Red Hat Bugzilla – Bug 26992
aacraid 2.2.17-14 2gb not bootable
Last modified: 2005-10-31 17:00:50 EST
From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 5.0) 2.2.17-14smp kernel rebuilt with following changes: 2GB max mem rather than 1 (as the machine in question has that much memory), add NFSv3 patches meant for 2.2.17. Installed the kernel on PowerEdge 4400 where filesystem is on raided drive (Perc 3/di, aacraid/percraid driver), built the initrd image. Would not use initrd image, ever (not even echo msgs in initrd image that should be displayed before attempted modprobe) and therefore kernel cannot proceed as it cannot access raided hard drives. The delay while decompressing initrd seems to be there but no actual use of files in initrd image. The same config used on a floppy again does not boot on the server but the same floppy will attempt inserting the module - and fail as other machines do not have perc raid card but module will show detection failure msgs. Reproducible: Always Steps to Reproduce: 1. Rebuild 2.2.17-14smp with NFSv3, 2GB (as rpm). 2. Install rpm (-ihv), build with mkinitrd (2.2.17-14smp... ) initrd image, correct lilo.conf, lilo. 3. Reboot and watch the failure occur on PowerEdge 4400 w/ 2GB, Perc 3/di, dual cpu. We had to revert back to 2.2.16 from Dell's support website (patched and rebuilt with NFSv3, 2GB mem and latest security patches) which I understand was prepared with RedHat's help. This kernel worked perfectly (version without latest security corrections had 187 day uptime - again, custom rebuilt by us to include NFSv3 and 2GB support - brought down to apply 2.2.17-14...). As this is production machine we didn't have time to try redhat's original 2.2.17-14 to see if 1GB to 2GB memory change had anything to do with the problem but I strongly suspect it does (NFSv3 patch should have nothing to do with this) or if even 2.2.17-14 has problems with this hardware config (note that the same floppy with appropriate config worked on different machines with less mem and no raid drive - raid drive shouldn't matter as it didn't reach inserting of module so then either memory or smp should be show stoppers). Oh yes, rebuilding done on uptodate redhat 6.2 machine.
Not relevant anymore - fixed it with custom 2.2.19-6.2.1 - again modified for 2GB memory (plus other fixes for NFS like smp fix) AND made sure this time BIGMEM is on as that was very very probably the problem last time (since only if BIGMEM is on relocate_the_initrd_before_kernel_overwrites_it fix is used even though the patch is always applied - uh those #ifdefine's...). New rpm's worked without any problems and machine has been up and running fine so far (2 days and going...) - including working aacraid (therefore initrd) and full 2gGB of mem.