drm/amdgpu: add prototype to dump ip state
PATCH v2 1/2 drm/amdgpu: add prototype to dump ip state
Add the prototype for all the ips of different
asics and set them to NULL for now and based on
the need will keep adding the function for each
ip eventually.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c | 1 + drivers/gpu/drm/amd/amdgpu/cik.c | 1 + drivers/gpu/drm/amd/amdgpu/cik_ih.c | 1 + drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 1 + drivers/gpu/drm/amd/amdgpu/cz_ih.c | 1 + drivers/gpu/drm/amd/amdgpu/dce_v10_0.c | 1 + drivers/gpu/drm/amd/amdgpu/dce_v11_0.c | 1 + drivers/gpu/drm/amd/amdgpu/dce_v6_0.c | 1 + drivers/gpu/drm/amd/amdgpu/dce_v8_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 1 + drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 1 + drivers/gpu/drm/amd/amdgpu/iceland_ih.c | 1 + drivers/gpu/drm/amd/amdgpu/ih_v6_0.c | 1 + drivers/gpu/drm/amd/amdgpu/ih_v6_1.c | 1 + drivers/gpu/drm/amd/amdgpu/ih_v7_0.c | 1 + drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c | 1 + drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c | 2 ++ drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c | 1 + drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c | 1 + drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c | 1 + drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c | 1 + drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c | 1 + drivers/gpu/drm/amd/amdgpu/mes_v10_1.c | 1 + drivers/gpu/drm/amd/amdgpu/mes_v11_0.c | 1 + drivers/gpu/drm/amd/amdgpu/navi10_ih.c | 1 + drivers/gpu/drm/amd/amdgpu/nv.c | 1 + drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 1 + drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 1 + drivers/gpu/drm/amd/amdgpu/si.c | 1 + drivers/gpu/drm/amd/amdgpu/si_dma.c | 1 + drivers/gpu/drm/amd/amdgpu/si_ih.c | 1 + drivers/gpu/drm/amd/amdgpu/soc15.c | 1 + drivers/gpu/drm/amd/amdgpu/soc21.c | 1 + drivers/gpu/drm/amd/amdgpu/tonga_ih.c | 1 + drivers/gpu/drm/amd/amdgpu/uvd_v3_1.c | 1 + drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c | 1 + drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c | 1 + drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vce_v2_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vce_v3_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 2 ++ drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vi.c | 1 + drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 1 + drivers/gpu/drm/amd/include/amd_shared.h | 1 + drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c | 1 + drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c | 1 + drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c | 1 + 64 files changed, 66 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c +1 -0
@@ -637,6 +637,7 @@ static const struct amd_ip_funcs acp_ip_funcs = { .soft_reset = acp_soft_reset, .set_clockgating_state = acp_set_clockgating_state, .set_powergating_state = acp_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version acp_ip_block = {
drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c +1 -0
@@ -875,6 +875,7 @@ static const struct amd_ip_funcs umsch_mm_v4_0_ip_funcs = { .hw_fini = umsch_mm_hw_fini, .suspend = umsch_mm_suspend, .resume = umsch_mm_resume, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version umsch_mm_v4_0_ip_block = {
drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c +1 -0
@@ -658,6 +658,7 @@ static const struct amd_ip_funcs amdgpu_vkms_ip_funcs = { .soft_reset = amdgpu_vkms_soft_reset, .set_clockgating_state = amdgpu_vkms_set_clockgating_state, .set_powergating_state = amdgpu_vkms_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version amdgpu_vkms_ip_block = {
drivers/gpu/drm/amd/amdgpu/cik.c +1 -0
@@ -2210,6 +2210,7 @@ static const struct amd_ip_funcs cik_common_ip_funcs = { .soft_reset = cik_common_soft_reset, .set_clockgating_state = cik_common_set_clockgating_state, .set_powergating_state = cik_common_set_powergating_state, + .dump_ip_state = NULL, }; static const struct amdgpu_ip_block_version cik_common_ip_block =
drivers/gpu/drm/amd/amdgpu/cik_ih.c +1 -0
@@ -435,6 +435,7 @@ static const struct amd_ip_funcs cik_ih_ip_funcs = { .soft_reset = cik_ih_soft_reset, .set_clockgating_state = cik_ih_set_clockgating_state, .set_powergating_state = cik_ih_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs cik_ih_funcs = {
drivers/gpu/drm/amd/amdgpu/cik_sdma.c +1 -0
@@ -1228,6 +1228,7 @@ static const struct amd_ip_funcs cik_sdma_ip_funcs = { .soft_reset = cik_sdma_soft_reset, .set_clockgating_state = cik_sdma_set_clockgating_state, .set_powergating_state = cik_sdma_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = {
drivers/gpu/drm/amd/amdgpu/cz_ih.c +1 -0
@@ -433,6 +433,7 @@ static const struct amd_ip_funcs cz_ih_ip_funcs = { .soft_reset = cz_ih_soft_reset, .set_clockgating_state = cz_ih_set_clockgating_state, .set_powergating_state = cz_ih_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs cz_ih_funcs = {
drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +1 -0
@@ -3333,6 +3333,7 @@ static const struct amd_ip_funcs dce_v10_0_ip_funcs = { .soft_reset = dce_v10_0_soft_reset, .set_clockgating_state = dce_v10_0_set_clockgating_state, .set_powergating_state = dce_v10_0_set_powergating_state, + .dump_ip_state = NULL, }; static void
drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +1 -0
@@ -3464,6 +3464,7 @@ static const struct amd_ip_funcs dce_v11_0_ip_funcs = { .soft_reset = dce_v11_0_soft_reset, .set_clockgating_state = dce_v11_0_set_clockgating_state, .set_powergating_state = dce_v11_0_set_powergating_state, + .dump_ip_state = NULL, }; static void
drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +1 -0
@@ -3154,6 +3154,7 @@ static const struct amd_ip_funcs dce_v6_0_ip_funcs = { .soft_reset = dce_v6_0_soft_reset, .set_clockgating_state = dce_v6_0_set_clockgating_state, .set_powergating_state = dce_v6_0_set_powergating_state, + .dump_ip_state = NULL, }; static void
drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +1 -0
@@ -3242,6 +3242,7 @@ static const struct amd_ip_funcs dce_v8_0_ip_funcs = { .soft_reset = dce_v8_0_soft_reset, .set_clockgating_state = dce_v8_0_set_clockgating_state, .set_powergating_state = dce_v8_0_set_powergating_state, + .dump_ip_state = NULL, }; static void
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +1 -0
@@ -9170,6 +9170,7 @@ static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .set_clockgating_state = gfx_v10_0_set_clockgating_state, .set_powergating_state = gfx_v10_0_set_powergating_state, .get_clockgating_state = gfx_v10_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c +1 -0
@@ -6174,6 +6174,7 @@ static const struct amd_ip_funcs gfx_v11_0_ip_funcs = { .set_clockgating_state = gfx_v11_0_set_clockgating_state, .set_powergating_state = gfx_v11_0_set_powergating_state, .get_clockgating_state = gfx_v11_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c +1 -0
@@ -3457,6 +3457,7 @@ static const struct amd_ip_funcs gfx_v6_0_ip_funcs = { .soft_reset = gfx_v6_0_soft_reset, .set_clockgating_state = gfx_v6_0_set_clockgating_state, .set_powergating_state = gfx_v6_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs gfx_v6_0_ring_funcs_gfx = {
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +1 -0
@@ -4977,6 +4977,7 @@ static const struct amd_ip_funcs gfx_v7_0_ip_funcs = { .soft_reset = gfx_v7_0_soft_reset, .set_clockgating_state = gfx_v7_0_set_clockgating_state, .set_powergating_state = gfx_v7_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +1 -0
@@ -6878,6 +6878,7 @@ static const struct amd_ip_funcs gfx_v8_0_ip_funcs = { .set_clockgating_state = gfx_v8_0_set_clockgating_state, .set_powergating_state = gfx_v8_0_set_powergating_state, .get_clockgating_state = gfx_v8_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +1 -0
@@ -6856,6 +6856,7 @@ static const struct amd_ip_funcs gfx_v9_0_ip_funcs = { .set_clockgating_state = gfx_v9_0_set_clockgating_state, .set_powergating_state = gfx_v9_0_set_powergating_state, .get_clockgating_state = gfx_v9_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c +1 -0
@@ -4016,6 +4016,7 @@ static const struct amd_ip_funcs gfx_v9_4_3_ip_funcs = { .set_clockgating_state = gfx_v9_4_3_set_clockgating_state, .set_powergating_state = gfx_v9_4_3_set_powergating_state, .get_clockgating_state = gfx_v9_4_3_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_compute = {
drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +1 -0
@@ -1115,6 +1115,7 @@ static const struct amd_ip_funcs gmc_v6_0_ip_funcs = { .soft_reset = gmc_v6_0_soft_reset, .set_clockgating_state = gmc_v6_0_set_clockgating_state, .set_powergating_state = gmc_v6_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_gmc_funcs gmc_v6_0_gmc_funcs = {
drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +1 -0
@@ -1354,6 +1354,7 @@ static const struct amd_ip_funcs gmc_v7_0_ip_funcs = { .soft_reset = gmc_v7_0_soft_reset, .set_clockgating_state = gmc_v7_0_set_clockgating_state, .set_powergating_state = gmc_v7_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_gmc_funcs gmc_v7_0_gmc_funcs = {
drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +1 -0
@@ -1717,6 +1717,7 @@ static const struct amd_ip_funcs gmc_v8_0_ip_funcs = { .set_clockgating_state = gmc_v8_0_set_clockgating_state, .set_powergating_state = gmc_v8_0_set_powergating_state, .get_clockgating_state = gmc_v8_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_gmc_funcs gmc_v8_0_gmc_funcs = {
drivers/gpu/drm/amd/amdgpu/iceland_ih.c +1 -0
@@ -425,6 +425,7 @@ static const struct amd_ip_funcs iceland_ih_ip_funcs = { .soft_reset = iceland_ih_soft_reset, .set_clockgating_state = iceland_ih_set_clockgating_state, .set_powergating_state = iceland_ih_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs iceland_ih_funcs = {
drivers/gpu/drm/amd/amdgpu/ih_v6_0.c +1 -0
@@ -748,6 +748,7 @@ static const struct amd_ip_funcs ih_v6_0_ip_funcs = { .set_clockgating_state = ih_v6_0_set_clockgating_state, .set_powergating_state = ih_v6_0_set_powergating_state, .get_clockgating_state = ih_v6_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs ih_v6_0_funcs = {
drivers/gpu/drm/amd/amdgpu/ih_v6_1.c +1 -0
@@ -753,6 +753,7 @@ static const struct amd_ip_funcs ih_v6_1_ip_funcs = { .set_clockgating_state = ih_v6_1_set_clockgating_state, .set_powergating_state = ih_v6_1_set_powergating_state, .get_clockgating_state = ih_v6_1_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs ih_v6_1_funcs = {
drivers/gpu/drm/amd/amdgpu/ih_v7_0.c +1 -0
@@ -749,6 +749,7 @@ static const struct amd_ip_funcs ih_v7_0_ip_funcs = { .set_clockgating_state = ih_v7_0_set_clockgating_state, .set_powergating_state = ih_v7_0_set_powergating_state, .get_clockgating_state = ih_v7_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs ih_v7_0_funcs = {
drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c +1 -0
@@ -759,6 +759,7 @@ static const struct amd_ip_funcs jpeg_v2_0_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = jpeg_v2_0_set_clockgating_state, .set_powergating_state = jpeg_v2_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs jpeg_v2_0_dec_ring_vm_funcs = {
drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c +2 -0
@@ -632,6 +632,7 @@ static const struct amd_ip_funcs jpeg_v2_5_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = jpeg_v2_5_set_clockgating_state, .set_powergating_state = jpeg_v2_5_set_powergating_state, + .dump_ip_state = NULL, };static const struct amd_ip_funcs jpeg_v2_6_ip_funcs = {
@@ -652,6 +653,7 @@ static const struct amd_ip_funcs jpeg_v2_6_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = jpeg_v2_5_set_clockgating_state, .set_powergating_state = jpeg_v2_5_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs jpeg_v2_5_dec_ring_vm_funcs = {
drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c +1 -0
@@ -557,6 +557,7 @@ static const struct amd_ip_funcs jpeg_v3_0_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = jpeg_v3_0_set_clockgating_state, .set_powergating_state = jpeg_v3_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs jpeg_v3_0_dec_ring_vm_funcs = {
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c +1 -0
@@ -719,6 +719,7 @@ static const struct amd_ip_funcs jpeg_v4_0_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = jpeg_v4_0_set_clockgating_state, .set_powergating_state = jpeg_v4_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs jpeg_v4_0_dec_ring_vm_funcs = {
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c +1 -0
@@ -1053,6 +1053,7 @@ static const struct amd_ip_funcs jpeg_v4_0_3_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = jpeg_v4_0_3_set_clockgating_state, .set_powergating_state = jpeg_v4_0_3_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs jpeg_v4_0_3_dec_ring_vm_funcs = {
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c +1 -0
@@ -762,6 +762,7 @@ static const struct amd_ip_funcs jpeg_v4_0_5_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = jpeg_v4_0_5_set_clockgating_state, .set_powergating_state = jpeg_v4_0_5_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs jpeg_v4_0_5_dec_ring_vm_funcs = {
drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c +1 -0
@@ -513,6 +513,7 @@ static const struct amd_ip_funcs jpeg_v5_0_0_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = jpeg_v5_0_0_set_clockgating_state, .set_powergating_state = jpeg_v5_0_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs jpeg_v5_0_0_dec_ring_vm_funcs = {
drivers/gpu/drm/amd/amdgpu/mes_v10_1.c +1 -0
@@ -1176,6 +1176,7 @@ static const struct amd_ip_funcs mes_v10_1_ip_funcs = { .hw_fini = mes_v10_1_hw_fini, .suspend = mes_v10_1_suspend, .resume = mes_v10_1_resume, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version mes_v10_1_ip_block = {
drivers/gpu/drm/amd/amdgpu/mes_v11_0.c +1 -0
@@ -1334,6 +1334,7 @@ static const struct amd_ip_funcs mes_v11_0_ip_funcs = { .hw_fini = mes_v11_0_hw_fini, .suspend = mes_v11_0_suspend, .resume = mes_v11_0_resume, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version mes_v11_0_ip_block = {
drivers/gpu/drm/amd/amdgpu/navi10_ih.c +1 -0
@@ -713,6 +713,7 @@ static const struct amd_ip_funcs navi10_ih_ip_funcs = { .set_clockgating_state = navi10_ih_set_clockgating_state, .set_powergating_state = navi10_ih_set_powergating_state, .get_clockgating_state = navi10_ih_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs navi10_ih_funcs = {
drivers/gpu/drm/amd/amdgpu/nv.c +1 -0
@@ -1131,4 +1131,5 @@ static const struct amd_ip_funcs nv_common_ip_funcs = { .set_clockgating_state = nv_common_set_clockgating_state, .set_powergating_state = nv_common_set_powergating_state, .get_clockgating_state = nv_common_get_clockgating_state, + .dump_ip_state = NULL, };
drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +1 -0
@@ -1113,6 +1113,7 @@ static const struct amd_ip_funcs sdma_v2_4_ip_funcs = { .soft_reset = sdma_v2_4_soft_reset, .set_clockgating_state = sdma_v2_4_set_clockgating_state, .set_powergating_state = sdma_v2_4_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = {
drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +1 -0
@@ -1553,6 +1553,7 @@ static const struct amd_ip_funcs sdma_v3_0_ip_funcs = { .set_clockgating_state = sdma_v3_0_set_clockgating_state, .set_powergating_state = sdma_v3_0_set_powergating_state, .get_clockgating_state = sdma_v3_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = {
drivers/gpu/drm/amd/amdgpu/si.c +1 -0
@@ -2706,6 +2706,7 @@ static const struct amd_ip_funcs si_common_ip_funcs = { .soft_reset = si_common_soft_reset, .set_clockgating_state = si_common_set_clockgating_state, .set_powergating_state = si_common_set_powergating_state, + .dump_ip_state = NULL, }; static const struct amdgpu_ip_block_version si_common_ip_block =
drivers/gpu/drm/amd/amdgpu/si_dma.c +1 -0
@@ -708,6 +708,7 @@ static const struct amd_ip_funcs si_dma_ip_funcs = { .soft_reset = si_dma_soft_reset, .set_clockgating_state = si_dma_set_clockgating_state, .set_powergating_state = si_dma_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs si_dma_ring_funcs = {
drivers/gpu/drm/amd/amdgpu/si_ih.c +1 -0
@@ -296,6 +296,7 @@ static const struct amd_ip_funcs si_ih_ip_funcs = { .soft_reset = si_ih_soft_reset, .set_clockgating_state = si_ih_set_clockgating_state, .set_powergating_state = si_ih_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs si_ih_funcs = {
drivers/gpu/drm/amd/amdgpu/soc15.c +1 -0
@@ -1501,4 +1501,5 @@ static const struct amd_ip_funcs soc15_common_ip_funcs = { .set_clockgating_state = soc15_common_set_clockgating_state, .set_powergating_state = soc15_common_set_powergating_state, .get_clockgating_state= soc15_common_get_clockgating_state, + .dump_ip_state = NULL, };
drivers/gpu/drm/amd/amdgpu/soc21.c +1 -0
@@ -985,4 +985,5 @@ static const struct amd_ip_funcs soc21_common_ip_funcs = { .set_clockgating_state = soc21_common_set_clockgating_state, .set_powergating_state = soc21_common_set_powergating_state, .get_clockgating_state = soc21_common_get_clockgating_state, + .dump_ip_state = NULL, };
drivers/gpu/drm/amd/amdgpu/tonga_ih.c +1 -0
@@ -486,6 +486,7 @@ static const struct amd_ip_funcs tonga_ih_ip_funcs = { .post_soft_reset = tonga_ih_post_soft_reset, .set_clockgating_state = tonga_ih_set_clockgating_state, .set_powergating_state = tonga_ih_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ih_funcs tonga_ih_funcs = {
drivers/gpu/drm/amd/amdgpu/uvd_v3_1.c +1 -0
@@ -819,6 +819,7 @@ static const struct amd_ip_funcs uvd_v3_1_ip_funcs = { .soft_reset = uvd_v3_1_soft_reset, .set_clockgating_state = uvd_v3_1_set_clockgating_state, .set_powergating_state = uvd_v3_1_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version uvd_v3_1_ip_block = {
drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c +1 -0
@@ -769,6 +769,7 @@ static const struct amd_ip_funcs uvd_v4_2_ip_funcs = { .soft_reset = uvd_v4_2_soft_reset, .set_clockgating_state = uvd_v4_2_set_clockgating_state, .set_powergating_state = uvd_v4_2_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = {
drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +1 -0
@@ -877,6 +877,7 @@ static const struct amd_ip_funcs uvd_v5_0_ip_funcs = { .set_clockgating_state = uvd_v5_0_set_clockgating_state, .set_powergating_state = uvd_v5_0_set_powergating_state, .get_clockgating_state = uvd_v5_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs uvd_v5_0_ring_funcs = {
drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +1 -0
@@ -1545,6 +1545,7 @@ static const struct amd_ip_funcs uvd_v6_0_ip_funcs = { .set_clockgating_state = uvd_v6_0_set_clockgating_state, .set_powergating_state = uvd_v6_0_set_powergating_state, .get_clockgating_state = uvd_v6_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs uvd_v6_0_ring_phys_funcs = {
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +1 -0
@@ -626,6 +626,7 @@ static const struct amd_ip_funcs vce_v2_0_ip_funcs = { .soft_reset = vce_v2_0_soft_reset, .set_clockgating_state = vce_v2_0_set_clockgating_state, .set_powergating_state = vce_v2_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs vce_v2_0_ring_funcs = {
drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +1 -0
@@ -913,6 +913,7 @@ static const struct amd_ip_funcs vce_v3_0_ip_funcs = { .set_clockgating_state = vce_v3_0_set_clockgating_state, .set_powergating_state = vce_v3_0_set_powergating_state, .get_clockgating_state = vce_v3_0_get_clockgating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs vce_v3_0_ring_phys_funcs = {
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +1 -0
@@ -1902,6 +1902,7 @@ static const struct amd_ip_funcs vcn_v1_0_ip_funcs = { .post_soft_reset = NULL /* vcn_v1_0_post_soft_reset */, .set_clockgating_state = vcn_v1_0_set_clockgating_state, .set_powergating_state = vcn_v1_0_set_powergating_state, + .dump_ip_state = NULL, };/*
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c +1 -0
@@ -2008,6 +2008,7 @@ static const struct amd_ip_funcs vcn_v2_0_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = vcn_v2_0_set_clockgating_state, .set_powergating_state = vcn_v2_0_set_powergating_state, + .dump_ip_state = NULL, };static const struct amdgpu_ring_funcs vcn_v2_0_dec_ring_vm_funcs = {
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +2 -0
@@ -1901,6 +1901,7 @@ static const struct amd_ip_funcs vcn_v2_5_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = vcn_v2_5_set_clockgating_state, .set_powergating_state = vcn_v2_5_set_powergating_state, + .dump_ip_state = NULL, };static const struct amd_ip_funcs vcn_v2_6_ip_funcs = {
@@ -1921,6 +1922,7 @@ static const struct amd_ip_funcs vcn_v2_6_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = vcn_v2_5_set_clockgating_state, .set_powergating_state = vcn_v2_5_set_powergating_state, + .dump_ip_state = NULL, }; const struct amdgpu_ip_block_version vcn_v2_5_ip_block =
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +1 -0
@@ -2230,6 +2230,7 @@ static const struct amd_ip_funcs vcn_v3_0_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = vcn_v3_0_set_clockgating_state, .set_powergating_state = vcn_v3_0_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version vcn_v3_0_ip_block = {
drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c +1 -0
@@ -2130,6 +2130,7 @@ static const struct amd_ip_funcs vcn_v4_0_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = vcn_v4_0_set_clockgating_state, .set_powergating_state = vcn_v4_0_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version vcn_v4_0_ip_block = {
drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c +1 -0
@@ -1660,6 +1660,7 @@ static const struct amd_ip_funcs vcn_v4_0_3_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = vcn_v4_0_3_set_clockgating_state, .set_powergating_state = vcn_v4_0_3_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version vcn_v4_0_3_ip_block = {
drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c +1 -0
@@ -1752,6 +1752,7 @@ static const struct amd_ip_funcs vcn_v4_0_5_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = vcn_v4_0_5_set_clockgating_state, .set_powergating_state = vcn_v4_0_5_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version vcn_v4_0_5_ip_block = {
drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c +1 -0
@@ -1328,6 +1328,7 @@ static const struct amd_ip_funcs vcn_v5_0_0_ip_funcs = { .post_soft_reset = NULL, .set_clockgating_state = vcn_v5_0_0_set_clockgating_state, .set_powergating_state = vcn_v5_0_0_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version vcn_v5_0_0_ip_block = {
drivers/gpu/drm/amd/amdgpu/vi.c +1 -0
@@ -2058,6 +2058,7 @@ static const struct amd_ip_funcs vi_common_ip_funcs = { .set_clockgating_state = vi_common_set_clockgating_state, .set_powergating_state = vi_common_set_powergating_state, .get_clockgating_state = vi_common_get_clockgating_state, + .dump_ip_state = NULL, }; static const struct amdgpu_ip_block_version vi_common_ip_block =
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +1 -0
@@ -3111,6 +3111,7 @@ static const struct amd_ip_funcs amdgpu_dm_funcs = { .soft_reset = dm_soft_reset, .set_clockgating_state = dm_set_clockgating_state, .set_powergating_state = dm_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version dm_ip_block = {
drivers/gpu/drm/amd/include/amd_shared.h +1 -0
@@ -321,6 +321,7 @@ struct amd_ip_funcs {int (*set_powergating_state)(void *handle,enum amd_powergating_state state);void (*get_clockgating_state)(void *handle,u64 *flags); +void (*dump_ip_state)(void *handle); };
drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c +1 -0
@@ -3316,6 +3316,7 @@ static const struct amd_ip_funcs kv_dpm_ip_funcs = { .soft_reset = kv_dpm_soft_reset, .set_clockgating_state = kv_dpm_set_clockgating_state, .set_powergating_state = kv_dpm_set_powergating_state, + .dump_ip_state = NULL, };const struct amdgpu_ip_block_version kv_smu_ip_block = {
drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c +1 -0
@@ -8060,6 +8060,7 @@ static const struct amd_ip_funcs si_dpm_ip_funcs = { .soft_reset = si_dpm_soft_reset, .set_clockgating_state = si_dpm_set_clockgating_state, .set_powergating_state = si_dpm_set_powergating_state, + .dump_ip_state = NULL, }; const struct amdgpu_ip_block_version si_smu_ip_block =
drivers/gpu/drm/amd/pm/powerplay/amd_powerplay.c +1 -0
@@ -302,6 +302,7 @@ static const struct amd_ip_funcs pp_ip_funcs = { .soft_reset = pp_sw_reset, .set_clockgating_state = pp_set_clockgating_state, .set_powergating_state = pp_set_powergating_state, + .dump_ip_state = NULL, }; const struct amdgpu_ip_block_version pp_smu_ip_block =
Signature · 1 line
2.34.1
PATCH v2 2/2 drm/amdgpu: Add support of gfx10 register dump
Adding initial set of registers for ipdump during devcoredump starting with gfx10 gc registers.
ip dump is triggered when gpu reset happens via devcoredump and the memory is allocated by each ip and is freed once the dump is complete by devcoredump.
--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 16 +++ .../gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 22 +++ drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 127 +++++++++++++++++- .../include/asic_reg/gc/gc_10_1_0_offset.h | 12 ++ 4 files changed, 176 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +16 -0
@@ -139,6 +139,18 @@ enum amdgpu_ss {AMDGPU_SS_DRV_UNLOAD }; +struct hwip_reg_entry { +u32 hwip; +u32 inst; +u32 seg; +u32 reg_offset; +}; + +struct reg_pair { +u32 offset; +u32 value; +}; +struct amdgpu_watchdog_timer {bool timeout_fatal_disable;uint32_t period;/* maxCycles = (1 << period), the number of cycles before a timeout */
@@ -1152,6 +1164,10 @@ struct amdgpu_device {bool debug_largebar;bool debug_disable_soft_recovery;bool debug_use_vram_fw_buf; + +/* IP register dump */ +struct reg_pair *ip_dump; +uint32_t num_regs; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +22 -0
@@ -261,6 +261,18 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,drm_printf (&p,"Faulty page starting at address: 0x%016llx\n" , fault_info->addr);drm_printf (&p,"Protection fault status register: 0x%x\n\n" , fault_info->status); +/* Add IP dump for each ip */ +if (coredump->adev->ip_dump !=NULL ) { +struct reg_pair *pair; + + pair = (struct reg_pair *)coredump->adev->ip_dump; +drm_printf (&p,"IP register dump\n" ); +drm_printf (&p,"Offset \t Value\n" ); +for (int i =0 ; i < coredump->adev->num_regs; i++) +drm_printf (&p,"0x%04x \t 0x%08x\n" , pair[i].offset, pair[i].value); +drm_printf (&p,"\n" ); + } +/* Add ring buffer information */ drm_printf (&p,"Ring buffer information\n" );for (int i =0 ; i < coredump->adev->num_rings; i++) {
@@ -299,6 +311,11 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,static void amdgpu_devcoredump_free (void *data) { +struct amdgpu_coredump_info *temp = data; + +kfree (temp->adev->ip_dump); + temp->adev->ip_dump =NULL ; + temp->adev->num_regs =0 ;kfree (data); }
@@ -337,6 +354,11 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost, coredump->adev = adev; +/* Trigger ip dump here to capture the value of registers */ +for (int i =0 ; i < adev->num_ip_blocks; i++) +if (adev->ip_blocks[i].version->funcs->dump_ip_state) + adev->ip_blocks[i].version->funcs->dump_ip_state ((void *)adev); +ktime_get_ts64 (&coredump->reset_time); dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +126 -1
145 lines not on this page · raw message
drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h +12 -0
@@ -4830,6 +4830,8 @@#define mmCP_ECC_FIRSTOCCURRENCE_RING2_BASE_IDX 0#define mmGB_EDC_MODE 0x1e1e#define mmGB_EDC_MODE_BASE_IDX 0 +#define mmCP_DEBUG 0x1e1f +#define mmCP_DEBUG_BASE_IDX 0#define mmCP_FETCHER_SOURCE 0x1e22#define mmCP_FETCHER_SOURCE_BASE_IDX 0#define mmCP_PQ_WPTR_POLL_CNTL 0x1e23
@@ -7778,6 +7780,8 @@#define mmCP_MES_DOORBELL_CONTROL5_BASE_IDX 1#define mmCP_MES_DOORBELL_CONTROL6 0x2841#define mmCP_MES_DOORBELL_CONTROL6_BASE_IDX 1 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x2842 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1#define mmCP_MES_GP0_LO 0x2843#define mmCP_MES_GP0_LO_BASE_IDX 1#define mmCP_MES_GP0_HI 0x2844
@@ -9332,10 +9336,16 @@#define mmRLC_LB_CNTR_INIT_1_BASE_IDX 1#define mmRLC_LB_CNTR_1 0x4c1c#define mmRLC_LB_CNTR_1_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_ADDR 0x4c1d +#define mmRLC_GPM_DEBUG_INST_ADDR_BASE_IDX 1#define mmRLC_JUMP_TABLE_RESTORE 0x4c1e#define mmRLC_JUMP_TABLE_RESTORE_BASE_IDX 1#define mmRLC_PG_DELAY_2 0x4c1f#define mmRLC_PG_DELAY_2_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_A 0x4c22 +#define mmRLC_GPM_DEBUG_INST_A_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_B 0x4c23 +#define mmRLC_GPM_DEBUG_INST_B_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_LSB 0x4c24#define mmRLC_GPU_CLOCK_COUNT_LSB_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_MSB 0x4c25
@@ -9720,6 +9730,8 @@#define mmRLC_SPM_THREAD_TRACE_CTRL_BASE_IDX 1#define mmRLC_LB_CNTR_2 0x4de7#define mmRLC_LB_CNTR_2_BASE_IDX 1 +#define mmRLC_LX6_CORE_PDEBUG_INST 0x4deb +#define mmRLC_LX6_CORE_PDEBUG_INST_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_CTRL 0x4df1#define mmRLC_CPAXI_DOORBELL_MON_CTRL_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_STAT 0x4df2
Signature · 1 line
2.34.1
Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
Quoted 9 lines · Sunil Khatri
Adding initial set of registers for ipdump during devcoredump starting with gfx10 gc registers.
ip dump is triggered when gpu reset happens via devcoredump and the memory is allocated by each ip and is freed once the dump is complete by devcoredump.
Quoted 4 lines
Signed-off-by Sunil Khatri <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 16 +++ .../gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 22 +++
I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Quoted 74 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 127 +++++++++++++++++- .../include/asic_reg/gc/gc_10_1_0_offset.h | 12 ++ 4 files changed, 176 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +16
-0@@ -139,6 +139,18 @@ enum amdgpu_ss {AMDGPU_SS_DRV_UNLOAD }; +struct hwip_reg_entry { +u32 hwip; +u32 inst; +u32 seg; +u32 reg_offset; +}; + +struct reg_pair { +u32 offset; +u32 value; +}; +struct amdgpu_watchdog_timer {bool timeout_fatal_disable;uint32_t period;/* maxCycles = (1 << period), the number of cycles before a timeout */ @@ -1152,6 +1164,10 @@ struct amdgpu_device {bool debug_largebar;bool debug_disable_soft_recovery;bool debug_use_vram_fw_buf; + +/* IP register dump */ +struct reg_pair *ip_dump; +uint32_t num_regs; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +17
-0@@ -261,6 +261,18 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,drm_printf (&p,"Faulty page starting at address: 0x%016llx\n" , fault_info->addr);drm_printf (&p,"Protection fault status register: 0x%x\n\n" , fault_info->status); +/* Add IP dump for each ip */ +if (coredump->adev->ip_dump !=NULL ) { +struct reg_pair *pair; + + pair = (struct reg_pair *)coredump->adev->ip_dump; +drm_printf (&p,"IP register dump\n" ); +drm_printf (&p,"Offset \t Value\n" ); +for (int i =0 ; i < coredump->adev->num_regs; i++) +drm_printf (&p,"0x%04x \t 0x%08x\n" , pair[i].offset, pair[i].value); +drm_printf (&p,"\n" ); + } +/* Add ring buffer information */ drm_printf (&p,"Ring buffer information\n" );for (int i =0 ; i < coredump->adev->num_rings; i++) {@@ -299,6 +311,11 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,static void amdgpu_devcoredump_free (void *data) { +struct amdgpu_coredump_info *temp = data; + +kfree (temp->adev->ip_dump); + temp->adev->ip_dump =NULL ; + temp->adev->num_regs =0 ;kfree (data); }
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost, · line 361
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +5
-0@@ -337,6 +354,11 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost, coredump->adev = adev; +/* Trigger ip dump here to capture the value of registers */ +for (int i =0 ; i < adev->num_ip_blocks; i++) +if (adev->ip_blocks[i].version->funcs->dump_ip_state) + adev->ip_blocks[i].version->funcs->dump_ip_state ((void *)adev); +
This seems too complicated. I think it would be easier to
Quoted 128 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +0
-0ktime_get_ts64 (&coredump->reset_time); dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +110
-0@@ -276,6 +276,99 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");MODULE_FIRMWARE ("amdgpu/gc_10_3_7_mec2.bin" );MODULE_FIRMWARE ("amdgpu/gc_10_3_7_rlc.bin" ); +static const struct hwip_reg_entry gc_reg_list_10_1[] = { + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS2) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS3) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_ERROR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_HPD_STATUS0) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPF_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPG_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_VM_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmPA_CL_CNTL_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRMI_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_DCACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_ICACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQG_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmTCP_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmWD_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_DEBUG) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC1_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC2_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_ME_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_PFP_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_COMMAND) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_MESSAGE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_1) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_3) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_4) }, + {SOC15_REG_ENTRY (GC ,0 , mmSMU_RLC_RESPONSE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_GPM_STAT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SPP_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_BOOTLOAD_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_INT_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_GENERAL_6) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_A) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_B) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_ADDR) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_LX6_CORE_PDEBUG_INST) } +}; +static const struct soc15_reg_golden golden_settings_gc_10_1[] = {SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCB_HW_CONTROL_4,0xffffffff ,0x00400014 ),SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCGTT_CPF_CLK_CTRL,0xfcff8fff ,0xf8000100 ),@@ -9154,6 +9247,38 @@ static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)amdgpu_ring_write (ring, gcr_cntl);/* GCR_CNTL */ } +static void gfx_v10_ip_dump (void *handle) +{ +struct amdgpu_device *adev = (struct amdgpu_device *)handle; +uint32_t reg_count =ARRAY_SIZE (gc_reg_list_10_1); +struct reg_pair *ptr, *gfx10_pair; +uint32_t new_count; + +/* +* Allocate memory based on the no of registers for each block. +* Since adev->ip_dump is common pointer and we add more +* memory to it as we move through different IP's of the ASIC. +* +* This memory needs to be freed after gpu reset once the ip +* dump is done in the devcoredump. +*/ + new_count = adev->num_regs + reg_count; + ptr =krealloc_array (adev->ip_dump, new_count,sizeof (struct reg_pair ),GFP_KERNEL );
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring) · line 9276
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +10
-0+if (ptr) { + adev->ip_dump = ptr; + gfx10_pair = &ptr[adev->num_regs]; + adev->num_regs = new_count; + }else { +DRM_ERROR ("Failed to allocate memory for IP Dump\n" ); +return ; + } + +for (uint32_t i =0 ; i < reg_count; i++) {
Some compilers don't like variables defined within code.
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring) · line 9280
Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4
-0+ gfx10_pair[i].offset = gc_reg_list_10_1[i].reg_offset; + gfx10_pair[i].value =RREG32 (SOC15_REG_ENTRY_OFFSET (gc_reg_list_10_1[i])); + } +}
This seems too complicated. Just pass drm_printer to the ip_dump callbacks and let each IP dump their state directly. E.g.,
for (i = 0; i < adev->num_ip_blocks; i++) {
if (adev->ip_blocks[i].version->funcs->dump_ip_state) {
drm_printf(&p, "$s IP state\n",adev->ip_blocks[i].version->funcs->name);
adev->ip_blocks[i].version->funcs->dump_ip_state((void *)adev, &p);
}
}Alex
Quoted 64 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1+static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .name ="gfx_v10_0" , .early_init = gfx_v10_0_early_init,@@ -9170,7 +9295,7 @@ static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .set_clockgating_state = gfx_v10_0_set_clockgating_state, .set_powergating_state = gfx_v10_0_set_powergating_state, .get_clockgating_state = gfx_v10_0_get_clockgating_state, - .dump_ip_state = NULL, + .dump_ip_state = gfx_v10_ip_dump, };static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h +12
-1@@ -4830,6 +4830,8 @@#define mmCP_ECC_FIRSTOCCURRENCE_RING2_BASE_IDX 0#define mmGB_EDC_MODE 0x1e1e#define mmGB_EDC_MODE_BASE_IDX 0 +#define mmCP_DEBUG 0x1e1f +#define mmCP_DEBUG_BASE_IDX 0#define mmCP_FETCHER_SOURCE 0x1e22#define mmCP_FETCHER_SOURCE_BASE_IDX 0#define mmCP_PQ_WPTR_POLL_CNTL 0x1e23@@ -7778,6 +7780,8 @@#define mmCP_MES_DOORBELL_CONTROL5_BASE_IDX 1#define mmCP_MES_DOORBELL_CONTROL6 0x2841#define mmCP_MES_DOORBELL_CONTROL6_BASE_IDX 1 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x2842 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1#define mmCP_MES_GP0_LO 0x2843#define mmCP_MES_GP0_LO_BASE_IDX 1#define mmCP_MES_GP0_HI 0x2844@@ -9332,10 +9336,16 @@#define mmRLC_LB_CNTR_INIT_1_BASE_IDX 1#define mmRLC_LB_CNTR_1 0x4c1c#define mmRLC_LB_CNTR_1_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_ADDR 0x4c1d +#define mmRLC_GPM_DEBUG_INST_ADDR_BASE_IDX 1#define mmRLC_JUMP_TABLE_RESTORE 0x4c1e#define mmRLC_JUMP_TABLE_RESTORE_BASE_IDX 1#define mmRLC_PG_DELAY_2 0x4c1f#define mmRLC_PG_DELAY_2_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_A 0x4c22 +#define mmRLC_GPM_DEBUG_INST_A_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_B 0x4c23 +#define mmRLC_GPM_DEBUG_INST_B_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_LSB 0x4c24#define mmRLC_GPU_CLOCK_COUNT_LSB_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_MSB 0x4c25@@ -9720,6 +9730,8 @@#define mmRLC_SPM_THREAD_TRACE_CTRL_BASE_IDX 1#define mmRLC_LB_CNTR_2 0x4de7#define mmRLC_LB_CNTR_2_BASE_IDX 1 +#define mmRLC_LX6_CORE_PDEBUG_INST 0x4deb +#define mmRLC_LX6_CORE_PDEBUG_INST_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_CTRL 0x4df1#define mmRLC_CPAXI_DOORBELL_MON_CTRL_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_STAT 0x4df2 --2.34.1
Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
Quoted 3 lines · Alex Deucher
I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Sure would push the v3 with the changes.
Regards
Sunil
Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
Quoted 10 lines · Sunil
On 4/12/2024 8:50 PM, Alex Deucher wrote:
I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Sure would push the v3 with the changes.
note my other comments further down as well.
Alex
Quoted 4 lines
Regards
Sunil
Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
In reply to Alex DeucherQuoted 9 lines · Alex Deucher
Quoted 8 lines · Sunil Khatri
Adding initial set of registers for ipdump during devcoredump starting with gfx10 gc registers.
ip dump is triggered when gpu reset happens via devcoredump and the memory is allocated by each ip and is freed once the dump is complete by devcoredump.
Quoted 7 lines
Quoted 4 lines
Signed-off-by Sunil Khatri <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 16 +++ .../gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 22 +++I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Sure
Quoted 77 lines
Quoted 77 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 127 +++++++++++++++++- .../include/asic_reg/gc/gc_10_1_0_offset.h | 12 ++ 4 files changed, 176 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +16
-0@@ -139,6 +139,18 @@ enum amdgpu_ss {AMDGPU_SS_DRV_UNLOAD }; +struct hwip_reg_entry { +u32 hwip; +u32 inst; +u32 seg; +u32 reg_offset; +}; + +struct reg_pair { +u32 offset; +u32 value; +}; +struct amdgpu_watchdog_timer {bool timeout_fatal_disable;uint32_t period;/* maxCycles = (1 << period), the number of cycles before a timeout */ @@ -1152,6 +1164,10 @@ struct amdgpu_device {bool debug_largebar;bool debug_disable_soft_recovery;bool debug_use_vram_fw_buf; + +/* IP register dump */ +struct reg_pair *ip_dump; +uint32_t num_regs; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +17
-0@@ -261,6 +261,18 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,drm_printf (&p,"Faulty page starting at address: 0x%016llx\n" , fault_info->addr);drm_printf (&p,"Protection fault status register: 0x%x\n\n" , fault_info->status); +/* Add IP dump for each ip */ +if (coredump->adev->ip_dump !=NULL ) { +struct reg_pair *pair; + + pair = (struct reg_pair *)coredump->adev->ip_dump; +drm_printf (&p,"IP register dump\n" ); +drm_printf (&p,"Offset \t Value\n" ); +for (int i =0 ; i < coredump->adev->num_regs; i++) +drm_printf (&p,"0x%04x \t 0x%08x\n" , pair[i].offset, pair[i].value); +drm_printf (&p,"\n" ); + } +/* Add ring buffer information */ drm_printf (&p,"Ring buffer information\n" );for (int i =0 ; i < coredump->adev->num_rings; i++) {@@ -299,6 +311,11 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,static void amdgpu_devcoredump_free (void *data) { +struct amdgpu_coredump_info *temp = data; + +kfree (temp->adev->ip_dump); + temp->adev->ip_dump =NULL ; + temp->adev->num_regs =0 ;kfree (data); }@@ -337,6 +354,11 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost, · line 361
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +5
-0coredump->adev = adev; +/* Trigger ip dump here to capture the value of registers */ +for (int i =0 ; i < adev->num_ip_blocks; i++) +if (adev->ip_blocks[i].version->funcs->dump_ip_state) + adev->ip_blocks[i].version->funcs->dump_ip_state ((void *)adev); +This seems too complicated. I think it would be easier to
This is how all other per IP functions are called. What do you suggest ?
Quoted 132 lines
Quoted 132 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +0
-0ktime_get_ts64 (&coredump->reset_time); dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +113
-0@@ -276,6 +276,99 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");MODULE_FIRMWARE ("amdgpu/gc_10_3_7_mec2.bin" );MODULE_FIRMWARE ("amdgpu/gc_10_3_7_rlc.bin" ); +static const struct hwip_reg_entry gc_reg_list_10_1[] = { + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS2) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS3) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_ERROR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_HPD_STATUS0) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPF_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPG_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_VM_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmPA_CL_CNTL_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRMI_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_DCACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_ICACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQG_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmTCP_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmWD_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_DEBUG) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC1_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC2_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_ME_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_PFP_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_COMMAND) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_MESSAGE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_1) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_3) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_4) }, + {SOC15_REG_ENTRY (GC ,0 , mmSMU_RLC_RESPONSE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_GPM_STAT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SPP_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_BOOTLOAD_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_INT_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_GENERAL_6) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_A) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_B) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_ADDR) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_LX6_CORE_PDEBUG_INST) } +}; +static const struct soc15_reg_golden golden_settings_gc_10_1[] = {SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCB_HW_CONTROL_4,0xffffffff ,0x00400014 ),SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCGTT_CPF_CLK_CTRL,0xfcff8fff ,0xf8000100 ),@@ -9154,6 +9247,38 @@ static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)amdgpu_ring_write (ring, gcr_cntl);/* GCR_CNTL */ } +static void gfx_v10_ip_dump (void *handle) +{ +struct amdgpu_device *adev = (struct amdgpu_device *)handle; +uint32_t reg_count =ARRAY_SIZE (gc_reg_list_10_1); +struct reg_pair *ptr, *gfx10_pair; +uint32_t new_count; + +/* +* Allocate memory based on the no of registers for each block. +* Since adev->ip_dump is common pointer and we add more +* memory to it as we move through different IP's of the ASIC. +* +* This memory needs to be freed after gpu reset once the ip +* dump is done in the devcoredump. +*/ + new_count = adev->num_regs + reg_count; + ptr =krealloc_array (adev->ip_dump, new_count,sizeof (struct reg_pair ),GFP_KERNEL ); +if (ptr) { + adev->ip_dump = ptr; + gfx10_pair = &ptr[adev->num_regs];
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring) · line 9276
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +7
-0+ adev->num_regs = new_count; + }else { +DRM_ERROR ("Failed to allocate memory for IP Dump\n" ); +return ; + } + +for (uint32_t i =0 ; i < reg_count; i++) {Some compilers don't like variables defined within code.
Sure
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring) · line 9280
Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4
-0+ gfx10_pair[i].offset = gc_reg_list_10_1[i].reg_offset; + gfx10_pair[i].value =RREG32 (SOC15_REG_ENTRY_OFFSET (gc_reg_list_10_1[i])); + } +}This seems too complicated. Just pass
drm_printerto theip_dumpcallbacks and let each IP dump their state directly. E.g.,
Quoted 8 lines
for (i = 0; i < adev->num_ip_blocks; i++) { if (adev->ip_blocks[i].version->funcs->dump_ip_state) { drm_printf(&p, "$s IP state\n",adev->ip_blocks[i].version->funcs->name);
adev->ip_blocks[i].version->funcs->dump_ip_state((void *)adev, &p); } }
I did try this but there is a problem in that approach. drm_printer is created on the buffer which is provided by "amdgpu_devcoredump_read" and "amdgpu_devcoredump_read" is only called when the user opens the file
cat /sys/class/devcoredump/devcd1/data and by that time gpu is reset and registers dont hold the value of interest. I tried this approach first and got all FFFF or reset values of the registers. We need to capture the values in "amdgpu_coredump" which is immediately called after GPU reset and values are what we need for debugging.
Crux is we need to capture data after gpu reset but read depends on when user open the file which is too late and hence we need to capture all the data in advance immediately after reset before recovery.
Regards
Sunil Khatri
Quoted 67 lines
Alex
Quoted 64 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1+static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .name ="gfx_v10_0" , .early_init = gfx_v10_0_early_init,@@ -9170,7 +9295,7 @@ static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .set_clockgating_state = gfx_v10_0_set_clockgating_state, .set_powergating_state = gfx_v10_0_set_powergating_state, .get_clockgating_state = gfx_v10_0_get_clockgating_state, - .dump_ip_state = NULL, + .dump_ip_state = gfx_v10_ip_dump, };static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h +12
-1@@ -4830,6 +4830,8 @@#define mmCP_ECC_FIRSTOCCURRENCE_RING2_BASE_IDX 0#define mmGB_EDC_MODE 0x1e1e#define mmGB_EDC_MODE_BASE_IDX 0 +#define mmCP_DEBUG 0x1e1f +#define mmCP_DEBUG_BASE_IDX 0#define mmCP_FETCHER_SOURCE 0x1e22#define mmCP_FETCHER_SOURCE_BASE_IDX 0#define mmCP_PQ_WPTR_POLL_CNTL 0x1e23@@ -7778,6 +7780,8 @@#define mmCP_MES_DOORBELL_CONTROL5_BASE_IDX 1#define mmCP_MES_DOORBELL_CONTROL6 0x2841#define mmCP_MES_DOORBELL_CONTROL6_BASE_IDX 1 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x2842 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1#define mmCP_MES_GP0_LO 0x2843#define mmCP_MES_GP0_LO_BASE_IDX 1#define mmCP_MES_GP0_HI 0x2844@@ -9332,10 +9336,16 @@#define mmRLC_LB_CNTR_INIT_1_BASE_IDX 1#define mmRLC_LB_CNTR_1 0x4c1c#define mmRLC_LB_CNTR_1_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_ADDR 0x4c1d +#define mmRLC_GPM_DEBUG_INST_ADDR_BASE_IDX 1#define mmRLC_JUMP_TABLE_RESTORE 0x4c1e#define mmRLC_JUMP_TABLE_RESTORE_BASE_IDX 1#define mmRLC_PG_DELAY_2 0x4c1f#define mmRLC_PG_DELAY_2_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_A 0x4c22 +#define mmRLC_GPM_DEBUG_INST_A_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_B 0x4c23 +#define mmRLC_GPM_DEBUG_INST_B_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_LSB 0x4c24#define mmRLC_GPU_CLOCK_COUNT_LSB_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_MSB 0x4c25@@ -9720,6 +9730,8 @@#define mmRLC_SPM_THREAD_TRACE_CTRL_BASE_IDX 1#define mmRLC_LB_CNTR_2 0x4de7#define mmRLC_LB_CNTR_2_BASE_IDX 1 +#define mmRLC_LX6_CORE_PDEBUG_INST 0x4deb +#define mmRLC_LX6_CORE_PDEBUG_INST_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_CTRL 0x4df1#define mmRLC_CPAXI_DOORBELL_MON_CTRL_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_STAT 0x4df2 --2.34.1
Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
Quoted 277 lines · Sunil
Quoted 9 lines · Alex Deucher
Quoted 8 lines · Sunil Khatri
Adding initial set of registers for ipdump during devcoredump starting with gfx10 gc registers.
ip dump is triggered when gpu reset happens via devcoredump and the memory is allocated by each ip and is freed once the dump is complete by devcoredump.
Quoted 7 lines
Quoted 4 lines
Signed-off-by Sunil Khatri <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 16 +++ .../gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 22 +++I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Sure
Quoted 77 lines
Quoted 77 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 127 +++++++++++++++++- .../include/asic_reg/gc/gc_10_1_0_offset.h | 12 ++ 4 files changed, 176 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +16
-0@@ -139,6 +139,18 @@ enum amdgpu_ss {AMDGPU_SS_DRV_UNLOAD }; +struct hwip_reg_entry { +u32 hwip; +u32 inst; +u32 seg; +u32 reg_offset; +}; + +struct reg_pair { +u32 offset; +u32 value; +}; +struct amdgpu_watchdog_timer {bool timeout_fatal_disable;uint32_t period;/* maxCycles = (1 << period), the number of cycles before a timeout */ @@ -1152,6 +1164,10 @@ struct amdgpu_device {bool debug_largebar;bool debug_disable_soft_recovery;bool debug_use_vram_fw_buf; + +/* IP register dump */ +struct reg_pair *ip_dump; +uint32_t num_regs; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +17
-0@@ -261,6 +261,18 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,drm_printf (&p,"Faulty page starting at address: 0x%016llx\n" , fault_info->addr);drm_printf (&p,"Protection fault status register: 0x%x\n\n" , fault_info->status); +/* Add IP dump for each ip */ +if (coredump->adev->ip_dump !=NULL ) { +struct reg_pair *pair; + + pair = (struct reg_pair *)coredump->adev->ip_dump; +drm_printf (&p,"IP register dump\n" ); +drm_printf (&p,"Offset \t Value\n" ); +for (int i =0 ; i < coredump->adev->num_regs; i++) +drm_printf (&p,"0x%04x \t 0x%08x\n" , pair[i].offset, pair[i].value); +drm_printf (&p,"\n" ); + } +/* Add ring buffer information */ drm_printf (&p,"Ring buffer information\n" );for (int i =0 ; i < coredump->adev->num_rings; i++) {@@ -299,6 +311,11 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,static void amdgpu_devcoredump_free (void *data) { +struct amdgpu_coredump_info *temp = data; + +kfree (temp->adev->ip_dump); + temp->adev->ip_dump =NULL ; + temp->adev->num_regs =0 ;kfree (data); }@@ -337,6 +354,11 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost, · line 361
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +5
-0coredump->adev = adev; +/* Trigger ip dump here to capture the value of registers */ +for (int i =0 ; i < adev->num_ip_blocks; i++) +if (adev->ip_blocks[i].version->funcs->dump_ip_state) + adev->ip_blocks[i].version->funcs->dump_ip_state ((void *)adev); +This seems too complicated. I think it would be easier to
This is how all other per IP functions are called. What do you suggest ?
Quoted 132 lines
Quoted 132 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +0
-0ktime_get_ts64 (&coredump->reset_time); dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +113
-0@@ -276,6 +276,99 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");MODULE_FIRMWARE ("amdgpu/gc_10_3_7_mec2.bin" );MODULE_FIRMWARE ("amdgpu/gc_10_3_7_rlc.bin" ); +static const struct hwip_reg_entry gc_reg_list_10_1[] = { + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS2) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS3) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_ERROR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_HPD_STATUS0) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPF_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPG_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_VM_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmPA_CL_CNTL_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRMI_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_DCACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_ICACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQG_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmTCP_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmWD_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_DEBUG) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC1_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC2_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_ME_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_PFP_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_COMMAND) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_MESSAGE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_1) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_3) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_4) }, + {SOC15_REG_ENTRY (GC ,0 , mmSMU_RLC_RESPONSE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_GPM_STAT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SPP_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_BOOTLOAD_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_INT_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_GENERAL_6) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_A) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_B) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_ADDR) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_LX6_CORE_PDEBUG_INST) } +}; +static const struct soc15_reg_golden golden_settings_gc_10_1[] = {SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCB_HW_CONTROL_4,0xffffffff ,0x00400014 ),SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCGTT_CPF_CLK_CTRL,0xfcff8fff ,0xf8000100 ),@@ -9154,6 +9247,38 @@ static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)amdgpu_ring_write (ring, gcr_cntl);/* GCR_CNTL */ } +static void gfx_v10_ip_dump (void *handle) +{ +struct amdgpu_device *adev = (struct amdgpu_device *)handle; +uint32_t reg_count =ARRAY_SIZE (gc_reg_list_10_1); +struct reg_pair *ptr, *gfx10_pair; +uint32_t new_count; + +/* +* Allocate memory based on the no of registers for each block. +* Since adev->ip_dump is common pointer and we add more +* memory to it as we move through different IP's of the ASIC. +* +* This memory needs to be freed after gpu reset once the ip +* dump is done in the devcoredump. +*/ + new_count = adev->num_regs + reg_count; + ptr =krealloc_array (adev->ip_dump, new_count,sizeof (struct reg_pair ),GFP_KERNEL ); +if (ptr) { + adev->ip_dump = ptr; + gfx10_pair = &ptr[adev->num_regs];Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring) · line 9276
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +7
-0+ adev->num_regs = new_count; + }else { +DRM_ERROR ("Failed to allocate memory for IP Dump\n" ); +return ; + } + +for (uint32_t i =0 ; i < reg_count; i++) {Some compilers don't like variables defined within code.
Sure
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring) · line 9280
Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4
-0+ gfx10_pair[i].offset = gc_reg_list_10_1[i].reg_offset; + gfx10_pair[i].value =RREG32 (SOC15_REG_ENTRY_OFFSET (gc_reg_list_10_1[i])); + } +}This seems too complicated. Just pass
drm_printerto theip_dumpcallbacks and let each IP dump their state directly. E.g.,Quoted 8 lines
for (i = 0; i < adev->num_ip_blocks; i++) { if (adev->ip_blocks[i].version->funcs->dump_ip_state) { drm_printf(&p, "$s IP state\n",adev->ip_blocks[i].version->funcs->name);
adev->ip_blocks[i].version->funcs->dump_ip_state((void *)adev, &p); } }I did try this but there is a problem in that approach.
drm_printeris created on the buffer which is provided by "amdgpu_devcoredump_read" and "amdgpu_devcoredump_read" is only called when the user opens the filecat
/sys/class/devcoredump/devcd1/dataand by that time gpu is reset and registers dont hold the value of interest. I tried this approach first and got all FFFF or reset values of the registers.
We need to capture the values in "amdgpu_coredump" which is immediately called after GPU reset and values are what we need for debugging.
Quoted 3 lines
Crux is we need to capture data after gpu reset but read depends on when user open the file which is too late and hence we need to capture all the data in advance immediately after reset before recovery.
Ah, ok. I would suggest pre-allocating the memory to store the dump state in each IP block in sw_init(). I guess we'd probably need two IP callbacks, dump_state() and print_state(). Then call the ip->dump_state() in the GPU recovery sequence before we try soft resets, then in devcore dump you can call the ip->print_state() callback to print the state to the devcore dump output.
Alex
Quoted 72 lines
Regards
Sunil Khatri
Quoted 67 lines
Alex
Quoted 64 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1+static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .name ="gfx_v10_0" , .early_init = gfx_v10_0_early_init,@@ -9170,7 +9295,7 @@ static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .set_clockgating_state = gfx_v10_0_set_clockgating_state, .set_powergating_state = gfx_v10_0_set_powergating_state, .get_clockgating_state = gfx_v10_0_get_clockgating_state, - .dump_ip_state = NULL, + .dump_ip_state = gfx_v10_ip_dump, };static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h +12
-1@@ -4830,6 +4830,8 @@#define mmCP_ECC_FIRSTOCCURRENCE_RING2_BASE_IDX 0#define mmGB_EDC_MODE 0x1e1e#define mmGB_EDC_MODE_BASE_IDX 0 +#define mmCP_DEBUG 0x1e1f +#define mmCP_DEBUG_BASE_IDX 0#define mmCP_FETCHER_SOURCE 0x1e22#define mmCP_FETCHER_SOURCE_BASE_IDX 0#define mmCP_PQ_WPTR_POLL_CNTL 0x1e23@@ -7778,6 +7780,8 @@#define mmCP_MES_DOORBELL_CONTROL5_BASE_IDX 1#define mmCP_MES_DOORBELL_CONTROL6 0x2841#define mmCP_MES_DOORBELL_CONTROL6_BASE_IDX 1 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x2842 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1#define mmCP_MES_GP0_LO 0x2843#define mmCP_MES_GP0_LO_BASE_IDX 1#define mmCP_MES_GP0_HI 0x2844@@ -9332,10 +9336,16 @@#define mmRLC_LB_CNTR_INIT_1_BASE_IDX 1#define mmRLC_LB_CNTR_1 0x4c1c#define mmRLC_LB_CNTR_1_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_ADDR 0x4c1d +#define mmRLC_GPM_DEBUG_INST_ADDR_BASE_IDX 1#define mmRLC_JUMP_TABLE_RESTORE 0x4c1e#define mmRLC_JUMP_TABLE_RESTORE_BASE_IDX 1#define mmRLC_PG_DELAY_2 0x4c1f#define mmRLC_PG_DELAY_2_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_A 0x4c22 +#define mmRLC_GPM_DEBUG_INST_A_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_B 0x4c23 +#define mmRLC_GPM_DEBUG_INST_B_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_LSB 0x4c24#define mmRLC_GPU_CLOCK_COUNT_LSB_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_MSB 0x4c25@@ -9720,6 +9730,8 @@#define mmRLC_SPM_THREAD_TRACE_CTRL_BASE_IDX 1#define mmRLC_LB_CNTR_2 0x4de7#define mmRLC_LB_CNTR_2_BASE_IDX 1 +#define mmRLC_LX6_CORE_PDEBUG_INST 0x4deb +#define mmRLC_LX6_CORE_PDEBUG_INST_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_CTRL 0x4df1#define mmRLC_CPAXI_DOORBELL_MON_CTRL_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_STAT 0x4df2 --2.34.1
Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
Quoted 275 lines · Alex Deucher
Quoted 271 lines · Sunil
Quoted 9 lines · Alex Deucher
Quoted 8 lines · Sunil Khatri
Adding initial set of registers for ipdump during devcoredump starting with gfx10 gc registers.
ip dump is triggered when gpu reset happens via devcoredump and the memory is allocated by each ip and is freed once the dump is complete by devcoredump.
Quoted 7 lines
Quoted 4 lines
Signed-off-by Sunil Khatri<[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 16 +++ .../gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 22 +++I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Sure
Quoted 76 lines
Quoted 76 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 127 +++++++++++++++++- .../include/asic_reg/gc/gc_10_1_0_offset.h | 12 ++ 4 files changed, 176 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +16
-0@@ -139,6 +139,18 @@ enum amdgpu_ss {AMDGPU_SS_DRV_UNLOAD }; +struct hwip_reg_entry { +u32 hwip; +u32 inst; +u32 seg; +u32 reg_offset; +}; + +struct reg_pair { +u32 offset; +u32 value; +}; +struct amdgpu_watchdog_timer {bool timeout_fatal_disable;uint32_t period;/* maxCycles = (1 << period), the number of cycles before a timeout */ @@ -1152,6 +1164,10 @@ struct amdgpu_device {bool debug_largebar;bool debug_disable_soft_recovery;bool debug_use_vram_fw_buf; + +/* IP register dump */ +struct reg_pair *ip_dump; +uint32_t num_regs; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +17
-0@@ -261,6 +261,18 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,drm_printf (&p,"Faulty page starting at address: 0x%016llx\n" , fault_info->addr);drm_printf (&p,"Protection fault status register: 0x%x\n\n" , fault_info->status); +/* Add IP dump for each ip */ +if (coredump->adev->ip_dump !=NULL ) { +struct reg_pair *pair; + + pair = (struct reg_pair *)coredump->adev->ip_dump; +drm_printf (&p,"IP register dump\n" ); +drm_printf (&p,"Offset \t Value\n" ); +for (int i =0 ; i < coredump->adev->num_regs; i++) +drm_printf (&p,"0x%04x \t 0x%08x\n" , pair[i].offset, pair[i].value); +drm_printf (&p,"\n" ); + } +/* Add ring buffer information */ drm_printf (&p,"Ring buffer information\n" );for (int i =0 ; i < coredump->adev->num_rings; i++) {@@ -299,6 +311,11 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,static void amdgpu_devcoredump_free (void *data) { +struct amdgpu_coredump_info *temp = data; + +kfree (temp->adev->ip_dump); + temp->adev->ip_dump =NULL ; + temp->adev->num_regs =0 ;kfree (data); }@@ -337,6 +354,11 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost, · line 361
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +5
-0coredump->adev = adev; +/* Trigger ip dump here to capture the value of registers */ +for (int i =0 ; i < adev->num_ip_blocks; i++) +if (adev->ip_blocks[i].version->funcs->dump_ip_state) + adev->ip_blocks[i].version->funcs->dump_ip_state ((void *)adev); +This seems too complicated. I think it would be easier to
This is how all other per IP functions are called. What do you suggest ?
Quoted 131 lines
Quoted 131 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +0
-0ktime_get_ts64 (&coredump->reset_time); dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +113
-0@@ -276,6 +276,99 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");MODULE_FIRMWARE ("amdgpu/gc_10_3_7_mec2.bin" );MODULE_FIRMWARE ("amdgpu/gc_10_3_7_rlc.bin" ); +static const struct hwip_reg_entry gc_reg_list_10_1[] = { + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS2) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS3) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_ERROR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_HPD_STATUS0) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPF_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPG_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_VM_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmPA_CL_CNTL_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRMI_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_DCACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_ICACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQG_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmTCP_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmWD_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_DEBUG) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC1_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC2_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_ME_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_PFP_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_COMMAND) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_MESSAGE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_1) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_3) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_4) }, + {SOC15_REG_ENTRY (GC ,0 , mmSMU_RLC_RESPONSE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_GPM_STAT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SPP_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_BOOTLOAD_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_INT_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_GENERAL_6) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_A) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_B) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_ADDR) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_LX6_CORE_PDEBUG_INST) } +}; +static const struct soc15_reg_golden golden_settings_gc_10_1[] = {SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCB_HW_CONTROL_4,0xffffffff ,0x00400014 ),SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCGTT_CPF_CLK_CTRL,0xfcff8fff ,0xf8000100 ),@@ -9154,6 +9247,38 @@ static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)amdgpu_ring_write (ring, gcr_cntl);/* GCR_CNTL */ } +static void gfx_v10_ip_dump (void *handle) +{ +struct amdgpu_device *adev = (struct amdgpu_device *)handle; +uint32_t reg_count =ARRAY_SIZE (gc_reg_list_10_1); +struct reg_pair *ptr, *gfx10_pair; +uint32_t new_count; + +/* +* Allocate memory based on the no of registers for each block. +* Since adev->ip_dump is common pointer and we add more +* memory to it as we move through different IP's of the ASIC. +* +* This memory needs to be freed after gpu reset once the ip +* dump is done in the devcoredump. +*/ + new_count = adev->num_regs + reg_count; + ptr =krealloc_array (adev->ip_dump, new_count,sizeof (struct reg_pair ),GFP_KERNEL ); +if (ptr) { + adev->ip_dump = ptr; + gfx10_pair = &ptr[adev->num_regs];Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring) · line 9276
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +7
-0+ adev->num_regs = new_count; + }else { +DRM_ERROR ("Failed to allocate memory for IP Dump\n" ); +return ; + } + +for (uint32_t i =0 ; i < reg_count; i++) {Some compilers don't like variables defined within code.
Sure
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring) · line 9280
Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4
-0+ gfx10_pair[i].offset = gc_reg_list_10_1[i].reg_offset; + gfx10_pair[i].value =RREG32 (SOC15_REG_ENTRY_OFFSET (gc_reg_list_10_1[i])); + } +}This seems too complicated. Just pass
drm_printerto theip_dumpcallbacks and let each IP dump their state directly. E.g.,Quoted 8 lines
for (i = 0; i < adev->num_ip_blocks; i++) { if (adev->ip_blocks[i].version->funcs->dump_ip_state) { drm_printf(&p, "$s IP state\n",adev->ip_blocks[i].version->funcs->name);
adev->ip_blocks[i].version->funcs->dump_ip_state((void *)adev, &p); } }I did try this but there is a problem in that approach.
drm_printeris created on the buffer which is provided by "amdgpu_devcoredump_read" and "amdgpu_devcoredump_read" is only called when the user opens the filecat
/sys/class/devcoredump/devcd1/dataand by that time gpu is reset and registers dont hold the value of interest. I tried this approach first and got all FFFF or reset values of the registers.
We need to capture the values in "amdgpu_coredump" which is immediately called after GPU reset and values are what we need for debugging.Quoted 3 lines
Crux is we need to capture data after gpu reset but read depends on when user open the file which is too late and hence we need to capture all the data in advance immediately after reset before recovery.
Quoted 6 lines
Ah, ok. I would suggest pre-allocating the memory to store the dump state in each IP block in
sw_init(). I guess we'd probably need two IP callbacks,dump_state()andprint_state(). Then call theip->dump_state()in the GPU recovery sequence before we try soft resets, then in devcore dump you can call theip->print_state()callback to print the state to the devcore dump output.
The problem is there is no standard size to proallocate memory statically. Each IP hold different set of registers and every time i add a new register set i have to update the memory size and allocate dynamically but this can be handled. Another problem i see with this approach is where to store the memory ptr of each IP. One way is to Directly have ptrs for each IP in amdgpu_device global struct or other way is to add ptr in IP specific struct withing amdgpu_device which are to be traversed in a hardcode way as they arent in an array or list.
For eg: we have to access each of these in a hardcode way in code as these structures arent in a list/array form to traverse for each ip.
/* nbio */structamdgpu_nbionbio;
/* hdp */structamdgpu_hdphdp;
/* smuio */structamdgpu_smuiosmuio;
/* mmhub */structamdgpu_mmhubmmhub;
/* gfxhub */structamdgpu_gfxhubgfxhub;
/* gfx */structamdgpu_gfxgfx;
/* sdma */structamdgpu_sdmasdma;
/* lsdma */structamdgpu_lsdmalsdma;
/* uvd */structamdgpu_uvduvd;
/* vce */structamdgpu_vcevce;
In current code the approach is to add memory based on need for each IP in the same ptr and keep adding registers to it and we can search with offset value in the devcoredump to debug.
Regards
Sunil
Quoted 73 lines
Alex
Quoted 70 lines
Regards
Sunil Khatri
Quoted 66 lines
Alex
Quoted 64 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1+static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .name ="gfx_v10_0" , .early_init = gfx_v10_0_early_init,@@ -9170,7 +9295,7 @@ static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .set_clockgating_state = gfx_v10_0_set_clockgating_state, .set_powergating_state = gfx_v10_0_set_powergating_state, .get_clockgating_state = gfx_v10_0_get_clockgating_state, - .dump_ip_state = NULL, + .dump_ip_state = gfx_v10_ip_dump, };static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h +12
-1@@ -4830,6 +4830,8 @@#define mmCP_ECC_FIRSTOCCURRENCE_RING2_BASE_IDX 0#define mmGB_EDC_MODE 0x1e1e#define mmGB_EDC_MODE_BASE_IDX 0 +#define mmCP_DEBUG 0x1e1f +#define mmCP_DEBUG_BASE_IDX 0#define mmCP_FETCHER_SOURCE 0x1e22#define mmCP_FETCHER_SOURCE_BASE_IDX 0#define mmCP_PQ_WPTR_POLL_CNTL 0x1e23@@ -7778,6 +7780,8 @@#define mmCP_MES_DOORBELL_CONTROL5_BASE_IDX 1#define mmCP_MES_DOORBELL_CONTROL6 0x2841#define mmCP_MES_DOORBELL_CONTROL6_BASE_IDX 1 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x2842 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1#define mmCP_MES_GP0_LO 0x2843#define mmCP_MES_GP0_LO_BASE_IDX 1#define mmCP_MES_GP0_HI 0x2844@@ -9332,10 +9336,16 @@#define mmRLC_LB_CNTR_INIT_1_BASE_IDX 1#define mmRLC_LB_CNTR_1 0x4c1c#define mmRLC_LB_CNTR_1_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_ADDR 0x4c1d +#define mmRLC_GPM_DEBUG_INST_ADDR_BASE_IDX 1#define mmRLC_JUMP_TABLE_RESTORE 0x4c1e#define mmRLC_JUMP_TABLE_RESTORE_BASE_IDX 1#define mmRLC_PG_DELAY_2 0x4c1f#define mmRLC_PG_DELAY_2_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_A 0x4c22 +#define mmRLC_GPM_DEBUG_INST_A_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_B 0x4c23 +#define mmRLC_GPM_DEBUG_INST_B_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_LSB 0x4c24#define mmRLC_GPU_CLOCK_COUNT_LSB_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_MSB 0x4c25@@ -9720,6 +9730,8 @@#define mmRLC_SPM_THREAD_TRACE_CTRL_BASE_IDX 1#define mmRLC_LB_CNTR_2 0x4de7#define mmRLC_LB_CNTR_2_BASE_IDX 1 +#define mmRLC_LX6_CORE_PDEBUG_INST 0x4deb +#define mmRLC_LX6_CORE_PDEBUG_INST_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_CTRL 0x4df1#define mmRLC_CPAXI_DOORBELL_MON_CTRL_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_STAT 0x4df2 --2.34.1
Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
Quoted 323 lines · Sunil
On 4/12/2024 10:42 PM, Alex Deucher wrote:
On Fri, Apr 12, 2024 at 1:05 PM Khatri, Sunil <[email protected]> wrote:
On 4/12/2024 8:50 PM, Alex Deucher wrote:
On Fri, Apr 12, 2024 at 10:00 AM Sunil Khatri <[email protected]> wrote:
Adding initial set of registers for ipdump during
devcoredump starting with gfx10 gc registers.ip dump is triggered when gpu reset happens via
devcoredump and the memory is allocated by each
ip and is freed once the dump is complete by
devcoredump.Signed-off-by Sunil Khatri <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 16 +++ .../gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 22 +++I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Sure
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 127 +++++++++++++++++- .../include/asic_reg/gc/gc_10_1_0_offset.h | 12 ++ 4 files changed, 176 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +16
-0@@ -139,6 +139,18 @@ enum amdgpu_ss {AMDGPU_SS_DRV_UNLOAD }; +struct hwip_reg_entry { +u32 hwip; +u32 inst; +u32 seg; +u32 reg_offset; +}; + +struct reg_pair { +u32 offset; +u32 value; +}; +struct amdgpu_watchdog_timer {bool timeout_fatal_disable;uint32_t period;/* maxCycles = (1 << period), the number of cycles before a timeout */ @@ -1152,6 +1164,10 @@ struct amdgpu_device {bool debug_largebar;bool debug_disable_soft_recovery;bool debug_use_vram_fw_buf; + +/* IP register dump */ +struct reg_pair *ip_dump; +uint32_t num_regs; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +22
-0@@ -261,6 +261,18 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,drm_printf (&p,"Faulty page starting at address: 0x%016llx\n" , fault_info->addr);drm_printf (&p,"Protection fault status register: 0x%x\n\n" , fault_info->status); +/* Add IP dump for each ip */ +if (coredump->adev->ip_dump !=NULL ) { +struct reg_pair *pair; + + pair = (struct reg_pair *)coredump->adev->ip_dump; +drm_printf (&p,"IP register dump\n" ); +drm_printf (&p,"Offset \t Value\n" ); +for (int i =0 ; i < coredump->adev->num_regs; i++) +drm_printf (&p,"0x%04x \t 0x%08x\n" , pair[i].offset, pair[i].value); +drm_printf (&p,"\n" ); + } +/* Add ring buffer information */ drm_printf (&p,"Ring buffer information\n" );for (int i =0 ; i < coredump->adev->num_rings; i++) {@@ -299,6 +311,11 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,static void amdgpu_devcoredump_free (void *data) { +struct amdgpu_coredump_info *temp = data; + +kfree (temp->adev->ip_dump); + temp->adev->ip_dump =NULL ; + temp->adev->num_regs =0 ;kfree (data); }@@ -337,6 +354,11 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost, coredump->adev = adev; +/* Trigger ip dump here to capture the value of registers */ +for (int i =0 ; i < adev->num_ip_blocks; i++) +if (adev->ip_blocks[i].version->funcs->dump_ip_state) + adev->ip_blocks[i].version->funcs->dump_ip_state ((void *)adev); +This seems too complicated. I think it would be easier to
This is how all other per IP functions are called. What do you suggest ?
ktime_get_ts64(&coredump->reset_time); dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +120
-0@@ -276,6 +276,99 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");MODULE_FIRMWARE ("amdgpu/gc_10_3_7_mec2.bin" );MODULE_FIRMWARE ("amdgpu/gc_10_3_7_rlc.bin" ); +static const struct hwip_reg_entry gc_reg_list_10_1[] = { + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS2) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS3) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_ERROR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_HPD_STATUS0) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPF_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPG_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_VM_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmPA_CL_CNTL_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRMI_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_DCACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_ICACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQG_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmTCP_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmWD_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_DEBUG) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC1_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC2_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_ME_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_PFP_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_COMMAND) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_MESSAGE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_1) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_3) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_4) }, + {SOC15_REG_ENTRY (GC ,0 , mmSMU_RLC_RESPONSE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_GPM_STAT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SPP_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_BOOTLOAD_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_INT_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_GENERAL_6) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_A) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_B) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_ADDR) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_LX6_CORE_PDEBUG_INST) } +}; +static const struct soc15_reg_golden golden_settings_gc_10_1[] = {SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCB_HW_CONTROL_4,0xffffffff ,0x00400014 ),SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCGTT_CPF_CLK_CTRL,0xfcff8fff ,0xf8000100 ),@@ -9154,6 +9247,38 @@ static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)amdgpu_ring_write (ring, gcr_cntl);/* GCR_CNTL */ } +static void gfx_v10_ip_dump (void *handle) +{ +struct amdgpu_device *adev = (struct amdgpu_device *)handle; +uint32_t reg_count =ARRAY_SIZE (gc_reg_list_10_1); +struct reg_pair *ptr, *gfx10_pair; +uint32_t new_count; + +/* +* Allocate memory based on the no of registers for each block. +* Since adev->ip_dump is common pointer and we add more +* memory to it as we move through different IP's of the ASIC. +* +* This memory needs to be freed after gpu reset once the ip +* dump is done in the devcoredump. +*/ + new_count = adev->num_regs + reg_count; + ptr =krealloc_array (adev->ip_dump, new_count,sizeof (struct reg_pair ),GFP_KERNEL ); +if (ptr) { + adev->ip_dump = ptr; + gfx10_pair = &ptr[adev->num_regs]; + adev->num_regs = new_count; + }else { +DRM_ERROR ("Failed to allocate memory for IP Dump\n" ); +return ; + } + +for (uint32_t i =0 ; i < reg_count; i++) {Some compilers don't like variables defined within code.
Sure
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4
-0+ gfx10_pair[i].offset = gc_reg_list_10_1[i].reg_offset; + gfx10_pair[i].value =RREG32 (SOC15_REG_ENTRY_OFFSET (gc_reg_list_10_1[i])); + } +}This seems too complicated. Just pass
drm_printerto theip_dumpcallbacks and let each IP dump their state directly. E.g.,for (i = 0; i < adev->num_ip_blocks; i++) { if (adev->ip_blocks[i].version->funcs->dump_ip_state) { drm_printf(&p, "$s IP state\n",adev->ip_blocks[i].version->funcs->name);
adev->ip_blocks[i].version->funcs->dump_ip_state((void *)adev, &p); } }I did try this but there is a problem in that approach.
drm_printeris created on the buffer which is provided by "amdgpu_devcoredump_read" and "amdgpu_devcoredump_read" is only called when the user opens the filecat
/sys/class/devcoredump/devcd1/dataand by that time gpu is reset and registers dont hold the value of interest. I tried this approach first and got all FFFF or reset values of the registers.
We need to capture the values in "amdgpu_coredump" which is immediately called after GPU reset and values are what we need for debugging.Crux is we need to capture data after gpu reset but read depends on when user open the file which is too late and hence we need to capture all the data in advance immediately after reset before recovery.
Ah, ok. I would suggest pre-allocating the memory to store the dump state in each IP block in
sw_init(). I guess we'd probably need two IP callbacks,dump_state()andprint_state(). Then call theip->dump_state()in the GPU recovery sequence before we try soft resets, then in devcore dump you can call theip->print_state()callback to print the state to the devcore dump output.The problem is there is no standard size to proallocate memory statically. Each IP hold different set of registers and every time i add a new register set i have to update the memory size and allocate dynamically but this can be handled. Another problem i see with this approach is where to store the memory ptr of each IP. One way is to Directly have ptrs for each IP in
amdgpu_deviceglobal struct or other way is to add ptr in IP specific struct withingamdgpu_devicewhich are to be traversed in a hardcode way as they arent in an array or list.For eg: we have to access each of these in a hardcode way in code as these structures arent in a list/array form to traverse for each ip.
/* nbio */
struct amdgpu_nbio nbio; /* hdp */ struct amdgpu_hdp hdp; /* smuio */ struct amdgpu_smuio smuio; /* mmhub */ struct amdgpu_mmhub mmhub; /* gfxhub */ struct amdgpu_gfxhub gfxhub; /* gfx */ struct amdgpu_gfx gfx; /* sdma */ struct amdgpu_sdma sdma; /* lsdma */ struct amdgpu_lsdma lsdma; /* uvd */ struct amdgpu_uvd uvd; /* vce */ struct amdgpu_vce vce;
Quoted 2 lines
In current code the approach is to add memory based on need for each IP in the same ptr and keep adding registers to it and we can search with offset value in the devcoredump to debug.
Each IP can store the data however they want. E.g., in the gfx code, we can do:
gfx_v10_0_sw_init()
{
...adev->gfx.ip_state = kcalloc(2, gc_reg_list_10_1, GFP_KERNEL);
...
}
gfx_v10_0_dump_state()
{
for (i = 0; i < ARRAY_SIZE(gc_reg_list_10_1); i += 2)
adev->gfx.ip_state[i] = gc_reg_list_10_1[i];
adev->gfx.ip_state[i+1] = RREG32(gc_reg_list_10_1[i]);
}}
gfx_v10_0_print_state()
{
for (i = 0; i < ARRAY_SIZE(gc_reg_list_10_1); i += 2)
drm_printf(&p, "0x%08x \t 0x%08x\n", adev->gfx.ip_state[i],
adev->gfx.ip_state[i+1]);
}
gfx_v10_0_sw_fini()
{
...kfree(adev->gfx.ip_state);
...
}
Then call dumps_state() in gpu_recover() before we actually reset the GPU. The driver will then have the state cached and then you call print_state() in amdgpu_devcoredump_read() to read out the cached state.
Alex
Quoted 75 lines
Regards
SunilAlex
Regards
Sunil Khatri
Alex
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1+static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .name ="gfx_v10_0" , .early_init = gfx_v10_0_early_init,@@ -9170,7 +9295,7 @@ static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .set_clockgating_state = gfx_v10_0_set_clockgating_state, .set_powergating_state = gfx_v10_0_set_powergating_state, .get_clockgating_state = gfx_v10_0_get_clockgating_state, - .dump_ip_state = NULL, + .dump_ip_state = gfx_v10_ip_dump, };static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h +12
-1@@ -4830,6 +4830,8 @@#define mmCP_ECC_FIRSTOCCURRENCE_RING2_BASE_IDX 0#define mmGB_EDC_MODE 0x1e1e#define mmGB_EDC_MODE_BASE_IDX 0 +#define mmCP_DEBUG 0x1e1f +#define mmCP_DEBUG_BASE_IDX 0#define mmCP_FETCHER_SOURCE 0x1e22#define mmCP_FETCHER_SOURCE_BASE_IDX 0#define mmCP_PQ_WPTR_POLL_CNTL 0x1e23@@ -7778,6 +7780,8 @@#define mmCP_MES_DOORBELL_CONTROL5_BASE_IDX 1#define mmCP_MES_DOORBELL_CONTROL6 0x2841#define mmCP_MES_DOORBELL_CONTROL6_BASE_IDX 1 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x2842 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1#define mmCP_MES_GP0_LO 0x2843#define mmCP_MES_GP0_LO_BASE_IDX 1#define mmCP_MES_GP0_HI 0x2844@@ -9332,10 +9336,16 @@#define mmRLC_LB_CNTR_INIT_1_BASE_IDX 1#define mmRLC_LB_CNTR_1 0x4c1c#define mmRLC_LB_CNTR_1_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_ADDR 0x4c1d +#define mmRLC_GPM_DEBUG_INST_ADDR_BASE_IDX 1#define mmRLC_JUMP_TABLE_RESTORE 0x4c1e#define mmRLC_JUMP_TABLE_RESTORE_BASE_IDX 1#define mmRLC_PG_DELAY_2 0x4c1f#define mmRLC_PG_DELAY_2_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_A 0x4c22 +#define mmRLC_GPM_DEBUG_INST_A_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_B 0x4c23 +#define mmRLC_GPM_DEBUG_INST_B_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_LSB 0x4c24#define mmRLC_GPU_CLOCK_COUNT_LSB_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_MSB 0x4c25@@ -9720,6 +9730,8 @@#define mmRLC_SPM_THREAD_TRACE_CTRL_BASE_IDX 1#define mmRLC_LB_CNTR_2 0x4de7#define mmRLC_LB_CNTR_2_BASE_IDX 1 +#define mmRLC_LX6_CORE_PDEBUG_INST 0x4deb +#define mmRLC_LX6_CORE_PDEBUG_INST_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_CTRL 0x4df1#define mmRLC_CPAXI_DOORBELL_MON_CTRL_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_STAT 0x4df2 --2.34.1
RE: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
[AMD Official Use Only - General]
-----Original Message-----
From: Alex Deucher <[email protected]>
Sent: Saturday, April 13, 2024 1:56 AM
To: Khatri, Sunil <[email protected]>
Cc: Khatri, Sunil <[email protected]>; Deucher, Alexander <[email protected]>; Koenig, Christian <[email protected]>; [email protected]
Subject: Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
Quoted 328 lines · Sunil
On 4/12/2024 10:42 PM, Alex Deucher wrote:
On Fri, Apr 12, 2024 at 1:05 PM Khatri, Sunil <[email protected]> wrote:
On 4/12/2024 8:50 PM, Alex Deucher wrote:
On Fri, Apr 12, 2024 at 10:00 AM Sunil Khatri <[email protected]> wrote:
Adding initial set of registers for ipdump during devcoredump starting with gfx10 gc registers.
ip dump is triggered when gpu reset happens via devcoredump and the memory is allocated by each ip and is freed once the dump is complete by devcoredump.
Signed-off-by Sunil Khatri <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 16 +++ .../gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 22 +++I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Sure
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 127 +++++++++++++++++- .../include/asic_reg/gc/gc_10_1_0_offset.h | 12 ++ 4 files changed, 176 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +12
-0@@ -139,6 +139,18 @@ enum amdgpu_ss {AMDGPU_SS_DRV_UNLOAD }; +struct hwip_reg_entry { +u32 hwip; +u32 inst; +u32 seg; +u32 reg_offset; +}; + +struct reg_pair { +u32 offset; +u32 value; +}; +struct amdgpu_watchdog_timer {bool timeout_fatal_disable;uint32_t period;/* maxCycles = (1 << period), the number of cycles before a timeout */ @@ -1152,6 +1164,10 @@ struct amdgpu_device { bool debug_largebar; bool debug_disable_soft_recovery; bool debug_use_vram_fw_buf;drivers/gpu/drm/amd/amdgpu/amdgpu.h +4
-0+ +/* IP register dump */ +struct reg_pair *ip_dump; +uint32_t num_regs; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device*adev, diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.cb/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c
index1129e5e5fb42..2079f67c9fac100644drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +12
-0@@ -261,6 +261,18 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,drm_printf (&p,"Faulty page starting at address: 0x%016llx\n" , fault_info->addr);drm_printf (&p,"Protection fault status register: 0x%x\n\n" , fault_info->status); +/* Add IP dump for each ip */ +if (coredump->adev->ip_dump !=NULL ) { +struct reg_pair *pair; + + pair = (struct reg_pair *)coredump->adev->ip_dump; +drm_printf (&p,"IP register dump\n" ); +drm_printf (&p,"Offset \t Value\n" ); +for (int i =0 ; i < coredump->adev->num_regs; i++) +drm_printf (&p,"0x%04x \t 0x%08x\n" , pair[i].offset, pair[i].value); +drm_printf (&p,"\n" ); + } +/* Add ring buffer information */ drm_printf (&p,"Ring buffer information\n" );for (int i = 0; i < coredump->adev->num_rings; i++) { @@-299,6 +311,11 @@ amdgpu_devcoredump_read(char *buffer,
loff_toffset,size_tcount,static void amdgpu_devcoredump_free(void *data) {drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +11
-0+struct amdgpu_coredump_info *temp = data; + +kfree (temp->adev->ip_dump); + temp->adev->ip_dump =NULL ; + temp->adev->num_regs =0 ;kfree (data); }@@ -337,6 +354,11 @@ void amdgpu_coredump(struct amdgpu_device *adev,bool vram_lost, coredump->adev = adev; +/* Trigger ip dump here to capture the value of registers */ +for (int i =0 ; i < adev->num_ip_blocks; i++) +if (adev->ip_blocks[i].version->funcs->dump_ip_state) + + adev->ip_blocks[i].version->funcs->dump_ip_state ((void *)adev); +This seems too complicated. I think it would be easier to
This is how all other per IP functions are called. What do you suggest ?
ktime_get_ts64(&coredump->reset_time); dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +92
-0@@ -276,6 +276,99 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");MODULE_FIRMWARE ("amdgpu/gc_10_3_7_mec2.bin" );MODULE_FIRMWARE ("amdgpu/gc_10_3_7_rlc.bin" ); +static const struct hwip_reg_entry gc_reg_list_10_1[] = { + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS2) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS3) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_ERROR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_HPD_STATUS0) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPF_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPG_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_VM_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmPA_CL_CNTL_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRMI_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_DCACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_ICACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQG_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmTCP_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmWD_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_DEBUG) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC1_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC2_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_ME_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_PFP_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_COMMAND) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_MESSAGE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_1) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_3) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_4) }, + {SOC15_REG_ENTRY (GC ,0 , mmSMU_RLC_RESPONSE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_GPM_STAT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SPP_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_BOOTLOAD_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_INT_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_GENERAL_6) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_A) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_B) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_GPM_DEBUG_INST_ADDR) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_LX6_CORE_PDEBUG_INST) } }; +static const struct soc15_reg_golden golden_settings_gc_10_1[] = {SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCB_HW_CONTROL_4,0xffffffff ,0x00400014 ), SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_CPF_CLK_CTRL,
0xfcff8fff,0xf8000100), @@ -9154,6 +9247,38 @@ static voidgfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)
amdgpu_ring_write(ring, gcr_cntl); /*GCR_CNTL*/
}drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +26
-0+static void gfx_v10_ip_dump (void *handle) { +struct amdgpu_device *adev = (struct amdgpu_device *)handle; +uint32_t reg_count =ARRAY_SIZE (gc_reg_list_10_1); +struct reg_pair *ptr, *gfx10_pair; +uint32_t new_count; + +/* +* Allocate memory based on the no of registers for each block. +* Since adev->ip_dump is common pointer and we add more +* memory to it as we move through different IP's of the ASIC. +* +* This memory needs to be freed after gpu reset once the ip +* dump is done in the devcoredump. +*/ + new_count = adev->num_regs + reg_count; + ptr =krealloc_array (adev->ip_dump, new_count,sizeof (struct reg_pair ),GFP_KERNEL ); +if (ptr) { + adev->ip_dump = ptr; + gfx10_pair = &ptr[adev->num_regs]; + adev->num_regs = new_count; + }else { +DRM_ERROR ("Failed to allocate memory for IP Dump\n" ); +return ; + } + +for (uint32_t i =0 ; i < reg_count; i++) {Some compilers don't like variables defined within code.
Sure
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4
-0+ gfx10_pair[i].offset = gc_reg_list_10_1[i].reg_offset; + gfx10_pair[i].value =RREG32 (SOC15_REG_ENTRY_OFFSET (gc_reg_list_10_1[i])); + } +}This seems too complicated. Just pass
drm_printerto theip_dumpcallbacks and let each IP dump their state directly. E.g.,for (i = 0; i < adev->num_ip_blocks; i++) { if (adev->ip_blocks[i].version->funcs->dump_ip_state) { drm_printf(&p, "$s IP state\n",adev->ip_blocks[i].version->funcs->name);
adev->ip_blocks[i].version->funcs->dump_ip_state((void *)adev, &p); } }I did try this but there is a problem in that approach.
drm_printeris created on the buffer which is provided by "amdgpu_devcoredump_read" and "amdgpu_devcoredump_read" is only called when the user opens the filecat
/sys/class/devcoredump/devcd1/dataand by that time gpu is reset and registers dont hold the value of interest. I tried this approach first and got all FFFF or reset values of the registers.
We need to capture the values in "amdgpu_coredump" which is immediately called after GPU reset and values are what we need for debugging.Crux is we need to capture data after gpu reset but read depends on when user open the file which is too late and hence we need to capture all the data in advance immediately after reset before recovery.
Ah, ok. I would suggest pre-allocating the memory to store the dump state in each IP block in
sw_init(). I guess we'd probably need two IP callbacks,dump_state()andprint_state(). Then call theip->dump_state()in the GPU recovery sequence before we try soft resets, then in devcore dump you can call theip->print_state()callback to print the state to the devcore dump output.The problem is there is no standard size to proallocate memory statically. Each IP hold different set of registers and every time i add a new register set i have to update the memory size and allocate dynamically but this can be handled. Another problem i see with this approach is where to store the memory ptr of each IP. One way is to Directly have ptrs for each IP in
amdgpu_deviceglobal struct or other way is to add ptr in IP specific struct withingamdgpu_devicewhich are to be traversed in a hardcode way as they arent in an array or list.For eg: we have to access each of these in a hardcode way in code as these structures arent in a list/array form to traverse for each ip.
/* nbio */
struct amdgpu_nbio nbio; /* hdp */ struct amdgpu_hdp hdp; /* smuio */ struct amdgpu_smuio smuio; /* mmhub */ struct amdgpu_mmhub mmhub; /* gfxhub */ struct amdgpu_gfxhub gfxhub; /* gfx */ struct amdgpu_gfx gfx; /* sdma */ struct amdgpu_sdma sdma; /* lsdma */ struct amdgpu_lsdma lsdma; /* uvd */ struct amdgpu_uvd uvd; /* vce */ struct amdgpu_vce vce;
Quoted 2 lines
In current code the approach is to add memory based on need for each IP in the same ptr and keep adding registers to it and we can search with offset value in the devcoredump to debug.
Each IP can store the data however they want. E.g., in the gfx code, we can do:
gfx_v10_0_sw_init()
{
...adev->gfx.ip_state = kcalloc(2, gc_reg_list_10_1, GFP_KERNEL);
...
}
gfx_v10_0_dump_state()
{
for (i = 0; i < ARRAY_SIZE(gc_reg_list_10_1); i += 2)
adev->gfx.ip_state[i] = gc_reg_list_10_1[i];
adev->gfx.ip_state[i+1] = RREG32(gc_reg_list_10_1[i]); } }gfx_v10_0_print_state()
{
for (i = 0; i < ARRAY_SIZE(gc_reg_list_10_1); i += 2)
drm_printf(&p, "0x%08x \t 0x%08x\n", adev->gfx.ip_state[i],
adev->gfx.ip_state[i+1]);
}
gfx_v10_0_sw_fini()
{
...kfree(adev->gfx.ip_state);
...
}
Then call dumps_state() in gpu_recover() before we actually reset the GPU. The driver will then have the state cached and then you call print_state() in amdgpu_devcoredump_read() to read out the cached state.
Sure Alex. Will update the patches as per this approach.
Regards
Sunil Khatri
Alex
Quoted 77 lines
Regards
SunilAlex
Regards
Sunil Khatri
Alex
+
static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .name = "gfx_v10_0", .early_init = gfx_v10_0_early_init, @@ -9170,7 +9295,7 @@static const struct
amd_ip_funcsgfx_v10_0_ip_funcs= {
.set_clockgating_state =gfx_v10_0_set_clockgating_state, .set_powergating_state =gfx_v10_0_set_powergating_state, .get_clockgating_state =gfx_v10_0_get_clockgating_state,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +1
-1- .dump_ip_state = NULL, + .dump_ip_state = gfx_v10_ip_dump, };static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {diff --git
a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.hb/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.hindex4908044f7409..4c8e7fdb6976100644drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h +12
-1@@ -4830,6 +4830,8 @@#define mmCP_ECC_FIRSTOCCURRENCE_RING2_BASE_IDX 0#define mmGB_EDC_MODE 0x1e1e#define mmGB_EDC_MODE_BASE_IDX 0 +#define mmCP_DEBUG 0x1e1f +#define mmCP_DEBUG_BASE_IDX 0#define mmCP_FETCHER_SOURCE 0x1e22#define mmCP_FETCHER_SOURCE_BASE_IDX 0#define mmCP_PQ_WPTR_POLL_CNTL 0x1e23@@ -7778,6 +7780,8 @@#define mmCP_MES_DOORBELL_CONTROL5_BASE_IDX 1#define mmCP_MES_DOORBELL_CONTROL6 0x2841#define mmCP_MES_DOORBELL_CONTROL6_BASE_IDX 1 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x2842 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1#define mmCP_MES_GP0_LO 0x2843#define mmCP_MES_GP0_LO_BASE_IDX 1#define mmCP_MES_GP0_HI 0x2844@@ -9332,10 +9336,16 @@#define mmRLC_LB_CNTR_INIT_1_BASE_IDX 1#define mmRLC_LB_CNTR_1 0x4c1c#define mmRLC_LB_CNTR_1_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_ADDR 0x4c1d +#define mmRLC_GPM_DEBUG_INST_ADDR_BASE_IDX 1#define mmRLC_JUMP_TABLE_RESTORE 0x4c1e#define mmRLC_JUMP_TABLE_RESTORE_BASE_IDX 1#define mmRLC_PG_DELAY_2 0x4c1f#define mmRLC_PG_DELAY_2_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_A 0x4c22 +#define mmRLC_GPM_DEBUG_INST_A_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_B 0x4c23 +#define mmRLC_GPM_DEBUG_INST_B_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_LSB 0x4c24#define mmRLC_GPU_CLOCK_COUNT_LSB_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_MSB 0x4c25@@ -9720,6 +9730,8 @@#define mmRLC_SPM_THREAD_TRACE_CTRL_BASE_IDX 1#define mmRLC_LB_CNTR_2 0x4de7#define mmRLC_LB_CNTR_2_BASE_IDX 1 +#define mmRLC_LX6_CORE_PDEBUG_INST 0x4deb +#define mmRLC_LX6_CORE_PDEBUG_INST_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_CTRL 0x4df1#define mmRLC_CPAXI_DOORBELL_MON_CTRL_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_STAT 0x4df2 --2.34.1
Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dump
Quoted 227 lines · Sunil
[AMD Official Use Only - General]
-----Original Message-----
From: Alex Deucher <[email protected]>
Sent: Saturday, April 13, 2024 1:56 AM
To: Khatri, Sunil <[email protected]>
Cc: Khatri, Sunil <[email protected]>; Deucher, Alexander <[email protected]>; Koenig, Christian <[email protected]>; [email protected]
Subject: Re: [PATCH v2 2/2] drm/amdgpu: Add support of gfx10 register dumpQuoted 216 lines · Sunil
On 4/12/2024 10:42 PM, Alex Deucher wrote:
On Fri, Apr 12, 2024 at 1:05 PM Khatri, Sunil <[email protected]> wrote:
On 4/12/2024 8:50 PM, Alex Deucher wrote:
On Fri, Apr 12, 2024 at 10:00 AM Sunil Khatri <[email protected]> wrote:
Adding initial set of registers for ipdump during devcoredump starting with gfx10 gc registers.
ip dump is triggered when gpu reset happens via devcoredump and the memory is allocated by each ip and is freed once the dump is complete by devcoredump.
Signed-off-by Sunil Khatri <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 16 +++ .../gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 22 +++I would split this into two patches, one to add the core infrastructure in devcoredump and one to add gfx10 support. The core support could be squashed into patch 1 as well.
Sure
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 127 +++++++++++++++++- .../include/asic_reg/gc/gc_10_1_0_offset.h | 12 ++ 4 files changed, 176 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +12
-0@@ -139,6 +139,18 @@ enum amdgpu_ss {AMDGPU_SS_DRV_UNLOAD }; +struct hwip_reg_entry { +u32 hwip; +u32 inst; +u32 seg; +u32 reg_offset; +}; + +struct reg_pair { +u32 offset; +u32 value; +}; +struct amdgpu_watchdog_timer {bool timeout_fatal_disable;uint32_t period;/* maxCycles = (1 << period), the number of cycles before a timeout */ @@ -1152,6 +1164,10 @@ struct amdgpu_device { bool debug_largebar; bool debug_disable_soft_recovery; bool debug_use_vram_fw_buf;drivers/gpu/drm/amd/amdgpu/amdgpu.h +4
-0+ +/* IP register dump */ +struct reg_pair *ip_dump; +uint32_t num_regs; }; static inline uint32_t amdgpu_ip_version(const struct amdgpu_device*adev, diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.cb/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c
index1129e5e5fb42..2079f67c9fac100644drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +12
-0@@ -261,6 +261,18 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,drm_printf (&p,"Faulty page starting at address: 0x%016llx\n" , fault_info->addr);drm_printf (&p,"Protection fault status register: 0x%x\n\n" , fault_info->status); +/* Add IP dump for each ip */ +if (coredump->adev->ip_dump !=NULL ) { +struct reg_pair *pair; + + pair = (struct reg_pair *)coredump->adev->ip_dump; +drm_printf (&p,"IP register dump\n" ); +drm_printf (&p,"Offset \t Value\n" ); +for (int i =0 ; i < coredump->adev->num_regs; i++) +drm_printf (&p,"0x%04x \t 0x%08x\n" , pair[i].offset, pair[i].value); +drm_printf (&p,"\n" ); + } +/* Add ring buffer information */ drm_printf (&p,"Ring buffer information\n" );for (int i = 0; i < coredump->adev->num_rings; i++) { @@-299,6 +311,11 @@ amdgpu_devcoredump_read(char *buffer,
loff_toffset,size_tcount,static void amdgpu_devcoredump_free(void *data) {drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +11
-0+struct amdgpu_coredump_info *temp = data; + +kfree (temp->adev->ip_dump); + temp->adev->ip_dump =NULL ; + temp->adev->num_regs =0 ;kfree (data); }@@ -337,6 +354,11 @@ void amdgpu_coredump(struct amdgpu_device *adev,bool vram_lost, coredump->adev = adev; +/* Trigger ip dump here to capture the value of registers */ +for (int i =0 ; i < adev->num_ip_blocks; i++) +if (adev->ip_blocks[i].version->funcs->dump_ip_state) + + adev->ip_blocks[i].version->funcs->dump_ip_state ((void *)adev); +This seems too complicated. I think it would be easier to
This is how all other per IP functions are called. What do you suggest ?
ktime_get_ts64(&coredump->reset_time); dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +83
-0@@ -276,6 +276,99 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");MODULE_FIRMWARE ("amdgpu/gc_10_3_7_mec2.bin" );MODULE_FIRMWARE ("amdgpu/gc_10_3_7_rlc.bin" ); +static const struct hwip_reg_entry gc_reg_list_10_1[] = { + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS2) }, + {SOC15_REG_ENTRY (GC ,0 , mmGRBM_STATUS3) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_STALLED_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STALLED_STAT1) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_BUSY_STAT2) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPF_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_ERROR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_GFX_HPD_STATUS0) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB0_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_RPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB1_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_BASE) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_RB2_WPTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_CMD_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB1_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_LO) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BASE_HI) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_IB2_BUFSZ) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPF_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCPG_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmGDS_VM_PROTECTION_FAULT) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmIA_UTCL1_STATUS_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmPA_CL_CNTL_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRMI_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_DCACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQC_ICACHE_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmSQG_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmTCP_UTCL0_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmWD_UTCL1_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmGCVM_L2_PROTECTION_FAULT_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_DEBUG) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_CNTL) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CE_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC1_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MEC2_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_MES_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_ME_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_PFP_INSTR_PNTR) }, + {SOC15_REG_ENTRY (GC ,0 , mmCP_CPC_STATUS) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_STAT) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_COMMAND) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_MESSAGE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_1) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_2) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_3) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_ARGUMENT_4) }, + {SOC15_REG_ENTRY (GC ,0 , mmSMU_RLC_RESPONSE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_SMU_SAFE_MODE) }, + {SOC15_REG_ENTRY (GC ,0 , mmRLC_RLCS_GPM_STAT_2) },
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin"); · line 369
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +8
-0+ { SOC15_REG_ENTRY(GC, 0, mmRLC_SPP_STATUS) }, + { SOC15_REG_ENTRY(GC, 0, mmRLC_RLCS_BOOTLOAD_STATUS) }, + { SOC15_REG_ENTRY(GC, 0, mmRLC_INT_STAT) }, + { SOC15_REG_ENTRY(GC, 0, mmRLC_GPM_GENERAL_6) }, + { SOC15_REG_ENTRY(GC, 0, mmRLC_GPM_DEBUG_INST_A) }, + { SOC15_REG_ENTRY(GC, 0, mmRLC_GPM_DEBUG_INST_B) }, + { SOC15_REG_ENTRY(GC, 0, mmRLC_GPM_DEBUG_INST_ADDR) }, + { SOC15_REG_ENTRY(GC, 0, mmRLC_LX6_CORE_PDEBUG_INST) } };
Might also be nice to store a string with the register name so we can print the names rather than the offsets in the devcoredump. Would be easier to read.
Alex
Quoted 222 lines
Quoted 104 lines
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +1
-0+static const struct soc15_reg_golden golden_settings_gc_10_1[] = {SOC15_REG_GOLDEN_VALUE (GC ,0 , mmCB_HW_CONTROL_4,0xffffffff ,0x00400014 ), SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_CPF_CLK_CTRL,
0xfcff8fff,0xf8000100), @@ -9154,6 +9247,38 @@ static voidgfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)
amdgpu_ring_write(ring, gcr_cntl); /*GCR_CNTL*/
}drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +26
-0+static void gfx_v10_ip_dump (void *handle) { +struct amdgpu_device *adev = (struct amdgpu_device *)handle; +uint32_t reg_count =ARRAY_SIZE (gc_reg_list_10_1); +struct reg_pair *ptr, *gfx10_pair; +uint32_t new_count; + +/* +* Allocate memory based on the no of registers for each block. +* Since adev->ip_dump is common pointer and we add more +* memory to it as we move through different IP's of the ASIC. +* +* This memory needs to be freed after gpu reset once the ip +* dump is done in the devcoredump. +*/ + new_count = adev->num_regs + reg_count; + ptr =krealloc_array (adev->ip_dump, new_count,sizeof (struct reg_pair ),GFP_KERNEL ); +if (ptr) { + adev->ip_dump = ptr; + gfx10_pair = &ptr[adev->num_regs]; + adev->num_regs = new_count; + }else { +DRM_ERROR ("Failed to allocate memory for IP Dump\n" ); +return ; + } + +for (uint32_t i =0 ; i < reg_count; i++) {Some compilers don't like variables defined within code.
Sure
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4
-0+ gfx10_pair[i].offset = gc_reg_list_10_1[i].reg_offset; + gfx10_pair[i].value =RREG32 (SOC15_REG_ENTRY_OFFSET (gc_reg_list_10_1[i])); + } +}This seems too complicated. Just pass
drm_printerto theip_dumpcallbacks and let each IP dump their state directly. E.g.,for (i = 0; i < adev->num_ip_blocks; i++) { if (adev->ip_blocks[i].version->funcs->dump_ip_state) { drm_printf(&p, "$s IP state\n",adev->ip_blocks[i].version->funcs->name);
adev->ip_blocks[i].version->funcs->dump_ip_state((void *)adev, &p); } }I did try this but there is a problem in that approach.
drm_printeris created on the buffer which is provided by "amdgpu_devcoredump_read" and "amdgpu_devcoredump_read" is only called when the user opens the filecat
/sys/class/devcoredump/devcd1/dataand by that time gpu is reset and registers dont hold the value of interest. I tried this approach first and got all FFFF or reset values of the registers.
We need to capture the values in "amdgpu_coredump" which is immediately called after GPU reset and values are what we need for debugging.Crux is we need to capture data after gpu reset but read depends on when user open the file which is too late and hence we need to capture all the data in advance immediately after reset before recovery.
Ah, ok. I would suggest pre-allocating the memory to store the dump state in each IP block in
sw_init(). I guess we'd probably need two IP callbacks,dump_state()andprint_state(). Then call theip->dump_state()in the GPU recovery sequence before we try soft resets, then in devcore dump you can call theip->print_state()callback to print the state to the devcore dump output.The problem is there is no standard size to proallocate memory statically. Each IP hold different set of registers and every time i add a new register set i have to update the memory size and allocate dynamically but this can be handled. Another problem i see with this approach is where to store the memory ptr of each IP. One way is to Directly have ptrs for each IP in
amdgpu_deviceglobal struct or other way is to add ptr in IP specific struct withingamdgpu_devicewhich are to be traversed in a hardcode way as they arent in an array or list.For eg: we have to access each of these in a hardcode way in code as these structures arent in a list/array form to traverse for each ip.
/* nbio */
struct amdgpu_nbio nbio; /* hdp */ struct amdgpu_hdp hdp; /* smuio */ struct amdgpu_smuio smuio; /* mmhub */ struct amdgpu_mmhub mmhub; /* gfxhub */ struct amdgpu_gfxhub gfxhub; /* gfx */ struct amdgpu_gfx gfx; /* sdma */ struct amdgpu_sdma sdma; /* lsdma */ struct amdgpu_lsdma lsdma; /* uvd */ struct amdgpu_uvd uvd; /* vce */ struct amdgpu_vce vce;Quoted 2 lines
In current code the approach is to add memory based on need for each IP in the same ptr and keep adding registers to it and we can search with offset value in the devcoredump to debug.
Each IP can store the data however they want. E.g., in the gfx code, we can do:
gfx_v10_0_sw_init()
{
...adev->gfx.ip_state=kcalloc(2, gc_reg_list_10_1, GFP_KERNEL);
...
}
gfx_v10_0_dump_state()
{
for (i = 0; i <ARRAY_SIZE(gc_reg_list_10_1); i += 2)adev->gfx.ip_state[i] = gc_reg_list_10_1[i]; adev->gfx.ip_state[i+1] = RREG32(gc_reg_list_10_1[i]); } }
gfx_v10_0_print_state()
{
for (i = 0; i <ARRAY_SIZE(gc_reg_list_10_1); i += 2)drm_printf(&p, "0x%08x \t 0x%08x\n", adev->gfx.ip_state[i],adev->gfx.ip_state[i+1]);
}
gfx_v10_0_sw_fini()
{
...kfree(adev->gfx.ip_state);
...
}Then call
dumps_state()ingpu_recover()before we actually reset the GPU. The driver will then have the state cached and then you callprint_state()inamdgpu_devcoredump_read()to read out the cached state.Sure Alex. Will update the patches as per this approach.
Regards
Sunil Khatri
AlexQuoted 77 lines
Regards
SunilAlex
Regards
Sunil Khatri
Alex
+
static const struct amd_ip_funcs gfx_v10_0_ip_funcs = { .name = "gfx_v10_0", .early_init = gfx_v10_0_early_init, @@ -9170,7 +9295,7 @@static const struct
amd_ip_funcsgfx_v10_0_ip_funcs= {
.set_clockgating_state =gfx_v10_0_set_clockgating_state, .set_powergating_state =gfx_v10_0_set_powergating_state, .get_clockgating_state =gfx_v10_0_get_clockgating_state,drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +1
-1- .dump_ip_state = NULL, + .dump_ip_state = gfx_v10_ip_dump, };static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {diff --git
a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.hb/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.hindex4908044f7409..4c8e7fdb6976100644drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h +12
-1@@ -4830,6 +4830,8 @@#define mmCP_ECC_FIRSTOCCURRENCE_RING2_BASE_IDX 0#define mmGB_EDC_MODE 0x1e1e#define mmGB_EDC_MODE_BASE_IDX 0 +#define mmCP_DEBUG 0x1e1f +#define mmCP_DEBUG_BASE_IDX 0#define mmCP_FETCHER_SOURCE 0x1e22#define mmCP_FETCHER_SOURCE_BASE_IDX 0#define mmCP_PQ_WPTR_POLL_CNTL 0x1e23@@ -7778,6 +7780,8 @@#define mmCP_MES_DOORBELL_CONTROL5_BASE_IDX 1#define mmCP_MES_DOORBELL_CONTROL6 0x2841#define mmCP_MES_DOORBELL_CONTROL6_BASE_IDX 1 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR 0x2842 +#define mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR_BASE_IDX 1#define mmCP_MES_GP0_LO 0x2843#define mmCP_MES_GP0_LO_BASE_IDX 1#define mmCP_MES_GP0_HI 0x2844@@ -9332,10 +9336,16 @@#define mmRLC_LB_CNTR_INIT_1_BASE_IDX 1#define mmRLC_LB_CNTR_1 0x4c1c#define mmRLC_LB_CNTR_1_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_ADDR 0x4c1d +#define mmRLC_GPM_DEBUG_INST_ADDR_BASE_IDX 1#define mmRLC_JUMP_TABLE_RESTORE 0x4c1e#define mmRLC_JUMP_TABLE_RESTORE_BASE_IDX 1#define mmRLC_PG_DELAY_2 0x4c1f#define mmRLC_PG_DELAY_2_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_A 0x4c22 +#define mmRLC_GPM_DEBUG_INST_A_BASE_IDX 1 +#define mmRLC_GPM_DEBUG_INST_B 0x4c23 +#define mmRLC_GPM_DEBUG_INST_B_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_LSB 0x4c24#define mmRLC_GPU_CLOCK_COUNT_LSB_BASE_IDX 1#define mmRLC_GPU_CLOCK_COUNT_MSB 0x4c25@@ -9720,6 +9730,8 @@#define mmRLC_SPM_THREAD_TRACE_CTRL_BASE_IDX 1#define mmRLC_LB_CNTR_2 0x4de7#define mmRLC_LB_CNTR_2_BASE_IDX 1 +#define mmRLC_LX6_CORE_PDEBUG_INST 0x4deb +#define mmRLC_LX6_CORE_PDEBUG_INST_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_CTRL 0x4df1#define mmRLC_CPAXI_DOORBELL_MON_CTRL_BASE_IDX 1#define mmRLC_CPAXI_DOORBELL_MON_STAT 0x4df2 --2.34.1