We cached the number of vma bound to the object in order to speed up
shrinker decisions. This has been superseded by being more proactive in
removing objects we cannot shrink from the shrinker lists, and so we can
drop the clumsy attempt at atomically counting the bind count and
comparing it to the number of pinned mappings of the object. This will
only get more clumsier with asynchronous binding and unbinding.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200401223924.16667-1-chris@chris-wilson.co.uk
This is really just an alias of mmap_gtt. The 'mmap offset' nomenclature
comes from the value returned by this ioctl which is the offset into the
device fd which userpace uses with mmap(2).
mmap_gtt was our initial mmap_offset implementation, this extends
our CPU mmap support to allow additional fault handlers that depends on
the object's backing pages.
Note that we multiplex mmap_gtt and mmap_offset through the same ioctl,
and use the zero extending behaviour of drm to differentiate between
them, when we inspect the flags.
To support multiple mmap types on an object we need to support multiple
mmap_offsets for an object (each offset in the global device address
space corresponding to a unique instance of the object for a file + mmap
type). As we drop the simplified drm core idea of a single mmap_offset,
we need to provide replacement hooks for the dumb mmap interface as
well.
Link: https://gitlab.freedesktop.org/mesa/mesa/merge_requests/1675
Testcase: igt/gem_mmap_offset
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20191204120032.3682839-1-chris@chris-wilson.co.uk
As the engine->kernel_context is used within the engine-pm barrier, we
have to be careful when emitting requests outside of the barrier, as the
strict timeline locking rules do not apply. Instead, we must ensure the
engine_park() cannot be entered as we build the request, which is
simplest by taking an explicit engine-pm wakeref around the request
construction.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191125105858.1718307-1-chris@chris-wilson.co.uk
Replace the struct_mutex requirement for pinning the i915_vma with the
local vm->mutex instead. Note that the vm->mutex is tainted by the
shrinker (we require unbinding from inside fs-reclaim) and so we cannot
allocate while holding that mutex. Instead we have to preallocate
workers to do allocate and apply the PTE updates after we have we
reserved their slot in the drm_mm (using fences to order the PTE writes
with the GPU work and with later unbind).
In adding the asynchronous vma binding, one subtle requirement is to
avoid coupling the binding fence into the backing object->resv. That is
the asynchronous binding only applies to the vma timeline itself and not
to the pages as that is a more global timeline (the binding of one vma
does not need to be ordered with another vma, nor does the implicit GEM
fencing depend on a vma, only on writes to the backing store). Keeping
the vma binding distinct from the backing store timelines is verified by
a number of async gem_exec_fence and gem_exec_schedule tests. The way we
do this is quite simple, we keep the fence for the vma binding separate
and only wait on it as required, and never add it to the obj->resv
itself.
Another consequence in reducing the locking around the vma is the
destruction of the vma is no longer globally serialised by struct_mutex.
A natural solution would be to add a kref to i915_vma, but that requires
decoupling the reference cycles, possibly by introducing a new
i915_mm_pages object that is own by both obj->mm and vma->pages.
However, we have not taken that route due to the overshadowing lmem/ttm
discussions, and instead play a series of complicated games with
trylocks to (hopefully) ensure that only one destruction path is called!
v2: Add some commentary, and some helpers to reduce patch churn.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191004134015.13204-4-chris@chris-wilson.co.uk
Currently, if there is time remaining before the start of the loop, we
do one full iteration over many possible different chunks within the
object. A full loop may take 50+s (depending on speed of indirect GTT
mmapings) and we try separately with LINEAR, X and Y -- at which point
igt times out. If we check more frequently, we will interrupt the loop
upon our timeout -- it is hard to argue for as this significantly reduces
the test coverage as we dramatically reduce the runtime. In practical
terms, the coverage we should prioritise is in using different fence
setups, forcing verification of the tile row computations over the
current preference of checking extracting chunks. Though the exhaustive
search is great given an infinite timeout, to improve our current
coverage, we also add a randomised smoketest of partial mmaps. So let's
do both, add a randomised smoketest of partial tiling chunks and the
exhaustive (though time limited) search for failures.
Even in adding another subtest, we should shave 100s off BAT! (With,
hopefully, no loss in coverage, at least over multiple runs.)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190910121009.13431-1-chris@chris-wilson.co.uk
Remove the accumulated optimisations that we have for i915_vma_retire
and reduce it to the bare essential of tracking the active object
reference. This allows us to only use atomic operations, and so will be
able to avoid the struct_mutex requirement.
The principal loss here is the shrinker MRU bumping, so now if we have
to shrink, we will do so in much more random order and more likely to
try and shrink recently used objects. That is a nuisance, but shrinking
active objects is a second step we try to avoid and will always be a
system-wide performance issue.
The other loss is here is in the automatic pruning of the
reservation_object when idling. This is not as large an issue as upon
reservation_object introduction as now adding new fences into the object
replaces already signaled fences, keeping the array compact. But we do
lose the auto-expiration of stale fences and unused arrays. That may be
a noticeable problem for which we need to re-implement autopruning.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621183801.23252-3-chris@chris-wilson.co.uk
Having introduced struct intel_gt (named the anonymous structure in i915)
we can start using it to compartmentalize our code better. It makes more
sense logically to have the code internally like this and it will also
help with future split between gt and display in i915.
v2:
* Keep ggtt flush before fb obj flush. (Chris)
v3:
* Fix refactoring fail.
* Always flush ggtt writes. (Chris)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-23-tvrtko.ursulin@linux.intel.com
