Several functions in the serial drivers can be called from bottom
half or interrupt context. They must use the GFP_ATOMIC flag for
calls to kmalloc() and usb_submit_urb().
  
Functions which must use GFP_ATOMIC:
1. All *_callback() functions.
2. Any code which is inside a spinlock.
3. write(), throttle(), unthrottle(), which may be called by
   the line discipline in bottom half context.
  
Functions which can use GFP_KERNEL:
1. open(), close(), startup(), shutdown(), set_termios().

  
Here is a patch to add support for F5U011 to catc.c driver. The
patch has been compile tested against 2.5.6 and 2.5.7pre1
(and tested against 2.5.5-dj1) and should apply cleanly.

  I've written a patch Vojtech and I discussed for enhancing the
hiddev code to optionally provide more detailed output on read().
The old functionality is still supported by default, and in
situations where HID usage codes are unique across reports, the
old method is still preferable due to its terseness.
    
  The new method provides the ability to determine exactly which
value has changed, in cases where the HID usage codes are not  
unique.  It also provides a means to optionally receive notification
when input reports are received from the device, whether or not
any of the values in the report have changed.

  The details of the changes are as follows:
  
  - All current code behaves identically

  - A new ioctl pair HIDIOCGFLAG/HIDIOCSFLAG gets and clears
    flags on the hiddev device.                             

  - If you set the flag HIDDEV_FLAG_UREF, the read() call switches
    from reading hiddev_event structures to hiddev_usage_ref
    structures.  The change takes effect immediately, even to
    already queued events that haven't been read() yet.  Here's
    an example of enabling FLAG_UREF:                          

    {
        int flag = HIDDEV_FLAG_UREF;
        if (ioctl(fd, HIDIOCSFLAG, &flag) != 0) {
                perror("ioctl");
                exit(1);
        }
    }
  
  - With the HIDDEV_FLAG_REPORT set (which is only allowed if
    HIDDEV_FLAG_UREF is also set), there is a special uref that
    will be read() in addition to the ones corresponding to
    changes in the device state: when uref.field_index is set to
    HID_FIELD_INDEX_NONE, this uref is a notification that the
    report referred to by report_type and report_id has been
    received from the device.  This can be useful in situations
    when the notification of the arrival of a report is useful
    even if there is no change in state.

- bind to 7/1/2 alternate setting by default, to fix printing with HP
  LaserJet 1200 and 2200
- ioctls needed by the GPL user-mode IEEE 1284.4 driver which is part of
  the HP OfficeJet Linux driver (http://hpoj.sourceforge.net):
  - dynamic switching between 7/1/[123] alternate settings
  - sending HP vendor-specific channel-change-request to support
    memory card readers on HP PhotoSmart printers
  - inquire more information about the peripheral, including
    /proc/bus/usb/xx/yy linkage to get even more information
- fix apparent array overflow (by 1 entry) in usblp_probe when more than
  the maximum number of USB printers are connected
- for the 2.2 version, added MODULE_{INC,DEC}_USE_COUNT to prevent rmmoding
  of printer.o (and subsequent OOPSes) while a USB printer device is open
- cleaned up the code in a few places by consolidating duplicated code

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

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

make driverfs initialize early, so that ACPI can come alive
in a world where you can register devices. 

into tp1.ruhr-uni-bochum.de:/home/kai/kernel/v2.5/linux-2.5.config_sound.x

For some reason the ACPI people continue to make the mistake
of thinking that they are the root of the system.  Disabuse
them of that notion.

Clean up BSD accounting locking code..

The appended patch brings the fixes applied in 2.4 to shmem.c to 2.5.

In Detail:
- Add needed checks for shmem_file_write and shmem_symlink
- Add Documentation/filesystems/tmpfs.txt and adjust Config.help
- Add uid and gid mount options
- Make the error messages more user friendly

Pick up binfmt_elf.c SET_PERSONALITY() fix from 2.4.18.

That preempt_count really is supposed to be unconditional,
architectures please take note and add to your thread info.

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

If you could only review one of the 9 patches, this would be the one.

- removes acpitable.c vestiges
- adds ACPI IRQ routing support to PCI (disableable via pci=noacpi option)
- adds code to get a <1MB page for sleep, and ACPI boot to setup.c
- allocates another page in the fixmap for ACPI
- changes driverfs a little to work better with ACPI.

This patch removes arch/i386/kernel/acpitable.c. As mentioned previously,
the new ACPI code integrates this, so it's no longer needed.

This updates the Configure.help, both in arch/i386, and in drivers/acpi.

This removes the old OSPM code. It lived under drivers/acpi/ospm/*, but
the new code just lives in drivers/acpi, and removes some unnecessary
abstraction that this old code had.

This is the update to the core interpreter code.

This is the config.in and makefile changes for the latest code. The most
(only) interesting thing probably is ACPI is no longer flagged experimental.

This patch updates the acpi IA32 arch-specific files. Part of this is
taking what was acpitable.c and implementing it with better integration
with the rest of the ACPI code.

This patch adds in the new drivers.

- Support for driverfs
- File/code layout more in the Linux style
- improvements to battery, processor, and thermal support

This is the first of 9 patches. We did a complete rewrite of the
Linux-specific code, so we wait for things to stabilize before submitting.
There will be more updates, but *much* smaller.

#1 - this updates the header file.

The patch below eliminates a case of boot_cpu_data corruption
on SMP x86 machines. This was first observed on SMP Athlons,
but it also affects SMP Intel boxes in a less serious way.

When the secondary processors boot and execute head.S:checkCPUtype,
the code performs a 32-bit write of a small constant to the
byte-sized variable boot_cpu_data.x86 (X86 in head.S). Since the
write is 32-bit, it also writes zeros to the following 3 bytes,
which clobbers the x86_vendor, x86_model, and x86_mask fields
previously set up by check_bugs()'s call to identify_cpu().
Thus, after smp_init(), boot_cpu_data will _always_ identify
the CPU as an Intel (X86_VENDOR_INTEL == 0 in processor.h) with
model 0 and stepping 0.

The effect in standard kernels is not catastrophic, since:
(a) most SMP x86 boxes are Intel
(b) most uses of x86_vendor occur before smp_init() or reference
    the SMP cpu_data[] array
(c) most post-boot references to boot_cpu_data occur in the
    cpu_has_XXX macros which only read the x86_capability[] array
However, third-party extensions (like my x86 performance-monitoring
conters driver) can get seriously confused by this mis-identification.

Don't print out FAT superblock warnings if the IO failed.

Seperates qnx4_sb_info from struct super_block.

Seperates smb_sb_info from struct super_block.

Seperates msdos_sb_info from struct super_block for msdos and vfat.
Umsdos is terminally broken and is not included.

NR_IRQS should be 16 when the IO apic is not configured, as the 8259 PIC
cannot generate any more interrupts.  It also fixes a bug where the IDT
gets populated with random addresses, since only 16 entry stubs are
created. 

This changes the F00F bug workaround code to use the fixmap facilities
instead of touching the page tables directly.  It also removes the
assumption that only 686's don't have the bug.  I have confirmation that
the patch works on buggy pentiums. 

It needs the following patch