Skip to contentSkip to conversation

scsi: megaraid_sas - stop cmd processing if hw_crit_error is set

PATCH In review Sumant Patro 2 participants 2 messages Feb 7, 2007

Subsystems (2)
SCSI Megaraid SCSI/SAS drivers
lore
Share & tools
lore

Sumant Patro

author
#0 lore raw

PATCH 1/5 scsi: megaraid_sas - stop cmd processing if hw_crit_error is set

Checks if hw_crit_error is set before processing cmds. Added in megasas_queue_command and command completion routines.

Signed-off-by Sumant Patro <[email protected]>

---
drivers/scsi/megaraid/megaraid_sas.c | 13 ++++++++++++-

 1 files changed, 12 insertions(+), 1 deletion(-)
drivers/scsi/megaraid/megaraid_sas.c +12 -1
diff -uprN 2.6.orig/drivers/scsi/megaraid/megaraid_sas.c 2.6.new-p1/drivers/scsi/megaraid/megaraid_sas.c
@@ -841,6 +841,11 @@ megasas_queue_command(struct scsi_cmnd *
 
 	instance = (struct megasas_instance *)
 	    scmd->device->host->hostdata;
+
+	/* Don't process if we have already declared adapter dead */
+	if(instance->hw_crit_error)
+		return SCSI_MLQUEUE_HOST_BUSY;
+
 	scmd->scsi_done = done;
 	scmd->result = 0;
 

@@ -1282,11 +1287,13 @@ megasas_deplete_reply_queue(struct megas
 	if(instance->instancet->clear_intr(instance->reg_set))
 		return IRQ_NONE;
 
+	if(instance->hw_crit_error)
+		goto out_done;
         /*
 	 * Schedule the tasklet for cmd completion
 	 */
 	tasklet_schedule(&instance->isr_tasklet);
-
+out_done:
 	return IRQ_HANDLED;
 }
 

@@ -1741,6 +1748,10 @@ static void megasas_complete_cmd_dpc(uns
 	struct megasas_cmd *cmd;
 	struct megasas_instance *instance = (struct megasas_instance *)instance_addr;
 
+	/* If we have already declared adapter dead, donot complete cmds */
+	if(instance->hw_crit_error)
+		return;
+
 	producer = *instance->producer;
 	consumer = *instance->consumer;
drivers/scsi/megaraid/megaraid_sas.c +12 -1
diff -uprN 2.6.orig/drivers/scsi/megaraid/megaraid_sas.c 2.6.new-p1/drivers/scsi/megaraid/megaraid_sas.c
@@ -841,6 +841,11 @@ megasas_queue_command(struct scsi_cmnd *
 
 	instance = (struct megasas_instance *)
 	    scmd->device->host->hostdata;
+
+	/* Don't process if we have already declared adapter dead */
+	if(instance->hw_crit_error)
+		return SCSI_MLQUEUE_HOST_BUSY;
+
 	scmd->scsi_done = done;
 	scmd->result = 0;
 

@@ -1282,11 +1287,13 @@ megasas_deplete_reply_queue(struct megas
 	if(instance->instancet->clear_intr(instance->reg_set))
 		return IRQ_NONE;
 
+	if(instance->hw_crit_error)
+		goto out_done;
         /*
 	 * Schedule the tasklet for cmd completion
 	 */
 	tasklet_schedule(&instance->isr_tasklet);
-
+out_done:
 	return IRQ_HANDLED;
 }
 

@@ -1741,6 +1748,10 @@ static void megasas_complete_cmd_dpc(uns
 	struct megasas_cmd *cmd;
 	struct megasas_instance *instance = (struct megasas_instance *)instance_addr;
 
+	/* If we have already declared adapter dead, donot complete cmds */
+	if(instance->hw_crit_error)
+		return;
+
 	producer = *instance->producer;
 	consumer = *instance->consumer;
#1 lore raw

Review

Quoted 2 lines · On Tue, 06 Feb 2007 13:49:24 -0800 Sumant Patro <[email protected]> wrote:
(quoted diff) +2 -0
+	/* Don't process if we have already declared adapter dead */
+	if(instance->hw_crit_error)

Preferred kernel coding style is "if (".

Quoted 1 line

[p1-crit_err.patch text/x-patch (1.3KB)]

argh. Please don't send two copies of the patch in the email. Because the result applies happily with `patch --dry-run' and then creates a great mess when you try to apply the patch for real.