Bug#641155: pu: package mdadm/3.1.4-1+8efb9d1+squeeze1
On Sat, 2012-01-14 at 20:23 +0400, Michael Tokarev wrote:
> On 14.01.2012 19:50, Adam D. Barratt wrote:
> > + * Fix checkarray script so that it does not die after scheduling the first
> > + device when there is no scheduling class specified; thanks to Mario
> > + 'BitKoenig' Holbe (closes: #611627).
> > The fix for this includes this change:
> > - ionice -p "$resync_pid" $arg
> > + ionice -p "$resync_pid" $arg || :
> > I assume this is because $arg is unset and the ionice invocation thus
> > fails? Wouldn't it better to check that the arguments are sane before
> > attempting to run ionice instead?
> No, it is not due to empty $arg. For the full details see the discussion
> in the bugreport mentioned:
> The change you're talking about is to fix one half of the
> race condition: if the thread exits between the ps call but
> before ionice, so ionice reports error and - due to set -e
> in effect - whole thing terminates.
Ah, thanks for the explanation.
> > + * Schedule start/stop of mdadm-raid before/after filesystems are
> > + checked&mounted/unmounted; thanks to Mario 'BitKoenig' Holbe
> > + (closes: #611632).
> > [...]
> > + * Make mdadm-raid init script depend on hostname; thanks to Mario
> > + 'BitKoenig' Holbe (closes: #610421).
> > Have these changes to the LSB headers been tested on squeeze systems?
> I tested these briefly, no extensive tests were done. That's
> probably my mistake, but the problem is that it's difficult to
> come with some real testcases where this may go wrong.
> But lemme recheck it again, I'll send a follow-up email.
Thanks. I'm a little paranoid when it comes to changes that affect
things like when RAID scripts run. :-)