On Wed, 24 Dec 2008, Ludovico Cavedon wrote: > Alan Stern wrote: > > On Wed, 24 Dec 2008, Ludovico Cavedon wrote: > >> -it is happening also with 2.6.28-rc8 (which should have the patch, right?) > > > > What patch? Do you mean the patch at the end of that email thread? It > > affects only Argosy USB drives, not your Western Digital. > > I thought also patches > http://marc.info/?l=linux-scsi&m=122443015406309&w=2 > were included, but later I realized I was wrong. A revised version of the first patch in that message is queued for 2.6.29. The second patch has not yet been merged in any form. > I have patched my kernel with them, and now I/O errors get reported! > > >> I am not able to understand if this is HD problem or a kernel problem. I > > > > Partly both. The HD (or more likely, its USB interface) is responsible > > for sending those unnecessary empty sense records. The kernel is > > responsible for not reporting an I/O error (assuming an error actually > > did take place). > > I think I found out what is happening on the HD side. The SMART self > test fails with a read error. The SMART log reports uncorrectable read > errors. However Reallocated_Event_Count is 0. Searching on the web, > looks like that these sectors have bad ECC, so the cause a read error, > however they are not bad sectors. Is this correct? I don't know. It sounds reasonable. The real issue is: Why doesn't the drive send back appropriate sense information to let the host know about the bad ECC? > My question is: how can this happen? and not just one sector, but at > least a dozen! > Bad HD? (it's new! 93 hrs of activity so far!) Maybe you can exchange it... > > 2.6.27 doesn't go into an endless loop? This may indicate that > > eventually the drive stops sending the Check Condition status. > > You are right. Most of the read attempts succeed after a a while. > However I found one sector that is causing an endless loop on 2.6.27 and > unpatched 2.6.28-rc8 > > > It would help to see the dmesg log for when one of these errors occurs. > > There are no additional messages. However here it is: > http://pastebin.com/mcfd54a3 Yeah, that's not very useful. > > It would also help to know what happens under Windows. Do the same > > "empty sense" errors occur? If they do, how does Windows handle them? > > I can try to use usb snoopy to log usb traffic under windows. > Do you know how I can ask Windows "read sector X"? I wish I knew! Perhaps Microsoft's KnowledgeBase site can tell you how. Alan Stern -- 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
