| Message ID | 1513855354-86603-3-git-send-email-hare@suse.de (mailing list archive) |
|---|---|
| State | Deferred |
| Headers | show |
On Thu, 2017-12-21 at 12:22 +0100, Hannes Reinecke wrote: > Some storage array set the 'Embedded enclosure' bit even though > no LUN is present, causing the first RECEIVE DIAGNOSTIC call to > be returned with sense code 'LOGICAL UNIT NOT SUPPORTED'. > This patch skips the annoying 'Failed to get diagnostic page 0x1' > messages for those cases. What disagnostic pages does this thing support? Can you do a receive diagnostic on page 0 to find out? I suspect a lot of embedded enclosure services are simple and support page 7 only. If it really refuses all diagnostic page requests (which would be a gross standards violation), then it should probably be blacklisted by inquiry string as unusable. James
On 12/22/2017 06:39 PM, James Bottomley wrote: > On Thu, 2017-12-21 at 12:22 +0100, Hannes Reinecke wrote: >> Some storage array set the 'Embedded enclosure' bit even though >> no LUN is present, causing the first RECEIVE DIAGNOSTIC call to >> be returned with sense code 'LOGICAL UNIT NOT SUPPORTED'. >> This patch skips the annoying 'Failed to get diagnostic page 0x1' >> messages for those cases. > > What disagnostic pages does this thing support? Can you do a receive > diagnostic on page 0 to find out? I suspect a lot of embedded > enclosure services are simple and support page 7 only. If it really > refuses all diagnostic page requests (which would be a gross standards > violation), then it should probably be blacklisted by inquiry string as > unusable. > If a LUN is connected it'll respond to the usual set (0, 1, 2, und 6). Sending a RECEIVE DIAGNOSTIC to page 0 will probably yield the same result (ie being returned with that sense code), but at least we won't try to access the other pages. But anyway, yes, the implementation is dodgy. I'm in contact with the vendor and try to straighten things out. Cheers, Hannes
diff --git a/drivers/scsi/ses.c b/drivers/scsi/ses.c index e8ffee1..c1f96b0 100644 --- a/drivers/scsi/ses.c +++ b/drivers/scsi/ses.c @@ -110,12 +110,17 @@ static int ses_recv_diag(struct scsi_device *sdev, int page_code, 0 }; unsigned char recv_page_code; + struct scsi_sense_hdr sshdr; ret = scsi_execute_req(sdev, cmd, DMA_FROM_DEVICE, buf, bufflen, - NULL, SES_TIMEOUT, SES_RETRIES, NULL); - if (unlikely(ret)) + &sshdr, SES_TIMEOUT, SES_RETRIES, NULL); + if (unlikely(ret)) { + if (status_byte(ret) == CHECK_CONDITION && + sshdr.asc == 0x25 && sshdr.ascq == 0x00) { + ret = -ENODEV; + } return ret; - + } recv_page_code = ((unsigned char *)buf)[0]; if (likely(recv_page_code == page_code)) @@ -674,9 +679,14 @@ static int ses_intf_add(struct device *cdev, page = 1; result = ses_recv_diag(sdev, page, hdr_buf, ses_alloc_size); - if (result) + if (result) { + if (result == -ENODEV && + sdev->inq_periph_qual != 0) { + kfree(hdr_buf); + return result; + } goto recv_failed; - + } len = (hdr_buf[2] << 8) + hdr_buf[3] + 4; buf = kzalloc(len, GFP_KERNEL); if (!buf)
Some storage array set the 'Embedded enclosure' bit even though no LUN is present, causing the first RECEIVE DIAGNOSTIC call to be returned with sense code 'LOGICAL UNIT NOT SUPPORTED'. This patch skips the annoying 'Failed to get diagnostic page 0x1' messages for those cases. Signed-off-by: Hannes Reinecke <hare@suse.com> --- drivers/scsi/ses.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-)