kern/150544: Panic, when viewing the list of ZFS snapshots

[Andriy Gapon]

The following reply was made to PR kern/150544; it has been noted by GNATS.

From: Andriy Gapon <avg at icyb.net.ua>
To: jhell <jhell at DataIX.net>
Cc: "Vladislav V. Prodan" <universite at ukr.net>, bug-followup at FreeBSD.org
Subject: Re: kern/150544: Panic, when viewing the list of ZFS snapshots
Date: Tue, 14 Sep 2010 09:19:09 +0300

 on 14/09/2010 08:56 jhell said the following:
 > Priority of this should be changed to *HIGH* & Severity changed to
 > *Critical*.
 > 
 > New synopsis: [ZFS][HIGH][CRIT] amd64 & i386 stable/8-ZFSv15 &
 > HEAD-ZFSv15, Panic, during ls(1) while in snapshot directories.
 > 
 > People BCC'd, pjd@ mm@ avg@ stable@ current@ to grab some more attention.
 > 
 > Backtraces: I have two available vmcore.37 & 38 along with core.txt.37 & 38.
 > 
 > Backtrace 37 attached.
 > 
 > Background: Because a normal user can access snapshot directories(.zfs)
 > they have the ability to crash a machine running HEAD or stable/8 with
 > ZFSv15 patches.
 > 
 > Workaround: Do not snapshot global readable directories or chmod go-rwx
 > /path/to where the snapshot directory (.zfs) is.
 > 
 > Systems effected thus far:
 > FreeBSD/i386 8.1-STABLE r212590M (ZFSv15 patches)
 > FreeBSD 9.0-CURRENT ? ?
 > Possibly 8.1-RELEASE (ZFSv15 patches)
 
 Looks like unimplemented vop_lock1.
 I think that default implementation usually comes in via default_vnodeops, but
 can be locally overridden.
 What kind of a vnode is this?  Can you print *vp in e.g. frame 8?
 
 -- 
 Andriy Gapon