Date: Fri, 24 May 1996 09:44:09 +0200 From: Martin Birgmeier <Martin.Birgmeier@nt.tuwien.ac.at> To: freebsd-bugs@freebsd.org Subject: NFS_DIRBLKSIZ = 4096 breaks diskless operation (only?) Message-ID: <199605240744.AAA06512@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
Dear FreeBSD developers, first let me thank you for the fine product you have developed! I truly enjoy being able to use it. Now on to the problem: I have an old 1.1.5.1 machine, onto which I have loaded 2.2-960501-SNAP; I've also access to a 2.2-current machine (located elsewhere), where I have compiled a kernel for another machine which I want to boot from the 1.1.5.1 machine. This nearly works, except for one thing: Using tcpdump on the 1.1.5.1 machine, I see that readdir nfs requests get fragmented. This seems to be because the macro NFS_DIRBLKSIZ is set to 4096, and this is used as the minimum in mountnfs(). But the diskless machine uses an old 3c503 (etherlink II, that is?) network card, for which I have to set rsize=1024. I would like to set readdirsize to 1024 as well, but alas, as already said, 4096 is the minimum enforced in mountnfs(). The net result is that I can read large files (e.g., the kernel) fine, but as soon as the diskless machine has to do directory searches, it will hang with `nfs server ... not responding'. I am not a kernel expert, but I think the setting of NFS_DIRBLKSIZ, or enforcing it in nfsmount(), is the source of this problem. I'd be really happy if you could help me with this! Greetings, and thanks again, Martin --- Martin Birgmeier Department of Communications and Radio Frequency Engineering Vienna University of Technology Gusshausstrasse 25/E389 A-1040 Vienna Austria Martin.Birgmeier@nt.tuwien.ac.at
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199605240744.AAA06512>