udsl_atm_proc_read may be called after USB disconnect.

This makes "test 10" verify that completions are returned in-order,
resolving a FIXME.  OHCI and EHCI do, but UHCI doesn't. (*)  That
simplified a cleanup of the queue fault tests to make it easier
to handle optional faults (with one or more failure modes).

It also returns a "lost subcase" that accidentally was not getting
run.  And in a case of pure paranoia, the unlink tests handle the
EBUSY return from an async urb unlink ... if that ever shows up I'd
expect it to be on an SMP box.


(*) I'd suspected as much given the first round of tests with UHCI;
     the diagnostics from "usbtest" made no sense otherwise.  This
     is just repeatable confirmation of the earlier bug report.  This
     could cause trouble in the usb_sg_*() I/O calls.

     In this case, "testusb -at10 -g4" reported that subcase 1 completed
     out of order (before subcase 0) ... without looking at details, I'd
     guess a list_add() vs list_add_tail() issue.

     Then after trying the queue cleanup code, I got diagnostics from
     uhci_destroy_urb_priv; then a kmalloc poisoning oops in uhci_irq,
     or "uhci_remove_pending_qhs+0x7c/0x1b0" in more detail.  Those
     looks to be the same "can't unlink from completions" errors that
     was also reported before in that code.

     Note that "testusb -at10" (like "testusb -at9") can be made to
     work with any USB device, using "usbtest" module options.

This gets rid of some potentially scarey messages I've recently
seen on disconnect, "bad hash" for a TD ... it's really a "no hash"
case, and this prevents the message in some cases where that's
not an issue.  Likely this is what Gary Gorgen noticed, but even
he had a different problem, this message shouldn't always appear.

It also slims down one other message, preventing it from appearing
in the routine "protocol stall" case (and cluttering logfiles).

This is the rest of the work to make the driver not care which
version of the OS it's using, so the difference between the 2.5
and 2.5 versions can just be a small patch with stuff that has
a real need to be different.

The 2.5 version branched from 2.4.5 or so, and since then
a couple hardware-specific tweaks were merged to 2.4; this
teaches 2.5 about NatSemi SUPERIO and PA-RISC quirks.

This also uses os/version neutral HCD calls to register the root
hub and find the HCD's bus.  It also adds os/version neutral
macros for its diagnostic macros.  Most of those changes have
been split out separately, but the macros and a few uses of them
weren't naturally splittable.

Also a couple minor cleanups, like removing CVS ids, having
only one copy of a routine used with the debug files, and
getting rid of some inline #ifdefs.

This potentially helps debugging, since otherwise a double fault
would generate a triple fault and then reboot the machine. Now
instead it can print out a note about where the problem happened,
unless all the kernel data structures are truly buggered.

it's a noop in 2.5 now that the behvaiour of implicitly exporting
all symbols is gone.

This patch removes the devfs interface code in mtrr that has been
stubbed out by an ifdef forever.  It's one of the few remaining users
of regular files on devfs so there's some urge for me to get rid of it :)

Use cdevname instead in the serial driver printk cut & pasted into about
a dozend places.

small updates to bring us nearer the lm_Sensors CVS (docs, comments,
missing statics, named initializers).  I still need to review their
actual code changes before submitting those..

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

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

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

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

into kernel.bkbits.net:/home/davem/sparc-2.5

into kernel.bkbits.net:/home/davem/sparc-2.5

into kernel.bkbits.net:/home/davem/net-2.5

into kanga.twiddle.net:/home/rth/linux/axp-2.5

the number of metadata buffers. Use the same techniques as the
regular linux buffer cache here.

use more hash buckets for holding xfs metadata, and use the same
hash algorithm as the regular buffer cache.

SGI Modid: 2.5.x-xfs:slinx:139997a

SGI Modid: 2.5.x-xfs:slinx:139992a

SGI Modid: 2.5.x-xfs:slinx:139854a

this was always there, but exposed by the last change in
this area and made much more likely.

SGI Modid: 2.5.x-xfs:slinx:139655a

SGI Modid: 2.5.x-xfs:slinx:139330a

with sector sizes smaller than those the device supports.  Tripped a BUG
in pagebuf, should now be resolved.

SGI Modid: 2.5.x-xfs:slinx:139328a