Greatly improves the chances of recovering the GPU from a CTXSW_TIMEOUT.
Tested with piglit's arb_shader_image_load_store-atomicity, which causes
GR to hang in such a way that recovery failed (CTXSW_TIMEOUT continually
re-triggers).
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This will serve as a basis for implementing some improvements to how
we recover the GPU from channel errors.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The previous commit simply changes the interface, but should result in
the same behaviour as previously. This commit has been split out from
it as it can result in a different channel being selected.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
There are instances (such as non-recoverable GPU page faults) where
NVKM decides that a channel's context is no longer viable, and will
be removed from the runlist.
This commit notifies the owner of the channel when this happens, so
it has the opportunity to take some kind of recovery action instead
of hanging.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Tested on a G92, seems to work. Confirmed by 8 mmiotraces.
Signed-off-by: Karol Herbst <karolherbst@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
To make the code clearer, use rb_entry() instead of container_of() to
deal with rbtree.
Signed-off-by: Geliang Tang <geliangtang@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Create instances for the FECS and GPCCS falcons and use the init() and
fini() hooks to reserve them for as long as GR controls them.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
gf100_gr_init_ctxctl() is basically two different functions (one for
use of internal firmware, the other for use of external firmware), but
its current layout makes it look more complex than it is. Split it to
better reflect that fact.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Add a fini() hook to the GR engine. This will be used by gf100+ to
properly release the FECS and GPCCS falcons.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Add a dummy PMU device so the PMU falcon is instanciated and can be used
by secure boot.
We could reuse gk20a's implementation here, but it would fight with
secboot over PMU falcon's ownership and secboot will reset the PMU,
preventing it from operating afterwards. Proper handout between secboot
and pmu is coming along with the actual gm20b PMU implementation, so
use this as a temporary solution.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The workaround appears to cause regressions on these boards, and from
inspection of RM traces, NVIDIA don't appear to do it on them either.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Tested-by: Roy Spliet <nouveau@spliet.org>
This has been on the TODO list for a while now, recovering from things
such as attempting to execute a push buffer or touch a semaphore in an
unmapped memory area.
The only thing required on the HW side here is that the offending
channel is removed from the runlist, and *not* a full reset of PFIFO.
This used to be a bit messier to handle before the rework to make use
of engine topology info, but is apparently now trivial.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
From visual inspection of traces, what we currently implement appears to
be identical to GP104. Seems to work well enough too.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This avoids an issue that occurs when we're attempting to preempt multiple
channels simultaneously. HW seems to ignore preempt requests while it's
still processing a previous one, which, well, makes sense.
Fixes random "fifo: SCHED_ERROR 0d []" + GPCCS page faults during parallel
piglit runs on (at least) GM107.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
Look for firmware files using the legacy ("nouveau/nvxx_fucxxxx") path
if they cannot be found in the new, "official" path. User setups were
broken by the switch, which is bad.
There are only 4 firmware files we may want to look up that way, so
hardcode them into the lookup function. All new firmware files should
use the standard "nvidia/<chip>/gr/" path.
Fixes: 8539b37ace ("drm/nouveau/gr: use NVIDIA-provided external firmwares")
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
