If we can't fit the buffer in the CPU visible portion of lmem, then the kernel will be unable to migrate the pages on fault on small-bar systems. Note that this doesn't restrict where the pages are allocated, but should just ensure we don't SIGBUS, if we need to migrate the pages. Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/7382 Signed-off-by: Matthew Auld <matthew.auld@xxxxxxxxx> Cc: Andrzej Hajda <andrzej.hajda@xxxxxxxxx> Cc: Nirmoy Das <nirmoy.das@xxxxxxxxx> --- tests/i915/gem_mmap_offset.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/i915/gem_mmap_offset.c b/tests/i915/gem_mmap_offset.c index 0adeb0b5..b27920de 100644 --- a/tests/i915/gem_mmap_offset.c +++ b/tests/i915/gem_mmap_offset.c @@ -750,7 +750,7 @@ static void always_clear(int i915, const struct gem_memory_region *r, int timeou struct thread_clear arg = { .i915 = i915, .region = r->ci, - .max = r->size / 2 >> 12, /* in pages */ + .max = r->cpu_size / 2 >> 12, /* in pages */ .timeout = timeout, }; const int ncpus = sysconf(_SC_NPROCESSORS_ONLN); -- 2.38.1
