From:  Hanna Linder <hannal@us.ibm.com>

From:  ookhoi@humilis.net

From:  Uros Bizjak <uros@kss-loka.si>

From: Adrian Bunk <bunk@fs.tum.de>

into home.transmeta.com:/home/torvalds/v2.5/linux

into home.transmeta.com:/home/torvalds/v2.5/linux

into raven.il.steeleye.com:/home/jejb/BK/scsi-for-linus-2.5

As people are using scsi_debug to simulate a large
number of disks to play with the dev_t expansion,
some problems have come to light.

Changelog:
   - change num_devs to num_tgts
       so now the user can independently modify the
       number of hosts (add_host), the number of targets
       per host (num_tgts) and the number of luns per
       target (max_luns)
   - num_tgts sets scsi_host::max_id after allowance is
     made for the initiator's target id
   - max_luns default changed to 1 and it is sysfs
     writeable
   - REPORT LUN response cleaned up. Could now possibly
     generate a response indicatimg there were up to
     16,384 logical units available
   - clean up sense buffer generation

Updated http://www.torque.net/sg/sdebug25.html

depending on whether the constant is a hexadecimal one as
opposed to a decimal one? 

Let's make it all explicit. There are probably more lurking
around, these were found during development of my C checker tool.

The queueing logic in scsi_lib looks really odd right now.

- Defers calling elv_next_request() until we actually think we can queue
  something. Always want to do that, request will have been marked
  REQ_STARTED after this, so block layer cannot touch it or merge to it.

- Kill the queue empty check. If elv_next_request() returned a req,
  there's one to queue.

Self explanatory, as per the discussion last week.

into home.transmeta.com:/home/torvalds/v2.5/linux

into nuts.ninka.net:/home/davem/src/BK/sparc-2.5

into nuts.ninka.net:/home/davem/src/BK/net-2.5

into home.transmeta.com:/home/torvalds/v2.5/linux

The SA1100 PCMCIA structure didn't lend itself well to the device
model.  With this reorganisation, we end up with a reasonable
structure which fits better with the driver model.  It is now
obvious that SA11x0-based socket drivers are separate from
SA1111-based socket drivers, and are treated as two separate drivers
by the driver model.

Trying to read the overlong target of a /proc/*/fd/N file descriptor
leads to a SIGSEGV inside the kernel, because the code doesn't check
for an error pointer return from d_path.

Bug 516.

Use schedule_delayed_work instead of a timer should fix this. Thanks
to Andrew Morton and Russell King.

(Added flush_scheduled_work() to ensure our delayed work completes
before we free the pcmcia_bus_socket structure. --rmk)

This adds support for i830 interrupt handling, and new improved
lock context keying. See per-file comments for more detail, as this
commit sadly mixes up a few different things (that's what you get
for not tracking the changes at a fine enough granularity).

into home.transmeta.com:/home/torvalds/v2.5/linux

into home.transmeta.com:/home/torvalds/v2.5/linux

Contributed by Andi Kleen

into home.transmeta.com:/home/torvalds/v2.5/linux

	o [CORRECT] Fix module refcounting (MOD_INC/DEC => .owner)
	o [FEATURE] Add hints to discovery (control channel)

	o [FEATURE] Make IrDA timers use mod_timer instead of
	add+del_timer

	o [FEATURE] Enable ZeroCopy Rx in irtty-sir/sir-dev
		(provided by the new SIR wrapper in 2.5.61).

	o [CORRECT] Don't do usb_clear_halt() on USB control pipe
	o [FEATURE] Cleanup and simplify the USB Rx path

	o [FEATURE] Fix the dynamic window code to properly send the pf bit.
		Increase perf by 40% for large packets at SIR.

	o [CRITICA] Fix remaining locking problem with discovery log
	o [CRITICA] Don't call expiry callback under spinlock
	o [FEATURE] Simplify/cleanup/optimise discovery/expiry code