On Wednesday 30 December 2009 01:13:01 Simon Riggs wrote:
> On Tue, 2009-12-29 at 11:13 -0500, Tom Lane wrote:
> > Andres Freund <andres@anarazel.de> writes:
> > > On Tuesday 29 December 2009 16:22:54 Tom Lane wrote:
> > >> This seems like a fairly bad idea. One of the intended use-cases is
> > >> to be able to manually "kill -INT" a misbehaving backend. Assuming
> > >> that there will be valid info about the signal in shared memory will
> > >> break that.
> > >
> > > Well. That already is the case now. MyProc->recoveryConflictMode is
> > > checked to recognize what kind of conflict is being resolved...
> >
> > In that case, HS has already broken it, and we need to fix it not make
> > it worse.
> >
> > My humble opinion is that SIGINT should not be overloaded with multiple
> > meanings. We already have a multiplexed signal mechanism, which is what
> > should be used for any additional signal reasons HS may need to
> > introduce.
>
> It's a revelation to me, but yes, I see it now and agree.
>
> I'm looking at Fujii-san's multiplexing patch from Jul 31 to rewrite
> this code using that mechanism. It sounds like it's a neat fit and it
> should get around the bug report from Kris also if it all works.
Hm. I just read a bit of that multiplexing facility (out of a different reason)
and I have some doubt about it being used unmodified for canceling backends:
procsignal.c:
/** Note: Since there's no locking, it's possible that the target* process detaches from shared memory and exits right
afterthis* test, before we set the flag and send signal. And the signal slot* might even be recycled by a new process,
soit's remotely possible* that we set a flag for a wrong process. That's OK, all the signals* are such that no harm is
doneif they're mistakenly fired.*/
procsignal.h:
...* Also, because of race conditions, it's important that all the signals be* defined so that no harm is done if a
processmistakenly receives one.*/
When cancelling a backend that behaviour could be a bit annoying ;-)
I guess locking procarray during sending the signal should be enough?
Andres
