The syscall optimization patches broke the OS INIT handler because
SAVE_RESET was addressing relative to r12, which contains the virtual
address of the stack pointer.  Fixed by addressing relative to r2/r3
instead.

Please find the attached patch that:

1. Moves user stack flag memory access before srlz.i;

2. Moves mov b6=r22 as late as possible. 

3. Changes (pSys) to (pLvSys) in skip_rbs_switch: section. IA32 syscall
set pSys=1 but pLvSys=0. It's not necessary to clear bank1 r16-r19 registers
for IA32 syscall.

The number for leave_syscall is 268 cycles with this patch. The number
is 295 cycles w/o this patch. It was 245 cycles with the original kee patched
kernel. The 23 cycles come from restoring b6 operation which didn't exist in
the orignal kee patch.

Only call get_scratch_regs() when pt is really needed.  The extraneous
calls to get_scratch_regs() can otherwise pick up the wrong address for pt.
For example, calling unw_access_ar(&info, UNW_AR_BSPSTORE,...) before pt_regs
had been reached could trigger this bug.

A bug-fix in IA-32 emulation ptrace code. The bug originally got
introduced with the addition of FPXREGS support in ptrace.

The bug is in ptrace get/set FPREGS routine. gdb by default will not use
FPREGS routines when FPXREGS routines are supported.  So we may not see
this bug during normal gdb operations. But, if gdb (or any other app)
directly tries to get/set FPREGS (probably an old version of gdb), it
will end with an segmentation fault due to this bug.

Attached patch fixes the issue. The patch is taken against 2.5.69. But
it applies to 2.4 tree as well.

the first patch  moves some definitions
out of perfmon.c to perfmon.h (similar to what is in 2.4). 

The second patch contains a fix for the generic built. I did 
it that way to avoid other problems such as SIMSCSI depends
on SCSI. so simply hardcoding this in the MAkefile could cause
other problems. Anyway, SIMSCI is not defaulting to y when
HP_SIM is defined so it is not so different.

Attached is a patch for interleaved discontigmem banks.

When memory banks are interleaved between nodes,
bank ids can be overwritten by other nodes.

I was building 2.5 and noticed that if you have raid5 configured as
a module, then the arch/ia64/lib/xor.S would not get built (gets a
make error since c file not there).

Attached patch fixes that so xor.S gets built if raid5 is built-in
or a module.

We need sometimes to load the physical address of a kernel
object.  Often we can convert the virtual address to physical
at execution time, but sometimes (either for performance reasons
or during error recovery) we cannot to this.  Patch the marked
bundles to load the physical address.

	kernel patch.  This patch makes it work again.  The MCA path is
	similarly broken.  Patch will follow later.

  Delete: arch/ia64/tools/print_offsets.awk
.del-Makefile~90bde6a95198c56c:
  Delete: arch/ia64/tools/Makefile

Delete references to arch/ia64/tools; use standard sed script to generate
offsets.h.

This gets the dependencies right for offsets.h

	run faster than break-based syscall stubs, even if there is
	no light-weight syscall handler.
	Adds a new boot command-line option "nolwsys" which can be used
	to turn off light-weight system call handlers.  Good for
	performance measurement and (potentially) for debugging.

	signals corrupting r14.

	fix bit rot in signal debug printk.

	Instead of using break, this one bubbles down into the kernel
	directly.  This code works, but isn't performance optimal yet.

	didn't preserve b6 and r11 anymore.  Unfortunately, preserving them costs
	a few cycles (~5 cycles in the cached case).  The uncached case is hopefully
	mostly unaffected because the number of cache-lines touched is the same
	(without preserving b6 and r11, the entry-patch _almost_ got away with
	touching a single 128-byte cacheline, but not quite, because r8 also
	had to be initialized).

	now cleared by the syscall exit path.

	whitespace changes only.

	Fix setting of current->thread.on_ustack flag in optimized syscall exit path.
	Tune break_fault for syscall execution.
	Break ia32_execve: the ia64_execve() hack that was there is too ugly for
	words; surely we can do better...

Here's another sn2 update.  It includes a bunch of misc. bits:
  o a bunch of cleanup from hch
  o addition of DMA routine wrappers
  o update of other PCI routines
  o topology.h prototype addition.

This should get the CONFIG_PREEMPT to work again.  We were trashing
r20 by accident.

fails:

shmctl(shmid, IPC_STAT, &shmid_ds) 

The patch corrects the definition of the ipc_perm32 structure.

This one is similar to msgctl. We should be calling sys_shmctl with
struct shmid64_ds.

In the absence of this patch:

shmctl(shmid, IPC_SET, &shmid_ds)

will fail.

IA-32 programs executing:

msgctl(id, IPC_SET, &buf)

currently fail with EPERM due to this bug.

Patch by Rohit Seth, Fengua Yu, and Arun Sharma:

Please find attached a patch for kernel entry exit optimization. This is
based on 2.5.69 kernel.

The main items covered by this patch are:
1) Support for 16 bytes instructions as per SDM2.1 (CSD/SSD in pt_regs)
2) f10-f11 are added as additional scratch registers for kernel's use.
3) Re-arrange pt_regs to access less cache lines in system call. Reduce
scratch register saving/restoring in system call path.
4) A few instruction reorg in low-level code.

	came out wrong.

Some patches from hch@infradead.org cleaning up sn2 code a bit,
including the removal of some unnecessary files.

This patch moves the kernel text and data into region 5 (0xa00...) by using
a translation register to pin the entire area (i.e., no TLB faults).
The 1st-order goal is to be able to boot a kernel even when there is
no usable memory in the 64-128MB range.  It is also a step towards enabling
text-replication on NUMA.