We need to be able to prevent memory from being freed while it's still
mapped in a GPU's address-space.
Will be used by upcoming MMU changes.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Map flags (access, kind, etc) are currently defined in either the VMA,
or the memory object, which turns out to not be ideal for things like
suballocated buffers, etc.
These will become per-map flags instead, so we need to support passing
these arguments in nvkm_memory_map().
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Upcoming changes will remove the nvkm_vmm pointer from nvkm_vma, instead
requiring it to be explicitly specified on each operation.
It's not currently possible to get this information for BAR1 mappings,
so let's fix that ahead of time.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Presumably we can never actually hit this return, but static checkers
complain that we should unlock before we return.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
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>
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>
It appears these don't map to PBDMAs (at least on Kepler, it may or may
be valid for Fermi - this hasn't been checked), but to runlists.
This drops the NVKM_ENGINE_FIFO data from the entries too, as resetting
all of PFIFO is *not* the way to handle such faults.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Bits 28:29 of RUNLIST_BASE specify the memory target of the runlist. Set
it to 0x3 (SYS_MEM_NONCOHERENT) if the runlist object resides in system
memory.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Fix the channel id bit mask in FIFO schedule timeout error handling.
FIFO_ENGINE_STATUS_NEXT_ID is bit 27:16 thus 0x0fff0000.
FIFO_ENGINE_STATUS_ID is bit 11:0 thus 0x00000fff.
Signed-off-by: Xia Yang <xiay@nvidia.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The CPU-side tracking of engine runlists was not protected by a lock,
leading to list corruption, eventually causing runlist_update() to
overrun the GPU-side runlist, triggering an OOPS.
Fixes some of the issues noticed during parallel piglit runs.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
