Re: kern/45272: mutt rebooting my machine

[David Holland <dholland-bugs%netbsd.org@localhost>]

The following reply was made to PR kern/45272; it has been noted by GNATS.

From: David Holland <dholland-bugs%netbsd.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/45272: mutt rebooting my machine
Date: Mon, 22 Aug 2011 00:58:24 +0000

 On Fri, Aug 19, 2011 at 12:45:01PM +0000, Martin Husemann wrote:
  >  >        An fsck after rm looked clean (at least I think so, despite the
  >  >        mentioning of "UNRESOLVED INCONSISTENCIES"):
  >  
  >  That sounds like a bug in fsck worth its own PR?
 
 Yes, not marking the fs clean (especially if it was already clean)
 shouldn't trigger that message.
 
  >  > >How-To-Repeat:
  >  >         With the file removed, it is hopefully not repeatable.
  >  
  >  Unfortunately this also leaves little to no hope for further investigation.
 
 ...none at all, I would say.
 
 My guess is that the block containing the message body has gone bad
 and some combination of hardware and drivers has a badly hosed error
 recovery path -- e.g. respond to error by retrying but then trying to
 use the error code as the DMA address or something else equally doomed.
 
 Unless the submitter wants to try reading the raw disk a chunk at a
 time to see if the problem can be repeated this way (and if so, find
 the offending region) I don't think there's much we can do with this
 PR but close it. :(
 
 Note that it takes more than just an ordinary software fault to get a
 spontaneous reboot; I'm not absolutely sure about amd64 but for i386
 AIUI you pretty much have to garbage up the GDT or TSS, which normally
 takes some doing.
 
 -- 
 David A. Holland
 dholland%netbsd.org@localhost