[Intel-gfx] [PATCH i-g-t] i915/module_load: Tidy up gem_exec_store workalike
Ramalingam C
ramalingam.c at intel.com
Fri Feb 5 17:43:54 UTC 2021
On 2021-02-05 at 10:10:05 +0000, Chris Wilson wrote:
> We emit a store on each GPU after loading the module to confirm the
> basic liveness of command submission. Trim away some of the chaff.
>
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Ramalingam C <ramalingam.c at intel.com>
Looks good to me
Reviewed-by: Ramalingam C <ramalingam.c at intel.com>
> ---
> tests/i915/i915_module_load.c | 146 ++++++++++++++--------------------
> 1 file changed, 58 insertions(+), 88 deletions(-)
>
> diff --git a/tests/i915/i915_module_load.c b/tests/i915/i915_module_load.c
> index 06522ba61..f1fb13914 100644
> --- a/tests/i915/i915_module_load.c
> +++ b/tests/i915/i915_module_load.c
> @@ -37,41 +37,45 @@
> #include "igt_sysfs.h"
> #include "igt_core.h"
>
> -static void store_all(int fd)
> +static void store_all(int i915)
> {
> - const unsigned int gen = intel_gen(intel_get_drm_devid(fd));
> - unsigned int permuted[I915_EXEC_RING_MASK + 1];
> - unsigned int engines[I915_EXEC_RING_MASK + 1];
> - struct drm_i915_gem_exec_object2 obj[2];
> - struct drm_i915_gem_relocation_entry reloc[2 * ARRAY_SIZE(engines)];
> - struct drm_i915_gem_execbuffer2 execbuf;
> - const struct intel_execution_engine2 *e;
> + const unsigned int gen = intel_gen(intel_get_drm_devid(i915));
> + uint32_t engines[I915_EXEC_RING_MASK + 1];
> uint32_t batch[16];
> - uint64_t offset;
> - unsigned nengine;
> - int value;
> - int i, j;
> -
> - memset(&execbuf, 0, sizeof(execbuf));
> - execbuf.buffers_ptr = (uintptr_t)obj;
> - execbuf.buffer_count = 2;
> -
> - memset(reloc, 0, sizeof(reloc));
> - memset(obj, 0, sizeof(obj));
> - obj[0].handle = gem_create(fd, 4096);
> - obj[1].handle = gem_create(fd, 4096);
> - obj[1].relocation_count = 1;
> -
> - offset = sizeof(uint32_t);
> + unsigned int sz = ALIGN(sizeof(batch) * ARRAY_SIZE(engines), 4096);
> + struct drm_i915_gem_relocation_entry reloc = {
> + .offset = sizeof(uint32_t),
> + .read_domains = I915_GEM_DOMAIN_RENDER,
> + .write_domain = I915_GEM_DOMAIN_RENDER,
> + };
> + struct drm_i915_gem_exec_object2 obj[2] = {
> + { .handle = gem_create(i915, sizeof(engines)) },
> + {
> + .handle = gem_create(i915, sz),
> + .relocation_count = 1,
> + .relocs_ptr = to_user_pointer(&reloc),
> + },
> + };
> + struct drm_i915_gem_execbuffer2 execbuf = {
> + .buffers_ptr = to_user_pointer(obj),
> + .buffer_count = 2,
> + };
> + const struct intel_execution_engine2 *e;
> + int reloc_sz = sizeof(uint32_t);
> + unsigned int nengine, value;
> + void *cs;
> + int i;
> +
> i = 0;
> batch[i] = MI_STORE_DWORD_IMM | (gen < 6 ? 1 << 22 : 0);
> if (gen >= 8) {
> batch[++i] = 0;
> batch[++i] = 0;
> + reloc_sz = sizeof(uint64_t);
> } else if (gen >= 4) {
> batch[++i] = 0;
> batch[++i] = 0;
> - offset += sizeof(uint32_t);
> + reloc.offset += sizeof(uint32_t);
> } else {
> batch[i]--;
> batch[++i] = 0;
> @@ -80,74 +84,43 @@ static void store_all(int fd)
> batch[++i] = MI_BATCH_BUFFER_END;
>
> nengine = 0;
> - intel_detect_and_clear_missed_interrupts(fd);
> - __for_each_physical_engine(fd, e) {
> - if (!gem_class_can_store_dword(fd, e->class))
> + cs = gem_mmap__device_coherent(i915, obj[1].handle, 0, sz, PROT_WRITE);
> + __for_each_physical_engine(i915, e) {
> + uint64_t addr;
> +
> + igt_assert(reloc.presumed_offset != -1);
> + addr = reloc.presumed_offset + reloc.delta;
> +
> + if (!gem_class_can_store_dword(i915, e->class))
> continue;
>
> - igt_assert(2 * (nengine + 1) * sizeof(batch) <= 4096);
> + engines[nengine] = nengine;
> + batch[value] = engines[nengine];
>
> - engines[nengine] = e->flags;
> + execbuf.flags = e->flags;
> if (gen < 6)
> - engines[nengine] |= I915_EXEC_SECURE;
> - execbuf.flags = engines[nengine];
> -
> - j = 2*nengine;
> - reloc[j].target_handle = obj[0].handle;
> - reloc[j].presumed_offset = ~0;
> - reloc[j].offset = j*sizeof(batch) + offset;
> - reloc[j].delta = nengine*sizeof(uint32_t);
> - reloc[j].read_domains = I915_GEM_DOMAIN_INSTRUCTION;
> - reloc[j].write_domain = I915_GEM_DOMAIN_INSTRUCTION;
> - obj[1].relocs_ptr = (uintptr_t)&reloc[j];
> -
> - batch[value] = 0xdeadbeef;
> - gem_write(fd, obj[1].handle, j*sizeof(batch),
> - batch, sizeof(batch));
> - execbuf.batch_start_offset = j*sizeof(batch);
> - gem_execbuf(fd, &execbuf);
> -
> - j = 2*nengine + 1;
> - reloc[j].target_handle = obj[0].handle;
> - reloc[j].presumed_offset = ~0;
> - reloc[j].offset = j*sizeof(batch) + offset;
> - reloc[j].delta = nengine*sizeof(uint32_t);
> - reloc[j].read_domains = I915_GEM_DOMAIN_INSTRUCTION;
> - reloc[j].write_domain = I915_GEM_DOMAIN_INSTRUCTION;
> - obj[1].relocs_ptr = (uintptr_t)&reloc[j];
> -
> - batch[value] = nengine;
> - gem_write(fd, obj[1].handle, j*sizeof(batch),
> - batch, sizeof(batch));
> - execbuf.batch_start_offset = j*sizeof(batch);
> - gem_execbuf(fd, &execbuf);
> -
> - nengine++;
> - }
> - gem_sync(fd, obj[1].handle);
> -
> - for (i = 0; i < nengine; i++) {
> - obj[1].relocs_ptr = (uintptr_t)&reloc[2*i];
> - execbuf.batch_start_offset = 2*i*sizeof(batch);
> - memcpy(permuted, engines, nengine*sizeof(engines[0]));
> - igt_permute_array(permuted, nengine, igt_exchange_int);
> - for (j = 0; j < nengine; j++) {
> - execbuf.flags = permuted[j];
> - gem_execbuf(fd, &execbuf);
> - }
> - obj[1].relocs_ptr = (uintptr_t)&reloc[2*i+1];
> - execbuf.batch_start_offset = (2*i+1)*sizeof(batch);
> - execbuf.flags = engines[i];
> - gem_execbuf(fd, &execbuf);
> + execbuf.flags |= I915_EXEC_SECURE;
> + execbuf.flags |= I915_EXEC_NO_RELOC | I915_EXEC_HANDLE_LUT;
> +
> + memcpy(cs + execbuf.batch_start_offset, batch, sizeof(batch));
> + memcpy(cs + reloc.offset, &addr, reloc_sz);
> + gem_execbuf(i915, &execbuf);
> +
> + if (++nengine == ARRAY_SIZE(engines))
> + break;
> +
> + reloc.delta += sizeof(uint32_t);
> + reloc.offset += sizeof(batch);
> + execbuf.batch_start_offset += sizeof(batch);
> }
> - gem_close(fd, obj[1].handle);
> + munmap(cs, sz);
> + gem_close(i915, obj[1].handle);
>
> - gem_read(fd, obj[0].handle, 0, engines, sizeof(engines));
> - gem_close(fd, obj[0].handle);
> + gem_read(i915, obj[0].handle, 0, engines, sizeof(engines));
> + gem_close(i915, obj[0].handle);
>
> for (i = 0; i < nengine; i++)
> igt_assert_eq_u32(engines[i], i);
> - igt_assert_eq(intel_detect_and_clear_missed_interrupts(fd), 0);
> }
>
> static int open_parameters(const char *module_name)
> @@ -196,11 +169,8 @@ static void gem_sanitycheck(void)
> err = 0;
> if (ioctl(i915, DRM_IOCTL_I915_GEM_SET_CACHING, &args))
> err = -errno;
> - if (err == -ENOENT) {
> - igt_fork_hang_detector(i915);
> + if (err == -ENOENT)
> store_all(i915);
> - igt_stop_hang_detector();
> - }
> errno = 0;
>
> close(i915);
> --
> 2.30.0
>
More information about the Intel-gfx
mailing list