Hi Andrew,
today I've 7 patches for you to consider. One is common code,
6 are s390 bug fixes. I've another one which adds experimental
cpu hotplug support for s390 which works on the -bk tree but
not one the -mm4 tree. I could track it down to two patches
in -mm4, if I remove sched-cleanup-improve-sched-fork-apis.patch
and sched-cleanup-init_idle.patch the hotplugging works on -mm4.
I'll have to investigate further what's wrong here.
The common code problem is in lib/radix-tree.c where an int
is shifted by more then 32 bits. We stumbled over this problem
with the efi partition code. After a dasd device is freshly
formatted it has 0 blocks. The efi partition code tries to load
the last block of the device which turns out to number -1. A
mapping for page at index (~0UL >> 3) related to a bdev inode
is established. Due to the bug in the radix tree code the page
isn't found when the bdev gets released and the nrpages counter
stays 1 -> "kernel BUG() at fs/inode.c:238!".
Have fun & blue skies,
Martin.