Le Fri, 18 Apr 2014 14:29:46 -0400,
Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> a écrit :
> On Fri, 18 Apr 2014, Matthieu CASTET wrote:
>
> > "echo mem > /sys/power/state; cat /dev/sdx > /dev/null" works.
>
> I guess this works because the initial open of /dev/sdx swallows the
> media-changed notification.
Ok
>
> > > The device reported a "Not-ready to ready change" because the power was
> > > lost and then restored. The SCSI layer interprets this as meaning that
> > > the media was changed, even though USB keys don't have changeable media
> > > -- evidently your key tells the computer that it does.
> >
> > yes, I tested with differents usb key and all of them report removable
> > (1
> > in /sys/devices/pci0000:00/*/usb*/*/*/host*/target*/*/block/sdx/removable).
>
> > > Since the SCSI layer thinks the media was changed, it has no choice but
> > > to fail the read request.
> > ok.
> >
> > I suppose the scsi layer can't ignore "Not-ready to ready change" event
> > at resume time ?
>
> But that event doesn't happen at resume time. It happens later, when
> the "cat" program issues its next read request.
Yes, but we can ignore the first "Not-ready to ready change" event
after resume.
In fact after reset, we call usb_stor_reset_resume that do a
scsi_report_bus_reset.
This will set the flag expecting_cc_ua in scsi layer and the scsi layer
will ignore the first unit attention.
But the commit dfcf7775815504d13a1d273073810058caf84b9d change the
logic and we don't ignore unit attention if "sshdr.asc == 0x28 &&
sshdr.ascq == 0x00" ("Not-ready to ready").
> In any case, the SCSI people will probably argue that ignoring such
> events would not be safe. And if it's not safe to ignore those events
> during normal operation then it's not safe to ignore them at resume
> time.
>
> If you want to discuss the matter any further, you should post to the
> linux-scsi mailing list. There's not much we can do about it at the
> USB level -- although perhaps we could turn off the erroneous
> "removable" indicator.
Thanks,
I will continue the discution on linux-scsi ML.
Matthieu CASTET
[1]
commit dfcf7775815504d13a1d273073810058caf84b9d
Author: TARUISI Hiroaki <taruishi.hiroak@xxxxxxxxxxxxxx>
Date: Thu Aug 11 20:25:20 2011 +0900
[SCSI] Fix out of spec CD-ROM problem with media change
Some CD-ROMs fail to report a media change correctly. The specific
one for this patch simply fails to respond to commands, then gives a
UNIT ATTENTION after being reset which returns ASC/ASCQ 28/00. This
is out of spec behaviour, but add a check in the eat CC/UA on reset
path to catch this case so the CD-ROM will function somewhat properly.
[jejb: fixed up white space and accepted without signoff]
Signed-off-by: James Bottomley <JBottomley@xxxxxxxxxxxxx>
diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
index a4b9cdb..dc6131e 100644
--- a/drivers/scsi/scsi_error.c
+++ b/drivers/scsi/scsi_error.c
@@ -293,8 +293,16 @@ static int scsi_check_sense(struct scsi_cmnd *scmd)
* so that we can deal with it there.
*/
if (scmd->device->expecting_cc_ua) {
- scmd->device->expecting_cc_ua = 0;
- return NEEDS_RETRY;
+ /*
+ * Because some device does not queue unit
+ * attentions correctly, we carefully check
+ * additional sense code and qualifier so as
+ * not to squash media change unit attention.
+ */
+ if (sshdr.asc != 0x28 || sshdr.ascq != 0x00) {
+ scmd->device->expecting_cc_ua = 0;
+ return NEEDS_RETRY;
+ }
}
/*
* if the device is in the process of becoming ready, we
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
