drm/amdgpu: support dpm level modification under virtualization v3

Under vega10 virtualuzation, smu ip block will not be added.
Therefore, we need add pp clk query and force dpm level function
at amdgpu_virt_ops to support the feature.

v2: add get_pp_clk existence check and use kzalloc to allocate buf

v3: return -ENOMEM for allocation failure and correct the coding style

Signed-off-by: Yintian Tao <yttao@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Yintian Tao 2019-04-09 20:33:20 +08:00 коммит произвёл Alex Deucher
Родитель b0960c3592
Коммит bb5a2bdf36
7 изменённых файлов: 165 добавлений и 0 удалений

Просмотреть файл

@ -2471,6 +2471,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
mutex_init(&adev->virt.vf_errors.lock); mutex_init(&adev->virt.vf_errors.lock);
hash_init(adev->mn_hash); hash_init(adev->mn_hash);
mutex_init(&adev->lock_reset); mutex_init(&adev->lock_reset);
mutex_init(&adev->virt.dpm_mutex);
amdgpu_device_check_arguments(adev); amdgpu_device_check_arguments(adev);

Просмотреть файл

@ -696,6 +696,10 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
if (adev->pm.dpm_enabled) { if (adev->pm.dpm_enabled) {
dev_info.max_engine_clock = amdgpu_dpm_get_sclk(adev, false) * 10; dev_info.max_engine_clock = amdgpu_dpm_get_sclk(adev, false) * 10;
dev_info.max_memory_clock = amdgpu_dpm_get_mclk(adev, false) * 10; dev_info.max_memory_clock = amdgpu_dpm_get_mclk(adev, false) * 10;
} else if (amdgpu_sriov_vf(adev) && amdgim_is_hwperf(adev) &&
adev->virt.ops->get_pp_clk) {
dev_info.max_engine_clock = amdgpu_virt_get_sclk(adev, false) * 10;
dev_info.max_memory_clock = amdgpu_virt_get_mclk(adev, false) * 10;
} else { } else {
dev_info.max_engine_clock = adev->clock.default_sclk * 10; dev_info.max_engine_clock = adev->clock.default_sclk * 10;
dev_info.max_memory_clock = adev->clock.default_mclk * 10; dev_info.max_memory_clock = adev->clock.default_mclk * 10;

Просмотреть файл

@ -327,6 +327,18 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
goto fail; goto fail;
} }
if (amdgpu_sriov_vf(adev)) {
if (amdgim_is_hwperf(adev) &&
adev->virt.ops->force_dpm_level) {
mutex_lock(&adev->pm.mutex);
adev->virt.ops->force_dpm_level(adev, level);
mutex_unlock(&adev->pm.mutex);
return count;
} else {
return -EINVAL;
}
}
if (current_level == level) if (current_level == level)
return count; return count;
@ -790,6 +802,10 @@ static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev,
struct drm_device *ddev = dev_get_drvdata(dev); struct drm_device *ddev = dev_get_drvdata(dev);
struct amdgpu_device *adev = ddev->dev_private; struct amdgpu_device *adev = ddev->dev_private;
if (amdgpu_sriov_vf(adev) && amdgim_is_hwperf(adev) &&
adev->virt.ops->get_pp_clk)
return adev->virt.ops->get_pp_clk(adev, PP_SCLK, buf);
if (is_support_sw_smu(adev)) if (is_support_sw_smu(adev))
return smu_print_clk_levels(&adev->smu, PP_SCLK, buf); return smu_print_clk_levels(&adev->smu, PP_SCLK, buf);
else if (adev->powerplay.pp_funcs->print_clock_levels) else if (adev->powerplay.pp_funcs->print_clock_levels)

Просмотреть файл

@ -375,4 +375,53 @@ void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
} }
} }
static uint32_t parse_clk(char *buf, bool min)
{
char *ptr = buf;
uint32_t clk = 0;
do {
ptr = strchr(ptr, ':');
if (!ptr)
break;
ptr+=2;
clk = simple_strtoul(ptr, NULL, 10);
} while (!min);
return clk * 100;
}
uint32_t amdgpu_virt_get_sclk(struct amdgpu_device *adev, bool lowest)
{
char *buf = NULL;
uint32_t clk = 0;
buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
if (!buf)
return -ENOMEM;
adev->virt.ops->get_pp_clk(adev, PP_SCLK, buf);
clk = parse_clk(buf, lowest);
kfree(buf);
return clk;
}
uint32_t amdgpu_virt_get_mclk(struct amdgpu_device *adev, bool lowest)
{
char *buf = NULL;
uint32_t clk = 0;
buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
if (!buf)
return -ENOMEM;
adev->virt.ops->get_pp_clk(adev, PP_MCLK, buf);
clk = parse_clk(buf, lowest);
kfree(buf);
return clk;
}

Просмотреть файл

@ -57,6 +57,8 @@ struct amdgpu_virt_ops {
int (*reset_gpu)(struct amdgpu_device *adev); int (*reset_gpu)(struct amdgpu_device *adev);
int (*wait_reset)(struct amdgpu_device *adev); int (*wait_reset)(struct amdgpu_device *adev);
void (*trans_msg)(struct amdgpu_device *adev, u32 req, u32 data1, u32 data2, u32 data3); void (*trans_msg)(struct amdgpu_device *adev, u32 req, u32 data1, u32 data2, u32 data3);
int (*get_pp_clk)(struct amdgpu_device *adev, u32 type, char *buf);
int (*force_dpm_level)(struct amdgpu_device *adev, u32 level);
}; };
/* /*
@ -83,6 +85,8 @@ enum AMDGIM_FEATURE_FLAG {
AMDGIM_FEATURE_GIM_LOAD_UCODES = 0x2, AMDGIM_FEATURE_GIM_LOAD_UCODES = 0x2,
/* VRAM LOST by GIM */ /* VRAM LOST by GIM */
AMDGIM_FEATURE_GIM_FLR_VRAMLOST = 0x4, AMDGIM_FEATURE_GIM_FLR_VRAMLOST = 0x4,
/* HW PERF SIM in GIM */
AMDGIM_FEATURE_HW_PERF_SIMULATION = (1 << 3),
}; };
struct amd_sriov_msg_pf2vf_info_header { struct amd_sriov_msg_pf2vf_info_header {
@ -252,6 +256,8 @@ struct amdgpu_virt {
struct amdgpu_vf_error_buffer vf_errors; struct amdgpu_vf_error_buffer vf_errors;
struct amdgpu_virt_fw_reserve fw_reserve; struct amdgpu_virt_fw_reserve fw_reserve;
uint32_t gim_feature; uint32_t gim_feature;
/* protect DPM events to GIM */
struct mutex dpm_mutex;
}; };
#define amdgpu_sriov_enabled(adev) \ #define amdgpu_sriov_enabled(adev) \
@ -278,6 +284,9 @@ static inline bool is_virtual_machine(void)
#endif #endif
} }
#define amdgim_is_hwperf(adev) \
((adev)->virt.gim_feature & AMDGIM_FEATURE_HW_PERF_SIMULATION)
bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev); bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev);
void amdgpu_virt_init_setting(struct amdgpu_device *adev); void amdgpu_virt_init_setting(struct amdgpu_device *adev);
uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg); uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg);
@ -295,5 +304,7 @@ int amdgpu_virt_fw_reserve_get_checksum(void *obj, unsigned long obj_size,
unsigned int key, unsigned int key,
unsigned int chksum); unsigned int chksum);
void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev); void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev);
uint32_t amdgpu_virt_get_sclk(struct amdgpu_device *adev, bool lowest);
uint32_t amdgpu_virt_get_mclk(struct amdgpu_device *adev, bool lowest);
#endif #endif

Просмотреть файл

@ -157,6 +157,82 @@ static void xgpu_ai_mailbox_trans_msg (struct amdgpu_device *adev,
xgpu_ai_mailbox_set_valid(adev, false); xgpu_ai_mailbox_set_valid(adev, false);
} }
static int xgpu_ai_get_pp_clk(struct amdgpu_device *adev, u32 type, char *buf)
{
int r = 0;
u32 req, val, size;
if (!amdgim_is_hwperf(adev) || buf == NULL)
return -EBADRQC;
switch(type) {
case PP_SCLK:
req = IDH_IRQ_GET_PP_SCLK;
break;
case PP_MCLK:
req = IDH_IRQ_GET_PP_MCLK;
break;
default:
return -EBADRQC;
}
mutex_lock(&adev->virt.dpm_mutex);
xgpu_ai_mailbox_trans_msg(adev, req, 0, 0, 0);
r = xgpu_ai_poll_msg(adev, IDH_SUCCESS);
if (!r && adev->fw_vram_usage.va != NULL) {
val = RREG32_NO_KIQ(
SOC15_REG_OFFSET(NBIO, 0,
mmBIF_BX_PF0_MAILBOX_MSGBUF_RCV_DW1));
size = strnlen((((char *)adev->virt.fw_reserve.p_pf2vf) +
val), PAGE_SIZE);
if (size < PAGE_SIZE)
strcpy(buf,((char *)adev->virt.fw_reserve.p_pf2vf + val));
else
size = 0;
r = size;
goto out;
}
r = xgpu_ai_poll_msg(adev, IDH_FAIL);
if(r)
pr_info("%s DPM request failed",
(type == PP_SCLK)? "SCLK" : "MCLK");
out:
mutex_unlock(&adev->virt.dpm_mutex);
return r;
}
static int xgpu_ai_force_dpm_level(struct amdgpu_device *adev, u32 level)
{
int r = 0;
u32 req = IDH_IRQ_FORCE_DPM_LEVEL;
if (!amdgim_is_hwperf(adev))
return -EBADRQC;
mutex_lock(&adev->virt.dpm_mutex);
xgpu_ai_mailbox_trans_msg(adev, req, level, 0, 0);
r = xgpu_ai_poll_msg(adev, IDH_SUCCESS);
if (!r)
goto out;
r = xgpu_ai_poll_msg(adev, IDH_FAIL);
if (!r)
pr_info("DPM request failed");
else
pr_info("Mailbox is broken");
out:
mutex_unlock(&adev->virt.dpm_mutex);
return r;
}
static int xgpu_ai_send_access_requests(struct amdgpu_device *adev, static int xgpu_ai_send_access_requests(struct amdgpu_device *adev,
enum idh_request req) enum idh_request req)
{ {
@ -375,4 +451,6 @@ const struct amdgpu_virt_ops xgpu_ai_virt_ops = {
.reset_gpu = xgpu_ai_request_reset, .reset_gpu = xgpu_ai_request_reset,
.wait_reset = NULL, .wait_reset = NULL,
.trans_msg = xgpu_ai_mailbox_trans_msg, .trans_msg = xgpu_ai_mailbox_trans_msg,
.get_pp_clk = xgpu_ai_get_pp_clk,
.force_dpm_level = xgpu_ai_force_dpm_level,
}; };

Просмотреть файл

@ -35,6 +35,10 @@ enum idh_request {
IDH_REL_GPU_FINI_ACCESS, IDH_REL_GPU_FINI_ACCESS,
IDH_REQ_GPU_RESET_ACCESS, IDH_REQ_GPU_RESET_ACCESS,
IDH_IRQ_FORCE_DPM_LEVEL = 10,
IDH_IRQ_GET_PP_SCLK,
IDH_IRQ_GET_PP_MCLK,
IDH_LOG_VF_ERROR = 200, IDH_LOG_VF_ERROR = 200,
}; };
@ -43,6 +47,8 @@ enum idh_event {
IDH_READY_TO_ACCESS_GPU, IDH_READY_TO_ACCESS_GPU,
IDH_FLR_NOTIFICATION, IDH_FLR_NOTIFICATION,
IDH_FLR_NOTIFICATION_CMPL, IDH_FLR_NOTIFICATION_CMPL,
IDH_SUCCESS,
IDH_FAIL,
IDH_EVENT_MAX IDH_EVENT_MAX
}; };