Date: Mon, 2 Jul 2007 19:04:31 +0200 From: Ed Schouten <ed@fxq.nl> To: Jeff Roberson <jroberson@chesapeake.net> Cc: FreeBSD Current <freebsd-current@freebsd.org> Subject: Re: New SCHED_SMP diff. Message-ID: <20070702170431.GG77379@hoeg.nl> In-Reply-To: <20070701224452.I552@10.0.0.1> References: <20070701224452.I552@10.0.0.1>
index | next in thread | previous in thread | raw e-mail
[-- Attachment #1 --] * Jeff Roberson <jroberson@chesapeake.net> wrote: > This diff also includes another patch that I intend to commit to current > which significantly improved write performance for mysql by removing all > use of Giant in fcntl(). Some time ago I filed a PR that F_DUPFD performs a lot of useless locking. It first performs lookups of the descriptor to make sure it exists, but afterwards it just calls do_dup(), which does the lookups again. The patch shaves off a pair of FILEDESC_XLOCK/FILEDESC_XUNLOCK and PROC_LOCK/PROC_UNLOCK calls. http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/105346 http://g-rave.nl/junk/freebsd-kern_descrip.diff The only reason why F_DUPFD performs these checks, is because it could return EINVAL in some cases, while dup() and dup2() only return EMFILE and EBADFD. This patch makes do_dup() return EINVAL when the offset is invalid, so that's why the dup()/dup2() syscalls are changed as well. Yours, -- Ed Schouten <ed@fxq.nl> WWW: http://g-rave.nl/ [-- Attachment #2 --] -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.7 (FreeBSD) iD4DBQFGiTAf52SDGA2eCwURAn6cAJi6O+DPFIAQJb6o/jpuy49eruM8AJ9Co9m4 8AuCi3j8Pp8DhU1xO2jZGA== =AiT/ -----END PGP SIGNATURE-----help
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070702170431.GG77379>