Date: Mon, 19 Oct 2009 22:55:15 +0200 From: Jedrzej Kalinowski <kalinoj1@iem.pw.edu.pl> To: freebsd-current@freebsd.org Subject: Many tftp requests for each kernel and module file during DISKLESS boot Message-ID: <4ADCD233.1070302@iem.pw.edu.pl>
index | next in thread | raw e-mail
Hello, I work with a freebsd-current DISKLESS boot setup at my university and try to make the diskless boot as fast as possible. I was wondering why my tftp log of a machine boot process looks like this: ... Oct 19 20:12:14 bsd in.tftpd[45140]: RRQ from xxx filename /boot/loader.conf.local Oct 19 20:12:14 bsd in.tftpd[45141]: RRQ from xxx filename /boot/kernel/kernel Oct 19 20:12:14 bsd in.tftpd[45142]: RRQ from xxx filename /boot/kernel/kernel Oct 19 20:12:14 bsd in.tftpd[45143]: RRQ from xxx filename /boot/kernel/kernel ... There are three requests for a kernel file and even more requests for modules when they get loaded. I added some additional output in loader files to trace the requests and now I see (when loading a kernel) that the file is tried to get loaded in elfN(loadfile) functions found in source files /usr/src/boot/common/load_elf.c and /usr/src/boot/common/load_elf_obj.c with no success at first, than it is successfully loaded by load_elf.c. Each load attempt results in tftp request being issued to the server. My question is - is there a simple way of getting rid of these requests? Are there any plans to include some sort of local caching in loader? I think that the way it works now is not the most optimal one, because of redundant tftp RRQs. Regards, Jędrzej Kalinowskihelp
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4ADCD233.1070302>