drm/ttm: Really use a separate LRU list for swapped- and pinned objects
Get series with b4
b4 am [email protected]Versions newest first · linked by author and title
- v6 Latest2 patches 2 repliesIn review
- v52 patches 0 repliesSuperseded
- v42 patches 8 repliesSuperseded
Earlier versions · 3
| Version | Posted | Patches | Replies | Status |
|---|---|---|---|---|
| v3 | 2 patches | 0 replies | Superseded | |
| v2 | 2 patches | 0 replies | Superseded | |
| v1 | 2 patches | 3 replies | Superseded |
Changes in v6 from the cover letter
- Move a lockdep assert.
- Tabify a line.
Patches in v6 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/2 | drm/ttm: Move swapped objects off the manager's LRU listdrivers/ drivers/ drivers/ +12 more | +109 | 0 replies | |
| 2/2 | drm/ttm: Move pinned objects off LRU lists when pinningdrivers/ | +5 | 0 replies |
Reviews trailers per patch
All 2 patches have a Reviewed-by or Acked-by.
| Patch | Christian König |
|---|---|
| 1/2 | R-b |
| 2/2 | R-b |
Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.
Cover letter [PATCH v6 0/2] drm/ttm: Really use a separate LRU list for swapped- and pinned objects
Resources of swapped objects were never moved off their LRU list when swapped, and also resources of pinned objects might remain on their LRU list unless the driver moved them to the device pinned list after pinning.
Rename the device "pinned" list to "unevictable" and ensure that resources of objects that are pinned or swapped are moved to that list.
RFC: Should we instead of a device-wide unevictable list, introduce an unevictable priority so that all objects remain with their resource's respective manager?
Patch 1/2 deals with swapped objects and also handles the problem of moving objects back to their manager's LRU list when populating.
Patch 2/2 deals with pinned objects.
v2:
- Address review comments by Christian König.
- Make TTM KUNIT tests build.
v3:
- Fix a couple of NULL pointer dereferences (Intel CI).
v4:
- Fix even more NULL pointer dereferences (Intel CI). The bulk move handling appears pretty fragile...
v5:
- Fix resources created for readback of swapped bos ending up on the unevictable list (Intel CI)
v6:
- Move a lockdep assert.
- Tabify a line.
Thomas Hellström (2):
drm/ttm: Move swapped objects off the manager's LRU list drm/ttm: Move pinned objects off LRU lists when pinning
drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 2 +- drivers/gpu/drm/i915/gem/i915_gem_ttm_move.c | 2 +- drivers/gpu/drm/i915/gem/i915_gem_ttm_pm.c | 4 +- drivers/gpu/drm/ttm/tests/ttm_bo_test.c | 4 +- drivers/gpu/drm/ttm/tests/ttm_resource_test.c | 6 +- drivers/gpu/drm/ttm/ttm_bo.c | 67 +++++++++++++++++-- drivers/gpu/drm/ttm/ttm_bo_util.c | 6 +- drivers/gpu/drm/ttm/ttm_bo_vm.c | 2 +- drivers/gpu/drm/ttm/ttm_device.c | 4 +- drivers/gpu/drm/ttm/ttm_resource.c | 28 ++++++-- drivers/gpu/drm/ttm/ttm_tt.c | 3 + drivers/gpu/drm/xe/xe_bo.c | 4 +- include/drm/ttm/ttm_bo.h | 2 + include/drm/ttm/ttm_device.h | 5 +- include/drm/ttm/ttm_tt.h | 5 ++ 15 files changed, 114 insertions(+), 30 deletions(-)
Signature · 1 line
2.46.0