On Wed, 2015-09-02 at 09:31 -0500, Brian King wrote:
> This patch fixes an issue seen with an IBM 2145 (SVC) where, following an error
> injection test which results in paths going offline, when they came
> back online, the path would timeout the REPORT_LUNS issued during the
> scan. This timeout situation continued until retries were expired, resulting in
> falling back to a sequential LUN scan. Then, since the target responds
> with PQ=1, PDT=0 for all possible LUNs, due to the way the sequential
> LUN scan code works, we end up adding 512 LUNs for each target, when there
> is really only a small handful of LUNs that are actually present.
>
> This patch doubles the timeout used on the REPORT_LUNS for each retry
> after a timeout is seen on a REPORT_LUNS. This patch solves the issue
> of 512 non existent LUNs showing up after this event. Running the test
> with this patch still showed that we were regularly hitting two timeouts,
> but the third, and final, REPORT_LUNS was always successful.
>
> Signed-off-by: Brian King <brking@xxxxxxxxxxxxxxxxxx>
> ---
>
> drivers/scsi/scsi_scan.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff -puN drivers/scsi/scsi_scan.c~scsi_report_luns_timeout_escalate drivers/scsi/scsi_scan.c
> --- linux/drivers/scsi/scsi_scan.c~scsi_report_luns_timeout_escalate 2015-09-02 08:49:07.268243497 -0500
> +++ linux-bjking1/drivers/scsi/scsi_scan.c 2015-09-02 08:49:07.272243461 -0500
> @@ -1304,6 +1304,7 @@ static int scsi_report_lun_scan(struct s
> struct scsi_device *sdev;
> struct Scsi_Host *shost = dev_to_shost(&starget->dev);
> int ret = 0;
> + int timeout = SCSI_TIMEOUT + 4 * HZ;
>
> /*
> * Only support SCSI-3 and up devices if BLIST_NOREPORTLUN is not set.
> @@ -1383,7 +1384,7 @@ retry:
>
> result = scsi_execute_req(sdev, scsi_cmd, DMA_FROM_DEVICE,
> lun_data, length, &sshdr,
> - SCSI_TIMEOUT + 4 * HZ, 3, NULL);
> + timeout, 3, NULL);
>
> SCSI_LOG_SCAN_BUS(3, sdev_printk (KERN_INFO, sdev,
> "scsi scan: REPORT LUNS"
> @@ -1392,6 +1393,8 @@ retry:
> retries, result));
> if (result == 0)
> break;
> + else if (host_byte(result) == DID_TIME_OUT)
> + timeout = timeout * 2;
> else if (scsi_sense_valid(&sshdr)) {
> if (sshdr.sense_key != UNIT_ATTENTION)
Actually, this is a bit pointless, isn't it; why retry, why not just set
the initial timeout? ... I could understand if retrying and printing a
message gave important or useful information, but it doesn't. How long
do you actually need? ... we can just up the initial timeout to that.
Currently we have a hacked 6s which looks arbitrary. Would 15s be
better? Nothing really times out anyway, so everything else will still
reply within the original 6s giving zero impact in the everyday case.
James
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
