linux-kernel - Re: reading errors on JMicron JM20337 USB-SATA

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <Pine.LNX.4.44L0.0908031142180.16208-100000@iolanthe.rowland.org>
Date:	Mon, 3 Aug 2009 11:45:26 -0400 (EDT)
From:	Alan Stern <stern@...land.harvard.edu>
To:	Artur Skawina <art.08.09@...il.com>
cc:	"Lev A. Melnikovsky" <melnikovsky@...l.ru>,
	Linux Kernel <linux-kernel@...r.kernel.org>,
	<linux-usb@...r.kernel.org>
Subject: Re: reading errors on JMicron JM20337 USB-SATA

On Mon, 3 Aug 2009, Artur Skawina wrote:

> > You are correct except for the term "indefinitely".  The retries _will_
> > stop if you wait long enough.  Unfortunately, because of all the nested
> > retry loops in the SCSI drivers and at the application level, you may
> > have to wait as long as half an hour.
> 
> iirc, i had stalls _way_ longer than that, probably because the reads
> eventually succeeded, only to stall on the next ones.
> 
> > I agree that this should be fixed.  But it is a SCSI issue, not a USB 
> > issue.  You could try bringing it up on the linux-scsi mailing list.
> 
> actually, the number of retries should probably be configurable, but i
> wouldn't lower them by default; losing data because of recoverable errors
> is bad. In this case the bridge may be at fault (by not passing along the
> error), but to make a significant difference you'd have to reduce the number
> of retries to something like zero, maybe one at most, and that's just too
> low for a default. 

As I understand it, the SCSI and block layers conspire to keep retrying
each command until a timeout expires, not until the number of retries 
reaches a limit.

But the situation is complicated, because some kinds of retries reset 
the timer.  And if the application repeats the I/O request then of 
course everything starts over again.

Alan Stern

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/