Re: [NetBSD 8.0] Another panic in ffs_newvnode()

["J. Hannken-Illjes" <hannken%eis.cs.tu-bs.de@localhost>]

> On 9. Jul 2017, at 12:09, BERTRAND Joël <joel.bertrand%systella.fr@localhost> wrote:
> 
> J. Hannken-Illjes a écrit :
>> Is the /usr file system UFS1 or UFS2, output of dumpfs could help.
> 
> 	FFSv2 with log (and async flag).

Ok, UFS2 with dynamic inode initialisation.  Did your machine ever
crash since this file system got newfs'd?

The tail of ffs_nodealloccg() looks bogus, we first write the cg
synchronous so the new value of cg_initediblk gets to disk and then
write the initialised inodes asynchronously to disk.  If the machine
crashes before this async write completes we just created a bunch of
garbage inodes.

Maybe this should become:

	if (ibp != NULL) {
		bwrite(ibp);
		bwrite(bp);
	} else

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