Subject: kern/21701: RAIDframe should stop device, not OS, on error
To: None <>
From: None <>
List: netbsd-bugs
Date: 05/28/2003 12:29:04
>Number:         21701
>Category:       kern
>Synopsis:       Kill the filesystem instead of panicing on an unrecoverable error
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    kern-bug-people
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Wed May 28 03:30:00 UTC 2003
>Originator:     Curt Sampson
>Release:        NetBSD 1.6K
System: NetBSD 1.6K NetBSD 1.6K (ANGELIC2-$Revision: 1.5 $) #0: Tue Nov 26 11:52:20 JST 2002 i386
Architecture: i386
Machine: i386

    Currently, if multiple components in a RAID device fail such that
    the RAID device can no longer reliably continue operation (e.g.,
    in a RAID-5 with one parity disk, two components fail), RAIDframe
    panics the system. This allows some hope of recovery if the second
    component did not fail due to a permenent hardware error. However,
    there's no need to go quite so far; simply deconfiguring that
    RAID device (thus forcably unmounting the filesystem) would also
    do the trick quite nicely, and leave the system up if it were a
    non-critical RAID device that failed.