Skip to contentSkip to conversation

lpfc 8.3.42: Fixed issue of task management commands having a fixed timeout

PATCH In review James Smart 2 participants 4 messages Sep 6, 2013 – Sep 9, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 17/18 lpfc 8.3.42: Fixed issue of task management commands having a fixed timeout

Fixed issue of task management commands having a fixed timeout

---

 lpfc.h      |    1 +
 lpfc_attr.c |    8 ++++++++
 lpfc_scsi.c |    2 +-
 3 files changed, 10 insertions(+), 1 deletion(-)
drivers/scsi/lpfc/lpfc_attr.c +8 -0
diff -upNr a/drivers/scsi/lpfc/lpfc_attr.c b/drivers/scsi/lpfc/lpfc_attr.c
@@ -4115,6 +4115,12 @@ LPFC_ATTR_RW(poll_tmo, 10, 1, 255,
 	     "Milliseconds driver will wait between polling FCP ring");
 
 /*
+# lpfc_task_mgmt_tmo: Maximum time to wait for task management commands
+# to complete in seconds. Value range is [5,180], default value is 60.
+*/
+LPFC_ATTR_RW(task_mgmt_tmo, 60, 5, 180,
+	     "Maximum time to wait for task management commands to complete");
+/*
 # lpfc_use_msi: Use MSI (Message Signaled Interrupts) in systems that
 #		support this feature
 #       0  = MSI disabled
@@ -4300,6 +4306,7 @@ struct device_attribute *lpfc_hba_attrs[
 	&dev_attr_issue_reset,
 	&dev_attr_lpfc_poll,
 	&dev_attr_lpfc_poll_tmo,
+	&dev_attr_lpfc_task_mgmt_tmo,
 	&dev_attr_lpfc_use_msi,
 	&dev_attr_lpfc_fcp_imax,
 	&dev_attr_lpfc_fcp_cpu_map,
@@ -5279,6 +5286,7 @@ lpfc_get_cfgparam(struct lpfc_hba *phba)
 	lpfc_topology_init(phba, lpfc_topology);
 	lpfc_link_speed_init(phba, lpfc_link_speed);
 	lpfc_poll_tmo_init(phba, lpfc_poll_tmo);
+	lpfc_task_mgmt_tmo_init(phba, lpfc_task_mgmt_tmo);
 	lpfc_enable_npiv_init(phba, lpfc_enable_npiv);
 	lpfc_fcf_failover_policy_init(phba, lpfc_fcf_failover_policy);
 	lpfc_enable_rrq_init(phba, lpfc_enable_rrq);
drivers/scsi/lpfc/lpfc.h +1 -0
diff -upNr a/drivers/scsi/lpfc/lpfc.h b/drivers/scsi/lpfc/lpfc.h
@@ -708,6 +708,7 @@ struct lpfc_hba {
 	uint32_t cfg_multi_ring_type;
 	uint32_t cfg_poll;
 	uint32_t cfg_poll_tmo;
+	uint32_t cfg_task_mgmt_tmo;
 	uint32_t cfg_use_msi;
 	uint32_t cfg_fcp_imax;
 	uint32_t cfg_fcp_cpu_map;
drivers/scsi/lpfc/lpfc_scsi.c +1 -1
diff -upNr a/drivers/scsi/lpfc/lpfc_scsi.c b/drivers/scsi/lpfc/lpfc_scsi.c
@@ -5015,7 +5015,7 @@ lpfc_send_taskmgmt(struct lpfc_vport *vp
 	lpfc_cmd = lpfc_get_scsi_buf(phba, rdata->pnode);
 	if (lpfc_cmd == NULL)
 		return FAILED;
-	lpfc_cmd->timeout = 60;
+	lpfc_cmd->timeout = phba->cfg_task_mgmt_tmo;
 	lpfc_cmd->rdata = rdata;
 
 	status = lpfc_scsi_prep_task_mgmt_cmd(vport, lpfc_cmd, lun_id,
#1 lore raw
Quoted 1 line · James Smart

Fixed issue of task management commands having a fixed timeout

I'm surprised about this change, since it appears a number of issues in the
send_taskmgmt() still exist that keep it from handling task management failures correctly. It also continues to have a number of smaller issues like for example dead code of the form

if (status != IOCB_STATUS)
else if (status == IOCB_BUSY)
else

See patch:

"lpfc should check return status for task mgmt IOCB"

http://marc.info/?l=linux-scsi&m=136242124409687

#2 lore raw

Jeremy,

Different issues....

The patch addresses only 1 topic - making an attribute to change the timeout for the TMF. This was a request by the distros to post a hack that was used on older releases to work around all the serialization and delay problems in the midlayer error recovery path - which allowed shortening delays from 10-12 minutes to 1-2 minutes. These are the same issues we discussed at the last storage forum and which Hannes has been working to correct upstream. The desire is to continue with a like tuning option in the driver for those customers that have used it in the past.

The other issue - we seem to have missed your prior post. I'll look into it shortly.

The two patches work on the same code, but are independent topics.

-- james

Quoted 21 lines · Jeremy Linton
Quoted 1 line · James Smart

Fixed issue of task management commands having a fixed timeout

I'm surprised about this change, since it appears a number of issues in the
send_taskmgmt() still exist that keep it from handling task management failures correctly. It also continues to have a number of smaller issues like for example dead code of the form

if (status != IOCB_STATUS)
else if (status == IOCB_BUSY)
else

See patch:

"lpfc should check return status for task mgmt IOCB"

http://marc.info/?l=linux-scsi&m=136242124409687

#3 lore raw
Quoted 2 lines · James Smart

The other issue - we seem to have missed your prior post. I'll look into it shortly.

Thanks,

Those patches were the result of various error injection test cases we were
performing. The ones that come to mind were, hung sequences, TM rejects, failures to respond to the TM, etc.