I had 2.6.29.4 gentoo kernel. I wanted to try 2.6.30 kernel. I installed gentoo-sources-2.6.30.ebuild (it is 2.6.30 kernel + fbcondecor patch) and compiled bzImage, compiled modules, installed modules. I copied bzImage into /boot dir. I have grub installed, boot partition is jfs. I added the entry into grub.conf, restarted system. The system is loaded ok. I changed the kernel config and recompiled the kernel. Next the same: I copied bzImage into /boot dir, restarted system. Grub was not able to load kernel. No any error messages, pc was restarted automatically. bzImage was copied ok I checked it. I booted with gentoo-livecd, mount my boot partition and copied bzImage again, restarted the system. The system is loaded ok. So I think the problem in new jfs code (The problem in the copying of bzImage under 2.6.30 kernel).
I saw the same problem on German forum for Arch Linux. So I think Gentoo is not reason.
Strange problem. I can't imagine any of the changes between 2.6.29 and 2.6.30 making any difference to how jfs will lay out a file on disk. Since you checked the copied bzImage, I'm guessing the problem is likely to be in grub, although I'm curious what what might have triggered it. Can you point me to the German Arch Linux forum? I'd like to see how similar the details are. Thanks, Shaggy
http://sidux.com/PNphpBB2-viewtopic-t-16583.html :) but it is for Sidux not Arch.
Created attachment 21943 [details] jfs: fix regression preventing coalescing of extents A change I made in 2.6.30 introduced a problem where each extent was created with a length of one block. I guess grub can't handle a file with too many extents. This one-line patch fixes the problem. Thanks for reporting this!