Date: Sun, 12 Apr 2009 16:33:24 +0100 (BST) From: Robert Watson <rwatson@FreeBSD.org> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: svn-src-head@FreeBSD.org, svn-src-all@FreeBSD.org, src-committers@FreeBSD.org, Luigi Rizzo <rizzo@iet.unipi.it> Subject: Re: per-cpu counters (Re: svn commit: r190967 - head/sys/netinet) Message-ID: <alpine.BSF.2.00.0904121629320.19879@fledge.watson.org> In-Reply-To: <30873.1239549114@critter.freebsd.dk> References: <30873.1239549114@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 12 Apr 2009, Poul-Henning Kamp wrote: > In message <alpine.BSF.2.00.0904121525540.19879@fledge.watson.org>, Robert > Wats on writes: > >> I have a project along these lines in progress, and will post the proposal >> to arch@ once I've finished prototyping it. In particular, it provides >> common implementations of "reset" and "report" in order to expose a single >> userspace version of the structure via sysctl. > > Please don't export them via sysctl. > > Export them via mmap(2) like we already do with the disk I/O statistics. > > That way monitoring the counters becomes a no-syscall operation. The sysctls already exist, and are used by both built-in and third-party monitoring tools. Given the 8.0 timeline and the list of other things to get done before 8.0, I'm happy to solve the per-CPU counter problem as it is necessary to do so, but I don't have time to solve the mmap'd per-CPU counter problem, as it's not necessary to do so. In the current design the per-CPU counter framework allocates the storage for the counters, so it should be easy to add that feature later. Robert N M Watson Computer Laboratory University of Cambridge
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.BSF.2.00.0904121629320.19879>