ORE stands for "Objects Raid Engine"

This patch is a mechanical rename of everything that was in ios.c
and its API declaration to an ore.c and an osd_ore.h header. The ore
engine will later be used by the pnfs objects layout driver.

* File ios.c => ore.c

* Declaration of types and API are moved from exofs.h to a new
  osd_ore.h

* All used types are prefixed by ore_ from their exofs_ name.

* Shift includes from exofs.h to osd_ore.h so osd_ore.h is
  independent, include it from exofs.h.

Other than a pure rename there are no other changes. Next patch
will move the ore into it's own module and will export the API
to be used by exofs and later the layout driver
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

Exofs raid engine was saving on memory space by having a single layout-info,
single pid, and a single device-table, global to the filesystem. Then passing
a credential and object_id info at the io_state level, private for each
inode. It would also devise this contraption of rotating the device table
view for each inode->ino to spread out the device usage.

This is not compatible with the pnfs-objects standard, demanding that
each inode can have it's own layout-info, device-table, and each object
component it's own pid, oid and creds.

So: Bring exofs raid engine to be usable for generic pnfs-objects use by:

* Define an exofs_comp structure that holds obj_id and credential info.

* Break up exofs_layout struct to an exofs_components structure that holds a
  possible array of exofs_comp and the array of devices + the size of the
  arrays.

* Add a "comps" parameter to get_io_state() that specifies the ids creds
  and device array to use for each IO.

  This enables to keep the layout global, but the device-table view, creds
  and IDs at the inode level. It only adds two 64bit to each inode, since
  some of these members already existed in another form.

* ios raid engine now access layout-info and comps-info through the passed
  pointers. Everything is pre-prepared by caller for generic access of
  these structures and arrays.

At the exofs Level:

* Super block holds an exofs_components struct that holds the device
  array, previously in layout. The devices there are in device-table
  order. The device-array is twice bigger and repeats the device-table
  twice so now each inode's device array can point to a random device
  and have a round-robin view of the table, making it compatible to
  previous exofs versions.

* Each inode has an exofs_components struct that is initialized at
  load time, with it's own view of the device table IDs and creds.
  When doing IO this gets passed to the io_state together with the
  layout.

While preforming this change. Bugs where found where credentials with the
wrong IDs where used to access the different SB objects (super.c). As well
as some dead code. It was never noticed because the target we use does not
check the credentials.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

ios.c will be moving to an external library, for use by the
objects-layout-driver. Remove from it some exofs specific functions.

Also g_attr_logical_length is used both by inode.c and ios.c
move definition to the later, to keep it independent
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

In future raid code we will need to know the IO offset/length
and if it's a read or write to determine some of the array
sizes we'll need.

So add a new exofs_get_rw_state() API for use when
writeing/reading. All other simple cases are left using the
old way.

The major change to this is that now we need to call
exofs_get_io_state later at inode.c::read_exec and
inode.c::write_exec when we actually know these things. So this
patch is kept separate so I can test things apart from other
changes.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

In the general raid-group case the truncate was wrong in that
it did not also fix the object length of the neighboring groups.

There are two bad cases in the old code:
1. Space that should be freed was not.
2. If a file That was big is truncated small, then made bigger
   again, the holes would not contain zeros but could expose old data.
   (If the growing of the file expands to more than a full
    groups cycle + group size (> S + T))
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

Small cleanup that unifies duplicated code used in both the
error and success cases
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

Since the beginning we realloced the sbi structure when a bigger
then one device table was specified. (I know that was really stupid).

Then much later when "register bdi" was added (By Jens) it was
registering the pointer to sbi->bdi before the realloc.

We never saw this problem because up till now the realloc did not
do anything since the device table was small enough to fit in the
original allocation. But once we starting testing with large device
tables (Bigger then 28) we noticed the crash of writeback operating
on a deallocated pointer.

* Avoid the all mess by allocating the device-table as a second array
  and get rid of the variable-sized structure and the rest of this
  mess.
* Take the chance to clean near by structures and comments.
* Add a needed dprint on startup to indicate the loaded layout.
* Also move the bdi registration to the very end because it will
  only fail in a low memory, which will probably fail before hand.
  There are many more likely causes to not load before that. This
  way the error handling is made simpler. (Just doing this would be
  enough to fix the BUG)
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

Now that pnfs-osd has hit mainline we can remove exofs's
private header. (And the FIXME comment)
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

exofs file system wants to use pnfs_osd_xdr.h file instead of
redefining pnfs-objects types in it's private "pnfs.h" headr.

Before we do the switch we must make sure pnfs_osd_xdr.h is
compilable also under NFS versions smaller than 4.1. Since now
it is needed regardless of version, by the exofs code.

nfs4_string is not the only nfs4 type out in the global scope.
Ack-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jwessel/linux-2.6-kgdb:
  sparc,kgdbts: fix compile regression with kgdb test suite

Commit 63ab25eb (kgdbts: unify/generalize gdb breakpoint adjustment)
introduced a compile regression on sparc.

kgdbts.c: In function 'check_and_rewind_pc':
kgdbts.c:307: error: implicit declaration of function 'instruction_pointer_set'

Simply add the correct macro definition for instruction pointer on the
Sparc architecture.
Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Acked-by: David S. Miller <davem@davemloft.net>

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  CIFS: Fix wrong length in cifs_iovec_read

Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Make Dell Latitude E6420 use reboot=pci
  x86: Make Dell Latitude E5420 use reboot=pci

Yet another variant of the Dell Latitude series which requires
reboot=pci.

From the E5420 bug report by Daniel J Blueman:

> The E6420 is affected also (same platform, different casing and
> features), which provides an external confirmation of the issue; I can
> submit a patch for that later or include it if you prefer:
> http://linux.koolsolutions.com/2009/08/04/howto-fix-linux-hangfreeze-during-reboots-and-restarts/Reported-by: Daniel J Blueman <daniel.blueman@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: <stable@kernel.org>

Rebooting on the Dell E5420 often hangs with the keyboard or ACPI
methods, but is reliable via the PCI method.

[ hpa: this was deferred because we believed for a long time that the
  recent reshuffling of the boot priorities in commit
  660e34ce fixed this platform.
  Unfortunately that turned out to be incorrect. ]
Signed-off-by: Daniel J Blueman <daniel.blueman@gmail.com>
Link: http://lkml.kernel.org/r/1305248699-2347-1-git-send-email-daniel.blueman@gmail.comSigned-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: <stable@kernel.org>

* 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/keithp/linux-2.6:
  drm/i915: Fix unfenced alignment on pre-G33 hardware
  drm/i915: Add quirk to disable SSC on Lenovo U160 LVDS

It seems to hurt performance in real life.  Yes, the inode will be used
later, but the conditional doesn't seem to predict all that well
(negative dentries are not uncommon) and it looks like the cost of
prefetching is simply higher than depending on the cache doing the right
thing.

As usual.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The compiler, at least for ix86 and m68k, validly warns that the
comparison:

	next <= (loff_t)-1

is always true (and it's always true also for x86-64 and probably all
other arches - as long as pgoff_t isn't wider than loff_t).  The
intention appears to be to avoid wrapping of "next", so rather than
eliminating the pointless comparison, fix the loop to indeed get exited
when "next" would otherwise wrap.

On m68k the following warning is observed:

  fs/fscache/page.c: In function '__fscache_uncache_all_inode_pages':
  fs/fscache/page.c:979: warning: comparison is always false due to limited range of data type
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reported-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: Suresh Jayaraman <sjayaraman@suse.de>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: Pavel Shilovsky <piastryyy@gmail.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>

Merge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  signal: align __lock_task_sighand() irq disabling and RCU
  softirq,rcu: Inform RCU of irq_exit() activity
  sched: Add irq_{enter,exit}() to scheduler_ipi()
  rcu: protect __rcu_read_unlock() against scheduler-using irq handlers
  rcu: Streamline code produced by __rcu_read_unlock()
  rcu: Fix RCU_BOOST race handling current->rcu_read_unlock_special
  rcu: decrease rcu_report_exp_rnp coupling with scheduler

Merge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched: Avoid creating superfluous NUMA domains on non-NUMA systems
  sched: Allow for overlapping sched_domain spans
  sched: Break out cpu_power from the sched_group structure

Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86. reboot: Make Dell Latitude E6320 use reboot=pci
  x86, doc only: Correct real-mode kernel header offset for init_size
  x86: Disable AMD_NUMA for 32bit for now

Merge branch 'rcu/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-2.6-rcu into core/urgent

The __lock_task_sighand() function calls rcu_read_lock() with interrupts
and preemption enabled, but later calls rcu_read_unlock() with interrupts
disabled.  It is therefore possible that this RCU read-side critical
section will be preempted and later RCU priority boosted, which means that
rcu_read_unlock() will call rt_mutex_unlock() in order to deboost itself, but
with interrupts disabled. This results in lockdep splats, so this commit
nests the RCU read-side critical section within the interrupt-disabled
region of code.  This prevents the RCU read-side critical section from
being preempted, and thus prevents the attempt to deboost with interrupts
disabled.

It is quite possible that a better long-term fix is to make rt_mutex_unlock()
disable irqs when acquiring the rt_mutex structure's ->wait_lock.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

The rcu_read_unlock_special() function relies on in_irq() to exclude
scheduler activity from interrupt level.  This fails because exit_irq()
can invoke the scheduler after clearing the preempt_count() bits that
in_irq() uses to determine that it is at interrupt level.  This situation
can result in failures as follows:

 $task			IRQ		SoftIRQ

 rcu_read_lock()

 /* do stuff */

 <preempt> |= UNLOCK_BLOCKED

 rcu_read_unlock()
   --t->rcu_read_lock_nesting

			irq_enter();
			/* do stuff, don't use RCU */
			irq_exit();
			  sub_preempt_count(IRQ_EXIT_OFFSET);
			  invoke_softirq()

					ttwu();
					  spin_lock_irq(&pi->lock)
					  rcu_read_lock();
					  /* do stuff */
					  rcu_read_unlock();
					    rcu_read_unlock_special()
					      rcu_report_exp_rnp()
					        ttwu()
					          spin_lock_irq(&pi->lock) /* deadlock */

   rcu_read_unlock_special(t);

Ed can simply trigger this 'easy' because invoke_softirq() immediately
does a ttwu() of ksoftirqd/# instead of doing the in-place softirq stuff
first, but even without that the above happens.

Cure this by also excluding softirqs from the
rcu_read_unlock_special() handler and ensuring the force_irqthreads
ksoftirqd/# wakeup is done from full softirq context.

[ Alternatively, delaying the ->rcu_read_lock_nesting decrement
  until after the special handling would make the thing more robust
  in the face of interrupts as well.  And there is a separate patch
  for that. ]

Cc: Thomas Gleixner <tglx@linutronix.de>
Reported-and-tested-by: Ed Tomlinson <edt@aei.ca>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

Ensure scheduler_ipi() calls irq_{enter,exit} when it does some actual
work. Traditionally we never did any actual work from the resched IPI
and all magic happened in the return from interrupt path.

Now that we do do some work, we need to ensure irq_{enter,exit} are
called so that we don't confuse things.

This affects things like timekeeping, NO_HZ and RCU, basically
everything with a hook in irq_enter/exit.

Explicit examples of things going wrong are:

  sched_clock_cpu() -- has a callback when leaving NO_HZ state to take
                    a new reading from GTOD and TSC. Without this
                    callback, time is stuck in the past.

  RCU -- needs in_irq() to work in order to avoid some nasty deadlocks
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

The addition of RCU read-side critical sections within runqueue and
priority-inheritance lock critical sections introduced some deadlock
cycles, for example, involving interrupts from __rcu_read_unlock()
where the interrupt handlers call wake_up().  This situation can cause
the instance of __rcu_read_unlock() invoked from interrupt to do some
of the processing that would otherwise have been carried out by the
task-level instance of __rcu_read_unlock().  When the interrupt-level
instance of __rcu_read_unlock() is called with a scheduler lock held
from interrupt-entry/exit situations where in_irq() returns false,
deadlock can result.

This commit resolves these deadlocks by using negative values of
the per-task ->rcu_read_lock_nesting counter to indicate that an
instance of __rcu_read_unlock() is in flight, which in turn prevents
instances from interrupt handlers from doing any special processing.
This patch is inspired by Steven Rostedt's earlier patch that similarly
made __rcu_read_unlock() guard against interrupt-mediated recursion
(see https://lkml.org/lkml/2011/7/15/326), but this commit refines
Steven's approach to avoid the need for preemption disabling on the
__rcu_read_unlock() fastpath and to also avoid the need for manipulating
a separate per-CPU variable.

This patch avoids need for preempt_disable() by instead using negative
values of the per-task ->rcu_read_lock_nesting counter.  Note that nested
rcu_read_lock()/rcu_read_unlock() pairs are still permitted, but they will
never see ->rcu_read_lock_nesting go to zero, and will therefore never
invoke rcu_read_unlock_special(), thus preventing them from seeing the
RCU_READ_UNLOCK_BLOCKED bit should it be set in ->rcu_read_unlock_special.
This patch also adds a check for ->rcu_read_unlock_special being negative
in rcu_check_callbacks(), thus preventing the RCU_READ_UNLOCK_NEED_QS
bit from being set should a scheduling-clock interrupt occur while
__rcu_read_unlock() is exiting from an outermost RCU read-side critical
section.

Of course, __rcu_read_unlock() can be preempted during the time that
->rcu_read_lock_nesting is negative.  This could result in the setting
of the RCU_READ_UNLOCK_BLOCKED bit after __rcu_read_unlock() checks it,
and would also result it this task being queued on the corresponding
rcu_node structure's blkd_tasks list.  Therefore, some later RCU read-side
critical section would enter rcu_read_unlock_special() to clean up --
which could result in deadlock if that critical section happened to be in
the scheduler where the runqueue or priority-inheritance locks were held.

This situation is dealt with by making rcu_preempt_note_context_switch()
check for negative ->rcu_read_lock_nesting, thus refraining from
queuing the task (and from setting RCU_READ_UNLOCK_BLOCKED) if we are
already exiting from the outermost RCU read-side critical section (in
other words, we really are no longer actually in that RCU read-side
critical section).  In addition, rcu_preempt_note_context_switch()
invokes rcu_read_unlock_special() to carry out the cleanup in this case,
which clears out the ->rcu_read_unlock_special bits and dequeues the task
(if necessary), in turn avoiding needless delay of the current RCU grace
period and needless RCU priority boosting.

It is still illegal to call rcu_read_unlock() while holding a scheduler
lock if the prior RCU read-side critical section has ever had either
preemption or irqs enabled.  However, the common use case is legal,
namely where then entire RCU read-side critical section executes with
irqs disabled, for example, when the scheduler lock is held across the
entire lifetime of the RCU read-side critical section.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

When creating sched_domains, stop when we've covered the entire
target span instead of continuing to create domains, only to
later find they're redundant and throw them away again.

This avoids single node systems from touching funny NUMA
sched_domain creation code and reduces the risks of the new
SD_OVERLAP code.
Requested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Anton Blanchard <anton@samba.org>
Cc: mahesh@linux.vnet.ibm.com
Cc: benh@kernel.crashing.org
Cc: linuxppc-dev@lists.ozlabs.org
Link: http://lkml.kernel.org/r/1311180177.29152.57.camel@twinsSigned-off-by: Ingo Molnar <mingo@elte.hu>

Allow for sched_domain spans that overlap by giving such domains their
own sched_group list instead of sharing the sched_groups amongst
each-other.

This is needed for machines with more than 16 nodes, because
sched_domain_node_span() will generate a node mask from the
16 nearest nodes without regard if these masks have any overlap.

Currently sched_domains have a sched_group that maps to their child
sched_domain span, and since there is no overlap we share the
sched_group between the sched_domains of the various CPUs. If however
there is overlap, we would need to link the sched_group list in
different ways for each cpu, and hence sharing isn't possible.

In order to solve this, allocate private sched_groups for each CPU's
sched_domain but have the sched_groups share a sched_group_power
structure such that we can uniquely track the power.
Reported-and-tested-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-08bxqw9wis3qti9u5inifh3y@git.kernel.orgSigned-off-by: Ingo Molnar <mingo@elte.hu>

In order to prepare for non-unique sched_groups per domain, we need to
carry the cpu_power elsewhere, so put a level of indirection in.
Reported-and-tested-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-qkho2byuhe4482fuknss40ad@git.kernel.orgSigned-off-by: Ingo Molnar <mingo@elte.hu>

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  ceph: fix file mode calculation

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/linux-arm-soc:
  davinci: DM365 EVM: fix video input mux bits
  ARM: davinci: Check for NULL return from irq_alloc_generic_chip
  arm: davinci: Fix low level gpio irq handlers' argument

I'm running a workload which triggers a lot of swap in a machine with 4
nodes.  After I kill the workload, I found a kswapd livelock.  Sometimes
kswapd3 or kswapd2 are keeping running and I can't access filesystem,
but most memory is free.

This looks like a regression since commit 08951e54 ("mm: vmscan:
correct check for kswapd sleeping in sleeping_prematurely").

Node 2 and 3 have only ZONE_NORMAL, but balance_pgdat() will return 0
for classzone_idx.  The reason is end_zone in balance_pgdat() is 0 by
default, if all zones have watermark ok, end_zone will keep 0.

Later sleeping_prematurely() always returns true.  Because this is an
order 3 wakeup, and if classzone_idx is 0, both balanced_pages and
present_pages in pgdat_balanced() are 0.  We add a special case here.
If a zone has no page, we think it's balanced.  This fixes the livelock.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Assume that /sys/kernel/debug/dummy64 is debugfs file created by
debugfs_create_x64().

	# cd /sys/kernel/debug
	# echo 0x1234567812345678 > dummy64
	# cat dummy64
	0x0000000012345678

	# echo 0x80000000 > dummy64
	# cat dummy64
	0xffffffff80000000

A value larger than INT_MAX cannot be written to the debugfs file created
by debugfs_create_u64 or debugfs_create_x64 on 32bit machine.  Because
simple_attr_write() uses simple_strtol() for the conversion.

To fix this, use simple_strtoll() instead.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  vfs: fix race in rcu lookup of pruned dentry
  Fix cifs_get_root()

[ Edited the last commit to get rid of a 'unused variable "seq"'
  warning due to Al editing the patch.  - Linus ]

Don't update *inode in __follow_mount_rcu() until we'd verified that
there is mountpoint there.  Kudos to Hugh Dickins for catching that
one in the first place and eventually figuring out the solution (and
catching a braino in the earlier version of patch).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Given some common flag combinations, particularly -Os, gcc will inline
rcu_read_unlock_special() despite its being in an unlikely() clause.
Use noinline to prohibit this misoptimization.

In addition, move the second barrier() in __rcu_read_unlock() so that
it is not on the common-case code path.  This will allow the compiler to
generate better code for the common-case path through __rcu_read_unlock().
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

The RCU_BOOST commits for TREE_PREEMPT_RCU introduced an other-task
write to a new RCU_READ_UNLOCK_BOOSTED bit in the task_struct structure's
->rcu_read_unlock_special field, but, as noted by Steven Rostedt, without
correctly synchronizing all accesses to ->rcu_read_unlock_special.
This could result in bits in ->rcu_read_unlock_special being spuriously
set and cleared due to conflicting accesses, which in turn could result
in deadlocks between the rcu_node structure's ->lock and the scheduler's
rq and pi locks.  These deadlocks would result from RCU incorrectly
believing that the just-ended RCU read-side critical section had been
preempted and/or boosted.  If that RCU read-side critical section was
executed with either rq or pi locks held, RCU's ensuing (incorrect)
calls to the scheduler would cause the scheduler to attempt to once
again acquire the rq and pi locks, resulting in deadlock.  More complex
deadlock cycles are also possible, involving multiple rq and pi locks
as well as locks from multiple rcu_node structures.

This commit fixes synchronization by creating ->rcu_boosted field in
task_struct that is accessed and modified only when holding the ->lock
in the rcu_node structure on which the task is queued (on that rcu_node
structure's ->blkd_tasks list).  This results in tasks accessing only
their own current->rcu_read_unlock_special fields, making unsynchronized
access once again legal, and keeping the rcu_read_unlock() fastpath free
of atomic instructions and memory barriers.

The reason that the rcu_read_unlock() fastpath does not need to access
the new current->rcu_boosted field is that this new field cannot
be non-zero unless the RCU_READ_UNLOCK_BLOCKED bit is set in the
current->rcu_read_unlock_special field.  Therefore, rcu_read_unlock()
need only test current->rcu_read_unlock_special: if that is zero, then
current->rcu_boosted must also be zero.

This bug does not affect TINY_PREEMPT_RCU because this implementation
of RCU accesses current->rcu_read_unlock_special with irqs disabled,
thus preventing races on the !SMP systems that TINY_PREEMPT_RCU runs on.
Maybe-reported-by: Dave Jones <davej@redhat.com>
Maybe-reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Steven Rostedt <rostedt@goodmis.org>

PREEMPT_RCU read-side critical sections blocking an expedited grace
period invoke rcu_report_exp_rnp().  When the last such critical section
has completed, rcu_report_exp_rnp() invokes the scheduler to wake up the
task that invoked synchronize_rcu_expedited() -- needlessly holding the
root rcu_node structure's lock while doing so, thus needlessly providing
a way for RCU and the scheduler to deadlock.

This commit therefore releases the root rcu_node structure's lock before
calling wake_up().
Reported-by: Ed Tomlinson <edt@aei.ca>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>