tests/amdgpu: add dispatch hang test

add compute/gfx dispatch hang test for gfx9

Signed-off-by: Flora Cui <flora.cui@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

diff --git a/tests/amdgpu/amdgpu_test.c b/tests/amdgpu/amdgpu_test.c
index 5aadd7e649ab20a84fcd017db161935a7947b5e2..08642203c71a55aa33311386c031046c74dc65be 100644 (file)
--- a/tests/amdgpu/amdgpu_test.c
+++ b/tests/amdgpu/amdgpu_test.c
@@ -452,6 +452,18 @@ static void amdgpu_disable_suites()
                                "sdma ring block test (set amdgpu.lockup_timeout=50)", CU_FALSE))
                fprintf(stderr, "test deactivation failed - %s\n", CU_get_error_msg());
 
+       /* This test was ran on GFX9 only */
+       //if (family_id < AMDGPU_FAMILY_AI || family_id > AMDGPU_FAMILY_RV)
+               if (amdgpu_set_test_active(DEADLOCK_TESTS_STR,
+                               "gfx ring bad dispatch test (set amdgpu.lockup_timeout=50)", CU_FALSE))
+                       fprintf(stderr, "test deactivation failed - %s\n", CU_get_error_msg());
+
+       /* This test was ran on GFX9 only */
+       //if (family_id < AMDGPU_FAMILY_AI || family_id > AMDGPU_FAMILY_RV)
+               if (amdgpu_set_test_active(DEADLOCK_TESTS_STR,
+                               "compute ring bad dispatch test (set amdgpu.lockup_timeout=50)", CU_FALSE))
+                       fprintf(stderr, "test deactivation failed - %s\n", CU_get_error_msg());
+
        if (amdgpu_set_test_active(BO_TESTS_STR, "Metadata", CU_FALSE))
                fprintf(stderr, "test deactivation failed - %s\n", CU_get_error_msg());
 

diff --git a/tests/amdgpu/amdgpu_test.h b/tests/amdgpu/amdgpu_test.h
index 36675ea3f70538e0a23e68ada41b43ee68c5876c..e1d9025372e764c1aa05a5434f2424673c6cbfbd 100644 (file)
--- a/tests/amdgpu/amdgpu_test.h
+++ b/tests/amdgpu/amdgpu_test.h
@@ -236,6 +236,7 @@ CU_BOOL suite_syncobj_timeline_tests_enable(void);
  */
 extern CU_TestInfo syncobj_timeline_tests[];
 
+void amdgpu_dispatch_hang_helper(amdgpu_device_handle device_handle, uint32_t ip_type);
 
 /**
  * Helper functions

diff --git a/tests/amdgpu/basic_tests.c b/tests/amdgpu/basic_tests.c
index b2468212004b424fd2c0b966d947f49eb60ba357..9dde7af0381e05b879ce7f86bd307b445bb31dc4 100644 (file)
--- a/tests/amdgpu/basic_tests.c
+++ b/tests/amdgpu/basic_tests.c
@@ -306,7 +306,8 @@ static  uint32_t shader_bin[] = {
 
 enum cs_type {
        CS_BUFFERCLEAR,
-       CS_BUFFERCOPY
+       CS_BUFFERCOPY,
+       CS_HANG
 };
 
 static const uint32_t bufferclear_cs_shader_gfx9[] = {
@@ -468,6 +469,14 @@ static const uint32_t cached_cmd_gfx9[] = {
        0xc0026900, 0x2b0, 0x0, 0x0, 0xc0016900, 0x2f8, 0x0
 };
 
+unsigned int memcpy_ps_hang[] = {
+        0xFFFFFFFF, 0xBEFE0A7E, 0xBEFC0304, 0xC0C20100,
+        0xC0800300, 0xC8080000, 0xC80C0100, 0xC8090001,
+        0xC80D0101, 0xBF8C007F, 0xF0800F00, 0x00010002,
+        0xBEFE040C, 0xBF8C0F70, 0xBF800000, 0xBF800000,
+        0xF800180F, 0x03020100, 0xBF810000
+};
+
 int amdgpu_bo_alloc_and_map_raw(amdgpu_device_handle dev, unsigned size,
                        unsigned alignment, unsigned heap, uint64_t alloc_flags,
                        uint64_t mapping_flags, amdgpu_bo_handle *bo, void **cpu,
@@ -2080,6 +2089,10 @@ static int amdgpu_dispatch_load_cs_shader(uint8_t *ptr,
                        shader = buffercopy_cs_shader_gfx9;
                        shader_size = sizeof(buffercopy_cs_shader_gfx9);
                        break;
+               case CS_HANG:
+                       shader = memcpy_ps_hang;
+                       shader_size = sizeof(memcpy_ps_hang);
+                       break;
                default:
                        return -1;
                        break;
@@ -2300,7 +2313,8 @@ static void amdgpu_memset_dispatch_test(amdgpu_device_handle device_handle,
 
 static void amdgpu_memcpy_dispatch_test(amdgpu_device_handle device_handle,
                                        uint32_t ip_type,
-                                       uint32_t ring)
+                                       uint32_t ring,
+                                       int hang)
 {
        amdgpu_context_handle context_handle;
        amdgpu_bo_handle bo_src, bo_dst, bo_shader, bo_cmd, resources[4];
@@ -2316,7 +2330,8 @@ static void amdgpu_memcpy_dispatch_test(amdgpu_device_handle device_handle,
        int bo_cmd_size = 4096;
        struct amdgpu_cs_request ibs_request = {0};
        struct amdgpu_cs_ib_info ib_info= {0};
-       uint32_t expired;
+       uint32_t expired, hang_state, hangs;
+       enum cs_type cs_type;
        amdgpu_bo_list_handle bo_list;
        struct amdgpu_cs_fence fence_status = {0};
 
@@ -2337,7 +2352,8 @@ static void amdgpu_memcpy_dispatch_test(amdgpu_device_handle device_handle,
        CU_ASSERT_EQUAL(r, 0);
        memset(ptr_shader, 0, bo_shader_size);
 
-       r = amdgpu_dispatch_load_cs_shader(ptr_shader, CS_BUFFERCOPY );
+       cs_type = hang ? CS_HANG : CS_BUFFERCOPY;
+       r = amdgpu_dispatch_load_cs_shader(ptr_shader, cs_type);
        CU_ASSERT_EQUAL(r, 0);
 
        r = amdgpu_bo_alloc_and_map(device_handle, bo_dst_size, 4096,
@@ -2423,14 +2439,21 @@ static void amdgpu_memcpy_dispatch_test(amdgpu_device_handle device_handle,
        r = amdgpu_cs_query_fence_status(&fence_status,
                                         AMDGPU_TIMEOUT_INFINITE,
                                         0, &expired);
-       CU_ASSERT_EQUAL(r, 0);
-       CU_ASSERT_EQUAL(expired, true);
 
-       /* verify if memcpy test result meets with expected */
-       i = 0;
-       while(i < bo_dst_size) {
-               CU_ASSERT_EQUAL(ptr_dst[i], ptr_src[i]);
-               i++;
+       if (!hang) {
+               CU_ASSERT_EQUAL(r, 0);
+               CU_ASSERT_EQUAL(expired, true);
+
+               /* verify if memcpy test result meets with expected */
+               i = 0;
+               while(i < bo_dst_size) {
+                       CU_ASSERT_EQUAL(ptr_dst[i], ptr_src[i]);
+                       i++;
+               }
+       } else {
+               r = amdgpu_cs_query_reset_state(context_handle, &hang_state, &hangs);
+               CU_ASSERT_EQUAL(r, 0);
+               CU_ASSERT_EQUAL(hang_state, AMDGPU_CTX_UNKNOWN_RESET);
        }
 
        r = amdgpu_bo_list_destroy(bo_list);
@@ -2464,7 +2487,7 @@ static void amdgpu_compute_dispatch_test(void)
 
        for (ring_id = 0; (1 << ring_id) & info.available_rings; ring_id++) {
                amdgpu_memset_dispatch_test(device_handle, AMDGPU_HW_IP_COMPUTE, ring_id);
-               amdgpu_memcpy_dispatch_test(device_handle, AMDGPU_HW_IP_COMPUTE, ring_id);
+               amdgpu_memcpy_dispatch_test(device_handle, AMDGPU_HW_IP_COMPUTE, ring_id, 0);
        }
 }
 
@@ -2481,7 +2504,25 @@ static void amdgpu_gfx_dispatch_test(void)
 
        for (ring_id = 0; (1 << ring_id) & info.available_rings; ring_id++) {
                amdgpu_memset_dispatch_test(device_handle, AMDGPU_HW_IP_GFX, ring_id);
-               amdgpu_memcpy_dispatch_test(device_handle, AMDGPU_HW_IP_GFX, ring_id);
+               amdgpu_memcpy_dispatch_test(device_handle, AMDGPU_HW_IP_GFX, ring_id, 0);
+       }
+}
+
+void amdgpu_dispatch_hang_helper(amdgpu_device_handle device_handle, uint32_t ip_type)
+{
+       int r;
+       struct drm_amdgpu_info_hw_ip info;
+       uint32_t ring_id;
+
+       r = amdgpu_query_hw_ip_info(device_handle, ip_type, 0, &info);
+       CU_ASSERT_EQUAL(r, 0);
+       if (!info.available_rings)
+               printf("SKIP ... as there's no ring for ip %d\n", ip_type);
+
+       for (ring_id = 0; (1 << ring_id) & info.available_rings; ring_id++) {
+               amdgpu_memcpy_dispatch_test(device_handle, ip_type, ring_id, 0);
+               amdgpu_memcpy_dispatch_test(device_handle, ip_type, ring_id, 1);
+               amdgpu_memcpy_dispatch_test(device_handle, ip_type, ring_id, 0);
        }
 }
 

diff --git a/tests/amdgpu/deadlock_tests.c b/tests/amdgpu/deadlock_tests.c
index 91368c15641a8fee1a4e14e595589aff34f675e0..58f93b50796fcd301ce6cd4c0954fdfd0d1b2afa 100644 (file)
--- a/tests/amdgpu/deadlock_tests.c
+++ b/tests/amdgpu/deadlock_tests.c
@@ -114,6 +114,8 @@ static void amdgpu_deadlock_compute(void);
 static void amdgpu_illegal_reg_access();
 static void amdgpu_illegal_mem_access();
 static void amdgpu_deadlock_sdma(void);
+static void amdgpu_dispatch_hang_gfx(void);
+static void amdgpu_dispatch_hang_compute(void);
 
 CU_BOOL suite_deadlock_tests_enable(void)
 {
@@ -178,6 +180,8 @@ CU_TestInfo deadlock_tests[] = {
        { "sdma ring block test (set amdgpu.lockup_timeout=50)", amdgpu_deadlock_sdma },
        { "illegal reg access test", amdgpu_illegal_reg_access },
        { "illegal mem access test (set amdgpu.vm_fault_stop=2)", amdgpu_illegal_mem_access },
+       { "gfx ring bad dispatch test (set amdgpu.lockup_timeout=50)", amdgpu_dispatch_hang_gfx },
+       { "compute ring bad dispatch test (set amdgpu.lockup_timeout=50,50)", amdgpu_dispatch_hang_compute },
        CU_TEST_INFO_NULL,
 };
 
@@ -478,3 +482,13 @@ static void amdgpu_illegal_mem_access()
 {
        bad_access_helper(0);
 }
+
+static void amdgpu_dispatch_hang_gfx(void)
+{
+       amdgpu_dispatch_hang_helper(device_handle, AMDGPU_HW_IP_GFX);
+}
+
+static void amdgpu_dispatch_hang_compute(void)
+{
+       amdgpu_dispatch_hang_helper(device_handle, AMDGPU_HW_IP_COMPUTE);
+}