Keep track of kswapds: it's OK that they get moved off a node when the
last CPU goes down, but when a CPU comes back, we should try to move
the kswapd back onto its node.

Workqueues need to bring up/destroy the per-cpu thread on cpu up/down.

1) Add a global list of workqueues, and keep the name in the structure
   (to name the newly created thread).

2) Remove BUG_ON in run_workqueue, since thread is dragged off CPU when
   it goes down.

3) Lock out cpu up/down in flush_workqueue, create_workqueue and
   destroy_workqueue.

4) Add notifier to add/destroy workqueue threads, and take over work.

Change ksoftirqd not to assume it's on the CPU: when a cpu goes down,
it will be rudely dragged off.  Since do_softirq() uses
smp_processor_id(), it's easiest to disable preemption, check that the
cpu is still up, then call do_softirq().

If the cpu is actually offline, wait for the notifier, which kills us.

Take over tasklets from dead cpu in the notifier.

Clean up redundant double assignment in CPU_UP callback.

Add hook for RCU to handle jobs on dead cpu.  Requires new
tasklet_kill_immediate for RCU to clean up its tasklet (which might
have been about to run, so tasklet_kill won't work).

Change the migration thread to directly use its cpu arg, rather than
smp_processor_id(): if a cpu goes up then down rapidly, it can be on
the wrong cpu just before it is stopped.

Add code to stop the migration thread on CPU_DEAD and CPU_UP_CANCELED.

Remove the (bogus) priority of the notifier.

We need the migration thread to be RT as soon as the CPU comes online:
for example, stop_machine() (another RT task) expects to yield to it.
Extract the core of setscheduler() and do that when the migration
thread is created.  rq lock is a precaution against the (theoretical)
possibility of someone else doing setscheduer on this thread at the
same time.

Currently the migration thread re-enables irqs, then calls
move_task_away which disables IRQs again and actually does the move.
This means there is a race where the migration thread gets preempted,
and the target CPU can go down.

Hold irqs disabled in migration thread across move_task_away(), which
now doesn't need to save flags (the other caller is the hotplug CPU
code, where irqs are also disabled).

Grab cpu lock around sched_migrate_task() and sys_sched_setaffinity().
This is a noop without CONFIG_HOTPLUG_CPU.

The sched_migrate_task may have a performance penalty on NUMA if lots
of exec rebalancing is happening, however this only applies to
CONFIG_NUMA and CONFIG_HOTPLUG_CPU, which noone does at the moment
anyway.

Also, the scheduler in -mm solves the race another way, so this will
vanish then.

Don't move tasks onto offline cpus in load_balance and wake_task (the
latter is caused by a completion run from stop_machine).

Note that cpu_is_offline() is a noop when CONFIG_HOTPLUG_CPU=n.

Add "online" sysfs attribute to cpus to bring them up and down.

Again, only under CONFIG_HOTPLUG_CPU.

Implement cpu_down(): uses stop_machine to freeze the machine, then
uses (arch-specific) __cpu_disable() and migrate_all_tasks().

Whole thing under CONFIG_HOTPLUG_CPU, so doesn't break archs which
don't define that.

From: Nikita Danilov <Nikita@Namesys.COM>

remove_suid() ignores return value of notify_change()->i_op->setattr().
This mean, that even if file system fails to clear suid bit,
generic_file_aio_write_nolock() proceeds with write, which is unsafe.

Actually, even ext2's ->setattr() can fail, when trying to update ACL, for
example.

Attached patch modifies remove_suid() to return result of ->setattr(), and
updates in-tree callers.

From: Stelian Pop <stelian@popies.net>

This patchlet is just a resync with my tree, it only increments the meye
driver version number and makes some small comment changes as suggested by
Randy Dunlap.

From: Andy Whitcroft <andyw@uk.ibm.com>

Whilst looking at the memory overcommit logic I noticed that the pointer to
the documentation from the *_vm_enough_memory calls is incorrect.  Also
that in one instance the routine does not have the expected pointers.

From: Stelian Pop <stelian@popies.net>

This patch removes the usage of __devinit in the srs methods of the sonypi
driver, because those functions are also called from sonypi_pm_callback().

Patch originally from Randy Dunlap.

From: Grant Grundler <grundler@parisc-linux.org>

This patch adds a comment to "Documentation/SubmittingDrivers" about the
importance of adding a Copyright notice in submitted code.

The parisc-linux port has neglected this in the past and I've been slowly
trying to correct that (along with proper GPL header).

While I make it sound like GPL is the "only" acceptable license, I'll leave
it up to lawyers to determine what other appropriate license could be used
for a new driver.

From: Carl Spalletta <ioanamitu@yahoo.com>

Fix a comment bug.

From: Matt Domsch <Matt_Domsch@dell.com>

Split EDD assembly code from setup.S into edd.S.  This will enable it to be
#included into x86-64 too.

From: Matt Domsch <Matt_Domsch@dell.com>

move edd.c from arch/i386/kernel to new dir drivers/firmware.  Fix up
makefiles and Kconfigs.

From: Matt Domsch <Matt_Domsch@dell.com>

Three patches to move the BIOS Enhanced Disk Drive code from i386-specific
locations into more generic locations, which will allow it to be used on
x86-64 as well.

move edd.h from include/asm-i386 to include/linux

From: DHollenbeck <dick@softplc.com>

This patch was needed against a pristine 2.6.4 kernel when compiling with
"gcc 3.4 _very recent_" using the -Os option.

Without this patch, modules would use a non-inline memcmp() and then not
find it in the kernel, causing depmod to complain and some modules not to
load.

From: Tim Schmielau <tim@physik3.uni-rostock.de>

BSD accounting was missed in the conversion from HZ to USER_HZ.  I thought
nobody cared, but apparently there are still users to it.

From: "Randy.Dunlap" <rddunlap@osdl.org>

From: Michael Veeck <michael.veeck@gmx.net>

Remove unnecessary min/max macros and changes calls to use kernel.h macros
instead.

From: "Randy.Dunlap" <rddunlap@osdl.org>

From: Michael Veeck <michael.veeck@gmx.net>

Remove unnecessary min/max macros and changes calls to use kernel.h macros
instead.

From: "Randy.Dunlap" <rddunlap@osdl.org>

From: Michael Veeck <michael.veeck@gmx.net>

Remove unnecessary min/max macros and changes calls to use kernel.h macros
instead.

From: "Randy.Dunlap" <rddunlap@osdl.org>

From: Michael Veeck <michael.veeck@gmx.net>

Patch (against 2.6.3) removes unnecessary min/max macros and changes calls to
use kernel.h macros instead.

From: "Randy.Dunlap" <rddunlap@osdl.org>

This trivial patch fixes the bug #320:

http://bugme.osdl.org/show_bug.cgi?id=320

The additional comment:

http://bugme.osdl.org/show_bug.cgi?id=320#c1

Is wrong, because it will send the command twice.

We only want to see if the command got success, thus is not necessary
to test against < 0 (if the return value is not 1, we got a error).

Note that I'm using the function (bad) style.

From: "Randy.Dunlap" <rddunlap@osdl.org>

Remove the rest of references to smp.tex
Documentation/cpufreq => Documentation/cpu-freq
DocBook/tulip.{pdf,ps,html} => DocBook/tulip-user.{pdf,ps,html}
Bunch of other typos.

From: "Randy.Dunlap" <rddunlap@osdl.org>

From: Luiz Fernando Capitulino <lcapitulino@prefeitura.sp.gov.br>

From: "Randy.Dunlap" <rddunlap@osdl.org>

From: Luiz Fernando Capitulino <lcapitulino@prefeitura.sp.gov.br>

From: Roman Zippel <zippel@linux-m68k.org>

When a boolean choice value has a dependency of 'm' it can be shortly
treated as a tristate symbol.  This fixes this and also add a small
optimization to precompute the value of the module symbol instead of
checking it all the time.

From: Roman Zippel <zippel@linux-m68k.org>

This patch is by Andreas Fester <Andreas.Fester@gmx.de> and saves and
restores various runtime options of qconf.

From: Roman Zippel <zippel@linux-m68k.org>

This disables some debug prints, which are more confusing than helpful for
normal users.

From: Roman Zippel <zippel@linux-m68k.org>

conf_write() now checks the target path whether it's a directory, so it
saves the config in the directory instead of renaming it.

From: Roman Zippel <zippel@linux-m68k.org>

The correct link path is needed at config time to find the correct library,
so let's ask gcc for the real path.

From: Brian Gerst <bgerst@didntduck.org>

Don't abuse empty_zero_page as temporary storage for boot parameters and
command line.  This is a holdover from the days before discardable init
sections.

From: Matt Mackall <mpm@selenic.com>

This keeps the alignment of the init task matched with the stack size.

From: Maneesh Soni <maneesh@in.ibm.com>

Fix a sysfs use-after-free crash.  The problem we have is of the kobject
going away while we have a live dentry (the corresponding sysfs directory)
still pointing to it throuh d_fsdata pointer.  The patch makes sure to keep
the kobject alive by taking a reference to it during the life-time of
corresponding dentry.


o The following pins the kobject when sysfs assigns dentry and inode to
  the kobject. This ensures that kobject is alive during the life time of
  the dentry and inode, and people holding ref. to the dentry can access the
  kobject without any problems.

o The ref. taken for the kobject is released through dentry->d_op->d_iput()
  call when the dentry ref. count drops to zero and it is being freed. For
  this sysfs_dentry_operations is introduced.

For testing one has to run the following test on a SMP box:

1) Do insmod/rmmod "dummy.o" network driver in a forever loop.

2) Parallely do "find /sys/class/net | xargs cat" also in a forever loop.

From: Maneesh Soni <maneesh@in.ibm.com>

The following patch fixes the dentry refcounting, during
sysfs_remove_group() and also adds the missing dput() for the "extra" ref
taken during sysfs_create() for the sub-directory dentry corresponding to
attribute group.

From: Maneesh Soni <maneesh@in.ibm.com>

I have re-done the patch fixing the race between sysfs_remove_dir() and
dcache_readdir().  If you recall, sysfs_remove_dir(kobj) manipulates the
->d_subdirs list for the dentry corresponding to the sysfs directory being
removed.  It can end up deleting the cursor dentry which is added to the
->d_subdirs list during a concurrent dcache_dir_open() ==> dcache_readdir()
for the same directory.  And as a result dcache_readdir() can loop for ever
holding dcache_lock.

The earlier patch which was included in -mm1 created problems which
resulted in list_del() BUG hits in prune_dcache().  The reason I think is
that in the main loop in sysfs_remove_dir(), dcache_lock is dropped and
re-acquired, and this could result in inconsistent ->d_subdirs list and
prune_dcache() may try to delete an already deleted dentry.  I have
corrected this in the new patch as below.

I could do sysfs_remove_dir() more neatly on sysfs backing store patch set
as there I don't use the ->d_subdirs list.  Instead the list of children
sysfs_dirent works out well.  But untill sysfs backing store patch is
picked up the existing code suffer from this race.  This can be easily
tested by running following two loops on a SMP box

# while true; do insmod drivers/net/dummy.ko; rmmod dummy; done
# while true; do find /sys/class/net > /dev/null; done


o This patch fixes sysfs_remove_dir race with dcache_readdir.  There is
  no need for sysfs_remove_dir to modify the d_subdirs list for the
  directory being deleted as it is taken care in the final dput.  Modifying
  this list results in inconsistent d_subdirs list and causes infinite loop
  in concurrently occurring dcache_readdir.

o The main loop is restarted every time, dcache_lock is re-acquired in
  order to maintain consistency.