[igt-dev] [PATCH v5 09/21] drm-uapi/xe: Reject bo creation of unaligned size
Rodrigo Vivi
rodrigo.vivi at intel.com
Thu Nov 30 20:06:45 UTC 2023
On Thu, Nov 30, 2023 at 06:45:24PM +0000, Francois Dugast wrote:
> Align with kernel commit ("drm/xe/uapi: Reject bo creation of unaligned size")
>
> Signed-off-by: Francois Dugast <francois.dugast at intel.com>
A grep per 4096 gives me the feeling that are more places we could replace
by this min alignment.
Anyway, let's move with this and convert any remaining cases later
Reviewed-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
> ---
> include/drm-uapi/xe_drm.h | 17 +++++++++--------
> tests/intel/xe_mmap.c | 22 ++++++++++++----------
> tests/intel/xe_prime_self_import.c | 28 +++++++++++++++++++++++++---
> tests/intel/xe_vm.c | 13 ++++++-------
> 4 files changed, 52 insertions(+), 28 deletions(-)
>
> diff --git a/include/drm-uapi/xe_drm.h b/include/drm-uapi/xe_drm.h
> index 6e97270dc..2a4e8743b 100644
> --- a/include/drm-uapi/xe_drm.h
> +++ b/include/drm-uapi/xe_drm.h
> @@ -206,11 +206,13 @@ struct drm_xe_query_mem_region {
> *
> * When the kernel allocates memory for this region, the
> * underlying pages will be at least @min_page_size in size.
> - *
> - * Important note: When userspace allocates a GTT address which
> - * can point to memory allocated from this region, it must also
> - * respect this minimum alignment. This is enforced by the
> - * kernel.
> + * Buffer objects with an allowable placement in this region must be
> + * created with a size aligned to this value.
> + * GPU virtual address mappings of (parts of) buffer objects that
> + * may be placed in this region must also have their GPU virtual
> + * address and range aligned to this value.
> + * Affected IOCTLS will return %-EINVAL if alignment restrictions are
> + * not met.
> */
> __u32 min_page_size;
> /**
> @@ -516,9 +518,8 @@ struct drm_xe_gem_create {
> __u64 extensions;
>
> /**
> - * @size: Requested size for the object
> - *
> - * The (page-aligned) allocated size for the object will be returned.
> + * @size: Size of the object to be created, must match region
> + * (system or vram) minimum alignment (&min_page_size).
> */
> __u64 size;
>
> diff --git a/tests/intel/xe_mmap.c b/tests/intel/xe_mmap.c
> index d5ca21b7a..63fdf46a8 100644
> --- a/tests/intel/xe_mmap.c
> +++ b/tests/intel/xe_mmap.c
> @@ -47,17 +47,18 @@
> static void
> test_mmap(int fd, uint32_t placement, uint32_t flags)
> {
> + size_t bo_size = xe_get_default_alignment(fd);
> uint32_t bo;
> void *map;
>
> igt_require_f(placement, "Device doesn't support such memory region\n");
>
> - bo = xe_bo_create(fd, 0, 4096, placement, flags);
> + bo = xe_bo_create(fd, 0, bo_size, placement, flags);
>
> - map = xe_bo_map(fd, bo, 4096);
> + map = xe_bo_map(fd, bo, bo_size);
> strcpy(map, "Write some data to the BO!");
>
> - munmap(map, 4096);
> + munmap(map, bo_size);
>
> gem_close(fd, bo);
> }
> @@ -156,13 +157,14 @@ static void trap_sigbus(uint32_t *ptr)
> */
> static void test_small_bar(int fd)
> {
> + size_t page_size = xe_get_default_alignment(fd);
> uint32_t visible_size = xe_visible_vram_size(fd, 0);
> uint32_t bo;
> uint64_t mmo;
> uint32_t *map;
>
> /* 2BIG invalid case */
> - igt_assert_neq(__xe_bo_create(fd, 0, visible_size + 4096,
> + igt_assert_neq(__xe_bo_create(fd, 0, visible_size + page_size,
> vram_memory(fd, 0),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM,
> &bo),
> @@ -172,12 +174,12 @@ static void test_small_bar(int fd)
> bo = xe_bo_create(fd, 0, visible_size / 4, vram_memory(fd, 0),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
> mmo = xe_bo_mmap_offset(fd, bo);
> - map = mmap(NULL, 4096, PROT_WRITE, MAP_SHARED, fd, mmo);
> + map = mmap(NULL, page_size, PROT_WRITE, MAP_SHARED, fd, mmo);
> igt_assert(map != MAP_FAILED);
>
> map[0] = 0xdeadbeaf;
>
> - munmap(map, 4096);
> + munmap(map, page_size);
> gem_close(fd, bo);
>
> /* Normal operation with system memory spilling */
> @@ -186,18 +188,18 @@ static void test_small_bar(int fd)
> system_memory(fd),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
> mmo = xe_bo_mmap_offset(fd, bo);
> - map = mmap(NULL, 4096, PROT_WRITE, MAP_SHARED, fd, mmo);
> + map = mmap(NULL, page_size, PROT_WRITE, MAP_SHARED, fd, mmo);
> igt_assert(map != MAP_FAILED);
>
> map[0] = 0xdeadbeaf;
>
> - munmap(map, 4096);
> + munmap(map, page_size);
> gem_close(fd, bo);
>
> /* Bogus operation with SIGBUS */
> - bo = xe_bo_create(fd, 0, visible_size + 4096, vram_memory(fd, 0), 0);
> + bo = xe_bo_create(fd, 0, visible_size + page_size, vram_memory(fd, 0), 0);
> mmo = xe_bo_mmap_offset(fd, bo);
> - map = mmap(NULL, 4096, PROT_WRITE, MAP_SHARED, fd, mmo);
> + map = mmap(NULL, page_size, PROT_WRITE, MAP_SHARED, fd, mmo);
> igt_assert(map != MAP_FAILED);
>
> trap_sigbus(map);
> diff --git a/tests/intel/xe_prime_self_import.c b/tests/intel/xe_prime_self_import.c
> index 9a263d326..8e7290e9e 100644
> --- a/tests/intel/xe_prime_self_import.c
> +++ b/tests/intel/xe_prime_self_import.c
> @@ -59,15 +59,20 @@ IGT_TEST_DESCRIPTION("Check whether prime import/export works on the same"
> static char counter;
> static int g_time_out = 5;
> static pthread_barrier_t g_barrier;
> -static size_t bo_size;
> +
> +static size_t get_min_bo_size(int fd1, int fd2)
> +{
> + return 4 * max(xe_get_default_alignment(fd1),
> + xe_get_default_alignment(fd2));
> +}
>
> static void
> check_bo(int fd1, uint32_t handle1, int fd2, uint32_t handle2)
> {
> + size_t bo_size = get_min_bo_size(fd1, fd2);
> char *ptr1, *ptr2;
> int i;
>
> -
> ptr1 = xe_bo_map(fd1, handle1, bo_size);
> ptr2 = xe_bo_map(fd2, handle2, bo_size);
>
> @@ -97,6 +102,7 @@ check_bo(int fd1, uint32_t handle1, int fd2, uint32_t handle2)
> static void test_with_fd_dup(void)
> {
> int fd1, fd2;
> + size_t bo_size;
> uint32_t handle, handle_import;
> int dma_buf_fd1, dma_buf_fd2;
>
> @@ -105,6 +111,8 @@ static void test_with_fd_dup(void)
> fd1 = drm_open_driver(DRIVER_XE);
> fd2 = drm_open_driver(DRIVER_XE);
>
> + bo_size = get_min_bo_size(fd1, fd2);
> +
> handle = xe_bo_create(fd1, 0, bo_size, vram_if_possible(fd1, 0),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
>
> @@ -131,6 +139,7 @@ static void test_with_fd_dup(void)
> static void test_with_two_bos(void)
> {
> int fd1, fd2;
> + size_t bo_size;
> uint32_t handle1, handle2, handle_import;
> int dma_buf_fd;
>
> @@ -139,6 +148,8 @@ static void test_with_two_bos(void)
> fd1 = drm_open_driver(DRIVER_XE);
> fd2 = drm_open_driver(DRIVER_XE);
>
> + bo_size = get_min_bo_size(fd1, fd2);
> +
> handle1 = xe_bo_create(fd1, 0, bo_size, vram_if_possible(fd1, 0),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
> handle2 = xe_bo_create(fd1, 0, bo_size, vram_if_possible(fd1, 0),
> @@ -171,12 +182,15 @@ static void test_with_two_bos(void)
> static void test_with_one_bo_two_files(void)
> {
> int fd1, fd2;
> + size_t bo_size;
> uint32_t handle_import, handle_open, handle_orig, flink_name;
> int dma_buf_fd1, dma_buf_fd2;
>
> fd1 = drm_open_driver(DRIVER_XE);
> fd2 = drm_open_driver(DRIVER_XE);
>
> + bo_size = get_min_bo_size(fd1, fd2);
> +
> handle_orig = xe_bo_create(fd1, 0, bo_size,
> vram_if_possible(fd1, 0),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
> @@ -205,12 +219,15 @@ static void test_with_one_bo_two_files(void)
> static void test_with_one_bo(void)
> {
> int fd1, fd2;
> + size_t bo_size;
> uint32_t handle, handle_import1, handle_import2, handle_selfimport;
> int dma_buf_fd;
>
> fd1 = drm_open_driver(DRIVER_XE);
> fd2 = drm_open_driver(DRIVER_XE);
>
> + bo_size = get_min_bo_size(fd1, fd2);
> +
> handle = xe_bo_create(fd1, 0, bo_size, vram_if_possible(fd1, 0),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
>
> @@ -279,6 +296,7 @@ static void *thread_fn_reimport_vs_close(void *p)
> pthread_t *threads;
> int r, i, num_threads;
> int fds[2];
> + size_t bo_size;
> int obj_count;
> void *status;
> uint32_t handle;
> @@ -298,6 +316,8 @@ static void *thread_fn_reimport_vs_close(void *p)
>
> fds[0] = drm_open_driver(DRIVER_XE);
>
> + bo_size = xe_get_default_alignment(fds[0]);
> +
> handle = xe_bo_create(fds[0], 0, bo_size,
> vram_if_possible(fds[0], 0),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
> @@ -336,6 +356,7 @@ static void *thread_fn_export_vs_close(void *p)
> struct drm_prime_handle prime_h2f;
> struct drm_gem_close close_bo;
> int fd = (uintptr_t)p;
> + size_t bo_size = xe_get_default_alignment(fd);
> uint32_t handle;
>
> pthread_barrier_wait(&g_barrier);
> @@ -463,6 +484,7 @@ static void test_llseek_size(void)
> static void test_llseek_bad(void)
> {
> int fd;
> + size_t bo_size;
> uint32_t handle;
> int dma_buf_fd;
>
> @@ -470,6 +492,7 @@ static void test_llseek_bad(void)
>
> fd = drm_open_driver(DRIVER_XE);
>
> + bo_size = 4 * xe_get_default_alignment(fd);
> handle = xe_bo_create(fd, 0, bo_size,
> vram_if_possible(fd, 0),
> DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
> @@ -510,7 +533,6 @@ igt_main
>
> igt_fixture {
> fd = drm_open_driver(DRIVER_XE);
> - bo_size = xe_get_default_alignment(fd);
> }
>
> for (i = 0; i < ARRAY_SIZE(tests); i++) {
> diff --git a/tests/intel/xe_vm.c b/tests/intel/xe_vm.c
> index c9b57b444..bfdff63f0 100644
> --- a/tests/intel/xe_vm.c
> +++ b/tests/intel/xe_vm.c
> @@ -1315,11 +1315,10 @@ test_munmap_style_unbind(int fd, struct drm_xe_engine_class_instance *eci,
> if (flags & MAP_FLAG_HAMMER_FIRST_PAGE) {
> t.fd = fd;
> t.vm = vm;
> -#define PAGE_SIZE 4096
> - t.addr = addr + PAGE_SIZE / 2;
> + t.addr = addr + page_size / 2;
> t.eci = eci;
> t.exit = &exit;
> - t.map = map + PAGE_SIZE / 2;
> + t.map = map + page_size / 2;
> t.barrier = &barrier;
> pthread_barrier_init(&barrier, NULL, 2);
> pthread_create(&t.thread, 0, hammer_thread, &t);
> @@ -1372,8 +1371,8 @@ test_munmap_style_unbind(int fd, struct drm_xe_engine_class_instance *eci,
> igt_assert_eq(data->data, 0xc0ffee);
> }
> if (flags & MAP_FLAG_HAMMER_FIRST_PAGE) {
> - memset(map, 0, PAGE_SIZE / 2);
> - memset(map + PAGE_SIZE, 0, bo_size - PAGE_SIZE);
> + memset(map, 0, page_size / 2);
> + memset(map + page_size, 0, bo_size - page_size);
> } else {
> memset(map, 0, bo_size);
> }
> @@ -1422,8 +1421,8 @@ try_again_after_invalidate:
> }
> }
> if (flags & MAP_FLAG_HAMMER_FIRST_PAGE) {
> - memset(map, 0, PAGE_SIZE / 2);
> - memset(map + PAGE_SIZE, 0, bo_size - PAGE_SIZE);
> + memset(map, 0, page_size / 2);
> + memset(map + page_size, 0, bo_size - page_size);
> } else {
> memset(map, 0, bo_size);
> }
> --
> 2.34.1
>
More information about the igt-dev
mailing list