SGI Modid: 2.5.x-xfs:slinx:152354a

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

Remove un-used commented module_exit functions from m68knommu
ColdFire and 68328 serial drivers. These drivers currently cannot
be configured as modules, and they have no exit functions.

Global SECURITY_INIT macro cannot be used inside .init section
for m68knommu linker script. It is a complete section of its own,
need to just list the components individually.

Conditionally copy the ROMfs filesystem on the NETtel/5307
target board only if using a ROMfs.

DragenEngine setup code updates:

 - Change interrupt handler return type to irqreturn_t
 - Allow configure time setting of boot parameters
 - Clean up warnings

Conditionally copy the ROMfs filesystem on the SecureEdgeMP3/5307
target board only if using a ROMfs.

Configuration updates for 68328 DragenEngine board. Fix up name
so that it is "DragenEngine" and clean up eeprom read.

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

   The intent patches broke behaviour w.r.t. following symlinks when
doing an open() with file creation. The problem occurs in open_namei()
because the LOOKUP_PARENT flag is no longer set when we do the call to
follow_link().

This is the updated versions of the patch Ingo sent some time ago to
implement a new tgkill() syscall which specifies the target thread
without any possibility of ambiguity or thread ID wrap races, by passing
in both the thread group _and_ the thread ID as the arguments.

This is really needed since many/most people still run with limited PID
ranges (maybe due to legacy apps breaking) and the PID reuse can cause
problems.

This allows suspend to work on UP machines, even if the kernel
is compiled for SMP.

Due to a typo in powernow-k7.c, the value which correspond
to the CPU core multiplicator and the VID value are swapped
when we go down to up in frequency step.

This adds a declaration that the macserial driver needs in order to
compile correctly, and removes some old SERIAL_DO_RESTART junk which
isn't used (SERIAL_DO_RESTART is never defined in this driver) and which
I think is incorrect anyway, since it looks to me like it would
potentially return an ERESTARTSYS error without a signal pending.

switch_mm and enter_lazy_tlb take a CPU arg, which is always
smp_processor_id().  This is misleading, and pointless if they use
per-cpu variables or other optimizations.  gcc will eliminate
redundant smp_processor_id() (in inline functions) anyway.

This removes that arg from all the architectures.

kstat_this_cpu() is defined in terms of per_cpu instead of __get_cpu_var.

This patch changes that, and uses it everywhere appropriate.  The sched.c
change puts it in a local variable, which helps gcc generate better code.

No longer need the 68360 specific trap init call. The generic
interrupt/trap code is now setup to do this itself.

Define the raw read and write access macros for m68knommu.
These rae use by MTD drivers in particular.

Clean up show_process_blocks() loop for non-mmu targets.

This patch includes the last peices of the flat laoder shared
library support. Define the shared lib limit and implement a
flag for doing kernel level tracing.

A couple of minor fixes for the 68328 interrupt setup code.

 - don't define the .romvec section for DragonEngine build
 - print newline at end of spurious interrupt count in show_interrupts()

doh - double unlock in the acct-is-on path. Noticed by Aneesh Kumar K.V
<aneesh.kumar@digital.com>

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

Conditionally copy the ROMfs filesystem on the Cleopatra/5307
target board only if using a ROMfs.

From: Ingo Molnar <mingo@elte.hu>

Most distributions turn on process accounting - but even the common
'accounting is off' case is horrible SMP-scalability-wise: it accesses a
global spinlock during every sys_exit() call, which bounces like mad on SMP
(and NUMA) systems.

(i also got rid of the unused return code.)

From: "lode leroy" <lode_leroy@hotmail.com>

I would like to submit a trivial enhancement to display the ip address of
the bootserver in /proc/net/pnp

This aids me in developing a diskless linux root image to know where it
comes from...

From: Jan Kara <jack@suse.cz>

This implements autoloading of quota modules. 

From: Andreas Gruenbacher <agruen@suse.de>

This patch removes the dependency on i_sem in the getxattr and
listxattr iops of ext2 and ext3. In addition, the global ext[23]_xattr
semaphores go away. Instead of i_sem and the global semaphore, mutual
exclusion is now ensured by per-inode xattr semaphores, and by locking
the buffers before modifying them. The detailed locking strategy is
described in comments in fs/ext[23]/xattr.c.

Due to this change it is no longer necessary to take i_sem in
ext[23]_permission() for retrieving acls, so the
ext[23]_permission_locked() functions go away.

Additionally, the patch fixes a race condition in ext[23]_permission:
Accessing inode->i_acl was protected by the BKL in 2.4; in 2.5 there no
longer is such protection. Instead, inode->i_acl (and inode->i_default_acl)
are now accessed under inode->i_lock. (This could be replaced by RCU in
the future.)

In the ext3 extended attribute code, an new uglines results from locking
at the buffer head level: The buffer lock must be held between testing
if an xattr block can be modified and the actual modification to prevent
races from happening. Before a block can be modified,
ext3_journal_get_write_access() must be called. But this requies an unlocked
buffer, so I call ext3_journal_get_write_access() before locking the
buffer. If it turns out that the buffer cannot be modified,
journal_release_buffer() is called. Calling ext3_journal_get_write_access
after the test but while the buffer is still locked would be much better.

From: Andreas Gruenbacher <agruen@suse.de>

Andrew Morton found that there is lock contention between extended
attribute operations (like reading ACLs, which `ls -l' needs to do)
and other operations on the same files. This is due to the fact that
all extended attribute syscalls take inode->i_sem before calling into
the filesystem code.

To fix this problem, this patch no longer takes inode->i_sem in the
getxattr and listxattr syscalls, and moves the lock taking code into
the file systems. (Another patch improves the locking strategy in
ext2 and ext3.)

From: Andreas Gruenbacher <agruen@suse.de>

It is common to update extended attributes without changing the value's
length. This patch optimizes this case. In addition to that, the current
code tries to recognize early when extended attribute blocks become
empty. This optimization is not of significant value, so this patch
removes it, and moves the empty block test further down.

From: Andreas Gruenbacher <agruen@suse.de>

The inode->i_bdev field is not the same as inode->i_sb->s_bdev or bh->b_bdev.
We must compare inode->i_sb->s_bdev with bh->b_bdev, or else equal extended
attribute block will not be found.

From: From: Andreas Gruenbacher <agruen@suse.de>

* Various minor cleanups and simplifications in the extended attributes
  and acl code.

* Use a smarter shortcut rule in ext[23]_permission(): If the mask
  contains permissions that are not also contained in the group
  file mode permission bits, those permissions can never be granted by
  an acl. (The previous shortcut rule was more coarse.)