[VDSO] [x86_32] v5-3-rc1 needs vdso32=0 to get systemd-journald running

[VDSO] [x86_32] v5-3-rc1 needs vdso32=0 to get systemd-journald running

[Paul Bolle]

My first attempts to boot v5.3-rc1 on my (ancient) ThinkPad X41 made systemd-
journald crash. I kept ending up with nasty my messages on the console:

         Starting Journal Service...
[...]
[    7.143552] systemd-journald[213]: Assertion 'clock_gettime(map_clock_id(clock_id), &ts) == 0' failed at ../src/basic/time-util.c:55, function now(). Aborting.
[FAILED] Failed to start Journal Service.
See 'systemctl status systemd-journald.service' for details.
[    7.220367] systemd-coredump[217]: Cannot resolve systemd-coredump user. Proceeding to dump core as root: No such process
[  OK  ] Stopped Journal Service.

And without systemd-journald I couldn't get userspace up and running.

A bit of tinkering showed that "vdso32=0" on the kernel command line allows me
to get a usable userspace.

Any idea where I should look next to pinpoint this?

Thanks,


Paul Bolle

Re: [VDSO] [x86_32] v5-3-rc1 needs vdso32=0 to get systemd-journald running

[Sean Christopherson]

On Fri, Jul 26, 2019 at 06:13:54PM +0200, Paul Bolle wrote:
> My first attempts to boot v5.3-rc1 on my (ancient) ThinkPad X41 made systemd-
> journald crash. I kept ending up with nasty my messages on the console:
> 
>          Starting Journal Service...
> [...]
> [    7.143552] systemd-journald[213]: Assertion 'clock_gettime(map_clock_id(clock_id), &ts) == 0' failed at ../src/basic/time-util.c:55, function now(). Aborting.
> [FAILED] Failed to start Journal Service.
> See 'systemctl status systemd-journald.service' for details.
> [    7.220367] systemd-coredump[217]: Cannot resolve systemd-coredump user. Proceeding to dump core as root: No such process
> [  OK  ] Stopped Journal Service.
> 
> And without systemd-journald I couldn't get userspace up and running.
> 
> A bit of tinkering showed that "vdso32=0" on the kernel command line allows me
> to get a usable userspace.
> 
> Any idea where I should look next to pinpoint this?

More than likely it's this:

https://lkml.kernel.org/r/20190719170343.GA13680@linux.intel.com

Re: [VDSO] [x86_32] v5-3-rc1 needs vdso32=0 to get systemd-journald running

[Paul Bolle]

Sean Christopherson schreef op vr 26-07-2019 om 09:20 [-0700]:
> More than likely it's this:
> 
> https://lkml.kernel.org/r/20190719170343.GA13680@linux.intel.com

Yes.

systemctl --version prints +SECCOMP so I guess systemd-journald has it enabled
too. So I now know which thread I should check. (The subject of that thread
contains "[5.2 REGRESSION]", but I'd say "[5.3-rc1 REGRESSION]" would more
accurate, but whatever.)

Thanks,


Paul Bolle

Re: [VDSO] [x86_32] v5-3-rc1 needs vdso32=0 to get systemd-journald running

[Sean Christopherson]

On Fri, Jul 26, 2019 at 07:12:41PM +0200, Paul Bolle wrote:
> Sean Christopherson schreef op vr 26-07-2019 om 09:20 [-0700]:
> > More than likely it's this:
> > 
> > https://lkml.kernel.org/r/20190719170343.GA13680@linux.intel.com
> 
> Yes.
> 
> systemctl --version prints +SECCOMP so I guess systemd-journald has it enabled
> too. So I now know which thread I should check. (The subject of that thread
> contains "[5.2 REGRESSION]", but I'd say "[5.3-rc1 REGRESSION]" would more
> accurate, but whatever.)

Doh, that's my bad.  The commit happened to show up right next to
the v5.2-rc6 tag in the log and I didn't bother checking to see if it was
actually in v5.2.