Re: POSIX shared memory redux

On Apr 11, 2011, at 7:25 PM, Tom Lane wrote:

> Robert Haas <robertmhaas(at)gmail(dot)com> writes:
>> On Sun, Apr 10, 2011 at 5:03 PM, A.M. <agentm(at)themactionfaction(dot)com> wrote:
>>> To ensure that no two postmasters can startup in the same data directory, I use fcntl range locking on the data directory lock file, which also works properly on (properly configured) NFS volumes. Whenever a postmaster or postmaster child starts, it acquires a read (non-exclusive) lock on the data directory's lock file. When a new postmaster starts, it queries if anything would block a write (exclusive) lock on the lock file which returns a lock-holding PID in the case when other postgresql processes are running.
>
>> This seems a lot leakier than what we do now (imagine, for example,
>> shared storage) and I'm not sure what the advantage is.
>
> BTW, the above-described solution flat out doesn't work anyway, because
> it has a race condition. Postmaster children have to reacquire the lock
> after forking, because fcntl locks aren't inherited during fork(). And
> that means you can't tell whether there's a just-started backend that
> hasn't yet acquired the lock. It's really critical for our purposes
> that SysV shmem segments are inherited at fork() and so there's no
> window where a just-forked backend isn't visible to somebody checking
> the state of the shmem segment.

Then you haven't looked at my patch because I address this race condition by ensuring that a lock-holding violator is the postmaster or a postmaster child. If such as condition is detected, the child exits immediately without touching the shared memory. POSIX shmem is inherited via file descriptors.

Cheers,
M