[Skiboot] [PATCH v4] occ/prd/opal-prd: Queue OCC_RESET event message to host in OpenPOWER

Shilpasri G Bhat shilpa.bhat at linux.vnet.ibm.com
Wed Jun 29 15:04:22 AEST 2016


During an OCC reset cycle the system is forced to Psafe pstate.
When OCC becomes active, the system has to be restored to its
last pstate as requested by host. So host needs to be notified
of OCC_RESET event or else system will continue to remian in
Psafe state until host requests a new pstate after the OCC
reset cycle.

This patch defines 'OPAL_PRD_MSG_TYPE_OCC_RESET_NOTIFY' to
notify OPAL when opal-prd issues OCC reset. OPAL will queue
OCC_RESET message to host when it receives opal_prd_msg of
type '*_OCC_RESET_NOTIFY'.

Signed-off-by: Shilpasri G Bhat <shilpa.bhat at linux.vnet.ibm.com>
---
- No changes from v3, rebased on master.

Changes from v2:
- Fixed the wrong message argument passed to write() in
  handle_prd_control() function in opal-prd.c
  - rc = write(ctx->fd, &msg, sizeof(omsg));
  + rc = write(ctx->fd, &omsg, sizeof(omsg));

Changes from v1:
- Modified commit log
- s/OPAL_PRD_MSG_TYPE_OCC_RESET_NOTIF/OPAL_PRD_MSG_TYPE_OCC_RESET_NOTIFY

 external/opal-prd/opal-prd.c |  6 +++++
 hw/occ.c                     | 56 ++++++++++++++++++++++++++------------------
 hw/prd.c                     |  4 ++++
 include/opal-api.h           |  1 +
 include/opal-internal.h      |  1 +
 5 files changed, 45 insertions(+), 23 deletions(-)

diff --git a/external/opal-prd/opal-prd.c b/external/opal-prd/opal-prd.c
index 18c6e49..67e0c62 100644
--- a/external/opal-prd/opal-prd.c
+++ b/external/opal-prd/opal-prd.c
@@ -1371,6 +1371,7 @@ static void handle_prd_control_run_cmd(struct control_msg *send_msg,
 static void handle_prd_control(struct opal_prd_ctx *ctx, int fd)
 {
 	struct control_msg msg, *recv_msg, *send_msg;
+	struct opal_prd_msg omsg;
 	bool enabled = false;
 	int rc, size;
 
@@ -1423,6 +1424,11 @@ static void handle_prd_control(struct opal_prd_ctx *ctx, int fd)
 		handle_prd_control_occ_actuation(send_msg, enabled);
 		break;
 	case CONTROL_MSG_TEMP_OCC_RESET:
+		omsg.hdr.type = OPAL_PRD_MSG_TYPE_OCC_RESET_NOTIFY;
+		omsg.hdr.size = htobe16(sizeof(omsg));
+		rc = write(ctx->fd, &omsg, sizeof(omsg));
+		if (rc != sizeof(omsg))
+			pr_log(LOG_WARNING, "FW: Failed to send OCC_RESET message: %m");
 		handle_prd_control_occ_reset(send_msg);
 		break;
 	case CONTROL_MSG_TEMP_OCC_ERROR:
diff --git a/hw/occ.c b/hw/occ.c
index 25e0e41..456c573 100644
--- a/hw/occ.c
+++ b/hw/occ.c
@@ -739,6 +739,38 @@ static void occ_do_load(u8 scope, u32 dbob_id __unused, u32 seq_id)
 	__occ_do_load(scope, dbob_id, seq_id);
 }
 
+int occ_msg_queue_occ_reset(void)
+{
+	struct opal_occ_msg occ_msg = { OCC_RESET, 0, 0 };
+	struct proc_chip *chip;
+	int rc;
+
+	lock(&occ_lock);
+	rc = _opal_queue_msg(OPAL_MSG_OCC, NULL, NULL, 3,
+			     (uint64_t *)&occ_msg);
+	if (rc) {
+		prlog(PR_INFO, "OCC: Failed to queue OCC_RESET message\n");
+		goto out;
+	}
+	/*
+	 * Set 'valid' byte of chip_occ_data to 0 since OCC
+	 * may not clear this byte on a reset.
+	 * OCC will set the 'valid' byte to 1 when it becomes
+	 * active again.
+	 */
+	for_each_chip(chip) {
+		struct occ_pstate_table *occ_data;
+
+		occ_data = chip_occ_data(chip);
+		occ_data->valid = 0;
+		chip->throttle = 0;
+	}
+	occ_reset = true;
+out:
+	unlock(&occ_lock);
+	return rc;
+}
+
 static void occ_do_reset(u8 scope, u32 dbob_id, u32 seq_id)
 {
 	struct fsp_msg *rsp, *stat;
@@ -787,8 +819,6 @@ static void occ_do_reset(u8 scope, u32 dbob_id, u32 seq_id)
 		rc = 0;
 	}
 	if (!rc) {
-		struct opal_occ_msg occ_msg = { OCC_RESET, 0, 0 };
-
 		/* Send a single success response for all chips */
 		stat = fsp_mkmsg(FSP_CMD_RESET_OCC_STAT, 2, 0, seq_id);
 		if (stat)
@@ -799,27 +829,7 @@ static void occ_do_reset(u8 scope, u32 dbob_id, u32 seq_id)
 				"OCC: Error %d queueing FSP OCC RESET"
 					" STATUS message\n", rc);
 		}
-		lock(&occ_lock);
-		rc = _opal_queue_msg(OPAL_MSG_OCC, NULL, NULL, 3,
-				     (uint64_t *)&occ_msg);
-		if (rc)
-			prlog(PR_INFO, "OCC: Failed to queue message %d\n",
-			      OCC_RESET);
-		/*
-		 * Set 'valid' byte of chip_occ_data to 0 since OCC
-		 * may not clear this byte on a reset.
-		 * OCC will set the 'valid' byte to 1 when it becomes
-		 * active again.
-		 */
-		for_each_chip(chip) {
-			struct occ_pstate_table *occ_data;
-
-			occ_data = chip_occ_data(chip);
-			occ_data->valid = 0;
-			chip->throttle = 0;
-		}
-		occ_reset = true;
-		unlock(&occ_lock);
+		occ_msg_queue_occ_reset();
 	} else {
 
 		/*
diff --git a/hw/prd.c b/hw/prd.c
index 27b8ad8..e520e13 100644
--- a/hw/prd.c
+++ b/hw/prd.c
@@ -182,6 +182,7 @@ static void send_next_pending_event(void)
 	} else if (event & EVENT_OCC_RESET) {
 		prd_msg.hdr.type = OPAL_PRD_MSG_TYPE_OCC_RESET;
 		prd_msg.occ_reset.chip = proc;
+		occ_msg_queue_occ_reset();
 	}
 
 	queue_prd_msg(&prd_msg, prd_msg_consumed);
@@ -339,6 +340,9 @@ static int64_t opal_prd_msg(struct opal_prd_msg *msg)
 	case OPAL_PRD_MSG_TYPE_ATTN_ACK:
 		rc = prd_msg_handle_attn_ack(msg);
 		break;
+	case OPAL_PRD_MSG_TYPE_OCC_RESET_NOTIFY:
+		rc = occ_msg_queue_occ_reset();
+		break;
 	default:
 		rc = OPAL_UNSUPPORTED;
 	}
diff --git a/include/opal-api.h b/include/opal-api.h
index 917d751..032086c 100644
--- a/include/opal-api.h
+++ b/include/opal-api.h
@@ -900,6 +900,7 @@ enum opal_prd_msg_type {
 	OPAL_PRD_MSG_TYPE_ATTN_ACK,	/* HBRT --> OPAL */
 	OPAL_PRD_MSG_TYPE_OCC_ERROR,	/* HBRT <-- OPAL */
 	OPAL_PRD_MSG_TYPE_OCC_RESET,	/* HBRT <-- OPAL */
+	OPAL_PRD_MSG_TYPE_OCC_RESET_NOTIFY, /* HBRT --> OPAL */
 };
 
 struct opal_prd_msg_header {
diff --git a/include/opal-internal.h b/include/opal-internal.h
index 5e41e10..6f866a8 100644
--- a/include/opal-internal.h
+++ b/include/opal-internal.h
@@ -81,5 +81,6 @@ extern void opal_del_host_sync_notifier(bool (*notify)(void *data));
  */
 struct OpalHMIEvent;
 extern int handle_hmi_exception(__be64 hmer, struct OpalHMIEvent *hmi_evt);
+extern int occ_msg_queue_occ_reset(void);
 
 #endif /* __OPAL_INTERNAL_H */
-- 
1.9.3



More information about the Skiboot mailing list