GitHub/MotorolaMobilityLLC/kernel-slsi.git
11 years agoblk-throttle: add throtl_qnode for dispatch fairness
Tejun Heo [Tue, 14 May 2013 20:52:38 +0000 (13:52 -0700)]
blk-throttle: add throtl_qnode for dispatch fairness

With flat hierarchy, there's only single level of dispatching
happening and fairness beyond that point is the responsibility of the
rest of the block layer and driver, which usually works out okay;
however, with the planned hierarchy support,
service_queue->bio_lists[] can be filled up by bios from a single
source.  While the limits would still be honored, it'd be very easy to
starve IOs from siblings or children.

To avoid such starvation, this patch implements throtl_qnode and
converts service_queue->bio_lists[] to lists of per-source qnodes
which in turn contains the bio's.  For example, when a bio is
dispatched from a child group, the bio doesn't get queued on
->bio_lists[] directly but it first gets queued on the group's qnode
which in turn gets queued on service_queue->queued[].  When
dispatching for the upper level, the ->queued[] list is consumed in
round-robing order so that the dispatch windows is consumed fairly by
all IO sources.

There are two ways a bio can come to a throtl_grp - directly queued to
the group or dispatched from a child.  For the former
throtl_grp->qnode_on_self[rw] is used.  For the latter, the child's
->qnode_on_parent[rw].

Note that this means that the child which is contributing a bio to its
parent should stay pinned until all its bios are dispatched to its
grand-parent.  This patch moves blkg refcnting from bio add/remove
spots to qnode activation/deactivation so that the blkg containing an
active qnode is always pinned.  As child pins the parent, this is
sufficient for keeping the relevant sub-tree pinned while bios are in
flight.

The starvation issue was spotted by Vivek Goyal.

v2: The original patch used the same throtl_grp->qnode_on_self/parent
    for reads and writes causing RWs to be queued incorrectly if there
    already are outstanding IOs in the other direction.  They should
    be throtl_grp->qnode_on_self/parent[2] so that READs and WRITEs
    can use different qnodes.  Spotted by Vivek Goyal.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: make throtl_pending_timer_fn() ready for hierarchy
Tejun Heo [Tue, 14 May 2013 20:52:38 +0000 (13:52 -0700)]
blk-throttle: make throtl_pending_timer_fn() ready for hierarchy

throtl_pending_timer_fn() currently assumes that the parent_sq is the
top level one and the bio's dispatched are ready to be issued;
however, this assumption will be wrong with proper hierarchy support.
This patch makes the following changes to make
throtl_pending_timer_fn() ready for hiearchy.

* If the parent_sq isn't the top-level one, update the parent
  throtl_grp's dispatch time and schedule the next dispatch as
  necessary.  If the parent's dispatch time is now, repeat the
  function for the parent throtl_grp.

* If the parent_sq is the top-level one, kick issue work_item as
  before.

* The debug message printed by throtl_log() now prints out the
  service_queue's nr_queued[] instead of the total nr_queued as the
  latter becomes uninteresting and misleading with hierarchical
  dispatch.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: make tg_dispatch_one_bio() ready for hierarchy
Tejun Heo [Tue, 14 May 2013 20:52:38 +0000 (13:52 -0700)]
blk-throttle: make tg_dispatch_one_bio() ready for hierarchy

tg_dispatch_one_bio() currently assumes that the parent_sq is the top
level one and the bio being dispatched is ready to be issued; however,
this assumption will be wrong with proper hierarchy support.  This
patch makes the following changes to make tg_dispatch_on_bio() ready
for hiearchy.

* throtl_data->nr_queued[] is incremented in blk_throtl_bio() instead
  of throtl_add_bio_tg() so that throtl_add_bio_tg() can be used to
  transfer a bio from a child tg to its parent.

* tg_dispatch_one_bio() is updated to distinguish whether its parent
  is another throtl_grp or the throtl_data.  If former, the bio is
  transferred to the parent throtl_grp using throtl_add_bio_tg().  If
  latter, the bio is ready to be issued and put on the top-level
  service_queue's bio_lists[] and throtl_data->nr_queued is
  decremented.

As all throtl_grps currently have the top level service_queue as their
->parent_sq, this patch in itself doesn't make any behavior
difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: make blk_throtl_bio() ready for hierarchy
Tejun Heo [Tue, 14 May 2013 20:52:38 +0000 (13:52 -0700)]
blk-throttle: make blk_throtl_bio() ready for hierarchy

Currently, blk_throtl_bio() issues the passed in bio directly if it's
within limits of its associated tg (throtl_grp).  This behavior
becomes incorrect with hierarchy support as the bio should be
accounted to and throttled by the ancestor throtl_grps too.

This patch makes the direct issue path of blk_throtl_bio() to loop
until it reaches the top-level service_queue or gets throttled.  If
the former, the bio can be issued directly; otherwise, it gets queued
at the first layer it was above limits.

As tg->parent_sq is always the top-level service queue currently, this
patch in itself doesn't make any behavior differences.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: make blk_throtl_drain() ready for hierarchy
Tejun Heo [Tue, 14 May 2013 20:52:37 +0000 (13:52 -0700)]
blk-throttle: make blk_throtl_drain() ready for hierarchy

The current blk_throtl_drain() assumes that all active throtl_grps are
queued on throtl_data->service_queue, which won't be true once
hierarchy support is implemented.

This patch makes blk_throtl_drain() perform post-order walk of the
blkg hierarchy draining each associated throtl_grp, which guarantees
that all bios will eventually be pushed to the top-level service_queue
in throtl_data.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: dispatch from throtl_pending_timer_fn()
Tejun Heo [Tue, 14 May 2013 20:52:37 +0000 (13:52 -0700)]
blk-throttle: dispatch from throtl_pending_timer_fn()

Currently, blk_throtl_dispatch_work_fn() is responsible for both
dispatching bio's from throtl_grp's according to their limits and then
issuing the dispatched bios.

This patch moves the dispatch part to throtl_pending_timer_fn() so
that the work item is kicked iff there are bio's to issue.  This is to
avoid work item execution at each step when hierarchy support is
enabled.  bio's will be dispatched towards the top-level service_queue
from the timers at each layer and the work item will only be used to
issue the bio's which reached the top-level service_queue.

While fetching bio's to issue from bio_lists[],
blk_throtl_dispatch_work_fn() fetches all READs before WRITEs.  While
the original code also dispatched READs first, if multiple throtl_grps
are dispatched on the same run, WRITEs from throtl_grp which is
dispatched first would precede READs from throtl_grps which are
dispatched later.  While this is a behavior change, given that the
previous code already prioritized READs and block layer generally
prioritizes and segregates READs from WRITEs, this isn't likely to
make any noticeable differences.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: implement dispatch looping
Tejun Heo [Tue, 14 May 2013 20:52:37 +0000 (13:52 -0700)]
blk-throttle: implement dispatch looping

throtl_select_dispatch() only dispatches throtl_quantum bios on each
invocation.  blk_throtl_dispatch_work_fn() in turn depends on
throtl_schedule_next_dispatch() scheduling the next dispatch window
immediately so that undue delays aren't incurred.  This effectively
chains multiple dispatch work item executions back-to-back when there
are more than throtl_quantum bios to dispatch on a given tick.

There is no reason to finish the current work item just to repeat it
immediately.  This patch makes throtl_schedule_next_dispatch() return
%false without doing anything if the current dispatch window is still
open and updates blk_throtl_dispatch_work_fn() repeat dispatching
after cpu_relax() on %false return.

This change will help implementing hierarchy support as dispatching
will be done from pending_timer and immediate reschedule of timer
function isn't supported and doesn't make much sense.

While this patch changes how dispatch behaves when there are more than
throtl_quantum bios to dispatch on a single tick, the behavior change
is immaterial.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: separate out throtl_service_queue->pending_timer from throtl_data-...
Tejun Heo [Tue, 14 May 2013 20:52:36 +0000 (13:52 -0700)]
blk-throttle: separate out throtl_service_queue->pending_timer from throtl_data->dispatch_work

Currently, throtl_data->dispatch_work is a delayed_work item which
handles both delayed dispatch and issuing bios.  The two tasks will be
separated to support proper hierarchy.  To prepare for that, this
patch separates out the timer into throtl_service_queue->pending_timer
from throtl_data->dispatch_work and make the latter a work_struct.

* As the timer is now per-service_queue, it's initialized and
  del_sync'd as its corresponding service_queue is created and
  destroyed.  The timer, when triggered, simply schedules
  throtl_data->dispathc_work for execution.

* throtl_schedule_delayed_work() is renamed to
  throtl_schedule_pending_timer() and takes @sq and @expires now.

* Simiarly, throtl_schedule_next_dispatch() now takes @sq, which
  should be the parent_sq of the service_queue which just got a new
  bio or updated.  As the parent_sq is always the top-level
  service_queue now, this doesn't change anything at this point.

This patch doesn't introduce any behavior differences.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: set REQ_THROTTLED from throtl_charge_bio() and gate stats update with it
Tejun Heo [Tue, 14 May 2013 20:52:36 +0000 (13:52 -0700)]
blk-throttle: set REQ_THROTTLED from throtl_charge_bio() and gate stats update with it

With proper hierarchy support, a bio can be dispatched multiple times
until it reaches the top-level service_queue and we don't want to
update dispatch stats at each step.  They are local stats and will be
kept local.  If recursive stats are necessary, they should be
implemented separately and definitely not by updating counters
recursively on each dispatch.

This patch moves REQ_THROTTLED setting to throtl_charge_bio() and gate
stats update with it so that dispatch stats are updated only on the
first time the bio is charged to a throtl_grp, which will always be
the throtl_grp the bio was originally queued to.

This means that REQ_THROTTLED would be set even for bios which don't
get throttled.  As we don't want bios to leave blk-throtl with the
flag set, move REQ_THROTLLED clearing to the end of blk_throtl_bio()
and clear if the bio is being issued directly.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: implement sq_to_tg(), sq_to_td() and throtl_log()
Tejun Heo [Tue, 14 May 2013 20:52:36 +0000 (13:52 -0700)]
blk-throttle: implement sq_to_tg(), sq_to_td() and throtl_log()

Now that both throtl_data and throtl_grp embed throtl_service_queue,
we can unify throtl_log() and throtl_log_tg().

* sq_to_tg() is added.  This returns the throtl_grp a service_queue is
  embedded in.  If the service_queue is the top-level one embedded in
  throtl_data, NULL is returned.

* sq_to_td() is added.  A service_queue is always associated with a
  throtl_data.  This function finds the associated td and returns it.

* throtl_log() is updated to take throtl_service_queue instead of
  throtl_data.  If the service_queue is one embedded in throtl_grp, it
  prints the same header as throtl_log_tg() did.  If it's one embedded
  in throtl_data, it behaves the same as before.  This renders
  throtl_log_tg() unnecessary.  Removed.

This change is necessary for hierarchy support as we're gonna be using
the same code paths to dispatch bios to intermediate service_queues
embedded in throtl_grps and the top-level service_queue embedded in
throtl_data.

This patch doesn't make any behavior changes.

v2: throtl_log() didn't print a space after blkg path.  Updated so
    that it prints a space after throtl_grp path.  Spotted by Vivek.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: add throtl_service_queue->parent_sq
Tejun Heo [Tue, 14 May 2013 20:52:36 +0000 (13:52 -0700)]
blk-throttle: add throtl_service_queue->parent_sq

To prepare for hierarchy support, this patch adds
throtl_service_queue->service_sq which points to the arent
service_queue.  Currently, for all service_queues embedded in
throtl_grps, it points to throtl_data->service_queue.  As
throtl_data->service_queue doesn't have a parent its parent_sq is set
to NULL.

There are a number of functions which take both throtl_grp *tg and
throtl_service_queue *parent_sq.  With this patch, the parent
service_queue can be determined from @tg and the @parent_sq arguments
are removed.

This patch doesn't make any behavior differences.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: generalize update_disptime optimization in blk_throtl_bio()
Tejun Heo [Tue, 14 May 2013 20:52:35 +0000 (13:52 -0700)]
blk-throttle: generalize update_disptime optimization in blk_throtl_bio()

When blk_throtl_bio() wants to queue a bio to a tg (throtl_grp), it
avoids invoking tg_update_disptime() and
throtl_schedule_next_dispatch() if the tg already has bios queued in
that direction.  As a new bio is appeneded after the existing ones, it
can't change the tg's next dispatch time or the parent's dispatch
schedule.

This optimization is currently open coded in blk_throtl_bio().
Whether the target biolist was occupied was recorded in a local
variable and later used to skip disptime update.  This patch moves
generalizes it so that throtl_add_bio_tg() sets a new flag
THROTL_TG_WAS_EMPTY if the biolist was empty before the new bio was
added.  tg_update_disptime() clears the flag automatically.
blk_throtl_bio() is updated to simply test the flag before updating
disptime.

This patch doesn't make any functional differences now but will enable
using the same optimization for recursive dispatch.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: dispatch to throtl_data->service_queue.bio_lists[]
Tejun Heo [Tue, 14 May 2013 20:52:35 +0000 (13:52 -0700)]
blk-throttle: dispatch to throtl_data->service_queue.bio_lists[]

throtl_service_queues will eventually form a tree which is anchored at
throtl_data->service_queue and queue bios will climb the tree to the
top service_queue to be executed.

This patch makes the dispatch paths in blk_throtl_dispatch_work_fn()
and blk_throtl_drain() to dispatch bios to
throtl_data->service_queue.bio_lists[] instead of the on-stack
bio_lists.  This will keep the final dispatch to the top level
service_queue share the same mechanism as dispatches through the rest
of the hierarchy.

As bio's should be issued in a sleepable context,
blk_throtl_dispatch_work_fn() transfers all dispatched bio's from the
service_queue bio_lists[] into an onstack one before dropping
queue_lock and issuing the bio's.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: move bio_lists[] and friends to throtl_service_queue
Tejun Heo [Tue, 14 May 2013 20:52:35 +0000 (13:52 -0700)]
blk-throttle: move bio_lists[] and friends to throtl_service_queue

throtl_service_queues will eventually form a tree which is anchored at
throtl_data->service_queue and queue bios will climb the tree to the
top service_queue to be executed.

This patch moves bio_lists[] and nr_queued[] from throtl_grp to its
service_queue to prepare for that.  As currently only the
throtl_data->service_queue is in use, this patch just ends up moving
throtl_grp->bio_lists[] and ->nr_queued[] to
throtl_grp->service_queue.bio_lists[] and ->nr_queued[] without making
any functional differences.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: add throtl_grp->service_queue
Tejun Heo [Tue, 14 May 2013 20:52:34 +0000 (13:52 -0700)]
blk-throttle: add throtl_grp->service_queue

Currently, there's single service_queue per queue -
throtl_data->service_queue.  All active throtl_grp's are queued on the
queue and dispatched according to their limits.  To support hierarchy,
this will be expanded such that active throtl_grp's form a tree
anchored at throtl_data->service_queue and chained through each
intermediate throtl_grp's service_queue.

This patch adds throtl_grp->service_queue to prepare for hierarchy
support.  The initialization function - throtl_service_queue_init() -
is added and replaces the macro initializer.  The newly added
tg->service_queue isn't used yet.  Following patches will do.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: reorganize throtl_service_queue passed around as argument
Tejun Heo [Tue, 14 May 2013 20:52:33 +0000 (13:52 -0700)]
blk-throttle: reorganize throtl_service_queue passed around as argument

throtl_service_queue will be the building block of hierarchy support
and will form a tree.  This patch updates its usages as arguments to
reduce confusion.

* When a service queue is used as the parent role - the host of the
  rbtree - use @parent_sq instead of @sq.

* For functions taking both @tg and @parent_sq, reorder them so that
  the order is (@tg, @parent_sq) not the other way around.  This makes
  the code follow the usual convention of specifying the primary
  target of the operation as the first argument.

This patch doesn't make any functional differences.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: pass around throtl_service_queue instead of throtl_data
Tejun Heo [Tue, 14 May 2013 20:52:33 +0000 (13:52 -0700)]
blk-throttle: pass around throtl_service_queue instead of throtl_data

throtl_service_queue will be used as the basic block to implement
hierarchy support.  Pass around throtl_service_queue *sq instead of
throtl_data *td in the following functions which will be used across
multiple levels of hierarchy.

* [__]throtl_enqueue/dequeue_tg()

* throtl_add_bio_tg()

* tg_update_disptime()

* throtl_select_dispatch()

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: add backlink pointer from throtl_grp to throtl_data
Tejun Heo [Tue, 14 May 2013 20:52:32 +0000 (13:52 -0700)]
blk-throttle: add backlink pointer from throtl_grp to throtl_data

Add throtl_grp->td so that the td (throtl_data) a given tg
(throtl_grp) belongs to can be determined, and remove @td argument
from functions which take both @td and @tg as the former now can be
determined from the latter.

This generally simplifies the code and removes a number of cases where
@td is passed as an argument without being actually used.  This will
also help hierarchy support implementation.

While at it, in multi-line conditions, move the logical operators
leading broken lines to the end of the previous line.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: simplify throtl_grp flag handling
Tejun Heo [Tue, 14 May 2013 20:52:32 +0000 (13:52 -0700)]
blk-throttle: simplify throtl_grp flag handling

blk-throttle is still using function-defining macros to define flag
handling functions, which went out style at least a decade ago.

Just define the flag as bitmask and use direct bit operations.

This patch doesn't make any functional changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: rename throtl_rb_root to throtl_service_queue
Tejun Heo [Tue, 14 May 2013 20:52:32 +0000 (13:52 -0700)]
blk-throttle: rename throtl_rb_root to throtl_service_queue

throtl_rb_root will be expanded to cover more roles for hierarchy
support.  Rename it to throtl_service_queue and make its fields more
descriptive.

* rb -> pending_tree
* left -> first_pending
* count -> nr_pending
* min_disptime -> first_pending_disptime

This patch is purely cosmetic.

Signed-off-by: Tejun Heo <tj@kernel.org
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: remove pointless throtl_nr_queued() optimizations
Tejun Heo [Tue, 14 May 2013 20:52:32 +0000 (13:52 -0700)]
blk-throttle: remove pointless throtl_nr_queued() optimizations

throtl_nr_queued() is used in several places to avoid performing
certain operations when the throtl_data is empty.  This usually is
useless as those paths usually aren't traveled if there's no bio
queued.

* throtl_schedule_delayed_work() skips scheduling dispatch work item
  if @td doesn't have any bios queued; however, the only case it can
  be called when @td is empty is from tg_set_conf() which isn't
  something we should be optimizing for.

* throtl_schedule_next_dispatch() takes a quick exit if @td is empty;
  however, right after that it triggers BUG if the service tree is
  empty.  The two conditions are equivalent and it can just test
  @st->count for the quick exit.

* blk_throtl_dispatch_work_fn() skips dispatch if @td is empty.  This
  work function isn't usually invoked when @td is empty.  The only
  possibility is from tg_set_conf() and when it happens the normal
  dispatching path can handle empty @td fine.  No need to add special
  skip path.

This patch removes the above three unnecessary optimizations, which
leave throtl_log() call in blk_throtl_dispatch_work_fn() the only user
of throtl_nr_queued().  Remove throtl_nr_queued() and open code it in
throtl_log().  I don't think we need td->nr_queued[] at all.  Maybe we
can remove it later.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: relocate throtl_schedule_delayed_work()
Tejun Heo [Tue, 14 May 2013 20:52:31 +0000 (13:52 -0700)]
blk-throttle: relocate throtl_schedule_delayed_work()

Move throtl_schedule_delayed_work() above its first user so that the
forward declaration can be removed.

This patch is pure relocaiton.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: collapse throtl_dispatch() into the work function
Tejun Heo [Tue, 14 May 2013 20:52:31 +0000 (13:52 -0700)]
blk-throttle: collapse throtl_dispatch() into the work function

blk-throttle is about to go through major restructuring to support
hierarchy.  Do cosmetic updates in preparation.

* s/throtl_data->throtl_work/throtl_data->dispatch_work/

* s/blk_throtl_work()/blk_throtl_dispatch_work_fn()/

* Collapse throtl_dispatch() into blk_throtl_dispatch_work_fn()

This patch is purely cosmetic.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: remove deferred config application mechanism
Tejun Heo [Tue, 14 May 2013 20:52:31 +0000 (13:52 -0700)]
blk-throttle: remove deferred config application mechanism

When bps or iops configuration changes, blk-throttle records the new
configuration and sets a flag indicating that the config has changed.
The flag is checked in the bio dispatch path and applied.  This
deferred config application was necessary due to limitations in blkcg
framework, which haven't existed for quite a while now.

This patch removes the deferred config application mechanism and
applies new configurations directly from tg_set_conf(), which is
simpler.

v2: Dropped unnecessary throtl_schedule_delayed_work() call from
    tg_set_conf() as suggested by Vivek Goyal.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblk-throttle: remove spurious throtl_enqueue_tg() call from throtl_select_dispatch()
Tejun Heo [Tue, 14 May 2013 20:52:31 +0000 (13:52 -0700)]
blk-throttle: remove spurious throtl_enqueue_tg() call from throtl_select_dispatch()

throtl_select_dispatch() calls throtl_enqueue_tg() right after
tg_update_disptime(), which always calls the function anyway.  The
call is, while harmless, unnecessary.  Remove it.

This patch doesn't introduce any behavior difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblkcg: move bulk of blkcg_gq release operations to the RCU callback
Tejun Heo [Tue, 14 May 2013 20:52:31 +0000 (13:52 -0700)]
blkcg: move bulk of blkcg_gq release operations to the RCU callback

Currently, when the last reference of a blkcg_gq is put, all then
release operations sans the actual freeing happen directly in
blkg_put().  As blkg_put() may be called under queue_lock, all
pd_exit_fn()s may be too.  This makes it impossible for pd_exit_fn()s
to use del_timer_sync() on timers which grab the queue_lock which is
an irq-safe lock due to the deadlock possibility described in the
comment on top of del_timer_sync().

This can be easily avoided by perfoming the release operations in the
RCU callback instead of directly from blkg_put().  This patch moves
the blkcg_gq release operations to the RCU callback.

As this leaves __blkg_release() with only call_rcu() invocation,
blkg_rcu_free() is renamed to __blkg_release_rcu(), exported and
call_rcu() invocation is now done directly from blkg_put() instead of
going through __blkg_release() which is removed.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblkcg: invoke blkcg_policy->pd_init() after parent is linked
Tejun Heo [Tue, 14 May 2013 20:52:31 +0000 (13:52 -0700)]
blkcg: invoke blkcg_policy->pd_init() after parent is linked

Currently, when creating a new blkcg_gq, each policy's pd_init_fn() is
invoked in blkg_alloc() before the parent is linked.  This makes it
difficult for policies to perform initializations which are dependent
on the parent.

This patch moves pd_init_fn() invocations to blkg_create() after the
parent blkg is linked where the new blkg is fully initialized.  As
this means that blkg_free() can't assume that pd's are initialized,
pd_exit_fn() invocations are moved to __blkg_release().  This
guarantees that pd_exit_fn() is also invoked with fully initialized
blkgs with valid parent pointers.

This will help implementing hierarchy support in blk-throttle.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblkcg: implement blkg_for_each_descendant_post()
Tejun Heo [Tue, 14 May 2013 20:52:31 +0000 (13:52 -0700)]
blkcg: implement blkg_for_each_descendant_post()

This will be used by blk-throttle hierarchy support.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblkcg: move blkg_for_each_descendant_pre() to block/blk-cgroup.h
Tejun Heo [Tue, 14 May 2013 20:52:30 +0000 (13:52 -0700)]
blkcg: move blkg_for_each_descendant_pre() to block/blk-cgroup.h

blk-throttle hierarchy support will make use of it.  Move
blkg_for_each_descendant_pre() from block/blk-cgroup.c to
block/blk-cgroup.h.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoblkcg: fix error return path in blkg_create()
Tejun Heo [Tue, 14 May 2013 20:52:30 +0000 (13:52 -0700)]
blkcg: fix error return path in blkg_create()

In blkg_create(), after lookup of parent fails, the control jumps to
error path with the error code encoded into @blkg.  The error path
doesn't use @blkg for the return value.  It returns ERR_PTR(ret).
Make lookup fail path set @ret instead of @blkg.

Note that the parent lookup is guaranteed to succeed at that point and
the condition check is purely for sanity and triggers WARN when fails.
As such, I don't think it's necessary to mark it for -stable.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
11 years agoLinux 3.10-rc1
Linus Torvalds [Sun, 12 May 2013 00:14:08 +0000 (17:14 -0700)]
Linux 3.10-rc1

11 years agoMerge tag 'trace-fixes-v3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rosted...
Linus Torvalds [Sun, 12 May 2013 00:04:59 +0000 (17:04 -0700)]
Merge tag 'trace-fixes-v3.10' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing/kprobes update from Steven Rostedt:
 "The majority of these changes are from Masami Hiramatsu bringing
  kprobes up to par with the latest changes to ftrace (multi buffering
  and the new function probes).

  He also discovered and fixed some bugs in doing so.  When pulling in
  his patches, I also found a few minor bugs as well and fixed them.

  This also includes a compile fix for some archs that select the ring
  buffer but not tracing.

  I based this off of the last patch you took from me that fixed the
  merge conflict error, as that was the commit that had all the changes
  I needed for this set of changes."

* tag 'trace-fixes-v3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing/kprobes: Support soft-mode disabling
  tracing/kprobes: Support ftrace_event_file base multibuffer
  tracing/kprobes: Pass trace_probe directly from dispatcher
  tracing/kprobes: Increment probe hit-count even if it is used by perf
  tracing/kprobes: Use bool for retprobe checker
  ftrace: Fix function probe when more than one probe is added
  ftrace: Fix the output of enabled_functions debug file
  ftrace: Fix locking in register_ftrace_function_probe()
  tracing: Add helper function trace_create_new_event() to remove duplicate code
  tracing: Modify soft-mode only if there's no other referrer
  tracing: Indicate enabled soft-mode in enable file
  tracing/kprobes: Fix to increment return event probe hit-count
  ftrace: Cleanup regex_lock and ftrace_lock around hash updating
  ftrace, kprobes: Fix a deadlock on ftrace_regex_lock
  ftrace: Have ftrace_regex_write() return either read or error
  tracing: Return error if register_ftrace_function_probe() fails for event_enable_func()
  tracing: Don't succeed if event_enable_func did not register anything
  ring-buffer: Select IRQ_WORK

11 years agoMerge tag 'stable/for-linus-3.10-rc0-tag-two' of git://git.kernel.org/pub/scm/linux...
Linus Torvalds [Sat, 11 May 2013 23:19:30 +0000 (16:19 -0700)]
Merge tag 'stable/for-linus-3.10-rc0-tag-two' of git://git./linux/kernel/git/konrad/xen

Pull Xen bug-fixes from Konrad Rzeszutek Wilk:
 - More fixes in the vCPU PVHVM hotplug path.
 - Add more documentation.
 - Fix various ARM related issues in the Xen generic drivers.
 - Updates in the xen-pciback driver per Bjorn's updates.
 - Mask the x2APIC feature for PV guests.

* tag 'stable/for-linus-3.10-rc0-tag-two' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
  xen/pci: Used cached MSI-X capability offset
  xen/pci: Use PCI_MSIX_TABLE_BIR, not PCI_MSIX_FLAGS_BIRMASK
  xen: clear IRQ_NOAUTOEN and IRQ_NOREQUEST
  xen: mask x2APIC feature in PV
  xen: SWIOTLB is only used on x86
  xen/spinlock: Fix check from greater than to be also be greater or equal to.
  xen/smp/pvhvm: Don't point per_cpu(xen_vpcu, 33 and larger) to shared_info
  xen/vcpu: Document the xen_vcpu_info and xen_vcpu
  xen/vcpu/pvhvm: Fix vcpu hotplugging hanging.

11 years agoMerge tag 'scsi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb...
Linus Torvalds [Sat, 11 May 2013 22:24:22 +0000 (15:24 -0700)]
Merge tag 'scsi-for-linus' of git://git./linux/kernel/git/jejb/scsi

Pull second SCSI update from James "Jaj B" Bottomley:
 "This is the final round of SCSI patches for the merge window.  It
  consists mostly of driver updates (bnx2fc, ibmfc, fnic, lpfc,
  be2iscsi, pm80xx, qla4x and ipr).

  There's also the power management updates that complete the patches in
  Jens' tree, an iscsi refcounting problem fix from the last pull, some
  dif handling in scsi_debug fixes, a few nice code cleanups and an
  error handling busy bug fix."

* tag 'scsi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (92 commits)
  [SCSI] qla2xxx: Update firmware link in Kconfig file.
  [SCSI] iscsi class, qla4xxx: fix sess/conn refcounting when find fns are used
  [SCSI] sas: unify the pointlessly separated enums sas_dev_type and sas_device_type
  [SCSI] pm80xx: thermal, sas controller config and error handling update
  [SCSI] pm80xx: NCQ error handling changes
  [SCSI] pm80xx: WWN Modification for PM8081/88/89 controllers
  [SCSI] pm80xx: Changed module name and debug messages update
  [SCSI] pm80xx: Firmware flash memory free fix, with addition of new memory region for it
  [SCSI] pm80xx: SPC new firmware changes for device id 0x8081 alone
  [SCSI] pm80xx: Added SPCv/ve specific hardware functionalities and relevant changes in common files
  [SCSI] pm80xx: MSI-X implementation for using 64 interrupts
  [SCSI] pm80xx: Updated common functions common for SPC and SPCv/ve
  [SCSI] pm80xx: Multiple inbound/outbound queue configuration
  [SCSI] pm80xx: Added SPCv/ve specific ids, variables and modify for SPC
  [SCSI] lpfc: fix up Kconfig dependencies
  [SCSI] Handle MLQUEUE busy response in scsi_send_eh_cmnd
  [SCSI] sd: change to auto suspend mode
  [SCSI] sd: use REQ_PM in sd's runtime suspend operation
  [SCSI] qla4xxx: Fix iocb_cnt calculation in qla4xxx_send_mbox_iocb()
  [SCSI] ufs: Correct the expected data transfersize
  ...

11 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux
Linus Torvalds [Sat, 11 May 2013 22:23:17 +0000 (15:23 -0700)]
Merge branch 'release' of git://git./linux/kernel/git/lenb/linux

Pull idle update from Len Brown:
 "Add support for new Haswell-ULT CPU idle power states"

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux:
  intel_idle: initial C8, C9, C10 support
  tools/power turbostat: display C8, C9, C10 residency

11 years agoMerge git://git.infradead.org/users/eparis/audit
Linus Torvalds [Sat, 11 May 2013 21:29:11 +0000 (14:29 -0700)]
Merge git://git.infradead.org/users/eparis/audit

Pull audit changes from Eric Paris:
 "Al used to send pull requests every couple of years but he told me to
  just start pushing them to you directly.

  Our touching outside of core audit code is pretty straight forward.  A
  couple of interface changes which hit net/.  A simple argument bug
  calling audit functions in namei.c and the removal of some assembly
  branch prediction code on ppc"

* git://git.infradead.org/users/eparis/audit: (31 commits)
  audit: fix message spacing printing auid
  Revert "audit: move kaudit thread start from auditd registration to kaudit init"
  audit: vfs: fix audit_inode call in O_CREAT case of do_last
  audit: Make testing for a valid loginuid explicit.
  audit: fix event coverage of AUDIT_ANOM_LINK
  audit: use spin_lock in audit_receive_msg to process tty logging
  audit: do not needlessly take a lock in tty_audit_exit
  audit: do not needlessly take a spinlock in copy_signal
  audit: add an option to control logging of passwords with pam_tty_audit
  audit: use spin_lock_irqsave/restore in audit tty code
  helper for some session id stuff
  audit: use a consistent audit helper to log lsm information
  audit: push loginuid and sessionid processing down
  audit: stop pushing loginid, uid, sessionid as arguments
  audit: remove the old depricated kernel interface
  audit: make validity checking generic
  audit: allow checking the type of audit message in the user filter
  audit: fix build break when AUDIT_DEBUG == 2
  audit: remove duplicate export of audit_enabled
  Audit: do not print error when LSMs disabled
  ...

11 years agoMerge branch 'for-3.10' of git://linux-nfs.org/~bfields/linux
Linus Torvalds [Fri, 10 May 2013 16:28:55 +0000 (09:28 -0700)]
Merge branch 'for-3.10' of git://linux-nfs.org/~bfields/linux

Pull nfsd fixes from Bruce Fields:
 "Small fixes for two bugs and two warnings"

* 'for-3.10' of git://linux-nfs.org/~bfields/linux:
  nfsd: fix oops when legacy_recdir_name_error is passed a -ENOENT error
  SUNRPC: fix decoding of optional gss-proxy xdr fields
  SUNRPC: Refactor gssx_dec_option_array() to kill uninitialized warning
  nfsd4: don't allow owner override on 4.1 CLAIM_FH opens

11 years agoMerge branch 'for_linus' of git://cavan.codon.org.uk/platform-drivers-x86
Linus Torvalds [Fri, 10 May 2013 16:27:40 +0000 (09:27 -0700)]
Merge branch 'for_linus' of git://cavan.codon.org.uk/platform-drivers-x86

Pull x86 platform drivers from Matthew Garrett:
 "Small set of updates, mainly trivial bugfixes and some small updates
  to deal with newer hardware.

  There's also a new driver that allows qemu guests to notify the
  hypervisor that they've just paniced, which seems useful."

* 'for_linus' of git://cavan.codon.org.uk/platform-drivers-x86:
  Add support for fan button on Ideapad Z580
  pvpanic: pvpanic device driver
  asus-nb-wmi: set wapf=4 for ASUSTeK COMPUTER INC. X75A
  drivers: platform: x86: Use PTR_RET function
  sony-laptop: SVS151290S kbd backlight and gfx switch support
  hp-wmi: add more definitions for new event_id's
  dell-laptop: Fix krealloc() misuse in parse_da_table()
  hp_accel: Ignore the error from lis3lv02d_poweron() at resume
  dell: add new dell WMI format for the AIO machines

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/signal
Linus Torvalds [Fri, 10 May 2013 16:21:05 +0000 (09:21 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/signal

Pull stray syscall bits from Al Viro:
 "Several syscall-related commits that were missing from the original"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/signal:
  switch compat_sys_sysctl to COMPAT_SYSCALL_DEFINE
  unicore32: just use mmap_pgoff()...
  unify compat fanotify_mark(2), switch to COMPAT_SYSCALL_DEFINE
  x86, vm86: fix VM86 syscalls: use SYSCALL_DEFINEx(...)

11 years agoMerge tag 'ecryptfs-3.10-rc1-ablkcipher' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 10 May 2013 16:20:01 +0000 (09:20 -0700)]
Merge tag 'ecryptfs-3.10-rc1-ablkcipher' of git://git./linux/kernel/git/tyhicks/ecryptfs

Pull eCryptfs update from Tyler Hicks:
 "Improve performance when AES-NI (and most likely other crypto
  accelerators) is available by moving to the ablkcipher crypto API.
  The improvement is more apparent on faster storage devices.

  There's no noticeable change when hardware crypto is not available"

* tag 'ecryptfs-3.10-rc1-ablkcipher' of git://git.kernel.org/pub/scm/linux/kernel/git/tyhicks/ecryptfs:
  eCryptfs: Use the ablkcipher crypto API

11 years agoMerge tag 'for-linus-20130509' of git://git.infradead.org/~dwmw2/random-2.6
Linus Torvalds [Fri, 10 May 2013 16:09:47 +0000 (09:09 -0700)]
Merge tag 'for-linus-20130509' of git://git.infradead.org/~dwmw2/random-2.6

Pull misc fixes from David Woodhouse:
 "This is some miscellaneous cleanups that don't really belong anywhere
  else (or were ignored), that have been sitting in linux-next for some
  time.  Two of them are fixes resulting from my audit of krealloc()
  usage that don't seem to have elicited any response when I posted
  them, and the other three are patches from Artem removing dead code."

* tag 'for-linus-20130509' of git://git.infradead.org/~dwmw2/random-2.6:
  pcmcia: remove RPX board stuff
  m68k: remove rpxlite stuff
  pcmcia: remove Motorola MBX860 support
  params: Fix potential memory leak in add_sysfs_param()
  dell-laptop: Fix krealloc() misuse in parse_da_table()

11 years agoMerge tag 'kvm-3.10-2' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 10 May 2013 16:08:21 +0000 (09:08 -0700)]
Merge tag 'kvm-3.10-2' of git://git./virt/kvm/kvm

Pull kvm fixes from Gleb Natapov:
 "Most of the fixes are in the emulator since now we emulate more than
  we did before for correctness sake we see more bugs there, but there
  is also an OOPS fixed and corruption of xcr0 register."

* tag 'kvm-3.10-2' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: emulator: emulate SALC
  KVM: emulator: emulate XLAT
  KVM: emulator: emulate AAM
  KVM: VMX: fix halt emulation while emulating invalid guest sate
  KVM: Fix kvm_irqfd_init initialization
  KVM: x86: fix maintenance of guest/host xcr0 state

11 years agoMerge tag 'dm-3.10-changes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/agk...
Linus Torvalds [Fri, 10 May 2013 16:02:50 +0000 (09:02 -0700)]
Merge tag 'dm-3.10-changes-2' of git://git./linux/kernel/git/agk/linux-dm

Pull device-mapper updates from Alasdair Kergon:
 "Allow devices that hold metadata for the device-mapper thin
  provisioning target to be extended easily; allow WRITE SAME on
  multipath devices; an assortment of little fixes and clean-ups."

* tag 'dm-3.10-changes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm: (21 commits)
  dm cache: set config value
  dm cache: move config fns
  dm thin: generate event when metadata threshold passed
  dm persistent metadata: add space map threshold callback
  dm persistent data: add threshold callback to space map
  dm thin: detect metadata device resizing
  dm persistent data: support space map resizing
  dm thin: open dev read only when possible
  dm thin: refactor data dev resize
  dm cache: replace memcpy with struct assignment
  dm cache: fix typos in comments
  dm cache policy: fix description of lookup fn
  dm: document iterate_devices
  dm persistent data: fix error message typos
  dm cache: tune migration throttling
  dm mpath: enable WRITE SAME support
  dm table: fix write same support
  dm bufio: avoid a possible __vmalloc deadlock
  dm snapshot: fix error return code in snapshot_ctr
  dm cache: fix error return code in cache_create
  ...

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Fri, 10 May 2013 16:00:39 +0000 (09:00 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/hid

Pull HID fixes from Jiri Kosina:

 - fix usage of sleeping lock in atomic context from Jiri Kosina

 - build fix for hid-steelseries under certain .config setups by Simon Wood

 - simple mismerge fix from Fernando Luis Vázquez Cao

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
  HID: debug: fix RCU preemption issue
  HID: hid-steelseries fix led class build issue
  HID: reintroduce fix-up for certain Sony RF receivers

11 years agoMerge branch 'postmerge' into for-linus
James Bottomley [Fri, 10 May 2013 14:54:01 +0000 (07:54 -0700)]
Merge branch 'postmerge' into for-linus

Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years agoMerge branch 'misc' into for-linus
James Bottomley [Fri, 10 May 2013 14:53:40 +0000 (07:53 -0700)]
Merge branch 'misc' into for-linus

Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years agoMerge tag 'sound-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
Linus Torvalds [Fri, 10 May 2013 14:51:56 +0000 (07:51 -0700)]
Merge tag 'sound-3.10' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "This contains small fixes since the previous pull request:

   - A few regression fixes and small updates of HD-audio

   - Yet another fix for Haswell HDMI audio

   - A copule of trivial fixes in ASoC McASP, DPAM and WM8994"

* tag 'sound-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  Revert "ALSA: hda - Don't set up active streams twice"
  ALSA: Add comment for control TLV API
  ALSA: hda - Apply pin-enablement workaround to all Haswell HDMI codecs
  ALSA: HDA: Fix Oops caused by dereference NULL pointer
  ALSA: mips/sgio2audio: Remove redundant platform_set_drvdata()
  ALSA: mips/hal2: Remove redundant platform_set_drvdata()
  ALSA: hda - Fix 3.9 regression of EAPD init on Conexant codecs
  sound: Fix make allmodconfig on MIPS
  ALSA: hda - Fix system panic when DMA > 40 bits for Nvidia audio controllers
  ALSA: atmel: Remove redundant platform_set_drvdata()
  ASoC: McASP: Fix receive clock polarity in DAIFMT_NB_NF mode.
  ASoC: wm8994: missing break in wm8994_aif3_hw_params()
  ASoC: McASP: Add pins output direction for rx clocks when configured in CBS_CFS format
  ASoC: dapm: use clk_prepare_enable and clk_disable_unprepare

11 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Fri, 10 May 2013 14:48:05 +0000 (07:48 -0700)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS updates from Ralf Baechle:

 - More work on DT support for various platforms

 - Various fixes that were to late to make it straight into 3.9

 - Improved platform support, in particular the Netlogic XLR and
   BCM63xx, and the SEAD3 and Malta eval boards.

 - Support for several Ralink SOC families.

 - Complete support for the microMIPS ASE which basically reencodes the
   existing MIPS32/MIPS64 ISA to use non-constant size instructions.

 - Some fallout from LTO work which remove old cruft and will generally
   make the MIPS kernel easier to maintain and resistant to compiler
   optimization, even in absence of LTO.

 - KVM support.  While MIPS has announced hardware virtualization
   extensions this KVM extension uses trap and emulate mode for
   virtualization of MIPS32.  More KVM work to add support for VZ
   hardware virtualizaiton extensions and MIPS64 will probably already
   be merged for 3.11.

Most of this has been sitting in -next for a long time.  All defconfigs
have been build or run time tested except three for which fixes are being
sent by other maintainers.

Semantic conflict with kvm updates done as per Ralf

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus: (118 commits)
  MIPS: Add new GIC clockevent driver.
  MIPS: Formatting clean-ups for clocksources.
  MIPS: Refactor GIC clocksource code.
  MIPS: Move 'gic_frequency' to common location.
  MIPS: Move 'gic_present' to common location.
  MIPS: MIPS16e: Add unaligned access support.
  MIPS: MIPS16e: Support handling of delay slots.
  MIPS: MIPS16e: Add instruction formats.
  MIPS: microMIPS: Optimise 'strnlen' core library function.
  MIPS: microMIPS: Optimise 'strlen' core library function.
  MIPS: microMIPS: Optimise 'strncpy' core library function.
  MIPS: microMIPS: Optimise 'memset' core library function.
  MIPS: microMIPS: Add configuration option for microMIPS kernel.
  MIPS: microMIPS: Disable LL/SC and fix linker bug.
  MIPS: microMIPS: Add vdso support.
  MIPS: microMIPS: Add unaligned access support.
  MIPS: microMIPS: Support handling of delay slots.
  MIPS: microMIPS: Add support for exception handling.
  MIPS: microMIPS: Floating point support.
  MIPS: microMIPS: Fix macro naming in micro-assembler.
  ...

11 years ago[SCSI] qla2xxx: Update firmware link in Kconfig file.
Chad Dupuis [Thu, 7 Mar 2013 15:30:22 +0000 (10:30 -0500)]
[SCSI] qla2xxx: Update firmware link in Kconfig file.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: Chad Dupuis <chad.dupuis@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] iscsi class, qla4xxx: fix sess/conn refcounting when find fns are used
Mike Christie [Mon, 6 May 2013 17:06:56 +0000 (12:06 -0500)]
[SCSI] iscsi class, qla4xxx: fix sess/conn refcounting when find fns are used

This fixes a bug where the iscsi class/driver did not do a put_device
when a sess/conn device was found. This also simplifies the interface
by not having to pass in some arguments that were duplicated and did
not need to be exported.

Reported-by: Zhao Hongjiang <zhaohongjiang@huawei.com>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] sas: unify the pointlessly separated enums sas_dev_type and sas_device_type
James Bottomley [Tue, 7 May 2013 21:44:06 +0000 (14:44 -0700)]
[SCSI] sas: unify the pointlessly separated enums sas_dev_type and sas_device_type

These enums have been separate since the dawn of SAS, mainly because the
latter is a procotol only enum and the former includes additional state
for libsas.  The dichotomy causes endless confusion about which one you
should use where and leads to pointless warnings like this:

drivers/scsi/mvsas/mv_sas.c: In function 'mvs_update_phyinfo':
drivers/scsi/mvsas/mv_sas.c:1162:34: warning: comparison between 'enum sas_device_type' and 'enum sas_dev_type' [-Wenum-compare]

Fix by eliminating one of them.  The one kept is effectively the sas.h
one, but call it sas_device_type and make sure the enums are all
properly namespaced with the SAS_ prefix.

Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: thermal, sas controller config and error handling update
Sakthivel K [Tue, 19 Mar 2013 12:38:40 +0000 (18:08 +0530)]
[SCSI] pm80xx: thermal, sas controller config and error handling update

Modified thermal configuration to happen after interrupt registration
Added SAS controller configuration during initialization
Added error handling logic to handle I_T_Nexus errors and variants

[jejb: fix up tabs and spaces issues]
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: NCQ error handling changes
Sakthivel K [Tue, 19 Mar 2013 12:38:08 +0000 (18:08 +0530)]
[SCSI] pm80xx: NCQ error handling changes

Handled NCQ errors in the low level driver as the FW
is not providing the faulty tag for NCQ errors for libsas
to recover.

[jejb: fix checkpatch issues]
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: WWN Modification for PM8081/88/89 controllers
Sakthivel K [Tue, 19 Mar 2013 12:37:35 +0000 (18:07 +0530)]
[SCSI] pm80xx: WWN Modification for PM8081/88/89 controllers

Individual WWN read operations based on controller.
PM8081 - Read WWN from Flash VPD.
PM8088/89 - Read WWN from EEPROM.
PM8001 - Read WWN from NVM.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: Changed module name and debug messages update
Sakthivel K [Tue, 19 Mar 2013 12:37:09 +0000 (18:07 +0530)]
[SCSI] pm80xx: Changed module name and debug messages update

Changed name in driver to pm80xx. Updated debug messages.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: Firmware flash memory free fix, with addition of new memory region...
Sakthivel K [Tue, 19 Mar 2013 12:36:40 +0000 (18:06 +0530)]
[SCSI] pm80xx: Firmware flash memory free fix, with addition of new memory region for it

Performing pci_free_consistent in tasklet had result in a core dump. So
allocated a new memory region for it. Fix for passing proper address
and operation in firmware flash update.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: SPC new firmware changes for device id 0x8081 alone
Sakthivel K [Tue, 19 Mar 2013 12:35:55 +0000 (18:05 +0530)]
[SCSI] pm80xx: SPC new firmware changes for device id 0x8081 alone

Additional bar shift for new SPC firmware, applicable to device
id 0x8081 only.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: Added SPCv/ve specific hardware functionalities and relevant changes...
Sakthivel K [Wed, 17 Apr 2013 11:07:02 +0000 (16:37 +0530)]
[SCSI] pm80xx: Added SPCv/ve specific hardware functionalities and relevant changes in common files

Implementation of SPCv/ve specific hardware functionality and
macros. Changing common functionalities wrt SPCv/ve operations.
Conditional checks for SPC specific operations.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: MSI-X implementation for using 64 interrupts
Sakthivel K [Tue, 19 Mar 2013 12:26:17 +0000 (17:56 +0530)]
[SCSI] pm80xx: MSI-X implementation for using 64 interrupts

Implementation of interrupt handlers and tasklets to support
upto 64 interrupt for the device.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: Updated common functions common for SPC and SPCv/ve
Sakthivel K [Wed, 27 Feb 2013 14:57:43 +0000 (20:27 +0530)]
[SCSI] pm80xx: Updated common functions common for SPC and SPCv/ve

Update of function prototype for common function to SPC and SPCv/ve.
Multiple queues implementation for IO.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: Multiple inbound/outbound queue configuration
Sakthivel K [Wed, 27 Feb 2013 14:55:25 +0000 (20:25 +0530)]
[SCSI] pm80xx: Multiple inbound/outbound queue configuration

Memory allocation and configuration of multiple inbound and
outbound queues.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] pm80xx: Added SPCv/ve specific ids, variables and modify for SPC
Sakthivel K [Wed, 17 Apr 2013 10:56:36 +0000 (16:26 +0530)]
[SCSI] pm80xx: Added SPCv/ve specific ids, variables and modify for SPC

Updated pci id table with device, vendor, subdevice and subvendor ids
for 8081, 8088, 8089 SAS/SATA controllers. Added SPCv/ve related macros.
Updated macros, hba info structure and other structures for SPCv/ve.
Update of structure and variable names for SPC hardware functionalities.

Signed-off-by: Sakthivel K <Sakthivel.SaravananKamalRaju@pmcs.com>
Signed-off-by: Anand Kumar S <AnandKumar.Santhanam@pmcs.com>
Acked-by: Jack Wang <jack_wang@usish.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] lpfc: fix up Kconfig dependencies
James Bottomley [Mon, 6 May 2013 16:49:25 +0000 (09:49 -0700)]
[SCSI] lpfc: fix up Kconfig dependencies

lpfc uses the generic checksum as well as the T10DIF one from the lib/
directory, so make sure they're selected.

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years ago[SCSI] Handle MLQUEUE busy response in scsi_send_eh_cmnd
Hannes Reinecke [Thu, 25 Apr 2013 06:10:00 +0000 (08:10 +0200)]
[SCSI] Handle MLQUEUE busy response in scsi_send_eh_cmnd

scsi_send_eh_cmnd() is calling queuecommand() directly, so
it needs to check the return value here.
The only valid return codes for queuecommand() are 'busy'
states, so we need to wait for a bit to allow the LLDD
to recover.

Based on an earlier patch from Wen Xiong.

[jejb: fix confusion between msec and jiffies values and other issues]
[bvanassche: correct stall_for interval]
Cc: Wen Xiong <wenxiong@linux.vnet.ibm.com>
Cc: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
11 years agoMerge tag 'arc-v3.10-rc1-part2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 10 May 2013 14:24:14 +0000 (07:24 -0700)]
Merge tag 'arc-v3.10-rc1-part2' of git://git./linux/kernel/git/vgupta/arc

Pull second set of arc arch updates from Vineet Gupta:
 "Aliasing VIPT dcache support for ARC

  I'm satisified with testing, specially with fuse which has
  historically given grief to VIPT arches (ARM/PARISC...)"

* tag 'arc-v3.10-rc1-part2' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARC: [TB10x] Remove GENERIC_GPIO
  ARC: [mm] Aliasing VIPT dcache support 4/4
  ARC: [mm] Aliasing VIPT dcache support 3/4
  ARC: [mm] Aliasing VIPT dcache support 2/4
  ARC: [mm] Aliasing VIPT dcache support 1/4
  ARC: [mm] refactor the core (i|d)cache line ops loops
  ARC: [mm] serious bug in vaddr based icache flush

11 years agoMerge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu
Linus Torvalds [Fri, 10 May 2013 14:22:35 +0000 (07:22 -0700)]
Merge branch 'for-next' of git://git./linux/kernel/git/gerg/m68knommu

Pull m68knommu updates from Greg Ungerer:
 "The bulk of the changes are generalizing the ColdFire v3 core support
  and adding in 537x CPU support.  Also a couple of other bug fixes, one
  to fix a reintroduction of a past bug in the romfs filesystem nommu
  support."

* 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu:
  m68knommu: enable Timer on coldfire 532x
  m68knommu: fix ColdFire 5373/5329 QSPI base address
  m68knommu: add support for configuring a Freescale M5373EVB board
  m68knommu: add support for the ColdFire 537x family of CPUs
  m68knommu: make ColdFire M532x platform support more v3 generic
  m68knommu: create and use a common M53xx ColdFire class of CPUs
  m68k: remove unused asm/dbg.h
  m68k: Set ColdFire ACR1 cache mode depending on kernel configuration
  romfs: fix nommu map length to keep inside filesystem
  m68k: clean up unused "config ROMVECSIZE"

11 years agoMerge tag 'for-linus' of git://github.com/realmz/blackfin-linux
Linus Torvalds [Fri, 10 May 2013 14:21:16 +0000 (07:21 -0700)]
Merge tag 'for-linus' of git://github.com/realmz/blackfin-linux

Pull blackfin updates from Steven Miao.

* tag 'for-linus' of git://github.com/realmz/blackfin-linux:
  bfin cache: dcplb map: add 16M dcplb map for BF60x
  blackfin: smp: fix smp build after drop asm/system.h
  blackfin: fix bootup core clock and system clock display
  Platform Nand: Set the GPIO for NAND read as input
  blackfin: rename vmImage to uImage after we move to buildroot
  blackfin: twi: Remove bogus #endif
  bf609: rsi: Add bf609 rsi MMR macro and board platform data.
  blackfin: dmc: Improve DDR2 write through in DMC effict controller.

11 years agoMerge branch 'next' of git://git.monstr.eu/linux-2.6-microblaze
Linus Torvalds [Fri, 10 May 2013 14:19:52 +0000 (07:19 -0700)]
Merge branch 'next' of git://git.monstr.eu/linux-2.6-microblaze

Pull microblaze updates from Michal Simek.

* 'next' of git://git.monstr.eu/linux-2.6-microblaze:
  microblaze: Enable IRQ in arch_cpu_idle
  microblaze: Fix uaccess_ok macro
  microblaze: Add support for new cpu versions and target architecture
  microblaze: Do not select OPT_LIB_ASM by default
  microblaze: Fix initrd support
  microblaze: Do not use r6 in head.S
  microblaze: pci: Remove duplicated header
  microblaze: Set the default irq_domain
  microblaze: pci: Remove duplicated include from pci-common.c

11 years agodm cache: set config value
Joe Thornber [Fri, 10 May 2013 13:37:21 +0000 (14:37 +0100)]
dm cache: set config value

Share configuration option processing code between the dm cache
ctr and message functions.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm cache: move config fns
Alasdair G Kergon [Fri, 10 May 2013 13:37:21 +0000 (14:37 +0100)]
dm cache: move config fns

Move process_config_option() in dm-cache-target.c to make the
next patch more readable.

Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm thin: generate event when metadata threshold passed
Joe Thornber [Fri, 10 May 2013 13:37:21 +0000 (14:37 +0100)]
dm thin: generate event when metadata threshold passed

Generate a dm event when the amount of remaining thin pool metadata
space falls below a certain level.

The threshold is taken to be a quarter of the size of the metadata
device with a minimum threshold of 4MB.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm persistent metadata: add space map threshold callback
Joe Thornber [Fri, 10 May 2013 13:37:20 +0000 (14:37 +0100)]
dm persistent metadata: add space map threshold callback

Add a threshold callback to dm persistent data space maps.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm persistent data: add threshold callback to space map
Joe Thornber [Fri, 10 May 2013 13:37:20 +0000 (14:37 +0100)]
dm persistent data: add threshold callback to space map

Add a threshold callback function to the persistent data space map
interface for a subsequent patch to use.

dm-thin and dm-cache are interested in knowing when they're getting
low on metadata or data blocks.  This patch introduces a new method
for registering a callback against a threshold.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm thin: detect metadata device resizing
Joe Thornber [Fri, 10 May 2013 13:37:19 +0000 (14:37 +0100)]
dm thin: detect metadata device resizing

Allow the dm thin pool metadata device to be extended.

Whenever a pool is resumed, detect whether the size of the metadata
device has increased, and if so, extend the metadata to use the new
space.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm persistent data: support space map resizing
Joe Thornber [Fri, 10 May 2013 13:37:19 +0000 (14:37 +0100)]
dm persistent data: support space map resizing

Support extending a dm persistent data metadata space map.

The extend itself is implemented by switching back to the boostrap
allocator and pointing to the new space.  The extra bitmap indexes are
then allocated from the new space, and finally we switch back to the
proper space map ops and tweak the reference counts.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm thin: open dev read only when possible
Joe Thornber [Fri, 10 May 2013 13:37:19 +0000 (14:37 +0100)]
dm thin: open dev read only when possible

If a thin pool is created in read-only-metadata mode then only open the
metadata device read-only.

Previously it was always opened with FMODE_READ | FMODE_WRITE.

(Note that dm_get_device() still allows read-only dm devices to be used
read-write at the moment: If I create a read-only linear device for the
metadata, via dmsetup load --readonly, then I can still create a rw pool
out of it.)

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm thin: refactor data dev resize
Joe Thornber [Fri, 10 May 2013 13:37:18 +0000 (14:37 +0100)]
dm thin: refactor data dev resize

Refactor device size functions in preparation for similar metadata
device resizing functions.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm cache: replace memcpy with struct assignment
Joe Thornber [Fri, 10 May 2013 13:37:18 +0000 (14:37 +0100)]
dm cache: replace memcpy with struct assignment

Use struct assignment rather than memcpy in dm cache.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm cache: fix typos in comments
Joe Thornber [Fri, 10 May 2013 13:37:18 +0000 (14:37 +0100)]
dm cache: fix typos in comments

Fix up some typos in dm-cache comments.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm cache policy: fix description of lookup fn
Alasdair G Kergon [Fri, 10 May 2013 13:37:17 +0000 (14:37 +0100)]
dm cache policy: fix description of lookup fn

Correct the documented requirement on the return code from dm cache policy
lookup functions stated in the policy module header file.

Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: document iterate_devices
Alasdair G Kergon [Fri, 10 May 2013 13:37:17 +0000 (14:37 +0100)]
dm: document iterate_devices

Document iterate_devices in device-mapper.h.

Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm persistent data: fix error message typos
Joe Thornber [Fri, 10 May 2013 13:37:17 +0000 (14:37 +0100)]
dm persistent data: fix error message typos

Fix some typos in dm-space-map-metadata.c error messages.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm cache: tune migration throttling
Joe Thornber [Fri, 10 May 2013 13:37:16 +0000 (14:37 +0100)]
dm cache: tune migration throttling

Tune the dm cache migration throttling.

i) Issue a tick every second, just in case there's no i/o going through.

ii) Drop the migration threshold right down to something suitable for
background work.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm mpath: enable WRITE SAME support
Mike Snitzer [Fri, 10 May 2013 13:37:16 +0000 (14:37 +0100)]
dm mpath: enable WRITE SAME support

Enable WRITE SAME support in dm multipath.  As far as multipath is
concerned it is just another write request.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Tested-by: Bharata B Rao <bharata.rao@gmail.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm table: fix write same support
Mike Snitzer [Fri, 10 May 2013 13:37:16 +0000 (14:37 +0100)]
dm table: fix write same support

If device_not_write_same_capable() returns true then the iterate_devices
loop in dm_table_supports_write_same() should return false.

Reported-by: Bharata B Rao <bharata.rao@gmail.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org # v3.8+
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm bufio: avoid a possible __vmalloc deadlock
Mikulas Patocka [Fri, 10 May 2013 13:37:15 +0000 (14:37 +0100)]
dm bufio: avoid a possible __vmalloc deadlock

This patch uses memalloc_noio_save to avoid a possible deadlock in
dm-bufio.  (it could happen only with large block size, at most
PAGE_SIZE << MAX_ORDER (typically 8MiB).

__vmalloc doesn't fully respect gfp flags. The specified gfp flags are
used for allocation of requested pages, structures vmap_area, vmap_block
and vm_struct and the radix tree nodes.

However, the kernel pagetables are allocated always with GFP_KERNEL.
Thus the allocation of pagetables can recurse back to the I/O layer and
cause a deadlock.

This patch uses the function memalloc_noio_save to set per-process
PF_MEMALLOC_NOIO flag and the function memalloc_noio_restore to restore
it. When this flag is set, all allocations in the process are done with
implied GFP_NOIO flag, thus the deadlock can't happen.

This should be backported to stable kernels, but they don't have the
PF_MEMALLOC_NOIO flag and memalloc_noio_save/memalloc_noio_restore
functions. So, PF_MEMALLOC should be set and restored instead.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: fix error return code in snapshot_ctr
Wei Yongjun [Fri, 10 May 2013 13:37:15 +0000 (14:37 +0100)]
dm snapshot: fix error return code in snapshot_ctr

Return -ENOMEM instead of success if unable to allocate pending
exception mempool in snapshot_ctr.

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Cc: stable@vger.kernel.org
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm cache: fix error return code in cache_create
Wei Yongjun [Fri, 10 May 2013 13:37:14 +0000 (14:37 +0100)]
dm cache: fix error return code in cache_create

Return -ENOMEM if memory allocation fails in cache_create
instead of 0 (to avoid NULL pointer dereference).

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Cc: stable@vger.kernel.org
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm stripe: fix regression in stripe_width calculation
Mike Snitzer [Fri, 10 May 2013 13:37:14 +0000 (14:37 +0100)]
dm stripe: fix regression in stripe_width calculation

Fix a regression in the calculation of the stripe_width in the
dm stripe target which led to incorrect processing of device limits.

The stripe_width is the stripe device length divided by the number of
stripes.  The group of commits in the range f14fa69 ("dm stripe: fix
size test") to eb850de ("dm stripe: support for non power of 2
chunksize") interfered with each other (a merging error) and led to the
stripe_width being set incorrectly to the stripe device length divided by
chunk_size * stripe_count.

For example, a stripe device's table with: 0 33553920 striped 3 512 ...
should result in a stripe_width of 11184640 (33553920 / 3), but due to
the bug it was getting set to 21845 (33553920 / (512 * 3)).

The impact of this bug is that device topologies that previously worked
fine with the stripe target are no longer considered valid.  In
particular, there is a higher risk of seeing this issue if one of the
stripe devices has a 4K logical block size.  Resulting in an error
message like this:
"device-mapper: table: 253:4: len=21845 not aligned to h/w logical block size 4096 of dm-1"

The fix is to swap the order of the divisions and to use a temporary
variable for the second one, so that width retains the intended
value.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org # 3.6+
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agoxen/pci: Used cached MSI-X capability offset
Bjorn Helgaas [Mon, 22 Apr 2013 23:12:28 +0000 (17:12 -0600)]
xen/pci: Used cached MSI-X capability offset

We now cache the MSI-X capability offset in the struct pci_dev, so no
need to find the capability again.

Acked-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
CC: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
11 years agoxen/pci: Use PCI_MSIX_TABLE_BIR, not PCI_MSIX_FLAGS_BIRMASK
Bjorn Helgaas [Mon, 22 Apr 2013 23:12:21 +0000 (17:12 -0600)]
xen/pci: Use PCI_MSIX_TABLE_BIR, not PCI_MSIX_FLAGS_BIRMASK

PCI_MSIX_FLAGS_BIRMASK is mis-named because the BIR mask is in the
Table Offset register, not the flags ("Message Control" per spec)
register.

Acked-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
CC: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
11 years agoRevert "ALSA: hda - Don't set up active streams twice"
Takashi Iwai [Fri, 10 May 2013 11:39:50 +0000 (13:39 +0200)]
Revert "ALSA: hda - Don't set up active streams twice"

This reverts commit affdb62b815b38261f09f9d4ec210a35c7ffb1f3.

The commit introduced a regression with AD codecs where the stream is
always clean up.  Since the patch is just a minor optimization and
reverting the commit fixes the issue, let's just revert it.

Reported-and-tested-by: Michael Burian <michael.burian@sbg.at>
Cc: <stable@vger.kernel.org> [v3.9+]
Signed-off-by: Takashi Iwai <tiwai@suse.de>
11 years agoARC: [TB10x] Remove GENERIC_GPIO
Vineet Gupta [Wed, 17 Apr 2013 12:07:31 +0000 (17:37 +0530)]
ARC: [TB10x] Remove GENERIC_GPIO

This tracks Alexandre Courbot's mainline GPIO rework

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Acked-by: Alexandre Courbot <acourbot@nvidia.com>
11 years agotracing/kprobes: Support soft-mode disabling
Masami Hiramatsu [Thu, 9 May 2013 05:44:54 +0000 (14:44 +0900)]
tracing/kprobes: Support soft-mode disabling

Support soft-mode disabling on kprobe-based dynamic events.
Soft-disabling is just ignoring recording if the soft disabled
flag is set.

Link: http://lkml.kernel.org/r/20130509054454.30398.7237.stgit@mhiramat-M0-7522
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Tom Zanussi <tom.zanussi@intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
11 years agotracing/kprobes: Support ftrace_event_file base multibuffer
Masami Hiramatsu [Thu, 9 May 2013 05:44:49 +0000 (14:44 +0900)]
tracing/kprobes: Support ftrace_event_file base multibuffer

Support multi-buffer on kprobe-based dynamic events by
using ftrace_event_file.

Link: http://lkml.kernel.org/r/20130509054449.30398.88343.stgit@mhiramat-M0-7522
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Tom Zanussi <tom.zanussi@intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
11 years agotracing/kprobes: Pass trace_probe directly from dispatcher
Masami Hiramatsu [Thu, 9 May 2013 05:44:41 +0000 (14:44 +0900)]
tracing/kprobes: Pass trace_probe directly from dispatcher

Pass the pointer of struct trace_probe directly from probe
dispatcher to handlers. This removes redundant container_of
macro uses. Same thing has already done in trace_uprobe.

Link: http://lkml.kernel.org/r/20130509054441.30398.69112.stgit@mhiramat-M0-7522
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Tom Zanussi <tom.zanussi@intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
11 years agotracing/kprobes: Increment probe hit-count even if it is used by perf
Masami Hiramatsu [Thu, 9 May 2013 05:44:36 +0000 (14:44 +0900)]
tracing/kprobes: Increment probe hit-count even if it is used by perf

Increment probe hit-count for profiling even if it is used
by perf tool. Same thing has already done in trace_uprobe.

Link: http://lkml.kernel.org/r/20130509054436.30398.21133.stgit@mhiramat-M0-7522
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Tom Zanussi <tom.zanussi@intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
11 years agotracing/kprobes: Use bool for retprobe checker
Masami Hiramatsu [Thu, 9 May 2013 05:44:32 +0000 (14:44 +0900)]
tracing/kprobes: Use bool for retprobe checker

Use bool instead of int for kretprobe checker.

Link: http://lkml.kernel.org/r/20130509054431.30398.38561.stgit@mhiramat-M0-7522
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Tom Zanussi <tom.zanussi@intel.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
11 years agoftrace: Fix function probe when more than one probe is added
Steven Rostedt (Red Hat) [Thu, 9 May 2013 23:37:36 +0000 (19:37 -0400)]
ftrace: Fix function probe when more than one probe is added

When the first function probe is added and the function tracer
is updated the functions are modified to call the probe.
But when a second function is added, it updates the function
records to have the second function also update, but it fails
to update the actual function itself.

This prevents the second (or third or forth and so on) probes
from having their functions called.

  # echo vfs_symlink:enable_event:sched:sched_switch > set_ftrace_filter
  # echo vfs_unlink:enable_event:sched:sched_switch > set_ftrace_filter
  # cat trace
 # tracer: nop
 #
 # entries-in-buffer/entries-written: 0/0   #P:4
 #
 #                              _-----=> irqs-off
 #                             / _----=> need-resched
 #                            | / _---=> hardirq/softirq
 #                            || / _--=> preempt-depth
 #                            ||| /     delay
 #           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
 #              | |       |   ||||       |         |
  # touch /tmp/a
  # rm /tmp/a
  # cat trace
 # tracer: nop
 #
 # entries-in-buffer/entries-written: 0/0   #P:4
 #
 #                              _-----=> irqs-off
 #                             / _----=> need-resched
 #                            | / _---=> hardirq/softirq
 #                            || / _--=> preempt-depth
 #                            ||| /     delay
 #           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
 #              | |       |   ||||       |         |
  # ln -s /tmp/a
  # cat trace
 # tracer: nop
 #
 # entries-in-buffer/entries-written: 414/414   #P:4
 #
 #                              _-----=> irqs-off
 #                             / _----=> need-resched
 #                            | / _---=> hardirq/softirq
 #                            || / _--=> preempt-depth
 #                            ||| /     delay
 #           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
 #              | |       |   ||||       |         |
           <idle>-0     [000] d..3  2847.923031: sched_switch: prev_comm=swapper/0 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=bash next_pid=2786 next_prio=120
            <...>-3114  [001] d..4  2847.923035: sched_switch: prev_comm=ln prev_pid=3114 prev_prio=120 prev_state=x ==> next_comm=swapper/1 next_pid=0 next_prio=120
             bash-2786  [000] d..3  2847.923535: sched_switch: prev_comm=bash prev_pid=2786 prev_prio=120 prev_state=S ==> next_comm=kworker/0:1 next_pid=34 next_prio=120
      kworker/0:1-34    [000] d..3  2847.923552: sched_switch: prev_comm=kworker/0:1 prev_pid=34 prev_prio=120 prev_state=S ==> next_comm=swapper/0 next_pid=0 next_prio=120
           <idle>-0     [002] d..3  2847.923554: sched_switch: prev_comm=swapper/2 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=sshd next_pid=2783 next_prio=120
             sshd-2783  [002] d..3  2847.923660: sched_switch: prev_comm=sshd prev_pid=2783 prev_prio=120 prev_state=S ==> next_comm=swapper/2 next_pid=0 next_prio=120

Still need to update the functions even though the probe itself
does not need to be registered again when added a new probe.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
11 years agoftrace: Fix the output of enabled_functions debug file
Steven Rostedt (Red Hat) [Thu, 9 May 2013 23:31:48 +0000 (19:31 -0400)]
ftrace: Fix the output of enabled_functions debug file

The enabled_functions debugfs file was created to be able to see
what functions have been modified from nops to calling a tracer.

The current method uses the counter in the function record.
As when a ftrace_ops is registered to a function, its count
increases. But that doesn't mean that the function is actively
being traced. /proc/sys/kernel/ftrace_enabled can be set to zero
which would disable it, as well as something can go wrong and
we can think its enabled when only the counter is set.

The record's FTRACE_FL_ENABLED flag is set or cleared when its
function is modified. That is a much more accurate way of knowing
what function is enabled or not.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>