2.4.22pre6aa2

[Andrea Arcangeli <andrea@suse.de>]

URL:

	http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.22pre6aa2.gz

changelog diff between 2.4.22pre6aa1 and 2.4.22pre6aa2:

Only in 2.4.22pre6aa1: 00_extraversion-26
Only in 2.4.22pre6aa2: 00_extraversion-27
Only in 2.4.22pre6aa1: 9900_aio-21-ppc-1
Only in 2.4.22pre6aa2: 9903_aio-22-ppc-1
Only in 2.4.22pre6aa1: 9999_sched_yield_scale-5
Only in 2.4.22pre6aa2: 9999_sched_yield_scale-6

	Rediffed.

Only in 2.4.22pre6aa1: 05_vm_23_per-cpu-pages-2
Only in 2.4.22pre6aa2: 05_vm_23_per-cpu-pages-3

	Minor cleanup.

Only in 2.4.22pre6aa1: 9900_aio-21.gz
Only in 2.4.22pre6aa2: 9900_aio-22.gz

	Add KM_SOFTIRQ0/1 so crypto can compile.

Only in 2.4.22pre6aa2: 9902_aio-poll-1

	Added the AIO polling functionality. NOTE: the ABI
	is IOCB_CMD_POLL = 5. IMHO rather than moving poll
	into aio, aio should be moved on top of epoll. using poll should always
	be avoided, but apps already uses aio poll, so at least
	for the short term it'll be useful. the better approch I suggested is
	to hook the iocb somehow into an epoll fd. that epoll
	feature is missing right now AFIK, but it shouldn't be hard to add and
	it will give persistence to the polling too. The efficient point to
	sleep is epoll, not io_getevents/poll (io_getevents is optimal only
	for waiting for I/O [be it storage or network or pipe I/O of course]).

Only in 2.4.22pre6aa1: 9999900_desktop-2
Only in 2.4.22pre6aa2: 9999900_desktop-3

	Convert max-timeslice/min-timeslice to usecs units
	(the day we'll be interested to schedule more frequently than 1 time
	per usec, I guess we'll be very happy to break this interface with
	userspace ;).

Only in 2.4.22pre6aa1: 9999900_ecc-20020904-1.gz
Only in 2.4.22pre6aa2: 9999900_ecc-20020904-2.gz

	Merged more recent updates from Chip.

Andrea