With full-ppgtt one of the main bottlenecks is the lookup of the VMA
underneath the object. For execbuf there is merit in having a very fast
direct lookup of ctx:handle to the vma using a hashtree, but that still
leaves a large number of other lookups. One way to speed up the lookup
would be to use a rhashtable, but that requires extra allocations and
may exhibit poor worse case behaviour. An alternative is to use an
embedded rbtree, i.e. no extra allocations and deterministic behaviour,
but at the slight cost of O(lgN) lookups (instead of O(1) for
rhashtable). The major of such tree will be very shallow and so not much
slower, and still scales much, much better than the current unsorted
list.
v2: Bump vma_compare() to return a long, as we return the result of
comparing two pointers.
References: https://bugs.freedesktop.org/show_bug.cgi?id=87726
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161101115400.15647-1-chris@chris-wilson.co.uk
In preparation to support many distinct timelines, we need to expand the
activity tracking on the GEM object to handle more than just a request
per engine. We already use the struct reservation_object on the dma-buf
to handle many fence contexts, so integrating that into the GEM object
itself is the preferred solution. (For example, we can now share the same
reservation_object between every consumer/producer using this buffer and
skip the manual import/export via dma-buf.)
v2: Reimplement busy-ioctl (by walking the reservation object), postpone
the ABI change for another day. Similarly use the reservation object to
find the last_write request (if active and from i915) for choosing
display CS flips.
Caveats:
* busy-ioctl: busy-ioctl only reports on the native fences, it will not
warn of stalls (in set-domain-ioctl, pread/pwrite etc) if the object is
being rendered to by external fences. It also will not report the same
busy state as wait-ioctl (or polling on the dma-buf) in the same
circumstances. On the plus side, it does retain reporting of which
*i915* engines are engaged with this object.
* non-blocking atomic modesets take a step backwards as the wait for
render completion blocks the ioctl. This is fixed in a subsequent
patch to use a fence instead for awaiting on the rendering, see
"drm/i915: Restore nonblocking awaits for modesetting"
* dynamic array manipulation for shared-fences in reservation is slower
than the previous lockless static assignment (e.g. gem_exec_lut_handle
runtime on ivb goes from 42s to 66s), mainly due to atomic operations
(maintaining the fence refcounts).
* loss of object-level retirement callbacks, emulated by VMA retirement
tracking.
* minor loss of object-level last activity information from debugfs,
could be replaced with per-vma information if desired
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-21-chris@chris-wilson.co.uk
We can remove the false coupling between RPM and struct mutex by the
observation that we can use the RPM wakeref as the barrier around user
mmap access. That is as we tear down the user's PTE atomically from
within rpm suspend and then to fault in new PTE requires the rpm
wakeref, means that no user access is possible through those PTE without
RPM being awake. Having made that observation, we can then remove the
presumption of having to take rpm outside of struct_mutex and so allow
fine grained acquisition of a wakeref around hw access rather than
having to remember to acquire the wakeref early on.
v2: Rejig placement of the new intel_runtime_pm_get() to be as tight
as possible around the GTT pread/pwrite.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161024124218.18252-2-chris@chris-wilson.co.uk
Since "Dynamic page table allocations" were introduced, our page tables
can grow (being dynamically allocated) with address space range usage.
Unfortunately, their lifetime is bound to vm. This is not a huge problem
when we're not using softpin - drm_mm is creating an upper bound on used
range by causing addresses for our VMAs to eventually be reused.
With softpin, long lived contexts can drain the system out of memory
even with a single "small" object. For example:
bo = bo_alloc(size);
while(true)
offset += size;
exec(bo, offset);
Will cause us to create new allocations until all memory in the system
is used for tracking GPU pages (even though almost all PTEs in this vm
are pointing to scratch).
Let's free unused page tables in clear_range to prevent this - if no
entries are used, we can safely free it and return this information to
the caller (so that higher-level entry is pointing to scratch).
v2: Document return value and free semantics (Joonas)
v3: No newlines in vars block (Joonas)
v4: Drop redundant local 'reduce' variable
v5: Handle CI fail with enable_ppgtt=2
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1476360162-24062-3-git-send-email-michal.winiarski@intel.com
With the possibility of addition of many more number of rings in future,
the drm_i915_private structure could bloat as an array, of type
intel_engine_cs, is embedded inside it.
struct intel_engine_cs engine[I915_NUM_ENGINES];
Though this is still fine as generally there is only a single instance of
drm_i915_private structure used, but not all of the possible rings would be
enabled or active on most of the platforms. Some memory can be saved by
allocating intel_engine_cs structure only for the enabled/active engines.
Currently the engine/ring ID is kept static and dev_priv->engine[] is simply
indexed using the enums defined in intel_engine_id.
To save memory and continue using the static engine/ring IDs, 'engine' is
defined as an array of pointers.
struct intel_engine_cs *engine[I915_NUM_ENGINES];
dev_priv->engine[engine_ID] will be NULL for disabled engine instances.
There is a text size reduction of 928 bytes, from 1028200 to 1027272, for
i915.o file (but for i915.ko file text size remain same as 1193131 bytes).
v2:
- Remove the engine iterator field added in drm_i915_private structure,
instead pass a local iterator variable to the for_each_engine**
macros. (Chris)
- Do away with intel_engine_initialized() and instead directly use the
NULL pointer check on engine pointer. (Chris)
v3:
- Remove for_each_engine_id() macro, as the updated macro for_each_engine()
can be used in place of it. (Chris)
- Protect the access to Render engine Fault register with a NULL check, as
engine specific init is done later in Driver load sequence.
v4:
- Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris)
- Kill the superfluous init_engine_lists().
v5:
- Cleanup the intel_engines_init() & intel_engines_setup(), with respect to
allocation of intel_engine_cs structure. (Chris)
v6:
- Rebase.
v7:
- Optimize the for_each_engine_masked() macro. (Chris)
- Change the type of 'iter' local variable to enum intel_engine_id. (Chris)
- Rebase.
v8: Rebase.
v9: Rebase.
v10:
- For index calculation use engine ID instead of pointer based arithmetic in
intel_engine_sync_index() as engine pointers are not contiguous now (Chris)
- For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas)
- Use for_each_engine macro for cleanup in intel_engines_init() and remove
check for NULL engine pointer in cleanup() routines. (Joonas)
v11: Rebase.
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com
Recently I have been applying an optimisation to avoid stalling and
clflushing GGTT objects based on their current binding. That is we only
set-to-gtt-domain upon first bind. However, on hibernation the objects
remain bound, but they are in the CPU domain. Currently (since commit
975f7ff42e ("drm/i915: Lazily migrate the objects after hibernation"))
we only flush scanout objects as all other objects are expected to be
flushed prior to use. That breaks down in the face of the runtime
optimisation above - and we need to flush all GGTT pinned objects
(essentially ringbuffers).
To reduce the burden of extra clflushes, we only flush those objects we
cannot discard from the GGTT. Everything pinned to the scanout, or
current contexts or ringbuffers will be flushed and rebound. Other
objects, such as inactive contexts, will be left unbound and in the CPU
domain until first use after resuming.
Fixes: 7abc98fadf ("drm/i915: Only change the context object's domain...")
Fixes: 57e8853181 ("drm/i915: Use VMA for ringbuffer tracking")
References: https://bugs.freedesktop.org/show_bug.cgi?id=94722
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: David Weinehall <david.weinehall@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909201957.2499-1-chris@chris-wilson.co.uk
In the next patch we want to handle reset directly by a locked waiter in
order to avoid issues with returning before the reset is handled. To
handle the reset, we must first know whether we hold the struct_mutex.
If we do not hold the struct_mtuex we can not perform the reset, but we do
not block the reset worker either (and so we can just continue to wait for
request completion) - otherwise we must relinquish the mutex.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-10-chris@chris-wilson.co.uk
As we never need to directly access the pages we allocate for scratch and
the pagetables, and always remap them into the GTT through the dma
remapper, we do not need to limit the allocations to lowmem i.e. we can
pass in the __GFP_HIGHMEM flag to the page allocation.
For backwards compatibility, e.g. certain old GPUs not liking highmem
for certain functions that may be accidentally mapped to the scratch
page by userspace, keep the GMCH probe as only allocating from DMA32.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20160822074431.26872-3-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Redo the fb rotation handling in order to:
- eliminate the NV12 special casing
- handle fb->offsets[] properly
- make the rotation handling easier for the plane code
To achieve these goals we reduce intel_rotation_info to only contain
(for each plane) the rotated view width,height,stride in tile units,
and the page offset into the object where the plane starts. Each plane
is handled exactly the same way, no special casing for NV12 or other
formats. We then store the computed rotation_info under
intel_framebuffer so that we don't have to recompute it again.
To handle fb->offsets[] we treat them as a linear offsets and convert
them to x/y offsets from the start of the relevant GTT mapping (either
normal or rotated). We store the x/y offsets under intel_framebuffer,
and for some extra convenience we also store the rotated pitch (ie.
tile aligned plane height). So for each plane we have the normal
x/y offsets, rotated x/y offsets, and the rotated pitch. The normal
pitch is available already in fb->pitches[].
While we're gathering up all that extra information, we can also easily
compute the storage requirements for the framebuffer, so that we can
check that the object is big enough to hold it.
When it comes time to deal with the plane source coordinates, we first
rotate the clipped src coordinates to match the relevant GTT view
orientation, then add to them the fb x/y offsets. Next we compute
the aligned surface page offset, and as a result we're left with some
residual x/y offsets. Finally, if required by the hardware, we convert
the remaining x/y offsets into a linear offset.
For gen2/3 we simply skip computing the final page offset, and just
convert the src+fb x/y offsets directly into a linear offset since
that's what the hardware wants.
After this all platforms, incluing SKL+, compute these things in exactly
the same way (excluding alignemnt differences).
v2: Use BIT(DRM_ROTATE_270) instead of ROTATE_270 when rotating
plane src coordinates
Drop some spurious changes that got left behind during
development
v3: Split out more changes to prep patches (Daniel)
s/intel_fb->plane[].foo.bar/intel_fb->foo[].bar/ for brevity
Rename intel_surf_gtt_offset to intel_fb_gtt_offset
Kill the pointless 'plane' parameter from intel_fb_gtt_offset()
v4: Fix alignment vs. alignment-1 when calling
_intel_compute_tile_offset() from intel_fill_fb_info()
Pass the pitch in tiles in
stad of pixels to intel_adjust_tile_offset() from intel_fill_fb_info()
Pass the full width/height of the rotated area to
drm_rect_rotate() for clarity
Use u32 for more offsets
v5: Preserve the upper_32_bits()/lower_32_bits() handling for the
fb ggtt offset (Sivakumar)
v6: Rebase due to drm_plane_state src/dst rects
Cc: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1470821001-25272-2-git-send-email-ville.syrjala@linux.intel.com
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The principal motivation for this was to try and eliminate the
struct_mutex from i915_gem_suspend - but we still need to hold the mutex
current for the i915_gem_context_lost(). (The issue there is that there
may be an indirect lockdep cycle between cpu_hotplug (i.e. suspend) and
struct_mutex via the stop_machine().) For the moment, enabling last
request tracking for the engine, allows us to do busyness checking and
waiting without requiring the struct_mutex - which is useful in its own
right.
As a side-effect of having a robust means for tracking engine busyness,
we can replace our other busyness heuristic, that of comparing against
the last submitted seqno. For paranoid reasons, we have a semi-ordered
check of that seqno inside the hangchecker, which we can now improve to
an ordered check of the engine's busyness (removing a locked xchg in the
process).
v2: Pass along "bool interruptible" as being unlocked we cannot rely on
i915->mm.interruptible being stable or even under our control.
v3: Replace check Ironlake i915_gpu_busy() with the common precalculated value
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1470388464-28458-6-git-send-email-chris@chris-wilson.co.uk
