summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c')
-rw-r--r--drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c114
1 files changed, 57 insertions, 57 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index b1f2684d854a..51c1745c8369 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -1304,7 +1304,7 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev)
u32 *hpd;
size_t mec_hpd_size;
- bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+ bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
/* take ownership of the relevant compute queues */
amdgpu_gfx_compute_queue_acquire(adev);
@@ -2001,7 +2001,8 @@ static int gfx_v8_0_sw_init(void *handle)
for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
- if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j))
+ if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+ k, j))
continue;
r = gfx_v8_0_compute_ring_init(adev,
@@ -2015,19 +2016,19 @@ static int gfx_v8_0_sw_init(void *handle)
}
}
- r = amdgpu_gfx_kiq_init(adev, GFX8_MEC_HPD_SIZE);
+ r = amdgpu_gfx_kiq_init(adev, GFX8_MEC_HPD_SIZE, 0);
if (r) {
DRM_ERROR("Failed to init KIQ BOs!\n");
return r;
}
- kiq = &adev->gfx.kiq;
- r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
+ kiq = &adev->gfx.kiq[0];
+ r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq, 0);
if (r)
return r;
/* create MQD for all compute queues as well as KIQ for SRIOV case */
- r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct vi_mqd_allocation));
+ r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct vi_mqd_allocation), 0);
if (r)
return r;
@@ -2050,9 +2051,9 @@ static int gfx_v8_0_sw_fini(void *handle)
for (i = 0; i < adev->gfx.num_compute_rings; i++)
amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
- amdgpu_gfx_mqd_sw_fini(adev);
- amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
- amdgpu_gfx_kiq_fini(adev);
+ amdgpu_gfx_mqd_sw_fini(adev, 0);
+ amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+ amdgpu_gfx_kiq_fini(adev, 0);
gfx_v8_0_mec_fini(adev);
amdgpu_gfx_rlc_fini(adev);
@@ -3394,7 +3395,8 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev)
}
static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev,
- u32 se_num, u32 sh_num, u32 instance)
+ u32 se_num, u32 sh_num, u32 instance,
+ int xcc_id)
{
u32 data;
@@ -3417,7 +3419,7 @@ static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev,
}
static void gfx_v8_0_select_me_pipe_q(struct amdgpu_device *adev,
- u32 me, u32 pipe, u32 q, u32 vm)
+ u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
{
vi_srbm_select(adev, me, pipe, q, vm);
}
@@ -3578,13 +3580,13 @@ gfx_v8_0_write_harvested_raster_configs(struct amdgpu_device *adev,
}
/* GRBM_GFX_INDEX has a different offset on VI */
- gfx_v8_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff, 0);
WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
}
/* GRBM_GFX_INDEX has a different offset on VI */
- gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
}
static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
@@ -3600,13 +3602,13 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
mutex_lock(&adev->grbm_idx_mutex);
for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
- gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
data = gfx_v8_0_get_rb_active_bitmap(adev);
active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
rb_bitmap_width_per_sh);
}
}
- gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
adev->gfx.config.backend_enable_mask = active_rbs;
adev->gfx.config.num_rbs = hweight32(active_rbs);
@@ -3629,7 +3631,7 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
/* cache the values for userspace */
for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
- gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
adev->gfx.config.rb_config[i][j].rb_backend_disable =
RREG32(mmCC_RB_BACKEND_DISABLE);
adev->gfx.config.rb_config[i][j].user_rb_backend_disable =
@@ -3640,7 +3642,7 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
RREG32(mmPA_SC_RASTER_CONFIG_1);
}
}
- gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
mutex_unlock(&adev->grbm_idx_mutex);
}
@@ -3787,7 +3789,7 @@ static void gfx_v8_0_constants_init(struct amdgpu_device *adev)
* making sure that the following register writes will be broadcasted
* to all the shaders
*/
- gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
WREG32(mmPA_SC_FIFO_SIZE,
(adev->gfx.config.sc_prim_fifo_size_frontend <<
@@ -3818,7 +3820,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
mutex_lock(&adev->grbm_idx_mutex);
for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
- gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
for (k = 0; k < adev->usec_timeout; k++) {
if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0)
break;
@@ -3826,7 +3828,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
}
if (k == adev->usec_timeout) {
gfx_v8_0_select_se_sh(adev, 0xffffffff,
- 0xffffffff, 0xffffffff);
+ 0xffffffff, 0xffffffff, 0);
mutex_unlock(&adev->grbm_idx_mutex);
DRM_INFO("Timeout wait for RLC serdes %u,%u\n",
i, j);
@@ -3834,7 +3836,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
}
}
}
- gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
mutex_unlock(&adev->grbm_idx_mutex);
mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -4281,7 +4283,6 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev)
/* start the ring */
amdgpu_ring_clear_ring(ring);
gfx_v8_0_cp_gfx_start(adev);
- ring->sched.ready = true;
return 0;
}
@@ -4292,7 +4293,7 @@ static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
WREG32(mmCP_MEC_CNTL, 0);
} else {
WREG32(mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
- adev->gfx.kiq.ring.sched.ready = false;
+ adev->gfx.kiq[0].ring.sched.ready = false;
}
udelay(50);
}
@@ -4314,12 +4315,12 @@ static void gfx_v8_0_kiq_setting(struct amdgpu_ring *ring)
static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev)
{
- struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+ struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
uint64_t queue_mask = 0;
int r, i;
for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) {
- if (!test_bit(i, adev->gfx.mec.queue_bitmap))
+ if (!test_bit(i, adev->gfx.mec_bitmap[0].queue_bitmap))
continue;
/* This situation may be hit in the future if a new HW
@@ -4595,14 +4596,13 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
struct vi_mqd *mqd = ring->mqd_ptr;
- int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
gfx_v8_0_kiq_setting(ring);
if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
/* reset MQD to a clean status */
- if (adev->gfx.mec.mqd_backup[mqd_idx])
- memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
+ if (adev->gfx.kiq[0].mqd_backup)
+ memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct vi_mqd_allocation));
/* reset ring buffer */
ring->wptr = 0;
@@ -4625,8 +4625,8 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
vi_srbm_select(adev, 0, 0, 0, 0);
mutex_unlock(&adev->srbm_mutex);
- if (adev->gfx.mec.mqd_backup[mqd_idx])
- memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
+ if (adev->gfx.kiq[0].mqd_backup)
+ memcpy(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(struct vi_mqd_allocation));
}
return 0;
@@ -4650,15 +4650,13 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)
if (adev->gfx.mec.mqd_backup[mqd_idx])
memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
- } else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
- /* reset MQD to a clean status */
+ } else {
+ /* restore MQD to a clean status */
if (adev->gfx.mec.mqd_backup[mqd_idx])
memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
/* reset ring buffer */
ring->wptr = 0;
amdgpu_ring_clear_ring(ring);
- } else {
- amdgpu_ring_clear_ring(ring);
}
return 0;
}
@@ -4678,21 +4676,22 @@ static int gfx_v8_0_kiq_resume(struct amdgpu_device *adev)
struct amdgpu_ring *ring;
int r;
- ring = &adev->gfx.kiq.ring;
+ ring = &adev->gfx.kiq[0].ring;
r = amdgpu_bo_reserve(ring->mqd_obj, false);
if (unlikely(r != 0))
return r;
r = amdgpu_bo_kmap(ring->mqd_obj, &ring->mqd_ptr);
- if (unlikely(r != 0))
+ if (unlikely(r != 0)) {
+ amdgpu_bo_unreserve(ring->mqd_obj);
return r;
+ }
gfx_v8_0_kiq_init_queue(ring);
amdgpu_bo_kunmap(ring->mqd_obj);
ring->mqd_ptr = NULL;
amdgpu_bo_unreserve(ring->mqd_obj);
- ring->sched.ready = true;
return 0;
}
@@ -4741,7 +4740,7 @@ static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev)
if (r)
return r;
- ring = &adev->gfx.kiq.ring;
+ ring = &adev->gfx.kiq[0].ring;
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
@@ -4808,7 +4807,7 @@ static int gfx_v8_0_hw_init(void *handle)
static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
{
int r, i;
- struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+ struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
r = amdgpu_ring_alloc(kiq_ring, 6 * adev->gfx.num_compute_rings);
if (r)
@@ -4902,7 +4901,7 @@ static int gfx_v8_0_hw_fini(void *handle)
pr_debug("For SRIOV client, shouldn't do anything.\n");
return 0;
}
- amdgpu_gfx_rlc_enter_safe_mode(adev);
+ amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
if (!gfx_v8_0_wait_for_idle(adev))
gfx_v8_0_cp_enable(adev, false);
else
@@ -4911,7 +4910,7 @@ static int gfx_v8_0_hw_fini(void *handle)
adev->gfx.rlc.funcs->stop(adev);
else
pr_err("rlc is busy, skip halt rlc\n");
- amdgpu_gfx_rlc_exit_safe_mode(adev);
+ amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
return 0;
}
@@ -5216,7 +5215,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
*(out++) = RREG32(mmSQ_IND_DATA);
}
-static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
{
/* type 0 wave data */
dst[(*no_fields)++] = 0;
@@ -5241,7 +5240,7 @@ static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
}
-static void gfx_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
uint32_t wave, uint32_t start,
uint32_t size, uint32_t *dst)
{
@@ -5263,6 +5262,7 @@ static int gfx_v8_0_early_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ adev->gfx.xcc_mask = 1;
adev->gfx.num_gfx_rings = GFX8_NUM_GFX_RINGS;
adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
AMDGPU_MAX_COMPUTE_RINGS);
@@ -5376,7 +5376,7 @@ static int gfx_v8_0_set_powergating_state(void *handle,
AMD_PG_SUPPORT_RLC_SMU_HS |
AMD_PG_SUPPORT_CP |
AMD_PG_SUPPORT_GFX_DMG))
- amdgpu_gfx_rlc_enter_safe_mode(adev);
+ amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
switch (adev->asic_type) {
case CHIP_CARRIZO:
case CHIP_STONEY:
@@ -5430,7 +5430,7 @@ static int gfx_v8_0_set_powergating_state(void *handle,
AMD_PG_SUPPORT_RLC_SMU_HS |
AMD_PG_SUPPORT_CP |
AMD_PG_SUPPORT_GFX_DMG))
- amdgpu_gfx_rlc_exit_safe_mode(adev);
+ amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
return 0;
}
@@ -5481,7 +5481,7 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev,
{
uint32_t data;
- gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
@@ -5535,7 +5535,7 @@ static bool gfx_v8_0_is_rlc_enabled(struct amdgpu_device *adev)
return true;
}
-static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
{
uint32_t data;
unsigned i;
@@ -5562,7 +5562,7 @@ static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev)
}
}
-static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
{
uint32_t data;
unsigned i;
@@ -5621,7 +5621,7 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
{
uint32_t temp, data;
- amdgpu_gfx_rlc_enter_safe_mode(adev);
+ amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
/* It is disabled by HW by default */
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
@@ -5717,7 +5717,7 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
gfx_v8_0_wait_for_rlc_serdes(adev);
}
- amdgpu_gfx_rlc_exit_safe_mode(adev);
+ amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
}
static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
@@ -5727,7 +5727,7 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
temp = data = RREG32(mmRLC_CGCG_CGLS_CTRL);
- amdgpu_gfx_rlc_enter_safe_mode(adev);
+ amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
temp1 = data1 = RREG32(mmRLC_CGTT_MGCG_OVERRIDE);
@@ -5810,7 +5810,7 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
gfx_v8_0_wait_for_rlc_serdes(adev);
- amdgpu_gfx_rlc_exit_safe_mode(adev);
+ amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
}
static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev,
bool enable)
@@ -6723,11 +6723,11 @@ static void gfx_v8_0_parse_sq_irq(struct amdgpu_device *adev, unsigned ih_data,
*/
if (from_wq) {
mutex_lock(&adev->grbm_idx_mutex);
- gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id);
+ gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id, 0);
sq_edc_source = REG_GET_FIELD(RREG32(mmSQ_EDC_INFO), SQ_EDC_INFO, SOURCE);
- gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
mutex_unlock(&adev->grbm_idx_mutex);
}
@@ -7001,7 +7001,7 @@ static void gfx_v8_0_set_ring_funcs(struct amdgpu_device *adev)
{
int i;
- adev->gfx.kiq.ring.funcs = &gfx_v8_0_ring_funcs_kiq;
+ adev->gfx.kiq[0].ring.funcs = &gfx_v8_0_ring_funcs_kiq;
for (i = 0; i < adev->gfx.num_gfx_rings; i++)
adev->gfx.gfx_ring[i].funcs = &gfx_v8_0_ring_funcs_gfx;
@@ -7116,7 +7116,7 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev)
mask = 1;
ao_bitmap = 0;
counter = 0;
- gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
if (i < 4 && j < 2)
gfx_v8_0_set_user_cu_inactive_bitmap(
adev, disable_masks[i * 2 + j]);
@@ -7137,7 +7137,7 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev)
cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
}
}
- gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
mutex_unlock(&adev->grbm_idx_mutex);
cu_info->number = active_cu_number;
generated by cgit v1.2.3-70-g09d2 (git 2.51.1) at 2025-11-04 06:47:22 +0000