Integrate linux-seco-imx/seco_lf-5.10.y-gpu-viv
MGS-7215 [#imx-3183] 0034-CL694252-KERNEL-SPACE-6.4.11_22Q2_NXP-IMX-3189-fix-t
6.4.11_22Q2_NXP: IMX-3189: fix the build error caused by conflicting types of "_QuerySignal" with gcc13.1.
Signed-off-by: IPD_SCM IPD_SCM@verisilicon.com
Signed-off-by: Yuan Tian yuan.tian@nxp.com
Upstream-Status: Inappropriate [other]
Upstream does not have the propriatery Vivante driver. Cherry picked from NXP downstream commit 9b70e0e11b04c18545093bd6d172762f65709365.
Signed-off-by: Max Krummenacher max.krummenacher@toradex.com
--
MA-21468-2 [#imx-3151] Fix GPU deadlock issue when do video stream stress test
CL696036 [KERNEL SPACE] 6.4.x_234062: Merge CL696034 from 64x for IMX-3151 - GPU deadlock issue when do video stream stress test. Disable TX clock gating for 8QM board to workaround HW bug#2385. By: xuan.huang.
This picks VSI 0002-CL696036-KERNEL-SPACE-6.4.x_234062-Merge-CL696034-fr.patch
Signed-off-by: Hui Wang hui.wang_7@nxp.com
--
MA-21468-1 [#imx-3151] Fix GPU deadlock issue when do video stream stress test
CL693528 [KERNEL SPACE] 6.4.x_234062: Merge CL693076 from 64x for IMX-3151 - GPU deadlock issue when do video stream stress test. Disable TX clock gating for QM_c0/m815/m865/m850D/m845s/8ULP/8.5 board to workaround HW bug#2385. By: xuan.huang.
This picks VSI 0001-CL693528-KERNEL-SPACE-6.4.x_234062-Merge-CL693076-fr.patch
Signed-off-by: Hui Wang hui.wang_7@nxp.com
--
MA-21513 [#imx-3206] Fix AXI BUS ERROR on more than 4G DDR board.
Meet GPU dump when run the below case on more than 4G DDR board. CtsNNAPITestCases TestRandomGraph/RandomGraphTest#LargeGraph_TENSOR_FLOAT16_Rank*
The above 4G address can be rejected when doing _UserMemoryAttach. But will be used by _UserMemoryPhysical as os->paddingPage is allocated with __GFP_HIGHMEM. Here change __GFP_HIGHMEM to __GFP_DMA32 to avoid the GPU dump.
Signed-off-by: Jessie Hao juan.hao@nxp.com
--
MGS-6664-4 [#imx-2955] Fix OpenCL regression issue
Since memory for sg_table doesn't have 4G limit, need to remove __GFP_DMA32 from normal_gfp when calling sg_alloc_table_from_pages, which causes error info "Unexpected gfp: 0x4 (GFP_DMA32). Fixing up to gfp: 0x12220 (__GFP_HIGH|__GFP_ATOMIC|__GFP_NOWARN|__GFP_NORETRY). Fix your code!" when doing test_integer_ops.
Signed-off-by: Guangliu Ding guangliu.ding@nxp.com
--
MGS-6409 [#imx-2928] Fix GPU hang when insmod galcore on 8qm_qnx
GPU hang on 8qm_qnx is caused by querying gpuTimeout not supported in gckOS_QueryOption under QNX environment, only supported under Linux environment. With this fix, gpuTimeout is rolled back to gcdGPU_TIMEOUT under QNX and other system.
Signed-off-by: Guangliu Ding guangliu.ding@nxp.com
--
MGS-6622 [#imx-2974] Fix build error using lower version toolchain
Fix following build error using 5.10, 5.4 or lower version toolchain. 5.15 toolchain is OK.
drivers/mxc/gpu-viv/hal/os/linux/kernel/allocator/default/gc_hal_kernel_allocator_gfp.c: In function ‘_GFPAlloc’: drivers/mxc/gpu-viv/hal/os/linux/kernel/allocator/default/gc_hal_kernel_allocator_gfp.c:568:9: error: a label can only be part of a statement and a declaration is not a statement 568 | void *addr = NULL;
Signed-off-by: Guangliu Ding guangliu.ding@nxp.com
--
MA-20587 Fix kernel panic when run camera monkey on 8ulp
Meet the below call stack when run camera monkey on 8ulp Call trace: __slab_free+0x238/0x294 kfree+0x21c/0x354 gckOS_TraceGpuMemory+0x1c4/0x418 gckVIDMEM_NODE_WrapUserMemory+0x194/0x2b4 _WrapUserMemory+0x5c/0x13c
This reverts commit aaa2d163e232009aaaa240c1242bd657bdac0b8d as a temp solution. Revert "MGS-6739 [#imx-2760] 0036-CL500952-KERNEL-SPACE-234062-merge-cl493474-494780-4"
Conflicts: drivers/mxc/gpu-viv/hal/kernel/inc/gc_hal_options.h
Change-Id: I4a451dd0bd0648fc956703ffe25920f11b073f46
--
MGS-6622 [#imx-2974] Change the order of memory zone to allocate gfp
GPU MMU can mapping 4G memory in total, currently 3G for flat mapping and the rest for dynamic mapping. The default order of memory zone to allocate gfp is from 4G above to within 4G expect for special cases with 4G limit flags, which may cause dynamic mapping area used out and GPU can't allocate gfp normally even though there is still available memory within 4G. So we set the default memory zone to allocate gfp is memory within 4G, then memory above 4G.
Signed-off-by: Guangliu Ding guangliu.ding@nxp.com
--
MGS-6813 [#imx-2968] Improve GPU subsystem driver for upstream compatibility
Upstream kernel support i.MX8MQ GPU with the unified device entry, Need improve GPU subsystem driver to support upstream gpu device.
This can avoid maintenance for the duplicated GPU device entry.
Signed-off-by: Xianzhong xianzhong.li@nxp.com
--
MGS-6409 [#imx-2928] Add configuration to enable runtime GPU timeout
Add gpuTimeout as module_param to replace gcdGPU_TIMEOUT. On one hand, gpuTimeout uses gcdGPU_TIMEOUT as initial value, which can be compatible with previous programs. On the other hand, gpuTimeout value can be changed conveniently through module params without rebuilding.
Note: gcdGPU_TIMEOUT still determines whether to enable GPU timeout.
Signed-off-by: Guangliu Ding guangliu.ding@nxp.com
--
MGS-6664-2 [#imx-2955] GPU AXI BUS error
___GFP_DMA32 mask misses for GPU non-contiguous GFP_Allocator when 4G memory limit flag is set.
Signed-off-by: Guangliu Ding guangliu.ding@nxp.com
--
MGS-6664-1 [#imx-2955] GPU AXI BUS error
Only 8MQ, 8MM, 8MN boards can't access 4G above memory and need to set gcvPLATFORM_FLAG_LIMIT_4G_ADDRESS flagbits. Totalram_pages check doesn't include kernel memory, reserved memory, causing this flagbits is not set and AXI BUS error on customer's board. So we remove this redundant check.
Signed-off-by: Guangliu Ding guangliu.ding@nxp.com
--
MGS-6681 [#imx-2760] 0066-CL483742-KERNEL-SPACE-6.4.x_234062-Merge-NXP-s-patch
6.4.x_234062: Merge NXP's patch to fix coverity test issue: Out-of-bounds-access in __free_pages. Not a bug, just in order to pass coverity test.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com (cherry picked from commit 9819d63480e389ae18bedbf7a237cffb445da181)
--
LF-6930 [#imx-2969] Skip checking MC idle bits in isGPUIdle function on 8ulp
Software workaround for MC idle bits not set for a long time on 8ULP. The reason is MC idle status depend on the undone memory request counter, which cannot decrease to zero due to HW bug of 8ULP GPU. Actually all modules finished the job and become idle properly. MC not-idle is a false alert if there is no HW activity in the whole pipeline. So that it is safe to skip the check for MC idle bit in SW workaround.
Signed-off-by: Xiaojun.Chen xiaojun.chen@versilicon.com Reviewed-by: Guangliu Ding guangliu.ding@nxp.com
--
MGS-6835 [#ccc] fix build error of incompatible-pointer-types
Fix build error when use imx_v6_v7 defconfig or enable CONFIG_PROVE_LOCKING.
drivers/mxc/gpu-viv/hal/kernel/gc_hal_kernel.c: In function ‘gckKERNEL_Construct’:
./drivers/mxc/gpu-viv/hal/os/linux/kernel/gc_hal_kernel_mutex.h:74:73: error: passing argument 3 of ‘gckOS_Allocate’ from incompatible pointer type [-Werror=incompatible-pointer-types]
74 | gceSTATUS _status = gckOS_Allocate(Os, gcmSIZEOF(struct key_mutex), &key_mut);
| ^~~~~~~~
| |
| struct key_mutex **
./drivers/mxc/gpu-viv/hal/kernel/inc/gc_hal_base.h:4427:18: note: in definition of macro ‘_gcmkONERROR’
4427 | status = func;
| ^~~~
drivers/mxc/gpu-viv/hal/kernel/gc_hal_kernel.c:584:9: note: in expansion of macro ‘gcmkONERROR’
584 | gcmkONERROR(gckOS_CreateMutex(Os, &kernel->db->dbMutex));
| ^~~~~~~~~~~
drivers/mxc/gpu-viv/hal/kernel/gc_hal_kernel.c:584:21: note: in expansion of macro ‘gckOS_CreateMutex’
584 | gcmkONERROR(gckOS_CreateMutex(Os, &kernel->db->dbMutex));
| ^~~~~~~~~~~~~~~~~
In file included from drivers/mxc/gpu-viv/hal/kernel/gc_hal_kernel_precomp.h:59,
from drivers/mxc/gpu-viv/hal/kernel/gc_hal_kernel.c:56:
./drivers/mxc/gpu-viv/hal/kernel/inc/gc_hal.h:293:22: note: expected ‘void **’ but argument is of type ‘struct key_mutex **’
293 | OUT gctPOINTER * Memory
| ~~~~~~~~~~~~~^~~~~~
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0046-KERNEL-SPACE-Update-gc_feature_database.h-to-544380
Update gc_feature_database.h to 544380
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0040-CL544904-KERNEL-SPACE-Chen-Yuan-cn8238-Back-out-CL54
(Chen Yuan cn8238)Back out #CL544518
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0038-CL544518-KERNEL-SPACE-Chen-Yuan-cn8238-add-gcregUnpa
(Chen Yuan cn8238)add gcregUnpackedSemantic register to provide interpolation information for unpacked components(v800)
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0031-CL534962-KERNEL-SPACE-Back-out-changelist-534588
Back out changelist 534588
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0029-CL534588-KERNEL-SPACE-Chen-Yuan-add-gcregUnpackedSem
(Chen Yuan)add gcregUnpackedSemantic register to provide interpolation location and mode information for unpacked components(currently used for clipdistance issue)
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0025-CL520026-KERNEL-SPACE-234062-hal-Merging-518316-add-
234062, hal, Merging 518316, add microseconds delay function, 518037, fix regression of 490023 //SW/Rel5x/dev6.4.x/projects.dev_6.4.11_22Q2/... to //SW/Rel5x/dev6.4.x/projects.dev_234062/...
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0022-IMX-2930-CL514087-KERNEL-SPACE-dev64x_234062-merge-N
dev64x_234062: merge NXP's patch [IMX2930]: Add support for CMA_LIMIT flag.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0014-CL511421-KERNEL-SPACE-dev64x_234062-Fix-the-build-is
dev64x_234062: Fix the build issue cause by CL511300.[Wang.Zijian]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6835 [#imx-2760] 0012-IMX-2878-CL511300-KERNEL-SPACE-dev64x_234062-Merge-C
dev64x_234062: Merge CL510259 from 64x. To solve the IMX-2878, allocate lock key for mutex when the CONFIG_PROVE_LOCKING is enabled.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0055-KERNEL-SPACE-Update-gc_feature_database.h-to-501179
Update gc_feature_database.h to 501179
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0051-CL504140-KERNEL-SPACE-234062-change-android-platform
234062:change android platform macro to gcdANDROID instead of ANDROID.[liu.shuai]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0049-IMX-2916-CL502884-KERNEL-SPACE-234062-align-the-flat
234062: align the flat-mapping address/size to gpu page size(IMX-2916)
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0044-CL502024-KERNEL-SPACE-set-gcdENABLE_VIDEO_MEMORY_TRA
set gcdENABLE_VIDEO_MEMORY_TRACE to 1 by default for android, fix.[liu.shuai]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0042-CL502022-KERNEL-SPACE-set-gcdENABLE_VIDEO_MEMORY_TRA
set gcdENABLE_VIDEO_MEMORY_TRACE to 1 by default for android.[liu.shuai]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0036-CL500952-KERNEL-SPACE-234062-merge-cl493474-494780-4
234062:merge cl493474/494780/495277/500513 from 64x, for video memory trace.[liu.shuai]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0031-CL499053-KERNEL-SPACE-Backout-CL498869
Backout CL498869
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0030-CL498981-KERNEL-SPACE-dev64x-fix-the-dump-command-bu
dev64x: fix the dump command buffer address dismatch
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0027-CL498869-KERNEL-SPACE-reg-Add-AQSecurtyNASID-registe
[reg]: Add AQSecurtyNASID registers for gc820 amazon.[Yandong]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0025-CL498655-KERNEL-SPACE-Rollback-HW-projects.dev-arch-
Rollback //HW/projects.dev/arch/XAQ2/reg/AQDma.r to revision 44 CL498504 caused all cases fail, revert it by request from SQA.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0023-CL498504-KERNEL-SPACE-reg-Add-AQSecurtyNASID-registe
[reg]: Add AQSecurtyNASID registers for gc820h.[Yandong]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0017-CL495140-KERNEL-SPACE-Rollback-HW-projects.dev-arch-
Rollback //HW/projects.dev/arch/XAQ2/reg/AQDrawingEngine.r to revision 52 As request from SQA.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6739 [#imx-2760] 0015-CL494632-KERNEL-SPACE-reg-Merge-registers-from-gc620
[reg]: Merge registers from gc620_v551.[Wang baojun]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6744 [#imx-2793] Revert GPU hang workaround for 8ULP A0.1 AXI issue
Verified ECO fix with GPU stress test in 8ULP A1, GPU hang issue solved
Revert "LF-5214 [#imx-2759] disable GPU command gate to avoid 8ULP es31 cts hang"
This reverts commit 7eedc4ed194129b6388cef94c4c9cef8605e026f.
Signed-off-by: Xianzhong xianzhong.li@nxp.com
--
MGS-6681 [#imx-2760] 0122-KERNEL-SPACE-Update-gc_feature_database.h-to-488823
Update gc_feature_database.h to 488823
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0114-CL489171-KERNEL-SPACE-6.4.x_234062-Add-error-check-f
6.4.x_234062: Add error check for the return value of gckVIDMEM_HANDLE_Dereference.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0113-BUG33142-CL489019-KERNEL-SPACE-dev64x_234062-fix-the
dev64x_234062: fix the bug_33142.[Wang.Zijian]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0106-IMX-2865-CL487094-KERNEL-SPACE-64x_234062-fix-cl4863
64x_234062: fix cl486335 for imx-2865
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0098-IMX-2865-CL486335-KERNEL-SPACE-64x_234062-imx-2865-m
64x_234062: imx-2865 merge
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0070-CL485673-KERNEL-SPACE-234062-fix-clock-gatting-bug-f
234062: fix clock gatting bug for 865.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0068-IMX-2865-CL484996-KERNEL-SPACE-64x_234062-merge-imx-
64x_234062: merge imx-2865 to 234062 for adding tracepointer gpu_mem_total.[liu.shuai]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0049-CL478430-KERNEL-SPACE-reg-Roll-back-to-revision-50.-
[reg]: Roll back to revision #50.[Yandong]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0047-CL478273-KERNEL-SPACE-reg-Delete-some-unused-registe
[reg]: Delete some unused registers.[Yandong]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0045-CL478258-KERNEL-SPACE-reg-Roll-back-to-revision-40.-
[reg]: Roll back to revision #40.[Yandong]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0044-CL478249-KERNEL-SPACE-reg-Roll-back-to-revision-84.-
[reg]: Roll back to revision #84.[Yandong]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0034-CL477604-KERNEL-SPACE-reg-Merge-registers-from-gc620
[reg]: Merge registers from gc620_v551.[Yandong]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0033-CL477198-KERNEL-SPACE-234062-merge-the-nxp-patch-to-
234062: merge the nxp patch to support devfer probe.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6681 [#imx-2760] 0007-CL470838-KERNEL-SPACE-64x_dev234062-Fix-insmod-error
64x_dev234062: Fix insmod error in IMX6Q board.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
LF-6205 [#imx-2897] openCV test ends in kernel panic
CL485410 [KERNEL SPACE] dev64x_234062: Merge CL484068 from dev64x. Check memory is allocated from reserved memory or not when wrap user memory. If the memory is allocated from the reserved memory, won't do cache operation for it. Add the gckOS as input parameter for import_page_map()/import_pfn_map() and import_physical_map() to transmit the info of reserved memory range by os->device. [Wang.Zijian]
Signed-off-by: Xiaojun.Chen xiaojun.chen@versilicon.com Signed-off-by: Eric Guo eric.guo@nxp.com
--
MGS-6628 [#imx-2760] 0050-KERNEL-SPACE-Update-gc_feature_database.h-to-469146
Update gc_feature_database.h to 469146
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6628 [#imx-2760] 0048-CL469248-KERNEL-SPACE-dev64x_234062-Fix-build-warnin
dev64x_234062: Fix build warning in android.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6628 [#imx-2760] 0038-BUG32447-CL468159-KERNEL-SPACE-234062-fix-memory-lea
234062:fix memory leak for bugzilla 32447[liu.shuai]
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6628 [#imx-2760] 0013-CL458157-KERNEL-SPACE-dev6.4.x_234062-Fix-build-erro
dev6.4.x_234062: Fix build error when enable capture only mode.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6571-2 [#imx-2806] add gbm_bo_get_fd_for_plane api for kernel space
[KERNEL SPACE] add gbm_bo_get_fd_for_plane api for weston 10.0
Signed-off-by: Ke Feng ke.feng@verisilicon.com Signed-off-by: Wujian Sun wujian.sun_1@nxp.com
--
MGS-6593 [#imx-2760] 0089-KERNEL-SPACE-Update-gc_feature_database.h-to-454235
Update gc_feature_database.h to 454235
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6593 [#imx-2760] 0040-CL442432-KERNEL-SPACE-v620_v2-Cmodel.-Correct-AHB-an
[v620_v2]Cmodel. Correct AHB and Load state register related Compressor/Decompressor, submitted by baojian
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6593 [#imx-2760] 0038-CL442208-KERNEL-SPACE-Reg-Regsiter.-Add-AHB-and-Load
Regsiter. Add AHB and Load state register related Compressor/Decompressor, submitted by baojian
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6593 [#imx-2760] 0032-CL441019-KERNEL-SPACE-dev64x_234062-merge-CL439888-a
dev64x_234062: merge CL439888 and CL440024 from dev64x, solve the kernel driver build issue in Linux5.16.0.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6593 [#imx-2760] 0021-CL439798-KERNEL-SPACE-234062-revert-438254
234062: revert 438254.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6593 [#imx-2760] 0003-IMX-2793-IMX-2759-CL438254-KERNEL-SPACE-234062-fix-b
234062: fix bug IMX-2793, IMX-2759. fix the HW MC not idle bug. fix the 8ulp clock unstable bug.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MA-19319 [#imx-2825] Add another thermal PM processing interface for android
There are several thermal sensor on some platforms. It causes GPU thermal_hot_pm_notifier callback function be called several times. It cannot bind to specific sensor and response to specific temperature point. Add proprietary thermal cooling device driver for GPU to process the thermal event for GPU.
Signed-off-by: Dandan Sun dandan.sun@nxp.com
--
MA-12288 [#imx-2729] support multiple thermal events processing for imx8mq
GPU clock will switch to minClock/64, 32/64, 64/64 of max GPU frequency according to thermal event. The minimum clock of GPU is variable and can be set through sysfs file /sys/bus/platform/drivers/galcore/gpu3DMinClock. This value should read every time when thermal event generated.
Signed-off-by: Zhang Bo bo.zhang@nxp.com
--
MGS-6469 [#imx-2756] Enable gpudump event when recovery is enabled
Add possibility to do gpu dump upon stall even when recovery enabled. Patch provided by VSI.
Date: Feb 08, 2022 Signed-off-by: Petr Cach petr.cach@nxp.com Reviewed by: Xianzhong Li xianzhong.li@nxp.com Reviewed by: Jozef Bastek jozef.bastek@nxp.com
--
MGS-6520 [#imx-2760] 0063-KERNEL-SPACE-Update-gc_feature_database.h-to-437745
Update gc_feature_database.h to 437745
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6520 [#imx-2760] 0057-BUG31845-CL437581-KERNEL-SPACE-6.4.x_234062-Merge-CL
6.4.x_234062: Merge CL435081 from 6.4.x. Fixed Bug_31845. Ensure the order of the chip features table. By: xuan.huang.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6520 [#imx-2760] 0017-BUG31480-CL431111-KERNEL-SPACE-6.4.x_234062-Merge-CL
6.4.x_234062: Merge CL430717 from 6.4.x. Fix Bug 31480. HW bug 2030, accoriding to the feature bit gcvFEATURE_BIT_PE_64BPP_LINEAR_FORMAT to deteminate whether to support 64BPP to draw clear. By: xuan.huang.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
LF-5214 [#imx-2759] disable GPU command switch to avoid 8ULP hang
GPU power management involve clock/power gating and command start/stop, the problem is caused by repeated GPU command start/stop quickly(<1ms).
this patch disabled separate GPU clock and command switch from idle state, GPU clock and command still can gate off along with power off together.
Signed-off-by: Xianzhong xianzhong.li@nxp.com Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
MGS-6497 [#imx-2759] disable MC not idle temporarily
CL433206 [KERNEL SPACE] After the internal command is executed, check the idle status of other modules to determine whether the command is executed.
Signed-off-by: Ke Feng ke.feng@verisilicon.com Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
Revert "MGS-6490 [#ccc] 0022-IMX-2729-CL402351-KERNEL-SPACE-IMX-2729-modify-GPU-t"
This reverts commit 2e3aedea774b0e54e4ea64cecfbb16fe8e8bf5da.
--
MGS-6490 [#ccc] 0022-IMX-2729-CL402351-KERNEL-SPACE-IMX-2729-modify-GPU-t
IMX-2729 modify GPU thermal_hot_pm_notify to support different thermal event Some use case of iMX8MQ only use limited heat sink, but the power consumption is too large when run some android application. Need to decrease the power consumption and keep system accessible. Here add half of Max GPU frequence work point when do thermal PM processing. GPU clock will switch to minClock/64, 32/64, 64/64 of max GPU frequency according to thermal event.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6490 [#ccc] 0068-CL424483-KERNEL-SPACE-6.4.x_234062-merge-CL423392-no
6.4.x_234062 merge CL423392, no alignment require for linear rendering target buffer.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6498 [#ccc] 0019-IMX-2693-CL426299-KERNEL-SPACE-64x_234062-imx-2693-I
64x_234062, imx-2693: Implemented YUV planar formats to YUY2 packed format conversion using compute shader to avoid UI freeze when playing youtube with ads on MP Android platform.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6490 [#ccc] 0052-IMX-2733-CL416520-KERNEL-SPACE-234062-deal-with-IMX-
234062: deal with IMX-2733, unified empty marco to 'nothing'
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6498 [#ccc] 0024-IMX-2667-IMX-2697-BUG29453-CL427346-KERNEL-SPACE-6.4
Antutu V8/V9 benchmark not work well on 8MN/8MP;
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6490 [#ccc] 0042-IMX-2740-IMX-2741-CL405339-KERNEL-SPACE-Integrate-NX
Integrate NXP QNX patches: IMX-2740 patch: qnx_improve_gpu_dump_logging.diff. IMX-2741 patch: qnx_fix_debugfs_crash.diff.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGG-325 [#imx-2773] _ReleaseVideoMemory: check retvals for gckVIDMEM_HANDLE_Dereference and gckVIDMEM_NODE_Dereference
gckVIDMEM_HANDLE_Dereference() and gckVIDMEM_NODE_Dereference() (both called by _ReleaseVideoMemory) can fail but their return values are not checked.
When gckVIDMEM_NODE_Dereference fails, gckVIDMEM_Free() might not get called so _ReleaseVideoMemory will return OK, but the video memory will still be allocated.
Add a gcmkONERROR around them so their return values are checked.
--
MGS-6498 [#ccc] 0032-KERNEL-SPACE-Update-gc_feature_database.h-to-427359
Update gc_feature_database.h to 427359
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6490 [#ccc] 0071-KERNEL-SPACE-Update-gc_feature_database.h-to-421636
Update gc_feature_database.h to 421636
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
LF-4817 [#imx-2750] fix 6124054 Out-of-bounds access
fix 6124054 Out-of-bounds access
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
MGS-6500 gpu: make gpu depend on drm
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4430 [#imx-2733] coverity fix in KERNEL_SPACE
fix Missing break in switch (MISSING_BREAK) when define gcmkASSERT() null
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4817 no stop timer
6.4.3.p4 release
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4817 rm duplicated line
6.4.3.p4 release
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4817 fix database leak
6.4.3.p4 release
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4817 refine gpu governor
6.4.3.p4 release
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
CL371289 [KERNEL SPACE] [234062]: verifyGraphOpt merge from 64x to 234062
Signed-off-by: Ke Feng ke.feng@verisilicon.com
--
Update version for 6.4.3.p4 release
Signed-off-by: Ke Feng ke.feng@verisilicon.com
--
Update gc_feature_database.h to 396789
Signed-off-by: Ke Feng ke.feng@verisilicon.com
--
CL395508 [KERNEL SPACE] 234062: Add Host Interface 0 probe counter reset.(Merging //SW/Rel5x/dev6.4.x/projects.dev/hal/kernel/arch/gc_hal_kernel_hardware_func.c to //SW/Rel5x/dev6.4.x/projects.dev_234062/hal/kernel/arch/gc_hal_kernel_hardware_func.c
Signed-off-by: Ke Feng ke.feng@verisilicon.com
--
CL395503 [KERNEL SPACE] 234062: enable HI store counter (Merging //SW/Rel5x/dev6.4.x/projects.dev/hal/kernel/arch/gc_hal_kernel_hardware_func.c to //SW/Rel5x/dev6.4.x/projects.dev_234062/hal/kernel/arch/gc_hal_kernel_hardware_func.c
Signed-off-by: Ke Feng ke.feng@verisilicon.com
--
CL395498 [KERNEL SPACE] 234062:enable tx/pa store counter (Merging //SW/Rel5x/dev6.4.x/projects.dev/hal/kernel/arch/gc_hal_kernel_hardware_func.c to //SW/Rel5x/dev6.4.x/projects.dev_234062/hal/kernel/arch/gc_hal_kernel_hardware_func.c
Signed-off-by: Ke Feng ke.feng@verisilicon.com
--
MGS-6435 [#ccc] 0006-IMX-2651-BUG30890-BUG31102-CL396388-KERNEL-SPACE-6.4
IMX-2651 BUG30890 BUG31102 CL396388 [KERNEL SPACE] 6.4.x & 6.4.x_234062: bug#30890: [NXP][android-webgl]evaluate android webgl with chrome V91. Refine and merge cl389548 again. Adjust fake hal format related logic for Android linear native buffer, to fix "rendering/multisample-corruption.html" fails,and fixed bug#31102.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6425 [#imx-2711] remove duplicated FindMdlMap
this call FindMdlMap without protect, also it conflicts with gckOS_UnmapMemoryEx which already call FindMdlMap with a mutex named mapsMutex.
[50194.783141] Unable to handle kernel paging request at virtual address dead0000000000e0 [50194.791132] Mem abort info: [50194.793932] ESR = 0x96000004 [50194.797022] EC = 0x25: DABT (current EL), IL = 32 bits [50194.808264] EA = 0, S1PTW = 0 [50194.811759] Data abort info: [50194.814689] ISV = 0, ISS = 0x00000004 [50194.818569] CM = 0, WnR = 0 [50194.821540] [dead0000000000e0] address between user and kernel address ranges [50194.828945] Internal error: Oops: 96000004 [#1 (closed)] PREEMPT SMP [50194.866817] CPU: 1 PID: 948515 Comm: [50194.878461] Hardware name: NXP i.MX8MPlus EVK board (DT) [50194.883774] pstate: 20000005 (nzCv daif -PAN -UAO -TCO BTYPE=--) [50194.890043] pc : FindMdlMap+0x30/0x50 [50194.893715] lr : gckKERNEL_DestroyProcessReservedUserMap+0x11c/0x1a0 [50194.903022] sp : ffff800012463a90 [50194.921456] x27: ffff0000c111a000 x26: 0000000000000000 [50194.926769] x25: ffff0000c7caae00 x24: 0000000000000005 [50194.932082] x23: ffff0000c0cbf600 x22: 00000000000e78f5 [50194.937395] x21: ffff0000c111a000 x20: 0000000010000000 [50194.942707] x19: ffff0000c0ca5800 x18: 00000000000000c0 [50194.948019] x17: 0000000000000000 x16: 0000000000000019 [50194.953331] x15: 0000000000000004 x14: ffff0000c0496048 [50194.958644] x13: 0000000000000000 x12: ffff0000d0664d08 [50194.963956] x11: ffff0000d0664b68 x10: ffff0000c3112482 [50194.969268] x9 : ffff0000c0496050 x8 : ffff0000cf6bb900 [50194.974581] x7 : 0000000000000001 x6 : ffff0000c7caa978 [50194.979893] x5 : 0000000000000005 x4 : 0000000000000000 [50194.985205] x3 : ffff0000c0cbf660 x2 : dead000000000100 [50194.990518] x1 : 00000000000e78f5 x0 : dead0000000000e0 [50194.995830] Call trace: [50194.998282] FindMdlMap+0x30/0x50 [50195.001601] gckKERNEL_DestroyProcessDB+0x328/0x420 [50195.006481] gckKERNEL_AttachProcessEx+0x40/0x174 [50195.011186] drv_release+0x60/0xb4 [50195.014589] __fput+0x78/0x230 [50195.017643] ____fput+0x10/0x20 [50195.020785] task_work_run+0x80/0x140 [50195.024448] do_exit+0x324/0x990 [50195.027677] do_group_exit+0x38/0xa0 [50195.031254] get_signal+0xe0/0x6b0 [50195.034657] do_notify_resume+0x180/0x9b0 [50195.038667] work_pending+0xc/0x440 [50195.042158] Code: f9401002 d1008040 eb02007f 540000c0 (b9400002)
Reviewed-by: Xianzhong xianzhong.li@nxp.com Reviewed-by: Nicusor Citu nicusor.citu@nxp.com Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
MGS-6425 [#imx-2711] use default list_head in linux
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4656 [#imx-2711] fix out-of-bounds access
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4656 [#imx-2711] fix coverity issue
Invalid type in argument to printf format specifier
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4436-4 [#imx-2664] avoid lowmem killer when allocate sgl table
sg_alloc_table_from_pages may trigger lowmem killer with GFP_KERNEL flag, disable GFP_KERNEL to avoid lowmem killer when allocate sgl table for GPU.
[ 1704.740010] Chrome_InProcGp invoked oom-killer: gfp_mask=0xcc0(GFP_KERNEL), order=0, oom_score_adj=0 [ 1704.749265] CPU: 3 PID: 1251 Comm: Chrome_InProcGp Tainted: G C 5.10.52-lts-5.10.y+g5788c4507376 #1 (closed) [ 1704.759618] Hardware name: NXP i.MX8MNano EVK board (DT) [ 1704.764929] Call trace: [ 1704.767385] dump_backtrace+0x0/0x1a0 [ 1704.771051] show_stack+0x18/0x70 [ 1704.774368] dump_stack+0xd0/0x12c [ 1704.777770] dump_header+0x48/0x1f4 [ 1704.781259] oom_kill_process+0x200/0x204 [ 1704.785268] out_of_memory+0x178/0x560 [ 1704.789018] __alloc_pages_nodemask+0xb34/0xc24 [ 1704.793549] alloc_pages_current+0x80/0x13c [ 1704.797752] __get_free_pages+0x14/0x50 [ 1704.801595] __sg_alloc_table_from_pages+0x22c/0x484 [ 1704.806559] sg_alloc_table_from_pages+0x24/0x3c [ 1704.811176] _GFPAlloc+0x2b0/0x670 [ 1704.814579] gckOS_AllocatePagedMemory+0xf4/0x1fc [ 1704.819302] gckVIDMEM_NODE_AllocateVirtual+0xc4/0x190 [ 1704.824446] gckKERNEL_AllocateVideoMemory+0x2f8/0x380 [ 1704.829584] _AllocateLinearMemory+0xb8/0x1b4 [ 1704.833943] gckKERNEL_Dispatch+0x20c/0x13d4 [ 1704.838229] gckDEVICE_Dispatch+0x80/0x1bc [ 1704.842328] drv_ioctl+0x1b4/0x270 [ 1704.845750] __arm64_sys_ioctl+0xa8/0xf0 [ 1704.849677] el0_svc_common.constprop.0+0x78/0x1a0 [ 1704.854469] do_el0_svc+0x24/0x90 [ 1704.857803] el0_svc+0x14/0x20 [ 1704.860860] el0_sync_handler+0x1a4/0x1b0 [ 1704.864869] el0_sync+0x180/0x1c0
Signed-off-by: Xianzhong xianzhong.li@nxp.com
--
LF-4436-3 [#imx-2664] add back scatter list to fix 6QP ocl cts failure
It is necessary to perform data cache sync-up with dma_map_sg operation
Revert "LF-4436-2 [#imx-2664] allocate scatter list for cacheable pages only"
This reverts commit fa4b5e7b6e9bc3c2d7363f5a7661b7174383923b.
Signed-off-by: Xianzhong xianzhong.li@nxp.com
--
LF-4656 [#imx-2711] disable systrace by default
this was introduced in MGS-6394 imx-2620, to support trace-cmd
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4656 [#ccc] declaring variable with initializer
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
LF-4656 [#ccc] fix dereference after null check
Coverity Issue:6124049
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
MGS-6408 [#imx-2696] fix Security vulnerabilities
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
MGS-6394 [#ccc] 0047-IMX-2620-CL377027-dev64x_234062-IMX-2620-trace-cmd-s
IMX-2620 CL377027 dev64x_234062:IMX-2620 trace-cmd support in the gpu lib on linux
Signed-off-by: Ke Feng ke.feng@verisilicon.com Signed-off-by: yuan.tian yuan.tian@nxp.com Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
MGS-6394 [#ccc] 0037-KERNEL-SPACE-Update-gc_feature_database.h-to-389622
Update gc_feature_database.h to 389622
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6394 [#ccc] 0071-Update-gc_feature_database.h-to-380216
Update gc_feature_database.h to 380216
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6394 [#ccc] 0014-IMX-2662-CL383711-KERNEL-SPACE-dev64x_234062-merge-t
dev64x_234062: merge the patch of IMX-2662.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
MGS-6394 [#ccc] 0002-CL381777-KERNEL-SPACE-Remove-WL_EGL_PLATFORM-re-defi
Remove WL_EGL_PLATFORM re-define in driver makefile as it is defined in Wayland System header "wayland-egl.h". Only use EGL_API_WL define for WL specific code in GPU drivers.
Signed-off-by: Ke Feng ke.feng@verisilicon.com
Signed-off-by: yuan.tian yuan.tian@nxp.com
--
LF-4373-2 [#ccc] fix 6qp opencl cts kernel panic
Only update timeout configuration for GPU3D/NPU only, Add hardware pointer check to avoid below kernel panic
[ 271.652290] Unable to handle kernel NULL pointer dereference at virtual address 00000010 [ 271.661137] pgd = 79b81a68 [ 271.664086] [00000010] *pgd=00000000 [ 271.667675] Internal error: Oops: 17 [#1 (closed)] PREEMPT SMP ARM [ 271.673075] Modules linked in: caam_jr caamkeyblob_desc caamhash_desc caamalg_desc crypto_engine rng_core authenc libdes mxc_v4l2_capture ipu_bg_overlay_sdc ipu_still ipu_prp_enc ipu_csi_enc ipu_fg_overlay_sdc imx_vdoa ov5640_camera_int ov5640_camera_mipi_int secvio caam v4l2_int_device error galcore(O) [ 271.700005] CPU: 1 PID: 1823 Comm: test_basic Tainted: G O 5.10.52-00002-g338ef9e708e5 #1 (closed) [ 271.709486] Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree) [ 271.716090] PC is at gckKERNEL_AttachProcessEx+0x84/0x1d8 [galcore] [ 271.722405] LR is at gckKERNEL_AttachProcessEx+0x54/0x1d8 [galcore] [ 271.728675] pc : [<7f010fe0>] lr : [<7f010fb0>] psr: 600f0013 [ 271.734944] sp : 8683bef0 ip : 00200001 fp : 411232e8 [ 271.740170] r10: 000000f8 r9 : 00000001 r8 : 82866550 [ 271.745397] r7 : 831489c0 r6 : 866a0300 r5 : 0000071f r4 : 833f8800 [ 271.751925] r3 : 00004e20 r2 : 00000000 r1 : 8358f800 r0 : 83064200 [ 271.758456] Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none [ 271.765594] Control: 10c5387d Table: 132a004a DAC: 00000051 [ 271.771342] Process test_basic (pid: 1823, stack limit = 0x6873d5cf) [ 271.855781] [<7f010fe0>] (gckKERNEL_AttachProcessEx [galcore]) from [<7f00d118>] (drv_release+0x5c/0xa4 [galcore]) [ 271.866182] [<7f00d118>] (drv_release [galcore]) from [<8025b7e0>] (__fput+0x78/0x234) [ 271.874113] [<8025b7e0>] (__fput) from [<8014dc68>] (task_work_run+0x90/0xbc) [ 271.881260] [<8014dc68>] (task_work_run) from [<8013397c>] (do_exit+0x378/0xa10) [ 271.888664] [<8013397c>] (do_exit) from [<8013407c>] (do_group_exit+0x3c/0xb4) [ 271.895893] [<8013407c>] (do_group_exit) from [<80134104>] (__wake_up_parent+0x0/0x18) [ 271.903814] Code: e594200c e3043e20 e59410f0 e5940004 (e592c010) [ 271.909989] ---[ end trace f095bb9104dab993 ]---
Signed-off-by: Xianzhong xianzhong.li@nxp.com
--
LF-3902 [#imx-2679] request_threaded_irq instead
- remove some dead task
- remove the extra semas
- increase the priority of the threadRoutine from 100 to 40
pass 8MP/8QM full mode test. 8MP suspend/resume ok at MGS-6236
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
MA-19350 [#imx-2662] Not use kernel_write/kernel_read on Android non-debug build
Google not allow to use kernel_write/kernel_read in drivers for secure reason. These symbols can’t be added into GKI (Generic Kernel Image) symbol list, to meet Google's request skip use them for non-debug build on Android.
Signed-off-by: Richard Liu xuegang.liu@nxp.com
--
LF-4373 [#ccc] Met random hang in Gfxbench vulkan testing
Increase gcdGPU_TIMEOUT value to avoid gpu false hang when running heavy workload on super-low GPU device.
Signed-off-by: Yuan Tian yuan.tian@nxp.com
--
LF-4436-2 [#imx-2664] allocate scatter list for cacheable pages only
sg_alloc_table_from_pages will consume system memory for gfp allocator, scatter list is used for the addtional operation on cacheable memory, check cacheable flag to allocate scatter list when necessary only.
[ 1704.740010] Chrome_InProcGp invoked oom-killer: gfp_mask=0xcc0(GFP_KERNEL), order=0, oom_score_adj=0 [ 1704.749265] CPU: 3 PID: 1251 Comm: Chrome_InProcGp Tainted: G C 5.10.52-lts-5.10.y+g5788c4507376 #1 (closed) [ 1704.759618] Hardware name: NXP i.MX8MNano EVK board (DT) [ 1704.764929] Call trace: [ 1704.767385] dump_backtrace+0x0/0x1a0 [ 1704.771051] show_stack+0x18/0x70 [ 1704.774368] dump_stack+0xd0/0x12c [ 1704.777770] dump_header+0x48/0x1f4 [ 1704.781259] oom_kill_process+0x200/0x204 [ 1704.785268] out_of_memory+0x178/0x560 [ 1704.789018] __alloc_pages_nodemask+0xb34/0xc24 [ 1704.793549] alloc_pages_current+0x80/0x13c [ 1704.797752] __get_free_pages+0x14/0x50 [ 1704.801595] __sg_alloc_table_from_pages+0x22c/0x484 [ 1704.806559] sg_alloc_table_from_pages+0x24/0x3c [ 1704.811176] _GFPAlloc+0x2b0/0x670 [ 1704.814579] gckOS_AllocatePagedMemory+0xf4/0x1fc [ 1704.819302] gckVIDMEM_NODE_AllocateVirtual+0xc4/0x190 [ 1704.824446] gckKERNEL_AllocateVideoMemory+0x2f8/0x380 [ 1704.829584] _AllocateLinearMemory+0xb8/0x1b4 [ 1704.833943] gckKERNEL_Dispatch+0x20c/0x13d4 [ 1704.838229] gckDEVICE_Dispatch+0x80/0x1bc [ 1704.842328] drv_ioctl+0x1b4/0x270 [ 1704.845750] __arm64_sys_ioctl+0xa8/0xf0 [ 1704.849677] el0_svc_common.constprop.0+0x78/0x1a0 [ 1704.854469] do_el0_svc+0x24/0x90 [ 1704.857803] el0_svc+0x14/0x20 [ 1704.860860] el0_sync_handler+0x1a4/0x1b0 [ 1704.864869] el0_sync+0x180/0x1c0
Signed-off-by: Xianzhong xianzhong.li@nxp.com
--
LF-4436-1 [#imx-2664] fix gpu memory limit to avoid system oom killer
Chromium v91 introduces more memory consumption to trigger out of memory, gpu memory limit become ineffective when free ram include more cma pages,
gpu memory limit apply gfp allocator only, need exclude cma pages. global_zone_page_state interface is introduced in 4.14 kernel.
enable gpu memory limit feature for cma and reserve allocator, skip gpu memory limit check for contiguous request by design.
[ 1251.328724] Chrome_InProcGp invoked oom-killer: gfp_mask=0xcc0(GFP_KERNEL), order=0, oom_score_adj=0 [ 1251.337928] CPU: 1 PID: 1160 Comm: Chrome_InProcGp Tainted: G C 5.10.52-lts-5.10.y+g5788c4507376 #1 (closed) [ 1251.348294] Hardware name: NXP i.MX8MNano EVK board (DT) [ 1251.456834] Mem-Info: [ 1251.459158] active_anon:447 inactive_anon:121521 isolated_anon:0 [ 1251.459158] active_file:2218 inactive_file:2748 isolated_file:0 [ 1251.459158] unevictable:0 dirty:28 writeback:0 [ 1251.459158] slab_reclaimable:3263 slab_unreclaimable:8837 [ 1251.459158] mapped:31664 shmem:38593 pagetables:2300 bounce:0 [ 1251.459158] free:96590 free_pcp:653 free_cma:78619
Signed-off-by: Xianzhong xianzhong.li@nxp.com
--
MGS-6358 [#imx-2679] remove unused stdarg.h
Signed-off-by: Jiyu Yang jiyu.yang@nxp.com
--
MGS-6400 [#imx-1870] fix vivante drm database leak
contiguous memory is part of video memory allocation, remove the duplicated database for contiguous memory, need set contiguous memory flag as database type.
Signed-off-by: Xianzhong xianzhong.li@nxp.com