Date: Sun, 17 Feb 2002 16:57:52 -0800 From: Terry Lambert <tlambert2@mindspring.com> To: Matthew Dillon <dillon@apollo.backplane.com> Cc: Poul-Henning Kamp <phk@critter.freebsd.dk>, Julian Elischer <julian@elischer.org>, Alfred Perlstein <bright@mu.org>, arch@FreeBSD.ORG, jhb@FreeBSD.ORG, peter@wemm.org, jake@locore.ca Subject: Re: gettimeofday() and crhold()/crfree() (was Re: gettimeofday()and copyout(). Is copyout() MPSAFE on non-i386 archs? ) Message-ID: <3C705190.8B56A024@mindspring.com> References: <5405.1013975811@critter.freebsd.dk> <200202172011.g1HKBsv88526@apollo.backplane.com> <3C7049A4.15412853@mindspring.com> <200202180030.g1I0UU309210@apollo.backplane.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Matthew Dillon wrote: > I'm just going to point something out here, and that is my proposed > system call fully supports the kernel saying, in effect, 'I don't want > you accessing this particular parameter from shared memory, use the > old way of doing it'. > > For the vast majority of processes in a system this is a perfectly > reasonable response. The shared memory feature would only need to be > enabled for those few processes, like web servers, databases, and > big threaded programs that really need it. > > So we can afford to waste some memory for the few processes that actually > need the feature as long as we don't waste any for the processes that > don't. When I was talking about the 4k/8k per process KVA mapping costs, I was not really concerned with the per process costs, I was merely documenting them, in case someone else was concerned. Personally, I think the number of clients goes up significantly faster than the number of processes or threads; obviously this would not be true if there were a thread per client, but I think that people who code that way will never beat my numbers on total clients per host, so I don't care about them anyway, since they are already shooting themselves before they even have a chance to enter the race. So actually, I would prefer that the call succeed always, and that the failure be fatal. The only reason to maintain support for system-call based access to the data is for legacy applications, IMO. -- Terry To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3C705190.8B56A024>