we never block on locks in the common case. I'd like to know
if this makes a difference to Oracle users. It is incremental
with lvm beta7 (that was just previously included into my tree)."
Date: Thu, 12 Jul 2001 10:16:35 +0200 From: Andrea Arcangeli Subject: 2.4.7pre6aa1 Diff between 2.4.7pre5aa1 and 2.4.7pre6aa1: Only in 2.4.7pre5aa1: 00_bh-async-2 Only in 2.4.7pre6aa1: 00_bh-async-3 Rediffed again due trivial rejects. Only in 2.4.7pre5aa1: 00_drop___unlock_buffer-1 Only in 2.4.7pre5aa1: 00_drop_end_buffer_write-1 Only in 2.4.7pre5aa1: 00_kiobuf-backout-get_bh-1 Only in 2.4.7pre5aa1: 00_linus-brelse-fix-1 Merged in mainline. Only in 2.4.7pre6aa1: 00_iput-debug-1 Minor debugging check. Only in 2.4.7pre6aa1: 00_lvm-0.9.1_beta7-4_rwsem-fast-path-1 Converted the semaphores in the lvm very fast path to rwsemaphores, so we never block on locks in the common case. I'd like to know if this makes a difference to Oracle users. It is incremental with lvm beta7 (that was just previously included into my tree). Only in 2.4.7pre5aa1: 00_rwsem-14 Only in 2.4.7pre6aa1: 00_rwsem-15 Temporarily turned off alpha optimizations because they don't fit into this framework, will turn them on ASAP. Only in 2.4.7pre5aa1: 00_softirq-fixes-4 Only in 2.4.7pre6aa1: 00_softirq-fixes-5 Dropped the definition of smp_mb__ for the atomic_t operations that are now in mainline (left the other parts). Only in 2.4.7pre6aa1: 40_blkdev-pagecache-5 Now fixed also initrd, and tested that reads and writes with part of the page beyond of the end of the device works (assuming userspace knows where the device ends without relying on the last readable/writeable byte, kernel doesn't destabilize if you write and read beyond the end though). btw, I also made a port of the blkdev in pagecache rev 5 against 2.4.7pre6+o_direct-10. So to test blkdev in pagecache you can also apply in order: ftp://ftp.us.kernel.org/pub/linux/kernel/people/andrea/patches/v2.4/2.4.7pre5/o_direct-10 ftp://ftp.us.kernel.org/pub/linux/kernel/people/andrea/patches/v2.4/2.4.7pre6/blkdev-pagecache-5 on top of 2.4.7pre6. Andrea