Lots of changes that have accumulated over the last weeks.

This makes it compile and boot again, Lots of bug fixes, including
security fixes.  Several speedups.

Only changes x86-64 specific files.

 - Use private copy of siginfo.h (for si_band)
 - Align 32bit vsyscall coredump (from Roland McGrath)
 - First steps towards 64bit vsyscall coredump (not working yet)
 - Use in kernel trampoline for signals
 - Merge APIC pm update from Pavel/Mikael
 - Security fix for ioperm (from i386)
 - Reenable vsyscall dumping for 32bit coredumps
 - Fix bugs in 32bit coredump that could lead to oopses.
 - Fix 64bit vsyscalls
 - Revert change in pci-gart.c: pci_alloc_consistent must use an
   0xffffffff mask hardcoded.
 - Fix bug in noexec= option handling
 - Export fake_node
 - Cleanups from Pavel
 - Disable 32bit vsyscall coredump again.  Still has some problems.
 - Implement new noexec= and noexec32= options to give a wide choice
   of support for non executable mappings for 32bit and 64bit processes.
   The default is now to honor PROT_EXEC, but mark stack and heap
   PROT_EXEC.
 - 32bit emulation changes from Pavel: use compat_* types.
 - (2.4) Use physical address for GART register.
 - Convert debugreg array to individual members and clean up ptrace
   access.  This saves 16 byte per task.
 - (2.4) Use new streamlined context switch code.  This avoids a
   pipeline stall and pushes the register saving to C code.
 - Save flags register in context switch
 - Clean up SMP early bootup.  Remove some unnecessary code.
 - (2.4) Process numa= option early
 - (2.4) Merge 2.4 clear_page, copy_*_user, copy_page, memcpy, memset.
   These are much faster.  clear/copy_page don't force the new page out
   of memory now which should speed up user processes.  Also full
   workaround for errata #91.
 - Some cleanup in pageattr.c code.
 - Fix warning in i387.h
 - Fix wrong PAGE_KERNEL_LARGE define.  This fixes a security hole and
   makes AGP work again.
 - Fix wrong segment exception handler to not crash.
 - Fix incorrect swapgs handling in bad iret exception handling
 - Clean up some boot printks
 - Micro optimize exception handling preamble.
 - New reboot handling.  Supports warm reboot and BIOS reboot vector
   reboot now.
 - (2.4) Use MTRRs by default in vesafb
 - Fix bug in put_dirty_page: use correct page permissions for the stack
 - Fix type of si_band in asm-generic/siginfo.h to match POSIX/glibc
   (needs checking with other architecture maintainers)
 - (2.4) Define ARCH_HAS_NMI_WATCHDOG
 - Minor cleanup in calling.h
 - IOMMU tuning: only flush the GART TLB when the IOMMU aperture area
   allocation wraps.  Also don't clear entries until needed.  This
   should increase IO performance for IOMMU devices greatly.  Still a
   bit experimental, handle with care.
 - Unmap the IOMMU aperture from kernel mapping to prevent unwanted CPU
   prefetches.
 - Make IOMMU_LEAK_TRACE depend on IOMMU_DEBUG
 - Fix minor bug in pci_alloc_consistent - always check against the dma
   mask of the device, not 0xffffffff.
 - Remove streamining mapping delayed flush in IOMMU: not needed anymore
   and didn't work correctly in 2.5 anyways.
 - Fix the bad pte warnings caused by the SMP/APIC bootup.
 - Forward port 2.4 fix: ioperm was changing the wrong io ports in some
   cases.
 - Minor cleanups
 - Some cleanups in pageattr.c (still buggy)
 - Fix some bugs in the AGP driver.
 - Forward port from 2.4: mask all reserved bits in debug register in
   ptrace.  Previously gdb could crash the kernel by passing invalid
   values.
 - Security fix: make sure FPU is in a defined state after an
   FXSAVE/FXRSTOR exception occurred.
 - Eats keys on panic (works around a buggy KVM)
 - Make user.h user includeable.
 - Disable sign compare warnings for gcc 3.3-hammer
 - Use DSO for 32bit vsyscalls and dump it in core dumps.  Add dwarf2
   information for the vsyscalls.
   Thanks to Richard Henderson for helping me with the nasty parts of
   it.  I had to do some changes over his patch and it's currently only
   lightly tested.  Handle with care.  This only affects 32bit programs
   that use a glibc 3.2 with sysenter support.
 - Security fixes for the 32bit ioctl handlers.  Also some simplications
   and speedups.
 - gcc 3.3-hammer compile fixes for inline assembly
 - Remove acpi.c file corpse.
 - Lots of warning fixes
 - Disable some Dprintks to make the bootup quieter again
 - Clean up ptrace a bit (together with warning fixes)
 - Merge with i386 (handle ACPI dynamic irq entries properly)
 - Disable change_page_attr in pci-gart for now.  Strictly that's
   incorrect, need to do more testing for the root cause of the current
   IOMMU problems.
 - Update defconfig
 - Disable first prefetch in copy_user that is likely to trigger Opteron
   Errata #91
 - More irqreturn_t fixes
 - Add pte_user and fix the vsyscall ptrace hack in generic code.
   It's still partly broken
 - Port verbose MCE handler from 2.4

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

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

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

into tetrachloride.(none):/mnt/raid/src/kernel/2.5/agpgart

The calling task must have a valid reclaim_state when running page
reclaim.  But I had forgotten about shrink_all_memory().

into kroah.com:/home/linux/linux/BK/pci-2.5

Move the PCI core to start using these, enabling proper reference counting
on struct pci_dev.

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

From Jeff Wiedemeier and Richard Henderson

into nuts.ninka.net:/home/davem/src/BK/net-2.5

into tetrachloride.(none):/mnt/raid/src/kernel/2.5/agpgart

Adjusts the IPTables targets to handle non-linear packets.
ipt_ULOG untested.

Adjusts the IPTables matches to handle non-linear packets.
Untested: ipt_ah and ipt_esp.

Adjusts the IPTables core to handle non-linear packets.  Extensions
done separately in next patch.  Also, comments about when to set
"me" field in struct ipt_match, ipt_table and ipt_target are
corrected.

Generic suspend/resume methods are no more.

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

Trivial bugfix: opening a file that doesn't have ->direct_IO() with
O_DIRECT passed in flags fails (as it should) but doesn't call
->release() even though ->open() had been successful.

IOW, we have a leak there - e.g.  open() of a tty with O_DIRECT will
leak a reference to tty and tty_driver.  Besides, quite a few drivers
are _not_ happy with struct file being freed without ->release() after
->open() has returned 0.

This converts /proc/pid/maps to use of seq_file, cleans the issues with
d_path() overflows as a side effect.

It's incremental to seq_path() patch.