With multi-device we are much more likely to have multiple
drm-gpusvm ranges pointing to the same struct mm range.
To avoid calling into drm_pagemap_populate_mm(), which is always
very costly, introduce a much less costly drm_gpusvm function,
drm_gpusvm_scan_mm() to scan the current migration state.
The device fault-handler and prefetcher can use this function to
determine whether migration is really necessary.
There are a couple of performance improvements that can be done
for this function if it turns out to be too costly. Those are
documented in the code.
v3:
- New patch.
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> # For merging through drm-xe.
Link: https://patch.msgid.link/20251219113320.183860-21-thomas.hellstrom@linux.intel.com
Compute kernels often issue memory copies immediately after completion.
If the memory being copied is an SVM pointer that was faulted into the
device and then bound via userptr, it is undesirable to move that
memory. Worse, if userptr is mixed between system and device memory, the
bind operation may be rejected.
Xe already has the necessary plumbing to support userptr with mixed
mappings. This update modifies GPUSVM's get_pages to correctly locate
pages in such mixed mapping scenarios.
v2:
- Rebase (Thomas Hellström)
v3:
- Remove Fixes tag.
v4:
- Break out from series since the other patch was merged.
- Update patch subject, ensure dri-devel and Maarten are CC'd.
Cc: Maarten Lankhorst <maarten.lankhorst@intel.com>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Link: https://lore.kernel.org/r/20251015120320.176338-1-thomas.hellstrom@linux.intel.com
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
When userptr is used on SVM-enabled VMs, a non-NULL
hmm_range::dev_private_owner value might mean that
hmm_range_fault() attempts to return device private pages.
Either that will fail, or the userptr code will not know
how to handle those.
Use NULL for hmm_range::dev_private_owner to migrate
such pages to system. In order to do that, move the
struct drm_gpusvm::device_private_page_owner field to
struct drm_gpusvm_ctx::device_private_page_owner so that
it doesn't remain immutable over the drm_gpusvm lifetime.
v2:
- Don't conditionally compile xe_svm_devm_owner().
- Kerneldoc xe_svm_devm_owner().
Fixes: 9e97874148 ("drm/xe/userptr: replace xe_hmm with gpusvm")
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://lore.kernel.org/r/20250930122752.96034-1-thomas.hellstrom@linux.intel.com
Export get/unmap/free pages API. We also need to tweak the SVM init to
allow skipping much of the unneeded parts.
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://lore.kernel.org/r/20250828142430.615826-15-matthew.auld@intel.com
Pull the pages stuff from the svm range into its own substructure, with
the idea of having the main pages related routines, like get_pages(),
unmap_pages() and free_pages() all operating on some lower level
structures, which can then be re-used for stuff like userptr.
v2:
- Move seq into pages struct (Matt B)
v3:
- Small kernel-doc fixes
Suggested-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20250828142430.615826-13-matthew.auld@intel.com
Sync with drm-misc-next which is necessary for changes in gpuvm
and gpusvm that will be used in xe.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
The drm_gpusvm_for_each_notifier, drm_gpusvm_for_each_notifier_safe and
drm_gpusvm_for_each_range_safe macros are useful for locating notifiers
and ranges within a user-specified range. By making these macros public,
we enable broader access and utility for developers who need to leverage
them in their implementations.
v2 (Matthew Brost)
- drop inline __drm_gpusvm_range_find
- /s/notifier_iter_first/drm_gpusvm_notifier_find
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20250819162058.2777306-5-himal.prasad.ghimiray@intel.com
Rename this struct to the more generic name drm_pagemap_addr so it can be
used in a broader context, such as DMA mappings of CPU memory.
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://lore.kernel.org/r/20250805140028.599361-2-francois.dugast@intel.com
Signed-off-by: Francois Dugast <francois.dugast@intel.com>
The migration functionality and track-keeping of per-pagemap VRAM
mapped to the CPU mm is not per GPU_vm, but rather per pagemap.
This is also reflected by the functions not needing the drm_gpusvm
structures. So move to drm_pagemap.
With this, drm_gpusvm shouldn't really access the page zone-device-data
since its meaning is internal to drm_pagemap. Currently it's used to
reject mapping ranges backed by multiple drm_pagemap allocations.
For now, make the zone-device-data a void pointer.
Alter the interface of drm_gpusvm_migrate_to_devmem() to ensure we don't
pass a gpusvm pointer.
Rename CONFIG_DRM_XE_DEVMEM_MIRROR to CONFIG_DRM_XE_PAGEMAP.
Matt is listed as author of this commit since he wrote most of the code,
and it makes sense to retain his git authorship.
Thomas mostly moved the code around.
v3:
- Kerneldoc fixes (CI)
- Don't update documentation about how the drm_pagemap
migration should be interpreted until upcoming
patches where the functionality is implemented.
(Matt Brost)
v4:
- More kerneldoc fixes around timeslice_ms
(Himal Ghimiray, Matt Brost)
v6:
- Fix an uninitialized pagemap pointer (CI)
Co-developed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20250619134035.170086-2-thomas.hellstrom@linux.intel.com
The drm_gpusvm_find_vma_start() function is used to determine the starting
address of a CPU VMA within a specified user range. If the range does not
contain any VMA, the function returns ULONG_MAX.
v2
- Rename function as drm_gpusvm_find_vma_start() (Matthew Brost)
- mmget/mmput
v3
- s/mmget/mmget_not_zero/
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20250513040228.470682-13-himal.prasad.ghimiray@intel.com
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Add timeslicing support to GPU SVM which will guarantee the GPU a
minimum execution time on piece of physical memory before migration back
to CPU. Intended to implement strict migration policies which require
memory to be in a certain placement for correct execution.
Required for shared CPU and GPU atomics on certain devices.
Fixes: 99624bdff8 ("drm/gpusvm: Add support for GPU Shared Virtual Memory")
Cc: stable@vger.kernel.org
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Link: https://lore.kernel.org/r/20250512135500.1405019-4-matthew.brost@intel.com
Mixing GPU and CPU atomics does not work unless a strict migration
policy of GPU atomics must be device memory. Enforce a policy of must be
in VRAM with a retry loop of 3 attempts, if retry loop fails abort
fault.
Removing always_migrate_to_vram modparam as we now have real migration
policy.
v2:
- Only retry migration on atomics
- Drop alway migrate modparam
v3:
- Only set vram_only on DGFX (Himal)
- Bail on get_pages failure if vram_only and retry count exceeded (Himal)
- s/vram_only/devmem_only
- Update xe_svm_range_is_valid to accept devmem_only argument
v4:
- Fix logic bug get_pages failure
v5:
- Fix commit message (Himal)
- Mention removing always_migrate_to_vram in commit message (Lucas)
- Fix xe_svm_range_is_valid to check for devmem pages
- Bail on devmem_only && !migrate_devmem (Thomas)
v6:
- Add READ_ONCE barriers for opportunistic checks (Thomas)
- Pair READ_ONCE with WRITE_ONCE (Thomas)
v7:
- Adjust comments (Thomas)
Fixes: 2f118c9491 ("drm/xe: Add SVM VRAM migration")
Cc: stable@vger.kernel.org
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Acked-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://lore.kernel.org/r/20250512135500.1405019-3-matthew.brost@intel.com
This commit adds a new flag, devmem_only, to the drm_gpusvm structure. The
purpose of this flag is to ensure that the get_pages function allocates
memory exclusively from the device's memory. If the allocation from
device memory fails, the function will return an -EFAULT error.
Required for shared CPU and GPU atomics on certain devices.
v3:
- s/vram_only/devmem_only/
Fixes: 99624bdff8 ("drm/gpusvm: Add support for GPU Shared Virtual Memory")
Cc: stable@vger.kernel.org
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20250512135500.1405019-2-matthew.brost@intel.com
This patch introduces support for GPU Shared Virtual Memory (SVM) in the
Direct Rendering Manager (DRM) subsystem. SVM allows for seamless
sharing of memory between the CPU and GPU, enhancing performance and
flexibility in GPU computing tasks.
The patch adds the necessary infrastructure for SVM, including data
structures and functions for managing SVM ranges and notifiers. It also
provides mechanisms for allocating, deallocating, and migrating memory
regions between system RAM and GPU VRAM.
This is largely inspired by GPUVM.
v2:
- Take order into account in check pages
- Clear range->pages in get pages error
- Drop setting dirty or accessed bit in get pages (Vetter)
- Remove mmap assert for cpu faults
- Drop mmap write lock abuse (Vetter, Christian)
- Decouple zdd from range (Vetter, Oak)
- Add drm_gpusvm_range_evict, make it work with coherent pages
- Export drm_gpusvm_evict_to_sram, only use in BO evict path (Vetter)
- mmget/put in drm_gpusvm_evict_to_sram
- Drop range->vram_alloation variable
- Don't return in drm_gpusvm_evict_to_sram until all pages detached
- Don't warn on mixing sram and device pages
- Update kernel doc
- Add coherent page support to get pages
- Use DMA_FROM_DEVICE rather than DMA_BIDIRECTIONAL
- Add struct drm_gpusvm_vram and ops (Thomas)
- Update the range's seqno if the range is valid (Thomas)
- Remove the is_unmapped check before hmm_range_fault (Thomas)
- Use drm_pagemap (Thomas)
- Drop kfree_mapping (Thomas)
- dma mapp pages under notifier lock (Thomas)
- Remove ctx.prefault
- Remove ctx.mmap_locked
- Add ctx.check_pages
- s/vram/devmem (Thomas)
v3:
- Fix memory leak drm_gpusvm_range_get_pages
- Only migrate pages with same zdd on CPU fault
- Loop over al VMAs in drm_gpusvm_range_evict
- Make GPUSVM a drm level module
- GPL or MIT license
- Update main kernel doc (Thomas)
- Prefer foo() vs foo for functions in kernel doc (Thomas)
- Prefer functions over macros (Thomas)
- Use unsigned long vs u64 for addresses (Thomas)
- Use standard interval_tree (Thomas)
- s/drm_gpusvm_migration_put_page/drm_gpusvm_migration_unlock_put_page (Thomas)
- Drop err_out label in drm_gpusvm_range_find_or_insert (Thomas)
- Fix kernel doc in drm_gpusvm_range_free_pages (Thomas)
- Newlines between functions defs in header file (Thomas)
- Drop shall language in driver vfunc kernel doc (Thomas)
- Move some static inlines from head to C file (Thomas)
- Don't allocate pages under page lock in drm_gpusvm_migrate_populate_ram_pfn (Thomas)
- Change check_pages to a threshold
v4:
- Fix NULL ptr deref in drm_gpusvm_migrate_populate_ram_pfn (Thomas, Himal)
- Fix check pages threshold
- Check for range being unmapped under notifier lock in get pages (Testing)
- Fix characters per line
- Drop WRITE_ONCE for zdd->devmem_allocation assignment (Thomas)
- Use completion for devmem_allocation->detached (Thomas)
- Make GPU SVM depend on ZONE_DEVICE (CI)
- Use hmm_range_fault for eviction (Thomas)
- Drop zdd worker (Thomas)
v5:
- Select Kconfig deps (CI)
- Set device to NULL in __drm_gpusvm_migrate_to_ram (Matt Auld, G.G.)
- Drop Thomas's SoB (Thomas)
- Add drm_gpusvm_range_start/end/size helpers (Thomas)
- Add drm_gpusvm_notifier_start/end/size helpers (Thomas)
- Absorb drm_pagemap name changes (Thomas)
- Fix driver lockdep assert (Thomas)
- Move driver lockdep assert to static function (Thomas)
- Assert mmap lock held in drm_gpusvm_migrate_to_devmem (Thomas)
- Do not retry forever on eviction (Thomas)
v6:
- Fix drm_gpusvm_get_devmem_page alignment (Checkpatch)
- Modify Kconfig (CI)
- Compile out lockdep asserts (CI)
v7:
- Add kernel doc for flags fields (CI, Auld)
Cc: Simona Vetter <simona.vetter@ffwll.ch>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: <dri-devel@lists.freedesktop.org>
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/20250306012657.3505757-7-matthew.brost@intel.com
