[USB] pxa2xx_udc updates, mostly for non-Lubbock hardware

  - IXP 42x UDC support (Greg Weeks)

  - remove Lubbock-specific build assumption (Guennadi Liakhovetski)

  - handle D+ pullup right on iPaqs, e7xx, etc (HH.org)

  - don't unbind() with irqs blocked; matches other controller drivers,
    and network layer expectations

  - handle some deferred ep0 responses better

  - support iso transfers (needs fifo size tracking)

[USB] ethernet gadget, locking tweaks

This problem showed pretty quickly on an SMP system.  Basically,
access to the freelist (tx more than rx) needs a spinlock.

Stop repeating some alloc_etherdev() work.  Disable DEBUG messages.

USB_SCANNER is obsolete, and it's now marked as BROKEN.

I there a good reason to keep it in the kernel?

Following is a trivial patch to update the ohci-omap.c in 2.6.2 to be in
sync with the OMAP tree. Basically the IRQ name was changed, which keeps
the driver from compiling.

It also includes a cosmetic change to replace inl/outl with readl/writel.

Thanks to Terry Markovich <terry@avionictools.com> for the information.

if submission of the interrupt URb fails in belkin_sa_open() the
previously submitted URB may remain live. The rest of the system
thinks that opening failed, therefore the URB should be unlinked
in the error case.

this driver does DMA to the stack via usb_bulk_msg().

2.6 has no -fsigned-char anymore. there are 3 checks for info->lun < 0

Following patch against 2.6.2-rc2 removes a reference to what appears
to be a deprecated member of struct usb_hcd from the function
ohci-sa111.c:usb_hcd_sa1111_remove().

Here is a revised version of the irqreturn_t propagation patch.
The only difference is now ohci-hcd returns IRQ_HANDLED in the remove case.

Greg, I think this now makes 2 distinct bugs in the scanner kernel
driver.  Maybe it should be protected with a BROKEN:

There's an unnecessary pci-ism in usbcore -- just an oversight
in some API conversion, and harmless on most systems.  This'll
be a good thing to remove for non-PCI HCD support though.

> Data point:  "visor.c" always passes GFP_ATOMIC there.

Here's a patch that makes cdc-acm use GFP_ATOMIC too,
and report consequent allocation failures.  Compiles
but otherwise untested.

I have a patch for a typo in the 2.6.0 kernel, drivers/usb/core/hub.c,

in hub_clear_tt_buffer, the value USB_DIR_IN | USB_RECIP_OTHER is used
for bmRequestType, when section 11.24.2 of the spec says it should
really be USB_RT_PORT, or 00100011b.

After limited testing, this change exhibited better recovery from TT
errors, here's the patch :

It replaces a call to device_unregister_wait() in the File-backed
Storage Gadget.

(moving the one used function into the usb.c file.)

drivers/usb/storage/usb.c: In function `usb_stor_release_resources':
drivers/usb/storage/usb.c:837: error: invalid lvalue in assignment

drivers/usb/misc/uss720.c: In function `parport_uss720_epp_read_data':
drivers/usb/misc/uss720.c:336: error: invalid lvalue in increment
drivers/usb/misc/uss720.c: In function `parport_uss720_epp_read_addr':
drivers/usb/misc/uss720.c:395: error: invalid lvalue in increment
drivers/usb/misc/uss720.c: In function `parport_uss720_epp_write_addr':
drivers/usb/misc/uss720.c:415: error: invalid lvalue in increment
drivers/usb/misc/uss720.c: In function `parport_uss720_ecp_write_addr':
drivers/usb/misc/uss720.c:472: error: invalid lvalue in increment

drivers/usb/input/hid-core.c: In function `fetch_item':
drivers/usb/input/hid-core.c:605: error: invalid lvalue in increment
drivers/usb/input/hid-core.c:612: error: invalid lvalue in increment

drivers/usb/gadget/net2280.c: In function `write_fifo':
drivers/usb/gadget/net2280.c:527: error: `typeof' applied to a bit-field
drivers/usb/gadget/net2280.c: In function `handle_ep_small':
drivers/usb/gadget/net2280.c:2042: error: `typeof' applied to a bit-field

into cheetah.(none):/home/davem/src/BK/sparc-2.6

into home.osdl.org:/home/torvalds/v2.5/linux

From: Dominik Brodowski <linux@dominikbrodowski.de>

This brown paper bag patch is needed to assure cpufreq_update_policy works
correctly.

Please apply, else the next ACPI patch will cause trouble with thermal
management [it needs cpufreq_update_policy to work properly].

Fix a horribly wrong memcpy instruction in cpufreq_update_policy which
caused it to oops.

From: Torsten Duwe <duwe@suse.de>

pmdisk.c uses struct new_utsname, so give it the header.

From: Christoph Hellwig <hch@lst.de>

Put kernel_flag back to where it used to be, near its comment and its
EXPORT_SYMBOL.

From: Andi Kleen <ak@muc.de>

Just fix two warnings on x86-64 that were recently introduced (one by me
and the other by the sort extable changes)

  Add comment for HdrS ver 0x201

into hopper.(none):/usr/src/kernel/sparc-2.6

into home.osdl.org:/home/torvalds/v2.5/linux

  - Fix kconfig botch (bk automerge can be rather flakey)
  - The gadgetfs patches going with the kconfig cleanups

into home.osdl.org:/home/torvalds/v2.5/linux