Re: [PATCH] fix race in AF_UNIX

From: David Miller
Date: Wed Jun 06 2007 - 01:41:32 EST


From: Miklos Szeredi <miklos@xxxxxxxxxx>
Date: Wed, 06 Jun 2007 07:26:52 +0200

> > Holding a global mutex over recvmsg() calls under AF_UNIX is pretty
> > much a non-starter, this will kill performance for multi-threaded
> > apps.
>
> That's an rwsem held for read. It's held for write in unix_gc() only
> for a short duration, and unix_gc() should only rarely be called. So
> I don't think there's any performance problem here.

It pulls a non-local cacheline into the local thread, that's extremely
expensive on SMP.

If everyone starts grabbing this thing during recvmsg() it's going to
become a really hot lock and kill performance, even if it's a read
side lock being taken.

That's why I said we need to investigate solutions involving
u->readlock, that already has to be taken and is local to the socket.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/