diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2023-06-29 11:00:17 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2023-06-29 11:00:17 -0700 |
commit | 1b722407a13b7f8658d2e26917791f32805980a2 (patch) | |
tree | 30aab582725a46e42843d75e2eb9ce4151f0f3ed /Documentation/gpu/rfc/xe.rst | |
parent | f8824e151fbfa0ac0a258015d606ea6f4a10251b (diff) | |
parent | 5ff2977b19769fd24b0cfbe7cbe4d5114b6106af (diff) | |
download | linux-stable-1b722407a13b7f8658d2e26917791f32805980a2.tar.gz linux-stable-1b722407a13b7f8658d2e26917791f32805980a2.tar.bz2 linux-stable-1b722407a13b7f8658d2e26917791f32805980a2.zip |
Merge tag 'drm-next-2023-06-29' of git://anongit.freedesktop.org/drm/drm
Pull drm updates from Dave Airlie:
"There is one set of patches to misc for a i915 gsc/mei proxy driver.
Otherwise it's mostly amdgpu/i915/msm, lots of hw enablement and lots
of refactoring.
core:
- replace strlcpy with strscpy
- EDID changes to support further conversion to struct drm_edid
- Move i915 DSC parameter code to common DRM helpers
- Add Colorspace functionality
aperture:
- ignore framebuffers with non-primary devices
fbdev:
- use fbdev i/o helpers
- add Kconfig options for fb_ops helpers
- use new fb io helpers directly in drivers
sysfs:
- export DRM connector ID
scheduler:
- Avoid an infinite loop
ttm:
- store function table in .rodata
- Add query for TTM mem limit
- Add NUMA awareness to pools
- Export ttm_pool_fini()
bridge:
- fsl-ldb: support i.MX6SX
- lt9211, lt9611: remove blanking packets
- tc358768: implement input bus formats, devm cleanups
- ti-snd65dsi86: implement wait_hpd_asserted
- analogix: fix endless probe loop
- samsung-dsim: support swapped clock, fix enabling, support var
clock
- display-connector: Add support for external power supply
- imx: Fix module linking
- tc358762: Support reset GPIO
panel:
- nt36523: Support Lenovo J606F
- st7703: Support Anbernic RG353V-V2
- InnoLux G070ACE-L01 support
- boe-tv101wum-nl6: Improve initialization
- sharp-ls043t1le001: Mode fixes
- simple: BOE EV121WXM-N10-1850, S6D7AA0
- Ampire AM-800480L1TMQW-T00H
- Rocktech RK043FN48H
- Starry himax83102-j02
- Starry ili9882t
amdgpu:
- add new ctx query flag to handle reset better
- add new query/set shadow buffer for rdna3
- DCN 3.2/3.1.x/3.0.x updates
- Enable DC_FP on loongarch
- PCIe fix for RDNA2
- improve DC FAMS/SubVP support for better power management
- partition support for lots of engines
- Take NUMA into account when allocating memory
- Add new DRM_AMDGPU_WERROR config parameter to help with CI
- Initial SMU13 overdrive support
- Add support for new colorspace KMS API
- W=1 fixes
amdkfd:
- Query TTM mem limit rather than hardcoding it
- GC 9.4.3 partition support
- Handle NUMA for partitions
- Add debugger interface for enabling gdb
- Add KFD event age tracking
radeon:
- Fix possible UAF
i915:
- new getparam for PXP support
- GSC/MEI proxy driver
- Meteorlake display enablement
- avoid clearing preallocated framebuffers with TTM
- implement framebuffer mmap support
- Disable sampler indirect state in bindless heap
- Enable fdinfo for GuC backends
- GuC loading and firmware table handling fixes
- Various refactors for multi-tile enablement
- Define MOCS and PAT tables for MTL
- GSC/MEI support for Meteorlake
- PMU multi-tile support
- Large driver kernel doc cleanup
- Allow VRR toggling and arbitrary refresh rates
- Support async flips on linear buffers on display ver 12+
- Expose CRTC CTM property on ILK/SNB/VLV
- New debugfs for display clock frequencies
- Hotplug refactoring
- Display refactoring
- I915_GEM_CREATE_EXT_SET_PAT for Mesa on Meteorlake
- Use large rings for compute contexts
- HuC loading for MTL
- Allow user to set cache at BO creation
- MTL powermanagement enhancements
- Switch to dedicated workqueues to stop using flush_scheduled_work()
- Move display runtime init under display/
- Remove 10bit gamma on desktop gen3 parts, they don't support it
habanalabs:
- uapi: return 0 for user queries if there was a h/w or f/w error
- Add pci health check when we lose connection with the firmware.
This can be used to distinguish between pci link down and firmware
getting stuck.
- Add more info to the error print when TPC interrupt occur.
- Firmware fixes
msm:
- Adreno A660 bindings
- SM8350 MDSS bindings fix
- Added support for DPU on sm6350 and sm6375 platforms
- Implemented tearcheck support to support vsync on SM150 and newer
platforms
- Enabled missing features (DSPP, DSC, split display) on sc8180x,
sc8280xp, sm8450
- Added support for DSI and 28nm DSI PHY on MSM8226 platform
- Added support for DSI on sm6350 and sm6375 platforms
- Added support for display controller on MSM8226 platform
- A690 GPU support
- Move cmdstream dumping out of fence signaling path
- a610 support
- Support for a6xx devices without GMU
nouveau:
- NULL ptr before deref fixes
armada:
- implement fbdev emulation as client
sun4i:
- fix mipi-dsi dotclock
- release clocks
vc4:
- rgb range toggle property
- BT601 / BT2020 HDMI support
vkms:
- convert to drmm helpers
- add reflection and rotation support
- fix rgb565 conversion
gma500:
- fix iomem access
shmobile:
- support renesas soc platform
- enable fbdev
mxsfb:
- Add support for i.MX93 LCDIF
stm:
- dsi: Use devm_ helper
- ltdc: Fix potential invalid pointer deref
renesas:
- Group drivers in renesas subdirectory to prepare for new platform
- Drop deprecated R-Car H3 ES1.x support
meson:
- Add support for MIPI DSI displays
virtio:
- add sync object support
mediatek:
- Add display binding document for MT6795"
* tag 'drm-next-2023-06-29' of git://anongit.freedesktop.org/drm/drm: (1791 commits)
drm/i915: Fix a NULL vs IS_ERR() bug
drm/i915: make i915_drm_client_fdinfo() reference conditional again
drm/i915/huc: Fix missing error code in intel_huc_init()
drm/i915/gsc: take a wakeref for the proxy-init-completion check
drm/msm/a6xx: Add A610 speedbin support
drm/msm/a6xx: Add A619_holi speedbin support
drm/msm/a6xx: Use adreno_is_aXYZ macros in speedbin matching
drm/msm/a6xx: Use "else if" in GPU speedbin rev matching
drm/msm/a6xx: Fix some A619 tunables
drm/msm/a6xx: Add A610 support
drm/msm/a6xx: Add support for A619_holi
drm/msm/adreno: Disable has_cached_coherent in GMU wrapper configurations
drm/msm/a6xx: Introduce GMU wrapper support
drm/msm/a6xx: Move CX GMU power counter enablement to hw_init
drm/msm/a6xx: Extend and explain UBWC config
drm/msm/a6xx: Remove both GBIF and RBBM GBIF halt on hw init
drm/msm/a6xx: Add a helper for software-resetting the GPU
drm/msm/a6xx: Improve a6xx_bus_clear_pending_transactions()
drm/msm/a6xx: Move a6xx_bus_clear_pending_transactions to a6xx_gpu
drm/msm/a6xx: Move force keepalive vote removal to a6xx_gmu_force_off()
...
Diffstat (limited to 'Documentation/gpu/rfc/xe.rst')
-rw-r--r-- | Documentation/gpu/rfc/xe.rst | 235 |
1 files changed, 235 insertions, 0 deletions
diff --git a/Documentation/gpu/rfc/xe.rst b/Documentation/gpu/rfc/xe.rst new file mode 100644 index 000000000000..2516fe141db6 --- /dev/null +++ b/Documentation/gpu/rfc/xe.rst @@ -0,0 +1,235 @@ +========================== +Xe – Merge Acceptance Plan +========================== +Xe is a new driver for Intel GPUs that supports both integrated and +discrete platforms starting with Tiger Lake (first Intel Xe Architecture). + +This document aims to establish a merge plan for the Xe, by writing down clear +pre-merge goals, in order to avoid unnecessary delays. + +Xe – Overview +============= +The main motivation of Xe is to have a fresh base to work from that is +unencumbered by older platforms, whilst also taking the opportunity to +rearchitect our driver to increase sharing across the drm subsystem, both +leveraging and allowing us to contribute more towards other shared components +like TTM and drm/scheduler. + +This is also an opportunity to start from the beginning with a clean uAPI that is +extensible by design and already aligned with the modern userspace needs. For +this reason, the memory model is solely based on GPU Virtual Address space +bind/unbind (‘VM_BIND’) of GEM buffer objects (BOs) and execution only supporting +explicit synchronization. With persistent mapping across the execution, the +userspace does not need to provide a list of all required mappings during each +submission. + +The new driver leverages a lot from i915. As for display, the intent is to share +the display code with the i915 driver so that there is maximum reuse there. + +As for the power management area, the goal is to have a much-simplified support +for the system suspend states (S-states), PCI device suspend states (D-states), +GPU/Render suspend states (R-states) and frequency management. It should leverage +as much as possible all the existent PCI-subsystem infrastructure (pm and +runtime_pm) and underlying firmware components such PCODE and GuC for the power +states and frequency decisions. + +Repository: + +https://gitlab.freedesktop.org/drm/xe/kernel (branch drm-xe-next) + +Xe – Platforms +============== +Currently, Xe is already functional and has experimental support for multiple +platforms starting from Tiger Lake, with initial support in userspace implemented +in Mesa (for Iris and Anv, our OpenGL and Vulkan drivers), as well as in NEO +(for OpenCL and Level0). + +During a transition period, platforms will be supported by both Xe and i915. +However, the force_probe mechanism existent in both drivers will allow only one +official and by-default probe at a given time. + +For instance, in order to probe a DG2 which PCI ID is 0x5690 by Xe instead of +i915, the following set of parameters need to be used: + +``` +i915.force_probe=!5690 xe.force_probe=5690 +``` + +In both drivers, the ‘.require_force_probe’ protection forces the user to use the +force_probe parameter while the driver is under development. This protection is +only removed when the support for the platform and the uAPI are stable. Stability +which needs to be demonstrated by CI results. + +In order to avoid user space regressions, i915 will continue to support all the +current platforms that are already out of this protection. Xe support will be +forever experimental and dependent on the usage of force_probe for these +platforms. + +When the time comes for Xe, the protection will be lifted on Xe and kept in i915. + +Xe driver will be protected with both STAGING Kconfig and force_probe. Changes in +the uAPI are expected while the driver is behind these protections. STAGING will +be removed when the driver uAPI gets to a mature state where we can guarantee the +‘no regression’ rule. Then force_probe will be lifted only for future platforms +that will be productized with Xe driver, but not with i915. + +Xe – Pre-Merge Goals +==================== + +Drm_scheduler +------------- +Xe primarily uses Firmware based scheduling (GuC FW). However, it will use +drm_scheduler as the scheduler ‘frontend’ for userspace submission in order to +resolve syncobj and dma-buf implicit sync dependencies. However, drm_scheduler is +not yet prepared to handle the 1-to-1 relationship between drm_gpu_scheduler and +drm_sched_entity. + +Deeper changes to drm_scheduler should *not* be required to get Xe accepted, but +some consensus needs to be reached between Xe and other community drivers that +could also benefit from this work, for coupling FW based/assisted submission such +as the ARM’s new Mali GPU driver, and others. + +As a key measurable result, the patch series introducing Xe itself shall not +depend on any other patch touching drm_scheduler itself that was not yet merged +through drm-misc. This, by itself, already includes the reach of an agreement for +uniform 1 to 1 relationship implementation / usage across drivers. + +GPU VA +------ +Two main goals of Xe are meeting together here: + +1) Have an uAPI that aligns with modern UMD needs. + +2) Early upstream engagement. + +RedHat engineers working on Nouveau proposed a new DRM feature to handle keeping +track of GPU virtual address mappings. This is still not merged upstream, but +this aligns very well with our goals and with our VM_BIND. The engagement with +upstream and the port of Xe towards GPUVA is already ongoing. + +As a key measurable result, Xe needs to be aligned with the GPU VA and working in +our tree. Missing Nouveau patches should *not* block Xe and any needed GPUVA +related patch should be independent and present on dri-devel or acked by +maintainers to go along with the first Xe pull request towards drm-next. + +DRM_VM_BIND +----------- +Nouveau, and Xe are all implementing ‘VM_BIND’ and new ‘Exec’ uAPIs in order to +fulfill the needs of the modern uAPI. Xe merge should *not* be blocked on the +development of a common new drm_infrastructure. However, the Xe team needs to +engage with the community to explore the options of a common API. + +As a key measurable result, the DRM_VM_BIND needs to be documented in this file +below, or this entire block deleted if the consensus is for independent drivers +vm_bind ioctls. + +Although having a common DRM level IOCTL for VM_BIND is not a requirement to get +Xe merged, it is mandatory to enforce the overall locking scheme for all major +structs and list (so vm and vma). So, a consensus is needed, and possibly some +common helpers. If helpers are needed, they should be also documented in this +document. + +ASYNC VM_BIND +------------- +Although having a common DRM level IOCTL for VM_BIND is not a requirement to get +Xe merged, it is mandatory to have a consensus with other drivers and Mesa. +It needs to be clear how to handle async VM_BIND and interactions with userspace +memory fences. Ideally with helper support so people don't get it wrong in all +possible ways. + +As a key measurable result, the benefits of ASYNC VM_BIND and a discussion of +various flavors, error handling and a sample API should be documented here or in +a separate document pointed to by this document. + +Userptr integration and vm_bind +------------------------------- +Different drivers implement different ways of dealing with execution of userptr. +With multiple drivers currently introducing support to VM_BIND, the goal is to +aim for a DRM consensus on what’s the best way to have that support. To some +extent this is already getting addressed itself with the GPUVA where likely the +userptr will be a GPUVA with a NULL GEM call VM bind directly on the userptr. +However, there are more aspects around the rules for that and the usage of +mmu_notifiers, locking and other aspects. + +This task here has the goal of introducing a documentation of the basic rules. + +The documentation *needs* to first live in this document (API session below) and +then moved to another more specific document or at Xe level or at DRM level. + +Documentation should include: + + * The userptr part of the VM_BIND api. + + * Locking, including the page-faulting case. + + * O(1) complexity under VM_BIND. + +Some parts of userptr like mmu_notifiers should become GPUVA or DRM helpers when +the second driver supporting VM_BIND+userptr appears. Details to be defined when +the time comes. + +Long running compute: minimal data structure/scaffolding +-------------------------------------------------------- +The generic scheduler code needs to include the handling of endless compute +contexts, with the minimal scaffolding for preempt-ctx fences (probably on the +drm_sched_entity) and making sure drm_scheduler can cope with the lack of job +completion fence. + +The goal is to achieve a consensus ahead of Xe initial pull-request, ideally with +this minimal drm/scheduler work, if needed, merged to drm-misc in a way that any +drm driver, including Xe, could re-use and add their own individual needs on top +in a next stage. However, this should not block the initial merge. + +This is a non-blocker item since the driver without the support for the long +running compute enabled is not a showstopper. + +Display integration with i915 +----------------------------- +In order to share the display code with the i915 driver so that there is maximum +reuse, the i915/display/ code is built twice, once for i915.ko and then for +xe.ko. Currently, the i915/display code in Xe tree is polluted with many 'ifdefs' +depending on the build target. The goal is to refactor both Xe and i915/display +code simultaneously in order to get a clean result before they land upstream, so +that display can already be part of the initial pull request towards drm-next. + +However, display code should not gate the acceptance of Xe in upstream. Xe +patches will be refactored in a way that display code can be removed, if needed, +from the first pull request of Xe towards drm-next. The expectation is that when +both drivers are part of the drm-tip, the introduction of cleaner patches will be +easier and speed up. + +Drm_exec +-------- +Helper to make dma_resv locking for a big number of buffers is getting removed in +the drm_exec series proposed in https://patchwork.freedesktop.org/patch/524376/ +If that happens, Xe needs to change and incorporate the changes in the driver. +The goal is to engage with the Community to understand if the best approach is to +move that to the drivers that are using it or if we should keep the helpers in +place waiting for Xe to get merged. + +This item ties into the GPUVA, VM_BIND, and even long-running compute support. + +As a key measurable result, we need to have a community consensus documented in +this document and the Xe driver prepared for the changes, if necessary. + +Dev_coredump +------------ + +Xe needs to align with other drivers on the way that the error states are +dumped, avoiding a Xe only error_state solution. The goal is to use devcoredump +infrastructure to report error states, since it produces a standardized way +by exposing a virtual and temporary /sys/class/devcoredump device. + +As the key measurable result, Xe driver needs to provide GPU snapshots captured +at hang time through devcoredump, but without depending on any core modification +of devcoredump infrastructure itself. + +Later, when we are in-tree, the goal is to collaborate with devcoredump +infrastructure with overall possible improvements, like multiple file support +for better organization of the dumps, snapshot support, dmesg extra print, +and whatever may make sense and help the overall infrastructure. + +Xe – uAPI high level overview +============================= + +...Warning: To be done in follow up patches after/when/where the main consensus in various items are individually reached. |