dead code")
Reviewed-by: Felix Kuehling <[email protected]>
Signed-off-by: Colin Ian King <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
|
|
Check memory allocation failure and return -ENOMEM in such a case.
'num_post_dep_syncobjs' still has to be set to 0 before the test in order
to have it initialized if 'amdgpu_cs_parser_fini()' is called to free
resources.
The calling graph would be, in such a case!
failure in amdgpu_cs_process_syncobj_out_dep()
---> error code returned by amdgpu_cs_dependencies()
--> amdgpu_cs_parser_fini() is called
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Christophe JAILLET <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
|
|
Function sii8620_mt_read_devcap_reg_recv() used to read array index
from a wrong msg register, which caused writing out of array
bounds. It led to writing on other fields of struct sii8620.
Signed-off-by: Maciej Purski <[email protected]>
Fixes: e9c6da270 ("drm/bridge/sii8620: add reading device capability
registers")
Signed-off-by: Andrzej Hajda <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
|
|
BANK_SELECT should always be FRAGMENT_SIZE + 3 due to 8-entry (2^3)
per cache line in L2 TLB for Vega10.
v2: agd: fix warning
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Roger He <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
|
|
The function is called only once and doesn't do anything special.
Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Roger He <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
|
|
Use ttm_bo_mem_space instead of manually allocating GART space.
This allows us to evict BOs when there isn't enought GART space any more.
Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
|
|
This isn't used since we don't map evicted BOs to GART any more.
Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Roger He <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
|
|
we observe abnormal number from:
/sys/devices/virtual/drm/amdttm/buffer_objects/bo_count
bo_count is atomic_inc which is "int" type,
shouldn't explicitly turn it to unsigned long.
Signed-off-by: Monk Liu <[email protected]>
Reviewed-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
|