Re: [PATCH i-g-t] tests/intel/xe_bo_shrink: Add a user-space shrinker test

[Matthew Auld <matthew.auld@intel.com>]

On 21/10/2026 13:33, Thomas Hellström wrote:
> Add a user-space counterpart of the xe_bo_shrink kunit test, now that
> user-space can mark buffer objects purgeable using the madvise IOCTL.
> 
> The test creates 64 MiB system buffer objects corresponding to twice
> the amount of free system memory, fills them with pseudo-random data
> through CPU mappings that are kept alive until readback, and verifies
> the contents after the overcommit has forced the shrinker to back up
> or purge them. As in the kunit test, the share of the memory that
> doesn't fit in free RAM and swap is marked purgeable (DONTNEED) and
> allocated first. Purgeable objects are only verified if found retained
> when marked WILLNEED again; purged objects are never accessed.
> 
> Subtests:
> - shrink: The kunit test equivalent.
> - purge: Ignore swap space when sizing the purgeable share, so that
>    the shrinker needs to purge rather than swap.
> - shrink-fault-inject: Enable TTM backup / restore fault injection
>    through debugfs, if available, making the shrinker back up only part
>    of the objects and making backup and restore of individual pages
>    fail. The number of injected faults is checked to be non-zero.
> - *-blit: Rather than restoring objects by CPU faults, restore them
>    through a GPU bind and read them back by a MEM_COPY blit to a dummy
>    object in a separate VM. On discrete devices, objects are created
>    with system and VRAM placements, prefetched to system memory on
>    creation and prefetched to VRAM before the copy, so that they are
>    restored by the migration to VRAM.

Do we use any compression here on igpu? IIRC that is still a pretty big 
test gap. We have the equivalent for VRAM via the CCS evict test, but I 
don't think we have anything for igpu + CCS via shrinker. Maybe the 
kunit does it, but IIUC that is never run in CI?

> - *-private: Create the objects private to the test VM rather than as
>    external objects.
> - *-fast: Allocate at most the free system memory, half of it
>    purgeable, capped to 8 GiB purgeable and 8 GiB non-purgeable, and
>    shrink using drop_caches rather than memory overcommit.
> 
> The non-fast subtests are slow and require a substantial amount of
> swap space for meaningful coverage, so block them in CI and run only
> the fast subtests.

Were there any issues with the test itself getting killed? I seem to 
vaguely remember something like that from the i915 shrinker test. Or is 
that not a concern here?