[PATCH 2/6] drm/xe/migrate: Add kunit to test clear functionality

Jahagirdar, Akshata akshata.jahagirdar at intel.com
Fri Jul 12 04:15:13 UTC 2024


On 7/11/2024 7:27 AM, Matthew Auld wrote:
> On 11/07/2024 10:18, Akshata Jahagirdar wrote:
>> This test verifies if the main and ccs data are cleared during bo 
>> creation.
>> The motivation to use Kunit instead of IGT is that, although we can 
>> verify
>> whether the data is zero following bo creation,
>> we cannot confirm whether the zero value after bo creation is the 
>> result of
>> our clear function or simply because the initial data present there 
>> was zero.
>
> One idea might be to pre-fill and release some amount of vram at the 
> start of the test. Then maybe spawn a number threads each allocating a 
> bit of vram (also various sizes) in a loop, each time dirtying the 
> pages and ccs before releasing it. There should be some amount of page 
> re-use to catch issues, plus is a lot more nasty with multiple threads 
> and varying sizes. Such a test can also be useful for lnl, just that 
> we use system memory instead of vram. Or maybe we already have 
> something like that?
>
If I am not wrong, think there is an igt testcase xe_evict_ccs which 
exercises similar testcases.
>>
>> Signed-off-by: Akshata Jahagirdar <akshata.jahagirdar at intel.com>
>> ---
>>   drivers/gpu/drm/xe/tests/xe_migrate.c      | 271 +++++++++++++++++++++
>>   drivers/gpu/drm/xe/tests/xe_migrate_test.c |   1 +
>>   drivers/gpu/drm/xe/tests/xe_migrate_test.h |   1 +
>>   3 files changed, 273 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/xe/tests/xe_migrate.c 
>> b/drivers/gpu/drm/xe/tests/xe_migrate.c
>> index 962f6438e219..6538c0e6b57e 100644
>> --- a/drivers/gpu/drm/xe/tests/xe_migrate.c
>> +++ b/drivers/gpu/drm/xe/tests/xe_migrate.c
>> @@ -359,3 +359,274 @@ void xe_migrate_sanity_kunit(struct kunit *test)
>>       xe_call_for_each_device(migrate_test_run_device);
>>   }
>>   EXPORT_SYMBOL_IF_KUNIT(xe_migrate_sanity_kunit);
>> +
>> +static struct dma_fence *blt_copy(struct xe_tile *tile,
>> +                struct xe_bo *src_bo, struct xe_bo *dst_bo,
>> +                bool copy_only_ccs, const char *str, struct kunit 
>> *test)
>> +{
>> +    struct xe_gt *gt = tile->primary_gt;
>> +    struct xe_migrate *m = tile->migrate;
>> +    struct xe_device *xe = gt_to_xe(gt);
>> +    struct dma_fence *fence = NULL;
>> +    u64 size = src_bo->size;
>> +    struct xe_res_cursor src_it, dst_it;
>> +    struct ttm_resource *src = src_bo->ttm.resource, *dst = 
>> dst_bo->ttm.resource;
>> +    u64 src_L0_ofs, dst_L0_ofs;
>> +    u32 src_L0_pt, dst_L0_pt;
>> +    u64 src_L0, dst_L0;
>> +    int err;
>> +    bool src_is_vram = mem_type_is_vram(src->mem_type);
>> +    bool dst_is_vram = mem_type_is_vram(dst->mem_type);
>> +
>> +    if (!src_is_vram)
>> +        xe_res_first_sg(xe_bo_sg(src_bo), 0, size, &src_it);
>> +    else
>> +        xe_res_first(src, 0, size, &src_it);
>> +
>> +    if (!dst_is_vram)
>> +        xe_res_first_sg(xe_bo_sg(dst_bo), 0, size, &dst_it);
>> +    else
>> +        xe_res_first(dst, 0, size, &dst_it);
>> +
>> +    while (size) {
>> +        u32 batch_size = 2; /* arb_clear() + MI_BATCH_BUFFER_END */
>> +        struct xe_sched_job *job;
>> +        struct xe_bb *bb;
>> +        u32 flush_flags = 0;
>> +        u32 update_idx;
>> +        u32 avail_pts = max_mem_transfer_per_pass(xe) / 
>> LEVEL0_PAGE_TABLE_ENCODE_SIZE;
>> +
>> +        src_L0 = xe_migrate_res_sizes(m, &src_it);
>> +        dst_L0 = xe_migrate_res_sizes(m, &dst_it);
>> +
>> +        src_L0 = min(src_L0, dst_L0);
>> +
>> +        batch_size += pte_update_size(m, src_is_vram, src_is_vram, 
>> src, &src_it, &src_L0,
>> +                          &src_L0_ofs, &src_L0_pt, 0, 0,
>> +                          avail_pts);
>> +
>> +        batch_size += pte_update_size(m, dst_is_vram, dst_is_vram, 
>> dst, &dst_it, &src_L0,
>> +                          &dst_L0_ofs, &dst_L0_pt, 0,
>> +                          avail_pts, avail_pts);
>> +
>> +        /* Add copy commands size here */
>> +        batch_size += ((copy_only_ccs) ? 0 : EMIT_COPY_DW) +
>> +            ((xe_device_has_flat_ccs(xe) && copy_only_ccs) ? 
>> EMIT_COPY_CCS_DW : 0);
>> +
>> +        bb = xe_bb_new(gt, batch_size, xe->info.has_usm);
>> +        if (IS_ERR(bb)) {
>> +            err = PTR_ERR(bb);
>> +            goto err_sync;
>> +        }
>> +
>> +        if (src_is_vram)
>> +            xe_res_next(&src_it, src_L0);
>> +        else
>> +            emit_pte(m, bb, src_L0_pt, src_is_vram, false,
>> +                 &src_it, src_L0, src);
>> +
>> +        if (dst_is_vram)
>> +            xe_res_next(&dst_it, src_L0);
>> +        else
>> +            emit_pte(m, bb, dst_L0_pt, dst_is_vram, false,
>> +                 &dst_it, src_L0, dst);
>> +
>> +        bb->cs[bb->len++] = MI_BATCH_BUFFER_END;
>> +        update_idx = bb->len;
>> +        if (!copy_only_ccs)
>> +            emit_copy(gt, bb, src_L0_ofs, dst_L0_ofs, src_L0, 
>> XE_PAGE_SIZE);
>> +
>> +        if (copy_only_ccs)
>> +            flush_flags = xe_migrate_ccs_copy(m, bb, src_L0_ofs,
>> +                    src_is_vram, dst_L0_ofs,
>> +                    dst_is_vram, src_L0, dst_L0_ofs, copy_only_ccs);
>> +
>> +        mutex_lock(&m->job_mutex);
>
> This is not allowed anymore it seems. CI is also complaing about this. 
> See: 50e52592fbe791d96ec2cb431d158cc6bc495be5
>
> Not sure what else has changed.
>
I see. I missed this. Thank you! Will fix this!
>> +
>> +        job = xe_bb_create_migration_job(m->q, bb,
>> +                         xe_migrate_batch_base(m, xe->info.has_usm),
>> +                         update_idx);
>> +        if (IS_ERR(job)) {
>> +            err = PTR_ERR(job);
>> +            goto err;
>> +        }
>> +
>> +        xe_sched_job_add_migrate_flush(job, flush_flags);
>> +
>> +        xe_sched_job_arm(job);
>> +        dma_fence_put(fence);
>> +        fence = dma_fence_get(&job->drm.s_fence->finished);
>> +        xe_sched_job_push(job);
>> +
>> +        dma_fence_put(m->fence);
>> +        m->fence = dma_fence_get(fence);
>> +
>> +        mutex_unlock(&m->job_mutex);
>> +
>> +        xe_bb_free(bb, fence);
>> +        size -= src_L0;
>> +        continue;
>> +
>> +err:
>> +        mutex_unlock(&m->job_mutex);
>> +        xe_bb_free(bb, NULL);
>> +
>> +err_sync:
>> +        if (fence) {
>> +            dma_fence_wait(fence, false);
>> +            dma_fence_put(fence);
>> +        }
>> +        return ERR_PTR(err);
>> +    }
>> +
>> +    return fence;
>> +}
>> +
>> +static void test_clear(struct xe_device *xe, struct xe_tile *tile,
>> +            struct xe_bo *sys_bo, struct xe_bo *vram_bo, struct 
>> kunit *test)
>> +{
>> +    struct dma_fence *fence;
>> +    u64 expected, retval;
>> +
>> +    expected = 0xd0d0d0d0d0d0d0d0;
>> +    xe_map_memset(xe, &sys_bo->vmap, 0, 0xd0, sys_bo->size);
>> +
>> +    fence = blt_copy(tile, sys_bo, vram_bo, false, "Blit copy from 
>> sysmem to vram", test);
>> +    if (!sanity_fence_failed(xe, fence, "Blit copy from sysmem to 
>> vram", test)) {
>> +        retval = xe_map_rd(xe, &vram_bo->vmap, 0, u64);
>> +        if (retval == expected)
>> +            KUNIT_FAIL(test, "Sanity check failed: VRAM must have 
>> compressed value\n");
>> +    }
>> +    dma_fence_put(fence);
>> +
>> +    fence = blt_copy(tile, vram_bo, sys_bo, false, "Blit copy from 
>> vram to sysmem", test);
>> +    if (!sanity_fence_failed(xe, fence, "Blit copy from vram to 
>> sysmem", test)) {
>> +        retval = xe_map_rd(xe, &sys_bo->vmap, 0, u64);
>> +        check(retval, expected, "Decompressed value must be equal to 
>> initial value", test);
>> +        retval = xe_map_rd(xe, &sys_bo->vmap, sys_bo->size - 8, u64);
>> +        check(retval, expected, "Decompressed value must be equal to 
>> initial value", test);
>> +    }
>> +    dma_fence_put(fence);
>> +
>> +    kunit_info(test, "Clear vram buffer object\n");
>> +    expected = 0x0000000000000000;
>> +    fence = xe_migrate_clear(tile->migrate, vram_bo, 
>> vram_bo->ttm.resource);
>> +
>> +    fence = blt_copy(tile, vram_bo, sys_bo,
>> +                false, "Blit copy from vram to sysmem", test);
>> +    if (!sanity_fence_failed(xe, fence, "Clear main buffer data", 
>> test)) {
>> +        retval = xe_map_rd(xe, &sys_bo->vmap, 0, u64);
>> +        check(retval, expected, "Clear main buffer first value", test);
>> +        retval = xe_map_rd(xe, &sys_bo->vmap, sys_bo->size - 8, u64);
>> +        check(retval, expected, "Clear main buffer last value", test);
>> +    }
>> +    dma_fence_put(fence);
>> +
>> +    fence = blt_copy(tile, vram_bo, sys_bo,
>> +            true, "Blit surf copy from vram to sysmem", test);
>> +    if (!sanity_fence_failed(xe, fence, "Clear ccs buffer data", 
>> test)) {
>> +        retval = xe_map_rd(xe, &sys_bo->vmap, 0, u64);
>> +        check(retval, expected, "Clear ccs data first value", test);
>> +        retval = xe_map_rd(xe, &sys_bo->vmap, sys_bo->size - 8, u64);
>> +        check(retval, expected, "Clear ccs data last value", test);
>> +    }
>> +    dma_fence_put(fence);
>> +}
>> +
>> +static void validate_ccs_test_run_tile(struct xe_device *xe, struct 
>> xe_tile *tile, struct kunit *test)
>> +{
>> +    struct xe_bo *sys_bo, *vram_bo;
>> +    unsigned int bo_flags = XE_BO_FLAG_VRAM_IF_DGFX(tile);
>> +    long ret;
>> +
>> +    sys_bo = xe_bo_create_user(xe, NULL, NULL, SZ_4M, 
>> DRM_XE_GEM_CPU_CACHING_WC,
>> +            ttm_bo_type_device, XE_BO_FLAG_SYSTEM | 
>> XE_BO_FLAG_NEEDS_CPU_ACCESS);
>> +
>> +    if (IS_ERR(sys_bo)) {
>> +        KUNIT_FAIL(test, "xe_bo_create() failed with err=%ld\n",
>> +               PTR_ERR(sys_bo));
>> +        return;
>> +    }
>> +
>> +    xe_bo_lock(sys_bo, false);
>> +    ret = xe_bo_validate(sys_bo, NULL, false);
>> +    if (ret) {
>> +        KUNIT_FAIL(test, "Failed to validate system bo for: %li\n", 
>> ret);
>> +        goto out_unlock;
>
> vram_bo is not initialized here.
>
>> +    }
>> +
>> +    ret = xe_bo_vmap(sys_bo);
>> +    if (ret) {
>> +        KUNIT_FAIL(test, "Failed to vmap system bo: %li\n", ret);
>> +        goto out_unlock;
>> +    }
>> +    xe_bo_unlock(sys_bo);
>> +
>> +    vram_bo = xe_bo_create_user(xe, NULL, NULL, SZ_4M, 
>> DRM_XE_GEM_CPU_CACHING_WC,
>> +                   ttm_bo_type_device, bo_flags | 
>> XE_BO_FLAG_NEEDS_CPU_ACCESS);
>> +    if (IS_ERR(vram_bo)) {
>> +        KUNIT_FAIL(test, "xe_bo_create() failed with err=%ld\n",
>> +               PTR_ERR(vram_bo));
>> +        return;
>
> Missing bo_put?
>
Since, bo creation failed, we don't call bo_put() again.
This is taken care by the xe_bo_create_user() function.
>
>> +    }
>> +
>> +    xe_bo_lock(vram_bo, false);
>> +    ret = xe_bo_validate(vram_bo, NULL, false);
>> +    if (ret) {
>> +        KUNIT_FAIL(test, "Failed to validate vram bo for: %li\n", ret);
>> +        goto out_unlock;
>> +    }
>> +
>> +    ret = xe_bo_vmap(vram_bo);
>> +    if (ret) {
>> +        KUNIT_FAIL(test, "Failed to vmap vram bo: %li\n", ret);
>> +        goto out_unlock;
>> +    }
>> +
>> +    test_clear(xe, tile, sys_bo, vram_bo, test);
>> +    xe_bo_unlock(vram_bo);
>> +
>> +    xe_bo_lock(vram_bo, false);
>> +    xe_bo_vunmap(vram_bo);
>> +    xe_bo_unlock(vram_bo);
>> +
>> +    xe_bo_lock(sys_bo, false);
>> +    xe_bo_vunmap(sys_bo);
>> +    xe_bo_unlock(sys_bo);
>> +out_unlock:
>
> There is no unlock here.
Will change it to out_put here.
>> +    xe_bo_put(vram_bo);
>> +    xe_bo_put(sys_bo);
>> +}
>> +
>> +static int validate_ccs_test_run_device(struct xe_device *xe)
>> +{
>> +    struct kunit *test = xe_cur_kunit();
>> +    struct xe_tile *tile;
>> +    int id;
>> +
>> +    if (!xe_device_has_flat_ccs(xe)) {
>> +        kunit_info(test, "Skipping non-flat-ccs device.\n");
>> +        return 0;
>> +    }
>> +
>> +    if (!(GRAPHICS_VER(xe) >= 20 && IS_DGFX(xe))) {
>> +        kunit_info(test, "Skipping non-xe2 discrete device %s.\n",
>> +               dev_name(xe->drm.dev));
>> +        return 0;
>> +    }
>
> So is there something else for xe2 igpu and ccs?
Yes, so there is another kunit test xe_ccs_migrate_kunit for xe2 igpu
that does data validation for ccs as well.
>> +
>> +    xe_pm_runtime_get(xe);
>> +
>> +    for_each_tile(tile, xe, id)
>> +        validate_ccs_test_run_tile(xe, tile, test);
>> +
>> +    xe_pm_runtime_put(xe);
>> +
>> +    return 0;
>> +}
>> +
>> +void xe_validate_ccs_kunit(struct kunit *test)
>> +{
>> +    xe_call_for_each_device(validate_ccs_test_run_device);
>> +}
>> +EXPORT_SYMBOL_IF_KUNIT(xe_validate_ccs_kunit);
>> diff --git a/drivers/gpu/drm/xe/tests/xe_migrate_test.c 
>> b/drivers/gpu/drm/xe/tests/xe_migrate_test.c
>> index eb0d8963419c..49b9ef060ad3 100644
>> --- a/drivers/gpu/drm/xe/tests/xe_migrate_test.c
>> +++ b/drivers/gpu/drm/xe/tests/xe_migrate_test.c
>> @@ -9,6 +9,7 @@
>>     static struct kunit_case xe_migrate_tests[] = {
>>       KUNIT_CASE(xe_migrate_sanity_kunit),
>> +    KUNIT_CASE(xe_validate_ccs_kunit),
>>       {}
>>   };
>>   diff --git a/drivers/gpu/drm/xe/tests/xe_migrate_test.h 
>> b/drivers/gpu/drm/xe/tests/xe_migrate_test.h
>> index 7c645c66824f..dc0ecfdfad39 100644
>> --- a/drivers/gpu/drm/xe/tests/xe_migrate_test.h
>> +++ b/drivers/gpu/drm/xe/tests/xe_migrate_test.h
>> @@ -9,5 +9,6 @@
>>   struct kunit;
>>     void xe_migrate_sanity_kunit(struct kunit *test);
>> +void xe_validate_ccs_kunit(struct kunit *test);
>>     #endif


More information about the Intel-xe mailing list