After compiling openmosix-sources, on a dual processor p3 and adding the kernel to grub, booting the kernel fails when init trys to start up /dev and /proc. Yes I did compile in /dev and /proc support as well as all other required kernel options. The same configuration works just fine when I downloaded 2.4.26 directly from kernel.org and then the openmosix patch from their website. The patch is intended for 2.4.26 so I think its just not playing well with 2.4.30. Reproducible: Always Steps to Reproduce: 1. emerge openmosix sources 2. configure and compile the kernel 3. add the kernel to grub 4. reboot into the new kernel
please, attache your kernel config
please, attach your kernel config
Created attachment 59434 [details] openmosix config file I am having the same problem trying to install openmosix-sources on two machines. Perhaps my config file will help you.
Created attachment 59464 [details] This is my config file for open-mosix sources, which makes a kernel which does not boot properly.
(In reply to comment #1) > please, attache your kernel config Sory it took so long I had finals week this last week.
are you able to boot gentoo-sources-2.4 with that configs?
(In reply to comment #6) > are you able to boot gentoo-sources-2.4 with that configs? gentoo-sources-2.4 is where it came from I just copied it over and added the open mosix options.
I have a very similar problem: The 2.4.30-r2 openmosix sources can not do nfsroot boot where they should be able to. The /sbin/rc script chokes on the check for /dev/null and the check for /proc. On logging in with sulogin at this point then /dev/null and /proc are there but the shell behaves oddly with the PATH not really working. This is on AMD Sempron but with a 386 kernel. I've tried lots of different .config setup but can't make it work. vanilla 2.4.26 + 2.4.26-1 worked correctly at this point.
Also there is something horible in fs/stat.c:sys_stat and sys_newstat asmlinkage long sys_stat(char * filename, struct __old_kernel_stat * statbuf) { struct nameidata nd; int error; if (!error) { error = do_revalidate(nd.dentry); if (!error) error = cp_old_stat(nd.dentry->d_inode, statbuf); path_release(&nd); } return error; }
Created attachment 59769 [details] The files open.c and stat.c need to be un - tarred into fs These changes made it boot for me. I think the problem is with the dfsa code. I wouldn't use my code directly since I am not a real kernel hacker.
I am having the same problem trying to install openmosix-sources 2.4.30. I'm trying to use the Andrew 'patch', but, like he said, I'm unable to have the console login at the end of the boot but I'm able only to connect to the node via ssh. In a production cluster this is not the best way to solve the problem.
fixed in .30-r3 thanks, everyone. and sorry for such delay (due to lack of x86 h/w).
I have to reopen the bug because the new openmosix-sources-2.4.30-r3 crash everytime I try to use openmosixtest and other stupid test written in python to stress the cluster.
did you recompile openmosix-user? what's in dmesg?
Yes, I recompiled openmosix-user. In dmesg I have the kernel stack.
so please, post it here as ksymoops output ksymoops < stackTraceFromDmesg.txt
Created attachment 60171 [details] ksymoops output I did ksymoops < /var/log/kernel
i can't read that binary file, unfortunately you should put one stack trace (or oops) in separate file and then pass it thru ksymoops
Created attachment 60174 [details] ksymoops output The last attachment was gzipped. The new one is text plain. I try to obtain a new stack because during the test the kernel crashes and I have to reset the computer manually.
thank you, that's enough information for me. i'll try to solve it within this week.
