gpu: nvgpu: get voltage, current, power and temperature

Add ioctls to retrieve voltage, current, power and temperature.
Add flags in GPU characteristics to indicate if feature is supported.

Jira DNVGPU-166

Change-Id: Idd5a767326c9d43630e8289ca7d2c27bb96a9f14
Signed-off-by: David Nieto <dmartineznie@nvidia.com>
Reviewed-on: http://git-master/r/1241862
Tested-by: Thomas Fleury <tfleury@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-by: Vijayakumar Subbu <vsubbu@nvidia.com>
Reviewed-on: http://git-master/r/1267153

3 files changed, 167 insertions, 1 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c b/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
index 3de84851..23fbdce0 100644
--- a/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/ctrl_gk20a.c
@@ -1063,7 +1063,6 @@ static int nvgpu_gpu_clk_set_info(struct gk20a *g,
         return ret;
 }
 
-
 static int nvgpu_gpu_clk_get_info(struct gk20a *g,
                 struct gk20a_ctrl_priv *priv,
                 struct nvgpu_gpu_clk_get_info_args *args)
@@ -1172,6 +1171,119 @@ static int nvgpu_gpu_clk_get_event_fd(struct gk20a *g,
 
         return nvgpu_clk_arb_install_event_fd(g, session, &args->event_fd);
 }
+
+static int nvgpu_gpu_get_voltage(struct gk20a *g,
+                struct nvgpu_gpu_get_voltage_args *args)
+{
+        int err = -EINVAL;
+
+        gk20a_dbg_fn("");
+
+        if (args->reserved)
+                return -EINVAL;
+
+        if (!(g->gpu_characteristics.flags & NVGPU_GPU_FLAGS_SUPPORT_GET_VOLTAGE))
+                return -EINVAL;
+
+        err = gk20a_busy(g->dev);
+        if (err)
+            return err;
+
+        switch (args->which) {
+        case NVGPU_GPU_VOLTAGE_CORE:
+                err = volt_get_voltage(g, CTRL_VOLT_DOMAIN_LOGIC, &args->voltage);
+                break;
+        case NVGPU_GPU_VOLTAGE_SRAM:
+                err = volt_get_voltage(g, CTRL_VOLT_DOMAIN_SRAM, &args->voltage);
+                break;
+        case NVGPU_GPU_VOLTAGE_BUS:
+                err = pmgr_pwr_devices_get_voltage(g, &args->voltage);
+                break;
+        default:
+                err = -EINVAL;
+        }
+
+        gk20a_idle(g->dev);
+
+        return err;
+}
+
+static int nvgpu_gpu_get_current(struct gk20a *g,
+                struct nvgpu_gpu_get_current_args *args)
+{
+        int err;
+
+        gk20a_dbg_fn("");
+
+        if (args->reserved[0] || args->reserved[1] || args->reserved[2])
+                return -EINVAL;
+
+        if (!(g->gpu_characteristics.flags & NVGPU_GPU_FLAGS_SUPPORT_GET_CURRENT))
+                return -EINVAL;
+
+        err = gk20a_busy(g->dev);
+        if (err)
+                return err;
+
+        err = pmgr_pwr_devices_get_current(g, &args->currnt);
+
+        gk20a_idle(g->dev);
+
+        return err;
+}
+
+static int nvgpu_gpu_get_power(struct gk20a *g,
+                struct nvgpu_gpu_get_power_args *args)
+{
+        int err;
+
+        gk20a_dbg_fn("");
+
+        if (args->reserved[0] || args->reserved[1] || args->reserved[2])
+                return -EINVAL;
+
+        if (!(g->gpu_characteristics.flags & NVGPU_GPU_FLAGS_SUPPORT_GET_POWER))
+                return -EINVAL;
+
+        err = gk20a_busy(g->dev);
+        if (err)
+                return err;
+
+        err = pmgr_pwr_devices_get_power(g, &args->power);
+
+        gk20a_idle(g->dev);
+
+        return err;
+}
+
+static int nvgpu_gpu_get_temperature(struct gk20a *g,
+                struct nvgpu_gpu_get_temperature_args *args)
+{
+        int err;
+        u32 temp_f24_8;
+
+        gk20a_dbg_fn("");
+
+        if (args->reserved[0] || args->reserved[1] || args->reserved[2])
+                return -EINVAL;
+
+        if (!g->ops.therm.get_internal_sensor_curr_temp)
+                return -EINVAL;
+
+        err = gk20a_busy(g->dev);
+        if (err)
+                return err;
+
+        err = g->ops.therm.get_internal_sensor_curr_temp(g, &temp_f24_8);
+
+        gk20a_idle(g->dev);
+
+        /* Convert from standard S24.8 fixed point to mC */
+        if (!err)
+                args->temperature = (temp_f24_8 * 1000) / 256;
+
+        return err;
+}
 #endif
 
 long gk20a_ctrl_dev_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
@@ -1461,6 +1573,26 @@ long gk20a_ctrl_dev_ioctl(struct file *filp, unsigned int cmd, unsigned long arg
                 err = nvgpu_gpu_clk_get_event_fd(g, priv,
                         (struct nvgpu_gpu_clk_get_event_fd_args *)buf);
                 break;
+
+        case NVGPU_GPU_IOCTL_GET_VOLTAGE:
+                err = nvgpu_gpu_get_voltage(g,
+                        (struct nvgpu_gpu_get_voltage_args *)buf);
+                break;
+
+        case NVGPU_GPU_IOCTL_GET_CURRENT:
+                err = nvgpu_gpu_get_current(g,
+                        (struct nvgpu_gpu_get_current_args *)buf);
+                break;
+
+        case NVGPU_GPU_IOCTL_GET_POWER:
+                err = nvgpu_gpu_get_power(g,
+                        (struct nvgpu_gpu_get_power_args *)buf);
+                break;
+
+        case NVGPU_GPU_IOCTL_GET_TEMPERATURE:
+                err = nvgpu_gpu_get_temperature(g,
+                        (struct nvgpu_gpu_get_temperature_args *)buf);
+                break;
 #endif
 
         default:
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.h b/drivers/gpu/nvgpu/gk20a/gk20a.h
index 564026a4..071111ab 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.h
@@ -588,6 +588,7 @@ struct gpu_ops {
                 int (*init_therm_setup_hw)(struct gk20a *g);
                 int (*elcg_init_idle_filters)(struct gk20a *g);
                 void (*therm_debugfs_init)(struct gk20a *g);
+                int (*get_internal_sensor_curr_temp)(struct gk20a *g, u32 *temp_f24_8);
         } therm;
         struct {
                 bool (*is_pmu_supported)(struct gk20a *g);
diff --git a/include/uapi/linux/nvgpu.h b/include/uapi/linux/nvgpu.h
index 13492e0d..f45be911 100644
--- a/include/uapi/linux/nvgpu.h
+++ b/include/uapi/linux/nvgpu.h
@@ -721,6 +721,31 @@ struct nvgpu_gpu_get_fbp_l2_masks_args {
         __u64 mask_buf_addr;
 };
 
+#define NVGPU_GPU_VOLTAGE_CORE          1
+#define NVGPU_GPU_VOLTAGE_SRAM          2
+#define NVGPU_GPU_VOLTAGE_BUS           3       /* input to regulator */
+
+struct nvgpu_gpu_get_voltage_args {
+        __u64 reserved;
+        __u32 which;            /* in: NVGPU_GPU_VOLTAGE_* */
+        __u32 voltage;          /* uV */
+};
+
+struct nvgpu_gpu_get_current_args {
+        __u32 reserved[3];
+        __u32 currnt;           /* mA */
+};
+
+struct nvgpu_gpu_get_power_args {
+        __u32 reserved[3];
+        __u32 power;            /* mW */
+};
+
+struct nvgpu_gpu_get_temperature_args {
+        __u32 reserved[3];
+        __u32 temperature;      /* mC */
+};
+
 #define NVGPU_GPU_IOCTL_ZCULL_GET_CTX_SIZE \
         _IOR(NVGPU_GPU_IOCTL_MAGIC, 1, struct nvgpu_gpu_zcull_get_ctx_size_args)
 #define NVGPU_GPU_IOCTL_ZCULL_GET_INFO \
@@ -792,6 +817,14 @@ struct nvgpu_gpu_get_fbp_l2_masks_args {
 #define NVGPU_GPU_IOCTL_GET_MEMORY_STATE \
         _IOWR(NVGPU_GPU_IOCTL_MAGIC, 33, \
                         struct nvgpu_gpu_get_memory_state_args)
+#define NVGPU_GPU_IOCTL_GET_VOLTAGE \
+        _IOWR(NVGPU_GPU_IOCTL_MAGIC, 33, struct nvgpu_gpu_get_voltage_args)
+#define NVGPU_GPU_IOCTL_GET_CURRENT \
+        _IOWR(NVGPU_GPU_IOCTL_MAGIC, 34, struct nvgpu_gpu_get_current_args)
+#define NVGPU_GPU_IOCTL_GET_POWER \
+        _IOWR(NVGPU_GPU_IOCTL_MAGIC, 35, struct nvgpu_gpu_get_power_args)
+#define NVGPU_GPU_IOCTL_GET_TEMPERATURE \
+        _IOWR(NVGPU_GPU_IOCTL_MAGIC, 36, struct nvgpu_gpu_get_temperature_args)
 #define NVGPU_GPU_IOCTL_GET_FBP_L2_MASKS \
         _IOWR(NVGPU_GPU_IOCTL_MAGIC, 38, struct nvgpu_gpu_get_fbp_l2_masks_args)
 #define NVGPU_GPU_IOCTL_LAST            \