This should apply to 2.5.59 too.  It seems to get rid of some pesky
hangs, on at least some hardware, but I won't have time to test it
on either VIA version ... maybe someone else will make the time?  :)

      New QH state prevents a re-activation race
	- nobody can un-halt a qh before its cleanup is done
	- resubmit-from-completion had this race (some usbtest cases)
	  as could some normal submit paths on busy endpoints (storage)
	- faster controllers would trip on this more consistently

      Queues of qtds
	- work harder to avoid ever modifing any qh in software
	- short reads block queue advance much less often
	- be more cautious with large (>~19KB) unaligned buffers

      Unlinking urbs
	- if qtd unlinked is at queue head, use its latest status
	  (main effect is reporting bytes from partial transfers)
	- another new qh state:  defer qh unlink if IAA is busy
	  (eliminates a busy-wait loop in a rare scenario)

      Enable features to improve bus utilization
	- PCI MWI ... can produce better write throughput; and by
	  using right cacheline size, sometimes read throughput too
	- USB NAK throttle ... sometimes reduces PCI access rates

      Other
	- async dump shows more funky qh+qtd states, and NAK count
	- cope with with some of the sprintf wierdness
	- periodic dump is usually smaller (so is that schedule)
      	- minor cleanups

Normally the CDC ACM devices have an subclass of 0, and the ACM subclass is
only applied to their first interface. But some have the subclass set on
the device itself, namely Motorola mobile phones. This patch takes those
devices into account.

The support for non-dma HCDs is likely the most interesting bit here.

    - makes dma calls behave sensibly when used with host controllers
      that don't use dma (including sl811).  usb_buffer_map() is a nop
      while scatterlist dma mappings fail (as they must).

    - make usb_sg_init() behave sensibly when used with non-dma hcs.
      the urbs are initted with transfer_buffer, not transfer_dma.
      this is the higher level analogue to usb_buffer_map(), so it
      needs to succeed unless there's a Real Error (tm).

    - moves two compatibility inlines from ehci.h into hcd.h so
      it'll be more practical to have the other hcds work in other
      environments (notably lk 2.4) too

    - remove URB_TIMEOUT_KILLED flag ... no device driver tests it;
      hcds don't really (uhci sets it, never reads it; sl811 doesn't
      enable the path that might set it), and it's not well defined.
      if any hcd needs such state, keep it in hc-private storage.

    - in usb_sg_wait(), use yield() instead of schedule() to let
      other activities free resources needed to continue.  (This
      was noted recently by Oliver.)

This is a minor cleanup to let per-request memory allocations block,
when the caller allows (it provided the bitmask).  The driver used
to work that way until something like 2.4.3; an update (a few months
back) to how the "dma_addr_t" hashes to a "struct ohci_td *" lets us
simplify things again.  Another benfit:  it blocks irqs for less time
on the submit path.  (The ehci driver already acts this way.)

  Some ethernet drivers other than those in .../drivers/net need generic
  MII code too and this cset shows how we do it for .../drivers/usb/net;
  For now only pegasus.c is using this feature, but as soon as we find
  more MII compliant controllers we'll put them in Makefile.mii too.
  Note: drivers which use the generic mii routines should bracket the
  code with #ifdef CONFIG_MII #endif since CONFIG_MII may not be present.
  See pegasus.c for more details.

This patch adds vendor/product ids for Artec, Canon, Compaq, Epson,
HP, and Microtek scanners. Further more, the device list was cleaned
up, sorted and duplicated entries have been removed.

  speedtouch: simplify the receive urb lifecycle: allocate them in the usb probe function,
  free them on disconnect.

  speedtouch: turn a micro race into a nano race.  The race is that an ATM device can
  be used the moment atm_dev_register returns, but you only get to fill out the
  atm_dev structure after atm_dev_register returns (this is a design flaw in the
  ATM layer).  Thus there is a small window during which you can be called with an
  incompletely set up data structure.  Workaround this by causing all ATM callbacks
  to fail if the dev_data field has not been set.  There is still a nano race if
  writing/reading the dev_data field is not atomic.  Is it atomic on all architectures?

  speedtouch: rework udsl_usb_probe error handling (for example, handle failure of
  atm_dev_register).  Do some trivial cleaning up while we're at it.

  speedtouch: roll udsl_atm_startdevice into udsl_usb_probe.

  speedtouch: use an array for rcvbufs rather than a pointer and dynamic allocation.

  speedtouch: roll udsl_atm_set_mac into udsl_usb_probe.

  speedtouch: trivial whitespace and debug message changes.

Added ids for the Dell Axim and Toshiba E740.

Thanks to Ian Molton and B.I.

  speedtouch: restore use of MODULE_DEVICE_TABLE to export usb info.  There may have
  been a problem with older 2.4 kernels, but there is none now.

Someone changed the "get string" logic to use short reads,
not long ones, a while back.  That broke many root hub
string accesses (not through tools like "lsusb"!) because
that logic didn't handle short reads quite right.

into kroah.com:/home/greg/linux/BK/gregkh-2.5

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

__virt_to_bus/__bus_to_virt depended on INTEGRATOR_HDR0_SDRAM_BASE
Unfortunately, this is defined in arch-integrator/platform.h, and
we really don't want to include it in memory.h.

We instead use BUS_OFFSET, which will eventually depend on the CPU
number in the system.

Only default BLK_DEV_IDEDMA on BLK_DEV_IDEDMA_ICS if ARCH_ACORN is
set, not if ARM is set.  There are PCI ARM systems out there!

Ensure that we clean up properly after initialisation error,
releasing all claimed resources in an orderly manner and
returning the correct error code.

Add cfbfillrect / cfbcopyarea / cfbimgblt objects for SA1100fb.
Remove redundant "pm" member.

Found a buglet in the marvel code -- doesn't change the number of IRQS
just the logic to get there.. This applies on top of the other marvel
code.

/jeff

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

to header files where they belong.

of AGP and SRMCONS patches.

From Jeff.Wiedemeier@hp.com.

(Titan / Marvel), Kconfig and headers.

From Jeff Wiedemeier.

Patch from Erich Focht

This adds a hook to rebalance globally across nodes every NODE_BALANCE_RATE
iterations of the rebalancer. This allows us to easily tune on an architecture
specific basis how often we wish to rebalance - machines with higher NUMA
ratios (more expensive off-node access) will want to do this less often.
It's currently set to 100 for NUMA-Q and 10 for other machines. If the
imbalance between nodes is > 125%, we'll rebalance them. The hook for this
is added to the NUMA definition of cpus_to_balance, so again, no impact
on non-NUMA machines.

Patch from Michael Hohnbaum

This adds a hook, sched_balance_exec(), to the exec code, to make it
place the exec'ed task on the least loaded queue. We have less state
to move at exec time than fork time, so this is the cheapest point
to cross-node migrate. Experience in Dynix/PTX and testing on Linux
has confirmed that this is the cheapest time to move tasks between nodes.

It also macro-wraps changes to nr_running, to allow us to keep track of
per-node nr_running as well. Again, no impact on non-NUMA machines.

Patch from Martin J. Bligh

This adds a small hook to the find_busiest_queue routine to allow us to
specify a mask of which CPUs to search over. In the NUMA case, it will
only balance inside the node (much cheaper to search, and stops tasks
from bouncing across nodes, which is very costly). The cpus_to_balance
routine is conditionally defined to ensure no impact to non-NUMA machines.

This is a tiny NUMA scheduler, but it needs the assistance of the second
and third patches in order to spread tasks across nodes.

Another left-over from ancient module code, it was supposed to return
non-zero if the module has a use count, but currently it always
evaluates to 0.

There are a few users of different types:
(1)  ioctl that perform a while(MOD_IN_USE) MOD_DEC_USE_COUNT loop.
     Just rip them out, we now have forced module unloading.
(2)  printk's that moan if the use-count in not zero in the exitfunc.
     Just rip them out, this can't happen.
(3)  if(MOD_IN_USE) MOD_DEC_USE_COUNT constructs in ->close of a few
     serial drivers.  Just remove the conditional, we did a
     MOD_INC_USE_COUNT in ->open.
(4)  This one is interesting:  drivers/sbus/char/display7seg.c uses
     the module use count to track openers.  Replace this with an
     atomic_t.

In addition remove tons of stale comments in network driver that aren't
understandable for anyone who doesn't know ancient Linux module semantics.