gpu: nvgpu: Add thermal alert event handling

* Added the thermal alert event handling
* Added the thermal alert event PMU RPC

JIRA DNVGPU-130
Bug 200231080

Change-Id: I54ab0dc35a940e70733238fd95669db28389619a
Signed-off-by: Lakshmanan M <lm@nvidia.com>
Reviewed-on: http://git-master/r/1248968
(cherry picked from commit a20593c89a374d64201e4720552160a65533ecc3)
Reviewed-on: http://git-master/r/1282076
GVS: Gerrit_Virtual_Submit
Reviewed-by: Vijayakumar Subbu <vsubbu@nvidia.com>

2 files changed, 125 insertions, 14 deletions

diff --git a/drivers/gpu/nvgpu/therm/thrmpmu.c b/drivers/gpu/nvgpu/therm/thrmpmu.c
index 0d0a4b3a..0f780730 100644
--- a/drivers/gpu/nvgpu/therm/thrmpmu.c
+++ b/drivers/gpu/nvgpu/therm/thrmpmu.c
@@ -77,9 +77,47 @@ exit:
         return status;
 }
 
-u32 therm_set_warn_temp_limit(struct gk20a *g)
+static u32 therm_pmu_cmd_post(struct gk20a *g, struct pmu_cmd *cmd,
+                struct pmu_msg *msg, struct pmu_payload *payload,
+                u32 queue_id, pmu_callback callback, void* cb_param,
+                u32 *seq_desc, unsigned long timeout)
 {
         u32 status;
+        struct therm_pmucmdhandler_params *handlerparams = NULL;
+
+        status = gk20a_pmu_cmd_post(g, cmd, msg, payload,
+                                queue_id,
+                                callback,
+                                cb_param,
+                                seq_desc,
+                                timeout);
+        if (status) {
+                gk20a_err(dev_from_gk20a(g),
+                        "unable to post therm cmd for unit %x cmd id %x size %x",
+                        cmd->hdr.unit_id, cmd->cmd.therm.cmd_type, cmd->hdr.size);
+                goto exit;
+        }
+
+        if (cb_param) {
+                handlerparams = (struct therm_pmucmdhandler_params*)cb_param;
+
+                pmu_wait_message_cond(&g->pmu,
+                                gk20a_get_gr_idle_timeout(g),
+                                &handlerparams->success, 1);
+
+                if (handlerparams->success == 0) {
+                        gk20a_err(dev_from_gk20a(g), "could not process cmd\n");
+                        status = -ETIMEDOUT;
+                        goto exit;
+                }
+        }
+
+exit:
+        return status;
+}
+
+static u32 therm_set_warn_temp_limit(struct gk20a *g)
+{
         u32 seqdesc = 0;
         struct pmu_cmd cmd = { {0} };
         struct pmu_msg msg = { {0} };
@@ -94,7 +132,69 @@ u32 therm_set_warn_temp_limit(struct gk20a *g)
         rpccall.b_supported = 0;
 
         cmd.hdr.unit_id = PMU_UNIT_THERM;
-        cmd.hdr.size = ((u32)sizeof(struct nv_pmu_therm_cmd) +
+        cmd.hdr.size = ((u32)sizeof(struct nv_pmu_therm_cmd_rpc) +
+                        (u32)sizeof(struct pmu_hdr));
+        cmd.cmd.therm.cmd_type = NV_PMU_THERM_CMD_ID_RPC;
+
+        msg.hdr.size = sizeof(struct pmu_msg);
+
+        payload.in.buf = (u8 *)&rpccall;
+        payload.in.size = (u32)sizeof(struct nv_pmu_therm_rpc);
+        payload.in.fb_size = PMU_CMD_SUBMIT_PAYLOAD_PARAMS_FB_SIZE_UNUSED;
+        payload.in.offset = NV_PMU_THERM_CMD_RPC_ALLOC_OFFSET;
+
+        payload.out.buf = (u8 *)&rpccall;
+        payload.out.size = (u32)sizeof(struct nv_pmu_therm_rpc);
+        payload.out.fb_size = PMU_CMD_SUBMIT_PAYLOAD_PARAMS_FB_SIZE_UNUSED;
+        payload.out.offset = NV_PMU_CLK_MSG_RPC_ALLOC_OFFSET;
+
+        /* Setup the handler params to communicate back results.*/
+        handlerparams.success = 0;
+        handlerparams.prpccall = &rpccall;
+
+        return therm_pmu_cmd_post(g, &cmd, NULL, &payload,
+                                PMU_COMMAND_QUEUE_LPQ,
+                                therm_pmucmdhandler,
+                                (void *)&handlerparams,
+                                &seqdesc, ~0);
+}
+
+static u32 therm_enable_slct_notification_request(struct gk20a *g)
+{
+        u32 seqdesc = 0;
+        struct pmu_cmd cmd = { {0} };
+
+        cmd.hdr.unit_id = PMU_UNIT_THERM;
+        cmd.hdr.size = ((u32)sizeof(struct nv_pmu_therm_cmd_hw_slowdown_notification) +
+                (u32)sizeof(struct pmu_hdr));
+
+        cmd.cmd.therm.cmd_type = NV_PMU_THERM_CMD_ID_HW_SLOWDOWN_NOTIFICATION;
+        cmd.cmd.therm.hw_slct_notification.request =
+                NV_RM_PMU_THERM_HW_SLOWDOWN_NOTIFICATION_REQUEST_ENABLE;
+
+        return therm_pmu_cmd_post(g, &cmd, NULL, NULL,
+                                PMU_COMMAND_QUEUE_LPQ,
+                                NULL,
+                                NULL,
+                                &seqdesc, ~0);
+}
+
+static u32 therm_send_slct_configuration_to_pmu(struct gk20a *g)
+{
+        u32 seqdesc = 0;
+        struct pmu_cmd cmd = { {0} };
+        struct pmu_msg msg = { {0} };
+        struct pmu_payload payload = { {0} };
+        struct nv_pmu_therm_rpc rpccall = {0};
+        struct therm_pmucmdhandler_params handlerparams = {0};
+
+        rpccall.function = NV_PMU_THERM_RPC_ID_SLCT;
+        rpccall.params.slct.mask_enabled =
+                (1 << NV_PMU_THERM_EVENT_THERMAL_1);
+        rpccall.b_supported = 0;
+
+        cmd.hdr.unit_id = PMU_UNIT_THERM;
+        cmd.hdr.size = ((u32)sizeof(struct nv_pmu_therm_cmd_rpc) +
                         (u32)sizeof(struct pmu_hdr));
         cmd.cmd.therm.cmd_type = NV_PMU_THERM_CMD_ID_RPC;
 
@@ -114,29 +214,40 @@ u32 therm_set_warn_temp_limit(struct gk20a *g)
         handlerparams.success = 0;
         handlerparams.prpccall = &rpccall;
 
-        status = gk20a_pmu_cmd_post(g, &cmd, NULL, &payload,
+        return therm_pmu_cmd_post(g, &cmd, NULL, &payload,
                                 PMU_COMMAND_QUEUE_LPQ,
                                 therm_pmucmdhandler,
                                 (void *)&handlerparams,
                                 &seqdesc, ~0);
+}
+
+u32 therm_configure_therm_alert(struct gk20a *g)
+{
+        u32 status;
+
+        status = therm_enable_slct_notification_request(g);
         if (status) {
                 gk20a_err(dev_from_gk20a(g),
-                        "unable to post pmgr cmd for unit %x cmd id %x size %x",
-                        cmd.hdr.unit_id, cmd.cmd.therm.cmd_type, cmd.hdr.size);
+                        "therm_enable_slct_notification_request-failed %d",
+                        status);
                 goto exit;
         }
 
-        pmu_wait_message_cond(&g->pmu,
-                        gk20a_get_gr_idle_timeout(g),
-                        &handlerparams.success, 1);
-
-        if (handlerparams.success == 0) {
-                gk20a_err(dev_from_gk20a(g), "could not process cmd\n");
-                status = -ETIMEDOUT;
+        status = therm_send_slct_configuration_to_pmu(g);
+        if (status) {
+                gk20a_err(dev_from_gk20a(g),
+                        "therm_send_slct_configuration_to_pmu-failed %d",
+                        status);
                 goto exit;
         }
 
+        status = therm_set_warn_temp_limit(g);
+        if (status) {
+                gk20a_err(dev_from_gk20a(g),
+                        "therm_set_warn_temp_limit-failed %d",
+                        status);
+                goto exit;
+        }
 exit:
         return status;
 }
-
diff --git a/drivers/gpu/nvgpu/therm/thrmpmu.h b/drivers/gpu/nvgpu/therm/thrmpmu.h
index e6f70411..8f0d7501 100644
--- a/drivers/gpu/nvgpu/therm/thrmpmu.h
+++ b/drivers/gpu/nvgpu/therm/thrmpmu.h
@@ -17,6 +17,6 @@
 
 u32 therm_send_pmgr_tables_to_pmu(struct gk20a *g);
 
-u32 therm_set_warn_temp_limit(struct gk20a *g);
+u32 therm_configure_therm_alert(struct gk20a *g);
 
 #endif