Age | Commit message (Collapse) | Author |
|
GPUSWERRATA-1564: Freed imported user memory can be accessed via vmap
GPUSWERRATA-1527: Userspace can trigger a data abort by using protected memory for kmap/kbase_vmap(), which can lead to DoS
Change-Id: I326d1a25cbbfb2574c48a6328ba5b0549bc14100
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4939
Bug: 270579983
Signed-off-by: Guus Sliepen <gsliepen@google.com>
|
|
Change-Id: I50c80f9d18055049af485a8a816af812ebcb4de8
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4940
Bug: 270580412
Signed-off-by: Guus Sliepen <gsliepen@google.com>
|
|
Bug: 263689926
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4859
Change-Id: I801d9add01ad36764bf4c0a38d858f7228802611
Signed-off-by: Siddharth Kapoor <ksiddharth@google.com>
|
|
Move the code to disable the MMU for a kctx (which also will flush the
GPU caches for that kctx's Address Space) from the end of kctx
termination to JM/CSF kctx-scheduler termination time.
This ensures the GPU cached writes for that kctx are clear before we
free the kctx's GPU memory pages in the region tracker, and before the
MMU tables are freed.
This in turn ensures the GPU cached writes do not land in freed memory
(which would be a security issue).
Change-Id: I2730a5762cb106ca2179cf9a2f789b8764f21901
(cherry picked from commit 09e14ccf27007dcb2cc0d4d64f5087ccc620a0b5)
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4810
Bug: 260123838
|
|
Our code didn't check if KBASE_REG_NO_USER_FREE was already set
in two code paths (in mali_kbase_csf_tiler_heap.c and
mali_kbase_csf.c). This could be used to maliciously clear
that KBASE_REG_NO_USER_FREE flag.
We add a new refcount for this no user free property, replacing the
KBASE_REG_NO_USER_FREE flag, as a stopgap solution.
In addition to this:
- fix a possible race condition in JIT alloc and tiler
heap init where another thread could take a no user free
reference while the DONT_NEED flag not yet being set
- fix another issue in JIT alloc where reg->flags
was updated without taking the appropriate lock
- move the no user free decref to remove_queue
to clean up context termination code
- refactor memory helpers
Also includes:
- GPUCORE-35469: Fix list del corruption issue in shrinker callback
- GPUCORE-35221 Defer the freeing of VA regions in the chunked tiler heap shrinker callback
- GPUCORE-35499: Fix GROUP_SUSPEND kcpu suspend handling to prevent UAF
- GPUCORE-35268: Fix UAF due to use of MEM_FLAGS_CHANGE ioctl for JIT allocs
(cherry picked from commit 7a1dc910a6a8c9c5aa06677c936c8ad6e9c369ab)
Bug: 260123539
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4801
Change-Id: I7e2349b135e61054f567bdf0577d27eb224d2b12
|
|
Some devices seem to get stuck when trying to turn L2 on in rare case.
As workaround, place a pause of 200us before writing L2_PWRON.
Test: build, boot
Test: stability test
Bug: 262266339
Change-Id: I6acfbb5fbe4497e4adacab01ceb6bfd90e646874
|
|
Change-Id: I86ecbdba8cdd3b18c0da137cc4194e7cb034101d
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4761
Bug: 243522189
Bug: 241217496
Test: Tested GPU reset as b/241217496#comment8
|
|
SBMerger: 478053055
Change-Id: Ia69ab924d682318f9b47efaf2389fd2caf2919d0
Signed-off-by: SecurityBot <android-nexus-securitybot@system.gserviceaccount.com>
|
|
This commit fixes a buffer vulnerability that is provided for internal
test/dump use in kcpu queue to get a CSG's suspend buffer content. The
change ensures that the user supplied GPU VA mapped dump receiving
buffer must meet some additional expected attributes so as to avoid the
risk the dump buffer is abused from the user side, yielding potential
vulnerability of use after free in kbase GPU mapped buffers from the
said kcpu dump arrangement for CSG suspend buffer content.
TI2: 906872 (PLAN-39472r8 R38P1 DDK Precommit)
TI2: 906873 (PLAN-39568r4 R38 GPU Core CS CSF, targeted R38 test, red
an unrelated issue, intermittent)
TI2: 907201 (PLAN-39568r4 R38 GPU Core CS CSF, rerun the targeted red)
TI2: 906018 (PLAN-2596r269 GPU Core CS CSF, 1 unrelated red)
TI2: 906307 (PLAN-2596r269 GPU Core CS CSF, rerun the red)
TI2: 906019 (PLAN-28130r51 GPUCORE CS Android-12-CSF-Minimal, the red
is a trunk issue, multiple nightly reds in history)
(cherry picked from commit 3646ab9f80d1e57b6cdff4e0651f3f07f7065954)
Bug: 254445909
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4746
Change-Id: I25275830b56941f597e6ebb7e38bf8764a5a2555
|
|
'group_id' member of the ioctl (KBASE_IOCTL_CS_TILER_HEAP_INIT) struct
must be validated before initializing CSF tiler heap.
Otherwise out-of-boundary of memory group pools array for the CSF tiler
heap could happen and will potentially lead to kernel panic.
TI2: 933204 (DDK Precommit)
TI2: 933199 (BASE_CSF_TEST)
Bug: 259061568
Test: verified fix using poc
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4766
Change-Id: I209a3d5152a34c278c17383e4aa9080aa9735822
(cherry picked from commit 55b44117111bf6a7e324301cbbf4f89669fa04c3)
|
|
android13-gs-pixel-5.10-tm-qpr2" into android13-gs-pixel-5.10-tm-qpr2
|
|
SBMerger: 478053055
Change-Id: Ic2a88fea0db3fefbbfd167bc6cfad56e050a8b24
Signed-off-by: SecurityBot <android-nexus-securitybot@system.gserviceaccount.com>
|
|
Customer reported an issue where an unexpected GPU page fault happened
due to Tiler trying to access the chunk that was already freed by the
Userspace. The issue was root caused to cacheline sharing between the
HeapContexts of 2 Tiler heaps of the same Kbase context.
The page fault occurred for an Application that made use of more than 1
GPU queue group where one of the group, and its corresponding Tiler heap
instance, is created and destroyed multiple times over the lifetime of
Application.
Kbase sub-allocates memory for a HeapContext from a 4KB page that is
mapped as cached on GPU side, and the memory for HeapContext is zeroed
on allocation through an uncached CPU mapping.
Since the size of HeapContext is 32 bytes, 2 HeapContexts (corresponding
to 2 Tiler heaps of the same context) can end up sharing the same GPU
cacheline (which is 64 bytes in size).
GPU page fault occurred as FW found a non NULL or stale value for the
'free_list_head' pointer in the HeapContext, even though the Heap was
newly created, and so FW assumed a free chunk is available and passed
the address of it to the Tiler and didn't raise an OoM event for Host.
The stale value was found as the zeroing of new HeapContext's memory on
allocation got lost due to the eviction of cacheline from L2 cache.
The cacheline became dirty when FW had updated the contents of older
HeapContext (sharing the cacheline with new HeapContext) on CSG suspend
operation.
This commit makes the GPU VA of HeapContext to be GPU cacheline aligned
to avoid cacheline sharing. The alignment would suffice and there is no
explicit cache flush needed when HeapContext is freed, as whole GPU
cache would anyways be flushed on Odin & Turse GPUs when the initial
chunks are freed just before the HeapContext is freed.
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4724/
Test: Boot to home
Bug: 259523790
Change-Id: Ie9e8bffcadbd2ca7705dcd44f9be76754e28138d
Signed-off-by: Jeremy Kemp <jeremykemp@google.com>
|
|
Through MEM_FLAGS_CHANGE ioctl Userspace can mark a native allocation as
evictable so that backing pages for it can be reclaimed by the shrinker
when system runs low on memory. The ioctl can also be used for the dma
buf imported memory to update the shareability attribute of its GPU
mapping.
The former case can be abused to mark an active JIT allocation as
evictable, due to which the active JIT allocation can be freed via the
shrinker. This would cause use after free on Kbase side when it handles
the JIT free softjob or kcpu command for the already freed allocation.
This commit restricts the use of MEM_FLAGS_CHANGE ioctl for JIT
allocations and for any other allocations that have NO_USER_FREE flag
set for them. There is no use case to support the ioctl in question for
such allocations.
A Base defect test has also been added to verify the changes.
Bug: 259695958
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4739
Change-Id: I9571060946b3d8e1c59e497392a480570d9b39c5
|
|
Through MEM_FLAGS_CHANGE ioctl Userspace can mark a native allocation as
evictable so that backing pages for it can be reclaimed by the shrinker
when system runs low on memory. The ioctl can also be used for the dma
buf imported memory to update the shareability attribute of its GPU
mapping.
The former case can be abused to mark an active JIT allocation as
evictable, due to which the active JIT allocation can be freed via the
shrinker. This would cause use after free on Kbase side when it handles
the JIT free softjob or kcpu command for the already freed allocation.
This commit restricts the use of MEM_FLAGS_CHANGE ioctl for JIT
allocations and for any other allocations that have NO_USER_FREE flag
set for them. There is no use case to support the ioctl in question for
such allocations.
A Base defect test has also been added to verify the changes.
Bug: 259695958
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4739
Change-Id: I9571060946b3d8e1c59e497392a480570d9b39c5
|
|
Bug: 258707257
Test: Run APC ioctl cracker and check APC log
Change-Id: I8a03cd0b23540cda9312c80ded1c5895ddcf0d56
|
|
This patch is a fix for:
- SW Errata: 2716399
- CVE: None
This patch fixes bad pointer access in
shrinker callback.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4612
Signed-off-by: Jack Diver <diverj@google.com>
Change-Id: I49f11ef55a8fb1c7f8af9a56b3391ee10feb5cbd
|
|
This patch is a fix for:
- SW Errata: 2715151
- CVE: CVE-2022-36449
This patch fixes potential use-after-free
after userbuf un-pin
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4611
Change-Id: I89aae381705466ca5971485d5b3b4ef48bc229d3
Signed-off-by: Jack Diver <diverj@google.com>
|
|
This patch is a fix for:
- SW Errata: 2713569
- CVE: CVE-2022-36449
This patch frees PGD pages after invalidation
of L2 cache & MMU TLB
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4610
Signed-off-by: Jack Diver <diverj@google.com>
Change-Id: I4204a351417d944a3c9efe077bca4e180a29270c
|
|
This patch is a fix for:
- SW Errata: 2713363
- CVE: CVE-2022-36449
Fixes overrun by kbase_read.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4609/1
Signed-off-by: Jack Diver <diverj@google.com>
Change-Id: I86d5af2b9d6a85a520987acb7c1e9a49c19f4d42
|
|
This patch is a fix for:
- SW Errata: 2712858
- CVE: CVE-2022-36449
It excludes MMU dumping and invalidates PGD
before free. For this fix to work, GPUCORE-32152
is needed which adds hooks for physical address
translation.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4607/1
Signed-off-by: Jack Diver <diverj@google.com>
Change-Id: I9d3718b57199b7e66a5b49730ac32c810a1fc9c9
|
|
This patch is a fix for:
- SW Errata: 2618270
- CVE: CVE-2022-28348
This fix adds an overflow check to
kbase_mem_alias.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4606/1
Signed-off-by: Jesse Hall <jessehall@google.com>
Change-Id: I33dcda73ec4e5871f5373d38ada29d8c6df7f311
|
|
This patch is a fix for:
- SW Errata: 2619157
- CVE: CVE-2022-28350
It fixes UAF issue with enqueue of
FENCE_SIGNAL kcpu command.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4605/1
Signed-off-by: Jesse Hall <jessehall@google.com>
Change-Id: Ibf2e760dee56190759c5bee047071644798f1aaa
|
|
This patch is a fix for SW errata 2696314 fixing race
with kctx->csf.event.error_list.
Bug: 240184939
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4604/1
Signed-off-by: Jesse Hall <jessehall@google.com>
Change-Id: I8c7ddaec720ede29720ade20e56363ce010503a8
|
|
This patch is a fix for SW errata 2695802 fixing handling of
remap or partial unmap for User IO pages mapping.
Bug: 240185324
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4598
Signed-off-by: Jesse Hall <jessehall@google.com>
Change-Id: I2124955de9027bb65c365930ef69d65af72ec4ac
|
|
This patch is a fix for SW errata 2645822 installing
fence fd at last stage
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4582
Signed-off-by: Jesse Hall <jessehall@google.com>
Change-Id: I6386a8ae56d9b1923b4e4faa0c80483ab41c0bf2
|
|
This patch is fix for SW errata 2637762 fixing possible
out of bounds array access.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4580
Signed-off-by: Jesse Hall <jessehall@google.com>
Change-Id: Ie8a48bbe903fe1127d298a33b9533753d0763452
|
|
Disable HOST_CONTROLS_SC_RAILS compile flags to disable IFPO
Bug: 257052747
Change-Id: Id9d87a365495b160f2f9c46015bfa0a12601c10f
Signed-off-by: Siddharth Kapoor <ksiddharth@google.com>
|
|
Disable GPU FW log, as it incurs performance penalty when left on
Bug: 228779790
Test: benchmark before and after setting this switch
Change-Id: If3b824eb3a7cbde502b3eef38befb9e9856eee50
Signed-off-by: Jack Diver <diverj@google.com>
(cherry picked from commit d4578bc2eb297fdb4c15bb54cd48d45c38ff1885)
|
|
partner/android13-gs-pixel-5.10-tm-qpr2""
This reverts commit 34e635317dc2a91076ac341df3867ac3bdb31ef1.
Bug: 228779790
Change-Id: Ic9d131af5568d7f55f610f255fa1c02925b18482
(cherry picked from commit 1c916e3f7c4d999f68e40c60fee6fe39418fcecd)
|
|
This patch is a fix for:
- SW Errata: 2716399
- CVE: None
This patch fixes bad pointer access in
shrinker callback.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4612
Change-Id: I1eccf3e0852f6e9d9b44d6527fa050ffdbe87caa
Signed-off-by: Jack Diver <diverj@google.com>
(cherry picked from commit 1a13e0c1c4e95f7584d99e7fcf84f79de96e8624)
|
|
This patch is a fix for:
- SW Errata: 2715151
- CVE: CVE-2022-36449
This patch fixes potential use-after-free
after userbuf un-pin
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4611
Change-Id: I3f0887a9465065cc60874c965d91ae2970b6f38a
Signed-off-by: Jack Diver <diverj@google.com>
(cherry picked from commit fb2e7ed2859a7cb88c50dc0d357bb27837aaa843)
|
|
This patch is a fix for:
- SW Errata: 2713569
- CVE: CVE-2022-36449
This patch frees PGD pages after invalidation
of L2 cache & MMU TLB
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4610
Change-Id: Ibb96f289330abb8c23630b06e26af8025b43e4ef
Signed-off-by: Jack Diver <diverj@google.com>
(cherry picked from commit 348355041f171b55a3b93f52821158689ae38b1c)
|
|
This patch is a fix for:
- SW Errata: 2713363
- CVE: CVE-2022-36449
Fixes overrun by kbase_read.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4609/1
Change-Id: I4fbed81f7f36906d228bf84667ebe0eea3e05a7f
Signed-off-by: Jack Diver <diverj@google.com>
(cherry picked from commit c7b5fc80ec4a8d89465cb004d22c24068125ea30)
|
|
This patch is a fix for:
- SW Errata: 2712858
- CVE: CVE-2022-36449
It excludes MMU dumping and invalidates PGD
before free. For this fix to work, GPUCORE-32152
is needed which adds hooks for physical address
translation.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4607/1
Change-Id: Ia0aab34bc483978243aff1bc6e9546385ce8c2eb
Signed-off-by: Jack Diver <diverj@google.com>
(cherry picked from commit 48ab9bf5fc5f938816b416a121a84128fabb4d5b)
|
|
This patch is a fix for:
- SW Errata: 2618270
- CVE: CVE-2022-28348
This fix adds an overflow check to
kbase_mem_alias.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4606/1
Change-Id: I548539e8dcf776e13e1b9be2033b64f4c271b5ef
Signed-off-by: Jesse Hall <jessehall@google.com>
(cherry picked from commit 55f1819bd982ca824ee16d19c6d2aa1bbc0ea706)
|
|
This patch is a fix for:
- SW Errata: 2619157
- CVE: CVE-2022-28350
It fixes UAF issue with enqueue of
FENCE_SIGNAL kcpu command.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4605/1
Change-Id: I094d1559fc27603fa64ef61a267a0205d67de8f7
Signed-off-by: Jesse Hall <jessehall@google.com>
(cherry picked from commit bc304db0cec36bcd040f0c265845c7bf6151245c)
|
|
This patch is a fix for SW errata 2696314 fixing race
with kctx->csf.event.error_list.
Bug: 240184939
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4604/1
Change-Id: Id89773bee8409b046a97dea5c3535d3df04c68b0
Signed-off-by: Jesse Hall <jessehall@google.com>
(cherry picked from commit 2a266f07e9903e9d1d3959e49dbb8e8aea820070)
|
|
This patch is a fix for SW errata 2695802 fixing handling of
remap or partial unmap for User IO pages mapping.
Bug: 240185324
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4598
Change-Id: I9c1365f748d39116afa279dff960b3f1ff01d1ee
Signed-off-by: Jesse Hall <jessehall@google.com>
(cherry picked from commit 828f8f33c56e5e4fc447783976ec56f41c2791ba)
|
|
This patch is a fix for SW errata 2645822 installing
fence fd at last stage
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4582
Change-Id: I15882589090eadfa8ca912b78d5ce7891e821937
Signed-off-by: Jesse Hall <jessehall@google.com>
(cherry picked from commit e29928e40415de6eba26a802d4feb1d9c68d9c5c)
|
|
This patch is fix for SW errata 2637762 fixing possible
out of bounds array access.
Bug: 251397485
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4580
Change-Id: I1597418f91390de7afc0c520d11fdfb739c1e69c
Signed-off-by: Jesse Hall <jessehall@google.com>
(cherry picked from commit 825d564553db3985e94a2e69e6168d596e58ec00)
|
|
This reverts commit 6dcd9736cbf84712dd7073dab4aea256e30517c2.
Reason for revert: UMD taking too long to merge
Bug: 228779790
Change-Id: I08b861ba3cfc8b025f653ef86b0a5ec643e5b13d
|
|
Bug: 228779790
Bug: 240184939
Bug: 240185324
Test: Boot to home
Test: SST run
Test: Camera, portrait, video record
Test: VK CTS dEQP-VK.protected_memory.stack#*
Test: CtsDisplayTestCases
Test: CtsOpenGLTestCases
Test: CtsSkQPTestCases
Test: CtsNativeHardwareTestCases
Test: CtsUiRenderingTestCases
Test: CtsGpuToolsHostTestCases
Test: CtsGpuProfilingDataTestCases
Test: CtsVrTestCases
Test: CtsCameraTestCases
Test: CtsDeqpTestCases
Signed-off-by: Jack Diver <diverj@google.com>
Change-Id: I555b0c61cd08c973aaf7f75dc92d0c7573fa7c91
|
|
In the module initialization for job manager, it should return
error if any of the initialization it calls return error.
The static variable is persistent even if the module is
unloaded. After module unload, sscd_dev contains garbage.
Made a constant copy of sscd_dev and sscd_dev be initialized
every time the module is loaded.
Bug: 218395503
Test: test patch simulating defer situation
Change-Id: I26070c6b2c8e826b2fc78fe337cdf0ca1d2f154b
|
|
This commit fixes the race between the ring of CSG doorbell to send the
acknowledgment for the CS_FAULT event and the ring of CSG doorbell for
sending the CSG start/resume request. Due to the race the update made to
GLB_DB_REQ for the start/resume request was lost and so the request
timed out.
The acknowledgment for the CS_FAULT event is sent in interrupt handler
whereas CSG requests are sent in the process context where Scheduler
mutex is held.
To serialize the access to GLB_DB_REQ between interrupt handler and
process context, this commit uses the interrupt_lock.
Change-Id: I51b2abad52952e0f27a05deb0519508e4c9bd0d2
Bug: 242161241
Bug: 235434294
Test: ~60h local SST.
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/4599
Note: Adapted for CONFIG_MALI_HOST_CONTROLS_SC_RAILS (378860272aef3).
(cherry picked from commit 65820c20e66e2c801445d7c37c67a86707ce0b13)
|
|
There was an issue on Turse platform where sometimes Kbase misses the
CSG idle event notification from FW. This happens when FW sends back to
back notifications for the same CSG for events like SYNC_UPDATE and IDLE
but Kbase gets a single IRQ and observes only the first event.
The issue was root caused to a barrier missing on Kbase side between the
write to JOB_IRQ_CLEAR register and the read from interface memory, i.e.
CSG_ACK. Without the barrier there is no guarantee about the ordering,
the write to JOB_IRQ_CLEAR can take effect after the read from interface
memory.
The ordering is needed considering the way FW & Kbase writes to the
JOB_IRQ_RAWSTAT & JOB_IRQ_CLEAR registers without any synchronization.
This commit adds dmb(osh) barrier after the write to JOB_IRQ_CLEAR to
resolve the issue.
TI2: 896668 (PLAN-12467r490 TGT CS Nightly, few CSF scenarios)
Bug: 243913790
Test: SST ~4600 hours
Provenance: https://code.ipdelivery.arm.com/c/GPU/mali-ddk/+/3841
Change-Id: I094a3b55c8ae28e8126057cdaf81990f62cd388e
(cherry picked from commit 220d89fd264b11a5b68290c3ca5a8c232e1d45db)
|
|
Without this patch, the platform work_begin callback will be called
multiple times for the same atom if a state update request is made
and the atom still resolves to submitted.
Bug: 247721380
Test: Boot g101 to home
Test: No per-UID-tis warnings
Signed-off-by: Jack Diver <diverj@google.com>
Change-Id: I535d1a1c2f983f0162dc0e18fe40544e61246d4c
|
|
Up to two Atoms can be submitted to the same Job slot.
To prevent warnings about multiple work begin/end for the same slot
we need to track the two atoms individually
Bug: 247721380
Test: Boot g101 to home
Test: No per-UID-tis warnings
Change-Id: I96dcc635f351c9ebd7161c1c619fc50f32bf190a
|
|
Refactor driver builds to support common codes for gs101 and gs201.
- Check SoC specific CONFIG flag in Kbuild
- Convert CONFIG_MALI_PIXEL_GPU_SSCD and
CONFIG_MALI_HOST_CONTROLS_SC_RAILS to ccflags
Bug: 245742922
Test: Not fully tested. DO NOT MERGE
Change-Id: I1cf7c19e2a2e63de17153c70ede806dc773398d3
Signed-off-by: Siddharth Kapoor <ksiddharth@google.com>
Signed-off-by: Lucas Wei <lucaswei@google.com>
|
|
worklog:
- commit 1b55a3b93f52 ("mali_kbase: use RT_MUTEX for some job submission
and context locks") replaced some locks with RT_MUTEX. Replace
corresponding locks to avoid build breakage.
Test: Build kernel
Test: Boot to home
Bug: 238399460
Signed-off-by: Lucas Wei <lucaswei@google.com>
Change-Id: Id0d3f05bd0b5ebacc29e126ba98db5bb789e91fe
|