There is an interesting refcounting loop:
struct intel_memory_region has a struct ttm_resource_manager,
ttm_resource_manager->move may hold a reference to i915_request,
i915_request may hold a reference to intel_context,
intel_context may hold a reference to drm_i915_gem_object,
drm_i915_gem_object may hold a reference to intel_memory_region.
Break this loop by dropping region reference counting.
In addition, Have regions with a manager moving fence make sure
that all region objects are released before freeing the region.
v6:
- Fix a code comment.
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211122214554.371864-4-thomas.hellstrom@linux.intel.com
We really only need memcpy restore for objects that affect the
operability of the migrate context. That is, primarily the page-table
objects of the migrate VM.
Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early
restores using memcpy and a way to assign LMEM page-table object flags
to be used by the vms.
Restore objects without this flag with the gpu blitter and only objects
carrying the flag using TTM memcpy.
Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and
defer for a later audit which vms actually need it. Most importantly, user-
allocated vms with pinned page-table objects can be restored using the
blitter.
Performance-wise memcpy restore is probably as fast as gpu restore if not
faster, but using gpu restore will help tackling future restrictions in
mappable LMEM size.
v4:
- Don't mark the aliasing ppgtt page table flags for early resume, but
rather the ggtt page table flags as intended. (Matthew Auld)
- The check for user buffer objects during early resume is pointless, since
they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld)
v5:
- Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored
before we fire up the migrate context.
Cc: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
It's been invariant since
commit ccbc1b9794
Author: Jason Ekstrand <jason@jlekstrand.net>
Date: Thu Jul 8 10:48:30 2021 -0500
drm/i915/gem: Don't allow changing the VM on running contexts (v4)
this just completes the deed. I've tried to split out prep work for
more careful review as much as possible, this is what's left:
- get_ppgtt gets simplified since we don't need to grab a temporary
reference - we can rely on the temporary reference for the gem_ctx
while we inspect the vm. The new vm_id still needs a full
i915_vm_open ofc. This also removes the final caller of context_get_vm_rcu
- A pile of selftests can now just look at ctx->vm instead of
rcu_dereference_protected( , true) or similar things.
- All callers of i915_gem_context_vm also disappear.
- I've changed the hugepage selftest to set scrub_64K without any
locking, because when we inspect that setting we're also not taking
any locks either. It works because it's a selftests that's careful
(single threaded gives you nice ordering) and not a live driver
where races can happen from anywhere.
These can only be split up further if we have some intermediate state
with a bunch more rcu_dereference_protected(ctx->vm, true), just to
shut up lockdep and sparse.
The conversion to __rcu happened in
commit a4e7ccdac3
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Fri Oct 4 14:40:09 2019 +0100
drm/i915: Move context management under GEM
Note that we're not breaking the actual bugfix in there: The real
bugfix is pushing the i915_vm_relase onto a separate worker, to avoid
locking inversion issues. The rcu conversion was just thrown in for
entertainment value on top (no vm lookup isn't even close to anything
that's a hotpath where removing the single spinlock can be measured).
v2: Rebase over the change to move the i915_vm_put() into
i915_gem_context_release().
v3: Trivial conflict against repainted shed.
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: "Thomas Hellström" <thomas.hellstrom@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Jason Ekstrand <jason@jlekstrand.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20210902142057.929669-9-daniel.vetter@ffwll.ch
For some specialised objects we might need something larger than the
regions min_page_size due to some hw restriction, and slightly more
hairy is needing something smaller with the guarantee that such objects
will never be inserted into any GTT, which is the case for the paging
structures.
This also fixes how we setup the BO page_alignment, if we later migrate
the object somewhere else. For example if the placements are {SMEM,
LMEM}, then we might get this wrong. Pushing the min_page_size behaviour
into the manager should fix this.
v2(Thomas): push the default page size behaviour into buddy_man, and let
the user override it with the page-alignment, which looks cleaner
v3: rebase on ttm sys changes
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210625103824.558481-1-matthew.auld@intel.com
The GEM object is grossly overweight for the practicality of tracking
large numbers of individual pages, yet it is currently our only
abstraction for tracking DMA allocations. Since those allocations need
to be reserved upfront before an operation, and that we need to break
away from simple system memory, we need to ditch using plain struct page
wrappers.
In the process, we drop the WC mapping as we ended up clflushing
everything anyway due to various issues across a wider range of
platforms. Though in a future step, we need to drop the kmap_atomic
approach which suggests we need to pre-map all the pages and keep them
mapped.
v2: Verify our large scratch page is suitably DMA aligned; and manually
clear the scratch since we are allocating plain struct pages full of
prior content.
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/20200729164219.5737-2-chris@chris-wilson.co.uk
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The history of i915_vma_close() is confusing, as is its use. As the
lifetime of the i915_vma is currently bounded by the object it is
attached to, we needed a means of identify when a vma was no longer in
use by userspace (via the user's fd). This is further complicated by
that only ppgtt vma should be closed at the user's behest, as the ggtt
were always shared.
Now that we attach the vma to a lut on the user's context, the open
count does indicate how many unique and open context/vm are referencing
this vma from the user. As such, we can and should just use the
open_count to track when the vma is still in use by userspace.
It's a poor man's replacement for reference counting.
Closes: https://gitlab.freedesktop.org/drm/intel/issues/1193
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Andi Shyti <andi.shyti@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200422190558.30509-1-chris@chris-wilson.co.uk
igt_ppgtt_pin_update() invokes i915_gem_context_get_vm_rcu(), which
returns a reference of the i915_address_space object to "vm" with
increased refcount.
When igt_ppgtt_pin_update() returns, "vm" becomes invalid, so the
refcount should be decreased to keep refcount balanced.
The reference counting issue happens in two exception handling paths of
igt_ppgtt_pin_update(). When i915_gem_object_create_internal() returns
IS_ERR, the refcnt increased by i915_gem_context_get_vm_rcu() is not
decreased, causing a refcnt leak.
Fix this issue by jumping to "out_vm" label when
i915_gem_object_create_internal() returns IS_ERR.
Fixes: a4e7ccdac3 ("drm/i915: Move context management under GEM")
Signed-off-by: Xiyu Yang <xiyuyang19@fudan.edu.cn>
Signed-off-by: Xin Tan <tanxin.ctf@gmail.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/1587361342-83494-1-git-send-email-xiyuyang19@fudan.edu.cn
When creating a handle, it is just that, an abstract handle. The fact
that we cannot currently support a handle larger than the size of the
backing storage is an artifact of our whole-object-at-a-time handling in
get_pages() and being an implementation limitation is best handled at
that point -- similar to shmem, where we only barf when asked to
populate the whole object if larger than RAM. (Pinning the whole object
at a time is major hindrance that we are likely to have to overcome in
the near future.) In the case of the buddy allocator, the late check is
preferable as the request size may often be smaller than the required
size.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191216122603.2598155-1-chris@chris-wilson.co.uk
As drm now exports a method to create an anonymous struct file around a
drm_device for internal use, make use of it to avoid our horrible hacks.
Danial suggested that the mock_file_put() wrapper was suitable for
drm-core, along with the mock_drm_getfile() [and that the vestigal
mock_drm_file() in this patch should perhaps be the drm interface
itself]. However, the eventual goal is to remove the mock_drm_file() and
use the struct file and fput() directly, in this patch we take a simple
transition in that direction.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20191107180601.30815-3-chris@chris-wilson.co.uk
The trouble with having a plain nesting flag for locks which do not
naturally nest (unlike block devices and their partitions, which is
the original motivation for nesting levels) is that lockdep will
never spot a true deadlock if you screw up.
This patch is an attempt at trying better, by highlighting a bit more
of the actual nature of the nesting that's going on. Essentially we
have two kinds of objects:
- objects without pages allocated, which cannot be on any lru and are
hence inaccessible to the shrinker.
- objects which have pages allocated, which are on an lru, and which
the shrinker can decide to throw out.
For the former type of object, memory allocations while holding
obj->mm.lock are permissible. For the latter they are not. And
get/put_pages transitions between the two types of objects.
This is still not entirely fool-proof since the rules might change.
But as long as we run such a code ever at runtime lockdep should be
able to observe the inconsistency and complain (like with any other
lockdep class that we've split up in multiple classes). But there are
a few clear benefits:
- We can drop the nesting flag parameter from
__i915_gem_object_put_pages, because that function by definition is
never going allocate memory, and calling it on an object which
doesn't have its pages allocated would be a bug.
- We strictly catch more bugs, since there's not only one place in the
entire tree which is annotated with the special class. All the
other places that had explicit lockdep nesting annotations we're now
going to leave up to lockdep again.
- Specifically this catches stuff like calling get_pages from
put_pages (which isn't really a good idea, if we can call get_pages
so could the shrinker). I've seen patches do exactly that.
Of course I fully expect CI will show me for the fool I am with this
one here :-)
v2: There can only be one (lockdep only has a cache for the first
subclass, not for deeper ones, and we don't want to make these locks
even slower). Still separate enums for better documentation.
Real fix: don't forget about phys objs and pin_map(), and fix the
shrinker to have the right annotations ... silly me.
v3: Forgot usertptr too ...
v4: Improve comment for pages_pin_count, drop the IMPORTANT comment
and instead prime lockdep (Chris).
v5: Appease checkpatch, no double empty lines (Chris)
v6: More rebasing over selftest changes. Also somehow I forgot to
push this patch :-/
Also format comments consistently while at it.
v7: Fix typo in commit message (Joonas)
Also drop the priming, with the lmem merge we now have allocations
while holding the lmem lock, which wreaks the generic priming I've
done in earlier patches. Should probably be resurrected when lmem is
fixed. See
commit 232a6ebae4
Author: Matthew Auld <matthew.auld@intel.com>
Date: Tue Oct 8 17:01:14 2019 +0100
drm/i915: introduce intel_memory_region
I'm keeping the priming patch locally so it wont get lost.
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: "Tang, CQ" <cq.tang@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v5)
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> (v6)
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191105090148.30269-1-daniel.vetter@ffwll.ch
[mlankhorst: Fix commit typos pointed out by Michael Ruhl]
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
