Date: Tue, 2 Aug 2005 10:28:55 -0400 (EDT) From: Daniel Eischen <deischen@freebsd.org> To: Poul-Henning Kamp <phk@haven.freebsd.dk> Cc: current@freebsd.org Subject: Re: pthreads: shouldn't nanosleep() be a cancellation point ? Message-ID: <Pine.GSO.4.43.0508021021450.5408-100000@sea.ntplx.net> In-Reply-To: <25679.1122992242@phk.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 2 Aug 2005, Poul-Henning Kamp wrote: > In message <Pine.GSO.4.43.0508021007350.5408-100000@sea.ntplx.net>, Daniel Eisc > hen writes: > > >Hmm, the same could be said for sleep() in libc also, but we jump > >through hoops to allow the thread libraries override sleep() with > >their own cancellable version. I think this is in case libc wants > >to use sleep(), usleep(), nanosleep() internally and not introduce > >cancellation points into functions that shouldn't have them. > > usleep() calls _nanosleep() but I wonder if it shouldn't be > redirected into the thead libraraies like sleep/nanosleep ? Yes, I think we should just add a cancellable version of usleep() to libpthread/thread/thr_sleep.c. syslog() is a consumer of usleep(), but I think it really wants to use _usleep() or _nanosleep(). By using usleep() it introduces a cancellation point while holding a lock. Do you want to do the mods to libpthread and libthr (in libthr/thread/thr_syscalls.c) or do you want me to do them? -- DE
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.43.0508021021450.5408-100000>