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

[Salvatore Bonaccorso <carnil@debian.org>]

Hi Bill,

On Sun, Dec 15, 2024 at 06:46:46PM -0800, Bill Brelsford wrote:
> Hi Salvatore,
> 
> On Sun Dec 15 2024 at 02:15 PM +0100, Salvatore Bonaccorso wrote:
> > I'm not able to reproduce it here. Can you please Install the dbgsym
> > packages as well and get more information by making sure the service
> > is stopped and start it by hand under debugger.
> > 
> > This might give some more clue for upstream.
> 
> I apparently need some help with gdb to get useful output. After
> running "find-dbgsym-packages /usr/sbin/rpc.nfsd", I installed
> libc6-dbg and nfs-kernel-server-dbgsym. Then I stopped
> /etc/init.d/nfs-kernel-server just before the call to rpc.nfsd
> and ran gdb:
> 
>    # gdb /usr/sbin/rpc.nfsd
>    GNU gdb (Debian 15.2-1) 15.2
>    ...
>    Reading symbols from /usr/sbin/rpc.nfsd...
>    Reading symbols from /usr/lib/debug/.build-id/4c/400a0c5314bb3884d7adbde7889a7bbc3a0eaa.debug...
>    (gdb) run
>    Starting program: /usr/sbin/rpc.nfsd 
>    [Thread debugging using libthread_db enabled]
>    Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
> 
>    Program terminated with signal SIGSEGV, Segmentation fault.
>    The program no longer exists.
>    (gdb) bt
>    No stack.
> 
> Suggestions?
> 
> Also, after rpc.nfsd fails, running it again (or, e.g., exportfs -r)
> hangs and can't be killed.
> 
> The same failure occurs on another similarly-configured system.

Let's try to tackle it from another angle. 

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)?

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.

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?

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

Regards,
Salvatore