[Pkg-sysvinit-devel] Bug#595431: Aborting fsck aborts all scripts in rcS.d

[Goswin von Brederlow]

Sven Joachim <svenjoac at gmx.de> writes:

> reassign 595431 sysvinit-utils
> found 595431 2.88-12
> thanks
>
> On 2010-09-03 22:39 +0200, Goswin von Brederlow wrote:
>
>> Package: insserv
>> Version: 1.14.0-2
>> Severity: critical
>>
>> Hi,
>>
>> during boot /etc/rcS.d/S13checkfs.sh starts a filesystem check (hasn't
>> been checked for 197 days) as well as giving some errors for missing
>> devices. Since I didn't want to wait for the fsck before fixing the
>> missing devices I aborted the check with crlt-c. This resulted in the
>> fsck to be aborted but then also skipped all further rcS.d scripts
>> saying:
>>
>> Running scripts in rcS.d/ took 41 seconds.
>> INIT: Entering runlevel: 2
>>
>> Given that filesystem weren't mounted or anything that didn't work out
>> well leaving the system unusable.
>
> I can reproduce this, but only if parallel booting is enabled (the
> default).
>
>> This is a serious regressions from before insserv. The old behaviour
>> was to display a message asking for the root password to get a shell
>> or ctrl-D to continue booting.
>
> Well, was it?  I think Ctrl-c should just abort the fsck and continue
> the boot (this happens here with CONCURRENCY=none in /etc/default/rcS,
> but I don't have any actual filesystem problems and booted with the
> forcefsck option).  Anyway, this seems to be a bug in startpar's signal
> handling, thus reassigning to sysvinit-utils.
>
> Sven

In my case there were fsck errors because a raid device didn't come up
and there it should give the prompt. At least that is the old behaviour.
Continue to boot with devices missing isn't a good plan.

But the cause will most likely be the same.

MfG
        Goswin