projects / people / pmueller / ipfire-2.x.git / blame
| Commit | Line | Data |
|---|---|---|
| 2cb7cef9 BS |
1 | From: Gerald Schaefer <geraldsc@de.ibm.com> |
| 2 | Subject: zfcp: Recover from stalled outbound queue | |
| 3 | References: bnc#518291,LTC#54460 | |
| 4 | ||
| 5 | Symptom: In some error scenarious, the channel can stall. To | |
| 6 | Linux this looks like a stall in the outbound queue. | |
| 7 | Linux has to recover this. | |
| 8 | Problem: There is no trigger to recover the queues if the outbound | |
| 9 | queue stalls. | |
| 10 | Solution: When trying to issue a non-SCSI command, zfcp waits for | |
| 11 | 5 seconds to get a free SBAL in the queue. Use this timeout | |
| 12 | as a trigger: If after 5 seconds there is no free SBAL, | |
| 13 | assume a stalled outbound queue and trigger recovery. | |
| 14 | ||
| 15 | Acked-by: John Jolly <jjolly@suse.de> | |
| 16 | --- | |
| 17 | drivers/s390/scsi/zfcp_dbf.c | 2 +- | |
| 18 | drivers/s390/scsi/zfcp_fsf.c | 5 ++++- | |
| 19 | 2 files changed, 5 insertions(+), 2 deletions(-) | |
| 20 | ||
| 21 | --- a/drivers/s390/scsi/zfcp_fsf.c 2009-06-23 14:24:48.000000000 +0200 | |
| 22 | +++ b/drivers/s390/scsi/zfcp_fsf.c 2009-06-23 14:26:00.000000000 +0200 | |
| 23 | @@ -668,8 +668,11 @@ static int zfcp_fsf_req_sbal_get(struct | |
| 24 | zfcp_fsf_sbal_check(adapter), 5 * HZ); | |
| 25 | if (ret > 0) | |
| 26 | return 0; | |
| 27 | - if (!ret) | |
| 28 | + if (!ret) { | |
| 29 | atomic_inc(&adapter->qdio_outb_full); | |
| 30 | + /* assume hanging outbound queue, try queue recovery */ | |
| 31 | + zfcp_erp_adapter_reopen(adapter, 0, 77, NULL); | |
| 32 | + } | |
| 33 | ||
| 34 | spin_lock_bh(&adapter->req_q_lock); | |
| 35 | return -EIO; | |
| 36 | --- a/drivers/s390/scsi/zfcp_dbf.c 2009-06-23 14:20:41.000000000 +0200 | |
| 37 | +++ b/drivers/s390/scsi/zfcp_dbf.c 2009-06-23 14:25:50.000000000 +0200 | |
| 38 | @@ -567,7 +567,7 @@ static const char *zfcp_rec_dbf_ids[] = | |
| 39 | [74] = "unit recovery follow-up after successful port recovery", | |
| 40 | [75] = "port recovery retry after failure", | |
| 41 | [76] = "unit recovery retry after failure", | |
| 42 | - [77] = "", | |
| 43 | + [77] = "adapter reopen from queue stall", | |
| 44 | [78] = "duplicate request id", | |
| 45 | [79] = "link down", | |
| 46 | [80] = "exclusive read-only unit access unsupported", |