On Wed, 2022-01-05 at 14:58 +0000, Matthew Auld wrote:
Purging can happen during swapping out, or directly invoked with the madvise ioctl. In such cases this doesn't involve a ttm move, which skips umapping the object.
Fixes: cf3e3e86d779 ("drm/i915: Use ttm mmap handling for ttm bo's.") Signed-off-by: Matthew Auld matthew.auld@intel.com Cc: Thomas Hellström thomas.hellstrom@linux.intel.com
drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 2 ++ 1 file changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c index 8d61d4538a64..f148e7e48f86 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c @@ -399,6 +399,8 @@ int i915_ttm_purge(struct drm_i915_gem_object *obj) if (obj->mm.madv == __I915_MADV_PURGED) return 0; + ttm_bo_unmap_virtual(bo);
ret = ttm_bo_validate(bo, &place, &ctx); if (ret) return ret;
The swap notifier and the move code both call i915_ttm_move_notify() to achieve this before calling i915_ttm_purge. This ensures both cpu- and gpu ptes are torn down, and also when we extend to dynamic dma-buf exporting, makes sure dma-buf importers unbind.
So I suggest we make a i915_ttm_truncate wrapper function that calls i915_ttm_move_notify() and then ttm_bo_purge(), and use that as the truncate callback as well as from those places we call i915_ttm_purge without a prior call to i915_ttm_move_notify(), which seems to be the ones you've identified in patch 3 and 4,
/Thomas