If a fence allocation fails in a blocking context, we will sleep on the
fence as a last resort. We can therefore allow ourselves to fail and
sleep on the fence instead of triggering a system-wide oom. This allows
us to throttle malicious clients that are consuming lots of system
resources by capping the amount of memory used by fences.
Testcase: igt/gem_shrink/execbufX
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171212180652.22061-2-chris@chris-wilson.co.uk
Use a priority stored in the context as the initial value when
submitting a request. This allows us to change the default priority on a
per-context basis, allowing different contexts to be favoured with GPU
time at the expense of lower importance work. The user can adjust the
context's priority via I915_CONTEXT_PARAM_PRIORITY, with more positive
values being higher priority (they will be serviced earlier, after their
dependencies have been resolved). Any prerequisite work for an execbuf
will have its priority raised to match the new request as required.
Normal users can specify any value in the range of -1023 to 0 [default],
i.e. they can reduce the priority of their workloads (and temporarily
boost it back to normal if so desired).
Privileged users can specify any value in the range of -1023 to 1023,
[default is 0], i.e. they can raise their priority above all overs and
so potentially starve the system.
Note that the existing schedulers are not fair, nor load balancing, the
execution is strictly by priority on a first-come, first-served basis,
and the driver may choose to boost some requests above the range
available to users.
This priority was originally based around nice(2), but evolved to allow
clients to adjust their priority within a small range, and allow for a
privileged high priority range.
For example, this can be used to implement EGL_IMG_context_priority
https://www.khronos.org/registry/egl/extensions/IMG/EGL_IMG_context_priority.txt
EGL_CONTEXT_PRIORITY_LEVEL_IMG determines the priority level of
the context to be created. This attribute is a hint, as an
implementation may not support multiple contexts at some
priority levels and system policy may limit access to high
priority contexts to appropriate system privilege level. The
default value for EGL_CONTEXT_PRIORITY_LEVEL_IMG is
EGL_CONTEXT_PRIORITY_MEDIUM_IMG."
so we can map
PRIORITY_HIGH -> 1023 [privileged, will failback to 0]
PRIORITY_MED -> 0 [default]
PRIORITY_LOW -> -1023
They also map onto the priorities used by VkQueue (and a VkQueue is
essentially a timeline, our i915_gem_context under full-ppgtt).
v2: s/CAP_SYS_ADMIN/CAP_SYS_NICE/
v3: Report min/max user priorities as defines in the uapi, and rebase
internal priorities on the exposed values.
Testcase: igt/gem_exec_schedule
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171003203453.15692-9-chris@chris-wilson.co.uk
We use INT_MIN to denote the priority of a request that has not been
submitted to the scheduler; we treat INT_MIN as an invalid priority and
initialise the request to it. Give the value a name so it stands out.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170928193910.17988-3-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com
No users now outside of i915_wait_request(), so we can make it private to
i915_gem_request.c, and assume the caller knows the seqno. In the
process, also remove i915_gem_request_started() as that was only ever
used by i915_spin_request().
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michal Winiarski <michal.winiarski@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170922120333.25535-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
2nd round of 4.14 features:
- prep for deferred fbdev setup
- refactor fixed 16.16 computations and skl+ wm code (Mahesh Kumar)
- more cnl paches (Rodrigo, Imre et al)
- tighten context cleanup and handling (Chris Wilson)
- fix interlaced handling on skl+ (Mahesh Kumar)
- small bits as usual
* tag 'drm-intel-next-2017-07-17' of git://anongit.freedesktop.org/git/drm-intel: (84 commits)
drm/i915: Update DRIVER_DATE to 20170717
drm/i915: Protect against deferred fbdev setup
drm/i915/fbdev: Always forward hotplug events
drm/i915/skl+: unify cpp value in WM calculation
drm/i915/skl+: WM calculation don't require height
drm/i915: Addition wrapper for fixed16.16 operation
drm/i915: cleanup fixed-point wrappers naming
drm/i915: Always perform internal fixed16 division in 64 bits
drm/i915: take-out common clamping code of fixed16 wrappers
drm/i915/cnl: Add missing type case.
drm/i915/cnl: Add max allowed Cannonlake DC.
drm/i915: Make DP-MST connector info work
drm/i915/cnl: Get DDI clock based on PLLs.
drm/i915/cnl: Inherit RPS stuff from previous platforms.
drm/i915/cnl: Gen10 render context size.
drm/i915/cnl: Don't trust VBT's alternate pin for port D for now.
drm/i915: Fix the kernel panic when using aliasing ppgtt
drm/i915/cnl: Cannonlake color init.
drm/i915/cnl: Add force wake for gen10+.
x86/gpu: CNL uses the same GMS values as SKL
...
-----BEGIN PGP SIGNATURE-----
iQIcBAABAgAGBQJZYseIAAoJEAx081l5xIa+85kP/0zKzKKVzZXSXG2TAGb5jNfk
Ex+TELG8tWk9KBxA7lEE5c0WEsnP79cNoXZLQu8wlUzO8+kwQK5Bz0zgNUkpSuo1
RthwdsxBQX1++UxB+HoSG+dOa7hkKVqlgQR3z9qyhsBXzetkJV0DoYcpMV0A1EWd
6Jzt+AvCShVkcW+21LqHPlc5EIVewrDMoA3oU6aYCLhyAOUTVvvQB2ML8YApH7TM
JrSrzCFHTrQEBbGUrZQhzR0sZzZzk9byntb/I/mdVbHeCyIHiL8sC4PfWSOyyazm
GkPnA8G3aFAY9haBRz9jG/VBr1yVb0mCBjkWQ1lGfIAOCDDSc+d7PDXdG+i4AewK
jZheXlrDIdGgmJLy4W3rdEqJvdf7UQHZOs8594OL19l4+FxCTrol1JSHSMeavCvr
8bUNil9Jb/ONU/wmp+q55U0k4TCTyerUA7gKnuaJAwBvd4n78/PKmQnbrWinDyJc
GQXp6zESk9bKt5DXSnVZuVf4POTzpuAsQkkfX1V2y145EHTQYfS3jLENWqEjyZUy
QtKCHZvRkJfGaFU4Pr+vBo9Iu1GlA5OiOv08QadldTT4OxUI0T6yaLDobHCQfKPE
sc3wCuCM+/dAnqoKDcGC4hAmF8zDdO0kw65P2m7uC6T9Jm1G35CioKbzo+fzUhuL
fg5TBpbp2Wwe2oPA5iBm
=2S5N
-----END PGP SIGNATURE-----
Merge tag 'drm-for-v4.13' of git://people.freedesktop.org/~airlied/linux
Pull drm updates from Dave Airlie:
"This is the main pull request for the drm, I think I've got one later
driver pull for mediatek SoC driver, I'm undecided on if it needs to
go to you yet.
Otherwise summary below:
Core drm:
- Atomic add driver private objects
- Deprecate preclose hook in modern drivers
- MST bandwidth tracking
- Use kvmalloc in more places
- Add mode_valid hook for crtc/encoder/bridge
- Reduce sync_file construction time
- Documentation updates
- New DRM synchronisation object support
New drivers:
- pl111 - pl111 CLCD display controller
Panel:
- Innolux P079ZCA panel driver
- Add NL12880B20-05, NL192108AC18-02D, P320HVN03 panels
- panel-samsung-s6e3ha2: Add s6e3hf2 panel support
i915:
- SKL+ watermark fixes
- G4x/G33 reset improvements
- DP AUX backlight improvements
- Buffer based GuC/host communication
- New getparam for (sub)slice infomation
- Cannonlake and Coffeelake initial patches
- Execbuf optimisations
radeon/amdgpu:
- Lots of Vega10 bug fixes
- Preliminary raven support
- KIQ support for compute rings
- MEC queue management rework
- DCE6 Audio support
- SR-IOV improvements
- Better radeon/amdgpu selection support
nouveau:
- HDMI stereoscopic support
- Display code rework for >= GM20x GPUs
msm:
- GEM rework for fine-grained locking
- Per-process pagetable work
- HDMI fixes for Snapdragon 820.
vc4:
- Remove 256MB CMA limit from vc4
- Add out-fence support
- Add support for cygnus
- Get/set tiling ioctls support
- Add T-format tiling support for scanout
zte:
- add VGA support.
etnaviv:
- Thermal throttle support for newer GPUs
- Restore userspace buffer cache performance
- dma-buf sync fix
stm:
- add stm32f429 display support
exynos:
- Rework vblank handling
- Fixup sw-trigger code
sun4i:
- V3s display engine support
- HDMI support for older SoCs
- Preliminary work on dual-pipeline SoCs.
rcar-du:
- VSP work
imx-drm:
- Remove counter load enable from PRE
- Double read/write reduction flag support
tegra:
- Documentation for the host1x and drm driver.
- Lots of staging ioctl fixes due to grate project work.
omapdrm:
- dma-buf fence support
- TILER rotation fixes"
* tag 'drm-for-v4.13' of git://people.freedesktop.org/~airlied/linux: (1270 commits)
drm: Remove unused drm_file parameter to drm_syncobj_replace_fence()
drm/amd/powerplay: fix bug fail to remove sysfs when rmmod amdgpu.
amdgpu: Set cik/si_support to 1 by default if radeon isn't built
drm/amdgpu/gfx9: fix driver reload with KIQ
drm/amdgpu/gfx8: fix driver reload with KIQ
drm/amdgpu: Don't call amd_powerplay_destroy() if we don't have powerplay
drm/ttm: Fix use-after-free in ttm_bo_clean_mm
drm/amd/amdgpu: move get memory type function from early init to sw init
drm/amdgpu/cgs: always set reference clock in mode_info
drm/amdgpu: fix vblank_time when displays are off
drm/amd/powerplay: power value format change for Vega10
drm/amdgpu/gfx9: support the amdgpu.disable_cu option
drm/amd/powerplay: change PPSMC_MSG_GetCurrPkgPwr for Vega10
drm/amdgpu: Make amdgpu_cs_parser_init static (v2)
drm/amdgpu/cs: fix a typo in a comment
drm/amdgpu: Fix the exported always on CU bitmap
drm/amdgpu/gfx9: gfx_v9_0_enable_gfx_static_mg_power_gating() can be static
drm/amdgpu/psp: upper_32_bits/lower_32_bits for address setup
drm/amd/powerplay/cz: print message if smc message fails
drm/amdgpu: fix typo in amdgpu_debugfs_test_ib_init
...
Once a client has requested a waitboost, we keep that waitboost active
until all clients are no longer waiting. This is because we don't
distinguish which waiter deserves the boost. However, with the advent of
fence signaling, the signaler threads appear as waiters to the RPS
interrupt handler. So instead of using a single boolean to track when to
keep the waitboost active, use a counter of all outstanding waitboosted
requests.
At this point, I have removed all vestiges of the rate limiting on
clients. Whilst this means that compositors should remain more fluid,
it also means that boosts are more prevalent. See commit b29c19b645
("drm/i915: Boost RPS frequency for CPU stalls") for a longer discussion
on the pros and cons of both approaches.
A drawback of this implementation is that it requires constant request
submission to keep the waitboost trimmed (as it is now cancelled when the
request is completed). This will be fine for a busy system, but near
idle the boosts may be kept for longer than desired (effectively tens of
vblanks worstcase) and there is a reliance on rc6 instead.
v2: Remove defunct rps.client_lock
Reported-by: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170628123548.9236-1-chris@chris-wilson.co.uk
Rename:
wait_queue_t => wait_queue_entry_t
'wait_queue_t' was always a slight misnomer: its name implies that it's a "queue",
but in reality it's a queue *entry*. The 'real' queue is the wait queue head,
which had to carry the name.
Start sorting this out by renaming it to 'wait_queue_entry_t'.
This also allows the real structure name 'struct __wait_queue' to
lose its double underscore and become 'struct wait_queue_entry',
which is the more canonical nomenclature for such data types.
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
More stuff for 4.13:
- skl+ wm fixes from Mahesh Kumar
- some refactor and tests for i915_sw_fence (Chris)
- tune execlist/scheduler code (Chris)
- g4x,g33 gpu reset improvements (Chris, Mika)
- guc code cleanup (Michal Wajdeczko, Michał Winiarski)
- dp aux backlight improvements (Puthikorn Voravootivat)
- buffer based guc/host communication (Michal Wajdeczko)
* tag 'drm-intel-next-2017-05-29' of git://anongit.freedesktop.org/git/drm-intel: (253 commits)
drm/i915: Update DRIVER_DATE to 20170529
drm/i915: Keep the forcewake timer alive for 1ms past the most recent use
drm/i915/guc: capture GuC logs if FW fails to load
drm/i915/guc: Introduce buffer based cmd transport
drm/i915/guc: Disable send function on fini
drm: Add definition for eDP backlight frequency
drm/i915: Drop AUX backlight enable check for backlight control
drm/i915: Consolidate #ifdef CONFIG_INTEL_IOMMU
drm/i915: Only GGTT vma may be pinned and prevent shrinking
drm/i915: Serialize GTT/Aperture accesses on BXT
drm/i915: Convert i915_gem_object_ops->flags values to use BIT()
drm/i915/selftests: Silence compiler warning in igt_ctx_exec
drm/i915/guc: Skip port assign on first iteration of GuC dequeue
drm/i915: Remove misleading comment in request_alloc
drm/i915/g33: Improve reset reliability
Revert "drm/i915: Restore lost "Initialized i915" welcome message"
drm/i915/huc: Update GLK HuC version
drm/i915: Check for allocation failure
drm/i915/guc: Remove action status and statistics from debugfs
drm/i915/g4x: Improve gpu reset reliability
...
All the requests at the same priority are executed in FIFO order. They
do not need to be stored in the rbtree themselves, as they are a simple
list within a level. If we move the requests at one priority into a list,
we can then reduce the rbtree to the set of priorities. This should keep
the height of the rbtree small, as the number of active priorities can not
exceed the number of active requests and should be typically only a few.
Currently, we have ~2k possible different priority levels, that may
increase to allow even more fine grained selection. Allocating those in
advance seems a waste (and may be impossible), so we opt for allocating
upon first use, and freeing after its requests are depleted. To avoid
the possibility of an allocation failure causing us to lose a request,
we preallocate the default priority (0) and bump any request to that
priority if we fail to allocate it the appropriate plist. Having a
request (that is ready to run, so not leading to corruption) execute
out-of-order is better than leaking the request (and its dependency
tree) entirely.
There should be a benefit to reducing execlists_dequeue() to principally
using a simple list (and reducing the frequency of both rbtree iteration
and balancing on erase) but for typical workloads, request coalescing
should be small enough that we don't notice any change. The main gain is
from improving PI calls to schedule, and the explicit list within a
level should make request unwinding simpler (we just need to insert at
the head of the list rather than the tail and not have to make the
rbtree search more complicated).
v2: Avoid use-after-free when deleting a depleted priolist
v3: Michał found the solution to handling the allocation failure
gracefully. If we disable all priority scheduling following the
allocation failure, those requests will be executed in fifo and we will
ensure that this request and its dependencies are in strict fifo (even
when it doesn't realise it is only a single list). Normal scheduling is
restored once we know the device is idle, until the next failure!
Suggested-by: Michał Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170517121007.27224-8-chris@chris-wilson.co.uk
Pull RCU updates from Ingo Molnar:
"The main changes are:
- Debloat RCU headers
- Parallelize SRCU callback handling (plus overlapping patches)
- Improve the performance of Tree SRCU on a CPU-hotplug stress test
- Documentation updates
- Miscellaneous fixes"
* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (74 commits)
rcu: Open-code the rcu_cblist_n_lazy_cbs() function
rcu: Open-code the rcu_cblist_n_cbs() function
rcu: Open-code the rcu_cblist_empty() function
rcu: Separately compile large rcu_segcblist functions
srcu: Debloat the <linux/rcu_segcblist.h> header
srcu: Adjust default auto-expediting holdoff
srcu: Specify auto-expedite holdoff time
srcu: Expedite first synchronize_srcu() when idle
srcu: Expedited grace periods with reduced memory contention
srcu: Make rcutorture writer stalls print SRCU GP state
srcu: Exact tracking of srcu_data structures containing callbacks
srcu: Make SRCU be built by default
srcu: Fix Kconfig botch when SRCU not selected
rcu: Make non-preemptive schedule be Tasks RCU quiescent state
srcu: Expedite srcu_schedule_cbs_snp() callback invocation
srcu: Parallelize callback handling
kvm: Move srcu_struct fields to end of struct kvm
rcu: Fix typo in PER_RCU_NODE_PERIOD header comment
rcu: Use true/false in assignment to bool
rcu: Use bool value directly
...
A group of Linux kernel hackers reported chasing a bug that resulted
from their assumption that SLAB_DESTROY_BY_RCU provided an existence
guarantee, that is, that no block from such a slab would be reallocated
during an RCU read-side critical section. Of course, that is not the
case. Instead, SLAB_DESTROY_BY_RCU only prevents freeing of an entire
slab of blocks.
However, there is a phrase for this, namely "type safety". This commit
therefore renames SLAB_DESTROY_BY_RCU to SLAB_TYPESAFE_BY_RCU in order
to avoid future instances of this sort of confusion.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: <linux-mm@kvack.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
[ paulmck: Add comments mentioning the old name, as requested by Eric
Dumazet, in order to help people familiar with the old name find
the new one. ]
Acked-by: David Rientjes <rientjes@google.com>
Introduce a new execobject.flag (EXEC_OBJECT_CAPTURE) that userspace may
use to indicate that it wants the contents of this buffer preserved in
the error state (/sys/class/drm/cardN/error) following a GPU hang
involving this batch.
Use this at your discretion, the contents of the error state. although
compressed, are allocated with GFP_ATOMIC (i.e. limited) and kept for all
eternity (until the error state is destroyed).
Based on an earlier patch by Ben Widawsky <ben@bwidawsk.net>
Testcase: igt/gem_exec_capture
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Matt Turner <mattst88@gmail.com>
Acked-by: Ben Widawsky <ben@bwidawsk.net>
Acked-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170415093902.22581-1-chris@chris-wilson.co.uk
The only time we need to emit a flush inside request emission is after
an execbuffer, for which we can use the full __i915_add_request(). All
other instances want the simpler i915_add_request() without flushing, so
remove the useless helper.
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/20170317114709.8388-1-chris@chris-wilson.co.uk
__i915_gem_request_started() asserts that the seqno is valid, but
i915_spin_request() was not checking before querying whether the request
had started.
Reported-by: Michał Winiarski <michal.winiarski@intel.com>
Fixes: 754c9fd576 ("drm/i915: Protect the request->global_seqno with the engine->timeline lock")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170308142238.22994-1-chris@chris-wilson.co.uk
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Adding to the tail of the client request list as the only other user is
in the throttle ioctl that iterates forwards over the list. It only
needs protection against deletion of a request as it reads it, it simply
won't see a new request added to the end of the list, or it would be too
early and rejected. We can further reduce the number of spinlocks
required when throttling by removing stale requests from the client_list
as we throttle.
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/20170302122525.19675-1-chris@chris-wilson.co.uk
As execlists and other non-semaphore multi-engine devices coordinate
between engines using interrupts, we can shave off a few 10s of
microsecond of scheduling latency by doing the fence signaling from the
interrupt as opposed to a RT kthread. (Realistically the delay adds
about 1% to an individual cross-engine workload.) We only signal the
first fence in order to limit the amount of work we move into the
interrupt handler. We also have to remember that our breadcrumbs may be
unordered with respect to the interrupt and so we still require the
waiter process to perform some heavyweight coherency fixups, as well as
traversing the tree of waiters.
v2: No need for early exit in irq handler - it breaks the flow between
patches and prevents the tracepoint
v3: Restore rcu hold across irq signaling of request
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170227205850.2828-2-chris@chris-wilson.co.uk
After the request is cancelled, we then need to remove it from the
global execution timeline and return it to the context timeline, the
inverse of submit_request().
v2: Move manipulation of struct intel_wait to helpers
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223074422.4125-12-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
A request is assigned a global seqno only when it is on the hardware
execution queue. The global seqno can be used to maintain a list of
requests on the same engine in retirement order, for example for
constructing a priority queue for waiting. Prior to its execution, or
if it is subsequently removed in the event of preemption, its global
seqno is zero. As both insertion and removal from the execution queue
may operate in IRQ context, it is not guarded by the usual struct_mutex
BKL. Instead those relying on the global seqno must be prepared for its
value to change between reads. Only when the request is complete can
the global seqno be stable (due to the memory barriers on submitting
the commands to the hardware to write the breadcrumb, if the HWS shows
that it has passed the global seqno and the global seqno is unchanged
after the read, it is indeed complete).
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/20170223074422.4125-9-chris@chris-wilson.co.uk
On reflection, we are only using the execute fence as a waitqueue on the
global_seqno and not using it for dependency tracking between fences
(unlike the submit and dma fences). By only treating it as a waitqueue,
we can then treat it similar to the other waitqueues during submit,
making the code simpler.
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/20170223074422.4125-8-chris@chris-wilson.co.uk
Replace the global device seqno with one for each engine, and account
for in-flight seqno on each separately. This is consistent with
dma-fence as each timeline has separate fence-contexts for each engine
and a seqno is only ordered within a fence-context (i.e. seqno do not
need to be ordered wrt to other engines, just ordered within a single
engine). This is required to enable request rewinding for preemption on
individual engines (we have to rewind the global seqno to avoid
overflow, and we do not have to rewind all engines just to preempt one.)
v2: Rename active_seqno to inflight_seqnos to more clearly indicate that
it is a counter and not equivalent to the existing seqno. Update
functions that operated on active_seqno similarly.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223074422.4125-3-chris@chris-wilson.co.uk
The i915_gem_active stuff doesn't like a NULL ->retire hook, but
the overlay code can set it to NULL. That obviously ends up oopsing.
Fix it by introducing a new helper to assign the retirement callback
that will switch out the NULL function pointer with
i915_gem_retire_noop.
Cc: stable@vger.kernel.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Fixes: 0d9bdd886f ("drm/i915: Convert intel_overlay to request tracking")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20161207175647.10018-1-chris@chris-wilson.co.uk
The requests conversion introduced a nasty bug where we could generate a
new request in the middle of constructing a request if we needed to idle
the system in order to evict space for a context. The request to idle
would be executed (and waited upon) before the current one, creating a
minor havoc in the seqno accounting, as we will consider the current
request to already be completed (prior to deferred seqno assignment) but
ring->last_retired_head would have been updated and still could allow
us to overwrite the current request before execution.
We also employed two different mechanisms to track the active context
until it was switched out. The legacy method allowed for waiting upon an
active context (it could forcibly evict any vma, including context's),
but the execlists method took a step backwards by pinning the vma for
the entire active lifespan of the context (the only way to evict was to
idle the entire GPU, not individual contexts). However, to circumvent
the tricky issue of locking (i.e. we cannot take struct_mutex at the
time of i915_gem_request_submit(), where we would want to move the
previous context onto the active tracker and unpin it), we take the
execlists approach and keep the contexts pinned until retirement.
The benefit of the execlists approach, more important for execlists than
legacy, was the reduction in work in pinning the context for each
request - as the context was kept pinned until idle, it could short
circuit the pinning for all active contexts.
We introduce new engine vfuncs to pin and unpin the context
respectively. The context is pinned at the start of the request, and
only unpinned when the following request is retired (this ensures that
the context is idle and coherent in main memory before we unpin it). We
move the engine->last_context tracking into the retirement itself
(rather than during request submission) in order to allow the submission
to be reordered or unwound without undue difficultly.
And finally an ulterior motive for unifying context handling was to
prepare for mock requests.
v2: Rename to last_retired_context, split out legacy_context tracking
for MI_SET_CONTEXT.
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/20161218153724.8439-3-chris@chris-wilson.co.uk
Track the priority of each request and use it to determine the order in
which we submit requests to the hardware via execlists.
The priority of the request is determined by the user (eventually via
the context) but may be overridden at any time by the driver. When we set
the priority of the request, we bump the priority of all of its
dependencies to match - so that a high priority drawing operation is not
stuck behind a background task.
When the request is ready to execute (i.e. we have signaled the submit
fence following completion of all its dependencies, including third
party fences), we put the request into a priority sorted rbtree to be
submitted to the hardware. If the request is higher priority than all
pending requests, it will be submitted on the next context-switch
interrupt as soon as the hardware has completed the current request. We
do not currently preempt any current execution to immediately run a very
high priority request, at least not yet.
One more limitation, is that this is first implementation is for
execlists only so currently limited to gen8/gen9.
v2: Replace recursive priority inheritance bumping with an iterative
depth-first search list.
v3: list_next_entry() for walking lists
v4: Explain how the dfs solves the recursion problem with PI.
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/20161114204105.29171-8-chris@chris-wilson.co.uk
The scheduler needs to know the dependencies of each request for the
lifetime of the request, as it may choose to reschedule the requests at
any time and must ensure the dependency tree is not broken. This is in
additional to using the fence to only allow execution after all
dependencies have been completed.
One option was to extend the fence to support the bidirectional
dependency tracking required by the scheduler. However the mismatch in
lifetimes between the submit fence and the request essentially meant
that we had to build a completely separate struct (and we could not
simply reuse the existing waitqueue in the fence for one half of the
dependency tracking). The extra dependency tracking simply did not mesh
well with the fence, and keeping it separate both keeps the fence
implementation simpler and allows us to extend the dependency tracking
into a priority tree (whilst maintaining support for reordering the
tree).
To avoid the additional allocations and list manipulations, the use of
the priotree is disabled when there are no schedulers to use it.
v2: Create a dedicated slab for i915_dependency.
Rename the lists.
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/20161114204105.29171-7-chris@chris-wilson.co.uk
Defer the transfer from the client's timeline onto the execution
timeline from the point of readiness to the point of actual submission.
For example, in execlists, a request is finally submitted to hardware
when the hardware is ready, and only put onto the hardware queue when
the request is ready. By deferring the transfer, we ensure that the
timeline is maintained in retirement order if we decide to queue the
requests onto the hardware in a different order than fifo.
v2: Rebased onto distinct global/user timeline lock classes.
v3: Play with the position of the spin_lock().
v4: Nesting finally resolved with distinct sw_fence lock classes.
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/20161114204105.29171-4-chris@chris-wilson.co.uk
In order to support deferred scheduling, we need to differentiate
between when the request is ready to run (i.e. the submit fence is
signaled) and when the request is actually run (a new execute fence).
This is typically split between the request itself wanting to wait upon
others (for which we use the submit fence) and the CPU wanting to wait
upon the request, for which we use the execute fence to be sure the
hardware is ready to signal completion.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161114204105.29171-3-chris@chris-wilson.co.uk
As a side product, had to split two other files;
- i915_gem_fence_reg.h
- i915_gem_object.h (only parts that needed immediate untanglement)
I tried to move code in as big chunks as possible, to make review
easier. i915_vma_compare was moved to a header temporarily.
v2:
- Use i915_gem_fence_reg.{c,h}
v3:
- Rebased
v4:
- Fix building when DEBUG_GEM is enabled by reordering a bit.
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1478861034-30643-1-git-send-email-joonas.lahtinen@linux.intel.com
Get rid of sloppy inline functions now that we don't have more users:
i915_gem_request_get_seqno
i915_gem_request_get_engine
v2:
- request->engine is always non-NULL (Chris)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1478589108-3702-1-git-send-email-joonas.lahtinen@linux.intel.com
Though we will have multiple timelines, we still have a single timeline
of execution. This we can use to provide an execution and retirement order
of requests. This keeps tracking execution of requests simple, and vital
for preserving a single waiter (i.e. so that we can order the waiters so
that only the earliest to wakeup need be woken). To accomplish this we
distinguish the seqno used to order requests per-context (external) and
that used internally for execution.
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/20161028125858.23563-26-chris@chris-wilson.co.uk
Our timelines are more than just a seqno. They also provide an ordered
list of requests to be executed. Due to the restriction of handling
individual address spaces, we are limited to a timeline per address
space but we use a fence context per engine within.
Our first step to introducing independent timelines per context (i.e. to
allow each context to have a queue of requests to execute that have a
defined set of dependencies on other requests) is to provide a timeline
abstraction for the global execution queue.
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/20161028125858.23563-23-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
Since we only use the more generic unlocked variant, just rename it as
the normal i915_gem_active_wait(). The temporary cost is that we need to
always acquire the reference in a RCU safe manner, but the benefit is
that we will combine the common paths.
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/20161028125858.23563-5-chris@chris-wilson.co.uk
Our low-level wait routine has evolved from our generic wait interface
that handled unlocked, RPS boosting, waits with time tracking. If we
push our GEM fence tracking to use reservation_objects (required for
handling multiple timelines), we lose the ability to pass the required
information down to i915_wait_request(). However, if we push the extra
functionality from i915_wait_request() to the individual callsites
(i915_gem_object_wait_rendering and i915_gem_wait_ioctl) that make use
of those extras, we can both simplify our low level wait and prepare for
extending the GEM interface for use of reservation_objects.
v2: Rewrite i915_wait_request() kerneldocs
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-4-chris@chris-wilson.co.uk
We will need to wait on DMA completion (as signaled via struct fence)
before executing our i915_gem_request. Therefore we want to expose a
method for adding the await on the fence itself to the request.
v2: Add a comment detailing a failure to handle a signal-on-any
fence-array.
v3: Pretend that magic numbers don't exist.
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/20161028125858.23563-1-chris@chris-wilson.co.uk
Now that we have fences in place to drive request submission, we can
employ those to queue requests after their dependencies as opposed to
stalling in the middle of an execbuf ioctl. (However, we still choose to
spin before enabling the IRQ as that is faster - though contentious.)
v2: Do the fence ordering first, where we can still fail.
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/20160909131201.16673-20-chris@chris-wilson.co.uk
We are about to specialize object synchronisation to enable nonblocking
execbuf submission. First we make a copy of the current object
synchronisation for execbuffer. The general i915_gem_object_sync() will
be removed following the removal of CS flips in the near future.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: John Harrison <john.c.harrison@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-16-chris@chris-wilson.co.uk
Drive final request submission from a callback from the fence. This way
the request is queued until all dependencies are resolved, at which
point it is handed to the backend for queueing to hardware. At this
point, no dependencies are set on the request, so the callback is
immediate.
A side-effect of imposing a heavier-irqsafe spinlock for execlist
submission is that we lose the softirq enabling after scheduling the
execlists tasklet. To compensate, we manually kickstart the softirq by
disabling and enabling the bh around the fence signaling.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: John Harrison <john.c.harrison@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-14-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
Emulate HW to track and manage ELSP queue. A set of SW ports are defined
and requests are assigned to these ports before submitting them to HW. This
helps in cleaning up incomplete requests during reset recovery easier
especially after engine reset by decoupling elsp queue management. This
will become more clear in the next patch.
In the engine reset case we want to resume where we left-off after skipping
the incomplete batch which requires checking the elsp queue, removing
element and fixing elsp_submitted counts in some cases. Instead of directly
manipulating the elsp queue from reset path we can examine these ports, fix
up ringbuffer pointers using the incomplete request and restart submissions
again after reset.
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1470414607-32453-3-git-send-email-arun.siluvery@linux.intel.com
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-6-chris@chris-wilson.co.uk
This issue here is (I think) purely theoretical, since a compiler
would need to be especially foolish to recompute the value of
i915_gem_request_completed right after it was already used. Hence the
additional barrier() is also not really a restriction.
But I believe this to be at least permissible, and since our rcu
trickery is a beast it's worth to annotate all the corner cases.
Chris proposed to instead just wrap a READ_ONCE around
request->fence.seqno in i915_gem_request_completed. But that has a
measurable impact on code size, and everywhere we hold a full
reference to the underlying request it's also not needed. And
personally I'd like to have just enough barriers and locking needed
for correctness, but not more - it makes it much easier in the future
to understand what's going on.
Since the busy ioctl has now fully embraced it's races there's no
point annotating it there too. We really only need it in
active_get_rcu, since that function _must_ deliver a correct snapshot
of the active fences (and not chase something else).
v2: Polish the comment a bit more (Chris).
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1471856122-466-1-git-send-email-daniel.vetter@ffwll.ch
Since contexts are not currently shared between userspace processes, we
have an exact correspondence between context creator and guilty batch
submitter. Therefore we can save some per-batch work by inspecting the
context->pid upon error instead. Note that we take the context's
creator's pid rather than the file's pid in order to better track fd
passed over sockets.
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/1471254551-25805-29-git-send-email-chris@chris-wilson.co.uk
Treat the VMA as the primary struct responsible for tracking bindings
into the GPU's VM. That is we want to treat the VMA returned after we
pin an object into the VM as the cookie we hold and eventually release
when unpinning. Doing so eliminates the ambiguity in pinning the object
and then searching for the relevant pin later.
v2: Joonas' stylistic nitpicks, a fun rebase.
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/1471254551-25805-27-git-send-email-chris@chris-wilson.co.uk
request->batch_obj is only set by execbuffer for the convenience of
debugging hangs. By moving that operation to the callsite, we can
simplify all other callers and future patches. We also move the
complications of reference handling of the request->batch_obj next to
where the active tracking is set up for the request.
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/1470832906-13972-2-git-send-email-chris@chris-wilson.co.uk
When using RCU lookup for the request, commit 0eafec6d32 ("drm/i915:
Enable lockless lookup of request tracking via RCU"), we acknowledge that
we may race with another thread that could have reallocated the request.
In order for the first thread not to blow up, the second thread must not
clear the request completed before overwriting it. In the RCU lookup, we
allow for the engine/seqno to be replaced but we do not allow for it to
be zeroed.
The choice we make is to either add extra checking to the RCU lookup, or
embrace the inherent races (as intended). It is more complicated as we
need to manually clear everything we depend upon being zero initialised,
but we benefit from not emiting the memset() to clear the entire
frequently allocated structure (that memset turns up in throughput
profiles). And at the same time, the lookup remains flexible for future
adjustments.
v2: Old style LRC requires another variable to be initialize. (The
danger inherent in not zeroing everything.)
v3: request->batch also needs to be cleared
v4: signaling.tsk is no long used unset, but pid still exists
Fixes: 0eafec6d32 ("drm/i915: Enable lockless lookup of request...")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: "Goel, Akash" <akash.goel@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-2-git-send-email-chris@chris-wilson.co.uk
In the debate as to whether the second read of active->request is
ordered after the dependent reads of the first read of active->request,
just give in and throw a smp_rmb() in there so that ordering of loads is
assured.
v2: Explain the manual smp_rmb()
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1470731014-6894-1-git-send-email-chris@chris-wilson.co.uk
