Date: Fri, 14 Mar 2003 02:04:06 -0800 From: David Schultz <das@FreeBSD.org> To: Mike Silbersack <silby@silby.com> Cc: Alfred Perlstein <bright@mu.org>, src-committers@FreeBSD.org, cvs-src@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/vm vm_map.c vm_map.h vm_pageout.c Message-ID: <20030314100406.GA20955@HAL9000.homeunix.com> In-Reply-To: <20030314010729.D988@odysseus.silby.com> References: <200303122313.h2CNDHMU046431@repoman.freebsd.org> <20030312175458.J32334@odysseus.silby.com> <20030313005115.GA11794@HAL9000.homeunix.com> <20030313154226.X682@odysseus.silby.com> <20030313224223.GP4145@elvis.mu.org> <20030314064614.GA20500@HAL9000.homeunix.com> <20030314010729.D988@odysseus.silby.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Thus spake Mike Silbersack <silby@silby.com>: > > On Thu, 13 Mar 2003, David Schultz wrote: > > > Thus spake Alfred Perlstein <bright@mu.org>: > > > Shooting from the hip here, but why not just forcefully toss away the > > > process's vmspace or a subset of it? Basically have the pageout > > > deamon deallocate the "to be killed" process's address space? > > > > Sharing would just make that a pain. You're not even guaranteed > > to get anything out of cannibalizing the address space if it's > > shared. I think I'll wait until someone presents conclusive > > evidence that the requirement that processes be swapped in before > > they are killed is really an issue in practice. > > David, we're just discussing the issue, not requiring you to fix it. As > such, there's no need to deny the existence of the problem I saw last time > I did OOM testing. > > It's OK to admit that there are bugs in the kernel which aren't going to > be fixed immediately, don't get overly excited about it. Hehe...I'm not trying to avoid work or sound pessimistic, I'm just saying that this might be more trouble than it's worth. The effect where the pagedaemon has to kill a process multiple times is just a race condition where the pagedaemon gets activated multiple times before the victim process finishes dying. A *real* problem would be if the system deadlocks because there's no memory to swap in the largest process' upages in order to kill it. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030314100406.GA20955>