IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Now that the tasklet completely controls scheduling of the requests, and
we postpone scheduling out the old requests, we can keep a hanging
virtual request bound to the engine on which it hung, and remove it from
te queue. On release, it will be returned to the same engine and remain
in its queue until it is scheduled; after which point it will become
eligible for transfer to a sibling. Instead, we could opt to resubmit the
request along the virtual engine on unhold, making it eligible for load
balancing immediately -- but that seems like a pointless optimisation
for a hanging context.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201224135544.1713-8-chris@chris-wilson.co.uk
Having recognised that we do not change the sibling until we schedule
out, we can then defer the decision to resubmit the virtual engine from
the unwind of the active queue to scheduling out of the virtual context.
This improves our resilence in virtual engine scheduling, and should
eliminate the rare cases of gem_exec_balance failing.
By keeping the unwind order intact on the local engine, we can preserve
data dependency ordering while doing a preempt-to-busy pass until we
have determined the new ELSP. This means that if we try to timeslice
between a virtual engine and a data-dependent ordinary request, the pair
will maintain their relative ordering and we will avoid the
resubmission, cancelling the timeslicing until further change.
The dilemma though is that we then may end up in a situation where the
'demotion' of the virtual request to an ordinary request in the engine
queue results in filling the ELSP[] with virtual requests instead of
spreading the load across the engines. To compensate for this, we mark
each virtual request and refuse to resubmit a virtual request in the
secondary ELSP slots, thus forcing subsequent virtual requests to be
scheduled out after timeslicing. By delaying the decision until we
schedule out, we will avoid unnecessary resubmission.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2079
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2098
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201224135544.1713-7-chris@chris-wilson.co.uk
The issue with stale virtual breadcrumbs remain. Now we have the problem
that if the irq-signaler is still referencing the stale breadcrumb as we
transfer it to a new sibling, the list becomes spaghetti. This is a very
small window, but that doesn't stop it being hit infrequently. To
prevent the lists being tangled (the iterator starting on one engine's
b->signalers but walking onto another list), always decouple the virtual
breadcrumb on schedule-out and make sure that the walker has stepped out
of the lists.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201224135544.1713-5-chris@chris-wilson.co.uk
Inside schedule_out, we do extra work upon idling the context, such as
updating the runtime, kicking off retires, kicking virtual engines.
However, if we are in a series of processing single requests per
contexts, we may find ourselves scheduling out the context, only to
immediately schedule it back in during dequeue. This is just extra work
that we can avoid if we keep the context marked as inflight across the
dequeue. This becomes more significant later on for minimising virtual
engine misses.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201224135544.1713-4-chris@chris-wilson.co.uk
Once a virtual engine has been bound to a sibling, it will remain bound
until we finally schedule out the last active request. We can not rebind
the context to a new sibling while it is inflight as the context save
will conflict, hence we wait. As we cannot then use any other sibliing
while the context is inflight, only kick the bound sibling while it
inflight and upon scheduling out the kick the rest (so that we can swap
engines on timeslicing if the previously bound engine becomes
oversubscribed).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201224135544.1713-3-chris@chris-wilson.co.uk
Rather than having special case code for opportunistically calling
process_csb() and performing a direct submit while holding the engine
spinlock for submitting the request, simply call the tasklet directly.
This allows us to retain the direct submission path, including the CS
draining to allow fast/immediate submissions, without requiring any
duplicated code paths, and most importantly greatly simplifying the
control flow by removing reentrancy. This will enable us to close a few
races in the virtual engines in the next few patches.
The trickiest part here is to ensure that paired operations (such as
schedule_in/schedule_out) remain under consistent locking domains,
e.g. when pulled outside of the engine->active.lock
v2: Use bh kicking, see commit 3c53776e29 ("Mark HI and TASKLET
softirq synchronous").
v3: Update engine-reset to be tasklet aware
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201224135544.1713-1-chris@chris-wilson.co.uk
We assume that both timestamps are driven off the same clock [reported
to userspace as I915_PARAM_CS_TIMESTAMP_FREQUENCY]. Verify that this is
so by reading the timestamp registers around a busywait (on an otherwise
idle engine so there should be no preemptions).
v2: Icelake (not ehl, nor tgl) seems to be using a fixed 80ns interval
for, and only for, CTX_TIMESTAMP -- or it may be GPU frequency and the
test is always running at maximum frequency?. As far as I can tell, this
isolated change in behaviour is undocumented.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201223122359.22562-1-chris@chris-wilson.co.uk
The caller determines if the failure is an error or not, so avoid
warning when we will try again and succeed. For example,
<7> [111.319321] [drm:intel_guc_fw_upload [i915]] GuC status 0x20
<3> [111.319340] i915 0000:00:02.0: [drm] *ERROR* GuC load failed: status = 0x00000020
<3> [111.319606] i915 0000:00:02.0: [drm] *ERROR* GuC load failed: status: Reset = 0, BootROM = 0x10, UKernel = 0x00, MIA = 0x00, Auth = 0x00
<7> [111.320045] [drm:__uc_init_hw [i915]] GuC fw load failed: -110; will reset and retry 2 more time(s)
<7> [111.322978] [drm:intel_guc_fw_upload [i915]] GuC status 0x8002f0ec
should not have been reported as a _test_ failure, as the GuC was
successfully loaded on the second attempt and the system remained
operational.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2797
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201214100949.11387-2-chris@chris-wilson.co.uk
Since we allow removing the timeline map at runtime, there is a risk
that rq->hwsp points into a stale page. To control that risk, we hold
the RCU read lock while reading *rq->hwsp, but we missed a couple of
important barriers. First, the unpinning / removal of the timeline map
must be after all RCU readers into that map are complete, i.e. after an
rcu barrier (in this case courtesy of call_rcu()). Secondly, we must
make sure that the rq->hwsp we are about to dereference under the RCU
lock is valid. In this case, we make the rq->hwsp pointer safe during
i915_request_retire() and so we know that rq->hwsp may become invalid
only after the request has been signaled. Therefore is the request is
not yet signaled when we acquire rq->hwsp under the RCU, we know that
rq->hwsp will remain valid for the duration of the RCU read lock.
This is a very small window that may lead to either considering the
request not completed (causing a delay until the request is checked
again, any wait for the request is not affected) or dereferencing an
invalid pointer.
Fixes: 3adac4689f ("drm/i915: Introduce concept of per-timeline (context) HWSP")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: <stable@vger.kernel.org> # v5.1+
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201218122421.18344-1-chris@chris-wilson.co.uk
Matthew Brost pointed out that the while-loop on a shared breadcrumb was
inherently fraught with danger as it competed with the other users of
the breadcrumbs. However, in order to completely drain the re-arming irq
worker, the while-loop is a necessity, despite my optimism that we could
force cancellation with a couple of irq_work invocations.
Given that we can't merely drop the while-loop, use an activity counter on
the breadcrumbs to detect when we are parking the breadcrumbs for the
last time.
Based on a patch by Matthew Brost.
Reported-by: Matthew Brost <matthew.brost@intel.com>
Suggested-by: Matthew Brost <matthew.brost@intel.com>
Fixes: 9d5612ca16 ("drm/i915/gt: Defer enabling the breadcrumb interrupt to after submission")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201217091524.10258-1-chris@chris-wilson.co.uk
Pull drm updates from Dave Airlie:
"Not a huge amount of big things here, AMD has support for a few new HW
variants (vangogh, green sardine, dimgrey cavefish), Intel has some
more DG1 enablement. We have a few big reworks of the TTM layers and
interfaces, GEM and atomic internal API reworks cross tree. fbdev is
marked orphaned in here as well to reflect the current reality.
core:
- documentation updates
- deprecate DRM_FORMAT_MOD_NONE
- atomic crtc enable/disable rework
- GEM convert drivers to gem object functions
- remove SCATTER_LIST_MAX_SEGMENT
sched:
- avoid infinite waits
ttm:
- remove AGP support
- don't modify caching for swapout
- ttm pinning rework
- major TTM reworks
- new backend allocator
- multihop support
vram-helper:
- top down BO placement fix
- TTM changes
- GEM object support
displayport:
- DP 2.0 DPCD prep work
- DP MST extended DPCD caps
fbdev:
- mark as orphaned
amdgpu:
- Initial Vangogh support
- Green Sardine support
- Dimgrey Cavefish support
- SG display support for renoir
- SMU7 improvements
- gfx9+ modiifier support
- CI BACO fixes
radeon:
- expose voltage via hwmon on SUMO
amdkfd:
- fix unique id handling
i915:
- more DG1 enablement
- bigjoiner support
- integer scaling filter support
- async flip support
- ICL+ DSI command mode
- Improve display shutdown
- Display refactoring
- eLLC machine fbdev loading fix
- dma scatterlist fixes
- TGL hang fixes
- eLLC display buffer caching on SKL+
- MOCS PTE seeting for gen9+
msm:
- Shutdown hook
- GPU cooling device support
- DSI 7nm and 10nm phy/pll updates
- sm8150/sm2850 DPU support
- GEM locking re-work
- LLCC system cache support
aspeed:
- sysfs output config support
ast:
- LUT fix
- new display mode
gma500:
- remove 2d framebuffer accel
panfrost:
- move gpu reset to a worker
exynos:
- new HDMI mode support
mediatek:
- MT8167 support
- yaml bindings
- MIPI DSI phy code moved
etnaviv:
- new perf counter
- more lockdep annotation
hibmc:
- i2c DDC support
ingenic:
- pixel clock reset fix
- reserved memory support
- allow both DMA channels at once
- different pixel format support
- 30/24/8-bit palette modes
tilcdc:
- don't keep vblank irq enabled
vc4:
- new maintainer added
- DSI registration fix
virtio:
- blob resource support
- host visible and cross-device support
- uuid api support"
* tag 'drm-next-2020-12-11' of git://anongit.freedesktop.org/drm/drm: (1754 commits)
drm/amdgpu: Initialise drm_gem_object_funcs for imported BOs
drm/amdgpu: fix size calculation with stolen vga memory
drm/amdgpu: remove amdgpu_ttm_late_init and amdgpu_bo_late_init
drm/amdgpu: free the pre-OS console framebuffer after the first modeset
drm/amdgpu: enable runtime pm using BACO on CI dGPUs
drm/amdgpu/cik: enable BACO reset on Bonaire
drm/amd/pm: update smu10.h WORKLOAD_PPLIB setting for raven
drm/amd/pm: remove one unsupported smu function for vangogh
drm/amd/display: setup system context for APUs
drm/amd/display: add S/G support for Vangogh
drm/amdkfd: Fix leak in dmabuf import
drm/amdgpu: use AMDGPU_NUM_VMID when possible
drm/amdgpu: fix sdma instance fw version and feature version init
drm/amd/pm: update driver if version for dimgrey_cavefish
drm/amd/display: 3.2.115
drm/amd/display: [FW Promotion] Release 0.0.45
drm/amd/display: Revert DCN2.1 dram_clock_change_latency update
drm/amd/display: Enable gpu_vm_support for dcn3.01
drm/amd/display: Fixed the audio noise during mode switching with HDCP mode on
drm/amd/display: Add wm table for Renoir
...
Tigerlake is plagued by spontaneous DMAR faults [reason 7, next page
table ptr is invalid] which lead to GPU hangs. These faults occur when
an iommu map is immediately reused. Adding further clflushes and
barriers around either the GTT PTE or iommu PTE updates do not prevent
the faults. So far the only effect has been from inducing a delay
between reuse of the iommu on the GPU, and applying the delay at the
iommu map allows for the smallest stable delay.
Note that such a delay is hideous and clearly does not fix the root cause,
and so should only be a bandaid until a complete solution is found. The
delay was determined by running igt/gem_exec_fence/parallel in a loop for
a few hours (unpatched MTBF is about 10s).
We have also seen such DMAR fault [reason 7] errors on other platforms,
notably gen9-gen11, but so far it has only been trivially and
consistently reproduced on Tigerlake.
v2: Leave a tell-tale to know when we apply the vt'd quirk, and as a
reminder to remove it again. Hopefully.
Testcase: igt/gem_exec_fence/parallel
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Acked-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201209164008.5487-2-chris@chris-wilson.co.uk
The use of "masked" in this function is due to its history. Once upon a
time it received a mask and a value as parameter. Since
commit eeec73f8a4 ("drm/i915/gt: Skip rmw for masked registers")
that is not true anymore and now there is a clear and a set parameter.
Depending on the case, that can still be thought as a mask and value,
but there are some subtle differences: what we clear doesn't need to be
the same bits we are setting, particularly when we are using masked
registers.
The fact that we also have "masked registers", i.e. registers whose mask
is stored in the upper 16 bits of the register, makes it even more
confusing, because "masked" in wa_write_masked_or() has little to do
with masked registers, but rather refers to the old mask parameter the
function received (that can also, but not exclusively, be used to write
to masked register).
Avoid the ambiguity and misnomer by renaming it to something else,
hopefully less confusing: wa_write_clr_set(), to designate that we are
doing both clr and set operations in the register.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20201209045246.2905675-2-lucas.demarchi@intel.com
When using masked registers, there is nothing to clear since a masked
register has the mask in the upper 16b: we can just write to the
location we want and use the mask to control what bits we are writing
to.
However that doesn't mean we don't want to read back the register and
check the value actually matched what we wanted to write, i.e. that
the WA stick. That should be an explicit opt-out for registers that are
either write-only or that are affected by hardware misbehavior.
Moreover both wa_masked_en() and wa_masked_dis() check the WA stick, so
skipping the check just because the field is more than 1 bit is
surprising and error-prone.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20201209045246.2905675-1-lucas.demarchi@intel.com
