diff options
Diffstat (limited to 'drivers/gpu/drm/tests/drm_buddy_test.c')
| -rw-r--r-- | drivers/gpu/drm/tests/drm_buddy_test.c | 136 |
1 files changed, 126 insertions, 10 deletions
diff --git a/drivers/gpu/drm/tests/drm_buddy_test.c b/drivers/gpu/drm/tests/drm_buddy_test.c index dd8fb9f8341a..5f40b5343bd8 100644 --- a/drivers/gpu/drm/tests/drm_buddy_test.c +++ b/drivers/gpu/drm/tests/drm_buddy_test.c @@ -21,6 +21,110 @@ static inline u64 get_size(int order, u64 chunk_size) return (1 << order) * chunk_size; } +static void drm_test_buddy_fragmentation_performance(struct kunit *test) +{ + struct drm_buddy_block *block, *tmp; + int num_blocks, i, ret, count = 0; + LIST_HEAD(allocated_blocks); + unsigned long elapsed_ms; + LIST_HEAD(reverse_list); + LIST_HEAD(test_blocks); + LIST_HEAD(clear_list); + LIST_HEAD(dirty_list); + LIST_HEAD(free_list); + struct drm_buddy mm; + u64 mm_size = SZ_4G; + ktime_t start, end; + + /* + * Allocation under severe fragmentation + * + * Create severe fragmentation by allocating the entire 4 GiB address space + * as tiny 8 KiB blocks but forcing a 64 KiB alignment. The resulting pattern + * leaves many scattered holes. Split the allocations into two groups and + * return them with different flags to block coalescing, then repeatedly + * allocate and free 64 KiB blocks while timing the loop. This stresses how + * quickly the allocator can satisfy larger, aligned requests from a pool of + * highly fragmented space. + */ + KUNIT_ASSERT_FALSE_MSG(test, drm_buddy_init(&mm, mm_size, SZ_4K), + "buddy_init failed\n"); + + num_blocks = mm_size / SZ_64K; + + start = ktime_get(); + /* Allocate with maximum fragmentation - 8K blocks with 64K alignment */ + for (i = 0; i < num_blocks; i++) + KUNIT_ASSERT_FALSE_MSG(test, drm_buddy_alloc_blocks(&mm, 0, mm_size, SZ_8K, SZ_64K, + &allocated_blocks, 0), + "buddy_alloc hit an error size=%u\n", SZ_8K); + + list_for_each_entry_safe(block, tmp, &allocated_blocks, link) { + if (count % 4 == 0 || count % 4 == 3) + list_move_tail(&block->link, &clear_list); + else + list_move_tail(&block->link, &dirty_list); + count++; + } + + /* Free with different flags to ensure no coalescing */ + drm_buddy_free_list(&mm, &clear_list, DRM_BUDDY_CLEARED); + drm_buddy_free_list(&mm, &dirty_list, 0); + + for (i = 0; i < num_blocks; i++) + KUNIT_ASSERT_FALSE_MSG(test, drm_buddy_alloc_blocks(&mm, 0, mm_size, SZ_64K, SZ_64K, + &test_blocks, 0), + "buddy_alloc hit an error size=%u\n", SZ_64K); + drm_buddy_free_list(&mm, &test_blocks, 0); + + end = ktime_get(); + elapsed_ms = ktime_to_ms(ktime_sub(end, start)); + + kunit_info(test, "Fragmented allocation took %lu ms\n", elapsed_ms); + + drm_buddy_fini(&mm); + + /* + * Reverse free order under fragmentation + * + * Construct a fragmented 4 GiB space by allocating every 8 KiB block with + * 64 KiB alignment, creating a dense scatter of small regions. Half of the + * blocks are selectively freed to form sparse gaps, while the remaining + * allocations are preserved, reordered in reverse, and released back with + * the cleared flag. This models a pathological reverse-ordered free pattern + * and measures how quickly the allocator can merge and reclaim space when + * deallocation occurs in the opposite order of allocation, exposing the + * cost difference between a linear freelist scan and an ordered tree lookup. + */ + ret = drm_buddy_init(&mm, mm_size, SZ_4K); + KUNIT_ASSERT_EQ(test, ret, 0); + + start = ktime_get(); + /* Allocate maximum fragmentation */ + for (i = 0; i < num_blocks; i++) + KUNIT_ASSERT_FALSE_MSG(test, drm_buddy_alloc_blocks(&mm, 0, mm_size, SZ_8K, SZ_64K, + &allocated_blocks, 0), + "buddy_alloc hit an error size=%u\n", SZ_8K); + + list_for_each_entry_safe(block, tmp, &allocated_blocks, link) { + if (count % 2 == 0) + list_move_tail(&block->link, &free_list); + count++; + } + drm_buddy_free_list(&mm, &free_list, DRM_BUDDY_CLEARED); + + list_for_each_entry_safe_reverse(block, tmp, &allocated_blocks, link) + list_move(&block->link, &reverse_list); + drm_buddy_free_list(&mm, &reverse_list, DRM_BUDDY_CLEARED); + + end = ktime_get(); + elapsed_ms = ktime_to_ms(ktime_sub(end, start)); + + kunit_info(test, "Reverse-ordered free took %lu ms\n", elapsed_ms); + + drm_buddy_fini(&mm); +} + static void drm_test_buddy_alloc_range_bias(struct kunit *test) { u32 mm_size, size, ps, bias_size, bias_start, bias_end, bias_rem; @@ -261,7 +365,6 @@ static void drm_test_buddy_alloc_range_bias(struct kunit *test) static void drm_test_buddy_alloc_clear(struct kunit *test) { unsigned long n_pages, total, i = 0; - DRM_RND_STATE(prng, random_seed); const unsigned long ps = SZ_4K; struct drm_buddy_block *block; const int max_order = 12; @@ -385,17 +488,28 @@ static void drm_test_buddy_alloc_clear(struct kunit *test) drm_buddy_fini(&mm); /* - * Create a new mm with a non power-of-two size. Allocate a random size, free as - * cleared and then call fini. This will ensure the multi-root force merge during - * fini. + * Create a new mm with a non power-of-two size. Allocate a random size from each + * root, free as cleared and then call fini. This will ensure the multi-root + * force merge during fini. */ - mm_size = 12 * SZ_4K; - size = max(round_up(prandom_u32_state(&prng) % mm_size, ps), ps); + mm_size = (SZ_4K << max_order) + (SZ_4K << (max_order - 2)); + KUNIT_EXPECT_FALSE(test, drm_buddy_init(&mm, mm_size, ps)); - KUNIT_ASSERT_FALSE_MSG(test, drm_buddy_alloc_blocks(&mm, 0, mm_size, - size, ps, &allocated, - DRM_BUDDY_TOPDOWN_ALLOCATION), - "buddy_alloc hit an error size=%u\n", size); + KUNIT_EXPECT_EQ(test, mm.max_order, max_order); + KUNIT_ASSERT_FALSE_MSG(test, drm_buddy_alloc_blocks(&mm, 0, SZ_4K << max_order, + 4 * ps, ps, &allocated, + DRM_BUDDY_RANGE_ALLOCATION), + "buddy_alloc hit an error size=%lu\n", 4 * ps); + drm_buddy_free_list(&mm, &allocated, DRM_BUDDY_CLEARED); + KUNIT_ASSERT_FALSE_MSG(test, drm_buddy_alloc_blocks(&mm, 0, SZ_4K << max_order, + 2 * ps, ps, &allocated, + DRM_BUDDY_CLEAR_ALLOCATION), + "buddy_alloc hit an error size=%lu\n", 2 * ps); + drm_buddy_free_list(&mm, &allocated, DRM_BUDDY_CLEARED); + KUNIT_ASSERT_FALSE_MSG(test, drm_buddy_alloc_blocks(&mm, SZ_4K << max_order, mm_size, + ps, ps, &allocated, + DRM_BUDDY_RANGE_ALLOCATION), + "buddy_alloc hit an error size=%lu\n", ps); drm_buddy_free_list(&mm, &allocated, DRM_BUDDY_CLEARED); drm_buddy_fini(&mm); } @@ -762,6 +876,7 @@ static struct kunit_case drm_buddy_tests[] = { KUNIT_CASE(drm_test_buddy_alloc_contiguous), KUNIT_CASE(drm_test_buddy_alloc_clear), KUNIT_CASE(drm_test_buddy_alloc_range_bias), + KUNIT_CASE(drm_test_buddy_fragmentation_performance), {} }; @@ -774,4 +889,5 @@ static struct kunit_suite drm_buddy_test_suite = { kunit_test_suite(drm_buddy_test_suite); MODULE_AUTHOR("Intel Corporation"); +MODULE_DESCRIPTION("Kunit test for drm_buddy functions"); MODULE_LICENSE("GPL"); |