Bug#1089976: nfs-kernel-server: Fails to start - segmentation fault

[Bill Brelsford <wb@k2di.net>]

Hi Salvatore,

On Mon Dec 16 2024 at 09:36 PM +0100, Salvatore Bonaccorso wrote:
> What is commont on thoe system where you see the failure? Are all not
> using systemd as init? Additionally can you give some more details on
> your setup, how the exports look? Can we boild down the setup
> minimally to trigger the issue (and so report upstream)?

The other system is an old one that I maintain as a backup; it also
uses sysvinit. The nfs setup for both is simple, with no changes to
/etc/nfs.conf or /etc/default/nfs-kernel-server. /etc/exports on
the newer system:

   / 10.20.40.80/29(rw,fsid=8208,no_root_squash,no_subtree_check,sync)
   /u 10.20.40.80/29(rw,fsid=8202,no_root_squash,no_subtree_check,sync)
   /u1 10.20.40.80/29(rw,fsid=8203,no_root_squash,no_subtree_check,sync)
   /n/ws 10.20.40.80/29(rw,fsid=8206,no_root_squash,no_subtree_check,sync)
   /n/wt 10.20.40.80/29(rw,fsid=8207,no_root_squash,no_subtree_check,sync)

It had been working fine on both systems prior to the 2.8.2-1
upgrade. It also works on 5 bookworm systems with the same setup.

> Can you additionally please test to downgrade to the 2.8.1-2 version
> and please report back if you see the problem there was well? For us
> it should behave actually same as 2.8.1-1 but I would like to double
> check.

I downgraded to 2.8.1-1 because it was available in the trixie
repository. Where can I get 2.8.1-2?

I also have trixie installed on both systems. Trixie now uses
2.8.2-1 so I upgraded one of them -- it also fails.

> the supported way is to run it with systemd, the ship'ed sysvinit
> scripts contain legacy, for instance with new version we would start
> the server with nfsdctl:
> 
> ExecStart=/bin/sh -c '/usr/sbin/nfsdctl autostart || /usr/sbin/rpc.nfsd'
> 
> If you start it through nfsdctl do you get the exports working?

I haven't installed systemd -- nothing I use has required it -- so
nfsdctl isn't available.

> Still, having more information on the underlying setup would be
> helpful.

Another possible clue: after upgrading (to 2.8.2-1), there is no
problem -- I can start and stop the daemon, export/unexport
filesystems, etc.  Everything seems normal -- until the system is
rebooted and rpc.nfsd is invoked.

Regards..  Bill