summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/xe/xe_vm.c
AgeCommit message (Collapse)Author
2024-03-28drm/xe: Move vma rebinding to the drm_exec locking loopThomas Hellström
Rebinding might allocate page-table bos, causing evictions. To support blocking locking during these evictions, perform the rebinding in the drm_exec locking loop. Also Reserve fence slots where actually needed rather than trying to predict how many fence slots will be needed over a complete wound-wait transaction. v2: - Remove a leftover call to xe_vm_rebind() (Matt Brost) - Add a helper function xe_vm_validate_rebind() (Matt Brost) v3: - Add comments and squash with previous patch (Matt Brost) Fixes: 24f947d58fe5 ("drm/xe: Use DRM GPUVM helpers for external- and evicted objects") Fixes: 29f424eb8702 ("drm/xe/exec: move fence reservation") Cc: Matthew Auld <matthew.auld@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240327091136.3271-5-thomas.hellstrom@linux.intel.com
2024-03-28drm/xe: Rework rebindingThomas Hellström
Instead of handling the vm's rebind fence separately, which is error prone if they are not strictly ordered, attach rebind fences as kernel fences to the vm's resv. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: <stable@vger.kernel.org> # v6.8+ Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240327091136.3271-3-thomas.hellstrom@linux.intel.com
2024-03-26drm/xe: Create a helper function to init job's user fenceNirmoy Das
Refactor xe_sync_entry_signal so it doesn't have to modify xe_sched_job struct instead create a new helper function to set user fence values for a job. v2: Move the sync type check to xe_sched_job_init_user_fence(Lucas) Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Matthew Auld <matthew.auld@intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Signed-off-by: Nirmoy Das <nirmoy.das@intel.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240321161142.4954-1-nirmoy.das@intel.com Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
2024-03-22drm/xe/devcoredump: Print errno if VM snapshot was not capturedJosé Roberto de Souza
My testing machine has only 8GB of RAM and while running piglit tests I can reach the OOM cache in xe_vm_snapshot_capture() snap allocaiton sometimes. So to differentiate the OOM from race between capture and UMDs unbinbind VMs here I'm adding a '[0].error: -12' to devcoredump. v2: - fix returned errno values Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Signed-off-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240307135229.41973-2-jose.souza@intel.com
2024-03-22drm/xe: Make devcoredump VM error state print consistentJosé Roberto de Souza
This makes VM error consistent with [x].length and [x].data. Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Signed-off-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240307135229.41973-1-jose.souza@intel.com
2024-03-21drm/xe: Add debug messages for MMU notifier and VMA invalidateMatthew Brost
Extra debug is useful when working on VM issues. Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Nirmoy Das <nirmoy.das@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240320194232.1910688-1-matthew.brost@intel.com
2024-03-19drm/xe/vm : Remove duplicate assignment of XE_VM_FLAG_LR_MODE flag.Himal Prasad Ghimiray
vm->flags are already assigned with passed flags. Remove the redundant assignment. Cc: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com> Reviewed: Matthew Brost <matthew.brost@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240307065213.1968688-1-himal.prasad.ghimiray@intel.com Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2024-03-19drm/xe/vm: fix xe_assert()Matthew Auld
The region can be used an index into the region_to_mem_type, so we should be asserting that it is less than the ARRAY_SIZE here. Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Nirmoy Das <nirmoy.das@intel.com> Reviewed-by: Nirmoy Das <nirmoy.das@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240318103616.26240-2-matthew.auld@intel.com
2024-03-13drm/xe: Invalidate userptr VMA on page pin faultMatthew Brost
Rather than return an error to the user or ban the VM when userptr VMA page pin fails with -EFAULT, invalidate VMA mappings. This supports the UMD use case of freeing userptr while still having bindings. Now that non-faulting VMs can invalidate VMAs, drop the usm prefix for the tile_invalidated member. v2: - Fix build error (CI) v3: - Don't invalidate VMA if in fault mode, rather kill VM (Thomas) - Update commit message with tile_invalidated name chagne (Thomas) - Wait VM bookkeep slots with VM resv lock (Thomas) v4: - Move list_del_init(&userptr.repin_link) after error check (Thomas) - Assert not in fault mode (Matthew) Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240312183907.933835-1-matthew.brost@intel.com
2024-03-04drm/xe: Add infrastructure for delayed LRC captureMaarten Lankhorst
Add a xe_guc_exec_queue_snapshot_capture_delayed and xe_lrc_snapshot_capture_delayed function to capture the contents of LRC in the next patch. Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240227131248.92910-2-maarten.lankhorst@linux.intel.com
2024-02-28drm/xe: Deny unbinds if uapi ufence pendingMika Kuoppala
If user fence was provided for MAP in vm_bind_ioctl and it has still not been signalled, deny UNMAP of said vma with EBUSY as long as unsignalled fence exists. This guarantees that MAP vs UNMAP sequences won't escape under the radar if we ever want to track the client's state wrt to completed and accessible MAPs. By means of intercepting the ufence release signalling. v2: find ufence with num_fences > 1 (Matt) v3: careful on clearing vma ufence (Matt) Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/1159 Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240215181152.450082-3-mika.kuoppala@linux.intel.com
2024-02-26drm/xe: get rid of MAX_BINDSPaulo Zanoni
Mesa has been issuing a single bind operation per ioctl since xe.ko changed to GPUVA due xe.ko bug #746. If I change Mesa to try again to issue every single bind operation it can in the same ioctl, it hits the MAX_BINDS assertion when running Vulkan conformance tests. Test dEQP-VK.sparse_resources.transfer_queue.3d.rgba32i.1024_128_8 issues 960 bind operations in a single ioctl, it's the most I could find in the conformance suite. I don't see a reason to keep the MAX_BINDS restriction: it doesn't seem to be preventing any specific issue. If the number is too big for the memory allocations, then those will fail. Nothing related to num_binds seems to be using the stack. Let's just get rid of it. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Testcase: dEQP-VK.sparse_resources.transfer_queue.3d.rgba32i.1024_128_8 References: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/746 Cc: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240215005353.1295420-1-paulo.r.zanoni@intel.com
2024-02-26drm/xe: Use vmalloc for array of bind allocation in bind IOCTLMatthew Brost
Use vmalloc in effort to allow a user pass in a large number of binds in an IOCTL (mesa use case). Also use array allocations rather open coding the size calculation. v2: Use __GFP_ACCOUNT for allocations (Thomas) Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240226155554.103384-1-matthew.brost@intel.com
2024-02-23drm/xe/uapi: Remove unused flagsFrancois Dugast
Those cases missed in previous uAPI cleanups were mostly accidentally brought in from i915 or created to exercise the possibilities of gpuvm but they are not used by userspace yet, so let's remove them. They can still be brought back later if needed. v2: - Fix XE_VM_FLAG_FAULT_MODE support in xe_lrc.c (Brian Welty) - Leave DRM_XE_VM_BIND_OP_UNMAP_ALL (José Roberto de Souza) - Ensure invalid flag values are rejected (Rodrigo Vivi) v3: Rebase after removal of persistent exec_queues (Francois Dugast) v4: Rodrigo: Rebase after the new dumpable flag. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Francois Dugast <francois.dugast@intel.com> Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240222232356.175431-1-rodrigo.vivi@intel.com
2024-02-21drm/xe: Do not include current dir for generated/xe_wa_oob.hDafna Hirschfeld
The generated file 'generated/xe_wa_oob.h' is included using: "generated/xe_wa_oob.h" which first look inside the source code. But the file resides in the build directory and should therefore be included using: <generated/xe_wa_oob.h> Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240221083622.1584492-1-dhirschfeld@habana.ai
2024-02-21drm/xe: Add debug prints for skipping rebindsMatthew Brost
Will help debug issues with VM binds. Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240221032743.3698849-1-matthew.brost@intel.com
2024-02-21drm/xe: Implement VM snapshot support for BO's and userptrMaarten Lankhorst
Since we cannot immediately capture the BO's and userptr, perform it in 2 stages. The immediate stage takes a reference to each BO and userptr, while a delayed worker captures the contents and then frees the reference. This is required because in signaling context, no locks can be taken, no memory can be allocated, and no waits on userspace can be performed. With the delayed worker, all of this can be performed very easily, without having to resort to hacks. Changes since v1: - Fix crash on NULL captured vm. - Use ascii85_encode to capture BO contents and save some space. - Add length to coredump output for each captured area. Changes since v2: - Dump each mapping on their own line, to simplify tooling. - Fix null pointer deref in xe_vm_snapshot_free. Changes since v3: - Don't add uninitialized value to snap->ofs. (Souza) - Use kernel types for u32 and u64. - Move snap_mutex destruction to final vm destruction. (Souza) Changes since v4: - Remove extra memset. (Souza) Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240221133024.898315-6-maarten.lankhorst@linux.intel.com
2024-02-21drm/xe: Add vm snapshot mutex for easily taking a vm snapshot during devcoredumpMaarten Lankhorst
The devcoredump is done in fence signaling context. Because of this, we cannot take any of the normal mutexes or we would invert. Normal: Take vm->lock, dma_fence_wait() Devcoredump: from dma_fence_wait() context, take vm->lock. This doesn't work, and we only care about integrity, so take the locks around additions and removals of vma's. Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240221133024.898315-5-maarten.lankhorst@linux.intel.com
2024-02-21drm/xe: Annotate each dumpable vma as suchMaarten Lankhorst
In preparation for snapshot dumping, mark each dumpable VMA as such, so we can walk over the VM later and dump it. Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240221133024.898315-4-maarten.lankhorst@linux.intel.com
2024-02-21drm/xe: Add uapi for dumpable bosMaarten Lankhorst
Add the flag XE_VM_BIND_FLAG_DUMPABLE to notify devcoredump that this mapping should be dumped. This is not hooked up, but the uapi should be ready before merging. It's likely easier to dump the contents of the bo's at devcoredump readout time, so it's better if the bos will stay unmodified after a hang. The NEEDS_CPU_MAPPING flag is removed as requirement. Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240221133024.898315-3-maarten.lankhorst@linux.intel.com
2024-02-20drm/xe: Return 2MB page size for compact 64k PTEsMatthew Brost
Compact 64k PTEs are only intended to be used within a single VMA which covers the entire 2MB range of the compact 64k PTEs. Add XE_VMA_PTE_COMPACT VMA flag to indicate compact 64k PTEs are used and update xe_vma_max_pte_size to return at least 2MB if set. v2: Include missing changes Fixes: 8f33b4f054fc ("drm/xe: Avoid doing rebinds") Fixes: c47794bdd63d ("drm/xe: Set max pte size when skipping rebinds") Reported-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/758 Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240219211942.3633795-4-matthew.brost@intel.com
2024-02-20drm/xe: Add XE_VMA_PTE_64K VMA flagMatthew Brost
Add XE_VMA_PTE_64K VMA flag to ensure skipping rebinds does not cross 64k page boundaries. Fixes: 8f33b4f054fc ("drm/xe: Avoid doing rebinds") Fixes: c47794bdd63d ("drm/xe: Set max pte size when skipping rebinds") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240219211942.3633795-3-matthew.brost@intel.com
2024-02-20drm/xe: Fix xe_vma_set_pte_sizeMatthew Brost
xe_vma_set_pte_size had a return value and did not set the 4k VMA flag. Both of these were incorrect. Fix these. Fixes: c47794bdd63d ("drm/xe: Set max pte size when skipping rebinds") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240219211942.3633795-2-matthew.brost@intel.com
2024-02-08drm/xe/vm: Avoid reserving zero fencesThomas Hellström
The function xe_vm_prepare_vma was blindly accepting zero as the number of fences and forwarded that to drm_exec_prepare_obj. However, that leads to an out-of-bounds shift in the dma_resv_reserve_fences() and while one could argue that the dma_resv code should be robust against that, avoid attempting to reserve zero fences. Relevant stack trace: [773.183188] ------------[ cut here ]------------ [773.183199] UBSAN: shift-out-of-bounds in ../include/linux/log2.h:57:13 [773.183241] shift exponent 64 is too large for 64-bit type 'long unsigned int' [773.183254] CPU: 2 PID: 1816 Comm: xe_evict Tainted: G U 6.8.0-rc3-xe #1 [773.183256] Hardware name: ASUS System Product Name/PRIME Z690-P D4, BIOS 2014 10/14/2022 [773.183257] Call Trace: [773.183258] <TASK> [773.183260] dump_stack_lvl+0xaf/0xd0 [773.183266] dump_stack+0x10/0x20 [773.183283] ubsan_epilogue+0x9/0x40 [773.183286] __ubsan_handle_shift_out_of_bounds+0x10f/0x170 [773.183293] dma_resv_reserve_fences.cold+0x2b/0x48 [773.183295] ? ww_mutex_lock+0x3c/0x110 [773.183301] drm_exec_prepare_obj+0x45/0x60 [drm_exec] [773.183313] xe_vm_prepare_vma+0x33/0x70 [xe] [773.183375] xe_vma_destroy_unlocked+0x55/0xa0 [xe] [773.183427] xe_vm_close_and_put+0x526/0x940 [xe] Fixes: 2714d5093620 ("drm/xe: Convert pagefaulting code to use drm_exec") Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240208132115.3132-1-thomas.hellstrom@linux.intel.com
2024-02-06drm/xe: Remove TEST_VM_ASYNC_OPS_ERRORMatthew Brost
TEST_VM_ASYNC_OPS_ERROR is broken and unused. Remove for now and will pull back in a later time when it is used, fixed, and properly hidden behind a Kconfig option. Also fixup the supported flags value. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240206045010.2981051-1-matthew.brost@intel.com
2024-02-06drm/xe/vm: don't ignore error when in_kthreadMatthew Auld
If GUP fails and we are in_kthread, we can have pinned = 0 and ret = 0. If that happens we call sg_alloc_append_table_from_pages() with n_pages = 0, which is not well behaved and can trigger: kernel BUG at include/linux/scatterlist.h:115! depending on if the pages array happens to be zeroed or not. Even if we don't hit that it crashes later when trying to dma_map the returned table. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240202171435.427630-2-matthew.auld@intel.com
2024-02-05drm/xe: Assume large page size if VMA not yet boundMatthew Brost
The calculation to determine max page size of a VMA during a REMAP operations assumes the VMA has been bound. This assumption is not true if the VMA is from an eariler operation in an array of binds. If a VMA has not been bound use the maximum page size which will ensure the previous / next REMAP operations are not incorrectly skipped. Fixes: 8f33b4f054fc ("drm/xe: Avoid doing rebinds") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240205231714.2956225-1-matthew.brost@intel.com
2024-02-02drm/xe: Pick correct userptr VMA to repin on REMAP op failureMatthew Brost
A REMAP op is composed of 3 VMA's - unmap, prev map, and next map. When op_execute fails with -EAGAIN we need to update the local VMA pointer to the current op state and then repin the VMA if it is a userptr. Fixes a failure seen in xe_vm.munmap-style-unbind-userptr-one-partial. Fixes: b06d47be7c83 ("drm/xe: Port Xe to GPUVA") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240201004849.2219558-3-matthew.brost@intel.com
2024-02-02drm/xe: Take a reference in xe_exec_queue_last_fence_get()Matthew Brost
Take a reference in xe_exec_queue_last_fence_get(). Also fix a reference counting underflow bug VM bind and unbind. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240201004849.2219558-2-matthew.brost@intel.com
2024-02-01drm/xe: Fix loop in vm_bind_ioctl_ops_unwindMatthew Brost
The logic for the unwind loop is incorrect resulting in an infinite loop. Fix to unwind to go from the last operations list to he first. Fixes: 617eebb9c480 ("drm/xe: Fix array of binds") Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240201175532.2303168-1-matthew.brost@intel.com
2024-02-01drm/xe/vm: Subclass userptr vmasThomas Hellström
The construct allocating only parts of the vma structure when the userptr part is not needed is very fragile. A developer could add additional fields below the userptr part, and the code could easily attempt to access the userptr part even if its not persent. So introduce xe_userptr_vma which subclasses struct xe_vma the proper way, and accordingly modify a couple of interfaces. This should also help if adding userptr helpers to drm_gpuvm. v2: - Fix documentation of to_userptr_vma() (Matthew Brost) - Fix allocation and freeing of vmas to clearer distinguish between the types. Closes: https://lore.kernel.org/intel-xe/0c4cc1a7-f409-4597-b110-81f9e45d1ffe@embeddedor.com/T/#u Fixes: a4cc60a55fd9 ("drm/xe: Only alloc userptr part of xe_vma for userptrs") Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240131091628.12318-1-thomas.hellstrom@linux.intel.com
2024-01-30drm/xe: Only allow 1 ufence per exec / bind IOCTLMatthew Brost
The way exec ufences are coded only 1 ufence per IOCTL will be signaled. It is possible to fix this but for current use cases 1 ufence per IOCTL is sufficient. Enforce a limit of 1 ufence per IOCTL (both exec and bind to be uniform). v2: - Add fixes tag (Thomas) Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Cc: Mika Kahola <mika.kahola@intel.com> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Brian Welty <brian.welty@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240124234413.1640825-1-matthew.brost@intel.com
2024-01-22Merge drm/drm-next into drm-xe-nextLucas De Marchi
Sync to v6.8-rc1. Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
2024-01-22drm/xe/vm: bugfix in xe_vm_create_ioctlMoti Haimovski
Fix xe_vm_create_ioctl routine not freeing the vm-id allocated to it when the function fails. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Moti Haimovski <mhaimovski@habana.ai> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Tomer Tayar <ttayar@habana.ai> Signed-off-by: Matt Roper <matthew.d.roper@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240122102424.4008095-1-mhaimovski@habana.ai
2024-01-19drm/xe: make gpuvm_ops constJani Nikula
Place the function pointers in rodata. Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240117122044.1544174-4-jani.nikula@intel.com Signed-off-by: Jani Nikula <jani.nikula@intel.com>
2024-01-15drm/xe: unlock on error path in xe_vm_add_compute_exec_queue()Dan Carpenter
Drop the "&vm->lock" before returning. Fixes: 24f947d58fe5 ("drm/xe: Use DRM GPUVM helpers for external- and evicted objects") Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> (cherry picked from commit cf46019e8550a810cc023af7aa020ba43103b44d) Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
2024-01-15drm/xe/vm: Fix an error pathThomas Hellström
If using the VM_BIND_OP_UNMAP_ALL without any bound vmas for the vm, we will end up dereferencing an uninitialized variable and leak a bo lock. Fix this. v2: - Updated commit message (Lucas De Marchi) Reported-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Closes: https://lore.kernel.org/intel-xe/jrwua7ckbiozfcaodx4gg2h4taiuxs53j5zlpf3qzvyhyiyl2d@pbs3plurokrj/ Suggested-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Fixes: b06d47be7c83 ("drm/xe: Port Xe to GPUVA") Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Acked-by: Lucas De Marchi <lucas.demarchi@intel.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231222175904.16732-1-thomas.hellstrom@linux.intel.com (cherry picked from commit 9d0c1c5618be02c5acda7e6bbb728007b0632984) Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
2024-01-10drm/xe: unlock on error path in xe_vm_add_compute_exec_queue()Dan Carpenter
Drop the "&vm->lock" before returning. Fixes: 24f947d58fe5 ("drm/xe: Use DRM GPUVM helpers for external- and evicted objects") Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com>
2024-01-09drm/xe/vm: Fix an error pathThomas Hellström
If using the VM_BIND_OP_UNMAP_ALL without any bound vmas for the vm, we will end up dereferencing an uninitialized variable and leak a bo lock. Fix this. v2: - Updated commit message (Lucas De Marchi) Reported-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Closes: https://lore.kernel.org/intel-xe/jrwua7ckbiozfcaodx4gg2h4taiuxs53j5zlpf3qzvyhyiyl2d@pbs3plurokrj/ Suggested-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Fixes: b06d47be7c83 ("drm/xe: Port Xe to GPUVA") Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Acked-by: Lucas De Marchi <lucas.demarchi@intel.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231222175904.16732-1-thomas.hellstrom@linux.intel.com
2023-12-26drm/xe: Fix UBSAN splat in add_preempt_fences()Matthew Brost
add_preempt_fences() calls dma_resv_reserve_fences() with num_fences == 0 resulting in the below UBSAN splat. Short circuit add_preempt_fences() if num_fences == 0. [ 58.652241] ================================================================================ [ 58.660736] UBSAN: shift-out-of-bounds in ./include/linux/log2.h:57:13 [ 58.667281] shift exponent 64 is too large for 64-bit type 'long unsigned int' [ 58.674539] CPU: 2 PID: 1170 Comm: xe_gpgpu_fill Not tainted 6.6.0-rc3-guc+ #630 [ 58.674545] Hardware name: Intel Corporation Tiger Lake Client Platform/TigerLake U DDR4 SODIMM RVP, BIOS TGLSFWI1.R00.3243.A01.2006102133 06/10/2020 [ 58.674547] Call Trace: [ 58.674548] <TASK> [ 58.674550] dump_stack_lvl+0x92/0xb0 [ 58.674555] __ubsan_handle_shift_out_of_bounds+0x15a/0x300 [ 58.674559] ? rcu_is_watching+0x12/0x60 [ 58.674564] ? software_resume+0x141/0x210 [ 58.674575] ? new_vma+0x44b/0x600 [xe] [ 58.674606] dma_resv_reserve_fences.cold+0x40/0x66 [ 58.674612] new_vma+0x4b3/0x600 [xe] [ 58.674638] xe_vm_bind_ioctl+0xffd/0x1e00 [xe] [ 58.674663] ? __pfx_xe_vm_bind_ioctl+0x10/0x10 [xe] [ 58.674680] drm_ioctl_kernel+0xc1/0x170 [ 58.674686] ? __pfx_xe_vm_bind_ioctl+0x10/0x10 [xe] [ 58.674703] drm_ioctl+0x247/0x4c0 [ 58.674709] ? find_held_lock+0x2b/0x80 [ 58.674716] __x64_sys_ioctl+0x8c/0xb0 [ 58.674720] do_syscall_64+0x3c/0x90 [ 58.674723] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 58.674727] RIP: 0033:0x7fce4bd1aaff [ 58.674730] Code: 00 48 89 44 24 18 31 c0 48 8d 44 24 60 c7 04 24 10 00 00 00 48 89 44 24 08 48 8d 44 24 20 48 89 44 24 10 b8 10 00 00 00 0f 05 <41> 89 c0 3d 00 f0 ff ff 77 1f 48 8b 44 24 18 64 48 2b 04 25 28 00 [ 58.674731] RSP: 002b:00007ffc57434050 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 [ 58.674734] RAX: ffffffffffffffda RBX: 00007ffc574340e0 RCX: 00007fce4bd1aaff [ 58.674736] RDX: 00007ffc574340e0 RSI: 0000000040886445 RDI: 0000000000000003 [ 58.674737] RBP: 0000000040886445 R08: 0000000000000002 R09: 00007ffc574341b0 [ 58.674739] R10: 000055de43eb3780 R11: 0000000000000246 R12: 00007ffc574340e0 [ 58.674740] R13: 0000000000000003 R14: 00007ffc574341b0 R15: 0000000000000001 [ 58.674747] </TASK> [ 58.674748] ================================================================================ Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Cc: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://lore.kernel.org/r/20231215230203.719244-1-matthew.brost@intel.com Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-22Merge tag 'drm-xe-next-2023-12-21-pr1-1' of ↵Dave Airlie
https://gitlab.freedesktop.org/drm/xe/kernel into drm-next Introduce a new DRM driver for Intel GPUs Xe, is a new driver for Intel GPUs that supports both integrated and discrete platforms. The experimental support starts with Tiger Lake. i915 will continue be the main production driver for the platforms up to Meteor Lake and Alchemist. Then the goal is to make this Intel Xe driver the primary driver for Lunar Lake and newer platforms. It uses most, if not all, of the key drm concepts, in special: TTM, drm-scheduler, drm-exec, drm-gpuvm/gpuva and others. Signed-off-by: Dave Airlie <airlied@redhat.com> [airlied: add an extra X86 check, fix a typo, fix drm_exec_init interface change]. From: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/ZYSwLgXZUZ57qGPQ@intel.com
2023-12-26drm/xe: Fix UBSAN splat in add_preempt_fences()Matthew Brost
add_preempt_fences() calls dma_resv_reserve_fences() with num_fences == 0 resulting in the below UBSAN splat. Short circuit add_preempt_fences() if num_fences == 0. [ 58.652241] ================================================================================ [ 58.660736] UBSAN: shift-out-of-bounds in ./include/linux/log2.h:57:13 [ 58.667281] shift exponent 64 is too large for 64-bit type 'long unsigned int' [ 58.674539] CPU: 2 PID: 1170 Comm: xe_gpgpu_fill Not tainted 6.6.0-rc3-guc+ #630 [ 58.674545] Hardware name: Intel Corporation Tiger Lake Client Platform/TigerLake U DDR4 SODIMM RVP, BIOS TGLSFWI1.R00.3243.A01.2006102133 06/10/2020 [ 58.674547] Call Trace: [ 58.674548] <TASK> [ 58.674550] dump_stack_lvl+0x92/0xb0 [ 58.674555] __ubsan_handle_shift_out_of_bounds+0x15a/0x300 [ 58.674559] ? rcu_is_watching+0x12/0x60 [ 58.674564] ? software_resume+0x141/0x210 [ 58.674575] ? new_vma+0x44b/0x600 [xe] [ 58.674606] dma_resv_reserve_fences.cold+0x40/0x66 [ 58.674612] new_vma+0x4b3/0x600 [xe] [ 58.674638] xe_vm_bind_ioctl+0xffd/0x1e00 [xe] [ 58.674663] ? __pfx_xe_vm_bind_ioctl+0x10/0x10 [xe] [ 58.674680] drm_ioctl_kernel+0xc1/0x170 [ 58.674686] ? __pfx_xe_vm_bind_ioctl+0x10/0x10 [xe] [ 58.674703] drm_ioctl+0x247/0x4c0 [ 58.674709] ? find_held_lock+0x2b/0x80 [ 58.674716] __x64_sys_ioctl+0x8c/0xb0 [ 58.674720] do_syscall_64+0x3c/0x90 [ 58.674723] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 58.674727] RIP: 0033:0x7fce4bd1aaff [ 58.674730] Code: 00 48 89 44 24 18 31 c0 48 8d 44 24 60 c7 04 24 10 00 00 00 48 89 44 24 08 48 8d 44 24 20 48 89 44 24 10 b8 10 00 00 00 0f 05 <41> 89 c0 3d 00 f0 ff ff 77 1f 48 8b 44 24 18 64 48 2b 04 25 28 00 [ 58.674731] RSP: 002b:00007ffc57434050 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 [ 58.674734] RAX: ffffffffffffffda RBX: 00007ffc574340e0 RCX: 00007fce4bd1aaff [ 58.674736] RDX: 00007ffc574340e0 RSI: 0000000040886445 RDI: 0000000000000003 [ 58.674737] RBP: 0000000040886445 R08: 0000000000000002 R09: 00007ffc574341b0 [ 58.674739] R10: 000055de43eb3780 R11: 0000000000000246 R12: 00007ffc574340e0 [ 58.674740] R13: 0000000000000003 R14: 00007ffc574341b0 R15: 0000000000000001 [ 58.674747] </TASK> [ 58.674748] ================================================================================ Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Cc: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://lore.kernel.org/r/20231215230203.719244-1-matthew.brost@intel.com Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-21drm/xe/uapi: Remove sync bindsMatthew Brost
Remove concept of async vs sync VM bind queues, rather make all binds async. The following bits have dropped from the uAPI: DRM_XE_ENGINE_CLASS_VM_BIND_ASYNC DRM_XE_ENGINE_CLASS_VM_BIND_SYNC DRM_XE_VM_CREATE_FLAG_ASYNC_DEFAULT DRM_XE_VM_BIND_FLAG_ASYNC To implement sync binds the UMD is expected to use the out-fence interface. v2: Send correct version v3: Drop drm_xe_syncs Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com> Cc: Francois Dugast <francois.dugast@intel.com> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Acked-by: José Roberto de Souza <jose.souza@intel.com> Acked-by: Mateusz Naklicki <mateusz.naklicki@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-21drm/xe: Allow num_batch_buffer / num_binds == 0 in IOCTLsMatthew Brost
The idea being out-syncs can signal indicating all previous operations on the bind queue are complete. An example use case of this would be support for implementing vkQueueWaitIdle easily. All in-syncs are waited on before signaling out-syncs. This is implemented by forming a composite software fence of in-syncs and installing this fence in the out-syncs and exec queue last fence slot. The last fence must be added as a dependency for jobs on user exec queues as it is possible for the last fence to be a composite software fence (unordered, ioctl with zero bb or binds) rather than hardware fence (ordered, previous job on queue). Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-21drm/xe: Use a flags field instead of bools for sync parseMatthew Brost
Use a flags field instead of severval bools for sync parse as it is easier to read and less bug prone. v2: Pull in header change from subsequent patch Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-21drm/xe: Use a flags field instead of bools for VMA createMatthew Brost
Use a flags field instead of severval bools for VMA create as it is easier to read and less bug prone. Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-21drm/xe: Use DRM_GPUVM_RESV_PROTECTED for gpuvmThomas Hellström
Use DRM_GPUVM_RESV_PROTECTED to use corse-grained locking for the evict and external object list. Since we are already holding the relevant RESV locks, for now at least, we don't need the fine-grained locking. Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231212100144.6833-3-thomas.hellstrom@linux.intel.com Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-21drm/xe: Use DRM GPUVM helpers for external- and evicted objectsThomas Hellström
Adapt to the DRM_GPUVM helpers moving removing a lot of complicated driver-specific code. For now this uses fine-grained locking for the evict list and external object list, which may incur a slight performance penalty in some situations. v2: - Don't lock all bos and validate on LR exec submissions (Matthew Brost) - Add some kerneldoc Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Acked-by: Matthew Brost <matthew.brost@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231212100144.6833-2-thomas.hellstrom@linux.intel.com Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-21drm/xe: Use NULL PTEs as scratch PTEsThomas Hellström
Currently scratch PTEs are write-enabled and points to a single scratch page. This has the side effect that buggy applications with out-of-bounds memory accesses may not notice the bad access since what's written may be read back. Instead use NULL PTEs as scratch PTEs. These always return 0 when reading, and writing has no effect. As a slight benefit, we can also use huge NULL PTEs. One drawback pointed out is that debugging may be hampered since previously when inspecting the content of the scratch page, it might be possible to detect writes to out-of-bound addresses and possibly also from where the out-of-bounds address originated. However since the scratch page-table structure is kept, it will be easy to add back the single RW-enabled scratch page under a debug define if needed. Also update the kerneldoc accordingly and move the function to create the scratch page-tables from xe_pt.c to xe_pt.h since it is accessing vm structure internals and this also makes it possible to make it static. v2: - Don't try to encode scratch PTEs larger than 1GiB. - Move xe_pt_create_scratch(), Update kerneldoc. v3: - Rebase. Cc: Brian Welty <brian.welty@intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Acked-by: Lucas De Marchi <lucas.demarchi@intel.com> #for general direction. Reviewed-by: Brian Welty <brian.welty@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231209151843.7903-3-thomas.hellstrom@linux.intel.com Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
2023-12-21drm/xe: Restrict huge PTEs to 1GiBThomas Hellström
Add a define for the highest level for which we can encode a huge PTE, and use it for page-table building. Also update an assert that checks that we don't try to encode for larger sizes. Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Brian Welty <brian.welty@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231209151843.7903-2-thomas.hellstrom@linux.intel.com Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
generated by cgit (git 2.34.1) at 2025-04-26 20:09:53 +0000