Date: Mon, 30 Jun 2003 10:59:45 -0400 (EDT) From: Daniel Eischen <eischen@vigrid.com> To: Petri Helenius <pete@he.iki.fi> Cc: freebsd-threads@freebsd.org Subject: Re: rtprio and kse Message-ID: <Pine.GSO.4.10.10306301055050.7047-100000@pcnet5.pcnet.com> In-Reply-To: <019601c33f13$015daa30$44d5473e@PETEX31>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 30 Jun 2003, Petri Helenius wrote: > > > > Theoretically any process with rt priority can monopolise the CPU if it > > spins, regardless of whether it's threaded or not.. > > > > Is not that what having RT-priority means? > > Yes, but in non-threaded environment with multiple processes > competing for CPU libc is not expected to have locks. In threaded > environment it does have them around things like malloc. My concern > is the realtime thread spinlocking waiting for a lock which is held > somewhere which does not get scheduled. These locks are not > visible to the appliation and thus one must either put additional > mutexes around them or know that libc stuff is rtprio-thread-safe. Libpthread silently converts libc spinlocks into non-spinning locks; there should be no deadlock or starvation issues in userland. I don't know if libthr ans any issues with spinlocks or not. -- Dan Eischen
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.10.10306301055050.7047-100000>