The current implementation of drm_sched_start uses a hardcoded
-ECANCELED to dispose of a job when the parent/hw fence is NULL.
This results in drm_sched_job_done being called with -ECANCELED for
each job with a NULL parent in the pending list, making it difficult
to distinguish between recovery methods, whether a queue reset or a
full GPU reset was used.
To improve this, we first try a soft recovery for timeout jobs and
use the error code -ENODATA. If soft recovery fails, we proceed with
a queue reset, where the error code remains -ENODATA for the job.
Finally, for a full GPU reset, we use error codes -ECANCELED or
-ETIME. This patch adds an error code parameter to drm_sched_start,
allowing us to differentiate between queue reset and GPU reset
failures. This enables user mode and test applications to validate
the expected correctness of the requested operation. After a
successful queue reset, the only way to continue normal operation is
to call drm_sched_job_done with the specific error code -ENODATA.
v1: Initial implementation by Jesse utilized amdgpu_device_lock_reset_domain
and amdgpu_device_unlock_reset_domain to allow user mode to track
the queue reset status and distinguish between queue reset and
GPU reset.
v2: Christian suggested using the error codes -ENODATA for queue reset
and -ECANCELED or -ETIME for GPU reset, returned to
amdgpu_cs_wait_ioctl.
v3: To meet the requirements, we introduce a new function
drm_sched_start_ex with an additional parameter to set
dma_fence_set_error, allowing us to handle the specific error
codes appropriately and dispose of bad jobs with the selected
error code depending on whether it was a queue reset or GPU reset.
v4: Alex suggested using a new name, drm_sched_start_with_recovery_error,
which more accurately describes the function's purpose.
Additionally, it was recommended to add documentation details
about the new method.
v5: Fixed declaration of new function drm_sched_start_with_recovery_error.(Alex)
v6 (chk): rebase on upstream changes, cleanup the commit message,
drop the new function again and update all callers,
apply the errno also to scheduler fences with hw fences
v7 (chk): rebased
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Vitaly Prosyak <vitaly.prosyak@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240826122541.85663-1-christian.koenig@amd.com
-----BEGIN PGP SIGNATURE-----
iQFSBAABCAA8FiEEq68RxlopcLEwq+PEeb4+QwBBGIYFAmbK2B8eHHRvcnZhbGRz
QGxpbnV4LWZvdW5kYXRpb24ub3JnAAoJEHm+PkMAQRiGFwkH/10QpUgzIfbFKbF+
5hwcvaqS5myxWwJ4PjN0eR1qGE6RzVO0Tb24+TVql+7pxu+iWm1kYgC3+/T5xJsP
ECAszdmPWSco1xaHrh2y3PyCJjaBiqFbIxdjPp7odjDpG9qarbcty8YpWs44u/gd
RDXzHUuScEShBhEt0ZhvE1pIDL8jJ8JL3yqOMZ+XaDxtJbjaHw4GHp8efxlBWc8N
jZKIVJi22q5NWG5T0tGtPWwzCm0ewA/JNMTEvE9leoSoAgO85NZ0ivxMC76q/tbj
BrYk5KnzfhJs4b/n/KtIwWaLTgLyXKGqHMaMq8sbXtp410aUdgnRJO2cl3fI+1vc
vxQfAfk=
=RemI
-----END PGP SIGNATURE-----
Merge v6.11-rc5 into drm-next
amdgpu pr conconflicts due to patches cherry-picked to -fixes, I might
as well catch up with a backmerge and handle them all. Plus both misc
and intel maintainers asked for a backmerge anyway.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
When enabling UBSAN on Raspberry Pi 5, we get the following warning:
[ 387.894977] UBSAN: array-index-out-of-bounds in drivers/gpu/drm/v3d/v3d_sched.c:320:3
[ 387.903868] index 7 is out of range for type '__u32 [7]'
[ 387.909692] CPU: 0 PID: 1207 Comm: kworker/u16:2 Tainted: G WC 6.10.3-v8-16k-numa #151
[ 387.919166] Hardware name: Raspberry Pi 5 Model B Rev 1.0 (DT)
[ 387.925961] Workqueue: v3d_csd drm_sched_run_job_work [gpu_sched]
[ 387.932525] Call trace:
[ 387.935296] dump_backtrace+0x170/0x1b8
[ 387.939403] show_stack+0x20/0x38
[ 387.942907] dump_stack_lvl+0x90/0xd0
[ 387.946785] dump_stack+0x18/0x28
[ 387.950301] __ubsan_handle_out_of_bounds+0x98/0xd0
[ 387.955383] v3d_csd_job_run+0x3a8/0x438 [v3d]
[ 387.960707] drm_sched_run_job_work+0x520/0x6d0 [gpu_sched]
[ 387.966862] process_one_work+0x62c/0xb48
[ 387.971296] worker_thread+0x468/0x5b0
[ 387.975317] kthread+0x1c4/0x1e0
[ 387.978818] ret_from_fork+0x10/0x20
[ 387.983014] ---[ end trace ]---
This happens because the UAPI provides only seven configuration
registers and we are reading the eighth position of this u32 array.
Therefore, fix the out-of-bounds read in `v3d_csd_job_run()` by
accessing only seven positions on the '__u32 [7]' array. The eighth
register exists indeed on V3D 7.1, but it isn't currently used. That
being so, let's guarantee that it remains unused and add a note that it
could be set in a future patch.
Fixes: 0ad5bc1ce4 ("drm/v3d: fix up register addresses for V3D 7.x")
Reported-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240809152001.668314-1-mcanal@igalia.com
This was basically just another one of amdgpus hacks. The parameter
allowed to restart the scheduler without turning fence signaling on
again.
That this is absolutely not a good idea should be obvious by now since
the fences will then just sit there and never signal.
While at it cleanup the code a bit.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240722083816.99685-1-christian.koenig@amd.com
If fetching of userspace memory fails during the main loop, all drm sync
objs looked up until that point will be leaked because of the missing
drm_syncobj_put.
Fix it by exporting and using a common cleanup helper.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Fixes: bae7cb5d68 ("drm/v3d: Create a CPU job extension for the reset performance query job")
Cc: Maíra Canal <mcanal@igalia.com>
Cc: Iago Toral Quiroga <itoral@igalia.com>
Cc: stable@vger.kernel.org # v6.8+
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240711135340.84617-4-tursulin@igalia.com
(cherry picked from commit 484de39fa5)
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
If fetching of userspace memory fails during the main loop, all drm sync
objs looked up until that point will be leaked because of the missing
drm_syncobj_put.
Fix it by exporting and using a common cleanup helper.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Fixes: 9ba0ff3e08 ("drm/v3d: Create a CPU job extension for the timestamp query job")
Cc: Maíra Canal <mcanal@igalia.com>
Cc: Iago Toral Quiroga <itoral@igalia.com>
Cc: stable@vger.kernel.org # v6.8+
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240711135340.84617-3-tursulin@igalia.com
(cherry picked from commit 753ce4fea6)
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
`args->cfg[4]` is configured in Indirect Dispatch using the number of
batches. Currently, for all V3D tech versions, `args->cfg[4]` equals the
number of batches subtracted by 1. But, for V3D 7.1.6 and later, we must not
subtract 1 from the number of batches.
Implement the fix by checking the V3D tech version and revision.
Fixes several `dEQP-VK.synchronization*` CTS tests related to Indirect Dispatch.
Fixes: 18b8413b25 ("drm/v3d: Create a CPU job extension for a indirect CSD job")
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240714145243.1223131-2-mcanal@igalia.com
Removing the intermediate buffer removes the last use of the
V3D_MAX_COUNTERS define, which will enable further driver cleanup.
While at it pull the 32 vs 64 bit copying decision outside the loop in
order to reduce the number of conditional instructions.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240711135340.84617-9-tursulin@igalia.com
Instead of statically reserving pessimistic space for the kperfmon_ids
array, make the userspace extension code allocate the exactly required
amount of space.
Apart from saving some memory at runtime, this also removes the need for
the V3D_MAX_PERFMONS macro whose removal will benefit further driver
cleanup.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240711135340.84617-8-tursulin@igalia.com
If fetching of userspace memory fails during the main loop, all drm sync
objs looked up until that point will be leaked because of the missing
drm_syncobj_put.
Fix it by exporting and using a common cleanup helper.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Fixes: bae7cb5d68 ("drm/v3d: Create a CPU job extension for the reset performance query job")
Cc: Maíra Canal <mcanal@igalia.com>
Cc: Iago Toral Quiroga <itoral@igalia.com>
Cc: stable@vger.kernel.org # v6.8+
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240711135340.84617-4-tursulin@igalia.com
If fetching of userspace memory fails during the main loop, all drm sync
objs looked up until that point will be leaked because of the missing
drm_syncobj_put.
Fix it by exporting and using a common cleanup helper.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Fixes: 9ba0ff3e08 ("drm/v3d: Create a CPU job extension for the timestamp query job")
Cc: Maíra Canal <mcanal@igalia.com>
Cc: Iago Toral Quiroga <itoral@igalia.com>
Cc: stable@vger.kernel.org # v6.8+
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240711135340.84617-3-tursulin@igalia.com
V3D_PERFCNT_NUM represents the maximum number of performance counters
for V3D 4.2, but not for V3D 7.1. This means that, if we use
V3D_PERFCNT_NUM, we might go out-of-bounds on V3D 7.1.
Therefore, use the number of performance counters on V3D 7.1 as the
maximum number of counters. This will allow us to create arrays on the
stack with reasonable size. Note that userspace must use the value
provided by DRM_V3D_PARAM_MAX_PERF_COUNTERS.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240512222655.2792754-6-mcanal@igalia.com
In V3D, the conclusion of a job is indicated by a IRQ. When a job
finishes, then we update the local and the global GPU stats of that
queue. But, while the GPU stats are being updated, a user might be
reading the stats from sysfs or fdinfo.
For example, on `gpu_stats_show()`, we could think about a scenario where
`v3d->queue[queue].start_ns != 0`, then an interrupt happens, we update
the value of `v3d->queue[queue].start_ns` to 0, we come back to
`gpu_stats_show()` to calculate `active_runtime` and now,
`active_runtime = timestamp`.
In this simple example, the user would see a spike in the queue usage,
that didn't match reality.
In order to address this issue properly, use a seqcount to protect read
and write sections of the code.
Fixes: 09a93cc4f7 ("drm/v3d: Implement show_fdinfo() callback for GPU usage stats")
Reported-by: Tvrtko Ursulin <tursulin@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240420213632.339941-7-mcanal@igalia.com
Given a set of GPU stats, that is, a `struct v3d_stats` related to a
queue in a given context, create a function that can update this set
of GPU stats.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240420213632.339941-5-mcanal@igalia.com
This will make it easier to instantiate the GPU stats variables and it
will create a structure where we can store all the variables that refer
to GPU stats.
Note that, when we created the struct `v3d_stats`, we renamed
`jobs_sent` to `jobs_completed`. This better express the semantics of
the variable, as we are only accounting jobs that have been completed.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240420213632.339941-4-mcanal@igalia.com
Currently, we manually perform all operations to update the GPU stats
variables. Apart from the code repetition, this is very prone to errors,
as we can see on commit 35f4f8c9fc ("drm/v3d: Don't increment
`enabled_ns` twice").
Therefore, create two functions to manage updating all GPU stats
variables. Now, the jobs only need to call for `v3d_job_update_stats()`
when the job is done and `v3d_job_start_stats()` when starting the job.
Co-developed-by: Tvrtko Ursulin <tursulin@igalia.com>
Signed-off-by: Tvrtko Ursulin <tursulin@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240420213632.339941-3-mcanal@igalia.com
A CPU job is a type of job that performs operations that requires CPU
intervention. A copy performance query job is a job that copy the complete
or partial result of a query to a buffer. In order to copy the result of
a performance query to a buffer, we need to get the values from the
performance monitors.
So, create a user extension for the CPU job that enables the creation
of a copy performance query job. This user extension will allow the creation
of a CPU job that copy the results of a performance query to a BO with the
possibility to indicate the availability with a availability bit.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231130164420.932823-19-mcanal@igalia.com
A CPU job is a type of job that performs operations that requires CPU
intervention. A reset performance query job is a job that resets the
performance queries by resetting the values of the perfmons. Moreover,
we also reset the syncobjs related to the availability of the query.
So, create a user extension for the CPU job that enables the creation
of a reset performance job. This user extension will allow the creation of
a CPU job that resets the perfmons values and resets the availability syncobj.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231130164420.932823-18-mcanal@igalia.com
A CPU job is a type of job that performs operations that requires CPU
intervention. A copy timestamp query job is a job that copy the complete
or partial result of a query to a buffer. As V3D doesn't provide any
mechanism to obtain a timestamp from the GPU, it is a job that needs
CPU intervention.
So, create a user extension for the CPU job that enables the creation
of a copy timestamp query job. This user extension will allow the creation
of a CPU job that copy the results of a timestamp query to a BO with the
possibility to indicate the timestamp availability with a availability bit.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231130164420.932823-17-mcanal@igalia.com
A CPU job is a type of job that performs operations that requires CPU
intervention. A reset timestamp job is a job that resets the timestamp
queries based on the value offset of the first query. As V3D doesn't
provide any mechanism to obtain a timestamp from the GPU, it is a job
that needs CPU intervention.
So, create a user extension for the CPU job that enables the creation
of a reset timestamp job. This user extension will allow the creation of
a CPU job that resets the timestamp value in the timestamp BO and resets
the availability syncobj.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231130164420.932823-16-mcanal@igalia.com
A CPU job is a type of job that performs operations that requires CPU
intervention. A timestamp query job is a job that calculates the
query timestamp and updates the query availability by signaling a
syncobj. As V3D doesn't provide any mechanism to obtain a timestamp
from the GPU, it is a job that needs CPU intervention.
So, create a user extension for the CPU job that enables the creation
of a timestamp query job. This user extension will allow the creation of
a CPU job that performs the timestamp query calculation and updates the
timestamp BO with the proper value.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231130164420.932823-15-mcanal@igalia.com
A CPU job is a type of job that performs operations that requires CPU
intervention. An indirect CSD job is a job that, when executed in the
queue, will map the indirect buffer, read the dispatch parameters, and
submit a regular dispatch. Therefore, it is a job that needs CPU
intervention.
So, create a user extension for the CPU job that enables the creation
of an indirect CSD. This user extension will allow the creation of a CSD
job linked to a CPU job. The CPU job will wait for the indirect CSD job
dependencies and, once they are signaled, it will update the CSD job
parameters.
Co-developed-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231130164420.932823-14-mcanal@igalia.com
Create tracepoints to track the three major events of a CPU job
lifetime:
1. Submission of a `v3d_submit_cpu` IOCTL
2. Beginning of the execution of a CPU job
3. Ending of the execution of a CPU job
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231130164420.932823-11-mcanal@igalia.com
Create a new type of job, a CPU job. A CPU job is a type of job that
performs operations that requires CPU intervention. The overall idea is
to use user extensions to enable different types of CPU job, allowing the
CPU job to perform different operations according to the type of user
extension. The user extension ID identify the type of CPU job that must
be dealt.
Having a CPU job is interesting for synchronization purposes as a CPU
job has a queue like any other V3D job and can be synchoronized by the
multisync extension.
Signed-off-by: Melissa Wen <mwen@igalia.com>
Co-developed-by: Maíra Canal <mcanal@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231130164420.932823-9-mcanal@igalia.com
The previous patch exposed the accumulated amount of active time per
client for each V3D queue. But this doesn't provide a global notion of
the GPU usage.
Therefore, provide the accumulated amount of active time for each V3D
queue (BIN, RENDER, CSD, TFU and CACHE_CLEAN), considering all the jobs
submitted to the queue, independent of the client.
This data is exposed through the sysfs interface, so that if the
interface is queried at two different points of time the usage percentage
of each of the queues can be calculated.
Co-developed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Signed-off-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Acked-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Melissa Wen <mwen@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230905213416.1290219-3-mcanal@igalia.com
This patch exposes the accumulated amount of active time per client
through the fdinfo infrastructure. The amount of active time is exposed
for each V3D queue: BIN, RENDER, CSD, TFU and CACHE_CLEAN.
In order to calculate the amount of active time per client, a CPU clock
is used through the function local_clock(). The point where the jobs has
started is marked and is finally compared with the time that the job had
finished.
Moreover, the number of jobs submitted to each queue is also exposed on
fdinfo through the identifier "v3d-jobs-<queue>".
Co-developed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Signed-off-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Acked-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Melissa Wen <mwen@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230905213416.1290219-3-mcanal@igalia.com
This patch updates a number of register addresses that have
been changed in Raspberry Pi 5 (V3D 7.1) and updates the
code to use the corresponding registers and addresses based
on the actual V3D version.
Signed-off-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231031073859.25298-3-itoral@igalia.com
In Xe, the new Intel GPU driver, a choice has made to have a 1 to 1
mapping between a drm_gpu_scheduler and drm_sched_entity. At first this
seems a bit odd but let us explain the reasoning below.
1. In Xe the submission order from multiple drm_sched_entity is not
guaranteed to be the same completion even if targeting the same hardware
engine. This is because in Xe we have a firmware scheduler, the GuC,
which allowed to reorder, timeslice, and preempt submissions. If a using
shared drm_gpu_scheduler across multiple drm_sched_entity, the TDR falls
apart as the TDR expects submission order == completion order. Using a
dedicated drm_gpu_scheduler per drm_sched_entity solve this problem.
2. In Xe submissions are done via programming a ring buffer (circular
buffer), a drm_gpu_scheduler provides a limit on number of jobs, if the
limit of number jobs is set to RING_SIZE / MAX_SIZE_PER_JOB we get flow
control on the ring for free.
A problem with this design is currently a drm_gpu_scheduler uses a
kthread for submission / job cleanup. This doesn't scale if a large
number of drm_gpu_scheduler are used. To work around the scaling issue,
use a worker rather than kthread for submission / job cleanup.
v2:
- (Rob Clark) Fix msm build
- Pass in run work queue
v3:
- (Boris) don't have loop in worker
v4:
- (Tvrtko) break out submit ready, stop, start helpers into own patch
v5:
- (Boris) default to ordered work queue
v6:
- (Luben / checkpatch) fix alignment in msm_ringbuffer.c
- (Luben) s/drm_sched_submit_queue/drm_sched_wqueue_enqueue
- (Luben) Update comment for drm_sched_wqueue_enqueue
- (Luben) Positive check for submit_wq in drm_sched_init
- (Luben) s/alloc_submit_wq/own_submit_wq
v7:
- (Luben) s/drm_sched_wqueue_enqueue/drm_sched_run_job_queue
v8:
- (Luben) Adjust var names / comments
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
Link: https://lore.kernel.org/r/20231031032439.1558703-3-matthew.brost@intel.com
Signed-off-by: Luben Tuikov <ltuikov89@gmail.com>
The GPU scheduler has now a variable number of run-queues, which are set up at
drm_sched_init() time. This way, each driver announces how many run-queues it
requires (supports) per each GPU scheduler it creates. Note, that run-queues
correspond to scheduler "priorities", thus if the number of run-queues is set
to 1 at drm_sched_init(), then that scheduler supports a single run-queue,
i.e. single "priority". If a driver further sets a single entity per
run-queue, then this creates a 1-to-1 correspondence between a scheduler and
a scheduled entity.
Cc: Lucas Stach <l.stach@pengutronix.de>
Cc: Russell King <linux+etnaviv@armlinux.org.uk>
Cc: Qiang Yu <yuq825@gmail.com>
Cc: Rob Clark <robdclark@gmail.com>
Cc: Abhinav Kumar <quic_abhinavk@quicinc.com>
Cc: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Cc: Danilo Krummrich <dakr@redhat.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Boris Brezillon <boris.brezillon@collabora.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Emma Anholt <emma@anholt.net>
Cc: etnaviv@lists.freedesktop.org
Cc: lima@lists.freedesktop.org
Cc: linux-arm-msm@vger.kernel.org
Cc: freedreno@lists.freedesktop.org
Cc: nouveau@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20231023032251.164775-1-luben.tuikov@amd.com
Remove redundant error message (since now it is very similar to what
we do in drm_sched_init) and centralize all error handling in a
unique place, as we follow the same steps in any case of failure.
Signed-off-by: Melissa Wen <mwen@igalia.com>
Acked-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Melissa Wen <melissa.srw@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220228181647.3794298-1-mwen@igalia.com
Add device pointer so scheduler's printing can use
DRM_DEV_ERROR() instead, which makes life easier under multiple GPU
scenario.
v2: amend all calls of drm_sched_init()
v3: fill dev pointer for all drm_sched_init() calls
Signed-off-by: Jiawei Gu <Jiawei.Gu@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220221095705.5290-1-Jiawei.Gu@amd.com
With the prep work out of the way this isn't tricky anymore.
Aside: The chaining of the various jobs is a bit awkward, with the
possibility of failure in bad places. I think with the
drm_sched_job_init/arm split and maybe preloading the
job->dependencies xarray this should be fixable.
v2: Rebase over renamed function names for adding dependencies.
Reviewed-by: Melissa Wen <mwen@igalia.com> (v1)
Acked-by: Emma Anholt <emma@anholt.net>
Cc: Melissa Wen <melissa.srw@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Emma Anholt <emma@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20210805104705.862416-11-daniel.vetter@ffwll.ch
Prep work for using the scheduler dependency handling. We need to call
drm_sched_job_init earlier so we can use the new drm_sched_job_await*
functions for dependency handling here.
v2: Slightly better commit message and rebase to include the
drm_sched_job_arm() call (Emma).
v3: Cleanup jobs under construction correctly (Emma)
v4: Rebase over perfmon patch
Reviewed-by: Melissa Wen <mwen@igalia.com> (v3)
Acked-by: Emma Anholt <emma@anholt.net>
Cc: Melissa Wen <melissa.srw@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Emma Anholt <emma@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20210805104705.862416-10-daniel.vetter@ffwll.ch
The V3D engine has several hardware performance counters that can of
interest for userspace performance analysis tools.
This exposes new ioctls to create and destroy performance monitor
objects, as well as to query the counter values.
Each created performance monitor object has an ID that can be attached
to CL/CSD submissions, so the driver enables the requested counters when
the job is submitted, and updates the performance monitor values when
the job is done.
It is up to the user to ensure all the jobs have been finished before
getting the performance monitor values. It is also up to the user to
properly synchronize BCL jobs when submitting jobs with different
performance monitors attached.
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: David Airlie <airlied@linux.ie>
Cc: Emma Anholt <emma@anholt.net>
To: dri-devel@lists.freedesktop.org
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Acked-by: Melissa Wen <mwen@igalia.com>
Signed-off-by: Melissa Wen <melissa.srw@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210608111541.461991-1-jasuarez@igalia.com
Mali Midgard/Bifrost GPUs have 3 hardware queues but only a global GPU
reset. This leads to extra complexity when we need to synchronize timeout
works with the reset work. One solution to address that is to have an
ordered workqueue at the driver level that will be used by the different
schedulers to queue their timeout work. Thanks to the serialization
provided by the ordered workqueue we are guaranteed that timeout
handlers are executed sequentially, and can thus easily reset the GPU
from the timeout handler without extra synchronization.
v5:
* Add a new paragraph to the timedout_job() method
v3:
* New patch
v4:
* Actually use the timeout_wq to queue the timeout work
Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Christian König <christian.koenig@amd.com>
Cc: Qiang Yu <yuq825@gmail.com>
Cc: Emma Anholt <emma@anholt.net>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210630062751.2832545-3-boris.brezillon@collabora.com
This patch does not change current behaviour.
The driver's job timeout handler now returns
status indicating back to the DRM layer whether
the device (GPU) is no longer available, such as
after it's been unplugged, or whether all is
normal, i.e. current behaviour.
All drivers which make use of the
drm_sched_backend_ops' .timedout_job() callback
have been accordingly renamed and return the
would've-been default value of
DRM_GPU_SCHED_STAT_NOMINAL to restart the task's
timeout timer--this is the old behaviour, and is
preserved by this patch.
v2: Use enum as the status of a driver's job
timeout callback method.
v3: Return scheduler/device information, rather
than task information.
Cc: Alexander Deucher <Alexander.Deucher@amd.com>
Cc: Andrey Grodzovsky <Andrey.Grodzovsky@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Lucas Stach <l.stach@pengutronix.de>
Cc: Russell King <linux+etnaviv@armlinux.org.uk>
Cc: Christian Gmeiner <christian.gmeiner@gmail.com>
Cc: Qiang Yu <yuq825@gmail.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Cc: Steven Price <steven.price@arm.com>
Cc: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Cc: Eric Anholt <eric@anholt.net>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Steven Price <steven.price@arm.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/415095/
Fixes the following W=1 kernel build warning(s):
drivers/gpu/drm/v3d/v3d_sched.c:75: warning: Function parameter or member 'sched_job' not described in 'v3d_job_dependency'
drivers/gpu/drm/v3d/v3d_sched.c:75: warning: Function parameter or member 's_entity' not described in 'v3d_job_dependency'
Cc: Eric Anholt <eric@anholt.net>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: dri-devel@lists.freedesktop.org
Cc: linux-media@vger.kernel.org
Cc: linaro-mm-sig@lists.linaro.org
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20201116174112.1833368-37-lee.jones@linaro.org
We already have it in v3d_dev->drm.dev with zero additional pointer
chasing. Personally I don't like duplicated pointers like this
because:
- reviewers need to check whether the pointer is for the same or
different objects if there's multiple
- compilers have an easier time too
But also a bit a bikeshed, so feel free to ignore.
Acked-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Eric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20200415074034.175360-10-daniel.vetter@ffwll.ch
We now destroy finished jobs from the worker thread to make sure that
we never destroy a job currently in timeout processing.
By this we avoid holding lock around ring mirror list in drm_sched_stop
which should solve a deadlock reported by a user.
v2: Remove unused variable.
v4: Move guilty job free into sched code.
v5:
Move sched->hw_rq_count to drm_sched_start to account for counter
decrement in drm_sched_stop even when we don't call resubmit jobs
if guily job did signal.
v6: remove unused variable
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109692
Acked-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1555599624-12285-3-git-send-email-andrey.grodzovsky@amd.com
It is the expectation of existing userspace (X11 + Mesa, in
particular) that jobs submitted to the kernel against a shared BO will
get implicitly synchronized by their submission order. If we want to
allow clever userspace to disable implicit synchronization, we should
do that under its own submit flag (as amdgpu and lima do).
Note that we currently only implicitly sync for the rendering pass,
not binning -- if you texture-from-pixmap in the binning vertex shader
(vertex coordinate generation), you'll miss out on synchronization.
Fixes flickering when multiple clients are running in parallel,
particularly GL apps and compositors.
v2: Fix a missing refcount on the CSD done fence for L2 cleaning.
Signed-off-by: Eric Anholt <eric@anholt.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20190416225856.20264-6-eric@anholt.net
Acked-by: Rob Clark <robdclark@gmail.com>