]> git.karo-electronics.de Git - karo-tx-linux.git/log
karo-tx-linux.git
11 years agobcache: Debug code improvements
Kent Overstreet [Wed, 11 Sep 2013 02:04:31 +0000 (19:04 -0700)]
bcache: Debug code improvements

Couple changes:
 * Consolidate bch_check_keys() and bch_check_key_order(), and move the
   checks that only check_key_order() could do to bch_btree_iter_next().

 * Get rid of CONFIG_BCACHE_EDEBUG - now, all that code is compiled in
   when CONFIG_BCACHE_DEBUG is enabled, and there's now a sysfs file to
   flip on the EDEBUG checks at runtime.

 * Dropped an old not terribly useful check in rw_unlock(), and
   refactored/improved a some of the other debug code.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Fix bch_ptr_bad()
Kent Overstreet [Thu, 25 Jul 2013 01:14:44 +0000 (18:14 -0700)]
bcache: Fix bch_ptr_bad()

Previously, bch_ptr_bad() could return false when there was a pointer to
a nonexistant device... it only filtered out keys with PTR_CHECK_DEV
pointers.

This behaviour was intended for multiple cache device support; for that,
just because the device for one of the pointers has gone away doesn't
mean we want to filter out the rest of the pointers.

But we don't yet explicitly filter/check individual pointers, so without
that this behaviour was wrong - a corrupt bkey with a bad device pointer
could cause us to deref a bad pointer. Doh.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Pull on disk data structures out into a separate header
Kent Overstreet [Mon, 5 Aug 2013 20:53:41 +0000 (13:53 -0700)]
bcache: Pull on disk data structures out into a separate header

Now, the on disk data structures are in a header that can be exported to
userspace - and having them all centralized is nice too.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Move sector allocator to alloc.c
Kent Overstreet [Thu, 25 Jul 2013 01:11:11 +0000 (18:11 -0700)]
bcache: Move sector allocator to alloc.c

Just reorganizing things a bit.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Break up struct search
Kent Overstreet [Wed, 11 Sep 2013 02:02:45 +0000 (19:02 -0700)]
bcache: Break up struct search

With all the recent refactoring around struct btree op struct search has
gotten rather large.

But we can now easily break it up in a different way - we break out
struct btree_insert_op which is for inserting data into the cache, and
that's now what the copying gc code uses - struct search is now specific
to request.c

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Convert bch_btree_insert() to bch_btree_map_leaf_nodes()
Kent Overstreet [Thu, 25 Jul 2013 01:07:22 +0000 (18:07 -0700)]
bcache: Convert bch_btree_insert() to bch_btree_map_leaf_nodes()

Last of the btree_map() conversions. Main visible effect is
bch_btree_insert() is no longer taking a struct btree_op as an argument
anymore - there's no fancy state machine stuff going on, it's just a
normal function.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Don't use op->insert_collision
Kent Overstreet [Thu, 25 Jul 2013 01:06:22 +0000 (18:06 -0700)]
bcache: Don't use op->insert_collision

When we convert bch_btree_insert() to bch_btree_map_leaf_nodes(), we
won't be passing struct btree_op to bch_btree_insert() anymore - so we
need a different way of returning whether there was a collision (really,
a replace collision).

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Kill op->replace
Kent Overstreet [Wed, 11 Sep 2013 01:52:54 +0000 (18:52 -0700)]
bcache: Kill op->replace

This is prep work for converting bch_btree_insert to
bch_btree_map_leaf_nodes() - we have to convert all its arguments to
actual arguments. Bunch of churn, but should be straightforward.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Drop some closure stuff
Kent Overstreet [Thu, 25 Jul 2013 00:26:22 +0000 (17:26 -0700)]
bcache: Drop some closure stuff

With a the recent bcache refactoring, some of the closure code isn't
needed anymore.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Kill op->cl
Kent Overstreet [Thu, 25 Jul 2013 01:04:18 +0000 (18:04 -0700)]
bcache: Kill op->cl

This isn't used for waiting asynchronously anymore - so this is a fairly
trivial refactoring.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Prune struct btree_op
Kent Overstreet [Thu, 25 Jul 2013 00:44:17 +0000 (17:44 -0700)]
bcache: Prune struct btree_op

Eventual goal is for struct btree_op to contain only what is necessary
for traversing the btree.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Clean up cache_lookup_fn
Kent Overstreet [Thu, 25 Jul 2013 00:41:13 +0000 (17:41 -0700)]
bcache: Clean up cache_lookup_fn

There was some looping in submit_partial_cache_hit() and
submit_partial_cache_hit() that isn't needed anymore - originally, we
wouldn't necessarily process the full hit or miss all at once because
when splitting the bio, we took into account the restrictions of the
device we were sending it to.

But, device bio size restrictions are now handled elsewhere, with a
wrapper around generic_make_request() - so that looping has been
unnecessary for awhile now and we can now do quite a bit of cleanup.

And if we trim the key we're reading from to match the subset we're
actually reading, we don't have to explicitly calculate bi_sector
anymore. Neat.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Convert bch_btree_read_async() to bch_btree_map_keys()
Kent Overstreet [Thu, 25 Jul 2013 00:41:08 +0000 (17:41 -0700)]
bcache: Convert bch_btree_read_async() to bch_btree_map_keys()

This is a fairly straightforward conversion, mostly reshuffling -
op->lookup_done goes away, replaced by MAP_DONE/MAP_CONTINUE. And the
code for handling cache hits and misses wasn't really btree code, so it
gets moved to request.c.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Move some stuff to btree.c
Kent Overstreet [Thu, 25 Jul 2013 00:37:59 +0000 (17:37 -0700)]
bcache: Move some stuff to btree.c

With the new btree_map() functions, we don't need to export the stuff
needed for traversing the btree anymore.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Add btree_map() functions
Kent Overstreet [Wed, 11 Sep 2013 01:48:51 +0000 (18:48 -0700)]
bcache: Add btree_map() functions

Lots of stuff has been open coding its own btree traversal - which is
generally pretty simple code, but there are a few subtleties.

This adds new new functions, bch_btree_map_nodes() and
bch_btree_map_keys(), which do the traversal for you. Everything that's
open coding btree traversal now (with the exception of garbage
collection) is slowly going to be converted to these two functions;
being able to write other code at a higher level of abstraction  is a
big improvement w.r.t. overall code quality.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Convert writeback to a kthread
Kent Overstreet [Thu, 25 Jul 2013 00:50:06 +0000 (17:50 -0700)]
bcache: Convert writeback to a kthread

This simplifies the writeback flow control quite a bit - previously, it
was conceptually two coroutines, refill_dirty() and read_dirty(). This
makes the code quite a bit more straightforward.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Convert gc to a kthread
Kent Overstreet [Thu, 25 Jul 2013 00:50:19 +0000 (17:50 -0700)]
bcache: Convert gc to a kthread

We needed a dedicated rescuer workqueue for gc anyways... and gc was
conceptually a dedicated thread, just one that wasn't running all the
time. Switch it to a dedicated thread to make the code a bit more
straightforward.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Convert bucket_wait to wait_queue_head_t
Kent Overstreet [Thu, 25 Jul 2013 00:29:09 +0000 (17:29 -0700)]
bcache: Convert bucket_wait to wait_queue_head_t

At one point we did do fancy asynchronous waiting stuff with
bucket_wait, but that's all gone (and bucket_wait is used a lot less
than it used to be). So use the standard primitives.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Convert try_wait to wait_queue_head_t
Kent Overstreet [Thu, 25 Jul 2013 00:27:07 +0000 (17:27 -0700)]
bcache: Convert try_wait to wait_queue_head_t

We never waited on c->try_wait asynchronously, so just use the standard
primitives.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Move keylist out of btree_op
Kent Overstreet [Thu, 25 Jul 2013 00:26:51 +0000 (17:26 -0700)]
bcache: Move keylist out of btree_op

Slowly working on pruning struct btree_op - the aim is for it to only
contain things that are actually necessary for traversing the btree.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Refactor journalling flow control
Kent Overstreet [Thu, 25 Jul 2013 00:26:39 +0000 (17:26 -0700)]
bcache: Refactor journalling flow control

Making things less asynchronous that don't need to be - bch_journal()
only has to block when the journal or journal entry is full, which is
emphatically not a fast path. So make it a normal function that just
returns when it finishes, to make the code and control flow easier to
follow.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Refactor read request code a bit
Kent Overstreet [Wed, 11 Sep 2013 00:06:17 +0000 (17:06 -0700)]
bcache: Refactor read request code a bit

More refactoring, and renaming.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Refactor request_write()
Kent Overstreet [Thu, 25 Jul 2013 00:24:52 +0000 (17:24 -0700)]
bcache: Refactor request_write()

Try to improve some of the naming a bit to be more consistent, and also
improve the flow of control in request_write() a bit.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Clean up keylist code
Kent Overstreet [Thu, 25 Jul 2013 00:24:25 +0000 (17:24 -0700)]
bcache: Clean up keylist code

More random refactoring.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Add explicit keylist arg to btree_insert()
Kent Overstreet [Wed, 11 Sep 2013 01:46:36 +0000 (18:46 -0700)]
bcache: Add explicit keylist arg to btree_insert()

Some refactoring - better to explicitly pass stuff around instead of
having it all in the "big bag of state", struct btree_op. Going to prune
struct btree_op quite a bit over time.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Convert btree_insert_check_key() to btree_insert_node()
Kent Overstreet [Wed, 11 Sep 2013 01:39:16 +0000 (18:39 -0700)]
bcache: Convert btree_insert_check_key() to btree_insert_node()

This was the main point of all this refactoring - now,
btree_insert_check_key() won't fail just because the leaf node happened
to be full.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Insert multiple keys at a time
Kent Overstreet [Thu, 25 Jul 2013 00:22:44 +0000 (17:22 -0700)]
bcache: Insert multiple keys at a time

We'll often end up with a list of adjacent keys to insert -
because bch_data_insert() may have to fragment the data it writes.

Originally, to simplify things and avoid having to deal with corner
cases bch_btree_insert() would pass keys from this list one at a time to
btree_insert_recurse() - mainly because the list of keys might span leaf
nodes, so it was easier this way.

With the btree_insert_node() refactoring, it's now a lot easier to just
pass down the whole list and have btree_insert_recurse() iterate over
leaf nodes until it's done.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Add btree_insert_node()
Kent Overstreet [Wed, 11 Sep 2013 01:41:15 +0000 (18:41 -0700)]
bcache: Add btree_insert_node()

The flow of control in the old btree insertion code was rather -
backwards; we'd recurse down the btree (in btree_insert_recurse()), and
then if we needed to split the keys to be inserted into the parent node
would be effectively returned up to btree_insert_recurse(), which would
notice there was more work to do and finish the insertion.

The main problem with this was that the full logic for btree insertion
could only be used by calling btree_insert_recurse; if you'd gotten to a
btree leaf some other way and had a key to insert, if it turned out that
node needed to be split you were SOL.

This inverts the flow of control so btree_insert_node() does _full_
btree insertion, including splitting - and takes a (leaf) btree node to
insert into as a parameter.

This means we can now _correctly_ handle cache misses - for cache
misses, we need to insert a fake "check" key into the btree when we
discover we have a cache miss - while we still have the btree locked.
Previously, if the btree node was full inserting a cache miss would just
fail.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Explicitly track btree node's parent
Kent Overstreet [Thu, 25 Jul 2013 00:20:19 +0000 (17:20 -0700)]
bcache: Explicitly track btree node's parent

This is prep work for the reworked btree insertion code.

The way we set b->parent is ugly and hacky... the problem is, when
btree_split() or garbage collection splits or rewrites a btree node, the
parent changes for all its (potentially already cached) children.

I may change this later and add some code to look through the btree node
cache and find all our cached child nodes and change the parent pointer
then...

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Remove unnecessary check in should_split()
Kent Overstreet [Thu, 25 Jul 2013 00:20:00 +0000 (17:20 -0700)]
bcache: Remove unnecessary check in should_split()

Checking i->seq was redundant, because since ages ago we always
initialize the new bset when advancing b->written

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Stripe size isn't necessarily a power of two
Kent Overstreet [Sat, 17 Aug 2013 09:13:15 +0000 (02:13 -0700)]
bcache: Stripe size isn't necessarily a power of two

Originally I got this right... except that the divides didn't use
do_div(), which broke 32 bit kernels. When I went to fix that, I forgot
that the raid stripe size usually isn't a power of two... doh

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Add on error panic/unregister setting
Kent Overstreet [Fri, 12 Jul 2013 02:42:51 +0000 (19:42 -0700)]
bcache: Add on error panic/unregister setting

Works kind of like the ext4 setting, to panic or remount read only on
errors.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Use blkdev_issue_discard()
Kent Overstreet [Thu, 25 Jul 2013 00:16:09 +0000 (17:16 -0700)]
bcache: Use blkdev_issue_discard()

The old asynchronous discard code was really a relic from when all the
allocation code was asynchronous - now that allocation runs out of a
dedicated thread there's no point in keeping around all that complicated
machinery.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Fix for handling overlapping extents when reading in a btree node
Kent Overstreet [Thu, 29 Aug 2013 01:01:24 +0000 (18:01 -0700)]
bcache: Fix for handling overlapping extents when reading in a btree node

btree_sort_fixup() was overly clever, because it was trying to avoid
pulling a key off the btree iterator in more than one place.

This led to a really obscure bug where we'd break early from the loop in
btree_sort_fixup() if the current key overlapped with keys in more than
one older set, and the next key it overlapped with was zero size.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Fix a shrinker deadlock
Kent Overstreet [Fri, 30 Aug 2013 21:12:24 +0000 (14:12 -0700)]
bcache: Fix a shrinker deadlock

GFP_NOIO means we could be getting called recursively - mca_alloc() ->
mca_data_alloc() - definitely can't use mutex_lock(bucket_lock) then.
Whoops.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Fix a dumb CPU spinning bug in writeback
Kent Overstreet [Fri, 23 Aug 2013 22:12:21 +0000 (15:12 -0700)]
bcache: Fix a dumb CPU spinning bug in writeback

schedule_timeout() != schedule_timeout_uninterruptible()

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Fix a flush/fua performance bug
Kent Overstreet [Wed, 14 Aug 2013 10:57:41 +0000 (03:57 -0700)]
bcache: Fix a flush/fua performance bug

bch_journal_meta() was missing the flush to make the journal write
actually go down (instead of waiting up to journal_delay_ms)...

Whoops

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Fix a writeback performance regression
Kent Overstreet [Wed, 31 Jul 2013 23:29:54 +0000 (16:29 -0700)]
bcache: Fix a writeback performance regression

Background writeback works by scanning the btree for dirty data and
adding those keys into a fixed size buffer, then for each dirty key in
the keybuf writing it to the backing device.

When read_dirty() finishes and it's time to scan for more dirty data, we
need to wait for the outstanding writeback IO to finish - they still
take up slots in the keybuf (so that foreground writes can check for
them to avoid races) - without that wait, we'll continually rescan when
we'll be able to add at most a key or two to the keybuf, and that takes
locks that starves foreground IO. Doh.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Correct printf()-style format length modifier
Geert Uytterhoeven [Mon, 29 Jul 2013 10:45:05 +0000 (12:45 +0200)]
bcache: Correct printf()-style format length modifier

drivers/md/bcache/btree.c: In function ‘bch_btree_node_read’:
drivers/md/bcache/btree.c:259: warning: format ‘%lu’ expects type ‘long unsigned int’, but argument 3 has type ‘size_t’

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Fix for when no journal entries are found
Kent Overstreet [Tue, 23 Jul 2013 01:13:44 +0000 (18:13 -0700)]
bcache: Fix for when no journal entries are found

The journal replay code didn't handle this case, causing it to go into
an infinite loop...

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Strip endline when writing the label through sysfs
Gabriel de Perthuis [Fri, 12 Jul 2013 11:35:44 +0000 (13:35 +0200)]
bcache: Strip endline when writing the label through sysfs

sysfs attributes with unusual characters have crappy failure modes
in Squeeze (udev 164); later versions of udev are unaffected.

This should make these characters more unusual.

Signed-off-by: Gabriel de Perthuis <g2p.code@gmail.com>
Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Fix a dumb journal discard bug
Kent Overstreet [Thu, 18 Jul 2013 17:50:55 +0000 (10:50 -0700)]
bcache: Fix a dumb journal discard bug

That switch statement was obviously wrong, leading to some sort of weird
spinning on rare occasion with discards enabled...

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Allocation kthread fixes
Kent Overstreet [Thu, 11 Jul 2013 01:31:58 +0000 (18:31 -0700)]
bcache: Allocation kthread fixes

The alloc kthread should've been using try_to_freeze() - and also there
was the potential for the alloc kthread to get woken up after it had
shut down, which would have been bad.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Fix GC_SECTORS_USED() calculation
Kent Overstreet [Fri, 12 Jul 2013 02:43:21 +0000 (19:43 -0700)]
bcache: Fix GC_SECTORS_USED() calculation

Part of the job of garbage collection is to add up however many sectors
of live data it finds in each bucket, but that doesn't work very well if
it doesn't reset GC_SECTORS_USED() when it starts. Whoops.

This wouldn't have broken anything horribly, but allocation tries to
preferentially reclaim buckets that are mostly empty and that's not
gonna work with an incorrect GC_SECTORS_USED() value.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Journal replay fix
Kent Overstreet [Fri, 12 Jul 2013 05:42:14 +0000 (22:42 -0700)]
bcache: Journal replay fix

The journal replay code starts by finding something that looks like a
valid journal entry, then it does a binary search over the unchecked
region of the journal for the journal entries with the highest sequence
numbers.

Trouble is, the logic was wrong - journal_read_bucket() returns true if
it found journal entries we need, but if the range of journal entries
we're looking for loops around the end of the journal - in that case
journal_read_bucket() could return true when it hadn't found the highest
sequence number we'd seen yet, and in that case the binary search did
the wrong thing. Whoops.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Shutdown fix
Kent Overstreet [Thu, 11 Jul 2013 04:03:25 +0000 (21:03 -0700)]
bcache: Shutdown fix

Stopping a cache set is supposed to make it stop attached backing
devices, but somewhere along the way that code got lost. Fixing this
mainly has the effect of fixing our reboot notifier.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Fix a sysfs splat on shutdown
Kent Overstreet [Thu, 11 Jul 2013 04:25:02 +0000 (21:25 -0700)]
bcache: Fix a sysfs splat on shutdown

If we stopped a bcache device when we were already detaching (or
something like that), bcache_device_unlink() would try to remove a
symlink from sysfs that was already gone because the bcache dev kobject
had already been removed from sysfs.

So keep track of whether we've removed stuff from sysfs.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Advertise that flushes are supported
Kent Overstreet [Thu, 11 Jul 2013 01:44:40 +0000 (18:44 -0700)]
bcache: Advertise that flushes are supported

Whoops - bcache's flush/FUA was mostly correct, but flushes get filtered
out unless we say we support them...

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: check for allocation failures
Dan Carpenter [Fri, 5 Jul 2013 06:05:46 +0000 (09:05 +0300)]
bcache: check for allocation failures

There is a missing NULL check after the kzalloc().

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
11 years agobcache: Fix a dumb race
Kent Overstreet [Thu, 11 Jul 2013 01:04:21 +0000 (18:04 -0700)]
bcache: Fix a dumb race

In the far-too-complicated closure code - closures can have destructors,
for probably dubious reasons; they get run after the closure is no
longer waiting on anything but before dropping the parent ref, intended
just for freeing whatever memory the closure is embedded in.

Trouble is, when remaining goes to 0 and we've got nothing more to run -
we also have to unlock the closure, setting remaining to -1. If there's
a destructor, that unlock isn't doing anything - nobody could be trying
to lock it if we're about to free it - but if the unlock _is needed...
that check for a destructor was racy. Argh.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.10
11 years agobcache: Use standard utility code
Kent Overstreet [Fri, 7 Jun 2013 01:15:57 +0000 (18:15 -0700)]
bcache: Use standard utility code

Some of bcache's utility code has made it into the rest of the kernel,
so drop the bcache versions.

Bcache used to have a workaround for allocating from a bio set under
generic_make_request() (if you allocated more than once, the bios you
already allocated would get stuck on current->bio_list when you
submitted, and you'd risk deadlock) - bcache would mask out __GFP_WAIT
when allocating bios under generic_make_request() so that allocation
could fail and it could retry from workqueue. But bio_alloc_bioset() has
a workaround now, so we can drop this hack and the associated error
handling.

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Update email address
Kent Overstreet [Mon, 1 Jul 2013 19:19:55 +0000 (12:19 -0700)]
bcache: Update email address

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Delete fuzz tester
Kent Overstreet [Thu, 16 May 2013 00:13:45 +0000 (17:13 -0700)]
bcache: Delete fuzz tester

This code has rotted and it hasn't been used in ages anyways.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: Document shrinker reserve better
Kent Overstreet [Mon, 3 Jun 2013 20:04:56 +0000 (13:04 -0700)]
bcache: Document shrinker reserve better

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
11 years agobcache: FUA fixes
Kent Overstreet [Thu, 27 Jun 2013 00:25:38 +0000 (17:25 -0700)]
bcache: FUA fixes

Journal writes need to be marked FUA, not just REQ_FLUSH. And btree node
writes have... weird ordering requirements.

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Refresh usage docs
Gabriel de Perthuis [Thu, 27 Jun 2013 00:12:07 +0000 (02:12 +0200)]
bcache: Refresh usage docs

Mention udev autoregistration, symlinks.  Write down some sysfs paths.

Signed-off-by: Gabriel de Perthuis <g2p.code@gmail.com>
Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Send label uevents
Gabriel de Perthuis [Sat, 8 Jun 2013 22:54:48 +0000 (00:54 +0200)]
bcache: Send label uevents

Signed-off-by: Gabriel de Perthuis <g2p.code@gmail.com>
Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Send a uevent with a cached device's UUID
Gabriel de Perthuis [Fri, 7 Jun 2013 21:27:01 +0000 (23:27 +0200)]
bcache: Send a uevent with a cached device's UUID

Signed-off-by: Gabriel de Perthuis <g2p.code@gmail.com>
11 years agodoc: Fix typo in documentation/bcache.txt
Masanari Iida [Sun, 19 May 2013 15:04:35 +0000 (00:04 +0900)]
doc: Fix typo in documentation/bcache.txt

Correct spelling typo in documentation/bcache.txt

Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Write out full stripes
Kent Overstreet [Wed, 5 Jun 2013 13:24:39 +0000 (06:24 -0700)]
bcache: Write out full stripes

Now that we're tracking dirty data per stripe, we can add two
optimizations for raid5/6:

 * If a stripe is already dirty, force writes to that stripe to
   writeback mode - to help build up full stripes of dirty data

 * When flushing dirty data, preferentially write out full stripes first
   if there are any.

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Track dirty data by stripe
Kent Overstreet [Wed, 5 Jun 2013 13:21:07 +0000 (06:21 -0700)]
bcache: Track dirty data by stripe

To make background writeback aware of raid5/6 stripes, we first need to
track the amount of dirty data within each stripe - we do this by
breaking up the existing sectors_dirty into per stripe atomic_ts

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Initialize sectors_dirty when attaching
Kent Overstreet [Sun, 12 May 2013 00:07:26 +0000 (17:07 -0700)]
bcache: Initialize sectors_dirty when attaching

Previously, dirty_data wouldn't get initialized until the first garbage
collection... which was a bit of a problem for background writeback (as
the PD controller keys off of it) and also confusing for users.

This is also prep work for making background writeback aware of raid5/6
stripes.

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Improve lazy sorting
Kent Overstreet [Sat, 11 May 2013 22:59:37 +0000 (15:59 -0700)]
bcache: Improve lazy sorting

The old lazy sorting code was kind of hacky - rewrite in a way that
mathematically makes more sense; the idea is that the size of the sets
of keys in a btree node should increase by a more or less fixed ratio
from smallest to biggest.

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Rip out pkey()/pbtree()
Kent Overstreet [Wed, 15 May 2013 03:33:16 +0000 (20:33 -0700)]
bcache: Rip out pkey()/pbtree()

Old gcc doesnt like the struct hack, and it is kind of ugly. So finish
off the work to convert pr_debug() statements to tracepoints, and delete
pkey()/pbtree().

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Fix/revamp tracepoints
Kent Overstreet [Fri, 26 Apr 2013 22:39:55 +0000 (15:39 -0700)]
bcache: Fix/revamp tracepoints

The tracepoints were reworked to be more sensible, and fixed a null
pointer deref in one of the tracepoints.

Converted some of the pr_debug()s to tracepoints - this is partly a
performance optimization; it used to be that with DEBUG or
CONFIG_DYNAMIC_DEBUG pr_debug() was an empty macro; but at some point it
was changed to an empty inline function.

Some of the pr_debug() statements had rather expensive function calls as
part of the arguments, so this code was getting run unnecessarily even
on non debug kernels - in some fast paths, too.

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Refactor btree io
Kent Overstreet [Thu, 25 Apr 2013 20:58:35 +0000 (13:58 -0700)]
bcache: Refactor btree io

The most significant change is that btree reads are now done
synchronously, instead of asynchronously and doing the post read stuff
from a workqueue.

This was originally done because we can't block on IO under
generic_make_request(). But - we already have a mechanism to punt cache
lookups to workqueue if needed, so if we just use that we don't have to
deal with the complexity of doing things asynchronously.

The main benefit is this makes the locking situation saner; we can hold
our write lock on the btree node until we're finished reading it, and we
don't need that btree_node_read_done() flag anymore.

Also, for writes, btree_write() was broken out into btree_node_write()
and btree_leaf_dirty() - the old code with the boolean argument was dumb
and confusing.

The prio_blocked mechanism was improved a bit too, now the only counter
is in struct btree_write, we don't mess with transfering a count from
struct btree anymore.

This required changing garbage collection to block prios at the start
and unblock when it finishes, which is cleaner than what it was doing
anyways (the old code had mostly the same effect, but was doing it in a
convoluted way)

And the btree iter btree_node_read_done() uses was converted to a real
mempool.

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Convert allocator thread to kthread
Kent Overstreet [Thu, 25 Apr 2013 02:01:12 +0000 (19:01 -0700)]
bcache: Convert allocator thread to kthread

Using a workqueue when we just want a single thread is a bit silly.

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: Warn when a device is already registered.
Gabriel de Perthuis [Sat, 4 May 2013 10:19:41 +0000 (12:19 +0200)]
bcache: Warn when a device is already registered.

Signed-off-by: Gabriel de Perthuis <g2p.code+bcache@gmail.com>
Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agobcache: fix a spurious gcc complaint, use scnprintf
Kent Overstreet [Wed, 29 May 2013 04:53:19 +0000 (21:53 -0700)]
bcache: fix a spurious gcc complaint, use scnprintf

An old version of gcc was complaining about using a const int as the
size of a stack allocated array. Which should be fine - but using
ARRAY_SIZE() is better, anyways.

Also, refactor the code to use scnprintf().

Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agomd: bcache: io.c: fix a potential NULL pointer dereference
Kumar Amit Mehta [Tue, 28 May 2013 07:31:15 +0000 (00:31 -0700)]
md: bcache: io.c: fix a potential NULL pointer dereference

bio_alloc_bioset returns NULL on failure. This fix adds a missing check
for potential NULL pointer dereferencing.

Signed-off-by: Kumar Amit Mehta <gmate.amit@gmail.com>
Signed-off-by: Kent Overstreet <koverstreet@google.com>
11 years agoLinux 3.10-rc7 v3.10-rc7
Linus Torvalds [Sat, 22 Jun 2013 19:47:31 +0000 (09:47 -1000)]
Linux 3.10-rc7

11 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Sat, 22 Jun 2013 19:44:45 +0000 (09:44 -1000)]
Merge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Arnd Bergmann:
 "These are two fixes that came in this week, one for a regression we
  introduced in 3.10 in the GIC interrupt code, and the other one fixes
  a typo in newly introduced code"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  irqchip: gic: call gic_cpu_init() as well in CPU_STARTING_FROZEN case
  ARM: dts: Correct the base address of pinctrl_3 on Exynos5250

11 years agoMerge tag 'driver-core-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 22 Jun 2013 19:02:44 +0000 (09:02 -1000)]
Merge tag 'driver-core-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core

Pull driver core fix from Greg Kroah-Hartman:
 "Here's a single patch for the firmware core that resolves a reported
  oops in the firmware core that people have been hitting."

* tag 'driver-core-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  firmware loader: fix use-after-free by double abort

11 years agoMerge tag 'usb-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sat, 22 Jun 2013 19:01:47 +0000 (09:01 -1000)]
Merge tag 'usb-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb

Pull USB fixes from Greg Kroah-Hartman:
 "Here are two USB patches for 3.10.

  One updates the Kconfig wording for CONFIG_USB_PHY to make it,
  hopefully, more obvious what this option is (I know you complained
  about this when it hit the tree.) The other is a new device id for a
  driver"

* tag 'usb-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  USB: serial: ti_usb_3410_5052: new device id for Abbot strip port cable
  usb: phy: Improve Kconfig help for CONFIG_USB_PHY

11 years agoMerge tag 'tty-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sat, 22 Jun 2013 19:00:28 +0000 (09:00 -1000)]
Merge tag 'tty-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty

Pul tty fixes from Greg Kroah-Hartman:
 "Here are two tty core fixes that resolve some regressions that have
  been reported recently.  Both tiny fixes, but needed"

* tag 'tty-3.10-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty: Fix transient pty write() EIO
  tty/vt: Return EBUSY if deallocating VT1 and it is busy

11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending
Linus Torvalds [Sat, 22 Jun 2013 18:54:06 +0000 (08:54 -1000)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending

Pull SCSI target fixes from Nicholas Bellinger:
 "Included is the recent tcm_qla2xxx residual underrun length fix from
  Roland, along with Joern's iscsi-target patch for session_lock
  breakage within iscsit_stop_time2retain_timer() code.  Both are CC'ed
  to stable.

  The remaining two are specific to recent iscsi-target + iser
  conversion changes.  One drops some left-over debug noise, and Andy's
  patch fixes configfs attribute handling during an explicit network
  portal feature bit disable when iser-target is unsupported."

* git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  iscsi-target: Remove left over v3.10-rc debug printks
  target/iscsi: Fix op=disable + error handling cases in np_store_iser
  tcm_qla2xxx: Fix residual for underrun commands that fail
  target/iscsi: don't corrupt bh_count in iscsit_stop_time2retain_timer()

11 years agoMerge branch 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Sat, 22 Jun 2013 18:43:17 +0000 (08:43 -1000)]
Merge branch 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media

Pull media fixes from Mauro Carvalho Chehab:
 "Another set of fixes for Kernel 3.10.

  This series contain:
   - two Kbuild fixes for randconfig
   - a buffer overflow when using rtl28xuu with r820t tuner
   - one clk fixup on exynos4-is driver"

* 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  [media] Fix build when drivers are builtin and frontend modules
  [media] s5p makefiles: don't override other selections on obj-[ym]
  [media] exynos4-is: Fix FIMC-IS clocks initialization
  [media] rtl28xxu: fix buffer overflow when probing Rafael Micro r820t tuner

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Sat, 22 Jun 2013 18:42:20 +0000 (08:42 -1000)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull vfs fixes from Al Viro:
 "Several fixes for bugs caught while looking through f_pos (ab)users"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  aout32 coredump compat fix
  splice: don't pass the address of ->f_pos to methods
  mconsole: we'd better initialize pos before passing it to vfs_read()...

11 years agoaout32 coredump compat fix
Al Viro [Sat, 22 Jun 2013 07:01:38 +0000 (11:01 +0400)]
aout32 coredump compat fix

dump_seek() does SEEK_CUR, not SEEK_SET; native binfmt_aout
handles it correctly (seeks by PAGE_SIZE - sizeof(struct user),
getting the current position to PAGE_SIZE), compat one seeks
by PAGE_SIZE and ends up at PAGE_SIZE + already written...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years agoMerge branch 'x86/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Linus Torvalds [Fri, 21 Jun 2013 16:33:48 +0000 (06:33 -1000)]
Merge branch 'x86/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 fixes from Peter Anvin:
 "This series fixes a couple of build failures, and fixes MTRR cleanup
  and memory setup on very specific memory maps.

  Finally, it fixes triggering backtraces on all CPUs, which was
  inadvertently disabled on x86."

* 'x86/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/efi: Fix dummy variable buffer allocation
  x86: Fix trigger_all_cpu_backtrace() implementation
  x86: Fix section mismatch on load_ucode_ap
  x86: fix build error and kconfig for ia32_emulation and binfmt
  range: Do not add new blank slot with add_range_with_merge
  x86, mtrr: Fix original mtrr range get for mtrr_cleanup

11 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 21 Jun 2013 16:33:06 +0000 (06:33 -1000)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm radeon fixes from Dave Airlie:
 "One core fix, but mostly radeon fixes for s/r and big endian UVD
  support, and a fix to stop the GPU being reset for no good reason, and
  crashing people's machines."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/radeon: update lockup tracking when scheduling in empty ring
  drm/prime: Honor requested file flags when exporting a buffer
  drm/radeon: fix UVD on big endian
  drm/radeon: fix write back suspend regression with uvd v2
  drm/radeon: do not try to uselessly update virtual memory pagetable

11 years agoMerge tag 'acpi-3.10-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 21 Jun 2013 16:31:10 +0000 (06:31 -1000)]
Merge tag 'acpi-3.10-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm

Pull ACPI fixes from Rafael Wysocki:

 - Fix for a regression causing a failure to turn on some devices on
   some systems during initialization introduced by a recent revert of
   an ACPI PM change that broke something else.  Fortunately, we know
   exactly what devices are affected, so we can add a fix just for them
   leaving everyone else alone.

 - ACPI power resources initialization fix preventing a NULL pointer
   from being dereferenced in the acpi_add_power_resource() error code
   path.

 - ACPI dock station driver fix that adds missing locking to
   write_undock().

 - ACPI resources allocation fix changing the scope of an old workaround
   so that it doesn't affect systems that aren't actually buggy.  This
   was reported a couple of days ago to fix DMA problems on some new
   platforms so we need it in -stable.  From Mika Westerberg.

* tag 'acpi-3.10-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI / LPSS: Power up LPSS devices during enumeration
  ACPI / PM: Fix error code path for power resources initialization
  ACPI / dock: Take ACPI scan lock in write_undock()
  ACPI / resources: call acpi_get_override_irq() only for legacy IRQ resources

11 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 21 Jun 2013 16:29:22 +0000 (06:29 -1000)]
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm

Pull KVM fixes from Paolo Bonzini:
 "Three one-line fixes for my first pull request; one for x86 host, one
  for x86 guest, one for PPC"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  x86: kvmclock: zero initialize pvclock shared memory area
  kvm/ppc/booke: Delay kvmppc_lazy_ee_enable
  KVM: x86: remove vcpu's CPL check in host-invoked XCR set

11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Fri, 21 Jun 2013 16:28:39 +0000 (06:28 -1000)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6

Pull crypto fix from Herbert Xu:
 "This fixes an unaligned crash in XTS mode when using aseni_intel"

* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: aesni_intel - fix accessing of unaligned memory

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Fri, 21 Jun 2013 16:27:40 +0000 (06:27 -1000)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client

Pull Ceph fix from Sage Weil:
 "This fixes a problem preventing the kernel and userland librbd
  libraries from sharing data with the new format 2 images"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: use the correct length for format 2 object names

11 years agoMerge tag 'efi-urgent' into x86/urgent
H. Peter Anvin [Fri, 21 Jun 2013 10:01:21 +0000 (03:01 -0700)]
Merge tag 'efi-urgent' into x86/urgent

 * Don't leak random kernel memory to EFI variable NVRAM when attempting
   to initiate garbage collection. Also, free the kernel memory when
   we're done with it instead of leaking - Ben Hutchings

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
11 years agox86/efi: Fix dummy variable buffer allocation
Ben Hutchings [Sun, 16 Jun 2013 20:27:12 +0000 (21:27 +0100)]
x86/efi: Fix dummy variable buffer allocation

1. Check for allocation failure
2. Clear the buffer contents, as they may actually be written to flash
3. Don't leak the buffer

Compile-tested only.

[ Tested successfully on my buggy ASUS machine - Matt ]

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Cc: stable@vger.kernel.org
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoiscsi-target: Remove left over v3.10-rc debug printks
Nicholas Bellinger [Thu, 20 Jun 2013 23:36:17 +0000 (16:36 -0700)]
iscsi-target: Remove left over v3.10-rc debug printks

Reported-by: Andy Grover <agrover@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agotarget/iscsi: Fix op=disable + error handling cases in np_store_iser
Andy Grover [Wed, 29 May 2013 19:05:59 +0000 (12:05 -0700)]
target/iscsi: Fix op=disable + error handling cases in np_store_iser

Writing 0 when iser was not previously enabled, so succeed but do
nothing so that user-space code doesn't need a try: catch block
when ib_isert logic is not available.

Also, return actual error from add_network_portal using PTR_ERR
during op=enable failure.

Signed-off-by: Andy Grover <agrover@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agoMerge branch 'drm-fixes-3.10' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Thu, 20 Jun 2013 22:52:19 +0000 (08:52 +1000)]
Merge branch 'drm-fixes-3.10' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

One user visible fix to stop misreport GPU hangs and subsequent resets.
* 'drm-fixes-3.10' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon: update lockup tracking when scheduling in empty ring

11 years agodrm/radeon: update lockup tracking when scheduling in empty ring
Jerome Glisse [Wed, 19 Jun 2013 14:02:28 +0000 (10:02 -0400)]
drm/radeon: update lockup tracking when scheduling in empty ring

There might be issue with lockup detection when scheduling on an
empty ring that have been sitting idle for a while. Thus update
the lockup tracking data when scheduling new work in an empty ring.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Tested-by: Andy Lutomirski <luto@amacapital.net>
Cc: stable@vger.kernel.org
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 20 Jun 2013 18:18:35 +0000 (08:18 -1000)]
Merge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "Two smaller fixes - plus a context tracking tracing fix that is a bit
  bigger"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tracing/context-tracking: Add preempt_schedule_context() for tracing
  sched: Fix clear NOHZ_BALANCE_KICK
  sched/x86: Construct all sibling maps if smt

11 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 20 Jun 2013 18:17:36 +0000 (08:17 -1000)]
Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "Four fixes.  The mmap ones are unfortunately larger than desired -
  fuzzing uncovered bugs that needed perf context life time management
  changes to fix properly"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86: Fix broken PEBS-LL support on SNB-EP/IVB-EP
  perf: Fix mmap() accounting hole
  perf: Fix perf mmap bugs
  kprobes: Fix to free gone and unused optprobes

11 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 20 Jun 2013 18:16:07 +0000 (08:16 -1000)]
Merge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull cpu idle fixes from Thomas Gleixner:
 - Add a missing irq enable. Fallout of the idle conversion
 - Fix stackprotector wreckage caused by the idle conversion

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  idle: Enable interrupts in the weak arch_cpu_idle() implementation
  idle: Add the stack canary init to cpu_startup_entry()

11 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 20 Jun 2013 18:15:13 +0000 (08:15 -1000)]
Merge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull timer fixes from Thomas Gleixner:
 - Fix inconstinant clock usage in virtual time accounting
 - Fix a build error in KVM caused by the NOHZ work
 - Remove a pointless timekeeping duty assignment which breaks NOHZ
 - Use a proper notifier return value to avoid random behaviour

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tick: Remove useless timekeeping duty attribution to broadcast source
  nohz: Fix notifier return val that enforce timekeeping
  kvm: Move guest entry/exit APIs to context_tracking
  vtime: Use consistent clocks among nohz accounting

11 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Thu, 20 Jun 2013 18:08:46 +0000 (08:08 -1000)]
Merge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc

Pull powerpc fix fro, Benjamin Herrenschmidt:
 "We accidentally broke hugetlbfs on Freescale embedded processors which
  use a slightly different page table layout than our server processors"

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc: Fix bad pmd error with book3E config

11 years agoMerge branch 'stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux...
Linus Torvalds [Thu, 20 Jun 2013 18:07:42 +0000 (08:07 -1000)]
Merge branch 'stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile

Pull tilepro fix from Chris Metcalf:
 "This change allows the older tilepro architecture to be correctly
  built by newer gccs, despite a change that caused gcc to start trying
  to use an out-of-line implementation for __builtin_ffsll().

  This should be inline again starting with gcc 4.7.4 and 4.8.2 or so,
  but meanwhile this change keeps things from breaking, with the only
  cost being a few bytes of code in the kernel to provide __ffsdi2 even
  for compilers that do inline it"

* 'stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  tilepro: work around module link error with gcc 4.7

11 years agoMerge tag 'arm64-stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmarinas...
Linus Torvalds [Thu, 20 Jun 2013 18:06:48 +0000 (08:06 -1000)]
Merge tag 'arm64-stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmarinas/linux-aarch64

Pull arm64 perf fix from Catalin Marinas:
 "Perf fix (user-mode PC recording)"

* tag 'arm64-stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmarinas/linux-aarch64:
  perf: arm64: Record the user-mode PC in the call chain.

11 years agosplice: don't pass the address of ->f_pos to methods
Al Viro [Thu, 20 Jun 2013 14:58:36 +0000 (18:58 +0400)]
splice: don't pass the address of ->f_pos to methods

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
11 years ago[media] Fix build when drivers are builtin and frontend modules
Mauro Carvalho Chehab [Thu, 20 Jun 2013 13:35:53 +0000 (10:35 -0300)]
[media] Fix build when drivers are builtin and frontend modules

There are a large number of reports that the media build is
not compiling when some drivers are compiled as builtin, while
the needed frontends are compiled as module.

On the last one of such reports:
From: kbuild test robot <fengguang.wu@intel.com>
Subject: saa7134-dvb.c:undefined reference to `zl10039_attach'

The .config file has:

CONFIG_VIDEO_SAA7134=y
CONFIG_VIDEO_SAA7134_DVB=y
# CONFIG_MEDIA_ATTACH is not set
CONFIG_DVB_ZL10039=m

And it produces all those errors:

   drivers/built-in.o: In function `set_type':
   tuner-core.c:(.text+0x2f263e): undefined reference to `tea5767_attach'
   tuner-core.c:(.text+0x2f273e): undefined reference to `tda9887_attach'
   drivers/built-in.o: In function `tuner_probe':
   tuner-core.c:(.text+0x2f2d20): undefined reference to `tea5767_autodetection'
   drivers/built-in.o: In function `av7110_attach':
   av7110.c:(.text+0x330bda): undefined reference to `ves1x93_attach'
   av7110.c:(.text+0x330bf7): undefined reference to `stv0299_attach'
   av7110.c:(.text+0x330c63): undefined reference to `tda8083_attach'
   av7110.c:(.text+0x330d09): undefined reference to `ves1x93_attach'
   av7110.c:(.text+0x330d33): undefined reference to `tda8083_attach'
   av7110.c:(.text+0x330d5d): undefined reference to `stv0297_attach'
   av7110.c:(.text+0x330dbe): undefined reference to `stv0299_attach'
   drivers/built-in.o: In function `tuner_attach_dtt7520x':
   ngene-cards.c:(.text+0x3381cb): undefined reference to `dvb_pll_attach'
   drivers/built-in.o: In function `demod_attach_lg330x':
   ngene-cards.c:(.text+0x33828a): undefined reference to `lgdt330x_attach'
   drivers/built-in.o: In function `demod_attach_stv0900':
   ngene-cards.c:(.text+0x3383d5): undefined reference to `stv090x_attach'
   drivers/built-in.o: In function `cineS2_probe':
   ngene-cards.c:(.text+0x338b7f): undefined reference to `drxk_attach'
   drivers/built-in.o: In function `configure_tda827x_fe':
   saa7134-dvb.c:(.text+0x346ae7): undefined reference to `tda10046_attach'
   drivers/built-in.o: In function `dvb_init':
   saa7134-dvb.c:(.text+0x347283): undefined reference to `mt352_attach'
   saa7134-dvb.c:(.text+0x3472cd): undefined reference to `mt352_attach'
   saa7134-dvb.c:(.text+0x34731c): undefined reference to `tda10046_attach'
   saa7134-dvb.c:(.text+0x34733c): undefined reference to `tda10046_attach'
   saa7134-dvb.c:(.text+0x34735c): undefined reference to `tda10046_attach'
   saa7134-dvb.c:(.text+0x347378): undefined reference to `tda10046_attach'
   saa7134-dvb.c:(.text+0x3473db): undefined reference to `tda10046_attach'
   drivers/built-in.o:saa7134-dvb.c:(.text+0x347502): more undefined references to `tda10046_attach' follow
   drivers/built-in.o: In function `dvb_init':
   saa7134-dvb.c:(.text+0x347812): undefined reference to `mt352_attach'
   saa7134-dvb.c:(.text+0x347951): undefined reference to `mt312_attach'
   saa7134-dvb.c:(.text+0x3479a9): undefined reference to `mt312_attach'
>> saa7134-dvb.c:(.text+0x3479c1): undefined reference to `zl10039_attach'

This is happening because a builtin module can't use directly a symbol
found on a module. By enabling CONFIG_MEDIA_ATTACH, the configuration
becomes valid, as dvb_attach() macro loads the module if needed, making
the symbol available to the builtin module.

While this bug started to appear after the patches that use IS_DEFINED
macro (like changeset 7b34be71db533f3e0cf93d53cf62d036cdb5418a), this
bug is a way ancient than that.

The thing is that, before the IS_DEFINED() patches, the logic used to be:

       && defined(MODULE))
struct dvb_frontend *zl10039_attach(struct dvb_frontend *fe,
u8 i2c_addr,
struct i2c_adapter *i2c);
static inline struct dvb_frontend *zl10039_attach(struct dvb_frontend *fe,
u8 i2c_addr,
struct i2c_adapter *i2c)
{
printk(KERN_WARNING "%s: driver disabled by Kconfig\n", __func__);
return NULL;
}

The above code, with the .config file used, was evoluting to FALSE
(instead of TRUE as it should be, as CONFIG_DVB_ZL10039 is 'm'),
and were adding the static inline code at saa7134-dvb, instead
of the external call. So, while it weren't producing any compilation
error, the code weren't working either.

So, as the overhead for using CONFIG_MEDIA_ATTACH is minimal, just
enable it, if MODULES is defined.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>