[PATCH] fix memory leak in dma_declare_coherent_memory()

When it goes to free1_out, dev->dma_mem has not been freed.

Signed-off-by: Yoichi Yuasa <yoichi_yuasa@tripeaks.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] Fix buffer overflow and races in capi debug functions

The CAPI trace debug functions were using a fixed size buffer, which can be
overflowed if wrong formatted CAPI messages were sent to the kernel capi
layer.  The code was also not protected against multiple callers.  This fix
bug 8028.

Additionally the patch make the CAPI trace functions optional.

Signed-off-by: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] adapt page_lock_anon_vma() to PREEMPT_RCU

page_lock_anon_vma() uses spin_lock() to block RCU.  This doesn't work with
PREEMPT_RCU, we have to do rcu_read_lock() explicitely.  Otherwise, it is
theoretically possible that slab returns anon_vma's memory to the system
before we do spin_unlock(&anon_vma->lock).

[ Hugh points out that this only matters for PREEMPT_RCU, which isn't merged
  yet, and may never be.  Regardless, this patch is conceptually the
  right thing to do, even if it doesn't matter at this point.  - Linus ]

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] Documentation: CPU load calculation description

Describes how/when the information exported to `/proc/stat' is calculated,
and possible problems with this approach.

Signed-off-by: Vassili Karpov <av1474@comtv.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] sched: fix SMT scheduler bug

The SMT scheduler incorrectly skips kernel threads even if they are
runnable (but they are preempted by a higher-prio user-space task which got
SMT-delayed by an even higher-priority task running on a sibling CPU).

Fix this for now by only doing the SMT-nice optimization if the
to-be-delayed task is the only runnable task.  (This should cover most of
the real-life cases anyway.)

This bug has been in the SMT scheduler since 2.6.17 or so, but has only
been noticed now by the active check in the dynticks code.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Michal Piotrowski <michal.k.k.piotrowski@gmail.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] ps3: introduce CONFIG_PS3_ADVANCED

ps3: Introduce CONFIG_PS3_ADVANCED, as suggested by Roman Zippel, and use
it to control questions about PS3 subsystems that may not be obvious for
the casual user.

This gets rid of the following warning on non-powerpc platforms: |
drivers/video/Kconfig:1604:warning: 'select' used by config symbol 'FB_PS3'
refer to undefined symbol 'PS3_PS3AV'

Signed-off-by: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Acked-by: Geoff Levand <geoffrey.levand@am.sony.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] ext[34]: EA block reference count racing fix

There are race issues around ext[34] xattr block release code.

ext[34]_xattr_release_block() checks the reference count of xattr block
(h_refcount) and frees that xattr block if it is the last one reference it.
 Unlike ext2, the check of this counter is unprotected by any lock.
ext[34]_xattr_release_block() will free the mb_cache entry before freeing
that xattr block.  There is a small window between the check for the re
h_refcount ==1 and the call to mb_cache_entry_free().  During this small
window another inode might find this xattr block from the mbcache and reuse
it, racing a refcount updates.  The xattr block will later be freed by the
first inode without notice other inode is still use it.  Later if that
block is reallocated as a datablock for other file, then more serious
problem might happen.

We need put a lock around places checking the refount as well to avoid
racing issue.  Another place need this kind of protection is in
ext3_xattr_block_set(), where it will modify the xattr block content in-
the-fly if the refcount is 1 (means it's the only inode reference it).

This will also fix another issue: the xattr block may not get freed at all
if no lock is to protect the refcount check at the release time.  It is
possible that the last two inodes could release the shared xattr block at
the same time.  But both of them think they are not the last one so only
decreased the h_refcount without freeing xattr block at all.

We need to call lock_buffer() after ext3_journal_get_write_access() to
avoid deadlock (because the later will call lock_buffer()/unlock_buffer
() as well).

Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Cc: Andreas Gruenbacher <agruen@suse.de>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] uml: pte_mkread fix

Fix the fact that pte_mkread set _PAGE_RW instead of _PAGE_USER (the logic is
copied from i386 in most place, so it is really as bad as you're thinking).

Thus currently page tables are more permissive than they should.

Such a change may trigger other latent bugs, so be careful with this.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] uml: host VDSO fix

This fixes a problem seen by a number of people running UML on newer host
kernels.  init would hang with an infinite segfault loop.

It turns out that the host kernel was providing a AT_SYSINFO_EHDR of
0xffffe000, which faked UML into believing that the host VDSO page could be
reused.  However, AT_SYSINFO pointed into the middle of the address space, and
was unmapped as a result.  Because UML was providing AT_SYSINFO_EHDR and
AT_SYSINFO to its own processes, these would branch to nowhere when trying to
use the VDSO.

The fix is to also check the location of AT_SYSINFO when deciding whether to
use the host's VDSO.

Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] uml: enable RAW

Add the RAW device driver options to the UML Kconfig.char file so that you may
use them in UML.

Signed-off-by: Allan Graves<allan.graves@gmail.com>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] FRV: Missing error defs

linux/irq.h uses EINVAL but does not #include linux/errno.h.  This results in
the compiler spitting out errors on some files.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] throttle_vm_writeout(): don't loop on GFP_NOFS and GFP_NOIO allocations

throttle_vm_writeout() is designed to wait for the dirty levels to subside.
But if the caller holds IO or FS locks, we might be holding up that writeout.

So change it to take a single nap to give other devices a chance to clean some
memory, then return.

Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Pete Zaitcev <zaitcev@redhat.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] ACPI: make bay depend on dock

Since the bay driver depends on the dock driver for proper notification,
make this driver depend on the dock driver.

Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Acked-by: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] Bug in MM_RB debugging

The code is seemingly trying to make sure that rb_next() brings us to
successive increasing vma entries.

But the two variables, prev and pend, used to perform these checks, are
never advanced.

Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Andrea Arcangeli <andrea@novell.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] uml: add back accidentally removed error

In the 2.6.20 hang patch, I accidentally threw out an error message.
This puts it back.

Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] uml: fix host LDT lookup initialization locking, try 2

Add some locking to host_ldt_entries to prevent racing when reading LDT
information from the host.

The locking is somewhat more careful than my previous attempt.  Now, only
the check of host_ldt_entries is locked.  The lock is dropped immediately
afterwards, and if the LDT needs initializing, that (and the memory
allocations needed) proceed outside the lock.

Also fixed some style violations.

Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] drivers/isdn/gigaset: build asyncdata.o into the gigaset module (fix)

a) Remove #define acrobatics that have become unnecessary by the move of
   asyncdata.o into the common part.

b) Correct the rule for building the common part into the kernel when
   some or all hardware specific parts are built as modules.

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Cc: Adrian Bunk <bunk@stusta.de>
Cc: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] eCryptfs: no path_release() after path_lookup() error

Dmitriy Monakhov wrote:
> if path_lookup() return non zero code we don't have to worry about
> 'nd' parameter, but ecryptfs_read_super does path_release(&nd) after
> path_lookup has failed, and dentry counter becomes negative

Do not do a path_release after a path_lookup error.

Signed-off-by: Michael Halcrow <mhalcrow@us.ibm.com>
Cc: Dmitriy Monakhov <dmonakhov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] eCryptfs: remove unnecessary flush_dcache_page()

Remove unnecessary flush_dcache_page() call. Thanks to Dmitriy
Monakhov for pointing this out.

Signed-off-by: Michael Halcrow <mhalcrow@us.ibm.com>
Cc: Dmitriy Monakhov <dmonakhov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] eCryptfs: set O_LARGEFILE when opening lower file

O_LARGEFILE should be set here when opening the lower file.

Signed-off-by: Michael Halcrow <mhalcrow@us.ibm.com>
Cc: Dmitriy Monakhov <dmonakhov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] rtc_cmos oops fix

Fix an oops on the rtc_device_unregister() path by waiting until the last
moment before nulling the rtc->ops vector.  Fix some potential oopses by
having the rtc_class_open()/rtc_class_close() interface increase the RTC's
reference count while an RTC handle is available outside the RTC framework.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] MAINTAINERS: Update email address

adaplas@pol.net is still alive, but is choking on the traffic.

Signed-off-by: Antonino Daplas <adaplas@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] add -mm testing in SubmitChecklist

Add -mm testing to SubmitChecklist.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] fix section mismatch warning in lockdep

lockdep_init() is marked __init but used in several places
outside __init code. This causes following warnings:
$ scripts/mod/modpost kernel/lockdep.o
WARNING: kernel/built-in.o - Section mismatch: reference to .init.text:lockdep_init from .text.lockdep_init_map after 'lockdep_init_map' (at offset 0x105)
WARNING: kernel/built-in.o - Section mismatch: reference to .init.text:lockdep_init from .text.lockdep_reset_lock after 'lockdep_reset_lock' (at offset 0x35)
WARNING: kernel/built-in.o - Section mismatch: reference to .init.text:lockdep_init from .text.__lock_acquire after '__lock_acquire' (at offset 0xb2)

The warnings are less obviously due to heavy inlining by gcc - this is not
altered.

Fix the section mismatch warnings by removing the __init marking, which
seems obviously wrong.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] Rename PG_checked to PG_owner_priv_1

Rename PG_checked to PG_owner_priv_1 to reflect its availablilty as a
private flag for use by the owner/allocator of the page.  In the case of
pagecache pages (which might be considered to be owned by the mm),
filesystems may use the flag.

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] fix the SYSCTL=n compilation

/home/bunk/linux/kernel-2.6/linux-2.6.20-mm2/kernel/sysctl.c:1411: error: conflicting types for 'register_sysctl_table'
/home/bunk/linux/kernel-2.6/linux-2.6.20-mm2/include/linux/sysctl.h:1042: error: previous declaration of 'register_sysctl_table' was here
make[2]: *** [kernel/sysctl.o] Error 1

Caused by commit 0b4d414714f0d2f922d39424b0c5c82ad900a381.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] Fix posix-cpu-timer breakage caused by stale p->last_ran value

Problem description at:
http://bugzilla.kernel.org/show_bug.cgi?id=8048

Commit b18ec80396834497933d77b81ec0918519f4e2a7
    [PATCH] sched: improve migration accuracy
optimized the scheduler time calculations, but broke posix-cpu-timers.

The problem is that the p->last_ran value is not updated after a context
switch.  So a subsequent call to current_sched_time() calculates with a
stale p->last_ran value, i.e.  accounts the full time, which the task was
scheduled away.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] eCryptfs: resolve lower page unlocking problem

eCryptfs lower file handling code has several issues:
  - Retval from prepare_write()/commit_write() wasn't checked to equality
    to AOP_TRUNCATED_PAGE.
  - In some places page wasn't unmapped and unlocked after error.

Signed-off-by: Michael Halcrow <mhalcrow@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] kernel-doc fixes for 2.6.20-git15 (non-drivers)

Fix kernel-doc warnings in 2.6.20-git15 (lib/, mm/, kernel/, include/).

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] kernel-doc: allow space after __attribute__

Allow space(s) between "__attribute__" and "((blah))" so that
kernel-doc does not complain like:

Warning(/tester/linsrc/linux-2.6.20-git15//kernel/timer.c:939): No description found for parameter 'read_persistent_clock(void'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] update timekeeping_is_continuous comment

Signed-off-by: Daniel Walker <dwalker@mvista.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] FRV: Add some missng lazy MMU hooks for NOMMU mode

Add some missing lazy MMU hooks for NOMMU mode.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] FRV: No ZONE_DMA

FRV does not require a ZONE_DMA, so all DMA'able pages that aren't highmem
should be in ZONE_NORMAL.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] md: add support for reshape of a raid6

i.e. one or more drives can be added and the array will re-stripe
while on-line.

Most of the interesting work was already done for raid5.  This just extends it
to raid6.

mdadm newer than 2.6 is needed for complete safety, however any version of
mdadm which support raid5 reshape will do a good enough job in almost all
cases (an 'echo repair > /sys/block/mdX/md/sync_action' is recommended after a
reshape that was aborted and had to be restarted with an such a version of
mdadm).

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] md: restart a (raid5) reshape that has been aborted due to a read/write error

An error always aborts any resync/recovery/reshape on the understanding that
it will immediately be restarted if that still makes sense.  However a reshape
currently doesn't get restarted.  With this patch it does.

To avoid restarting when it is not possible to do work, we call into the
personality to check that a reshape is ok, and strengthen raid5_check_reshape
to fail if there are too many failed devices.

We also break some code out into a separate function: remove_and_add_spares as
the indent level for that code was getting crazy.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] md: clean out unplug and other queue function on md shutdown

The mddev and queue might be used for another array which does not set these,
so they need to be cleared.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] md: move warning about creating a raid array on partitions of the one device

md tries to warn the user if they e.g.  create a raid1 using two partitions of
the same device, as this does not provide true redundancy.

However it also warns if a raid0 is created like this, and there is nothing
wrong with that.

At the place where the warning is currently printer, we don't necessarily know
what level the array will be, so move the warning from the point where the
device is added to the point where the array is started.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] md: RAID6: clean up CPUID and FPU enter/exit code

- Use kernel_fpu_begin() and kernel_fpu_end()
- Use boot_cpu_has() for feature testing even in userspace

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] md: fix raid10 recovery problem.

There are two errors that can lead to recovery problems with raid10
when used in 'far' more (not the default).

Due to a '>' instead of '>=' the wrong block is located which would result in
garbage being written to some random location, quite possible outside the
range of the device, causing the newly reconstructed device to fail.

The device size calculation had some rounding errors (it didn't round when it
should) and so recovery would go a few blocks too far which would again cause
a write to a random block address and probably a device error.

The code for working with device sizes was fairly confused and spread out, so
this has been tided up a bit.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] Fix failure paths in modules init in umem.c

If register_blkdev() or alloc-disk fail in mm_init() after
pci_register_driver() succeeds, then mm_pci_driver is not unregistered
properly:

Cc: Philip Guo <pg@cs.stanford.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] drivers/video/sm501fb.c: make 4 functions static

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] mm/{,tiny-}shmem.c cleanups

shmem_{nopage,mmap} are no longer used in ipc/shm.c

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] make ipc/shm.c:shm_nopage() static

shm_nopage() can become static.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] arch/arm26/kernel/entry.S: remove dead code

CONFIG_ALIGNMENT_TRAP is never set on arm26.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Ian Molton <spyro@f2s.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[PATCH] x86_64/i386 irq: Fix !CONFIG_SMP compilation

When removing set_native_irq I missed the fact that it was
called in a couple of places that were compiled even when
SMP support is disabled.  And since the irq_desc[].affinity
field only exists in SMP things broke.

Thanks to Simon Arlott <simon@arlott.org> for spotting this.

There are a couple of ways to fix this but the simplest one
is to just remove the assignments.  The affinity field is only
used to display a value to the user, and nothing on either i386
or x86_64 reads it or depends on it being any particlua value,
so skipping the assignment is safe.  The assignment that
is being removed is just for the initial affinity value before
the user explicitly sets it.  The irq_desc array initializes
this field to CPU_MASK_ALL so the field is initialized to
a reasonable value in the SMP case without being set.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Linux 2.6.21-rc2

Too many changes for comfort since -rc1.  Some missed merges, and some
just annoyingly big fixes since.  This is not how an -rc2 should look.
Need to really calm things down!

[SPARC64]: Fix PCI interrupts on E450 et al.

When the PCI controller OBP node lacks an interrupt-map
and interrupt-map-mask property, we need to form the
INO by hand.  The PCI swizzle logic was not doing that
properly.

This was a regression added by the of_device code.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge /pub/scm/linux/kernel/git/davej/agpgart

* master.kernel.org:/pub/scm/linux/kernel/git/davej/agpgart:
  [AGPGART] fix compile errors

Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/netdev-2.6

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6: (52 commits)
  netxen: do_rom_fast_write error handling
  natsemi: Fix detection of vanilla natsemi cards
  net: remove a collection of unneeded #undef REALLY_SLOW_IO stuff
  chelsio: Fix non-NAPI compile
  cxgb3 - Feed Rx free list with pages
  cxgb3 - Recovery from HW starvation of response queue entries.
  cxgb3 - Unmap offload packets when they are freed
  cxgb3 - FW version update
  cxgb3 - private ioctl cleanup
  cxgb3 - manage sysfs attributes per port
  S2IO: Restoring the mac address in s2io_reset
  S2IO: Avoid printing the Enhanced statistics for Xframe I card.
  S2IO: Making LED off during LINK_DOWN notification.
  S2IO: Added a loadable parameter to enable or disable vlan stripping in frame.
  S2IO: Optimized the delay to wait for command completion
  S2IO: Fixes for MSI and MSIX
  qla3xxx: Bumping driver version number
  qla3xxx: Kernic Panic on pSeries under stress conditions
  qla3xxx: bugfix tx reset after stress conditions.
  qla3xxx: Check return code from pci_map_single() in ql_release_to_lrg_buf_free_list(), ql_populate_free_queue(), ql_alloc_large_buffers(), and ql3xxx_send()
  ...

Merge branch 'jffs-maint' of /linux/kernel/git/jgarzik/misc-2.6

* 'jffs-maint' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/misc-2.6:
  remove the jffs MAINTAINERS entry

netxen: do_rom_fast_write error handling

Compiler warning spots real error!

The function do_rom_fast_read called in do_rom_fast_write can fail
and leave data1 unset. This causes a compile warning.
The correct thing is to propagate the error out.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

remove the jffs MAINTAINERS entry

This patch removes the MAINTAINERS entry for the removed jffs
filesystem.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

natsemi: Fix detection of vanilla natsemi cards

Bob Tracy <rct@gherkin.frus.com> reported that the addition of support
for Aculab E1/T1 cPCI carrier cards broke detection of vanilla natsemi
cards.  This patch fixes that: the problem is that the driver-specific
ta in the PCI device table is an index into a second table and this
had not been updated for the vanilla cards.

This patch fixes the problem minimally.

Signed-Off-By: Mark Brown <broonie@sirena.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

net: remove a collection of unneeded #undef REALLY_SLOW_IO stuff

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

chelsio: Fix non-NAPI compile

Chelsio without NAPI enabled has been broken (won't compile) since
3de00b89 ("chelsio: NAPI speed improvement"):

    drivers/net/chelsio/sge.c: In function `t1_interrupt`:
    drivers/net/chelsio/sge.c:1716: error: `Q` undeclared (first use in this function)

The change below seems to add back in the declaration and
initialization of `Q` that was removed by mistake, and at least makes
the driver compile for me, although I have no hardware and hence no
way to test whether this actually works.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

cxgb3 - Feed Rx free list with pages

Populate Rx free list with pages.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

cxgb3 - Recovery from HW starvation of response queue entries.

Improve the traffic recovery after the HW ran out of response queue entries.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

cxgb3 - Unmap offload packets when they are freed

Offload packets may be DMAed long after their SGE Tx descriptors are done
so they must remain mapped until they are freed rather than until their
descriptors are freed.  Unmap such packets through an skb destructor.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

cxgb3 - FW version update

Update FW version to 3.2

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

cxgb3 - private ioctl cleanup

Clean up some private ioctls.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

cxgb3 - manage sysfs attributes per port

sysfs attributes are now managed per port, no longer per adapter.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

S2IO: Restoring the mac address in s2io_reset

- Restore in s2io_reset, the mac address assigned during s2io_open.
  Earlier, it was getting overwritten to the factory default (read from the
  eeprom) and subsequently dropping received frames.

- Fixed the typo in calling rtnl_unlock in s2io_set_link function.

Signed-off-by: Sivakumar Subramani <sivakumar.subramani@neterion.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

S2IO: Avoid printing the Enhanced statistics for Xframe I card.

- Enhanced Statistics are supported only for Xframe II (Herculas) card. Add
  condition check such Enhanced statistics will included only in the case of
  Xframe II card.

Signed-off-by: Sivakumar Subramani <sivakumar.subramani@neterion.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

S2IO: Making LED off during LINK_DOWN notification.

- Turning off LED for LINK_DOWN notification
- Return from rxd_owner_bit_reset function if call to set_rxd_buffer_pointer
  fails with ENOMEM

Signed-off-by: Sivakumar Subramani <sivakumar.subramani@neterion.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

S2IO: Added a loadable parameter to enable or disable vlan stripping in frame.

- Added code to not to strip vlan tag when driver is in promiscuous mode
- Added module loadable parameter 'vlan_tag_strip" through which user can
  enable or disable vlan stripping irrespective of mode
  ( promiscuous or non-promiscuous ).

Signed-off-by: Sivakumar Subramani <sivakumar.subramani@neterion.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

S2IO: Optimized the delay to wait for command completion

- Optimized delay to wait for command completion so as to reduce the
  initialization wait time.
- Disable differentiated services steering. By default RMAC is configured to
  steer traffic with certain DS codes to other queues. Driver must initialize
  the DS memory to 0 to make sure that DS steering will not be used by default.

Signed-off-by: Sivakumar Subramani <sivakumar.subramani@neterion.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

S2IO: Fixes for MSI and MSIX

- Added debug statements to print a debug message if the MSI/MSI-X vector (or)
  data is zero.

- This patch removes the code that will enable NAPI for the case of single
  ring and MSI-X / MSI case. There are some issue in the enabling NAPI with
  MSI/MSI-X.  So we are turning off NAPI in the case of MSI/MSI-X.

Signed-off-by: Sivakumar Subramani <sivakumar.subramani@neterion.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Bumping driver version number

Signed-off-by: Benjamin Li <benjamin.li@qlogic.com>
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Kernic Panic on pSeries under stress conditions

To reproduce this panic consistently, we run an intensive network
application like 'netperf'. After waiting for a couple of seconds,
you will see a stack trace and a kernel panic where we are calling
pci_unmap_single() in ql_poll().

Changes:
1)  Check the flags on the Response MAC IO Control block to check for
errors
2)  Ensure that if we are on the 4022 we only use one segment
3)  Before, we were reading the memory mapped producer index register
everytime we iterated in the loop when clearing the queue.  We should
only be iterating to a known point, not as the producer index
is being updated.

Signed-off-by: Benjamin Li <benjamin.li@qlogic.com>
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: bugfix tx reset after stress conditions.

To Reproduce the Problem:
To reproduce this panic consistently, we run an intensive network
application like 'netperf' and then switch to a different console.
After waiting for a couple of seconds, you will see a tx reset has occured.

Reason:
We enable interrupts even if we were not running.

Solution:
Now we will enable interrupts only after we are ready to give up the poll
routine.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Check return code from pci_map_single() in ql_release_to_lrg_buf_free_list(), ql_populate_free_queue(), ql_alloc_large_buffers(), and ql3xxx_send()

pci_map_single() could fail.  We need to properly check the return
code from pci_map_single().  If we can not properly map this address,
then we should cleanup and return the proper return code.

Signed-off-by: Benjamin Li <benjamin.li@qlogic.com>
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: bugfix: Fixed jumbo frame handling for 3032 chip.

The scatter/gather lists were not being build correctly.  When
large frames spanned several buffers the chip would panic.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Clean up receive process.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Dynamically size the rx buffer queue based on the MTU.

This change removes use of constants for rx buffer queue size
and instead calculates the queue length based on what he MTU
is set to.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Remove API to change MTU.

This network device driver shares the same hardware as the qla4xxx
iSCSI driver. Changing the MTU via the device interface will
cause qla4xxx to crash as there is no way to make notification.
Users wishing to change the MTU must do so using an iSCSI
utility such as Qlogic SanSurfer. This forces the user to
unload/reload this network device driver after the MTU
value has been changed in flash.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Fix deadlock issue on error paths

1) Fix deadlock issue when in QL_RESET_ACTIVE state and traversing
   through the Link State Machine
2) Fix deadlock issue when ethtool would call ql_get_settings()
3) Fix deadlock issue when adaptor is ifup'ed but adaptor fails to initialize

Signed-off-by: Benjamin Li <benjamin.li@qlogic.com>
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Changed to use netdev_alloc_skb() from dev_alloc_skb

Signed-off-by: Benjamin Li <benjamin.li@qlogic.com>
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Remove unnecessary memset() in qla3xxx_send()

We do not need to zero out the 64 byte MAC request I/O control block.
  By zeroing out the control block and setting it to proper fields is
redundant work.  This is because in the qla3xxx_send() function we will
already set the proper fields in this structure.  The unused fields are
not looked at by the hardware and do not need to be zeroed out.

Signed-off-by: Benjamin Li <benjamin.li@qlogic.com>
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx: Return proper error codes when the 4022/4032 is being probed

The return code was not properly set when when allocating memory or mapping
memory failed.  Depending on the stack, the return code would sometimes
return 0, which indicates everything was ok, when in fact there was an error.
  This would cause trouble when the module was removed.  Now, we will pass
back the proper return code when an error occurs during the PCI probe.

Signed-off-by: Benjamin Li <benjamin.li@qlogic.com>
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Alignment in mv643xx_eth

The driver contains this little piece of candy:

#if defined(CONFIG_DMA_NONCOHERENT) || defined(CONFIG_NOT_COHERENT_CACHE)
#define ETH_DMA_ALIGN           L1_CACHE_BYTES
#else
#define ETH_DMA_ALIGN           8
#endif

Any reason why we're not using dma_get_cache_alignment() instead?

  Ralf

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

sgiseeq: Don't include unnecessary headerfiles.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

ucc_geth: use of_get_mac_address()

Update ucc_geth_probe() to use function of_get_mac_address() to obtain the MAC
address.

Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

skge: comma consistency

Use comma's consistently.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

skge: fix transmitter flow control

It looks like the skge driver inherited another bug from the sk98lin code.
If I send from 1000mbit port to a machine on 100mbit port, the switch should
be doing hardware flow control, but no pause frames show up in the statistics.

This is the analog of the recent sky2 fixes. The device needs to listen
for multicast pause frames and then not discard them.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

myri10ge: workaround buggy adopted firmwares

Work around a bug which occurs when adopting firmware versions
1.4.4 though 1.4.11 where broadcasts are filtered as if they
were multicasts.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

NetXen: Driver ifconfig fix

Signed-off-by: Amit S. Kale <amitkale@netxen.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

NetXen: Driver unload fixes

Signed-off-by: Amit S. Kale <amitkale@netxen.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: maintainership

Update driver support contact info.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Cc: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Cc: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: janitorial, typos

Janitorial patch. Undo long lines, fix typo in err msg.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Cc: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Cc: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: transmit race

Multiple threads performing a transmit can race into
the spidernet tx ring cleanup code. This puts the
relevant check under a lock.

Signed-off-by: Linas Vepstas <lins@austin.ibm.com>
Cc: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Cc: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: fix racy double-free of skb

It appears that under certain circumstances, a race will result
in a double-free of an skb. This patch null's out the skb pointer
upon the skb free, avoiding the inadvertent deref of bogus data.
The next patch fixes the actual race.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Cc: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Cc: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: separate hardware state from driver state.

This patch separates the hardware descriptor state from the
driver descriptor state, per (old) suggestion from Ben Herrenschmidt.
This compiles and boots and seems to work.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Cc: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Cc: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: move medium variable into card struct

This moves the medium variable into the spidernet card structure.
It renames the GMII_ variables to BCM54XX specific ones.

Signed-off-by: Jens Osterkamp <jens@de.ibm.com>
Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: remove txram full logging

This patches removes logging for SPIDER_NET_GTMFLLINT interrupts.
Since the interrupts are not irregular, and they happen frequently
when using 100Mbps network switches.

Signed-off-by: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: spidernet: add support for Celleb

This patch adds or changes some HW specific settings for spider_net on
Celleb.

Signed-off-by: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: load firmware when open

This patch moves calling init_firmware() from spider_net_probe() to
spider_net_open() so as to use the driver by built-in.

Signed-off-by: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: autoneg support for Celleb

Add auto negotiation support for Celleb.

Signed-off-by: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

spidernet: compile break.

As of 2.6.20-git4, the spider_net driver does not compile.
This appears to be due to some archaic usage involving kobjects.

It also fixes a nasty double-free during ifdown of the interface.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Cc: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Cc: Kou Ishizaki <kou.ishizaki@toshiba.co.jp>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

sungem_phy: support bcm5461 phy, autoneg.

This version moves the medium variable to the card specific structure and
changes the GMII_* to BCM54XX_* #defines.

This patch adds improved version of enable_fiber for both the 5421 and
the 5461 phy. It is now possible to specify with these wether you want
autonegotiation or not. This is needed for bladecenter switches where
some expect autonegotiation and some dont seem to like this at all.
Depending on this flag it sets phy->autoneg accordingly for the fiber mode.

More importantly it implements proper read_link and poll_link functions
for both phys which can handle both copper and fiber mode by determining
the medium first and then branching to the required functions. For fiber
they all work fine, for copper they are not tested but return the result
of the genmii_* function anyway which is supposed to work.

The patch moves the genmii_* functions around to avoid foreward declarations.

Signed-off-by: Jens Osterkamp <jens@de.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

forcedeth: fix checksum feature in mcp65

This patch removes checksum offload feature in mcp65 chipsets as they
are not supported in hw.

Signed-Off-By: Ayaz Abdulla <aabdulla@nvidia.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>