One impact of commit 047a1b877e ("dma-buf & drm/amdgpu: remove
dma_resv workaround") is that it stores many, many more fences. Whereas
adding an exclusive fence used to remove the shared fence list, that
list is now preserved and the write fences included into the list. Not
just a single write fence, but now a write/read fence per context. That
causes us to have to track more fences than before (albeit half of those
are redundant), and we trigger more interrupts for multi-engine
workloads.
As part of reducing the impact from handling more signaling, we observe
we only need to kick the signal worker after adding a fence iff we have
good cause to believe that there is work to be done in processing the
fence i.e. we either need to enable the interrupt or the request is
already complete but we don't know if we saw the interrupt and so need
to check signaling.
References: 047a1b877e ("dma-buf & drm/amdgpu: remove dma_resv workaround")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Karolina Drobnik <karolina.drobnik@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/d7b953c7a4ba747c8196a164e2f8c5aef468d048.1657289332.git.karolina.drobnik@intel.com
We employ a "waitboost" heuristic to detect when userspace is stalled
waiting for results from earlier execution. Under latency sensitive work
mixed between the gpu/cpu, the GPU is typically under-utilised and so
RPS sees that low utilisation as a reason to downclock the frequency,
causing longer stalls and lower throughput. The user left waiting for
the results is not impressed.
On applying commit 047a1b877e ("dma-buf & drm/amdgpu: remove dma_resv
workaround") it was observed that deinterlacing h264 on Haswell
performance dropped by 2-5x. The reason being that the natural workload
was not intense enough to trigger RPS (using HW evaluation intervals) to
upclock, and so it was depending on waitboosting for the throughput.
Commit 047a1b877e ("dma-buf & drm/amdgpu: remove dma_resv workaround")
changes the composition of dma-resv from keeping a single write fence +
multiple read fences, to a single array of multiple write and read
fences (a maximum of one pair of write/read fences per context). The
iteration order was also changed implicitly from all-read fences then
the single write fence, to a mix of write fences followed by read
fences. It is that ordering change that belied the fragility of
waitboosting.
Currently, a waitboost is inspected at the point of waiting on an
outstanding fence. If the GPU is backlogged such that we haven't yet
stated the request we need to wait on, we force the GPU to upclock until
the completion of that request. By changing the order in which we waited
upon requests, we ended up waiting on those requests in sequence and as
such we saw that each request was already started and so not a suitable
candidate for waitboosting.
Instead of asking whether to boost each fence in turn, we can look at
whether boosting is required for the dma-resv ensemble prior to waiting
on any fence, making the heuristic more robust to the order in which
fences are stored in the dma-resv.
Reported-by: Thomas Voegtle <tv@lio96.de>
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/6284
Fixes: 047a1b877e ("dma-buf & drm/amdgpu: remove dma_resv workaround")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Karolina Drobnik <karolina.drobnik@intel.com>
Tested-by: Thomas Voegtle <tv@lio96.de>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/07e05518d9f6620d20cc1101ec1849203fe973f9.1657289332.git.karolina.drobnik@intel.com
drm-misc-next for $kernel-version:
UAPI Changes:
Cross-subsystem Changes:
Core Changes:
* crtc: Remove unnessary include statements from drm_crtc.h, plus
fallout in drivers
* edid: More use of struct drm_edid; implement HF-EEODB extension
Driver Changes:
* bridge:
* anx7625: Implement HDP timeout via callback; Cleanups
* fsl-ldb: Drop DE flip; Modesetting fixes
* imx: Depend on ARCH_MXC
* sil8620: Fix off-by-one
* ti-sn65dsi86: Convert to atomic modesetting
* ingenic: Fix display at maximum resolution
* panel:
* simple: Add support for HannStar HSD101PWW2, plus DT bindings; Add
support for ETML0700Y5DHA, plus DT bindings
* rockchip: Fixes
* vc4: Cleanups
* vmwgfx: Cleanups
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/YsaHq1pvE699NtOM@linux-uq9g
Although all DSS belong to a single pool on Xe_HP platforms (i.e.,
they're not organized into slices from a topology point of view), we do
still need to pass 'group' and 'instance' targets when steering register
accesses to a specific instance of a per-DSS multicast register. The
rules for how to determine group and instance IDs (which previously used
legacy terms "slice" and "subslice") varies by platform. Some platforms
determine steering by gslice membership, some platforms by cslice
membership, and future platforms may have other rules.
Since looping over each DSS and performing steered unicast register
accesses is a relatively common pattern, let's add a dedicated iteration
macro to handle this (and replace the platform-specific "instdone" loop
we were using previously. This will avoid the calling code needing to
figure out the details about how to obtain steering IDs for a specific
DSS.
Most of the places where we use this new loop are in the GPU errorstate
code at the moment, but we do have some additional features coming in
the future that will also need to loop over each DSS and steer some
register accesses accordingly.
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Atwood <matthew.s.atwood@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220701232006.1016135-1-matthew.d.roper@intel.com
The global sseu config is applicable only to gen11 platforms where
concurrent media, render and OA use cases may cause some subslices to be
turned off and hence lose NOA configuration. Ideally we want to return
ENODEV for non-gen11 platforms, however, this has shipped with gfx12, so
disable only for gfx12.50+.
v2: gfx12 is already shipped with this, disable for gfx12.50+ (Lionel)
v3: (Matt)
- Update commit message and replace "12.5" with "12.50"
- Replace DRM_DEBUG() with driver specific drm_dbg()
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220707193002.2859653-2-umesh.nerlige.ramappa@intel.com
Both error-capture and relay-logging mechanism use the GuC
log infrastructure. That means the KMD must send a log flush
complete notification back to GuC after reading the data out.
This call is currently being sent synchronously.
However, synchronous H2Gs cause problems when the system is
backed up. There is no need for this to be synchronous. The
KMD wasn't even looking at the return status from it. So make
it asynchronous and then there is no issue about time outs.
Signed-off-by: Alan Previn <alan.previn.teres.alexis@intel.com>
Reviewed-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220607002314.1451656-2-alan.previn.teres.alexis@intel.com
JSP is based on ICP and we don't really need to differentiate
between the two. So let's just delcare JSP to be ICP.
The only slight change here is for Wa_14011294188 which we
used to apply for JSP but now we'll only apply to MCC. This
should be fine since the issue being dealt with was introduced
in TGP and inherited into MCC. JSP being derived from ICP
should not need this workaround.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220630150600.24611-4-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Currently shrinkers are anonymous objects. For debugging purposes they
can be identified by count/scan function names, but it's not always
useful: e.g. for superblock's shrinkers it's nice to have at least an
idea of to which superblock the shrinker belongs.
This commit adds names to shrinkers. register_shrinker() and
prealloc_shrinker() functions are extended to take a format and arguments
to master a name.
In some cases it's not possible to determine a good name at the time when
a shrinker is allocated. For such cases shrinker_debugfs_rename() is
provided.
The expected format is:
<subsystem>-<shrinker_type>[:<instance>]-<id>
For some shrinkers an instance can be encoded as (MAJOR:MINOR) pair.
After this change the shrinker debugfs directory looks like:
$ cd /sys/kernel/debug/shrinker/
$ ls
dquota-cache-16 sb-devpts-28 sb-proc-47 sb-tmpfs-42
mm-shadow-18 sb-devtmpfs-5 sb-proc-48 sb-tmpfs-43
mm-zspool:zram0-34 sb-hugetlbfs-17 sb-pstore-31 sb-tmpfs-44
rcu-kfree-0 sb-hugetlbfs-33 sb-rootfs-2 sb-tmpfs-49
sb-aio-20 sb-iomem-12 sb-securityfs-6 sb-tracefs-13
sb-anon_inodefs-15 sb-mqueue-21 sb-selinuxfs-22 sb-xfs:vda1-36
sb-bdev-3 sb-nsfs-4 sb-sockfs-8 sb-zsmalloc-19
sb-bpf-32 sb-pipefs-14 sb-sysfs-26 thp-deferred_split-10
sb-btrfs:vda2-24 sb-proc-25 sb-tmpfs-1 thp-zero-9
sb-cgroup2-30 sb-proc-39 sb-tmpfs-27 xfs-buf:vda1-37
sb-configfs-23 sb-proc-41 sb-tmpfs-29 xfs-inodegc:vda1-38
sb-dax-11 sb-proc-45 sb-tmpfs-35
sb-debugfs-7 sb-proc-46 sb-tmpfs-40
[roman.gushchin@linux.dev: fix build warnings]
Link: https://lkml.kernel.org/r/Yr+ZTnLb9lJk6fJO@castle
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lkml.kernel.org/r/20220601032227.4076670-4-roman.gushchin@linux.dev
Signed-off-by: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Kent Overstreet <kent.overstreet@gmail.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
If the move or clear operation somehow fails, and the memory underneath
is not cleared, like when moving to lmem, then we currently fallback to
memcpy or memset. However with small-BAR systems this fallback might no
longer be possible. For now we use the set_wedged sledgehammer if we
ever encounter such a scenario, and mark the object as borked to plug
any holes where access to the memory underneath can happen. Add some
basic selftests to exercise this.
v2:
- In the selftests make sure we grab the runtime pm around the reset.
Also make sure we grab the reset lock before checking if the device
is wedged, since the wedge might still be in-progress and hence the
bit might not be set yet.
- Don't wedge or put the object into an unknown state, if the request
construction fails (or similar). Just returning an error and
skipping the fallback should be safe here.
- Make sure we wedge each gt. (Thomas)
- Peek at the unknown_state in io_reserve, that way we don't have to
export or hand roll the fault_wait_for_idle. (Thomas)
- Add the missing read-side barriers for the unknown_state. (Thomas)
- Some kernel-doc fixes. (Thomas)
v3:
- Tweak the ordering of the set_wedged, also add FIXME.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Kenneth Graunke <kenneth@whitecape.org>
Cc: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220629174350.384910-11-matthew.auld@intel.com
On small BAR configurations, when dealing with I915_MEMORY_CLASS_DEVICE
allocations, we assume that by default, all userspace allocations should
be placed in the non-CPU visible portion. Note that dumb buffers are
not included here, since these are not "GPU accelerated" and likely need
CPU access. We choose to just always set GPU_ONLY, and let the backend
figure out if that should be ignored or not, for example on full BAR
systems.
In a later patch userspace will be able to provide a hint if CPU access
to the buffer is needed.
v2(Thomas)
- Apply GPU_ONLY on all discrete devices, but only if the BO can be
placed in LMEM. Down in the depths this should be turned into a noop,
where required, and as an annotation it still make some sense. If we
apply it regardless of the placements then we end up needing to check
the placements during exec capture. Also it's slightly inconsistent
since the NEEDS_CPU_ACCESS can only be applied on objects that can be
placed in LMEM. The other annoyance would be gem_create_ext vs plain
gem_create, if we were to always apply GPU_ONLY.
Testcase: igt@gem-create@create-ext-cpu-access-sanity-check
Testcase: igt@gem-create@create-ext-cpu-access-big
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Kenneth Graunke <kenneth@whitecape.org>
Cc: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220629174350.384910-5-matthew.auld@intel.com
