drm/syncobj: add IOCTL to register an eventfd

Introduce a new DRM_IOCTL_SYNCOBJ_EVENTFD IOCTL which signals an
eventfd from a syncobj.

This is useful for Wayland compositors to handle wait-before-s

drm/syncobj: add IOCTL to register an eventfd

Introduce a new DRM_IOCTL_SYNCOBJ_EVENTFD IOCTL which signals an
eventfd from a syncobj.

This is useful for Wayland compositors to handle wait-before-submit.
Wayland clients can send a timeline point to the compositor
before the point has materialized yet, then compositors can wait
for the point to materialize via this new IOCTL.

The existing DRM_IOCTL_SYNCOBJ_TIMELINE_WAIT IOCTL is not suitable
because it blocks. Compositors want to integrate the wait with
their poll(2)-based event loop.

Requirements for new uAPI:

- User-space patch: https://gitlab.freedesktop.org/wlroots/wlroots/-/merge_requests/4262
- IGT: https://lists.freedesktop.org/archives/igt-dev/2023-July/057893.html

v2:
- Wait for fence when flags is zero
- Improve documentation (Pekka)
- Rename IOCTL (Christian)
- Fix typo in drm_syncobj_add_eventfd() (Christian)

v3:
- Link user-space + IGT patches
- Add reference from overview docs

v4: fix IOCTL number conflict with GETFB2 (Nicholas Choi, Vitaly Prosyak)

Signed-off-by: Simon Ser <[email protected]>
Reviewed-by: Christian König <[email protected]>
Acked-by: Pekka Paalanen <[email protected]>
Cc: Jason Ekstrand <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Bas Nieuwenhuizen <[email protected]>
Cc: Daniel Stone <[email protected]>
Cc: James Jones <[email protected]>
Cc: Austin Shafer <[email protected]>
Cc: Vitaly Prosyak <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

show more ...

drm/syncobj: add new drm_syncobj_add_point interface v4

Use the dma_fence_chain object to create a timeline of fence objects
instead of just replacing the existing fence.

v2: rebase and cleanup
v3:

drm/syncobj: add new drm_syncobj_add_point interface v4

Use the dma_fence_chain object to create a timeline of fence objects
instead of just replacing the existing fence.

v2: rebase and cleanup
v3: fix garbage collection parameters
v4: add unorder point check, print a warn calltrace

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Lionel Landwerlin <[email protected]>
Link: https://patchwork.freedesktop.org/patch/295780/?series=58813&rev=1

show more ...

drm: forward declare struct drm_file in drm_syncobj.h

drm_syncobj.h uses struct drm_file pointers, forward declare struct
drm_file to make the header self-contained. This prepares for dropping
drmP.

drm: forward declare struct drm_file in drm_syncobj.h

drm_syncobj.h uses struct drm_file pointers, forward declare struct
drm_file to make the header self-contained. This prepares for dropping
drmP.h from files including drm_syncobj.h.

Switch from "" to <> includes while at it.

v2: forward declare instead of including drm_file.h (Daniel)

[Updated commit message per Laurent's review while applying.]

Cc: Sam Ravnborg <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Laurent Pinchart <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Reviewed-by: Laurent Pinchart <[email protected]>
Signed-off-by: Jani Nikula <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

show more ...

drm/syncobj: remove drm_syncobj_cb and cleanup

This completes "drm/syncobj: Drop add/remove_callback from driver
interface" and cleans up the implementation a bit.

Signed-off-by: Christian König <c

drm/syncobj: remove drm_syncobj_cb and cleanup

This completes "drm/syncobj: Drop add/remove_callback from driver
interface" and cleans up the implementation a bit.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Chunming Zhou <[email protected]>
Link: https://patchwork.freedesktop.org/patch/266255/

show more ...

drm: revert "expand replace_fence to support timeline point v2"

This reverts commit 9a09a42369a4a37a959c051d8e1a1f948c1529a4.

The whole interface isn't thought through. Since this function can't
fa

drm: revert "expand replace_fence to support timeline point v2"

This reverts commit 9a09a42369a4a37a959c051d8e1a1f948c1529a4.

The whole interface isn't thought through. Since this function can't
fail we actually can't allocate an object to store the sync point.

Sorry, I should have taken the lead on this from the very beginning and
reviewed it more thoughtfully. Going to propose a new interface as a
follow up change.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Chunming Zhou <[email protected]>
Link: https://patchwork.freedesktop.org/patch/265580/

show more ...

drm: Revert syncobj timeline changes.

Daniel suggested I submit this, since we're still seeing regressions
from it. This is a revert to before 48197bc564c7 ("drm: add syncobj
timeline support v9")

drm: Revert syncobj timeline changes.

Daniel suggested I submit this, since we're still seeing regressions
from it. This is a revert to before 48197bc564c7 ("drm: add syncobj
timeline support v9") and its followon fixes.

Fixes this on first V3D testcase execution:

[ 48.767088] ============================================
[ 48.772410] WARNING: possible recursive locking detected
[ 48.777739] 4.19.0-rc6+ #489 Not tainted
[ 48.781668] --------------------------------------------
[ 48.786993] shader_runner/3284 is trying to acquire lock:
[ 48.792408] ce309d7f (&(&array->lock)->rlock){....}, at: dma_fence_add_callback+0x30/0x23c
[ 48.800714]
[ 48.800714] but task is already holding lock:
[ 48.806559] c5952bd3 (&(&array->lock)->rlock){....}, at: dma_fence_add_callback+0x30/0x23c
[ 48.814862]
[ 48.814862] other info that might help us debug this:
[ 48.821410] Possible unsafe locking scenario:
[ 48.821410]
[ 48.827338] CPU0
[ 48.829788] ----
[ 48.832239] lock(&(&array->lock)->rlock);
[ 48.836434] lock(&(&array->lock)->rlock);
[ 48.840640]
[ 48.840640] *** DEADLOCK ***
[ 48.840640]
[ 48.846582] May be due to missing lock nesting notation
[ 130.763560] 1 lock held by cts-runner/3270:
[ 130.767745] #0: 7834b793 (&(&array->lock)->rlock){-...}, at: dma_fence_add_callback+0x30/0x23c
[ 130.776461]
stack backtrace:
[ 130.780825] CPU: 1 PID: 3270 Comm: cts-runner Not tainted 4.19.0-rc6+ #486
[ 130.787706] Hardware name: Broadcom STB (Flattened Device Tree)
[ 130.793645] [<c021269c>] (unwind_backtrace) from [<c020db1c>] (show_stack+0x10/0x14)
[ 130.801404] [<c020db1c>] (show_stack) from [<c0c2c4b0>] (dump_stack+0xa8/0xd4)
[ 130.808642] [<c0c2c4b0>] (dump_stack) from [<c0281a84>] (__lock_acquire+0x848/0x1a68)
[ 130.816483] [<c0281a84>] (__lock_acquire) from [<c02835d8>] (lock_acquire+0xd8/0x22c)
[ 130.824326] [<c02835d8>] (lock_acquire) from [<c0c49948>] (_raw_spin_lock_irqsave+0x54/0x68)
[ 130.832777] [<c0c49948>] (_raw_spin_lock_irqsave) from [<c086bf54>] (dma_fence_add_callback+0x30/0x23c)
[ 130.842183] [<c086bf54>] (dma_fence_add_callback) from [<c086d4c8>] (dma_fence_array_enable_signaling+0x58/0xec)
[ 130.852371] [<c086d4c8>] (dma_fence_array_enable_signaling) from [<c086c00c>] (dma_fence_add_callback+0xe8/0x23c)
[ 130.862647] [<c086c00c>] (dma_fence_add_callback) from [<c06d8774>] (drm_syncobj_wait_ioctl+0x518/0x614)
[ 130.872143] [<c06d8774>] (drm_syncobj_wait_ioctl) from [<c06b8458>] (drm_ioctl_kernel+0xb0/0xf0)
[ 130.880940] [<c06b8458>] (drm_ioctl_kernel) from [<c06b8818>] (drm_ioctl+0x1d8/0x390)
[ 130.888782] [<c06b8818>] (drm_ioctl) from [<c03a4510>] (do_vfs_ioctl+0xb0/0x8ac)
[ 130.896187] [<c03a4510>] (do_vfs_ioctl) from [<c03a4d40>] (ksys_ioctl+0x34/0x60)
[ 130.903593] [<c03a4d40>] (ksys_ioctl) from [<c0201000>] (ret_fast_syscall+0x0/0x28)

Cc: Chunming Zhou <[email protected]>
Cc: Christian König <[email protected]>
Cc: Daniel Vetter <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Acked-by: Christian König <[email protected]>
Signed-off-by: Christian König <[email protected]>
Link: https://patchwork.freedesktop.org/patch/261044/

show more ...

drm/syncobj: disable the timeline UAPI for now v2

Until we have sorted out all problems.

v2: return -EINVAL during create if flag is set.

Signed-off-by: Christian König <[email protected]>

drm/syncobj: disable the timeline UAPI for now v2

Until we have sorted out all problems.

v2: return -EINVAL during create if flag is set.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/260937/

show more ...

drm: fix deadlock of syncobj v6

v2:
add a mutex between sync_cb execution and free.
v3:
clearly separating the roles for pt_lock and cb_mutex (Chris)
v4:
the cb_mutex should be taken outside of the

drm: fix deadlock of syncobj v6

v2:
add a mutex between sync_cb execution and free.
v3:
clearly separating the roles for pt_lock and cb_mutex (Chris)
v4:
the cb_mutex should be taken outside of the pt_lock around
this if() block. (Chris)
v5:
fix a corner case
v6:
tidy drm_syncobj_fence_get_or_add_callback up. (Chris)

Tested by syncobj_basic and syncobj_wait of igt.

Signed-off-by: Chunming Zhou <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Chris Wilson <[email protected]>
Cc: Christian König <[email protected]>
Cc: [email protected]
Reviewed-by: Chris Wilson <[email protected]>
Signed-off-by: Christian König <[email protected]>
Link: https://patchwork.kernel.org/patch/10652893/

show more ...

drm: add syncobj timeline support v9

This patch is for VK_KHR_timeline_semaphore extension, semaphore is called syncobj in kernel side:
This extension introduces a new type of syncobj that has an in

drm: add syncobj timeline support v9

This patch is for VK_KHR_timeline_semaphore extension, semaphore is called syncobj in kernel side:
This extension introduces a new type of syncobj that has an integer payload
identifying a point in a timeline. Such timeline syncobjs support the
following operations:
* CPU query - A host operation that allows querying the payload of the
timeline syncobj.
* CPU wait - A host operation that allows a blocking wait for a
timeline syncobj to reach a specified value.
* Device wait - A device operation that allows waiting for a
timeline syncobj to reach a specified value.
* Device signal - A device operation that allows advancing the
timeline syncobj to a specified value.

v1:
Since it's a timeline, that means the front time point(PT) always is signaled before the late PT.
a. signal PT design:
Signal PT fence N depends on PT[N-1] fence and signal opertion fence, when PT[N] fence is signaled,
the timeline will increase to value of PT[N].
b. wait PT design:
Wait PT fence is signaled by reaching timeline point value, when timeline is increasing, will compare
wait PTs value with new timeline value, if PT value is lower than timeline value, then wait PT will be
signaled, otherwise keep in list. syncobj wait operation can wait on any point of timeline,
so need a RB tree to order them. And wait PT could ahead of signal PT, we need a sumission fence to
perform that.

v2:
1. remove unused DRM_SYNCOBJ_CREATE_TYPE_NORMAL. (Christian)
2. move unexposed denitions to .c file. (Daniel Vetter)
3. split up the change to drm_syncobj_find_fence() in a separate patch. (Christian)
4. split up the change to drm_syncobj_replace_fence() in a separate patch.
5. drop the submission_fence implementation and instead use wait_event() for that. (Christian)
6. WARN_ON(point != 0) for NORMAL type syncobj case. (Daniel Vetter)

v3:
1. replace normal syncobj with timeline implemenation. (Vetter and Christian)
a. normal syncobj signal op will create a signal PT to tail of signal pt list.
b. normal syncobj wait op will create a wait pt with last signal point, and this wait PT is only signaled by related signal point PT.
2. many bug fix and clean up
3. stub fence moving is moved to other patch.

v4：
1. fix RB tree loop with while(node=rb_first(...)). (Christian)
2. fix syncobj lifecycle. (Christian)
3. only enable_signaling when there is wait_pt. (Christian)
4. fix timeline path issues.
5. write a timeline test in libdrm

v5: (Christian)
1. semaphore is called syncobj in kernel side.
2. don't need 'timeline' characters in some function name.
3. keep syncobj cb.

v6: (Christian)
1. merge syncobj_timeline to syncobj structure.
2. simplify some check sentences.
3. some misc change.
4. fix CTS failed issue.

v7: (Christian)
1. error handling when creating signal pt.
2. remove timeline naming in func.
3. export flags in find_fence.
4. allow reset timeline.

v8:
1. use wait_event_interruptible without timeout
2. rename _TYPE_INDIVIDUAL to _TYPE_BINARY

v9:
1. rename signal_pt->base to signal_pt->fence_array to avoid misleading
2. improve kerneldoc

individual syncobj is tested by ./deqp-vk -n dEQP-VK*semaphore*
timeline syncobj is tested by ./amdgpu_test -s 9

Signed-off-by: Chunming Zhou <[email protected]>
Signed-off-by: Christian König <[email protected]>
Cc: Christian Konig <[email protected]>
Cc: Dave Airlie <[email protected]>
Cc: Daniel Rakos <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Bas Nieuwenhuizen <[email protected]>
Cc: Jason Ekstrand <[email protected]>
Reviewed-by: Christian König <[email protected]>
Acked-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/257258/

show more ...

drm: add flags to drm_syncobj_find_fence

flags can be used by driver to decide whether need to block wait submission.

Signed-off-by: Chunming Zhou <[email protected]>
SIgned-off-by: Christian Kön

drm: add flags to drm_syncobj_find_fence

flags can be used by driver to decide whether need to block wait submission.

Signed-off-by: Chunming Zhou <[email protected]>
SIgned-off-by: Christian König <[email protected]>
Reviewed-by: Christian König <[email protected]>
Link: https://patchwork.kernel.org/patch/10641339/

show more ...

drm: expand replace_fence to support timeline point v2

we can place a fence to a timeline point after expanded.
v2: change func parameter order

Signed-off-by: Chunming Zhou <[email protected]>
Re

drm: expand replace_fence to support timeline point v2

we can place a fence to a timeline point after expanded.
v2: change func parameter order

Signed-off-by: Chunming Zhou <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Christian König <[email protected]>
Link: https://patchwork.freedesktop.org/patch/246543/

show more ...

drm: expand drm_syncobj_find_fence to support timeline point v2

we can fetch timeline point fence after expanded.
v2: The parameter fence is the result of the function and should come last.

Signed-

drm: expand drm_syncobj_find_fence to support timeline point v2

we can fetch timeline point fence after expanded.
v2: The parameter fence is the result of the function and should come last.

Signed-off-by: Chunming Zhou <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Christian König <[email protected]>
Link: https://patchwork.freedesktop.org/patch/246541/

show more ...

drm/syncobj: Drop add/remove_callback from driver interface

This is used for handling future fences. Currently no driver use
these, and I think given the new timeline fence proposed by KHR it
would

drm/syncobj: Drop add/remove_callback from driver interface

This is used for handling future fences. Currently no driver use
these, and I think given the new timeline fence proposed by KHR it
would be better to have a more abstract interface for future fences.
Could be something simple like a struct dma_future_fence plus a
function to add a callback or wait for the fence to materialize.

Then syncobj (and anything else really) could grow new functions to
expose these two drivers. Normal dma_fence would then keep the nice
guarantee that they will always signal (and through ordering, be
deadlock free). dma_future_fence would then be the tricky one.

This also fixes sphinx complaining about the kerneldoc.

Cc: Jason Ekstrand <[email protected]>
Cc: Dave Airlie <[email protected]>
Cc: Chris Wilson <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

show more ...

drm/syncobj: some kerneldoc polish

Complete a few missing bits, fix up the existing xcross-references and
add a bunch more.

v2: Fix typos (Alex).

Cc: Dave Airlie <[email protected]> via lists.free

drm/syncobj: some kerneldoc polish

Complete a few missing bits, fix up the existing xcross-references and
add a bunch more.

v2: Fix typos (Alex).

Cc: Dave Airlie <[email protected]> via lists.freedesktop.org
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

show more ...

drm/syncobj: Mark up the fence as an RCU protected pointer

We take advantage of that syncobj->fence is an RCU-protected pointer, and
so sparse complains that it is lacking annotation.

Cc: Dave Airl

drm/syncobj: Mark up the fence as an RCU protected pointer

We take advantage of that syncobj->fence is an RCU-protected pointer, and
so sparse complains that it is lacking annotation.

Cc: Dave Airlie <[email protected]>
Cc: Jason Ekstrand <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: Alex Deucher <[email protected]>
Cc: Christian König <[email protected]>
Cc: Sumit Semwal <[email protected]>
Signed-off-by: Chris Wilson <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>
Acked-by: Christian König <[email protected]>
Signed-off-by: Ville Syrjälä <[email protected]>

show more ...

drm/syncobj: add a new helper drm_syncobj_get_fd

Reviewed-by: Dave Airlie <[email protected]>
Signed-off-by: Marek Olšák <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/syncobj: extract two helpers from drm_syncobj_create

For amdgpu.

drm_syncobj_create is renamed to drm_syncobj_create_as_handle, and new
helpers drm_syncobj_create and drm_syncobj_get_handle are

drm/syncobj: extract two helpers from drm_syncobj_create

For amdgpu.

drm_syncobj_create is renamed to drm_syncobj_create_as_handle, and new
helpers drm_syncobj_create and drm_syncobj_get_handle are added.

Reviewed-by: Dave Airlie <[email protected]>
Signed-off-by: Marek Olšák <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

show more ...

drm/syncobj: Add a callback mechanism for replace_fence (v3)

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 f

drm/syncobj: Add a callback mechanism for replace_fence (v3)

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]>

show more ...

drm/syncobj: Add a race-free drm_syncobj_fence_get helper (v2)

The atomic exchange operation in drm_syncobj_replace_fence is sufficient
for the case where it races with itself. However, if you have

drm/syncobj: Add a race-free drm_syncobj_fence_get helper (v2)

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]>

show more ...

drm/syncobj: Rename fence_get to find_fence

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]>
Ac

drm/syncobj: Rename fence_get to find_fence

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]>

show more ...

drm: Remove unused drm_file parameter to drm_syncobj_replace_fence()

the drm_file parameter is unused, so remove it.

Signed-off-by: Chris Wilson <[email protected]>
Cc: Dave Airlie <airlied@

drm: Remove unused drm_file parameter to drm_syncobj_replace_fence()

the drm_file parameter is unused, so remove it.

Signed-off-by: Chris Wilson <[email protected]>
Cc: Dave Airlie <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>

show more ...

drm: introduce sync objects (v4)

Sync objects are new toplevel drm object, that contain a
pointer to a fence. This fence can be updated via command
submission ioctls via drivers.

There is also a ge

drm: introduce sync objects (v4)

Sync objects are new toplevel drm object, that contain a
pointer to a fence. This fence can be updated via command
submission ioctls via drivers.

There is also a generic wait obj API modelled on the vulkan
wait API (with code modelled on some amdgpu code).

These objects can be converted to an opaque fd that can be
passes between processes.

v2: rename reference/unreference to put/get (Chris)
fix leaked reference (David Zhou)
drop mutex in favour of cmpxchg (Chris)
v3: cleanups from danvet, rebase on drm_fops rename
check fd_flags is 0 in ioctls.
v4: export find/free, change replace fence to take a
syncobj. In order to support lookup first, replace
later semantics which seem in the end to be cleaner.

Reviewed-by: Sean Paul <[email protected]>
Reviewed-by: Chris Wilson <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>

show more ...