Re: panic ffs_valloc dup_alloc

[christos%zoulas.com@localhost (Christos Zoulas)]

On Dec 14, 10:37pm, martians%sdf.org@localhost ("Stephen M. Jones") wrote:
-- Subject: Re: panic ffs_valloc dup_alloc

| > Stephen M. Jones <martians%sdf.org@localhost> wrote:
| > >> > In some cases the filesystem corruption can only be corrected by a newfs.
| > >> 
| > >> Does it happen with FFSv2? I also had lot of FS corruption lately. it
| > >> only hapened on some specific machines (moving VM to another host fixed
| > >> the problem), hence I suspected some problems with SATA controllers, I
| > >> also had the feeling FFSv2 behaves better than FFSv1.
| > >
| > >This is FFSv1 with various SATA controllers, I can get more specific.
| > >In two cases it was the root file system (ffsv1) that became corrupt and
| > >although several passes (and clearing of DUPs) with fsck marked a file
| > >system as 'clean' the machine would ultimately kernel panic.  newfs seems
| > >to be the only way out of that.
| > >
| > >Can a root file system be ffsv2? I've been going with defaults to keep things
| > >as generic as possible.  
| > 
| > Yes, it can... OTOH, if you are going to do that, use netbsd-8 instead...
| 
| Yikes, these are production systems.  Is NetBSD-8 ready for that?  Is the
| patch you posted to address the dup alloc for ffs1 sources in the release
| of NetBSD-7.1?

No the patch is not there and it does not per say fix anything but improves
the information we get when we panic. It also fixes an NFS related panic
and corruption. If you are exporting via NFS you might get that issue.

| We can probably manage an upgrade (to NetBSD-8) on Sunday to minimize the
| impact on 54585 users.  What is the largest NetBSD deployment at the moment?

I don't know any bigger than yours :-) Why don't you boot an 8 kernel with
a 7 userland and see how that goes?

christos