Re: [SSI-devel] Re: [Etherboot-developers] BUG: Large ramdisks break ELF image
Brought to you by:
brucewalker,
rogertsang
From: <ebi...@ln...> - 2003-11-28 08:09:17
|
Chirag Kantharia <chi...@hp...> writes: > On Thu, Nov 27, 2003 at 11:56:34AM -0700, Eric W. Biederman wrote: > | >From the output it looks like your initrd is being mounted > | ro only in the mkelfImage case. Mount fails since it > | can't write to /etc/mtab and the failures continue from there. > | > | That is the kernels default behavior so except for not adding > | a "rw" early on the command line I don't see a problem. > > That's right! (silly me). > > I've been able to boot secondary nodes in OpenSSI cluster, using > mkelfImage (but not with mkelf-linux). Does mkelfImage does something > extra, that mkelf-linux doesn't do? mkelfImage does things differently. mkelfImage is a little more aggressive in design. It enters the kernel at it's 32bit entry point. It does not run all of the kernel's 16bit entry point code, it makes it's own BIOS calls. mkelfImage hard codes the location of the ramdisk when the image is made instead of dynamically positioning it. In general mkelf-linux should be the more conservative design. And it should work on a wider class of PCs. So it would be interesting to find out how mkelf-linux fails. And now there is something to contrast it with. The best I can suggest for debugging is serial console from both test cases so we can contrast them. I don't know how mkelf-linux fails. You might want to look in the archives (both etherboot-developers and openssi) there are some suggestions of possible failure reasons earlier in this thread. I don't think either archive has fully captured the log. Eric |