summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdgpu
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu.h12
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_device.c38
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c8
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c16
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c7
-rw-r--r--drivers/gpu/drm/amd/amdgpu/cz_dpm.c70
-rw-r--r--drivers/gpu/drm/amd/amdgpu/dce_v10_0.c4
-rw-r--r--drivers/gpu/drm/amd/amdgpu/dce_v11_0.c4
-rw-r--r--drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c46
-rw-r--r--drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c47
-rw-r--r--drivers/gpu/drm/amd/amdgpu/vce_v3_0.c48
11 files changed, 231 insertions, 69 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 01657830b470..31b00f91cfcd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1614,6 +1614,9 @@ struct amdgpu_uvd {
#define AMDGPU_MAX_VCE_HANDLES 16
#define AMDGPU_VCE_FIRMWARE_OFFSET 256
+#define AMDGPU_VCE_HARVEST_VCE0 (1 << 0)
+#define AMDGPU_VCE_HARVEST_VCE1 (1 << 1)
+
struct amdgpu_vce {
struct amdgpu_bo *vcpu_bo;
uint64_t gpu_addr;
@@ -1626,6 +1629,7 @@ struct amdgpu_vce {
const struct firmware *fw; /* VCE firmware */
struct amdgpu_ring ring[AMDGPU_MAX_VCE_RINGS];
struct amdgpu_irq_src irq;
+ unsigned harvest_config;
};
/*
@@ -1862,6 +1866,12 @@ typedef void (*amdgpu_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
typedef uint32_t (*amdgpu_block_rreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
typedef void (*amdgpu_block_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t, uint32_t);
+struct amdgpu_ip_block_status {
+ bool valid;
+ bool sw;
+ bool hw;
+};
+
struct amdgpu_device {
struct device *dev;
struct drm_device *ddev;
@@ -2004,7 +2014,7 @@ struct amdgpu_device {
const struct amdgpu_ip_block_version *ip_blocks;
int num_ip_blocks;
- bool *ip_block_enabled;
+ struct amdgpu_ip_block_status *ip_block_status;
struct mutex mn_lock;
DECLARE_HASHTABLE(mn_hash, 7);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index d79009b65867..99f158e1baff 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -1191,8 +1191,9 @@ static int amdgpu_early_init(struct amdgpu_device *adev)
return -EINVAL;
}
- adev->ip_block_enabled = kcalloc(adev->num_ip_blocks, sizeof(bool), GFP_KERNEL);
- if (adev->ip_block_enabled == NULL)
+ adev->ip_block_status = kcalloc(adev->num_ip_blocks,
+ sizeof(struct amdgpu_ip_block_status), GFP_KERNEL);
+ if (adev->ip_block_status == NULL)
return -ENOMEM;
if (adev->ip_blocks == NULL) {
@@ -1203,18 +1204,18 @@ static int amdgpu_early_init(struct amdgpu_device *adev)
for (i = 0; i < adev->num_ip_blocks; i++) {
if ((amdgpu_ip_block_mask & (1 << i)) == 0) {
DRM_ERROR("disabled ip block: %d\n", i);
- adev->ip_block_enabled[i] = false;
+ adev->ip_block_status[i].valid = false;
} else {
if (adev->ip_blocks[i].funcs->early_init) {
r = adev->ip_blocks[i].funcs->early_init((void *)adev);
if (r == -ENOENT)
- adev->ip_block_enabled[i] = false;
+ adev->ip_block_status[i].valid = false;
else if (r)
return r;
else
- adev->ip_block_enabled[i] = true;
+ adev->ip_block_status[i].valid = true;
} else {
- adev->ip_block_enabled[i] = true;
+ adev->ip_block_status[i].valid = true;
}
}
}
@@ -1227,11 +1228,12 @@ static int amdgpu_init(struct amdgpu_device *adev)
int i, r;
for (i = 0; i < adev->num_ip_blocks; i++) {
- if (!adev->ip_block_enabled[i])
+ if (!adev->ip_block_status[i].valid)
continue;
r = adev->ip_blocks[i].funcs->sw_init((void *)adev);
if (r)
return r;
+ adev->ip_block_status[i].sw = true;
/* need to do gmc hw init early so we can allocate gpu mem */
if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC) {
r = amdgpu_vram_scratch_init(adev);
@@ -1243,11 +1245,12 @@ static int amdgpu_init(struct amdgpu_device *adev)
r = amdgpu_wb_init(adev);
if (r)
return r;
+ adev->ip_block_status[i].hw = true;
}
}
for (i = 0; i < adev->num_ip_blocks; i++) {
- if (!adev->ip_block_enabled[i])
+ if (!adev->ip_block_status[i].sw)
continue;
/* gmc hw init is done early */
if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC)
@@ -1255,6 +1258,7 @@ static int amdgpu_init(struct amdgpu_device *adev)
r = adev->ip_blocks[i].funcs->hw_init((void *)adev);
if (r)
return r;
+ adev->ip_block_status[i].hw = true;
}
return 0;
@@ -1265,7 +1269,7 @@ static int amdgpu_late_init(struct amdgpu_device *adev)
int i = 0, r;
for (i = 0; i < adev->num_ip_blocks; i++) {
- if (!adev->ip_block_enabled[i])
+ if (!adev->ip_block_status[i].valid)
continue;
/* enable clockgating to save power */
r = adev->ip_blocks[i].funcs->set_clockgating_state((void *)adev,
@@ -1287,7 +1291,7 @@ static int amdgpu_fini(struct amdgpu_device *adev)
int i, r;
for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
- if (!adev->ip_block_enabled[i])
+ if (!adev->ip_block_status[i].hw)
continue;
if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC) {
amdgpu_wb_fini(adev);
@@ -1300,14 +1304,16 @@ static int amdgpu_fini(struct amdgpu_device *adev)
return r;
r = adev->ip_blocks[i].funcs->hw_fini((void *)adev);
/* XXX handle errors */
+ adev->ip_block_status[i].hw = false;
}
for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
- if (!adev->ip_block_enabled[i])
+ if (!adev->ip_block_status[i].sw)
continue;
r = adev->ip_blocks[i].funcs->sw_fini((void *)adev);
/* XXX handle errors */
- adev->ip_block_enabled[i] = false;
+ adev->ip_block_status[i].sw = false;
+ adev->ip_block_status[i].valid = false;
}
return 0;
@@ -1318,7 +1324,7 @@ static int amdgpu_suspend(struct amdgpu_device *adev)
int i, r;
for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
- if (!adev->ip_block_enabled[i])
+ if (!adev->ip_block_status[i].valid)
continue;
/* ungate blocks so that suspend can properly shut them down */
r = adev->ip_blocks[i].funcs->set_clockgating_state((void *)adev,
@@ -1336,7 +1342,7 @@ static int amdgpu_resume(struct amdgpu_device *adev)
int i, r;
for (i = 0; i < adev->num_ip_blocks; i++) {
- if (!adev->ip_block_enabled[i])
+ if (!adev->ip_block_status[i].valid)
continue;
r = adev->ip_blocks[i].funcs->resume(adev);
if (r)
@@ -1582,8 +1588,8 @@ void amdgpu_device_fini(struct amdgpu_device *adev)
amdgpu_fence_driver_fini(adev);
amdgpu_fbdev_fini(adev);
r = amdgpu_fini(adev);
- kfree(adev->ip_block_enabled);
- adev->ip_block_enabled = NULL;
+ kfree(adev->ip_block_status);
+ adev->ip_block_status = NULL;
adev->accel_working = false;
/* free i2c buses */
amdgpu_i2c_fini(adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index ae43b58c9733..4afc507820c0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -449,7 +449,7 @@ out:
* vital here, so they are not reported back to userspace.
*/
static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
- struct amdgpu_bo_va *bo_va)
+ struct amdgpu_bo_va *bo_va, uint32_t operation)
{
struct ttm_validate_buffer tv, *entry;
struct amdgpu_bo_list_entry *vm_bos;
@@ -485,7 +485,9 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
if (r)
goto error_unlock;
- r = amdgpu_vm_bo_update(adev, bo_va, &bo_va->bo->tbo.mem);
+
+ if (operation == AMDGPU_VA_OP_MAP)
+ r = amdgpu_vm_bo_update(adev, bo_va, &bo_va->bo->tbo.mem);
error_unlock:
mutex_unlock(&bo_va->vm->mutex);
@@ -580,7 +582,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
}
if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE))
- amdgpu_gem_va_update_vm(adev, bo_va);
+ amdgpu_gem_va_update_vm(adev, bo_va, args->operation);
drm_gem_object_unreference_unlocked(gobj);
return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index 52dff75aac6f..bc0fac618a3f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -180,16 +180,16 @@ int amdgpu_ib_schedule(struct amdgpu_device *adev, unsigned num_ibs,
if (vm) {
/* do context switch */
amdgpu_vm_flush(ring, vm, ib->sync.last_vm_update);
- }
- if (vm && ring->funcs->emit_gds_switch)
- amdgpu_ring_emit_gds_switch(ring, ib->vm->ids[ring->idx].id,
- ib->gds_base, ib->gds_size,
- ib->gws_base, ib->gws_size,
- ib->oa_base, ib->oa_size);
+ if (ring->funcs->emit_gds_switch)
+ amdgpu_ring_emit_gds_switch(ring, ib->vm->ids[ring->idx].id,
+ ib->gds_base, ib->gds_size,
+ ib->gws_base, ib->gws_size,
+ ib->oa_base, ib->oa_size);
- if (ring->funcs->emit_hdp_flush)
- amdgpu_ring_emit_hdp_flush(ring);
+ if (ring->funcs->emit_hdp_flush)
+ amdgpu_ring_emit_hdp_flush(ring);
+ }
old_ctx = ring->current_ctx;
for (i = 0; i < num_ibs; ++i) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
index 5533434c7a8f..9736892bcdf9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -235,7 +235,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
for (i = 0; i < adev->num_ip_blocks; i++) {
if (adev->ip_blocks[i].type == type &&
- adev->ip_block_enabled[i]) {
+ adev->ip_block_status[i].valid) {
ip.hw_ip_version_major = adev->ip_blocks[i].major;
ip.hw_ip_version_minor = adev->ip_blocks[i].minor;
ip.capabilities_flags = 0;
@@ -274,7 +274,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
for (i = 0; i < adev->num_ip_blocks; i++)
if (adev->ip_blocks[i].type == type &&
- adev->ip_block_enabled[i] &&
+ adev->ip_block_status[i].valid &&
count < AMDGPU_HW_IP_INSTANCE_MAX_COUNT)
count++;
@@ -416,7 +416,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
return n ? -EFAULT : 0;
}
case AMDGPU_INFO_DEV_INFO: {
- struct drm_amdgpu_info_device dev_info;
+ struct drm_amdgpu_info_device dev_info = {};
struct amdgpu_cu_info cu_info;
dev_info.device_id = dev->pdev->device;
@@ -459,6 +459,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
memcpy(&dev_info.cu_bitmap[0], &cu_info.bitmap[0], sizeof(cu_info.bitmap));
dev_info.vram_type = adev->mc.vram_type;
dev_info.vram_bit_width = adev->mc.vram_width;
+ dev_info.vce_harvest_config = adev->vce.harvest_config;
return copy_to_user(out, &dev_info,
min((size_t)size, sizeof(dev_info))) ? -EFAULT : 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/cz_dpm.c b/drivers/gpu/drm/amd/amdgpu/cz_dpm.c
index 1a2d419cbf16..ace870afc7d4 100644
--- a/drivers/gpu/drm/amd/amdgpu/cz_dpm.c
+++ b/drivers/gpu/drm/amd/amdgpu/cz_dpm.c
@@ -494,29 +494,67 @@ static void cz_dpm_fini(struct amdgpu_device *adev)
amdgpu_free_extended_power_table(adev);
}
+#define ixSMUSVI_NB_CURRENTVID 0xD8230044
+#define CURRENT_NB_VID_MASK 0xff000000
+#define CURRENT_NB_VID__SHIFT 24
+#define ixSMUSVI_GFX_CURRENTVID 0xD8230048
+#define CURRENT_GFX_VID_MASK 0xff000000
+#define CURRENT_GFX_VID__SHIFT 24
+
static void
cz_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
struct seq_file *m)
{
+ struct cz_power_info *pi = cz_get_pi(adev);
struct amdgpu_clock_voltage_dependency_table *table =
&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
- u32 current_index =
- (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) &
- TARGET_AND_CURRENT_PROFILE_INDEX__CURR_SCLK_INDEX_MASK) >>
- TARGET_AND_CURRENT_PROFILE_INDEX__CURR_SCLK_INDEX__SHIFT;
- u32 sclk, tmp;
- u16 vddc;
-
- if (current_index >= NUM_SCLK_LEVELS) {
- seq_printf(m, "invalid dpm profile %d\n", current_index);
+ struct amdgpu_uvd_clock_voltage_dependency_table *uvd_table =
+ &adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table;
+ struct amdgpu_vce_clock_voltage_dependency_table *vce_table =
+ &adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table;
+ u32 sclk_index = REG_GET_FIELD(RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX),
+ TARGET_AND_CURRENT_PROFILE_INDEX, CURR_SCLK_INDEX);
+ u32 uvd_index = REG_GET_FIELD(RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX_2),
+ TARGET_AND_CURRENT_PROFILE_INDEX_2, CURR_UVD_INDEX);
+ u32 vce_index = REG_GET_FIELD(RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX_2),
+ TARGET_AND_CURRENT_PROFILE_INDEX_2, CURR_VCE_INDEX);
+ u32 sclk, vclk, dclk, ecclk, tmp;
+ u16 vddnb, vddgfx;
+
+ if (sclk_index >= NUM_SCLK_LEVELS) {
+ seq_printf(m, "invalid sclk dpm profile %d\n", sclk_index);
} else {
- sclk = table->entries[current_index].clk;
- tmp = (RREG32_SMC(ixSMU_VOLTAGE_STATUS) &
- SMU_VOLTAGE_STATUS__SMU_VOLTAGE_CURRENT_LEVEL_MASK) >>
- SMU_VOLTAGE_STATUS__SMU_VOLTAGE_CURRENT_LEVEL__SHIFT;
- vddc = cz_convert_8bit_index_to_voltage(adev, (u16)tmp);
- seq_printf(m, "power level %d sclk: %u vddc: %u\n",
- current_index, sclk, vddc);
+ sclk = table->entries[sclk_index].clk;
+ seq_printf(m, "%u sclk: %u\n", sclk_index, sclk);
+ }
+
+ tmp = (RREG32_SMC(ixSMUSVI_NB_CURRENTVID) &
+ CURRENT_NB_VID_MASK) >> CURRENT_NB_VID__SHIFT;
+ vddnb = cz_convert_8bit_index_to_voltage(adev, (u16)tmp);
+ tmp = (RREG32_SMC(ixSMUSVI_GFX_CURRENTVID) &
+ CURRENT_GFX_VID_MASK) >> CURRENT_GFX_VID__SHIFT;
+ vddgfx = cz_convert_8bit_index_to_voltage(adev, (u16)tmp);
+ seq_printf(m, "vddnb: %u vddgfx: %u\n", vddnb, vddgfx);
+
+ seq_printf(m, "uvd %sabled\n", pi->uvd_power_gated ? "dis" : "en");
+ if (!pi->uvd_power_gated) {
+ if (uvd_index >= CZ_MAX_HARDWARE_POWERLEVELS) {
+ seq_printf(m, "invalid uvd dpm level %d\n", uvd_index);
+ } else {
+ vclk = uvd_table->entries[uvd_index].vclk;
+ dclk = uvd_table->entries[uvd_index].dclk;
+ seq_printf(m, "%u uvd vclk: %u dclk: %u\n", uvd_index, vclk, dclk);
+ }
+ }
+
+ seq_printf(m, "vce %sabled\n", pi->vce_power_gated ? "dis" : "en");
+ if (!pi->vce_power_gated) {
+ if (vce_index >= CZ_MAX_HARDWARE_POWERLEVELS) {
+ seq_printf(m, "invalid vce dpm level %d\n", vce_index);
+ } else {
+ ecclk = vce_table->entries[vce_index].ecclk;
+ seq_printf(m, "%u vce ecclk: %u\n", vce_index, ecclk);
+ }
}
}
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
index 6e77964f1b64..e70a26f587a0 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
@@ -2632,6 +2632,7 @@ static void dce_v10_0_crtc_dpms(struct drm_crtc *crtc, int mode)
struct drm_device *dev = crtc->dev;
struct amdgpu_device *adev = dev->dev_private;
struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+ unsigned type;
switch (mode) {
case DRM_MODE_DPMS_ON:
@@ -2640,6 +2641,9 @@ static void dce_v10_0_crtc_dpms(struct drm_crtc *crtc, int mode)
dce_v10_0_vga_enable(crtc, true);
amdgpu_atombios_crtc_blank(crtc, ATOM_DISABLE);
dce_v10_0_vga_enable(crtc, false);
+ /* Make sure VBLANK interrupt is still enabled */
+ type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id);
+ amdgpu_irq_update(adev, &adev->crtc_irq, type);
drm_vblank_post_modeset(dev, amdgpu_crtc->crtc_id);
dce_v10_0_crtc_load_lut(crtc);
break;
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
index 7f7abb0e0be5..dcb402ee048a 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
@@ -2631,6 +2631,7 @@ static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode)
struct drm_device *dev = crtc->dev;
struct amdgpu_device *adev = dev->dev_private;
struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+ unsigned type;
switch (mode) {
case DRM_MODE_DPMS_ON:
@@ -2639,6 +2640,9 @@ static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode)
dce_v11_0_vga_enable(crtc, true);
amdgpu_atombios_crtc_blank(crtc, ATOM_DISABLE);
dce_v11_0_vga_enable(crtc, false);
+ /* Make sure VBLANK interrupt is still enabled */
+ type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id);
+ amdgpu_irq_update(adev, &adev->crtc_irq, type);
drm_vblank_post_modeset(dev, amdgpu_crtc->crtc_id);
dce_v11_0_crtc_load_lut(crtc);
break;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 2c188fb9fd22..2db6ab0a543d 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -2561,7 +2561,7 @@ static bool gfx_v7_0_ring_emit_semaphore(struct amdgpu_ring *ring,
* sheduling on the ring. This function schedules the IB
* on the gfx ring for execution by the GPU.
*/
-static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring,
+static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
struct amdgpu_ib *ib)
{
bool need_ctx_switch = ring->current_ctx != ib->ctx;
@@ -2569,15 +2569,10 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring,
u32 next_rptr = ring->wptr + 5;
/* drop the CE preamble IB for the same context */
- if ((ring->type == AMDGPU_RING_TYPE_GFX) &&
- (ib->flags & AMDGPU_IB_FLAG_PREAMBLE) &&
- !need_ctx_switch)
+ if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch)
return;
- if (ring->type == AMDGPU_RING_TYPE_COMPUTE)
- control |= INDIRECT_BUFFER_VALID;
-
- if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX)
+ if (need_ctx_switch)
next_rptr += 2;
next_rptr += 4;
@@ -2588,7 +2583,7 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring,
amdgpu_ring_write(ring, next_rptr);
/* insert SWITCH_BUFFER packet before first IB in the ring frame */
- if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) {
+ if (need_ctx_switch) {
amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
amdgpu_ring_write(ring, 0);
}
@@ -2611,6 +2606,35 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring,
amdgpu_ring_write(ring, control);
}
+static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
+ struct amdgpu_ib *ib)
+{
+ u32 header, control = 0;
+ u32 next_rptr = ring->wptr + 5;
+
+ control |= INDIRECT_BUFFER_VALID;
+ next_rptr += 4;
+ amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+ amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
+ amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
+ amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff);
+ amdgpu_ring_write(ring, next_rptr);
+
+ header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
+
+ control |= ib->length_dw |
+ (ib->vm ? (ib->vm->ids[ring->idx].id << 24) : 0);
+
+ amdgpu_ring_write(ring, header);
+ amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+ (2 << 0) |
+#endif
+ (ib->gpu_addr & 0xFFFFFFFC));
+ amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF);
+ amdgpu_ring_write(ring, control);
+}
+
/**
* gfx_v7_0_ring_test_ib - basic ring IB test
*
@@ -5555,7 +5579,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
.get_wptr = gfx_v7_0_ring_get_wptr_gfx,
.set_wptr = gfx_v7_0_ring_set_wptr_gfx,
.parse_cs = NULL,
- .emit_ib = gfx_v7_0_ring_emit_ib,
+ .emit_ib = gfx_v7_0_ring_emit_ib_gfx,
.emit_fence = gfx_v7_0_ring_emit_fence_gfx,
.emit_semaphore = gfx_v7_0_ring_emit_semaphore,
.emit_vm_flush = gfx_v7_0_ring_emit_vm_flush,
@@ -5571,7 +5595,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
.get_wptr = gfx_v7_0_ring_get_wptr_compute,
.set_wptr = gfx_v7_0_ring_set_wptr_compute,
.parse_cs = NULL,
- .emit_ib = gfx_v7_0_ring_emit_ib,
+ .emit_ib = gfx_v7_0_ring_emit_ib_compute,
.emit_fence = gfx_v7_0_ring_emit_fence_compute,
.emit_semaphore = gfx_v7_0_ring_emit_semaphore,
.emit_vm_flush = gfx_v7_0_ring_emit_vm_flush,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index 1c7c992dea37..9e1d4ddbf475 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -3753,7 +3753,7 @@ static void gfx_v8_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
amdgpu_ring_write(ring, 0x20); /* poll interval */
}
-static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring,
+static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
struct amdgpu_ib *ib)
{
bool need_ctx_switch = ring->current_ctx != ib->ctx;
@@ -3761,15 +3761,10 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring,
u32 next_rptr = ring->wptr + 5;
/* drop the CE preamble IB for the same context */
- if ((ring->type == AMDGPU_RING_TYPE_GFX) &&
- (ib->flags & AMDGPU_IB_FLAG_PREAMBLE) &&
- !need_ctx_switch)
+ if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch)
return;
- if (ring->type == AMDGPU_RING_TYPE_COMPUTE)
- control |= INDIRECT_BUFFER_VALID;
-
- if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX)
+ if (need_ctx_switch)
next_rptr += 2;
next_rptr += 4;
@@ -3780,7 +3775,7 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring,
amdgpu_ring_write(ring, next_rptr);
/* insert SWITCH_BUFFER packet before first IB in the ring frame */
- if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) {
+ if (need_ctx_switch) {
amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
amdgpu_ring_write(ring, 0);
}
@@ -3803,6 +3798,36 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring,
amdgpu_ring_write(ring, control);
}
+static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
+ struct amdgpu_ib *ib)
+{
+ u32 header, control = 0;
+ u32 next_rptr = ring->wptr + 5;
+
+ control |= INDIRECT_BUFFER_VALID;
+
+ next_rptr += 4;
+ amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+ amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
+ amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
+ amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff);
+ amdgpu_ring_write(ring, next_rptr);
+
+ header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
+
+ control |= ib->length_dw |
+ (ib->vm ? (ib->vm->ids[ring->idx].id << 24) : 0);
+
+ amdgpu_ring_write(ring, header);
+ amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+ (2 << 0) |
+#endif
+ (ib->gpu_addr & 0xFFFFFFFC));
+ amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF);
+ amdgpu_ring_write(ring, control);
+}
+
static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
u64 seq, unsigned flags)
{
@@ -4224,7 +4249,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
.get_wptr = gfx_v8_0_ring_get_wptr_gfx,
.set_wptr = gfx_v8_0_ring_set_wptr_gfx,
.parse_cs = NULL,
- .emit_ib = gfx_v8_0_ring_emit_ib,
+ .emit_ib = gfx_v8_0_ring_emit_ib_gfx,
.emit_fence = gfx_v8_0_ring_emit_fence_gfx,
.emit_semaphore = gfx_v8_0_ring_emit_semaphore,
.emit_vm_flush = gfx_v8_0_ring_emit_vm_flush,
@@ -4240,7 +4265,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
.get_wptr = gfx_v8_0_ring_get_wptr_compute,
.set_wptr = gfx_v8_0_ring_set_wptr_compute,
.parse_cs = NULL,
- .emit_ib = gfx_v8_0_ring_emit_ib,
+ .emit_ib = gfx_v8_0_ring_emit_ib_compute,
.emit_fence = gfx_v8_0_ring_emit_fence_compute,
.emit_semaphore = gfx_v8_0_ring_emit_semaphore,
.emit_vm_flush = gfx_v8_0_ring_emit_vm_flush,
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
index d62c4002e39c..d1064ca3670e 100644
--- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
@@ -35,6 +35,8 @@
#include "oss/oss_2_0_d.h"
#include "oss/oss_2_0_sh_mask.h"
#include "gca/gfx_8_0_d.h"
+#include "smu/smu_7_1_2_d.h"
+#include "smu/smu_7_1_2_sh_mask.h"
#define GRBM_GFX_INDEX__VCE_INSTANCE__SHIFT 0x04
#define GRBM_GFX_INDEX__VCE_INSTANCE_MASK 0x10
@@ -112,6 +114,10 @@ static int vce_v3_0_start(struct amdgpu_device *adev)
mutex_lock(&adev->grbm_idx_mutex);
for (idx = 0; idx < 2; ++idx) {
+
+ if (adev->vce.harvest_config & (1 << idx))
+ continue;
+
if(idx == 0)
WREG32_P(mmGRBM_GFX_INDEX, 0,
~GRBM_GFX_INDEX__VCE_INSTANCE_MASK);
@@ -190,10 +196,52 @@ static int vce_v3_0_start(struct amdgpu_device *adev)
return 0;
}
+#define ixVCE_HARVEST_FUSE_MACRO__ADDRESS 0xC0014074
+#define VCE_HARVEST_FUSE_MACRO__SHIFT 27
+#define VCE_HARVEST_FUSE_MACRO__MASK 0x18000000
+
+static unsigned vce_v3_0_get_harvest_config(struct amdgpu_device *adev)
+{
+ u32 tmp;
+ unsigned ret;
+
+ if (adev->flags & AMDGPU_IS_APU)
+ tmp = (RREG32_SMC(ixVCE_HARVEST_FUSE_MACRO__ADDRESS) &
+ VCE_HARVEST_FUSE_MACRO__MASK) >>
+ VCE_HARVEST_FUSE_MACRO__SHIFT;
+ else
+ tmp = (RREG32_SMC(ixCC_HARVEST_FUSES) &
+ CC_HARVEST_FUSES__VCE_DISABLE_MASK) >>
+ CC_HARVEST_FUSES__VCE_DISABLE__SHIFT;
+
+ switch (tmp) {
+ case 1:
+ ret = AMDGPU_VCE_HARVEST_VCE0;
+ break;
+ case 2:
+ ret = AMDGPU_VCE_HARVEST_VCE1;
+ break;
+ case 3:
+ ret = AMDGPU_VCE_HARVEST_VCE0 | AMDGPU_VCE_HARVEST_VCE1;
+ break;
+ default:
+ ret = 0;
+ }
+
+ return ret;
+}
+
static int vce_v3_0_early_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ adev->vce.harvest_config = vce_v3_0_get_harvest_config(adev);
+
+ if ((adev->vce.harvest_config &
+ (AMDGPU_VCE_HARVEST_VCE0 | AMDGPU_VCE_HARVEST_VCE1)) ==
+ (AMDGPU_VCE_HARVEST_VCE0 | AMDGPU_VCE_HARVEST_VCE1))
+ return -ENOENT;
+
vce_v3_0_set_ring_funcs(adev);
vce_v3_0_set_irq_funcs(adev);
generated by cgit (git 2.34.1) at 2024-06-09 10:35:46 +0000