2.4.10pre12aa1

[Andrea Arcangeli]

Again many thanks especially to Hugh, Andrew, Linus and Al for the help
with the recent fixes.

Only in 2.4.10pre12aa1: 00_GFP_ATOMIC-1

	Lower the gfp_atomic low watermark and make sure that __GFP_WAIT
	allocations won't eat the atomic GAP, in the attempt to reduce
	the GFP_ATOMIC allocation faliures.

Only in 2.4.10pre11aa1: 00_builtin_expect-1

	Dropped, likely/unlikely in mainline are nicer.

Only in 2.4.10pre12aa1: 00_compile-sysrq-1

	Fixed small compile troubles with sysrq (wakeup_bdflush() takes
	no arguments in mainline).

Only in 2.4.10pre11aa1: 00_copy-user-lat-4
Only in 2.4.10pre12aa1: 00_copy-user-lat-5

	Cleanup: use unlikely and __set_current_state.

Only in 2.4.10pre12aa1: 00_flush-inode-reschedule-1

	Reschedule during inode flushes under mem pressure.

Only in 2.4.10pre11aa1: 00_highmem-debug-1
Only in 2.4.10pre12aa1: 00_highmem-debug-2

	Rediffed due trivial rejects.

Only in 2.4.10pre11aa1: 00_max-threads-backout-1

	Mainline allows 1/8 of ram in tasks, that sounds ok too, so
	dropped the patch that made it 1/4 instead of 1/16.

Only in 2.4.10pre11aa1: 00_module-gfp-4
Only in 2.4.10pre12aa1: 00_module-gfp-5

	Rediffed due trivial rejects.

Only in 2.4.10pre12aa1: 00_netconsole-code-1
Only in 2.4.10pre12aa1: 00_netconsole-misc-1

	Splitted the netconsole code out of Ingo Molnar's latest Tux patch
	(J4). (from Ingo Molnar)

Only in 2.4.10pre11aa1: 00_numa-sched-8

	Rediffed due trivial rejects.

Only in 2.4.10pre12aa1: 00_ramdisk-secure-1

	Fix double BLKFLSBUF bug spotted by Al, and fixed a security problem
	with the pagecache backed ramdisk.

Only in 2.4.10pre12aa1: 00_ramdisk-secure-1-const-1

	Incremental cleanup, we may need to write to 'src' too.

Only in 2.4.10pre11aa1: 00_rwsem-19
Only in 2.4.10pre12aa1: 00_rwsem-fair-20

	Dropped the asm version of the rwsemaphores that was left disabled
	anyways in -aa and made all the fast paths inline to be competitive in
	the microbenchmark. Also made the rw sempahores fair again, see below.

Only in 2.4.10pre12aa1: 00_rwsem-fair-20-recursive-1

	Fixed the deadlock with recursive down_read, by introducing a
	down_read_recursive that gets a rw_sem_recursor structure as
	second paremeter that knows the state of the recursion. Such
	second parameter is of course allocated in the task structure,
	only page fault, coredump and /proc need to use the new recursive
	down. This way the deadlock is fixed cleanly and there is no
	unfariness.

Only in 2.4.10pre12aa1: 00_shrink_cache-BUG-1

	Fix from Andrew Morton to avoid freed buffers to be visible in the
	LRUs, it cures a BUG().

Only in 2.4.10pre12aa1: 00_swap-writepage-1

	Optimization resurrected by Hugh Dickins to avoid doing
	unnecessary I/O flushes on orphaned swapcache pages.
	The page->buffer case is fine too in his patch unlike what
	I commented in the last email.

Only in 2.4.10pre12aa1: 00_swapin-wrprotect-1

	Don't mark pte writeable during swapins unless VM_WRITE is set in the
	vma->vm_flags. Fix from Linus, spotted by Hugh Dickins.

Only in 2.4.10pre12aa1: 00_swapout-1

	Fix from Hugh Dickins to cure swapout.

Only in 2.4.10pre12aa1: 00_tty_register_ldisc-1

	Export a symbol needed by modules.

Only in 2.4.10pre11aa1: 00_vm-aa-2

	Merged in mainline.

Only in 2.4.10pre12aa1: 00_vm-writepage-reference-1

	Fixed race condition where page structures can be still modified after
	all the I/O is been started (so potentially the page could go away
	under those later references).

Only in 2.4.10pre12aa1: 10_numa-sched-9

	Rediffed due trivial rejects.

Only in 2.4.10pre12aa1: 56_uml-rb-mmap-1

	Make uml to compile again.

Only in 2.4.10pre11aa1: 60_tux-2.4.9-ac10-H7
Only in 2.4.10pre12aa1: 60_tux-2.4.9-ac10-J4

	Picked last update from www.redhat.com/~mingo/ .

Andrea