Date: Sun, 13 Jun 2004 13:07:28 -0500 From: Jon Noack <noackjr@alumni.rice.edu> To: =?ISO-8859-1?Q?Magnus_Carlebj=F6rk?= <magnus@carlebjork.se> Cc: luigi@iet.unipi.it Subject: Re: Device polling Message-ID: <40CC97E0.5010003@alumni.rice.edu> In-Reply-To: <Pine.BSF.4.58L0.0406110957020.49590@crimp.gluerecord.net> References: <Pine.BSF.4.58L0.0406110957020.49590@crimp.gluerecord.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 06/11/04 03:05, Magnus Carlebj=F6rk wrote: > Hi. I have been using Luigi's polling code extensively i several multip= rocessor > systems for over a year now, and it works flawlessly and efficiently. N= icely > done Luigi! >=20 > In the beginning of /usr/src/sys/kern/kern_poll.c there are a few lines= thas > says that polling is incompatible with SMP. This is not true, polling I= S > compatible with SMP, it just does not utilize all of it's features. I just tested this on my SMP all-in-one home server (Web, Mail, NFS,=20 Samba, Squid, etc.). It's been up for over 24 hours with no apparent=20 issues. The machine is used pretty heavily, with NFS mounted home=20 directories and CVS mirror (see below) -- a CVS update of the src tree=20 over NFS has done a good job of breaking fragile setups in the past.=20 Everything seemed OK. That said, peak performance (as tested by iperf)=20 took a nosedive: with 32-bit em adapters (gige), tcp bandwidth dropped=20 from over 360Mbps to around 200 Mbps. If anyone has any suggestions for=20 more in-depth testing, I'd be willing to try them. If I have the time I=20 may also try the latest netperf patch and see how that affects things. > I'm getting tired of having the builkernel break everytime because of t= his > define in kern_poll.c. Would it be possible to change that so that is j= ust puts > a note in the syslog instead? Create an update script that automatically applies a patch after the=20 CVSup. Alternatively, use CVSup to mirror the repository and then check=20 out the src directory (with regular CVS from the local mirror). That=20 way you can make local changes and still keep updated. I was introduced=20 to this latter method recently (on this list) and it works quite well.=20 See the CVSup page in the Handbook and=20 /usr/share/examples/cvsup/cvs-supfile. Jon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?40CC97E0.5010003>