Tom Hochstein [Tue, 14 May 2019 23:40:58 +0000 (18:40 -0500)]
imx-gpu-apitrace: Update
013499c6 wrappers/egltrace: drop the hardcoded library path 0697c15a gltrace: avoid recursion caused by viv extensions 5b56f5a5 specs/glapi: consistent signatures for viv extensions f6e1a557 glext.h: declare the viv extensions 4cb5ad01 retrace,cmake: link with vivante when using waffle 95ba8bd3 retrace,cmake: look for libwaffle in waffle_library_dirs ca7f1071 retrace,waffle: set wayland as default WAFFLE_PLATFORM 6ea46b12 retrace,waffle: use waffle error reporting api
Signed-off-by: Tom Hochstein <tom.hochstein@nxp.com>
Tom Hochstein [Tue, 14 May 2019 23:36:49 +0000 (18:36 -0500)]
libdrm: Update
95645843 headers: Sync with drm-next 49cca25f MA-13450 Switch layer memory allocation to virtual pool to improve performance. 8edffe33 MA-13354 Fix imx_drm head file license issue for libdrm-imx. e1658ff7 MA-13122 Add DRM_VIV_GEM_CMA_LIMIT support for TS buffer.
Signed-off-by: Tom Hochstein <tom.hochstein@nxp.com>
Tom Hochstein [Mon, 13 May 2019 15:08:05 +0000 (10:08 -0500)]
imx-gpu-viv: Fix i.MX 8 and 8X binary compatibility
The removal of gl32.h for 8QXP can cause build breaks in a multi-machine
build folder like the following:
| .../build/tmp/work/imx8qxpmek-poky-linux/gstreamer1.0-plugins-good-qt/1.14.4.imx-r0/recipe-sysroot/usr/include/QtGui/qopengl.h:105:13: fatal error: GLES3/gl32.h: No such file or directory
| # include <GLES3/gl32.h>
To make the ABI safe for 8 and 8X, which have a shared PACKAGE_ARCH,
provide gl32.h in both cases. The lack of support can then be handled
by a runtime error.
Signed-off-by: Tom Hochstein <tom.hochstein@nxp.com>
Tom Hochstein [Wed, 8 May 2019 17:34:38 +0000 (12:34 -0500)]
imx-gpu-viv: Upgrade to 6.2.4.p4.0
Key Features of the Vivante 6.x Driver Series
- Full support for Khronos OpenGL ES 3.2 API. The new OpenGL ES 3.2 and OpenGL ES Shading Language 3.20 specifications bring AEP,
plus additional functionality, into core OpenGL ES. Vivante OpenGL ES 3.2 implementation has passed Khronos OpenGL ES 3.2 Conformance Test.
And Vivante GC7000XS* GPU core has been certified by Khronos as an OpenGL ES 3.2 compliant product.
- Backward Compatible with 5.x and 4.x Drivers. Existing OpenGL ES 3.x/2.0 applications can run on these drivers without any changes.
Known Issues will be addressed in the GA release:
- Vulkan 1.1 CTS failure
- OpenCV some Tests failure
- MGS-4743 [MX7ULP_EVK]std::bad_alloc reported when running glmark2-es2-wayland. Rootcause GCC toolchain issue
- MGS-4728 RenderToTexture rendered abnormally
6.2.4 p4
=======================
Further enhancement
- commit worker to support dual GPUs
- Enablement of Robust feature Android CTS
- Vulkan driver fixes to comply with Android CTS
- OpenCL general improvements
Improvements and Bug fixes
MGS-4638 [#imx-1373] fix gpu kernel build error with kasan config
MGS-4629 [#imx-1369] Fix the GPU hang issue in vxuThreshold()
MGS-4634 [#imx-1374] weston failed to start as 64 bit kernel with 32bit graphics lib case
MGS-4609 iMX8QM - glsl memfault, mat4 access
MGS-4601 fix transform feedback random failure
MGS-4531 disable loop unroll for opencl kernel
MGS-3944fix frag-depth for WebGL 2.0 CTS
MGS-4547-2 [#imx-1330] second fix for pow performance issue
MGS-4547fix built-in pow performance issue
MGS-4611 fix 8mm es20 cts failure with 3.2.5.1
MGS-4448 fix vulkan sascha deferredmultisampling demo, deferredmultisampling
MA-11528 Support enable framebuffer direct rendering
MGS-4550 vxSobel3x3Node() segmentation fault when output_x or output_y is NULL
MGS-4569 fix GPU hang when play 4K video
MGS-4394 Out of memory issue on MQ when running opencv_perf_imgproc
MGS-4541 fix galcore kernel panic in debugfs
MGS-4551 fix gpu hang by wrong idle check
MGS-4444 fix opencv_perf_photo failure
MGS-4523-1 Add support for gcc 8.2
Increase GL_CHIP_NAME_LEN to 32
MGS-4483 ACL, v18.11, arm_benchmark, GPU hang , at case 3646 and 4406
MGS-4476-1 Fix GBM used too much memory in p4 driver
MGS-3596-1 Disable surface compress if g2d-renderer
MGS-4513 avoid oom killer in gpu kernel driver
MGS-4376 fix low performance with CMA allocator
MA-13763 Fix fence signal memory leakage issue.
MGS-4494 fix GL_SRG8_ALPHA8 renderable texture
MGS-4326 support YUY2 to NV21/NV61
Fix Video performance dropped with in_fence feature
Fix memory leakage issue when render buffer is reused to bind to FBO.
Fixed WILDCAT-41 YUY2 conversion & composition in Weston
Refine hardware query idle.
i.MX8M GLSL shader conversion output distorted need to set all calculations related to mod() source to high precision to avoid precision loss in dual16 mode. The case "Ocean surface simulation" Out_of_momory when run compubenchcl_v2.0
Support for GBM asynchrounous support
Add RGBA_FP16 format support to fix CtsGraphicsTestCases crash issue: fix R8G8B8_UNORM related fail
When destory the wayland surface or display, we should destroy the source until the swap thread handles all workers.
Fix SoftISP denoise performance fix
Enable OES_EGL_image_external_essl3 extension for Android/QNX.
Added two new API in GBM driver to support 18.1.9 gbm version.
Fix in driver to support Vulkan-validation-layer
Fix for vcompiler preprocessor fails to parse function-like macros with zero parameter
Query clock frequency dynamically.
Fix filterblit vertical strip issue
i.MX8MQ Android DRM no-resolve feature for KMS
Fixed the arithmetic assign's left hand side to be memory store instead of simple MOV.
QNX - problem rendering to KHR image on iMX8QXP B0. Flush the tile status and decompress the pixmap buffers, as pixmap buffer could be get by QNX system calls.
Dont' enable gcdENABLE_DRM when kernel doesn't have CONFIG_DRM.
Fix galBlit2 & galBlit & stretchBlit rotation issue.
Improve to comply with eglslsanbox-player
6.2.4 p3
=====================
Skipped
6.2.4 p2 GPU Release for i.MX6, i.MX7, i.MX8
===========================================
GA Release Updated (incremental fixes from Beta release
=====================================================
Support Android P. Fixes including new cases added in DEQP, Vulkan CTS
Improvements and Bug fixes
MGS-4326 support YUY2 to NV12/YV12/NV16
MA-13159 Fix crash caused by changes in copyimagebuffer
MA-13159 Enable Vulkan CTS 64, 128 bpp msaa formats
MA-13159 Fix Vulkan SPIRV shader compilation issues
MA-13159 [#imx-1215] Fix dEQP-GLES3.functional.shaders.indexing.moredynamic.* CTS fail
MGS-3083 [#imx-662] fix kernel build error with kasan config
MGS-4184 [#imx-1158] Display image is blurred when run SVG demo in VG3d environment.
MLK-20450 [#imx-1251] Fix Linux NO SMP kernel - GPU
MGS-4087 [#imx-1095] Fix the GPU hang issue when run opencv_perf_imgproc on QXP
MGS-4384-81 vkmark: The screen flash the gray squares when run vkmark(cube)
MGS-4347 App weston-touch-calibrator is stucked in wl_egl_window_destroy
MA-13450 Switch layer memory allocation to virtual pool to improve performance.
MGS-4376 fix low performance with CMA allocator
MA-13203 Fix kernel panic when do video plaback with dual display on 8mq.
MGS-4340 clpeak, kernel latency got a negtive number randomly, 1/3
MGS-3612 Fix Lower performance of SoftISP running on MX8QXP
MA-12928-2 [#imx-1199] Use gcdANDROID MARCO to control android build
MA-13067-17 [#imx-1215] Fix dEQP-VK.renderpass.suballocation.attachment.4.532 fail
MA-13166-1 [#imx-1229] Fix CtsNativeHardwareTestCases CTS crash issue
MGS-4161 Fix glslsandbox some items fail
MGS-3732 [#imx-1224] OpenCL C, select() issue.
MGS-4407 [QNX] gl-dlls.so is missing reported in slog2info when graphics apps started 402abada7 MGS-4406 [QNX] - fix free error message when slaying screen b01759433 MGS-4392 [QNX] - Screen hang when slay - problem terminating resource manager 105078615 MGS-4390 [#imx-1249] [QNX] - memory fault when playing glslsandbox case 196
============================
Improvements and Bug fixes
GPU3D:
MGS-4092 eglslsanbox-player cause galcore hang
MGS-4152 fix ES31 builtin cts failures
MGS-4193 fix too many uniforms for webgl 1.0.2 cts
MGS-3434 fix webgl color-adjust item draw error
MGS-4151-28 pointer variable fix for CL compiler
MGS-4151-27 infinite loop fix for CL compiler
MGS-4151-26 texture border clamp fix for webGL2.0
MGS-4151-23 android gles3 CTS fix
the vulkan saschawillems (pipelines) render error on imx8qm wayland/BG4CT O80 CL163426 correct the scissor rect, intersect scissor with viewport.
MGS-4151-8 merge Vulkan color/depth address fix
EGL
MGS-4247 Video flicker with green splash
MGS-4134 [QNX] problem rendering to KHR image on iMX8QXP B0
MGS-4170-1 [QNX] - update memory allocation to correctly configure cache
MGS-3501-2 Enable asynchrounous swap in libGBM
disable gbm import feature for EGL_IMAGE instead of WL_BUFFER
QNX - optimization of _GetWindowSize and gckOS_WriteMemory()
OpenVX
MGS-3579 fix the OpenVX sample segmentation fault issue.
MGS-4151-25 OpenVX fix for OpenCV crash
MGS-4151-22 openVX host image fix
MGS-4151-2 merge openvx HW caps initialization
MGS-4151-1 merge openvx cmdBuffer release
MGS-4151-5 merge OpenVX fix for vxNonLinearFilterNode
openvx api return failure error if VXCreateContext() is called after eglInitialize() in GLES application
OpenVX error when release user kernel.
G2D-GPU
MGS-4135 Weston crash when run two videos with g2d compositor
MGS-4128 Fix g2d_multiblit_test case segmentation fault.
MGS-4091 fix g2d YUY2 scaling issue
MGS-4091 fix g2d YUY2 output issue
MGS-4113 Fix G2d to work with new ION
General
MGS-4066 fix oom with OpenVG memory leak test
MGS-4151-24 debugfs fix for kernel panic
MGS-4151-21 max invocation size fix
MGS-4151-20 es20 fbo fix for subimage
MGS-4151-19 vCompiler fix for es20 tutorial
MGS-4151-18 glmark2 draw error fix
MGS-4151-17 compiler fix for Android 8.0 cts
MGS-4151-16 shader MOV optimization fix
MGS-4151-15 es20 fbo renderer fix
MGS-4151-14 shader recompilation fix
MGS-4151-13 GL framebuffer attachment fix
MGS-4151-12 arithmetic expression fix
MGS-4151-11 webGL 2.0 conformance fix
MGS-4151-10 vulkan saschawillems fix
MGS-4151-9 merge kernel panic fix in GFP oom
MGS-4151-7 merge gckHARDWARE_QueryIdle fix
MGS-4151-6 merge register read fix for clock-off
MGS-4151-4 merge openVG A8 format fix
MGS-4151-3 merge veglCreateWaylandBufferFromImage fix
MGS-4222 [QNX] - add option to disable depth compression
MGS-4223 QNX - add a disable y inversion option
MGS-4137 [QNX] - problem running OpenGL/OpenCL apps in user mode
MA-12629 Fix UI display not correct on 7ulp
MGS-4170 [QNX] - update memory allocation to correctly configure cache
MGS-3255-47 remove redundant surface unlock
MGS-3255-46 add sync commit mode to debug worker thread
MA-12612 Fix WebGL conformance 2.0.1 freeze on android
MGS-4076 Fix opencv_test_imgproc GPU hang issue
MGS-3255-45 optimize pixel copy with 3d blit
MA-12411 Enable in-fence function for DRM driver on IMX8 platform.
MA-12560 Fix UI display issue on imx845s
MA-12481 Enable direct rendering without tile status on android 845s
MA-12227 Add framebuffer compression support on android
MGS-4130 Cannot pass the stress test of low bus suspend/resume.
MGS-4147 Don't try to de-register OPP entries when we don't have them in the first place
MGS-4135 display blurred when run weston-nested
MGS-4120 Enable GPU workaround on 4GB LPDDR4 platform
MGS-3785 test_buffers in OCL1.2 CTS crash on 850D
MGS-3527 Rename fence to dma_fence
MGS-4125 fix 6sx suspend/resume stuck issue
MGS-3951 fix multiple cts hang issue
MGS-4111 Fix displaying wrong governor values on 8QP
MA-10844 GPU driver pass build after refine some NXP git path
MGS-3527 enable android native fence for Weston sync
MGS-3951 fix multiple cts hang on mscale and qxp
MGS-3255 flush commit worker to fix QM cts crash
MGS-4043 [QNX] Sync QNX resource manager with latest Linux version
MGS-3501 Enable asynchrounous swap in libGBM
MGS-4105 fix OPP failure log in gpu kernel
MGS-4104 Enable direct rendering on wayland
MGS-4103 fix gpu axi bus error on 6sx
screen mess in fullscreen mode with fb modifier.
"Could not open egl display" when run monkey test case for long time. CL150804 galcore: add MAP_NORESERVE flag for virtual memory
Refined gcoOS_GetTLS logic to remove recursive call
Fixed CtsMonkeyTestCases crash issue
Upgrade OpenGL ES CTS with opengl-es-cts-3.2.4.0 version CL151117 Fixed CTS dEQP-GLES2.functional.fbo.render.*_depth_components*_stencil* by excluding separate depth and stencil buffer config.
test global bandwidth of clpeak, get an error, clEnqueueNDRangeKernel (-54), local work size larger than Kernel maxWorkGroupSize
relax the maxworkgroupsize when threre's no barrier or private memory in the kernel.
[imx6sl_wayland]:fail when run vgcts CL151588 fix vgcts failed on 6SL wayland. backbuffer should be re-create if config format is changed.#19212.
ISP in imx-gpu-sdk/OpenCL met GPU hang with QM beta2 (regression introduced in 23Nov release)
OCL1.2: test_image_streams: 2 tests failed on 8QXP FB.
Add VX_CONTEXT_NONLINEAR_MAX_DIMENSION support to vxQueryContext
[NXP][MX8_MSCALE_BETA]HDMI hot plug, cause Weston crash, blank screen CL151952 Fixed IMX-844 weston crash when HDMI hotplug
CL151953 Prepare code for GPU PM and ISR refinement
6.2.4: fix events stuck issue when clock off CL151955 Refine GPU PM logic to be more robust.
fix synchronization between isr and clock state,fix kernel panic issue in bug
Correct hw-event synchronization between pm and other threads
Support aliased attribute in compiler. Check the aliased attribute for ES11 only.
the high dynamic range crash on imx8qm/imx8mscale android o CL153605 fixed #20270, changed SPV_MAX_OPERAND_NUM to 30
Fix:OpenCL precompiled kernel doesn't work
[8mscale_wayland[cl12_conform]:random fail when run "images/kernel_read_write/test_image_streams max_images"
fixed potential mmu setup command allocate fail, since kernel might not support CONFIG_ZONE_DMA32.
Key Features
============
- GPU Frequency Scaling governor
- OpenCL improvements
- OpenVX 1.1 Support
- Wayland-weston DRM Support
- Wayland-weston with 2D accelerated renderer
- X11 with DRI3 support with multiple display support
- DRMFB (Experimental feature: DRM framework wrapper for Vivante FB APIs)
6.2.4 Improvements
==================
- Fixed multiple Vulkan CTS 1.0.2.* failures.
- Improved vProfiler support in driver to have more accurate HW counter values.
- Fixed Linux kernel panic issue when try to put_page for reserved memory.
- Enhanced CL kernel compiler support for OpenCV library and applications.
- Opened Wayland server side buffer protocol in EGL driver to enable WL applications.
- Refined flat mapping kernel memory allocation. Use a single allocation for all STLBs.
- Fixed a kernel driver issue to enable repeated galcore insmod/rmmod to work with mmu.
- Fixed QNX driver issues to pass ES CTS and CL CTS.
- Fixed the 3DVG SDK app up-side down rendering issue
- Add Wayland support for EGL_EXT_swap_buffers_with_damage and EGL_EXT_buffer_age
- Support the tiled and compression texture with EXT_image_dma_buf_import_modifiers
- Stream cache optimization.
- VBOs, Index buffer cache optimizations for performance
6.2.4 Changes and Bug Fixes
===========================
General:gpu idle cycle counter register has two values in gpu 5.x
WL-EGL: [6sx-sdb/6sx-ard][6.2.2]Memory leak found in test destory_simple-egl2
General: Fix GPU clock issue with built-in mode on QM/QXP
General: Allow same process to open more than one fd with the driver
General: iMX6Q : Dithering is only performed on first frame
General: [8QXP]repeated insmod/rmmod galcore w/ power management off causes hang upon reloading
General: Replace CONFIG_PM_RUNTIME with CONFIG_PM
General: [imx8qm] Met segmentation fault when run GLES samples stress on GPU0, GPU1 and bridged mode
General: Use default implementation for cache operations.
General: fix the suspend/resume hang issue on imx6
General: fix power-off issue for ctrl-C exit
General: Fix power-off random failure when GPU become idle
GBM: gbm->surface_create is null before call eglGetDisplay caused glmark2 segmentation fault.
GBM: Add map/unmap functions
GBM: Enable GBM EGL Support in FB,WLD,X11 backends
GBM: kmscube does not work
GBM: [Imx8][wayland]gbm_surface_lock_front_buffer failed in Weston
GBM: [8qm][weson]Error log found after weston stopped.
GBM: Enable GBM EGL Support in FB,WLD,X11 backends
OpenCL: hevc_cl_test has performance drop compared with imx8 qxp. 25%
FB-EGL: GPU installs a signal hander, which can overwrite user's signal handler
X11-EGL: [MX8/imx6QP] X11: glxs app test some frames render incorrectly.
X11-EGL: server down when run sdk app multi times in background
WL-EGL: [wayland]The fps of weston-simple-egl was changed according to the back buffer number
WL-EGL: tutorials crash on wayland with ivi-shell as core
WL-EGL: weston-simple-egl full screeen met core dump
EGL: fix cma_release dump with ctrl-c
EGL: gplay-1.0 met "Failed to convert multiview video buffer"
OpenCL: opencl high cpu usage
OpenCL: clpeak, met GPU freeze on 6q with 6.2.2 driver.
OpenCL: opencv_perf_objdetect failed with clBuildProgram()
OpenCL: math_brute_force of OpenCL 1.2 CTS, crashes with 6.2.3 driver on qxp & mscale. randomly, like 40%
OpenCL: some data types of CL_R format are not supported in driver 6.2.3
[i.MX_Android]use kzalloc instead of kmalloc during fence_init.
[i.MX_Android] DEQP GLES3.1 CTS cases fail on mx6qp board
[i.MX_Android]GPU kernel driver build error when enable KASAN check
[i.MX_android] YUV buffer with GRALLOC_USAGE_HW_TEXTURE usage will allocate as titled RGB buffer
[i.MX_android] Add more reserve memory in private_handle_t for imx8 on android O
[i.MX_android] dEQP-EGL CTS fail and hang up issue on mx8qm board
[i.MX_android] dEQP-VK CTS cause GPU dump on android 8.0 mx8qm board
[i.MX_android] GPU lock up issue on Android7.1 + 4.9.y kernel with GPU 6.2.2 driver
[i.MX_android] CTS testNonDefaultLauncherCantStart fail on GPU6.2.2.p1
[i.MX_android] Fix cts testOpenGlEsVersion fail on mx6qp
[i.MX_android]Fix monkey test reboot issue due to drm gem bug
[i.MX_android]Export Buffer size and set virtual address when lock
[i.MX_android]Support more YUV formats in drm gralloc
QNX - iMX8QM - Deadlock seen with Screen API test
QNX -[iMX8] Fix NV12 handling
QNX - gpu hang when running es32 conformance test suite
QNX OpenCL buffer mapping doesn't work
QNX - [iMX8] Rotation using blitter seems broken
QNX - OpenVG apps failing on iXM8QXP with 6.2.4
QNX CAR apps report EGL errors
QNX MMU faults on iMX8DV
6.2.4 Known Issues
==================
MGS-3453 [8QM] X11:WebEngine test met core dump and segmentation fault
MGS-3434 [8QXP/8QM] webgl color-adjust item draw error.
MGS-3423 es11 test failed with 'could not find encoder'
MGS-3421 weston:G2D: rotate(transform) not working
MGS-3396 [8QM] Xwld: met blurred screen when use G2D compositor. 100%
MGS-3373 [imx8qxp] KPA met gpu hang
MGS-3354 [imx8qm/8mq] Vulkan cts can't pass with 22 failed due to Khronos spec error
MGS-3330 [8QM/8QXP]GPU top tool met GPU dump when monitor process in background. 100%
MGS-3276 [imx8qm] x11 2D acceleration performance test met fatal IO error
Key Features of the Vivante 6.2.2P0 Driver Series
Full support for Khronos Vulkan 1.0 API and WSI extension APIs Vulkan is a new generation graphics
and compute API that provides high-efficiency, cross-platform access to modern GPUs used in a wide
variety of devices. Vivante’s Vulkan 1.0 implementation has passed Khronos Vulkan 1.0 Conformance Test.
And Vivante GC7000XS* GPU core has been certified by Khronos as a Vulkan 1.0 compliant product.
Full support for Khronos OpenGL ES 3.2 API
The new OpenGL ES 3.2 and OpenGL ES Shading Language 3.20 specifications bring AEP, plus additional
functionality, into core OpenGL ES. Vivante OpenGL ES 3.2 implementation has passed Khronos
OpenGL ES 3.2 Conformance Test. And Vivante GC7000XS* GPU core has been certified by Khronos as an
OpenGL ES 3.2 compliant product.
Draw calls specifying a base vertex parameter
Floating-point framebuffers
Robust buffer access control
Support for querying CONTEXT_FLAGS, as needed by debug and robust buffer access functionality.
Khronos OpenVX 1.0.1 API
Khronos OpenCL 1.2 API
Improvements:
1. Added mutex protection for referencing gctSIGNAL in gckOS_MapSignal to fix a MT race issue.
2. Streamlined GPU address calculation base on MC20 (0/1) and MMU (0/1) combinations.
3. Fixed multiple Android HWC 2.0 rendering issues. Refined HWC 2.0 driver implementation.
4. Improved HWC2.0 composition performance by composing damaged regions only.
6. Enabled offline/online compiler IR assembly dump function for users.
7. Implemented the direct rendering support (no-resolve) for Wayland platform.
8. Added EGL_EXT_buffer_age extension for Wayland and fbdev platforms.
9. Updated wayland-viv protocol to support tile status sync from client to server.
10. Improved OpenCL 1.2 builtin function support with native GPU instructions.
11. Enabled OpenCL 1.2 API trace dump function controlled by VIV_TRACE environment variable.
12. Support for OpenGL4.0
13. Cleaned up driver code issues reported by Klocwork and Coverity.
Changes and Bug Fixes
VG3D: MGS-2972 fix 3D OpenVG render target orientation
EGL: MGS-2790 6sx-sdb Met memory dump when do xwld stress test
EGL: MGS-2885 fix recursive call to dequeue buffer issue on wayland
EGL: MGS-2785 Meet segmentation fault when test qt5nmapper on wayland
EGL: MGS-2734 fix x11 memory leak issue
Gralloc:MA-9561 fix crash issue on new gralloc1 interface.
2D: MGS-2900 fix filterblt memory leak
VG2D/6SL: MGS-2882 fix video memory allocator
OCL: MGS-2732 clCopyImage with max_images in OCL1.1 cts, crashes.
GLES: MGS-2747 enable the full config for ES3 CTS
GLES: MGS-1832 fix Mirada performance issue. optimize cpu load for fence check
WEBGL: MGS-2779 webgl 1.0.2 conformance test met out of memory
General:MGS-2923 fix gc2000 gpu hang when flush TS cache
General:MGS-2907 fix wrong cache invalidate with dma_map_area
General:MGS-2902 fix wrong release_mem_region
General:MGS-2848 fix gpu hang for flat mapping optimization
General:MGS-2848 Optimize MMUv1 FlatMapping performances
General:MGS-2857 fix spinlock in gpu kernel driver
General:MGS-2805 Enable gcdENABLE_VG by default.
General:MGS-2827 fix serious performance drop on imx6q
General:MGS-2832 fix axi bus error with correct baseaddress
General:MA-9483 Refine android native fence code to match 4.9 kernel
General:MA-9473 Fix gpu random crash issue caused by NULL pointer
General:MGS-2804 fix flat mapping overflow issue
General:MGS-2786 fix gpu kernel panic with debugfs
General:GS-2776 fix galcore device open failure
General:MGS-2767 i.MX7ULP GPU suspend/resume problem
Coverity:MGS-2858 Double free in __driUtilUpdateExtraDrawableInfo
Known Issues
OpenCL 1.1 CTS failures. Will be addresed in 6.2.3
MGS-3027 GPU: X11 manywin draw error. 100%
MGS-2935 [imx6QP] X11: glxs app test some frames render incorrectly. 100%
MGS-2860 [MX7ULP/imx6SLEVK] X11/Xwld: XWindow 2D Acceleration performance test met "X Error of failed request: BadMatch (invalid parameter attributes)" error. 100%
MGS-2829 [imx6qp/imx6d] xwld: mm06 Samurai and Proxycon performance regression comparing with L4.1 using 5.0.11p8
MGS-2310 [imx6qp/imx6dq/imx6dl/imx6sx] webgl: two items performance data is lower than L4.1_GA
MGS-1828 [imx6qp_sd/imx6Q/imx6dl] X11/xwld: VGMARK on vg3d have some items performance drop comparing with L4.1 GA
MGS-1350 [imx6qp-sd] X/FB: Some Basemark test items performance drop comparing with L3.14.52GA which use 5.0.11p7
6.2.2.p0 New features:
- Added mutex protection for referencing gctSIGNAL in gckOS_MapSignal to fix a MT race issue.
- Streamlined GPU address calculation base on MC20 (0/1) and MMU (0/1) combinations.
- Fixed multiple Android HWC 2.0 rendering issues. Refined HWC 2.0 driver implementation.
- Improved HWC2.0 composition performance by composing damaged regions only.
- Enabled offline/online compiler IR assembly dump function for users.
- Implemented the direct rendering support (no-resolve) for Wayland platform.
- Added EGL_EXT_buffer_age extension for Wayland and fbdev platforms.
- Updated wayland-viv protocol to support tile status sync from client to server.
- Improved OpenCL 1.2 builtin function support with native GPU instructions.
- Enabled OpenCL 1.2 API trace dump function controlled by VIV_TRACE environment variable.
- Support for OpenGL4.0
- Cleaned up driver code issues reported by Klocwork and Coverity.
6.2.2.p0 Known Issues
- MGS-3074 [6qp] apitrace does not work
- MGS-3027 GPU: X11 manywin draw error.
- MGS-2935 [imx6QP] X11: glxs app test some frames render incorrectly.
- MGS-2860 [MX7ULP/imx6SLEVK] X11/Xwld: XWindow 2D Acceleration performance test met "X Error of failed request: BadMatch (invalid parameter attributes)" error.
- MGS-2829 [imx6qp/imx6d] xwld: mm06 Samurai and Proxycon performance regression comparing with L4.1 using 5.0.11p8
- MGS-2310 [imx6qp/imx6dq/imx6dl/imx6sx] webgl: two items performance data is lower than L4.1_GA
- MGS-1828 [imx6qp_sd/imx6Q/imx6dl] X11/xwld: VGMARK on vg3d have some items performance drop comparing with L4.1 GA
- MGS-1350 [imx6qp-sd] X/FB: Some Basemark test items performance drop comparing with L3.14.52GA which use 5.0.11p7
6.2.0.p4
- 3D: MGS-2551 [#imx-406] webgl case shader-with-too-many-uniforms.html failed on 7ulp
6.2.0.p3
- Unified driver for i.MX6X, i.MX7ULP. However current release focused in i.MX7ULP. Full integration testing not performedd so there could be some bugs.
6.2.0.p2 New features:
- Unified driver that includes OpenGL ES, Vulkan, and OpenVX.
- Multi-GPU combined mode and individual mode support. More information in the i.MX Graphics user guide
- Linux DMA-BUF buffer sharing.
- Brand new OpenGL 2.1, 4.0 driver implementation which is based on ES3 driver code.
- Improved vProfiler support to include many new HW counters.
- Compiler optimizations that improve benchmark/application performance.
- Resolved Khronos GLES3.x CTS failures in all EGLConfig runs.
- Fixed a YUV420 buffer allocation issue that caused 4K video OOM on Android N.
- Fixed several GLES driver memory leak problems that cause random failures in Android CTS.
- Corrected EGL buffer age calculation to resolve a GooglePlay UI flickering issue.
- Fixed a MMU exception by limiting the number of sampler prefetch in GPU.
- Fixed driver code issues reported by static code analysis tools.
- Add support for GL_LUMINANCE8_ALPHA8_EXT texture in the directTexVIV extension.
- Fixed Xserver crash issues on Yocto X11 desktop.
- Clamp 3Dblit rectangle size to surface area to fix a GPU hang in Wayland tests.
- Set HW type correctly in Wayland server side to fix a memory leak in 2D composition.
6.2.0.p2 Known Issues
- MGS-2513:[imx7ulp] RTC suspend/resume stress test met resume failed. GPU driver issue.
- MGS-2512:[imx7ulp] xwld:mm07 some items show nothing on display.
- MGS-2511:[imx7ulp] xwld: QT5 app show wrong resolution.
- MGS-2503:[imx7ulp] X11/Xwld: XWindow 2D Acceleration performance test met "X Error of failed request: BadMatch (invalid parameter attributes)" error.
Signed-off-by: Tom Hochstein <tom.hochstein@nxp.com>
Ideally a SOC layer should not have general purpose recipes like dpdk
and may be get the recipes from
http://git.yoctoproject.org/cgit/cgit.cgi/meta-dpdk working on imx SOCs
instead
This patch just limits the spill over to non nxp bsp layers
Joshua Watt [Wed, 1 May 2019 21:49:30 +0000 (16:49 -0500)]
imx-gpu-viv: Fix GLESv1 packaging
The GLESv1 library was being packaged with the name "libgles" instead of
the canonical "libglesv1". The latter is the name that recipes used when
RDEPENDing on GLESv1. Correct this so that the package is correctly
named "libglesv1", but RPROVIDES the old "libgles" name in case there
are recipes still depending on it.
Khem Raj [Wed, 28 Aug 2019 20:07:06 +0000 (13:07 -0700)]
qemu-qoriq: Mark is specific to imx/qoriq
It has started to fail now that we have libssh dependency in 4.1 instead
of libssh2 in earlier versions, so this recipe needs to be upgraded to
use 4.1 release as well, but I dont have way to test it out.
Andrey Zhizhikin [Fri, 23 Aug 2019 09:31:14 +0000 (09:31 +0000)]
mesa: switch mesa to swrast renderer for mx8mm machine, resolve build issue
After mesa update to 19.1.0, the OSMesa is not anymore explicitly set to
'classic' rather than defaults to 'none'. If the DRI driver is included
in the build and neither gl nor 'classic' osmesa is used - this causes
the build to fail.
This commit disables the gallium for mx8mm, sets the dri drivers to
swrast and explicitly defines the 'classic' osmesa to be built.
gallium is disabled since current support state of i.MX8M Mini GPU in
gallium is unknown.
Daniel Klauer [Thu, 1 Aug 2019 09:15:01 +0000 (11:15 +0200)]
atf: Fix builds of native tools during build process
atf builds some native tools such as fiptool during its build process,
this should use Yocto's native toolchain options instead of just the
host's.
This way the fiptool build will actually use the openssl-native dependency
from Yocto's native sysroot, instead of the host's openssl (which may not
exist).
Signed-off-by: Daniel Klauer <daniel.klauer@gin.de>
Daniel Klauer [Thu, 1 Aug 2019 09:06:56 +0000 (11:06 +0200)]
Split fiptool build into separate recipe
Previously, ddr-phy built native fiptool during its own build process,
but without using Yocto's native toolchain settings. This results in
unexpected host dependencies. For example, fiptool depends on openssl,
which may or may not be installed on the host.
So now fiptool is built in a separate recipe (atf-tools-native) which can
easily depend on openssl.
Signed-off-by: Daniel Klauer <daniel.klauer@gin.de>
Alban Bedel [Thu, 15 Aug 2019 09:39:18 +0000 (11:39 +0200)]
atf: Only enable and depends on optee when supported
Use the new optee machine feature to only enable it when it is supported
by the machine and the distro. We use a package config switch to only
add the dependency when it is needed. This prevent a build failure on
machines that don't support optee.
Alban Bedel [Thu, 15 Aug 2019 09:28:57 +0000 (11:28 +0200)]
optee: Add optee as a machine feature to all supported boards
Currently we assume that all qoriq arm64 boards support optee, however
that might not be the case for boards added by other layers building
on top of meta-freescale. To better support this case add a machine
feature to indicate if a given machine has optee support.
Add the new optee feature to all qoriq arm64 except ls2080ardb as it
doesn't seems to be supported by optee.
Fabio Berton [Mon, 12 Aug 2019 14:35:54 +0000 (11:35 -0300)]
Fix WKS_FILE to use files with .in extension
meta-freescale commit 6be9d197386b5c3bd72023981df805d42f87684c
renamed imx-uboot-spl-bootpart.wks to imx-uboot-spl-bootpart.wks.in
The .in extension in wks files allows bitbake variables to be used in
kickstarter files. Set WKS_FILES for all machines to match this new
filename.
This is the successor to libimxvpuapi. Since at this moment, there is
software still using the former, its recipe remains in place. Both
versions (1 and 2) can coexist on the same rootfs.
Version 2 introduces support for i.MX8 (mx8m and mx8mm for now), and also
still supports i.MX6.
Signed-off-by: Carlos Rafael Giani <crg7475@mailbox.org>
libimxdmabuffer provides an API for allocating and handling physically
contiguous buffers ("DMA buffers") on imx6, imx7, imx8 machines with the
imx-kernel. The underlying allocation can be backed by the PxP, IPU, ION,
DWL, G2D APIs.
The API is backend agnostic. The same structures and functions can be used
with the underlying PxP and IPU allocators for example. Furthermore, the
library defines a "default" allocator (which one is the "default" is
determined by the library and by the build configuration).
By using this API, libraries can use compatible types for exchanging
DMA buffers in userspace, and can also use the same API and support mx6,
mx7, and mx8 machines, without requiring platform specific code changes.
Signed-off-by: Carlos Rafael Giani <crg7475@mailbox.org>
*update to lsdk 1906 tag
include the following changes: 0123ced - Merge pull request #966 in GITAM/dpdk from 18.11-qoriq-dev to 18.11-qoriq 9e58741 - Merge pull request #965 in GITAM/dpdk from 18.11-qoriq-dev-ppfe to 18.11-qoriq-dev 87b8c52 - net/ppfe: add checksum offload in capability ae172a2 - bus/dpaa: fix incorrect rte_free of malloc memory
*update to lsdk 1906 tag
include the following changes: ff7bf7b - Merge pull request #32 in DQNS/vpp from ~NXA12342/vpp:19.01.1_devel to 19.01.1 5c2b539 - PVT: ipsec: Fixed capability matching for openssl based sha1 authentication a24588f - PVT: dpdk_plugin: introducing ipsec protocol offload 33d0d81 - Fixed Flexbuild integration issues c200b3f - dpdk-ipsec: Fixed ramdom Crash in esp traffic 2a0a031 - Enhanced the VPP startup configuration 05bce64 - Merge pull request #30 in DQNS/vpp from ~NXA19696/vpp:19.01.1 to 19.01.1 c63f8ad - Fixed RPM packaging of VPP libs w.r.t Tag format 85e9d4d - Merge pull request #29 in DQNS/vpp from ~NXA19696/vpp:19.01.1 to 19.01.1 26d2455 - Adding README for NXP platforms 9df88fa - Handled review comments d66c318 - Fixed Debian packing support for arm64 platform 513a05a - Fix for using Best mem pool ops as per platform d0f5fa4 - Updating path for cmake cross compile file fe17e96 - More Compilation Fixes e64c6ca - Cross compilation & RPM pkging known issues cbd68cb - 19.01.1 Release Notes 3262adc - cmake: fix out-of-git-tree build d302022 - Fix vpp crashing when attempting to run in kubernetes Pod 6342e48 - tcp: allow future acks if in window 46ee286 - IPSEC: crash on SA dump when SA are bound to tunnels bad0bcd - Fix crash in barrier sync when vlib_worker_threads is zero 372a33e - buffer chain linearization
*update PV to "19.01"
*refresh patch
*update DEPENDS
include the following changes: b3d08bb - update version number and log 7c4187c - add new portInfo lua data 0c39019 - Cleanup build, add plugin support and meson build not working f6e9e33 - fix crash not resetting tty and other fixes 4199555 - Merge branch 'bug-fix' 3942af7 - fix Update to version 3.6.5 176df0a - fix ldflags order for lua lib 8c746d0 - minor message missing a word 5e29e43 - reorder the defines on the LDFLAGS line 6ed0bde - convert to DPDK network structs and fix offload code. 0e8ac37 - fix lua builds oni ubuntu 18.10 5b537c9 - change port command to use uint16_t 96776e5 - update .cfg file information b5133f1 - add page stats screen 5b883b2 - add more color to static info a0c5819 - convert all rte_delay_ms to rte_delay_us_sleep 83b1bfa - remove need for rte_timer_manage 4dcc2db - use faster timer routines then rte_timer_manage cca5864 - remove cycles.h header 0d5d9ab - change cli_use_timers routine to run rte_timer_manage 7283ba7 - remove crc-strip option b44f9e6 - remove init stats structure a82bbf5 - fix formatting with spaces 81623f1 - update readme, version and changelog
*update to lsdk 1906 tag
include the following changes: 0123ced - Merge pull request #966 in GITAM/dpdk from 18.11-qoriq-dev to 18.11-qoriq 9e58741 - Merge pull request #965 in GITAM/dpdk from 18.11-qoriq-dev-ppfe to 18.11-qoriq-dev 87b8c52 - net/ppfe: add checksum offload in capability ae172a2 - bus/dpaa: fix incorrect rte_free of malloc memory 866adf5 - Merge pull request #961 in GITAM/dpdk from 18.11-qoriq-dev to 18.11-qoriq 0b36cdd - Merge pull request #953 in GITAM/dpdk from DPDK-1790 to 18.11-qoriq-dev 7dfce10 - PVT: nxp: supress warnings from performance script 4e26967 - PVT: nxp/README: update with loopback and no prefetch devargs option 71f3ebd - PVT: net/dpaa2: add optional non-prefetch rx mode 025c689 - PVT: eth/dpaa2: reset frc and ctrl in sg tx fd 6e8d2fe - net/ppfe: fix no ports available issue 28bfd64 - PVT: examples: l3fwd: Fix spec creation for traffic split scenario 279901c - Merge pull request #946 in GITAM/dpdk from DPDK-1799-pdcp-security to 18.11-qoriq-dev 066ff8d - crypto/dpaa_sec: update dpovrd for hfn override in PDCP 73c6c39 - Merge pull request #947 in GITAM/dpdk from DPDK-1795-ecpri to 18.11-qoriq-dev ef48961 - bus/dpaa: fix static initialization to 0 29235c3 - Merge pull request #948 in GITAM/dpdk from 18.11-qoriq-dev to 18.11-qoriq c3a19c0 - PVT: test: restructure PDCP test cases into new file 627fc3d - Merge pull request #943 in GITAM/dpdk from feature/DPDK-1595-ovs-2.11-migration to 18.11-qoriq-dev 7b9db20 - PVT: examples/l3fwd: add support for ethtype based traffic splitting
*Update to lsdk 1906 tag
include the following changes: abc80ef - netdev-dpdk: have env based configurable number of packet buffers 2410fad - rhel: Use PIDFile on forking systemd service files a631ef7 - dpif-netdev: fix meter at high packet rate. 6b14ed0 - dpif-netdev: Fix double parsing of packets when EMC disabled. 0118706 - dpif-netdev: Update comment about flow installation race. 3c54cf1 - netlink linux: fix to append the netnsid netlink attr. 9a72286 - conntrack: Fix minimum connections to clean. 9b2e4ff - datapath: Revert "datapath: Fix template leak in error cases." 62c7a9d - compiler: Fix compilation when using VStudio 2015/2017 2560908 - datapath: fix flow actions reallocation fae3311 - Docs: fix conntrack flow ct_state input daeab19 - netdev-dpdk: Fix printing masks with wrong byte order. 24909e6 - Prepare for 2.11.2. 022c173 - Set release date for 2.11.1. 044a61d - netdev-tc-offloads: Fix probe tc block support 434ffa9 - rhel: Include all header files in the Fedora's devel package 359a7d2 - reconnect.c: Don't transition back to ACTIVE when forced to RECONNECT. 2924b48 - OVN: Make periodic RAs consistent with RA responder. 54f2d11 - OVN: Always send prefix option in RAs 76f2c3b - OVN: Use offset instead of pointer into ofpbuf 6ffb917 - ofproto: fix the bug of bucket counter is not updated b1f9b4e - netdev-dpdk: Print netdev name for txq mapping. 9068a55 - dpif-netdev-perf: Fix millisecond stats precision with slower TSC. 3c440c2 - ifupdown.sh: Add missing "--may-exist" option b88dafd - dpif-netdev-perf: Fix double update of perf histograms. a55b827 - dpdk: Stop dumping memzones to stdout. 68b2304 - dpctl: Drop parser debug information. 5dea53a - netdev-tc-offloads: Properly get the block id on flow del/get 4c26dfd - netdev-tc-offloads: Improve log message for icmpv6 offload not supported d8bb89f - conntrack: Replace structure copy by memcpy().
*include the following changes: 79df057 - vfio/fsl-mc: Check command version on dprc and dpmcp cmdif 475d962 - vfio/fsl-mc: Add support for DPRC command version-2 271b7fa - vfio/fsl-mc: Add interface for device region range 23f6ed8 - vfio/fsl-mc: Setup address space for QBman CE-SH region 89add80 - vfio/fsl-mc: Fix mc-command version check
*update to lsdk 1906 tag
include the following changes: cd3c353 - updated the year in file copyright headers. 4d5cc8d - app/device_record: Stores the client's device record in a file. 322a6d6 - APP,TA,LIB: Creating 2 separate objects for public and private key for KeyPair bc0553a - removed dead code under compile time flag SK_INVALID_HANDLE f2fddd6 - added support for sk_digestinit sk_digestupdate sk_digestfinal. f509695 - APP_TA_LIB: Adding SK_ATTR_PRIVATE attribute for objects 6791873 - SOBJ_LIB: Exposing SK_CreateObject for PKCS#11 Library f3da9c0 - TA: Miscellaneous Changes 07ef069 - APP_ENG: Adding support for OAEP Encryption in engine and engine app 9fdbf7a - LIB_TA: Adding support for OAEP Encryption Scheme 1652924 - TA: Not returning TEE_ERROR_BAD_PARAMETERS for unmatched attributes 2a49505 - LIB: Exposing SK_GenerateKeyPair and SK_EraseObject Functions bbcbe6c - BUG: Fix for PLATSEC-666 e944456 - BUG: Fix for PLATSEC-628 41a0eaa - Making secure object components ompatible to OpenSSL 1.1.0 and 1.0.2 f72cf1a - Fix for PLATSEC-581
*update to lsdk 1906 tag
include the following changes: a243fb2 - corrected the start year ac5d4db - Updated the year in file header. d642bdd - Changes for making SK_GenerateKeyPair API compatible
include the following changes: 7b5819e - Restructure copy_name() to follow best practice 8bd9964 - Set algorithm mask only when type is provided 81bd24f - Don't overwrite compiler env. if alredy set 0248373 - Update dependencies - Linux kernel and iproute2 headers 547b464 - Fix potential segfault when omitting type argument 274eaba - Honor max length of alg/driver name in strncpy 9d0ae95 - Improve usage information 6ab1c3c - crconf: Fix error output on crconf_del_alg. d8a0a6f - crconf: Add a basic manpage 27a173f - crconf: Update headers to the Linux mainline 873058f - crconf - Linux crypto layer configuraton tool
*update to lsdk 1906 tag
include the following changes: 1b9e9ac - .gitmodules: use codeaurora url e451edf - perf-test: Reject large input files 3bdbdf3 - perf-test: Error out in stateless decomp mismatch 64fc3b1 - dce-internals: Efficient FQ full scheme d1add39 - Merge pull request #32 in DPAA2/dce from ~NXA07713/dce:jira/QSDK-5623-update-dce-license-files to master b3c27ed - LICENSE: Remove GPL License 68a6dab - Merge pull request #31 in DPAA2/dce from ~NXA07713/dce:jira/QSDK-5623-update-dce-license-files to master 0f33f16 - LICENSE: Update License year and full BSD-3 text 0175bb4 - Merge pull request #30 in DPAA2/dce from ~NXA07713/dce:jira/QSDK-5621-dpdcei-object-fails-to-unassign-out-of-child-dprc to master 361e4d0 - dpdcei: Cleanup hanldes in dpdcei_deactivate
*update to lsdk 1906 tag
include the following changes: be7dd8e - Increment SPC Tool version 2dafd3a - Refactoring following support for multiple protocols 727329a - Multiple protocols support
*update to lsdk 1906 tag
include the following changes: 18c7760 - Add DPSECI object in default DPL for LX2160 platform 598afb9 - Add DPL files for LS1088/LS2088 with DPSECI objects included 8da7d1d - api: add MC 10.16.0 APIs 19a70bb - api: add MC 10.15.0 APIs
*update to lsdk 1906 tag
include the following changes: 17f94e4 - plat/nxp/fsl_sec: Remove infinite loop from DEBUG Build 2d18325 - plat/nxp: Change MMU mapping of QSPI, NOR flash to MEMORY from DEVICE 5b8b9a3 - plat/nxp: Add lx2160aqds support 332c6ee - Merge branch 'LSDK' of ssh://bitbucket.sw.nxp.com/gitam/atf into LSDK 1e4b1e6 - Merge pull request #41 in GITAM/atf from ~NXA19332/atf-fork:LSDK to LSDK 693673c - plat: nxp: ls1028rdb: optimize DDR performance 8976e00 - Merge pull request #36 in GITAM/atf from ~NXA27427/atf:jira/PLATSEC-822 to LSDK af21b29 - common: include: correct setbits_le32 macro 3947e8e - plat/nxp/tools: Adds flag to check CRC&STOP or STOP command. c8af318 - nxp: ddr: ls1012a: fixes random hang issue fea3df8 - Merge pull request #30 in GITAM/atf from ~NXF49783/atf:LSDK to LSDK 1df8fb4 - Merge pull request #38 in GITAM/atf from ls1028_brup to LSDK 7e3f1fc - nxp: drivers: ddr: phy-gen2: add PllCtrl settings 07a2317 - nxp/ls1046: Add support for frwy board 25443b9 - plat: nxp: ddr: update ODT and output driver impedance b68a2d7 - csf_hdr_parser/input_blx_ch2: Corrects OEMUID 0 input attribute name. 5b14d6d - plat/nxp/drivers/sfp: Reset the data value at the set GPIO pin. 89963ec - Merge branch 'LSDK' of ssh://bitbucket.sw.nxp.com/gitam/atf into LSDK b54f200 - nxp/ls1028a: Enable core timebase
*update to lsdk 1906 tag
include the following changes: 1296390 - update to commit 7918c743857 ee84dcf - update binaries for LSDK 19.06 e4f1a6f - update binaries for LSDK 19.06
*update to lsdk 1906 tag
include the following changes: 93358b4 - rcw.py: Update comments for loadwochecksum command de68a41 - ls2088aqds: Corrected MEM2_PLL_RAT for DP-DDR frequency ac05bc4 - ls1046afrwy: fix a compiling error issue a25dde7 - Add init rcw support for ls1028a platform de0baa2 - rcw.py: Add Support to Load RCW without checksum b99ec29 - ls1046a: ls1046afrwy: frwy-ls1046a platform support 498de8d - LX2160: Add workaround of errata A-011270
Alban Bedel [Tue, 25 Jun 2019 09:44:57 +0000 (11:44 +0200)]
atf: Fix the build on machines that don't support UEFI
The uefi package silently build nothing on machines that are not
supported. On machines without UEFI support this lead to a build
failure in atf as it always try to build fip_uefi.bin. Fix the
compile step to only build fip_uefi.bin if there is an UEFI image.
Signed-off-by: Alban Bedel <alban.bedel@avionic-design.de>
Neena Busireddy [Fri, 22 Feb 2019 22:03:02 +0000 (16:03 -0600)]
xserver-xorg: Fix X server 1.20.1 crash [YOCIMX-3280]
One of our patches which we upstreamed has been removed from the poky thud.
http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/meta/recipes-graphics/xorg-xserver?h=thud&id=9d2cbf041403f5012b251de6c2dc0683755398ed
Add the patch back to fix the crash.
Signed-off-by: Neena Busireddy <neenareddy.busireddy@nxp.com>
Backport from https://source.codeaurora.org/external/imx/meta-fsl-bsp-release/commit/imx/meta-bsp/recipes-graphics/xorg-xserver?h=thud-4.19.35-1.0.0&id=7b9dbcf65929446ddbd5c8245617a624eb30b9ff
Limit the patch to machines which use xf86-video-imx-vivante. Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Andrey Zhizhikin [Tue, 18 Jun 2019 14:45:32 +0000 (14:45 +0000)]
mx8mm: use cortexa53-crypto tune for imx8m mini per default
Current implementation of i.MX8M Mini is utilizing generic aarch64 tune
(arch-arm64) provided from OE meta layer.
According to IMX8MMRM [1] this CPU contains Cortex-A53 cores, so this commit
sets a default tune to cortexa53-crypto.
Note, that according to [1] ARM Crypto extensions are supported, therefore it
is beneficial to set the tune to -crypto, as this also enables NEON and VFP
support in GCC (via -march and -mcpu feature modifiers, see [2]).
