On Fri, Jul 18, 2008 at 02:43:11AM +0200, Bernhard Walle wrote: > * Neil Horman [2008-08-14 08:18]: > > > > That being said, Bernhard, I suppose it would be worthwhile standardizing some > > configuration settings, seeing as despite our implementation differences, we > > seem to largely support the same feature set. Have you given any consideration > > to defining a standard kdump feature set and configuration syntax? > > I have nothing against consideration in that area. > > SUSE does configuration in /etc/sysconfig/kdump. I think sysconfig is > also on RedHat, so that is the same? > Its close, /etc/sysconfig/kdump is used to apply options to the kexec utility, while /etc/kdump.conf directs how the initrd should be constructed. I would like to move to using a url style syntax for specifying dump targets like you have, but there are series of other options that I need to take into consideration (see http://download.fedora.redhat.com/pub/fedora/linux/development/source/SRPMS/kexec-tools-1.102pre-15.fc10.src.rpm > > > [1] mkdumprd on SUSE is only a script that calls mkinitrd with right > > > parameters while the mkdumprd on RedHat builds a initramfs for > > > kdump itself with busybox > > I was looking at this, I assume that your initramfs then acts very simmilar to > > your default initramfs, which mounts the rootfs and runs init, and your kdump > > capture operations are then implemented in the kdump init script. Is that > > correct? > > For kdump, an own initramfs is built. That does the copy stuff in > initramfs, so that works without rootfs and reboots before init is > executed. > > So the 'copy' script is not the init script but ?init/boot-kdump.sh? > script if you ?hg clone http://freehg.org/u/bwalle/kdump/?. > Unfortunately, freehg has no tree view in its web interface. > Ahh, so you do all your dump manipulation with your kdumptool binary, I see now. I handle all of that with various shell scripts in the initramfs. > Having that said, for testing one must do: > > - install packages > SUSE: > o kexec-tools > o kdump > o makedumpfile > RH: ? > Same here. > - set crashkernel parameter > => can be done with GRUB configuration manually with editing > /boot/grub/menu.lst, should be the same on RH and SUSE > Agreed. > - enable kdump (that would be ?chkconfig kdump on?) > => should be the same on RH and SUSE > Agreed. > - tweak some configuration settings like the place where it should be > saved > => KDUMP_SAVEDIR in /etc/sysconfig/kdump > => KDUMP_DUMPLEVEL for makedumpfile in /etc/sysconfig/kdump > => KDUMP_DUMPFORMAT (?compressed? / ?ELF? / ??) > in /etc/sysconfig/kdump > Agreed. I think this is where we diverge though. We seem to support much of the same dump targets but our configuaration directives are different. For me, I would need to configure: <dump type> <type data> <core_collector> <cp|makedumpfile> [options] > - trigger kdump > => echo c > /proc/sysrq-trigger > > Most other stuff is ?nice to have? but not ?mandatory? I think? > Actually, you're right. Looking over it, I do have a set of additional options, but none are central to dump collection (link delay specification for spanning tree, kdump pre/post scripts, etc). > > > Bernhard > -- > Bernhard Walle, SUSE LINUX Products GmbH, Architecture Development > > ?Make everything as simple as possible, but not simpler.? > -- Albert Einstein > > _______________________________________________ > kexec mailing list > kexec at lists.infradead.org > http://lists.infradead.org/mailman/listinfo/kexec -- /*************************************************** *Neil Horman *Senior Software Engineer *Red Hat, Inc. *nhorman at redhat.com *gpg keyid: 1024D / 0x92A74FA1 *http://pgp.mit.edu ***************************************************/
