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

This backs out James' sysfs support for fbdev again.  It introduces a
big, race for every driver not converted to framebuffer_{alloc,release}
(that is every driver but Ben's new radeonfb).

I've left in framebuffer_{alloc,release} as stubs so drivers can be
converted to it gradually and once all drivers are done it can be
enabled again.

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

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

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

Theres still more to do here, but at least the ifdef mess is gone. No
more checking for NULL before calling functions, that was playing with
fire. Oh yeah and lots more deletions :)

Clean up the debugger hooks, it was way too easy to screw up.
And we did. And Linus hit it.

- create CONFIG_DEBUGGER so we can enable kernel debugging options but not
  have any trace of debugger gunk.
- remove a bunch of xmon prototypes so no one gets the urge to call them
- Use die() instead of panic in a number of places, it gives us much better
  debug information.
- Get rid of the ifdef madness

Heres a patch I've had for a while, it removes a bunch of debugger code
which is good :) The next patch will sanitise it (and the rest of the
debugger hooks).

Various xmon cleanups

- recover from bad SPR read/write (we get a program check)
- remove some old code (bat and segment register stuff)
- update the help text to match reality
- add a "press ? for help" when xmon first appears to make rusty happy
- protect against flushing bad parts of memory from Milton
- dont print iseries specific stuff on pseries in SPR dump (S)
- add code to dump the segment table or SLB
- remove a number of functions that wouldnt work on LPAR

- Add thread_info to pointer, its a useful piece of information.
- Do the kallsyms lookup on the link register
- Remove extra newline on one call to die()

From: Paul Mackerras <paulus@samba.org>

Recent changes in include/linux/*.h meant that likely()
isn't defined here (since we don't set __KERNEL__), and thus
we don't get some prototypes and we can't use do_div.  This
fixes the resulting compile errors and warnings.

Remove %L handling from sprintf - we don't need it, and it
meant we needed do_div from asm/div64.h, which gives problems
when __KERNEL__ isn't defined.  Also add a prototype for
strlen to kill a warning.

into kernel.bkbits.net:/home/davem/sctp-2.6

into kernel.bkbits.net:/home/davem/sparc-2.6

into kernel.bkbits.net:/home/davem/netfix-2.6

into nuts.davemloft.net:/disk1/BK/sctp-2.6

Also, un-inline large local_irq functions.

into intel.com:/home/lenb/src/linux-acpi-test-2.6.3

and we do _not_ want the broken MACZILOG serial driver.

This turned what _should_ have been a nice
easy oops into a nightmare of debugging.

Kernel debuggers suck.

I missed a kfree -> framebuffer_release() in the new radeonfb.

Thanks to Luca for noticing it.

Some drivers need clear_user_page, is should be exported

This avoids a link error if PPC_PMAC is set and the user forgets to set
ADB_PMU.  (The PMU driver is mandatory for pmac)

Also update copyrights for 2004.

From: Tom Rini <trini@kernel.crashing.org>

From: Randy Vinson <rvinson@mvista.com>

- Update the IBM Spruce defconfig

From: Matt Porter <mporter@kernel.crashing.org>

The following patch enhances and fixes a number of bugs in PPC44x MMU
handling.

From: Matt Porter <mporter@kernel.crashing.org>

The following patch removes some obsolete PPC4xx platforms that Paul and I
have discussed and some minor cleanup.

From: Tom Rini <trini@kernel.crashing.org>

On IBM 40x and IBM 4xx (or more specifically, all Book E processors), the
Save/Restor Registers 2 and 3 Critical Save and Restore Registers 0 and 1
are logically and functionally equivalent.  And since the 40x is the early
variant on the Book E model, make generic 4xx/BookE code refer to
CSRR0/CSRR1, and map these to SRR2/SRR3 on 40x.

From: Tom Rini <trini@kernel.crashing.org>

- Convert PPC_PREP to using the todc_time.c functions instead of its own
  copy of them.

From: Tom Rini <trini@kernel.crashing.org>

From: Randy Vinson <rvinson@mvista.com>

- Fix compilation of IBM Spruce when CONFIG_SERIAL_TEXT_DEBUG=n

From: Tom Rini <trini@kernel.crashing.org>

From: Randy Vinson <rvinson@mvista.com>

- Fixup IBM Spruce support (GEN550, general fixes and cleanups).
- Forward-port the INTERACTIVE_CONSOLE bits from 2.4.
- Forward-port the bootinfo code.
- Add a weak get_mem_size() function.

From: Nikita Danilov <Nikita@Namesys.COM>

Some stage in reiserfs balancing (fix_nodes() function) has to be performed
without ever scheduling.  If it schedules, it has to be restarted.  As we
don't want to restart often, we first try to do atomic allocation, and if
it fails, GFP_NOFS allocation is done, and fix_nodes() restarted.

From: Ulrich Drepper <drepper@redhat.com>

The len parameter for fadvise has a signed type and negative values passed
must be rejected.  The attached patch does the job.

From: Bjorn Helgaas <bjorn.helgaas@hp.com>

The current serial console code only works for ports that are either
defined in SERIAL_PORT_DFNS (and set up by serial8250_isa_init_ports()) or
registered by early_serial_setup().

On ia64, SERIAL_PORT_DFNS is empty because we discover everything via ACPI
and PCI.  And we only use early_serial_setup() for one port described by
the HCDP firmware table.

This patch against 2.6.3-rc2 makes it work for any valid port.  If we don't
know about the port early, we just return -ENODEV from the setup()
function, which leaves the serial console disabled.  After the driver has
found all the ports, we try to register the serial console again if it
hasn't been enabled already.

I think the "port->type == PORT_UNKNOWN" test is cleaner than the
"port->ops" test -- it more clearly gets to the point of "do we know about
this port".

From: Hugh Dickins <hugh@veritas.com>

Fix a screwup in the merging of swap extents: not only do they have to be
contiguous on-disk, they have to be logically contiguous in file offset too.