From: "Wang, X" <x.wang@intel.com>
To: Brian Nguyen <brian3.nguyen@intel.com>, <igt-dev@lists.freedesktop.org>
Subject: Re: [PATCH 2/4] tests/xe: Add random page reclaim subtest
Date: Mon, 13 Apr 2026 15:14:48 -0700 [thread overview]
Message-ID: <e64727e3-3eff-472b-bd4b-191ff9c61d88@intel.com> (raw)
In-Reply-To: <20260406184226.1294486-8-brian3.nguyen@intel.com>
On 4/6/2026 11:42, Brian Nguyen wrote:
> Add a subtest case providing a random distribution in both page sizes
> and page ordering to increase coverage of random pages. To ensure the
> page size chosen are properly used in PRL, filler pages are added in
> between the random chosen to provide proper alignment.
>
> Signed-off-by: Brian Nguyen <brian3.nguyen@intel.com>
> Suggested-by: Xin Wang <x.wang@intel.com>
> ---
> tests/intel/xe_page_reclaim.c | 151 ++++++++++++++++++++++++++++++++++
> 1 file changed, 151 insertions(+)
>
> diff --git a/tests/intel/xe_page_reclaim.c b/tests/intel/xe_page_reclaim.c
> index acc237d43..7741063d2 100644
> --- a/tests/intel/xe_page_reclaim.c
> +++ b/tests/intel/xe_page_reclaim.c
> @@ -164,6 +164,153 @@ static void test_vma_ranges_basic_mixed(int fd)
> compare_prl_stats(&stats_before, &stats_after, &expected_stats);
> }
>
> +/* Helper to calculate alignment filler pages needed */
> +struct alignment_fillers {
> + uint64_t gap_to_64k;
> + uint64_t remaining_gap;
> +};
> +
> +static struct alignment_fillers calculate_alignment_fillers(uint64_t start_addr,
> + uint64_t current_offset,
> + uint64_t page_size)
> +{
> + struct alignment_fillers fillers = {0};
> + uint64_t current_addr = start_addr + current_offset;
> + uint64_t misalignment = current_addr % page_size;
> + uint64_t gap, misalignment_64k;
> +
> + if (!misalignment)
> + return fillers;
> +
> + gap = page_size - misalignment;
> +
> + /*
> + * Fill the alignment gap using a two-level strategy to match the
> + * GPU page table hierarchy (4K → 64K → page_size):
> + * 1. Use 4K pages to reach the next 64K boundary (if the current
> + * address is not already 64K-aligned).
> + * 2. Use 64K pages for the remaining gap up to page_size alignment
> + * (this remainder is always a multiple of 64K).
> + */
> + misalignment_64k = current_addr % SZ_64K;
> + if (misalignment_64k) {
> + /* Not 64K aligned, fill with 4K pages up to next 64K boundary */
> + fillers.gap_to_64k = SZ_64K - misalignment_64k;
> + if (fillers.gap_to_64k > gap)
> + fillers.gap_to_64k = gap;
> + }
> +
> + fillers.remaining_gap = gap - fillers.gap_to_64k;
> + return fillers;
> +}
> +
> +/**
> + * SUBTEST: random
> + * Description: Create a random mix of page sizes (4K, 64K, 2M) with
> + * proper alignment handling. Larger pages are aligned by inserting
> + * filler pages (64K and 4K) as needed to test random page size
> + * combinations in page reclamation.
> + */
> +static void test_vma_range_random(int fd)
> +{
> + struct xe_prl_stats stats_before, stats_after, expected_stats = { 0 };
> + const int num_random_pages = 10; /* Total number of random pages to generate */
> + struct alignment_fillers fillers;
> + uint64_t *sizes;
> + uint64_t *random_pages;
> + uint64_t start_addr = 1ull << 30; /* Start at 1GB aligned */
> + int count_4k = 0, count_64k = 0, count_2m = 0;
> + uint64_t current_offset = 0;
> + uint64_t remainder, j;
> + int idx = 0;
> + int i, rand_val;
> +
> + /* Generate random page sizes */
> + random_pages = calloc(num_random_pages, sizeof(uint64_t));
> + igt_assert(random_pages);
> +
> + for (i = 0; i < num_random_pages; i++) {
> + rand_val = random() % 100;
> +
> + /* Weight the distribution: 50% 4K, 30% 64K, 20% 2M */
> + if (rand_val < 50)
> + random_pages[i] = SZ_4K;
> + else if (rand_val < 80)
> + random_pages[i] = SZ_64K;
> + else
> + random_pages[i] = SZ_2M;
> + }
> +
> + /* Over-allocate: worst case is 47 pages per random page (15×4K + 31×64K + 1×2M) */
> + sizes = calloc(num_random_pages * 47, sizeof(uint64_t));
> + igt_assert(sizes);
> +
> + /* Populate sizes array in a single pass while tracking counts */
> + for (i = 0; i < num_random_pages; i++) {
> + fillers = calculate_alignment_fillers(start_addr,
> + current_offset,
> + random_pages[i]);
> +
> + if (fillers.gap_to_64k || fillers.remaining_gap) {
> + /* Fill gap to 64K boundary with 4K pages */
> + for (j = 0; j < fillers.gap_to_64k; j += SZ_4K) {
> + sizes[idx++] = SZ_4K;
> + current_offset += SZ_4K;
> + count_4k++;
> + }
> +
> + /* Fill remainder with 64K pages */
> + remainder = fillers.remaining_gap;
> + while (remainder >= SZ_64K) {
> + sizes[idx++] = SZ_64K;
> + current_offset += SZ_64K;
> + remainder -= SZ_64K;
> + count_64k++;
> + }
> +
> + /* After 64K alignment, remainder should always be 0 */
> + igt_assert_eq(remainder, 0);
> + }
> +
> + sizes[idx++] = random_pages[i];
> + current_offset += random_pages[i];
> +
> + switch (random_pages[i]) {
> + case SZ_4K:
> + count_4k++;
> + break;
> + case SZ_64K:
> + count_64k++;
> + break;
> + case SZ_2M:
> + count_2m++;
> + break;
> + }
> + }
> +
> + igt_assert_f(idx < OVERFLOW_PRL_SIZE,
> + "Random test generated %d entries, exceeds PRL limit %d\n",
> + idx, OVERFLOW_PRL_SIZE);
> +
> + /* Set expected stats based on tracked counts */
> + expected_stats.prl_4k_entry_count = count_4k;
> + expected_stats.prl_64k_entry_count = count_64k;
> + expected_stats.prl_2m_entry_count = count_2m;
> + expected_stats.prl_issued_count = 1;
> + expected_stats.prl_aborted_count = 0;
> +
> + igt_debug("Random test generated: %d total pages (%d 4K, %d 64K, %d 2M)\n",
> + idx, count_4k, count_64k, count_2m);
> +
> + stats_before = get_prl_stats(fd, 0);
> + test_vma_ranges_list(fd, sizes, idx, start_addr);
> + stats_after = get_prl_stats(fd, 0);
> +
> + free(random_pages);
> + free(sizes);
> + compare_prl_stats(&stats_before, &stats_after, &expected_stats);
> +}
> +
> /**
> * SUBTEST: prl-invalidate-full
> * Description: Create 512 4K page entries at the maximum page reclaim list
> @@ -410,11 +557,15 @@ int igt_main()
>
> igt_require_f(xe_gt_stats_get_count(fd, 0, "prl_4k_entry_count") >= 0,
> "gt_stats is required for Page Reclamation tests.\n");
> + igt_srandom();
> }
>
> igt_subtest("basic-mixed")
> test_vma_ranges_basic_mixed(fd);
>
> + igt_subtest("random")
> + test_vma_range_random(fd);
> +
Reviewed-by: Xin Wang <x.wang@intel.com>
> igt_subtest("prl-invalidate-full")
> test_vma_ranges_prl_entries(fd, OVERFLOW_PRL_SIZE, 0, 1);
>
next prev parent reply other threads:[~2026-04-13 22:15 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-04-06 18:42 [PATCH 0/4] tests/xe: Add xe_page_reclaim test suite Brian Nguyen
2026-04-06 18:42 ` [PATCH 1/4] tests/xe: Add page reclaim test Brian Nguyen
2026-04-13 22:14 ` Wang, X
2026-04-06 18:42 ` [PATCH 2/4] tests/xe: Add random page reclaim subtest Brian Nguyen
2026-04-13 22:14 ` Wang, X [this message]
2026-04-06 18:42 ` [PATCH 3/4] tests/xe: Add transient display PRL skip Brian Nguyen
2026-04-13 22:15 ` Wang, X
2026-04-06 18:42 ` [PATCH 4/4] tests/xe: Add large VMA range tests for better coverage Brian Nguyen
2026-04-13 22:16 ` Wang, X
2026-04-06 19:29 ` ✓ Xe.CI.BAT: success for tests/xe: Add xe_page_reclaim test suite Patchwork
2026-04-06 19:45 ` ✓ i915.CI.BAT: " Patchwork
2026-04-06 21:45 ` ✓ i915.CI.Full: " Patchwork
2026-04-07 0:23 ` ✗ Xe.CI.FULL: failure " Patchwork
2026-04-07 19:15 ` [PATCH 0/4] " Summers, Stuart
2026-04-07 22:02 ` Nguyen, Brian3
2026-04-13 22:12 ` Wang, X
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=e64727e3-3eff-472b-bd4b-191ff9c61d88@intel.com \
--to=x.wang@intel.com \
--cc=brian3.nguyen@intel.com \
--cc=igt-dev@lists.freedesktop.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox