Re: slru.c race condition (was Re: [HACKERS] TRAP: FailedAssertion("!((itemid)->lp_flags

Tom Lane wrote:
> Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> > Well, then what real options do we have? It seems the patch is just
> > required for all branches.
>
> I think it would be possible to fix it in a less invasive way by taking
> and releasing the ControlLock an extra time in SimpleLruReadPage and
> SimpleLruWritePage. What's indeterminate about that is the performance
> cost. In situations where there's not a lot of SLRU I/O traffic it's
> presumably negligible, but in a case like Jim's where there's evidently
> a *whole* lot of traffic, it might be a killer.

To me a performance problem is much harder get reports on and to locate
than a real fix to the problem. I think if a few people eyeball the
patch, it is OK for application. Are backpatches significantly
different?

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073