Subject: Re: fss+quota panic in netbsd-3
To: Manuel Bouyer <bouyer@antioche.eu.org>
From: Juergen Hannken-Illjes <hannken@eis.cs.tu-bs.de>
List: tech-kern
Date: 11/25/2006 18:34:29
On Sat, Nov 25, 2006 at 06:09:11PM +0100, Manuel Bouyer wrote:
> Hi,
> I've got the panic below (note it's not frequent; I got it once in
> several months, on a system backed up every night):
> chkdquot: missing dquot: tag VT_UFS(1) type VREG(1), usecount 1, writecount 0, r
> efcount 1, flags (080<VLOCKSWORK,ONWORKLIST>)
>         tag VT_UFS, ino 7, on dev 18, 54 flags 0x0, effnlink 1, nlink 1
>         mode 0100600, owner 0, group 0, size 0 lock type vnlock: EXCL (count 1)
> by pid 3420.1
> panic: missing dquot
> Begin traceback...
> chkdquot(d7c52ef4,1b0000,ceb465dd,c02930d1,c241be00) at netbsd:chkdquot+0x4a
> chkdq(d7c52ef4,10,0,d04aa544,0) at netbsd:chkdq+0x16f
> ffs_alloc(d7c52ef4,328158,0,9f96c8,0) at netbsd:ffs_alloc+0x175
> ffs_balloc_ufs1(ceb4b474,0,ceb4b4dc,c01fc3e4,ceb4b484) at netbsd:ffs_balloc_ufs1
> +0xcde
> ufs_gop_alloc(cf01f0b4,502b0000,6,8,0) at netbsd:ufs_gop_alloc+0xbf
> ufs_balloc_range(cf01f0b4,502b0000,6,8,0) at netbsd:ufs_balloc_range+0x2a3
> ffs_write(ceb4b774,1238,ceb4b75c,c0273bbb,cf064710) at netbsd:ffs_write+0x46a
> vn_rdwr(1,cf01f0b4,ceb4b92c,8,502b0000) at netbsd:vn_rdwr+0xe1
> ffs_snapshot(c3649000,cf01f0b4,ceb4bb6c,1fff,10) at netbsd:ffs_snapshot+0x24e
> fss_create_files(c03b14d0,ceb4bea4,ceb4bce4,cf3de4d4,c3649000) at netbsd:fss_cre
> ate_files+0x556
> fss_create_snapshot(c03b14d0,ceb4bea4,cf3de4d4,40,ceb4bd44) at netbsd:fss_create
> _snapshot+0x1d
> fss_ioctl(a302,800c4600,ceb4bea4,3,cf3de4d4) at netbsd:fss_ioctl+0x13e
> spec_ioctl(ceb4bd84,1,ceb4be0c,c01fdfb8,ceb4bd94) at netbsd:spec_ioctl+0x6e
> vn_ioctl(d16f3014,800c4600,ceb4bea4,cf3de4d4,0) at netbsd:vn_ioctl+0xd7
> sys_ioctl(d04ad7cc,ceb4bf64,ceb4bf5c,0,0) at netbsd:sys_ioctl+0x12e
> syscall_plain() at netbsd:syscall_plain+0x18f
> 
> 
> This is a netbsd-3 kernel from march, with sys/dev/fss.c 1.13.2.3 but not
> 1.13.2.4. If 1.13.2.4 is likely to fix it It's not worth investigating,
> otherwise I have a core dump ...

Rev 1.13.2.4 will not fix it.  This revision adds missing FSS_LOCK() for
I/O errors.

-- 
Juergen Hannken-Illjes - hannken@eis.cs.tu-bs.de - TU Braunschweig (Germany)