I've seen this problem twice on armel. Both times, the box was running unstable, but still had the stable 2.6.32 kernel.
-rw-r--r-- 1 root root 7.1M Jan 28 21:36 initrd.img-2.6.32-5-iop32x -rw-r--r-- 1 root root 1.2M Nov 26 2010 vmlinuz-2.6.32-5-iop32x AFAICS, what happened here was the udev or another upgrade caused the initramfs to be rebuilt, with the new udev, and written to flash (this happened on a Thecus N2100 and a Sheevaplug). Of course, recovery was tricky for these embedded devices, but once I got into the system, upgrading the kernel to 3.2.0-1 fixed it. Ironically, I had kept the old kernel because I did not want to deal with any breakage from unstable's kernel. I wonder if perhaps udev should install a update-initramfs hook that refuses to let the initramfs be built if it's for this old kernel. Otherwise partial upgrades from stable will tend to encounter the problem. Looks like the ia64 kernel has been fixed to have the missing syscall. If there was an update-initramfs hook it could have the proper kernel version for ia64 too. -- see shy jo
signature.asc
Description: Digital signature
