We need to move to a more flexible timeline that doesn't assume one
fence context per engine, and so allow for a single timeline to be used
across a combination of engines. This means that preallocating a fence
context per engine is now a hindrance, and so we want to introduce the
singular timeline. From the code perspective, this has the notable
advantage of clearing up a lot of mirky semantics and some clumsy
pointer chasing.
By splitting the timeline up into a single entity rather than an array
of per-engine timelines, we can realise the goal of the previous patch
of tracking the timeline alongside the ring.
v2: Tweak wait_for_idle to stop the compiling thinking that ret may be
uninitialised.
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/20180502163839.3248-2-chris@chris-wilson.co.uk
In the future, we want to move a request between engines. To achieve
this, we first realise that we have two timelines in effect here. The
first runs through the GTT is required for ordering vma access, which is
tracked currently by engine. The second is implied by sequential
execution of commands inside the ringbuffer. This timeline is one that
maps to userspace's expectations when submitting requests (i.e. given the
same context, batch A is executed before batch B). As the rings's
timelines map to userspace and the GTT timeline an implementation
detail, move the timeline from the GTT into the ring itself (per-context
in logical-ring-contexts/execlists, or a global per-engine timeline for
the shared ringbuffers in legacy submission.
The two timelines are still assumed to be equivalent at the moment (no
migrating requests between engines yet) and so we can simply move from
one to the other without adding extra ordering.
v2: Reinforce that one isn't allowed to mix the engine execution
timeline with the client timeline from userspace (on the ring).
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/20180502163839.3248-1-chris@chris-wilson.co.uk
In commit 9b6586ae9f ("drm/i915: Keep a global seqno per-engine"), we
moved from a global inflight counter to per-engine counters in the
hope that will be easy to run concurrently in future. However, with the
advent of the desire to move requests between engines, we do need a
global counter to preserve the semantics that no engine wraps in the
middle of a submit. (Although this semantic is now only required for gen7
semaphore support, which only supports greater-then comparisons!)
v2: Keep a global counter of all requests ever submitted and force the
reset when it wraps.
References: 9b6586ae9f ("drm/i915: Keep a global seqno per-engine")
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/20180430131503.5375-1-chris@chris-wilson.co.uk
We can convert engine stats from a spinlock to seqlock to ensure interrupt
processing is never even a tiny bit delayed by parallel readers.
There is a smidgen bit more cost on the write lock side, and an extremely
unlikely chance that readers will have to retry a few times in face of
heavy interrupt load. But it should be extremely unlikely given how
lightweight read side section is compared to the interrupt processing
side, and also compared to the rest of the code paths which can lead into
it. Furthermore, writer is the ones doing the real, latency sensitive
work, while readers are only informative.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20180426074716.7352-1-tvrtko.ursulin@linux.intel.com
If we have more than a few, possibly several thousand request in the
queue, don't show the central portion, just the first few and the last
being executed and/or queued. The first few should be enough to help
identify a problem in execution, and most often comparing the first/last
in the queue is enough to identify problems in the scheduling.
We may need some fine tuning to set MAX_REQUESTS_TO_SHOW for common
debug scenarios, but for the moment if we can avoiding spending more
than a few seconds dumping the GPU state that will avoid a nasty
livelock (where hangcheck spends so long dumping the state, it fires
again and starts to dump the state again in parallel, ad infinitum).
v2: Remember to print last not the stale rq iter after the loop.
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/20180424081600.27544-1-chris@chris-wilson.co.uk
Today we only want to pass along the priority to engine->schedule(), but
in the future we want to have much more control over the various aspects
of the GPU during a context's execution, for example controlling the
frequency allowed. As we need an ever growing number of parameters for
scheduling, move those into a struct for convenience.
v2: Move the anonymous struct into its own function for legibility and
ye olde gcc.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180418184052.7129-3-chris@chris-wilson.co.uk
This has grown to be a sizable amount of code, so move it to
its own file before we try to refactor anything. For the moment,
we are leaving behind the WA BB code and the WAs that get applied
(incorrectly) in init_clock_gating, but we will deal with it later.
v2: Use intel_ prefix for code that deals with the hardware (Chris)
v3: Rebased
v4:
- Rebased
- New license header
v5:
- Rebased
- Added some organisational notes to the file (Chris)
v6: Include DOC section in the documentation build (Jani)
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[ickle: appease checkpatch, mostly]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1523376767-18480-1-git-send-email-oscar.mateo@intel.com
We can refine our current execlists->queue_priority if we inspect
ELSP[1] rather than the head of the unsubmitted queue. Currently, we use
the unsubmitted queue and say that if a subsequent request is more
important than the current queue, we will rerun the submission tasklet
to evaluate the need for preemption. However, we only want to preempt if
we need to jump ahead of a currently executing request in ELSP. The
second reason for running the submission tasklet is amalgamate requests
into the active context on ELSP[0] to avoid a stall when ELSP[0] drains.
(Though repeatedly amalgamating requests into the active context and
triggering many lite-restore is off question gain, the goal really is to
put a context into ELSP[1] to cover the interrupt.) So if instead of
looking at the head of the queue, we look at the context in ELSP[1] we
can answer both of the questions more accurately -- we don't need to
rerun the submission tasklet unless our new request is important enough
to feed into, at least, ELSP[1].
v2: Add some comments from the discussion with Tvrtko.
v3: More commentary to cross-reference queue_request()
References: f6322eddaf ("drm/i915/preemption: Allow preemption between submission ports")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180411103929.27374-1-chris@chris-wilson.co.uk
The only usage outside the intel_lrc.c file is in the ringbuffer
init, but the irq mask calculated there is then overwritten for
all engines that have a non-zero shift, so we can drop it.
This change is not aimed at code saving but at removing from
intel_engines information that does not apply to all gens that have
the engine. When checking without the temporary WARN_ON, code size
is basically unchanged.
v2: make the irq_shifts array static const
v3: rebase, move irq_shifts array to logical_ring_default_irqs
v4: move array inside the if and use u8 for it (Chris)
Suggested-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
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/20180314182653.26981-4-daniele.ceraolospurio@intel.com
Starting from Gen11 the context descriptor format has been updated in
the HW. The hw_id field has been considerably reduced in size and engine
class and instance fields have been added.
There is a slight name clashing issue because the field that we call
hw_id is actually called SW Context ID in the specs for Gen11+.
With the current size of the hw_id field we can have a maximum of 2k
contexts at any time, but we could use the sw_counter field (which is sw
defined) to increase that because the HW requirement is that
engine_id + sw id + sw_counter is a unique number.
GuC uses a similar method to support more contexts but does its tracking
at lrc level. To avoid doing an implementation that will need to be
reworked once GuC support lands, defer it for now and mark it as TODO.
v2: rebased, add documentation, fix GEN11_ENGINE_INSTANCE_SHIFT
v3: rebased, bring back lost code from i915_gem_context.c
v4: make TODO comment more generic
v5: be consistent with bit ordering, add extra checks (Chris)
Cc: Oscar Mateo <oscar.mateo@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Oscar Mateo <oscar.mateo@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180302161501.28594-3-mika.kuoppala@linux.intel.com
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Although we protect the request itself, we don't lock inside
intel_engine_dump() and so the request maybe retired as we peek into it.
One consequence is that the request->ctx may be freed before we
dereference it, leading to a use-after-free. Replace the hw_id we are
peeking from inside request->ctx with the request->fence.context, with
which we can still track from which context the request originated
(although to tie to HW reports requires a little more legwork, but is
good enough to follow the GEM traces).
[52640.729670] general protection fault: 0000 [#2] SMP
[52640.729694] Dumping ftrace buffer:
[52640.729701] (ftrace buffer empty)
[52640.729705] Modules linked in: vgem snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic x86_pkg_\
temp_thermal intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul snd_hda_intel snd_hda_codec snd_hwdep gha\
sh_clmulni_intel snd_hda_core snd_pcm mei_me mei i915 r8169 mii prime_numbers i2c_hid
[52640.729748] CPU: 2 PID: 4335 Comm: gem_exec_schedu Tainted: G UD W 4.16.0-rc3+ #7
[52640.729759] Hardware name: Acer Aspire E5-575G/Ironman_SK , BIOS V1.12 08/02/2016
[52640.729803] RIP: 0010:print_request+0x2b/0xb0 [i915]
[52640.729811] RSP: 0018:ffffc90001453c18 EFLAGS: 00010206
[52640.729820] RAX: 6b6b6b6b6b6b6b6b RBX: ffff8801e0292d40 RCX: 0000000000000006
[52640.729829] RDX: ffffc90001453c60 RSI: ffff8801e0292d40 RDI: 0000000000000003
[52640.729838] RBP: ffffc90001453d80 R08: 0000000000000000 R09: 0000000000000001
[52640.729847] R10: ffffc90001453bd0 R11: ffffc90001453c73 R12: ffffc90001453c60
[52640.729856] R13: ffffc90001453d80 R14: ffff8801d5a683c8 R15: ffff8801e0292d40
[52640.729866] FS: 00007f1ee50548c0(0000) GS:ffff8801e8200000(0000) knlGS:0000000000000000
[52640.729876] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[52640.729884] CR2: 00007f1ee5077000 CR3: 00000001d9411004 CR4: 00000000003606e0
[52640.729893] Call Trace:
[52640.729922] intel_engine_print_registers+0x623/0x890 [i915]
[52640.729948] intel_engine_dump+0x4a3/0x590 [i915]
[52640.729957] ? seq_printf+0x3a/0x50
[52640.729977] i915_engine_info+0xb8/0xe0 [i915]
[52640.729984] ? drm_mode_gamma_get_ioctl+0xf0/0xf0
[52640.729990] seq_read+0xd5/0x410
[52640.729997] full_proxy_read+0x4b/0x70
[52640.730004] __vfs_read+0x1e/0x120
[52640.730009] ? do_sys_open+0x134/0x220
[52640.730015] ? kmem_cache_free+0x174/0x2b0
[52640.730021] vfs_read+0xa1/0x150
[52640.730026] SyS_read+0x40/0xa0
[52640.730032] do_syscall_64+0x65/0x1a0
[52640.730038] entry_SYSCALL_64_after_hwframe+0x42/0xb7
Reported-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180228094732.28462-1-chris@chris-wilson.co.uk
Sometimes we need to boost the priority of an in-flight request, which
may lead to the situation where the second submission port then contains
a higher priority context than the first and so we need to inject a
preemption event. To do so we must always check inside
execlists_dequeue() whether there is a priority inversion between the
ports themselves as well as the head of the priority sorted queue, and we
cannot just skip dequeuing if the queue is empty.
As Michał noted, this doesn't simply extend to handling more than 2-port
submission, as we may need to reorder within the array of executing
requests which themselves are lower priority than the first. A task for
later!
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180222142229.14517-1-chris@chris-wilson.co.uk
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Gen11 removes the Resource Streamer, which frees up a big chunk of
the context image. BSpec indicates 12544 DWORDs (13 pages), plus
one page for PPHWSP.
Please notice that, when looking at the BSpec context image table,
the right filter has to be applied as some rows are excluded for
specific GENs. Also, some rows apply per-subslice (for the
calculation above, we have supposed I915_MAX_SUBSLICES = 8).
v2: Rebase.
v3: Use the right size as per the BSpec.
v4:
- Rebased on top of the default context size (Rodrigo)
- Clarify in the commit message where the subslice calculation
comes from.
v5: s/12538/12544/ (Daniele)
BSpec: 18907
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Ben Widawsky <benjamin.widawsky@intel.com> (older version)
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1515711307-28979-2-git-send-email-oscar.mateo@intel.com
In order to prevent a race condition where we may end up overaccounting
the active state and leaving the busy-stats believing the GPU is 100%
busy, lock out the tasklet while we reconstruct the busy state. There is
no direct spinlock guard for the execlists->port[], so we need to
utilise tasklet_disable() as a synchronous barrier to prevent it, the
only writer to execlists->port[], from running at the same time as the
enable.
Fixes: 4900727d35 ("drm/i915/pmu: Reconstruct active state on starting busy-stats")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180115092041.13509-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
We have a hole in our busy-stat accounting if the pmu is enabled during
a long running batch, the pmu will not start accumulating busy-time
until the next context switch. This then fails tests that are only
sampling a single batch.
v2: Count each active port just once (context in/out events are only on
the first and last assignment to a port).
v3: Avoid hardcoding knowledge of 2 submission ports
Fixes: 30e17b7847 ("drm/i915: Engine busy time tracking")
Testcase: igt/perf_pmu/busy-start
Testcase: igt/perf_pmu/busy-double-start
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/20180111073031.14614-1-chris@chris-wilson.co.uk
Geminilake requires the 3D driver to select whether barriers are
intended for compute shaders, or tessellation control shaders, by
whacking a "Barrier Mode" bit in SLICE_COMMON_ECO_CHICKEN1 when
switching pipelines. Failure to do this properly can result in GPU
hangs.
Unfortunately, this means it needs to switch mid-batch, so only
userspace can properly set it. To facilitate this, the kernel needs
to whitelist the register.
The workarounds page currently tags this as applying to Broxton only,
but that doesn't make sense. The documentation for the register it
references says the bit userspace is supposed to toggle only exists on
Geminilake. Empirically, the Mesa patch to toggle this bit appears to
fix intermittent GPU hangs in tessellation control shader barrier tests
on Geminilake; we haven't seen those hangs on Broxton.
v2: Mention WA #0862 in the comment (it doesn't have a name).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180105085905.9298-1-kenneth@whitecape.org
Looking at a CI failure with an ominous line of
[ 362.550715] hangcheck current seqno ffffff6b, last ffffff8c, hangcheck ffffff6b [6016 ms], inflight 118
with no apparent cause for the seqno to be negative, left me wondering
if someone had scribbled over the HWSP. So include the HWSP in the
engine dump to see if there are more signs of random scribbling.
v2: Fix row pointer, i is now incremented by 8 so doesn't need scaling
by 8, and we don't need to keep volatile here as the status_page isn't
marked up as volatile itself.
v3: Use hexdump, with suppression of identical lines. (Tvrtko)
Which results in
HWSP:
00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
*
00000040 00000001 00000000 00000018 00000002 00000001 00000000 00000018 00000000
00000060 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000003
00000080 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
*
000000c0 00000002 00000000 00000000 00000000 00000000 00000000 00000000 00000000
000000e0 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
*
instead of 128 lines of mostly 0s.
v4: Tidy up the locals
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>
Link: https://patchwork.freedesktop.org/patch/msgid/20171222182521.18106-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Sagar noticed the check can be consolidated between the engine stats
implementation and the PMU.
My first choice was a static inline helper but that got into include
ordering mess quickly fast so I went with a macro instead. At some point
we should perhaps looking into taking out the non-ringubffer bits from
intel_ringbuffer.h into a new intel_engine.h or something.
v2: Use engine->flags. (Chris Wilson)
v3: Rebase and mark GuC as not yet supported. (Chris Wilson)
v4: Move flag setting to intel_engines_reset_default_submission.
(Chris Wilson)
v5: Move flag setting to logical_ring_setup.
v6: intel_engines_reset_default_submission is the wrong place to set the
flag - it needs to be in execlists_set_default_submission. (Sagar)
v7: Flag setting in logical_ring_setup is not required. (Chris)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Suggested-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com> (v6)
Link: https://patchwork.freedesktop.org/patch/msgid/20171129102805.22690-1-tvrtko.ursulin@linux.intel.com
