| Age | Commit message (Collapse) | Author |
|---|
|
Original rationale for those had been the reduced cost of mntput()
for the stuff that is mounted somewhere. Mount refcount increments and
decrements are frequent; what's worse, they tend to concentrate on the
same instances and cacheline pingpong is quite noticable.
As the result, mount refcounts are per-cpu; that allows a very cheap
increment. Plain decrement would be just as easy, but decrement-and-test
is anything but (we need to add the components up, with exclusion against
possible increment-from-zero, etc.).
Fortunately, there is a very common case where we can tell that decrement
won't be the final one - if the thing we are dropping is currently
mounted somewhere. We have an RCU delay between the removal from mount
tree and dropping the reference that used to pin it there, so we can
just take rcu_read_lock() and check if the victim is mounted somewhere.
If it is, we can go ahead and decrement without and further checks -
the reference we are dropping is not the last one. If it isn't, we
get all the fun with locking, carefully adding up components, etc.,
but the majority of refcount decrements end up taking the fast path.
There is a major exception, though - pipes and sockets. Those live
on the internal filesystems that are not going to be mounted anywhere.
They are not going to be _un_mounted, of course, so having to take the
slow path every time a pipe or socket gets closed is really obnoxious.
Solution had been to mark them as long-lived ones - essentially faking
"they are mounted somewhere" indicator.
With minor modification that works even for ones that do eventually get
dropped - all it takes is making sure we have an RCU delay between
clearing the "mounted somewhere" indicator and dropping the reference.
There are some additional twists (if you want to drop a dozen of such
internal mounts, you'd be better off with clearing the indicator on
all of them, doing an RCU delay once, then dropping the references),
but in the basic form it had been
* use kern_mount() if you want your internal mount to be
a long-term one.
* use kern_unmount() to undo that.
Unfortunately, the things did rot a bit during the mount API reshuffling.
In several cases we have lost the "fake the indicator" part; kern_unmount()
on the unmount side remained (it doesn't warn if you use it on a mount
without the indicator), but all benefits regaring mntput() cost had been
lost.
To get rid of that bitrot, let's add a new helper that would work
with fs_context-based API: fc_mount_longterm(). It's a counterpart
of fc_mount() that does, on success, mark its result as long-term.
It must be paired with kern_unmount() or equivalents.
Converted:
1) mqueue (it used to use kern_mount_data() and the umount side
is still as it used to be)
2) hugetlbfs (used to use kern_mount_data(), internal mount is
never unmounted in this one)
3) i915 gemfs (used to be kern_mount() + manual remount to set
options, still uses kern_unmount() on umount side)
4) v3d gemfs (copied from i915)
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
|
|
Replace multi-line SPDX license headers with single-line
equivalents (// SPDX-License-Identifier: MIT or /* ... */ for
headers), as preferred by current kernel coding style.
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Reviewed-by: Nitin Gote <nitin.r.gote@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: MikoĊaj Wasiak <mikolaj.wasiak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250327232629.2939-1-andi.shyti@linux.intel.com
|
|
If i915 does not want to use huge pages there is a) no point in setting up
the private mount and b) should former fail, it is misleading to log THP
support is disabled in the caller, which does not even know if callee
tried to enable it.
Fix both by restructuring the flow in i915_gemfs_init and at the same time
note the failure to set it up in all cases.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Eero Tamminen <eero.t.tamminen@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220429100414.647857-2-tvrtko.ursulin@linux.intel.com
|
|
We have a statement from HW designers that the GPU read regression when
using 2M pages was fixed from Icelake onwards, which was also confirmed
by bencharking Eero did last year:
"""
When IOMMU is disabled, enabling THP causes following perf changes on
TGL-H (GT1):
10-15% SynMark Batch[0-3]
5-10% MemBW GPU texture, SynMark ShMapVsm
3-5% SynMark TerrainFly* + Geom* + Fill* + CSCloth + Batch4
1-3% GpuTest Triangle, SynMark TexMem* + DeferredAA + Batch[5-7]
+ few others
-7% MemBW GPU blend
In the above 3D benchmark names, * means all the variants of tests with
the same prefix. For example "SynMark TexMem*", means both TexMem128 &
TexMem512 tests in the synthetic (Intel internal) SynMark test suite.
In the (public, but proprietary) GfxBench & GLB(enchmark) test suites,
there are both onscreen and offscreen variants of each test. Unless
explicitly stated otherwise, numbers are for both variants.
All tests are run with FullHD monitor. All tests are fullscreen except
for GLB and GpuTest ones, which are run in 1/2 screen window (GpuTest
triangle is run both in fullscreen and 1/2 screen window).
"""
Since the only regression is MemBW GPU blend, against many more gains,
it sounds it is time to enable THP on Gen11+.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
References: https://gitlab.freedesktop.org/drm/intel/-/issues/430
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Eero Tamminen <eero.t.tamminen@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220429100414.647857-1-tvrtko.ursulin@linux.intel.com
|
|
Continuation of the effort to declutter i915_drv.h.
Also, component specific helpers which consult the iommu/virtualization
helpers moved to respective component source/header files as appropriate.
v2:
* s/dev_priv/i915/ in intel_scanout_needs_vtd_wa. (Lucas)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220329090204.2324499-1-tvrtko.ursulin@linux.intel.com
[tursulin: fixup conflict in i915_drv.h]
|
|
With both integrated and discrete Intel GPUs in a system, the current
global check of intel_iommu_gfx_mapped, as done from intel_vtd_active()
may not be completely accurate.
In this patch we add i915 parameter to intel_vtd_active() in order to
prepare it for multiple GPUs and we also change the check away from Intel
specific intel_iommu_gfx_mapped (global exported by the Intel IOMMU
driver) to probing the presence of IOMMU on a specific device using
device_iommu_mapped().
This will return true both for IOMMU pass-through and address translation
modes which matches the current behaviour. If in the future we wanted to
distinguish between these two modes we could either use
iommu_get_domain_for_dev() and check for __IOMMU_DOMAIN_PAGING bit
indicating address translation, or ask for a new API to be exported from
the IOMMU core code.
v2:
* Check for dmar translation specifically, not just iommu domain. (Baolu)
v3:
* Go back to plain "any domain" check for now, rewrite commit message.
v4:
* Use device_iommu_mapped. (Robin, Baolu)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Lu Baolu <baolu.lu@linux.intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211126141424.493753-1-tvrtko.ursulin@linux.intel.com
|
|
vfs_kernel_mount() modifies the passed in mount options, leaving us with
"huge", instead of "huge=within_size". Normally this shouldn't matter
with the usual module load/unload flow, however with the core_hotunplug
IGT we are hitting the following, when re-probing the memory regions:
i915 0000:00:02.0: [drm] Transparent Hugepage mode 'huge'
tmpfs: Bad value for 'huge'
[drm] Unable to create a private tmpfs mount, hugepage support will be disabled(-22).
References: https://gitlab.freedesktop.org/drm/intel/-/issues/4651
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211126110843.2028582-1-matthew.auld@intel.com
|
|
Usage of Transparent Hugepages was disabled in 9987da4b5dcf
("drm/i915: Disable THP until we have a GPU read BW W/A"), but since it
appears majority of performance regressions reported with an enabled IOMMU
can be almost eliminated by turning them on, lets just do that.
To err on the side of safety we keep the current default in cases where
IOMMU is not active, and only when it is default to the "huge=within_size"
mode. Although there probably would be wins to enable them throughout,
more extensive testing across benchmarks and platforms would need to be
done.
With the patch and IOMMU enabled my local testing on a small Skylake part
shows OglVSTangent regression being reduced from ~14% (IOMMU on versus
IOMMU off) to ~2% (same comparison but with THP on).
More detailed testing done in the below referenced Gitlab issue by Eero:
Skylake GT4e:
Performance drops from enabling IOMMU:
30-35% SynMark CSDof
20-25% Unigine Heaven, MemBW GPU write, SynMark VSTangent
~20% GLB Egypt (1/2 screen window)
10-15% GLB T-Rex (1/2 screen window)
8-10% GfxBench T-Rex, MemBW GPU blit
7-8% SynMark DeferredAA + TerrainFly* + ZBuffer
6-7% GfxBench Manhattan 3.0 + 3.1, SynMark TexMem128 & CSCloth
5-6% GfxBench CarChase, Unigine Valley
3-5% GfxBench Vulkan & GL AztecRuins + ALU2, MemBW GPU texture,
SynMark Fill*, Deferred, TerrainPan*
1-2% Most of the other tests
With the patch drops become:
20-25% SynMark TexMem*
15-20% GLB Egypt (1/2 screen window)
10-15% GLB T-Rex (1/2 screen window)
4-7% GfxBench T-Rex, GpuTest Triangle
1-8% GfxBench ALU2 (offscreen 1%, onscreen 8%)
3% GfxBench Manhattan 3.0, SynMark CSDof
2-3% Unigine Heaven + Valley, MemBW GPU texture
1-3 GfxBench Manhattan 3.1 + CarChase + Vulkan & GL AztecRuins
Broxton:
Performance drops from IOMMU, without patch:
30% MemBW GPU write
25% SynMark ZBuffer + Fill*
20% MemBW GPU blit
15% MemBW GPU blend, GpuTest Triangle
10-15% MemBW GPU texture
10% GLB Egypt, Unigine Heaven (had hangs), SynMark TerrainFly*
7-9% GLB T-Rex, GfxBench Manhattan 3.0 + T-Rex,
SynMark Deferred* + TexMem*
6-8% GfxBench CarChase, Unigine Valley,
SynMark CSCloth + ShMapVsm + TerrainPan*
5-6% GfxBench Manhattan 3.1 + GL AztecRuins,
SynMark CSDof + TexFilterTri
2-4% GfxBench ALU2, SynMark DrvRes + GSCloth + ShMapPcf + Batch[0-5] +
TexFilterAniso, GpuTest GiMark + 32-bit Julia
And with patch:
15-20% MemBW GPU texture
10% SynMark TexMem*
8-9% GLB Egypt (1/2 screen window)
4-5% GLB T-Rex (1/2 screen window)
3-6% GfxBench Manhattan 3.0, GpuTest FurMark,
SynMark Deferred + TexFilterTri
3-4% GfxBench Manhattan 3.1 + T-Rex, SynMark VSInstancing
2-4% GpuTest Triangle, SynMark DeferredAA
2-3% Unigine Heaven + Valley
1-3% SynMark Terrain*
1-2% GfxBench CarChase, SynMark TexFilterAniso + ZBuffer
Tigerlake-H:
20-25% MemBW GPU texture
15-20% GpuTest Triangle
13-15% SynMark TerrainFly* + DeferredAA + HdrBloom
8-10% GfxBench Manhattan 3.1, SynMark TerrainPan* + DrvRes
6-7% GfxBench Manhattan 3.0, SynMark TexMem*
4-8% GLB onscreen Fill + T-Rex + Egypt (more in onscreen than
offscreen versions of T-Rex/Egypt)
4-6% GfxBench CarChase + GLES AztecRuins + ALU2, GpuTest 32-bit Julia,
SynMark CSDof + DrvState
3-5% GfxBench T-Rex + Egypt, Unigine Heaven + Valley, GpuTest Plot3D
1-7% Media tests
2-3% MemBW GPU blit
1-3% Most of the rest of 3D tests
With the patch:
6-8% MemBW GPU blend => the only regression in these tests (compared
to IOMMU without THP)
4-6% SynMark DrvState (not impacted) + HdrBloom (improved)
3-4% GLB T-Rex
~3% GLB Egypt, SynMark DrvRes
1-3% GfxBench T-Rex + Egypt, SynMark TexFilterTri
1-2% GfxBench CarChase + GLES AztecRuins, Unigine Valley,
GpuTest Triangle
~1% GfxBench Manhattan 3.0/3.1, Unigine Heaven
Perf of several tests actually improved with IOMMU + THP, compared to no
IOMMU / no THP:
10-15% SynMark Batch[0-3]
5-10% MemBW GPU texture, SynMark ShMapVsm
3-4% SynMark Fill* + Geom*
2-3% SynMark TexMem512 + CSCloth
1-2% SynMark TexMem128 + DeferredAA
As a summary across all platforms, these are the benchmarks where enabling
THP on top of IOMMU enabled brings regressions:
* Skylake GT4e:
20-25% SynMark TexMem*
(whereas all MemBW GPU tests either improve or are not affected)
* Broxton J4205:
7% MemBW GPU texture
2-3% SynMark TexMem*
* Tigerlake-H:
7% MemBW GPU blend
Other benchmarks show either lowering of regressions or improvements.
v2:
* Add Kconfig dependency to transparent hugepages and some help text.
* Move to helper for easier handling of kernel build options.
v3:
* Drop Kconfig. (Daniel)
v4:
* Add some benchmark results to commit message.
v5:
* Add explicit regression summary to commit message. (Eero)
References: b901bb89324a ("drm/i915/gemfs: enable THP")
References: 9987da4b5dcf ("drm/i915: Disable THP until we have a GPU read BW W/A")
References: https://gitlab.freedesktop.org/drm/intel/-/issues/430
Co-developed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Eero Tamminen <eero.t.tamminen@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210909114448.508493-1-tvrtko.ursulin@linux.intel.com
|
|
The filesystem reconfigure API is undergoing a transition, breaking our
current code. As we only set the default options, we can simply remove
the call to s_op->remount_fs(). In the future, when HW permits, we can
try re-enabling huge page support, albeit as suggested with new per-file
controls.
Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Reported-by: Sedat Dilek <sedat.dilek@gmail.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190808172226.18306-1-chris@chris-wilson.co.uk
|
|
Continuing the theme of separating out the GEM clutter.
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/20190528092956.14910-8-chris@chris-wilson.co.uk
|
