firefly-linux-kernel-4.4.55.git
10 years agodm thin: fix rcu_read_lock being held in code that can sleep
Joe Thornber [Tue, 8 Apr 2014 10:29:01 +0000 (11:29 +0100)]
dm thin: fix rcu_read_lock being held in code that can sleep

Commit c140e1c4e23 ("dm thin: use per thin device deferred bio lists")
introduced the use of an rculist for all active thin devices.  The use
of rcu_read_lock() in process_deferred_bios() can result in a BUG if a
dm_bio_prison_cell must be allocated as a side-effect of bio_detain():

 BUG: sleeping function called from invalid context at mm/mempool.c:203
 in_atomic(): 1, irqs_disabled(): 0, pid: 6, name: kworker/u8:0
 3 locks held by kworker/u8:0/6:
   #0:  ("dm-" "thin"){.+.+..}, at: [<ffffffff8106be42>] process_one_work+0x192/0x550
   #1:  ((&pool->worker)){+.+...}, at: [<ffffffff8106be42>] process_one_work+0x192/0x550
   #2:  (rcu_read_lock){.+.+..}, at: [<ffffffff816360b5>] do_worker+0x5/0x4d0

We can't process deferred bios with the rcu lock held, since
dm_bio_prison_cell allocation may block if the bio-prison's cell mempool
is exhausted.

To fix:

- Introduce a refcount and completion field to each thin_c

- Add thin_get/put methods for adjusting the refcount.  If the refcount
  hits zero then the completion is triggered.

- Initialise refcount to 1 when creating thin_c

- When iterating the active_thins list we thin_get() whilst the rcu
  lock is held.

- After the rcu lock is dropped we process the deferred bios for that
  thin.

- When destroying a thin_c we thin_put() and then wait for the
  completion -- to avoid a race between the worker thread iterating
  from that thin_c and destroying the thin_c.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm thin: irqsave must always be used with the pool->lock spinlock
Joe Thornber [Tue, 8 Apr 2014 10:08:41 +0000 (11:08 +0100)]
dm thin: irqsave must always be used with the pool->lock spinlock

Commit c140e1c4e23 ("dm thin: use per thin device deferred bio lists")
incorrectly stopped disabling irqs when taking the pool's spinlock.

Irqs must be disabled when taking the pool's spinlock otherwise a thread
could spin_lock(), then get interrupted to service thin_endio() in
interrupt context, which would then deadlock in spin_lock_irqsave().

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm cache: fix a lock-inversion
Joe Thornber [Thu, 3 Apr 2014 15:16:44 +0000 (16:16 +0100)]
dm cache: fix a lock-inversion

When suspending a cache the policy is walked and the individual policy
hints written to the metadata via sync_metadata().  This led to this
lock order:

      policy->lock
        cache_metadata->root_lock

When loading the cache target the policy is populated while the metadata
lock is held:

      cache_metadata->root_lock
         policy->lock

Fix this potential lock-inversion (ABBA) deadlock in sync_metadata() by
ensuring the cache_metadata root_lock is held whilst all the hints are
written, rather than being repeatedly locked while policy->lock is held
(as was the case with each callout that policy_walk_mappings() made to
the old save_hint() method).

Found by turning on the CONFIG_PROVE_LOCKING ("Lock debugging: prove
locking correctness") build option.  However, it is not clear how the
LOCKDEP reported paths can lead to a deadlock since the two paths,
suspending a target and loading a target, never occur at the same time.
But that doesn't mean the same lock-inversion couldn't have occurred
elsewhere.

Reported-by: Marian Csontos <mcsontos@redhat.com>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
10 years agodm thin: sort the per thin deferred bios using an rb_tree
Mike Snitzer [Fri, 21 Mar 2014 22:33:41 +0000 (18:33 -0400)]
dm thin: sort the per thin deferred bios using an rb_tree

A thin-pool will allocate blocks using FIFO order for all thin devices
which share the thin-pool.  Because of this simplistic allocation the
thin-pool's space can become fragmented quite easily; especially when
multiple threads are requesting blocks in parallel.

Sort each thin device's deferred_bio_list based on logical sector to
help reduce fragmentation of the thin-pool's ondisk layout.

The following tables illustrate the realized gains/potential offered by
sorting each thin device's deferred_bio_list.  An "io size"-sized random
read of the device would result in "seeks/io" fragments being read, with
an average "distance/seek" between each fragment.

Data was written to a single thin device using multiple threads via
iozone (8 threads, 64K for both the block_size and io_size).

unsorted:

     io size   seeks/io distance/seek
  --------------------------------------
          4k    0.000   0b
         16k    0.013   11m
         64k    0.065   11m
        256k    0.274   10m
          1m    1.109   10m
          4m    4.411   10m
         16m    17.097  11m
         64m    60.055  13m
        256m    148.798 25m
          1g    809.929 21m

sorted:

     io size   seeks/io distance/seek
  --------------------------------------
          4k    0.000   0b
         16k    0.000   1g
         64k    0.001   1g
        256k    0.003   1g
          1m    0.011   1g
          4m    0.045   1g
         16m    0.181   1g
         64m    0.747   1011m
        256m    3.299   1g
          1g    14.373  1g

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
10 years agodm thin: use per thin device deferred bio lists
Mike Snitzer [Fri, 21 Mar 2014 01:17:14 +0000 (21:17 -0400)]
dm thin: use per thin device deferred bio lists

The thin-pool previously only had a single deferred_bios list that would
collect bios for all thin devices in the pool.  Split this per-pool
deferred_bios list out to per-thin deferred_bios_list -- doing so
enables increased parallelism when processing deferred bios.  And now
that each thin device has it's own deferred_bios_list we can sort all
bios in the list using logical sector.  The requeue code in error
handling path is also cleaner as a side-effect.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
10 years agodm thin: simplify pool_is_congested
Mike Snitzer [Thu, 20 Mar 2014 12:36:47 +0000 (08:36 -0400)]
dm thin: simplify pool_is_congested

The pool is congested if the pool is in PM_OUT_OF_DATA_SPACE mode.  This
is more explicit/clear/efficient than inferring whether or not the pool
is congested by checking if retry_on_resume_list is empty.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
10 years agodm thin: fix dangling bio in process_deferred_bios error path
Mike Snitzer [Fri, 28 Mar 2014 06:15:02 +0000 (02:15 -0400)]
dm thin: fix dangling bio in process_deferred_bios error path

If unable to ensure_next_mapping() we must add the current bio, which
was removed from the @bios list via bio_list_pop, back to the
deferred_bios list before all the remaining @bios.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Cc: stable@vger.kernel.org
10 years agodm mpath: print more useful warnings in multipath_message()
Jose Castillo [Wed, 29 Jan 2014 16:52:45 +0000 (17:52 +0100)]
dm mpath: print more useful warnings in multipath_message()

The warning message "Unrecognised multipath message received" is
displayed in two different situations in multipath_message(): when the
number of arguments passed is invalid and when the string passed in
argv[0] is not recognized.

Make it easier to identify where the problem is by making these warnings
more specific with additional context for each case.

Signed-off-by: Jose Castillo <jcastillo@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm-mpath: do not activate failed paths
Hannes Reinecke [Fri, 28 Feb 2014 14:33:49 +0000 (15:33 +0100)]
dm-mpath: do not activate failed paths

activate_path() is run without a lock, so the path might be
set to failed before activate_path() had a chance to run.
This patch add a check for ->active in activate_path() to
avoid unnecessary overhead by calling functions which are known
to be failing.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm mpath: remove extra nesting in map function
Mike Snitzer [Fri, 28 Feb 2014 14:33:48 +0000 (15:33 +0100)]
dm mpath: remove extra nesting in map function

Return early for case when no path exists, and when the
pathgroup isn't ready. This eliminates the need for
extra nesting for the the common case.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
10 years agodm mpath: remove map_io()
Hannes Reinecke [Fri, 28 Feb 2014 14:33:47 +0000 (15:33 +0100)]
dm mpath: remove map_io()

multipath_map() is now just a wrapper around map_io(), so we
can rename map_io() to multipath_map().

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
10 years agodm mpath: reduce memory pressure when requeuing
Hannes Reinecke [Fri, 28 Feb 2014 14:33:46 +0000 (15:33 +0100)]
dm mpath: reduce memory pressure when requeuing

When multipath needs to requeue I/O in the block layer the per-request
context shouldn't be allocated, as it will be freed immediately
afterwards anyway.  Avoiding this memory allocation will reduce memory
pressure during requeuing.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
10 years agodm mpath: remove process_queued_ios()
Hannes Reinecke [Fri, 28 Feb 2014 14:33:45 +0000 (15:33 +0100)]
dm mpath: remove process_queued_ios()

process_queued_ios() has served 3 functions:
  1) select pg and pgpath if none is selected
  2) start pg_init if requested
  3) dispatch queued IOs when pg is ready

Basically, a call to queue_work(process_queued_ios) can be replaced by
dm_table_run_md_queue_async(), which runs request queue and ends up
calling map_io(), which does 1), 2) and 3).

Exception is when !pg_ready() (which means either pg_init is running or
requested), then multipath_busy() prevents map_io() being called from
request_fn.

If pg_init is running, it should be ok as long as pg_init_done() does
the right thing when pg_init is completed, I.e.: restart pg_init if
!pg_ready() or call dm_table_run_md_queue_async() to kick map_io().

If pg_init is requested, we have to make sure the request is detected
and pg_init will be started.  pg_init is requested in 3 places:
  a) __choose_pgpath() in map_io()
  b) __choose_pgpath() in multipath_ioctl()
  c) pg_init retry in pg_init_done()
a) is ok because map_io() calls __pg_init_all_paths(), which does 2).
b) needs a call to __pg_init_all_paths(), which does 2).
c) needs a call to __pg_init_all_paths(), which does 2).

So this patch removes process_queued_ios() and ensures that
__pg_init_all_paths() is called at the appropriate locations.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
10 years agodm mpath: push back requests instead of queueing
Hannes Reinecke [Fri, 28 Feb 2014 14:33:44 +0000 (15:33 +0100)]
dm mpath: push back requests instead of queueing

There is no reason why multipath needs to queue requests internally for
queue_if_no_path or pg_init; we should rather push them back onto the
request queue.

And while we're at it we can simplify the conditional statement in
map_io() to make it easier to read.

Since mpath no longer does internal queuing of I/O the table info no
longer emits the internal queue_size.  Instead it displays 1 if queuing
is being used or 0 if it is not.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
10 years agodm table: add dm_table_run_md_queue_async
Mike Snitzer [Fri, 28 Feb 2014 14:33:43 +0000 (15:33 +0100)]
dm table: add dm_table_run_md_queue_async

Introduce dm_table_run_md_queue_async() to run the request_queue of the
mapped_device associated with a request-based DM table.

Also add dm_md_get_queue() wrapper to extract the request_queue from a
mapped_device.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
10 years agodm mpath: do not call pg_init when it is already running
Hannes Reinecke [Fri, 28 Feb 2014 14:33:42 +0000 (15:33 +0100)]
dm mpath: do not call pg_init when it is already running

This patch moves condition checks as a preparation of following
patches and has no effect on behaviour.
process_queued_ios() is the only caller of __pg_init_all_paths()
and 2 condition checks are moved from outside to inside without
side effects.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
10 years agodm: use RCU_INIT_POINTER instead of rcu_assign_pointer in __unbind
Monam Agarwal [Sun, 23 Mar 2014 18:28:27 +0000 (23:58 +0530)]
dm: use RCU_INIT_POINTER instead of rcu_assign_pointer in __unbind

Replace rcu_assign_pointer(p, NULL) with RCU_INIT_POINTER(p, NULL).

The rcu_assign_pointer() ensures that the initialization of a structure
is carried out before storing a pointer to that structure.  And in the
case of the NULL pointer, there is no structure to initialize.  So,
rcu_assign_pointer(p, NULL) can be safely converted to
RCU_INIT_POINTER(p, NULL).

Signed-off-by: Monam Agarwal <monamagarwal123@gmail.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm: stop using bi_private
Mikulas Patocka [Tue, 4 Mar 2014 23:24:49 +0000 (18:24 -0500)]
dm: stop using bi_private

Device mapper uses the bio structure's bi_private field as a pointer
to dm_target_io or dm_rq_clone_bio_info.  But a bio structure is
embedded in the dm_target_io and dm_rq_clone_bio_info structures, so the
pointer to the structure that contains the bio can be found with the
container_of() macro.

Remove the use of bi_private and use container_of() instead.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm: remove dm_get_mapinfo
Mikulas Patocka [Mon, 3 Mar 2014 22:43:31 +0000 (17:43 -0500)]
dm: remove dm_get_mapinfo

Remove dm_get_mapinfo() because no target uses it.  Targets can allocate
per-bio data using ti->per_bio_data_size, this is much more flexible
than union map_info.

Leave union map_info only for the request-based multipath target's use.
Also delete the unused "unsigned long long ll" field of union map_info.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm: make dm_table_alloc_md_mempools static
Mikulas Patocka [Thu, 13 Feb 2014 18:43:32 +0000 (13:43 -0500)]
dm: make dm_table_alloc_md_mempools static

Make the function dm_table_alloc_md_mempools static because it is not
called from another file.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm: take care to copy the space map roots before locking the superblock
Joe Thornber [Thu, 27 Mar 2014 14:13:23 +0000 (14:13 +0000)]
dm: take care to copy the space map roots before locking the superblock

In theory copying the space map root can fail, but in practice it never
does because we're careful to check what size buffer is needed.

But make certain we're able to copy the space map roots before
locking the superblock.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org # drop dm-era and dm-cache changes as needed
10 years agodm transaction manager: fix corruption due to non-atomic transaction commit
Joe Thornber [Thu, 27 Mar 2014 14:13:20 +0000 (14:13 +0000)]
dm transaction manager: fix corruption due to non-atomic transaction commit

The persistent-data library used by dm-thin, dm-cache, etc is
transactional.  If anything goes wrong, such as an io error when writing
new metadata or a power failure, then we roll back to the last
transaction.

Atomicity when committing a transaction is achieved by:

a) Never overwriting data from the previous transaction.
b) Writing the superblock last, after all other metadata has hit the
   disk.

This commit and the following commit ("dm: take care to copy the space
map roots before locking the superblock") fix a bug associated with (b).
When committing it was possible for the superblock to still be written
in spite of an io error occurring during the preceeding metadata flush.
With these commits we're careful not to take the write lock out on the
superblock until after the metadata flush has completed.

Change the transaction manager's semantics for dm_tm_commit() to assume
all data has been flushed _before_ the single superblock that is passed
in.

As a prerequisite, split the block manager's block unlocking and
flushing by simplifying dm_bm_flush_and_unlock() to dm_bm_flush().  Now
the unlocking must be done separately.

This issue was discovered by forcing io errors at the crucial time
using dm-flakey.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
10 years agodm cache: remove remainder of distinct discard block size
Heinz Mauelshagen [Thu, 27 Mar 2014 19:14:10 +0000 (15:14 -0400)]
dm cache: remove remainder of distinct discard block size

Discard block size not being equal to cache block size causes data
corruption by erroneously avoiding migrations in issue_copy() because
the discard state is being cleared for a group of cache blocks when it
should not.

Completely remove all code that enabled a distinction between the
cache block size and discard block size.

Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm cache: prevent corruption caused by discard_block_size > cache_block_size
Mike Snitzer [Thu, 20 Mar 2014 14:11:15 +0000 (10:11 -0400)]
dm cache: prevent corruption caused by discard_block_size > cache_block_size

If the discard block size is larger than the cache block size we will
not properly quiesce IO to a region that is about to be discarded.  This
results in a race between a cache migration where no copy is needed, and
a write to an adjacent cache block that's within the same large discard
block.

Workaround this by limiting the discard_block_size to cache_block_size.
Also limit the max_discard_sectors to cache_block_size.

A more comprehensive fix that introduces range locking support in the
bio_prison and proper quiescing of a discard range that spans multiple
cache blocks is already in development.

Reported-by: Morgan Mears <Morgan.Mears@netapp.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Acked-by: Heinz Mauelshagen <heinzm@redhat.com>
Cc: stable@vger.kernel.org
10 years agodm bitset: only flush the current word if it has been dirtied
Joe Thornber [Mon, 3 Mar 2014 15:37:24 +0000 (10:37 -0500)]
dm bitset: only flush the current word if it has been dirtied

This change offers a big performance boost for dm-era.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agodm: add era target
Joe Thornber [Mon, 3 Mar 2014 15:23:15 +0000 (10:23 -0500)]
dm: add era target

dm-era is a target that behaves similar to the linear target.  In
addition it keeps track of which blocks were written within a user
defined period of time called an 'era'.  Each era target instance
maintains the current era as a monotonically increasing 32-bit
counter.

Use cases include tracking changed blocks for backup software, and
partially invalidating the contents of a cache to restore cache
coherency after rolling back a vendor snapshot.

dm-era is primarily expected to be paired with the dm-cache target.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agoLinux 3.14-rc8
Linus Torvalds [Tue, 25 Mar 2014 02:31:17 +0000 (19:31 -0700)]
Linux 3.14-rc8

10 years agoMerge branch 'parisc-3.14' of git://git.kernel.org/pub/scm/linux/kernel/git/deller...
Linus Torvalds [Tue, 25 Mar 2014 00:36:58 +0000 (17:36 -0700)]
Merge branch 'parisc-3.14' of git://git./linux/kernel/git/deller/parisc-linux

Pull parisc updates from Helge Deller:
 - revert parts of the latest patch regarding font selection with STICON
   console
 - wire up the utimes() syscall for parisc
 - remove the unused parisc tmpalias code and unnecessary arch*relax
   defines

* 'parisc-3.14' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: locks: remove redundant arch_*_relax operations
  parisc: wire up sys_utimes
  parisc: Remove unused CONFIG_PARISC_TMPALIAS code
  partly revert commit 8a10bc9: parisc/sti_console: prefer Linux fonts over built-in ROM fonts

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc
Linus Torvalds [Tue, 25 Mar 2014 00:30:44 +0000 (17:30 -0700)]
Merge git://git./linux/kernel/git/davem/sparc

Pull sparc fixes from David Miller:

 1) Do serial locking in a way that makes things clear that these are
    IRQ spinlocks.

 2) Conversion to generic idle loop broke first generation Niagara
    machines, need to have %pil interrupts enabled during cpu yield
    hypervisor call.

 3) Do not use magic constants for iterations over tsb tables, from Doug
    Wilson.

 4) Fix erroneous truncation of 64-bit system call return values to
    32-bit.  From Dave Kleikamp.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  sparc64: Make sure %pil interrupts are enabled during hypervisor yield.
  sparc64:tsb.c:use array size macro rather than number
  sparc64: don't treat 64-bit syscall return codes as 32-bit
  sparc: serial: Clean up the locking for -rt

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Tue, 25 Mar 2014 00:07:24 +0000 (17:07 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) OpenVswitch's lookup_datapath() returns error pointers, so don't
    check against NULL.  From Jiri Pirko.

 2) pfkey_compile_policy() code path tries to do a GFP_KERNEL allocation
    under RCU locks, fix by using GFP_ATOMIC when necessary.  From
    Nikolay Aleksandrov.

 3) phy_suspend() indirectly passes uninitialized data into the ethtool
    get wake-on-land implementations.  Fix from Sebastian Hesselbarth.

 4) CPSW driver unregisters CPTS twice, fix from Benedikt Spranger.

 5) If SKB allocation of reply packet fails, vxlan's arp_reduce() defers
    a NULL pointer.  Fix from David Stevens.

 6) IPV6 neigh handling in vxlan doesn't validate the destination
    address properly, and it builds a packet with the src and dst
    reversed.  Fix also from David Stevens.

 7) Fix spinlock recursion during subscription failures in TIPC stack,
    from Erik Hugne.

 8) Revert buggy conversion of davinci_emac to devm_request_irq, from
    Chrstian Riesch.

 9) Wrong flags passed into forwarding database netlink notifications,
    from Nicolas Dichtel.

10) The netpoll neighbour soliciation handler checks wrong ethertype,
    needs to be ETH_P_IPV6 rather than ETH_P_ARP.  Fix from Li RongQing.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (34 commits)
  tipc: fix spinlock recursion bug for failed subscriptions
  vxlan: fix nonfunctional neigh_reduce()
  net: davinci_emac: Fix rollback of emac_dev_open()
  net: davinci_emac: Replace devm_request_irq with request_irq
  netpoll: fix the skb check in pkt_is_ns
  net: micrel : ks8851-ml: add vdd-supply support
  ip6mr: fix mfc notification flags
  ipmr: fix mfc notification flags
  rtnetlink: fix fdb notification flags
  tcp: syncookies: do not use getnstimeofday()
  netlink: fix setsockopt in mmap examples in documentation
  openvswitch: Correctly report flow used times for first 5 minutes after boot.
  via-rhine: Disable device in error path
  ATHEROS-ATL1E: Convert iounmap to pci_iounmap
  vxlan: fix potential NULL dereference in arp_reduce()
  cnic: Update version to 2.5.20 and copyright year.
  cnic,bnx2i,bnx2fc: Fix inconsistent use of page size
  cnic: Use proper ulp_ops for per device operations.
  net: cdc_ncm: fix control message ordering
  ipv6: ip6_append_data_mtu do not handle the mtu of the second fragment properly
  ...

10 years agotipc: fix spinlock recursion bug for failed subscriptions
Erik Hugne [Mon, 24 Mar 2014 15:56:38 +0000 (16:56 +0100)]
tipc: fix spinlock recursion bug for failed subscriptions

If a topology event subscription fails for any reason, such as out
of memory, max number reached or because we received an invalid
request the correct behavior is to terminate the subscribers
connection to the topology server. This is currently broken and
produces the following oops:

[27.953662] tipc: Subscription rejected, illegal request
[27.955329] BUG: spinlock recursion on CPU#1, kworker/u4:0/6
[27.957066]  lock: 0xffff88003c67f408, .magic: dead4ead, .owner: kworker/u4:0/6, .owner_cpu: 1
[27.958054] CPU: 1 PID: 6 Comm: kworker/u4:0 Not tainted 3.14.0-rc6+ #5
[27.960230] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
[27.960874] Workqueue: tipc_rcv tipc_recv_work [tipc]
[27.961430]  ffff88003c67f408 ffff88003de27c18 ffffffff815c0207 ffff88003de1c050
[27.962292]  ffff88003de27c38 ffffffff815beec5 ffff88003c67f408 ffffffff817f0a8a
[27.963152]  ffff88003de27c58 ffffffff815beeeb ffff88003c67f408 ffffffffa0013520
[27.964023] Call Trace:
[27.964292]  [<ffffffff815c0207>] dump_stack+0x45/0x56
[27.964874]  [<ffffffff815beec5>] spin_dump+0x8c/0x91
[27.965420]  [<ffffffff815beeeb>] spin_bug+0x21/0x26
[27.965995]  [<ffffffff81083df6>] do_raw_spin_lock+0x116/0x140
[27.966631]  [<ffffffff815c6215>] _raw_spin_lock_bh+0x15/0x20
[27.967256]  [<ffffffffa0008540>] subscr_conn_shutdown_event+0x20/0xa0 [tipc]
[27.968051]  [<ffffffffa000fde4>] tipc_close_conn+0xa4/0xb0 [tipc]
[27.968722]  [<ffffffffa00101ba>] tipc_conn_terminate+0x1a/0x30 [tipc]
[27.969436]  [<ffffffffa00089a2>] subscr_conn_msg_event+0x1f2/0x2f0 [tipc]
[27.970209]  [<ffffffffa0010000>] tipc_receive_from_sock+0x90/0xf0 [tipc]
[27.970972]  [<ffffffffa000fa79>] tipc_recv_work+0x29/0x50 [tipc]
[27.971633]  [<ffffffff8105dbf5>] process_one_work+0x165/0x3e0
[27.972267]  [<ffffffff8105e869>] worker_thread+0x119/0x3a0
[27.972896]  [<ffffffff8105e750>] ? manage_workers.isra.25+0x2a0/0x2a0
[27.973622]  [<ffffffff810648af>] kthread+0xdf/0x100
[27.974168]  [<ffffffff810647d0>] ? kthread_create_on_node+0x1a0/0x1a0
[27.974893]  [<ffffffff815ce13c>] ret_from_fork+0x7c/0xb0
[27.975466]  [<ffffffff810647d0>] ? kthread_create_on_node+0x1a0/0x1a0

The recursion occurs when subscr_terminate tries to grab the
subscriber lock, which is already taken by subscr_conn_msg_event.
We fix this by checking if the request to establish a new
subscription was successful, and if not we initiate termination of
the subscriber after we have released the subscriber lock.

Signed-off-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agovxlan: fix nonfunctional neigh_reduce()
David Stevens [Mon, 24 Mar 2014 14:39:58 +0000 (10:39 -0400)]
vxlan: fix nonfunctional neigh_reduce()

The VXLAN neigh_reduce() code is completely non-functional since
check-in. Specific errors:

1) The original code drops all packets with a multicast destination address,
even though neighbor solicitations are sent to the solicited-node
address, a multicast address. The code after this check was never run.
2) The neighbor table lookup used the IPv6 header destination, which is the
solicited node address, rather than the target address from the
neighbor solicitation. So neighbor lookups would always fail if it
got this far. Also for L3MISSes.
3) The code calls ndisc_send_na(), which does a send on the tunnel device.
The context for neigh_reduce() is the transmit path, vxlan_xmit(),
where the host or a bridge-attached neighbor is trying to transmit
a neighbor solicitation. To respond to it, the tunnel endpoint needs
to do a *receive* of the appropriate neighbor advertisement. Doing a
send, would only try to send the advertisement, encapsulated, to the
remote destinations in the fdb -- hosts that definitely did not do the
corresponding solicitation.
4) The code uses the tunnel endpoint IPv6 forwarding flag to determine the
isrouter flag in the advertisement. This has nothing to do with whether
or not the target is a router, and generally won't be set since the
tunnel endpoint is bridging, not routing, traffic.

The patch below creates a proxy neighbor advertisement to respond to
neighbor solicitions as intended, providing proper IPv6 support for neighbor
reduction.

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'davinci_emac'
David S. Miller [Mon, 24 Mar 2014 19:32:34 +0000 (15:32 -0400)]
Merge branch 'davinci_emac'

Christian Riesch says:

====================
net: davinci_emac: Fix interrupt requests and error handling

since commit 6892b41d9701283085b655c6086fb57a5d63fa47 (Linux 3.11) the
davinci_emac driver is broken. After doing ifconfig down, ifconfig up,
requesting the interrupts for the driver fails. The interface remains dead
until the board is rebooted.

The first patch in this patchset reverts commit
6892b41d9701283085b655c6086fb57a5d63fa47 partially and makes the driver
useable again.

During the work on the first patch, a number of bugs in the error handling
of the driver's ndo_open code were found. The second patch fixes these bugs.

I believe the first patch meets the rules for stable kernels, I therefore added
the stable tag to this patch. The second patch is just cleanup, the code
that is fixed by this patch is only executed in case of an error.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: davinci_emac: Fix rollback of emac_dev_open()
Christian Riesch [Mon, 24 Mar 2014 12:46:27 +0000 (13:46 +0100)]
net: davinci_emac: Fix rollback of emac_dev_open()

If an error occurs during the initialization in emac_dev_open() (the
driver's ndo_open function), interrupts, DMA descriptors etc. must be freed.
The current rollback code is buggy in several ways.

  1) Freeing the interrupts. The current code will not free all interrupts
     that were requested by the driver. Furthermore,  the code tries to do a
     platform_get_resource(priv->pdev, IORESOURCE_IRQ, -1) in its last
     iteration.

     This patch fixes these bugs.

  2) Wrong order of err: and rollback: labels. If the setup of the PHY in
     the code fails, the interrupts that have been requested before are
     not freed:

        request irq
                if requesting irqs fails, goto rollback
        setup phy
                if phy setup fails, goto err
        return 0

     rollback:
        free irqs
     err:

     This patch brings the code into the correct order.

  3) The code calls napi_enable() and emac_int_enable(), but does not
     undo both in case of an error.

     This patch adds calls of emac_int_disable() and napi_disable() to the
     rollback code.

  4) RX DMA descriptors are not freed in case of an error: Right before
     requesting the irqs, the function creates DMA descriptors for the
     RX channel. These RX descriptors are never freed when we jump to either
     rollback or err.

     This patch adds code for freeing the DMA descriptors in the case of
     an initialization error. This required a modification of
     cpdma_ctrl_stop() in davinci_cpdma.c: We must be able to call this
     function to free the DMA descriptors while the DMA channels are
     in IDLE state (before cpdma_ctlr_start() was called).

Tested on a custom board with the Texas Instruments AM1808.

Signed-off-by: Christian Riesch <christian.riesch@omicron.at>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: davinci_emac: Replace devm_request_irq with request_irq
Christian Riesch [Mon, 24 Mar 2014 12:46:26 +0000 (13:46 +0100)]
net: davinci_emac: Replace devm_request_irq with request_irq

In commit 6892b41d9701283085b655c6086fb57a5d63fa47

Author: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Date:   Tue Jun 25 21:24:51 2013 +0530
net: davinci: emac: Convert to devm_* api

the call of request_irq is replaced by devm_request_irq and the call
of free_irq is removed. But since interrupts are requested in
emac_dev_open, doing ifconfig up/down on the board requests the
interrupts again each time, causing devm_request_irq to fail. The
interface is dead until the device is rebooted.

This patch reverts said commit partially: It changes the driver back
to use request_irq instead of devm_request_irq, puts free_irq back in
place, but keeps the remaining changes of the original patch.

Reported-by: Jon Ringle <jon@ringle.org>
Signed-off-by: Christian Riesch <christian.riesch@omicron.at>
Cc: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetpoll: fix the skb check in pkt_is_ns
Li RongQing [Fri, 21 Mar 2014 12:53:57 +0000 (20:53 +0800)]
netpoll: fix the skb check in pkt_is_ns

Neighbor Solicitation is ipv6 protocol, so we should check
skb->protocol with ETH_P_IPV6

Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Cc: WANG Cong <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agosparc64: Make sure %pil interrupts are enabled during hypervisor yield.
David S. Miller [Mon, 24 Mar 2014 18:45:12 +0000 (14:45 -0400)]
sparc64: Make sure %pil interrupts are enabled during hypervisor yield.

In arch_cpu_idle() we must enable %pil based interrupts before
potentially invoking the hypervisor cpu yield call.

As per the Hypervisor API documentation for cpu_yield:

Interrupts which are blocked by some mechanism other that
pstate.ie (for example %pil) are not guaranteed to cause
a return from this service.

It seems that only first generation Niagara chips are hit by this
bug.  My best guess is that later chips implement this in hardware
and wake up anyways from %pil events, whereas in first generation
chips the yield is implemented completely in hypervisor code and
requires %pil to be enabled in order to wake properly from this
call.

Fixes: 87fa05aeb3a5 ("sparc: Use generic idle loop")
Reported-by: Fabio M. Di Nitto <fabbione@fabbione.net>
Reported-by: Jan Engelhardt <jengelh@inai.de>
Tested-by: Jan Engelhardt <jengelh@inai.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: micrel : ks8851-ml: add vdd-supply support
Nishanth Menon [Fri, 21 Mar 2014 06:52:48 +0000 (01:52 -0500)]
net: micrel : ks8851-ml: add vdd-supply support

Few platforms use external regulator to keep the ethernet MAC supplied.
So, request and enable the regulator for driver functionality.

Fixes: 66fda75f47dc (regulator: core: Replace direct ops->disable usage)
Reported-by: Russell King <rmk+kernel@arm.linux.org.uk>
Suggested-by: Markus Pargmann <mpa@pengutronix.de>
Signed-off-by: Nishanth Menon <nm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoparisc: locks: remove redundant arch_*_relax operations
Will Deacon [Fri, 21 Feb 2014 17:34:57 +0000 (17:34 +0000)]
parisc: locks: remove redundant arch_*_relax operations

Now that the arch_{spin,read,write}_relax macros default to cpu_relax(),
remove the redundant definitions for parisc.

Cc: Helge Deller <deller@gmx.de>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Helge Deller <deller@gmx.de>
10 years agoparisc: wire up sys_utimes
Helge Deller [Sun, 23 Mar 2014 14:24:40 +0000 (15:24 +0100)]
parisc: wire up sys_utimes

We seem to be nearly the only platform which does not provide the
sys_utimes syscall.  Adding it now makes our life much easier with
userspace applications (like dietlibc and e2fsprogs) since we then
behave like all other platforms too and don't need extra patches which
are hard to get upstream anyway because we are not a mainstream
architecture.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org # v3.13
10 years agoparisc: Remove unused CONFIG_PARISC_TMPALIAS code
John David Anglin [Sat, 1 Mar 2014 22:41:22 +0000 (17:41 -0500)]
parisc: Remove unused CONFIG_PARISC_TMPALIAS code

The attached change removes the unused and experimental
CONFIG_PARISC_TMPALIAS code. It doesn't work and I don't believe it will
ever be used.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Signed-off-by: Helge Deller <deller@gmx.de>
10 years agopartly revert commit 8a10bc9: parisc/sti_console: prefer Linux fonts over built-in...
Helge Deller [Sun, 23 Mar 2014 15:39:52 +0000 (16:39 +0100)]
partly revert commit 8a10bc9: parisc/sti_console: prefer Linux fonts over built-in ROM fonts

STI console is used on parisc and m68k HP machines. This patch partly reverts
my previous commit and as such restores the fonts for the m68k machines.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org # v3.13
10 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 22 Mar 2014 16:13:59 +0000 (09:13 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "Two oneliner 'perf bench' tooling fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf bench: Fix NULL pointer dereference in "perf bench all"
  perf bench numa: Make no args mean 'run all tests'

10 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Sat, 22 Mar 2014 16:13:13 +0000 (09:13 -0700)]
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "Only two patches this time, one to fix ethernet probe order on at91
  (better fix with proper device aliasing will be done for 3.15, this is
  stop-gap), and one update to MAINTAINERS due to Freescale moving their
  repo to kernel.org"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: at91: fix network interface ordering for sama5d36
  MAINTAINERS: update IMX kernel git tree

10 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 21 Mar 2014 05:32:28 +0000 (22:32 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Some final few intel fixes, all regressions, all stable cc, and one
  exynos oops fixer.

  The biggest is probably the intel display error irqs one, but it seems
  to fix a few crashes on startup, and one use after free in drm core"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/exynos: Fix (more) freeing issues in exynos_drm_drv.c
  drm/i915: Disable stolen memory when DMAR is active
  Revert "drm/i915: don't touch the VDD when disabling the panel"
  drm: Fix use-after-free in the shadow-attache exit code
  drm/i915: Don't enable display error interrupts from the start
  drm/i915: Fix scanline counter fixup on BDW
  drm/i915: Add a workaround for HSW scanline counter weirdness
  drm/i915: Fix PSR programming

10 years agoblock: free q->flush_rq in blk_init_allocated_queue error paths
Dave Jones [Thu, 20 Mar 2014 21:03:58 +0000 (15:03 -0600)]
block: free q->flush_rq in blk_init_allocated_queue error paths

Commit 7982e90c3a57 ("block: fix q->flush_rq NULL pointer crash on
dm-mpath flush") moved an allocation to blk_init_allocated_queue(), but
neglected to free that allocation on the error paths that follow.

Signed-off-by: Dave Jones <davej@fedoraproject.org>
Acked-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agofutex: revert back to the explicit waiter counting code
Linus Torvalds [Fri, 21 Mar 2014 05:11:17 +0000 (22:11 -0700)]
futex: revert back to the explicit waiter counting code

Srikar Dronamraju reports that commit b0c29f79ecea ("futexes: Avoid
taking the hb->lock if there's nothing to wake up") causes java threads
getting stuck on futexes when runing specjbb on a power7 numa box.

The cause appears to be that the powerpc spinlocks aren't using the same
ticket lock model that we use on x86 (and other) architectures, which in
turn result in the "spin_is_locked()" test in hb_waiters_pending()
occasionally reporting an unlocked spinlock even when there are pending
waiters.

So this reinstates Davidlohr Bueso's original explicit waiter counting
code, which I had convinced Davidlohr to drop in favor of figuring out
the pending waiters by just using the existing state of the spinlock and
the wait queue.

Reported-and-tested-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Original-code-by: Davidlohr Bueso <davidlohr@hp.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMerge tag 'trace-fixes-v3.14-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 21 Mar 2014 05:09:30 +0000 (22:09 -0700)]
Merge tag 'trace-fixes-v3.14-rc7' of git://git./linux/kernel/git/rostedt/linux-trace

Pull trace fix from Steven Rostedt:
 "Vaibhav Nagarnaik discovered that since 3.10 a clean-up patch made the
  array index in the trace event format bogus.

  He supplied an elegant solution that uses __stringify() and also
  removes the need for the event_storage and event_storage_mutex and
  also cuts off a few K of overhead from the trace events"

* tag 'trace-fixes-v3.14-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Fix array size mismatch in format string

10 years agomm: fix swapops.h:131 bug if remap_file_pages raced migration
Hugh Dickins [Fri, 21 Mar 2014 04:52:17 +0000 (21:52 -0700)]
mm: fix swapops.h:131 bug if remap_file_pages raced migration

Add remove_linear_migration_ptes_from_nonlinear(), to fix an interesting
little include/linux/swapops.h:131 BUG_ON(!PageLocked) found by trinity:
indicating that remove_migration_ptes() failed to find one of the
migration entries that was temporarily inserted.

The problem comes from remap_file_pages()'s switch from vma_interval_tree
(good for inserting the migration entry) to i_mmap_nonlinear list (no good
for locating it again); but can only be a problem if the remap_file_pages()
range does not cover the whole of the vma (zap_pte() clears the range).

remove_migration_ptes() needs a file_nonlinear method to go down the
i_mmap_nonlinear list, applying linear location to look for migration
entries in those vmas too, just in case there was this race.

The file_nonlinear method does need rmap_walk_control.arg to do this;
but it never needed vma passed in - vma comes from its own iteration.

Reported-and-tested-by: Dave Jones <davej@redhat.com>
Reported-and-tested-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jesse/openvswitch
David S. Miller [Thu, 20 Mar 2014 21:29:02 +0000 (17:29 -0400)]
Merge branch 'fixes' of git://git./linux/kernel/git/jesse/openvswitch

Jesse Gross says:

====================
Open vSwitch

Four small fixes for net/3.14. I realize that these are late in the
cycle - just got back from vacation.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoip6mr: fix mfc notification flags
Nicolas Dichtel [Wed, 19 Mar 2014 16:47:51 +0000 (17:47 +0100)]
ip6mr: fix mfc notification flags

Commit 812e44dd1829 ("ip6mr: advertise new mfc entries via rtnl") reuses the
function ip6mr_fill_mroute() to notify mfc events.
But this function was used only for dump and thus was always setting the
flag NLM_F_MULTI, which is wrong in case of a single notification.

Libraries like libnl will wait forever for NLMSG_DONE.

CC: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipmr: fix mfc notification flags
Nicolas Dichtel [Wed, 19 Mar 2014 16:47:50 +0000 (17:47 +0100)]
ipmr: fix mfc notification flags

Commit 8cd3ac9f9b7b ("ipmr: advertise new mfc entries via rtnl") reuses the
function ipmr_fill_mroute() to notify mfc events.
But this function was used only for dump and thus was always setting the
flag NLM_F_MULTI, which is wrong in case of a single notification.

Libraries like libnl will wait forever for NLMSG_DONE.

CC: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agortnetlink: fix fdb notification flags
Nicolas Dichtel [Wed, 19 Mar 2014 16:47:49 +0000 (17:47 +0100)]
rtnetlink: fix fdb notification flags

Commit 3ff661c38c84 ("net: rtnetlink notify events for FDB NTF_SELF adds and
deletes") reuses the function nlmsg_populate_fdb_fill() to notify fdb events.
But this function was used only for dump and thus was always setting the
flag NLM_F_MULTI, which is wrong in case of a single notification.

Libraries like libnl will wait forever for NLMSG_DONE.

CC: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: syncookies: do not use getnstimeofday()
Eric Dumazet [Thu, 20 Mar 2014 04:02:21 +0000 (21:02 -0700)]
tcp: syncookies: do not use getnstimeofday()

While it is true that getnstimeofday() uses about 40 cycles if TSC
is available, it can use 1600 cycles if hpet is the clocksource.

Switch to get_jiffies_64(), as this is more than enough, and
go back to 60 seconds periods.

Fixes: 8c27bd75f04f ("tcp: syncookies: reduce cookie lifetime to 128 seconds")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetlink: fix setsockopt in mmap examples in documentation
stephen hemminger [Thu, 20 Mar 2014 04:54:20 +0000 (21:54 -0700)]
netlink: fix setsockopt in mmap examples in documentation

The documentation for how to use netlink mmap interface is incorrect.
The calls to setsockopt() require an additional argument.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Thu, 20 Mar 2014 17:57:52 +0000 (10:57 -0700)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS fixes from Ralf Baechle:
 "Another set of five fixes.  The most interesting one is a fix for race
  condition in the local_irq_disable() implementation used by .S code
  for pre-MIPS R2 processors only.  It leaves a race that's hard but not
  impossible to hit; the others fairly obvious"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: Make local_irq_disable macro safe for non-Mipsr2
  MIPS: Octeon: Fix warning in of_device_alloc on cn3xxx
  MIPS: ftrace: Tweak safe_load()/safe_store() macros
  MIPS: BCM47XX: Check all (32) GPIOs when looking for a pin
  MIPS: Fix possible build error with transparent hugepages enabled

10 years agoMerge tag 'sound-3.14' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
Linus Torvalds [Thu, 20 Mar 2014 17:56:37 +0000 (10:56 -0700)]
Merge tag 'sound-3.14' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Just two minor bug fixes: a fix for a regression in oxygen driver that
  was introduced in 3.14-rc1, and a stable fix for the return value of
  compress offload open callback"

* tag 'sound-3.14' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: compress: Pass through return value of open ops callback
  ALSA: oxygen: Xonar DG(X): fix Stereo Upmixing regression

10 years agoopenvswitch: Correctly report flow used times for first 5 minutes after boot.
Ben Pfaff [Thu, 20 Mar 2014 17:45:21 +0000 (10:45 -0700)]
openvswitch: Correctly report flow used times for first 5 minutes after boot.

The kernel starts out its "jiffies" timer as 5 minutes below zero, as
shown in include/linux/jiffies.h:

  /*
   * Have the 32 bit jiffies value wrap 5 minutes after boot
   * so jiffies wrap bugs show up earlier.
   */
  #define INITIAL_JIFFIES ((unsigned long)(unsigned int) (-300*HZ))

The loop in ovs_flow_stats_get() starts out with 'used' set to 0, then
takes any "later" time.  This means that for the first five minutes after
boot, flows will always be reported as never used, since 0 is greater than
any time already seen.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
10 years agotracing: Fix array size mismatch in format string
Vaibhav Nagarnaik [Fri, 14 Feb 2014 03:51:48 +0000 (19:51 -0800)]
tracing: Fix array size mismatch in format string

In event format strings, the array size is reported in two locations.
One in array subscript and then via the "size:" attribute. The values
reported there have a mismatch.

For e.g., in sched:sched_switch the prev_comm and next_comm character
arrays have subscript values as [32] where as the actual field size is
16.

name: sched_switch
ID: 301
format:
        field:unsigned short common_type;       offset:0;       size:2; signed:0;
        field:unsigned char common_flags;       offset:2;       size:1; signed:0;
        field:unsigned char common_preempt_count;       offset:3;       size:1;signed:0;
        field:int common_pid;   offset:4;       size:4; signed:1;

        field:char prev_comm[32];       offset:8;       size:16;        signed:1;
        field:pid_t prev_pid;   offset:24;      size:4; signed:1;
        field:int prev_prio;    offset:28;      size:4; signed:1;
        field:long prev_state;  offset:32;      size:8; signed:1;
        field:char next_comm[32];       offset:40;      size:16;        signed:1;
        field:pid_t next_pid;   offset:56;      size:4; signed:1;
        field:int next_prio;    offset:60;      size:4; signed:1;

After bisection, the following commit was blamed:
92edca0 tracing: Use direct field, type and system names

This commit removes the duplication of strings for field->name and
field->type assuming that all the strings passed in
__trace_define_field() are immutable. This is not true for arrays, where
the type string is created in event_storage variable and field->type for
all array fields points to event_storage.

Use __stringify() to create a string constant for the type string.

Also, get rid of event_storage and event_storage_mutex that are not
needed anymore.

also, an added benefit is that this reduces the overhead of events a bit more:

   text    data     bss     dec     hex filename
8424787 2036472 1302528 11763787         b3804b vmlinux
8420814 2036408 1302528 11759750         b37086 vmlinux.patched

Link: http://lkml.kernel.org/r/1392349908-29685-1-git-send-email-vnagarnaik@google.com
Cc: Laurent Chavey <chavey@google.com>
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: Vaibhav Nagarnaik <vnagarnaik@google.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
10 years agoMIPS: Make local_irq_disable macro safe for non-Mipsr2
Jim Quinlan [Wed, 27 Nov 2013 20:34:50 +0000 (15:34 -0500)]
MIPS: Make local_irq_disable macro safe for non-Mipsr2

For non-mipsr2 processors, the local_irq_disable contains an mfc0-mtc0
pair with instructions inbetween.  With preemption enabled, this sequence
may get preempted and effect a stale value of CP0_STATUS when executing
the mtc0 instruction.  This commit avoids this scenario by incrementing
the preempt count before the mfc0 and decrementing it after the mtc9.

[ralf@linux-mips.org: This patch is sorting out the part that were missed
by e97c5b6098 [MIPS: Make irqflags.h functions preempt-safe for non-mipsr2
cpus.]  I also re-enabled the inclusion of <asm/asm-offsets.h> at the top
of <asm/asmmacro.h>].

Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
Cc: linux-mips@linux-mips.org
Cc: cernekee@gmail.com
Patchwork: https://patchwork.linux-mips.org/patch/6164/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
10 years agoMerge branch 'exynos-drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git...
Dave Airlie [Thu, 20 Mar 2014 03:20:00 +0000 (13:20 +1000)]
Merge branch 'exynos-drm-fixes' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-fixes

 Just fixed resource release issue at open fail.

* 'exynos-drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm/exynos: Fix (more) freeing issues in exynos_drm_drv.c

10 years agodrm/exynos: Fix (more) freeing issues in exynos_drm_drv.c
Daniel Kurtz [Mon, 17 Mar 2014 03:28:06 +0000 (11:28 +0800)]
drm/exynos: Fix (more) freeing issues in exynos_drm_drv.c

The following commit [0] fixed a use-after-free, but left the subdrv open
in the error path.

[0] commit 6ca605f7c70895a35737435f17ae9cc5e36f1466
drm/exynos: Fix freeing issues in exynos_drm_drv.c

Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agomm: fix bad rss-counter if remap_file_pages raced migration
Hugh Dickins [Wed, 19 Mar 2014 00:38:38 +0000 (17:38 -0700)]
mm: fix bad rss-counter if remap_file_pages raced migration

Fix some "Bad rss-counter state" reports on exit, arising from the
interaction between page migration and remap_file_pages(): zap_pte()
must count a migration entry when zapping it.

And yes, it is possible (though very unusual) to find an anon page or
swap entry in a VM_SHARED nonlinear mapping: coming from that horrid
get_user_pages(write, force) case which COWs even in a shared mapping.

Signed-off-by: Hugh Dickins <hughd@google.com>
Tested-by: Sasha Levin sasha.levin@oracle.com>
Tested-by: Dave Jones davej@redhat.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMerge tag 'pci-v3.14-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Wed, 19 Mar 2014 23:15:54 +0000 (16:15 -0700)]
Merge tag 'pci-v3.14-fixes-3' of git://git./linux/kernel/git/helgaas/pci

Pull PCI resource management fix from Bjorn Helgaas:
 "This is a fix for an AGP regression exposed by e501b3d87f00 ("agp:
  Support 64-bit APBASE"), which we merged in v3.14-rc1.

  We've warned about the conflict between the GART and PCI resources and
  cleared out the PCI resource for a long time, but after e501b3d87f00,
  we still *use* that cleared-out PCI resource.  I think the GART
  resource is incorrect, so this patch removes it"

* tag 'pci-v3.14-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  Revert "[PATCH] Insert GART region into resource map"

10 years agoMerge branch 'for-3.14-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj...
Linus Torvalds [Wed, 19 Mar 2014 23:13:40 +0000 (16:13 -0700)]
Merge branch 'for-3.14-fixes' of git://git./linux/kernel/git/tj/cgroup

Pull cgroup fix from Tejun Heo:
 "One really late cgroup patch to fix error path in create_css().
  Hitting this bug would be pretty rare but still possible and it gets
  delayed we'd need to backport it through -stable anyway.  It only
  updates error path in create_css() and has low chance of new
  breakages"

* 'for-3.14-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  cgroup: fix a failure path in create_css()

10 years agoMIPS: Octeon: Fix warning in of_device_alloc on cn3xxx
Andreas Herrmann [Wed, 19 Mar 2014 22:03:30 +0000 (23:03 +0100)]
MIPS: Octeon: Fix warning in of_device_alloc on cn3xxx

Starting with commit 3da5278727a895d49a601f67fd49dffa0b80f9a5 (of/irq:
Rework of_irq_count()) the following warning is triggered on octeon
cn3xxx:

[    0.887281] WARNING: CPU: 0 PID: 1 at drivers/of/platform.c:171 of_device_alloc+0x228/0x230()
[    0.895642] Modules linked in:
[    0.898689] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.14.0-rc7-00012-g9ae51f2-dirty #41
[    0.906860] Stack : c8b439581166d96e ffffffff816b0000 0000000040808000 ffffffff81185ddc
[    0.906860]    0000000000000000 0000000000000000 0000000000000000 000000000000000b
[    0.906860]    000000000000000a 000000000000000a 0000000000000000 0000000000000000
[    0.906860]    ffffffff81740000 ffffffff81720000 ffffffff81615900 ffffffff816b0177
[    0.906860]    ffffffff81727d10 800000041f868fb0 0000000000000001 0000000000000000
[    0.906860]    0000000000000000 0000000000000038 0000000000000001 ffffffff81568484
[    0.906860]    800000041f86faa8 ffffffff81145ddc 0000000000000000 ffffffff811873f4
[    0.906860]    800000041f868b88 800000041f86f9c0 0000000000000000 ffffffff81569c9c
[    0.906860]    0000000000000000 0000000000000000 0000000000000000 0000000000000000
[    0.906860]    0000000000000000 ffffffff811205e0 0000000000000000 0000000000000000
[    0.906860]    ...
[    0.971695] Call Trace:
[    0.974139] [<ffffffff811205e0>] show_stack+0x68/0x80
[    0.979183] [<ffffffff81569c9c>] dump_stack+0x8c/0xe0
[    0.984196] [<ffffffff81145efc>] warn_slowpath_common+0x84/0xb8
[    0.990110] [<ffffffff81436888>] of_device_alloc+0x228/0x230
[    0.995726] [<ffffffff814368d8>] of_platform_device_create_pdata+0x48/0xd0
[    1.002593] [<ffffffff81436a94>] of_platform_bus_create+0x134/0x1e8
[    1.008837] [<ffffffff81436af8>] of_platform_bus_create+0x198/0x1e8
[    1.015064] [<ffffffff81436cc4>] of_platform_bus_probe+0xa4/0x100
[    1.021149] [<ffffffff81100570>] do_one_initcall+0xd8/0x128
[    1.026701] [<ffffffff816e2a10>] kernel_init_freeable+0x144/0x210
[    1.032753] [<ffffffff81564bc4>] kernel_init+0x14/0x110
[    1.037973] [<ffffffff8111bb44>] ret_from_kernel_thread+0x14/0x1c

With this commit the kernel starts mapping the interrupts listed for
gpio-controller node. irq_domain_ops for CIU (octeon_irq_ciu_map and
octeon_irq_ciu_xlat) refuse to handle the GPIO lines (returning -EINVAL)
and this is causing above warning in of_device_alloc().

Modify irq_domain_ops for CIU and CIU2 to "gracefully handle" GPIO
lines (neither return error code nor call octeon_irq_set_ciu_mapping
for it). This should avoid the warning.

(As before the real setup for GPIO lines will happen using
irq_domain_ops of gpio-controller.)

This patch is based on Wei's patch v2 (see
http://marc.info/?l=linux-mips&m=139511814813247).

Signed-off-by: Andreas Herrmann <andreas.herrmann@caviumnetworks.com>
Reported-by: Yang Wei <wei.yang@windriver.com>
Acked-by: David Daney <david.daney@cavium.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/6624/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
10 years agoMIPS: ftrace: Tweak safe_load()/safe_store() macros
Viller Hsiao [Tue, 18 Mar 2014 07:39:34 +0000 (15:39 +0800)]
MIPS: ftrace: Tweak safe_load()/safe_store() macros

Due to name collision in ftrace safe_load and safe_store macros,
these macros cannot take expressions as operands.

For example, compiler will complain for a macro call like the following:
  safe_store_code(new_code2, ip + 4, faulted);

  arch/mips/include/asm/ftrace.h:61:6: note: in definition of macro 'safe_store'
     : [dst] "r" (dst), [src] "r" (src)\
        ^
  arch/mips/kernel/ftrace.c:118:2: note: in expansion of macro 'safe_store_code'
    safe_store_code(new_code2, ip + 4, faulted);
    ^
  arch/mips/kernel/ftrace.c:118:32: error: undefined named operand 'ip + 4'
    safe_store_code(new_code2, ip + 4, faulted);
                                  ^
  arch/mips/include/asm/ftrace.h:61:6: note: in definition of macro 'safe_store'
     : [dst] "r" (dst), [src] "r" (src)\
        ^
  arch/mips/kernel/ftrace.c:118:2: note: in expansion of macro 'safe_store_code'
    safe_store_code(new_code2, ip + 4, faulted);
    ^

This build error is triggered by a4671094 [MIPS: ftrace: Fix icache flush
range error].  Tweak variable naming in those macros to allow flexible
operands.

Signed-off-by: Viller Hsiao <villerhsiao@gmail.com>
Cc: linux-mips@linux-mips.org
Cc: rostedt@goodmis.org
Cc: fweisbec@gmail.com
Cc: mingo@redhat.com
Cc: Qais.Yousef@imgtec.com
Patchwork: https://patchwork.linux-mips.org/patch/6622/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
10 years agoMerge tag 'drm-intel-fixes-2014-03-19' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Wed, 19 Mar 2014 21:17:31 +0000 (07:17 +1000)]
Merge tag 'drm-intel-fixes-2014-03-19' of git://anongit.freedesktop.org/drm-intel into drm-fixes

two more fixes, both regressions.

* tag 'drm-intel-fixes-2014-03-19' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: Disable stolen memory when DMAR is active
  Revert "drm/i915: don't touch the VDD when disabling the panel"

10 years agovia-rhine: Disable device in error path
Roger Luethi [Tue, 18 Mar 2014 17:14:01 +0000 (18:14 +0100)]
via-rhine: Disable device in error path

Currently, via-rhine fails to call pci_disable_device() for errors
in rhine_init_one().

Reported-by: Huqiu Liu <liuhq11@mails.tsinghua.edu.cn>
Signed-off-by: Roger Luethi <rl@hellgate.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 19 Mar 2014 16:05:38 +0000 (09:05 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Thomas Gleixner:
 "A single bugfix: make the scheduler clock on Vybrid SoCs count
  forward"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  clocksource: vf_pit_timer: use complement for sched_clock reading

10 years agoALSA: compress: Pass through return value of open ops callback
Charles Keepax [Wed, 19 Mar 2014 12:59:39 +0000 (12:59 +0000)]
ALSA: compress: Pass through return value of open ops callback

The snd_compr_open function would always return 0 even if the compressed
ops open function failed, obviously this is incorrect. Looks like this
was introduced by a small typo in:

commit a0830dbd4e42b38aefdf3fb61ba5019a1a99ea85
ALSA: Add a reference counter to card instance

This patch returns the value from the compressed op as it should.

Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com>
Acked-by: Vinod Koul <vinod.koul@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
10 years agoMIPS: BCM47XX: Check all (32) GPIOs when looking for a pin
Rafał Miłecki [Thu, 13 Feb 2014 16:48:12 +0000 (17:48 +0100)]
MIPS: BCM47XX: Check all (32) GPIOs when looking for a pin

Broadcom boards support 32 GPIOs and NVRAM may have entires for higher
ones too. Example:
gpio23=wombo_reset

Signed-off-by: Rafa? Mi?ecki <zajec5@gmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Cc: linux-mips@linux-mips.org
Cc: Rafał Miłecki <zajec5@gmail.com>
Patchwork: https://patchwork.linux-mips.org/patch/6547/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
10 years agodrm/i915: Disable stolen memory when DMAR is active
Chris Wilson [Tue, 18 Mar 2014 12:50:50 +0000 (14:50 +0200)]
drm/i915: Disable stolen memory when DMAR is active

We have reports of heavy screen corruption if we try to use the stolen
memory reserved by the BIOS whilst the DMA-Remapper is active. This
quirk may be only specific to a few machines or BIOSes, but first lets
apply the big hammer and always disable use of stolen memory when DMAR
is active.

v2 by Jani: Rebase on -fixes, only look at intel_iommu_gfx_mapped.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68535
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable@vger.kernel.org
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
10 years agoRevert "drm/i915: don't touch the VDD when disabling the panel"
Jani Nikula [Mon, 17 Mar 2014 13:20:59 +0000 (15:20 +0200)]
Revert "drm/i915: don't touch the VDD when disabling the panel"

This reverts
commit dff392dbd258381a6c3164f38420593f2d291e3b
Author: Paulo Zanoni <paulo.r.zanoni@intel.com>
Date:   Fri Dec 6 17:32:41 2013 -0200

    drm/i915: don't touch the VDD when disabling the panel

which didn't take into account

commit 6cb49835da0426f69a2931bc2a0a8156344b0e41
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Sun May 20 17:14:50 2012 +0200

    drm/i915: enable vdd when switching off the eDP panel

and

commit 35a38556d900b9cb5dfa2529c93944b847f8a8a4
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Sun Aug 12 22:17:14 2012 +0200

    drm/i915: reorder edp disabling to fix ivb MacBook Air

Unsurprisingly, various MacBooks failed.

Effectively the same has already been done in drm-intel-next-queued.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=74628
Tested-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
10 years agocgroup: fix a failure path in create_css()
Li Zefan [Tue, 18 Mar 2014 09:02:36 +0000 (17:02 +0800)]
cgroup: fix a failure path in create_css()

If online_css() fails, we should remove cgroup files belonging
to css->ss.

Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
10 years agoRevert "[PATCH] Insert GART region into resource map"
Bjorn Helgaas [Tue, 18 Mar 2014 20:26:12 +0000 (14:26 -0600)]
Revert "[PATCH] Insert GART region into resource map"

This reverts commit 56dd669a138c, which makes the GART visible in
/proc/iomem.  This fixes a regression: e501b3d87f00 ("agp: Support 64-bit
APBASE") exposed an existing problem with a conflict between the GART
region and a PCI BAR region.

The GART addresses are bus addresses, not CPU addresses, and therefore
should not be inserted in iomem_resource.

On many machines, the GART region is addressable by the CPU as well as by
an AGP master, but CPU addressability is not required by the spec.  On some
of these machines, the GART is mapped by a PCI BAR, and in that case, the
PCI core automatically inserts it into iomem_resource, just as it does for
all BARs.

Inserting it here means we'll have a conflict if the PCI core later tries
to claim the GART region, so let's drop the insertion here.

The conflict indirectly causes X failures, as reported by Jouni in the
bugzilla below.  We detected the conflict even before e501b3d87f00, but
after it the AGP code (fix_northbridge()) uses the PCI resource (which is
zeroed because of the conflict) instead of reading the BAR again.

Conflicts:
arch/x86_64/kernel/aperture.c

Fixes: e501b3d87f00 agp: Support 64-bit APBASE
Link: https://bugzilla.kernel.org/show_bug.cgi?id=72201
Reported-and-tested-by: Jouni Mettälä <jtmettala@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
10 years agoATHEROS-ATL1E: Convert iounmap to pci_iounmap
Peter Senna Tschudin [Tue, 18 Mar 2014 16:11:24 +0000 (17:11 +0100)]
ATHEROS-ATL1E: Convert iounmap to pci_iounmap

Use pci_iounmap instead of iounmap when the virtual mapping was done
with pci_iomap. A simplified version of the semantic patch that finds this
issue is as follows: (http://coccinelle.lip6.fr/)

// <smpl>
@r@
expression addr;
@@
addr = pci_iomap(...)

@rr@
expression r.addr;
@@
* iounmap(addr)
// </smpl>

Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agovxlan: fix potential NULL dereference in arp_reduce()
David Stevens [Tue, 18 Mar 2014 16:32:29 +0000 (12:32 -0400)]
vxlan: fix potential NULL dereference in arp_reduce()

This patch fixes a NULL pointer dereference in the event of an
skb allocation failure in arp_reduce().

Signed-Off-By: David L Stevens <dlstevens@us.ibm.com>
Acked-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'cnic-net'
David S. Miller [Tue, 18 Mar 2014 20:03:44 +0000 (16:03 -0400)]
Merge branch 'cnic-net'

Michael Chan says:

====================
cnic bug fixes for net-next

Michael Chan (3):
  cnic: Use proper ulp_ops for per device operations.
  cnic,bnx2i,bnx2fc: Fix inconsistent use of page size
  cnic: Update version to 2.5.20 and copyright year.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocnic: Update version to 2.5.20 and copyright year.
Michael Chan [Tue, 18 Mar 2014 03:19:08 +0000 (19:19 -0800)]
cnic: Update version to 2.5.20 and copyright year.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocnic,bnx2i,bnx2fc: Fix inconsistent use of page size
Michael Chan [Tue, 18 Mar 2014 03:19:07 +0000 (19:19 -0800)]
cnic,bnx2i,bnx2fc: Fix inconsistent use of page size

The bnx2/bnx2x rings are made up of linked pages.  However there is an
upper limit on the page size as some the page size settings are 16-bit
in the hardware/firmware interface.  In the current code, some parts
use BNX2_PAGE_SIZE which has a 16K upper limit and some parts use
PAGE_SIZE.  On archs with >= 64K PAGE_SIZE, it generates some compile
warnings.  Define a new CNIC_PAGE_SZIE which has an upper limit of
16K and use it consistently in all relevant parts.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocnic: Use proper ulp_ops for per device operations.
Michael Chan [Tue, 18 Mar 2014 03:19:06 +0000 (19:19 -0800)]
cnic: Use proper ulp_ops for per device operations.

For per device operations, cnic needs to dereference the RCU protected
cp->ulp_ops instead of the global cnic_ulp_tbl.  In 2 locations,
cnic_send_nlmsg() and cnic_copy_ulp_stats(), it was referencing the
global table.  If the device has been unregistered and these functions
are still being called (very unlikely scenarios), it could lead to NULL
pointer dereference.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: cdc_ncm: fix control message ordering
Bjørn Mork [Mon, 17 Mar 2014 15:25:18 +0000 (16:25 +0100)]
net: cdc_ncm: fix control message ordering

This is a context modified revert of commit 6a9612e2cb22
("net: cdc_ncm: remove ncm_parm field") which introduced
a NCM specification violation, causing setup errors for
some devices. These errors resulted in the device and
host disagreeing about shared settings, with complete
failure to communicate as the end result.

The NCM specification require that many of the NCM specific
control reuests are sent only while the NCM Data Interface
is in alternate setting 0. Reverting the commit ensures that
we follow this requirement.

Fixes: 6a9612e2cb22 ("net: cdc_ncm: remove ncm_parm field")
Reported-and-tested-by: Pasi Kärkkäinen <pasik@iki.fi>
Reported-by: Thomas Schäfer <tschaefer@t-online.de>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: ip6_append_data_mtu do not handle the mtu of the second fragment properly
lucien [Mon, 17 Mar 2014 04:51:01 +0000 (12:51 +0800)]
ipv6: ip6_append_data_mtu do not handle the mtu of the second fragment properly

In ip6_append_data_mtu(), when the xfrm mode is not tunnel(such as
transport),the ipsec header need to be added in the first fragment, so the mtu
will decrease to reserve space for it, then the second fragment come, the mtu
should be turn back, as the commit 0c1833797a5a6ec23ea9261d979aa18078720b74
said.  however, in the commit a493e60ac4bbe2e977e7129d6d8cbb0dd236be, it use
*mtu = min(*mtu, ...) to change the mtu, which lead to the new mtu is alway
equal with the first fragment's. and cannot turn back.

when I test through  ping6 -c1 -s5000 $ip (mtu=1280):
...frag (0|1232) ESP(spi=0x00002000,seq=0xb), length 1232
...frag (1232|1216)
...frag (2448|1216)
...frag (3664|1216)
...frag (4880|164)

which should be:
...frag (0|1232) ESP(spi=0x00001000,seq=0x1), length 1232
...frag (1232|1232)
...frag (2464|1232)
...frag (3696|1232)
...frag (4928|116)

so delete the min() when change back the mtu.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Fixes: 75a493e60ac4bb ("ipv6: ip6_append_data_mtu did not care about pmtudisc and frag_size")
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoisdn/capi: Make Middleware depend on CAPI2.0
Paul Bolle [Sun, 16 Mar 2014 17:44:21 +0000 (18:44 +0100)]
isdn/capi: Make Middleware depend on CAPI2.0

The Kconfig symbol ISDN_CAPI_MIDDLEWARE is only used in capi.c. Setting
it without setting ISDN_CAPI_CAPI20 is therefor useless. Make it depend
on ISDN_CAPI_CAPI20 and put its entry after ISDN_CAPI_CAPI20's entry.

Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Tue, 18 Mar 2014 18:32:08 +0000 (11:32 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull another kvm fix from Paolo Bonzini:
 "A fix for a PowerPC bug that was introduced during the 3.14 merge
  window"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: PPC: Book3S HV: Fix register usage when loading/saving VRSAVE
  KVM: PPC: Book3S HV: Remove bogus duplicate code

10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec
David S. Miller [Tue, 18 Mar 2014 16:42:33 +0000 (12:42 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/klassert/ipsec

Steffen Klassert says:

====================
1) Fix a sleep in atomic when pfkey_sadb2xfrm_user_sec_ctx()
   is called from pfkey_compile_policy().
   Fix from Nikolay Aleksandrov.

2) security_xfrm_policy_alloc() can be called in process and atomic
   context. Add an argument to let the callers choose the appropriate
   way. Fix from Nikolay Aleksandrov.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoALSA: oxygen: Xonar DG(X): fix Stereo Upmixing regression
Clemens Ladisch [Tue, 18 Mar 2014 08:31:18 +0000 (09:31 +0100)]
ALSA: oxygen: Xonar DG(X): fix Stereo Upmixing regression

The code introduced in commit 1f91ecc14dee ("ALSA: oxygen: modify
adjust_dg_dac_routing function") accidentally disregarded the old value
of the playback routing register, so it broke the "Stereo Upmixing"
mixer control.

The unmuted parts of the channel routing are the same for all settings
of the output destination, so it suffices to revert that part of the
patch.

Fixes: 1f91ecc14dee ('ALSA: oxygen: modify adjust_dg_dac_routing function')
Tested-by: Roman Volkov <v1ron@mail.ru>
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
10 years agoMerge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Tue, 18 Mar 2014 08:21:42 +0000 (09:21 +0100)]
Merge tag 'perf-urgent-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull two 'perf bench' fixes from Arnaldo:

  * Make 'perf bench mem' (i.e. no args) mean 'run all tests' so that we can run
    all tests, not stopping at the numa ones.  (Arnaldo Carvalho de Melo)

  * Fix NULL pointer dereference after last test in in "perf bench all" (Patrick Palka)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
10 years agoMIPS: Fix possible build error with transparent hugepages enabled
Alex Smith [Tue, 21 Jan 2014 11:22:35 +0000 (11:22 +0000)]
MIPS: Fix possible build error with transparent hugepages enabled

If CONFIG_TRANSPARENT_HUGEPAGE is enabled, but CONFIG_HUGETLB_PAGE is not,
it is possible to end up with a configuration that fails to build with the
following error:

include/linux/huge_mm.h:125:2: error: #error "hugepages can't be allocated by the buddy allocator"

This is due to CONFIG_FORCE_MAX_ZONEORDER defaulting to 11. It already has
ranges that change the valid values when HUGETLB_PAGE is enabled, but this
is not done for TRANSPARENT_HUGEPAGE. Fix by changing the HUGETLB_PAGE
dependencies to MIPS_HUGE_TLB_SUPPORT, which includes both
TRANSPARENT_HUGEPAGE and HUGETLB_PAGE.

Signed-off-by: Alex Smith <alex.smith@imgtec.com>
Reviewed-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/6391/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
10 years agodrm: Fix use-after-free in the shadow-attache exit code
Daniel Vetter [Thu, 30 Jan 2014 16:58:38 +0000 (17:58 +0100)]
drm: Fix use-after-free in the shadow-attache exit code

This regression has been introduced in

commit b3f2333de8e81b089262b26d52272911523e605f
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Wed Dec 11 11:34:31 2013 +0100

    drm: restrict the device list for shadow attached drivers

Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge tag 'drm-intel-fixes-2014-03-17' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Tue, 18 Mar 2014 03:06:17 +0000 (13:06 +1000)]
Merge tag 'drm-intel-fixes-2014-03-17' of git://anongit.freedesktop.org/drm-intel into drm-fixes

Two 3.14 specific fixes, two cc: stable.

* tag 'drm-intel-fixes-2014-03-17' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: Don't enable display error interrupts from the start
  drm/i915: Fix scanline counter fixup on BDW
  drm/i915: Add a workaround for HSW scanline counter weirdness
  drm/i915: Fix PSR programming

10 years agonet: cpsw: do not register cpts twice
Benedikt Spranger [Sun, 16 Mar 2014 15:36:29 +0000 (16:36 +0100)]
net: cpsw: do not register cpts twice

commit f280e89a (drivers: net: cpsw: fix for cpsw crash when build as modules)
moved cpts_register()/cpts_unregister() to  ndo_open()/ndo_stop(), but failed
to remove cpts_register in cpsw_probe() which leads to a double registration
and the following debug object splat.

[   18.991902] ODEBUG: init active (active state 0) object type: timer_list hint: delayed_work_timer_fn+0x0/0x2c
[   19.082249] [<c0059e80>] (init_timer_key) from [<c04965d4>] (cpts_register+0x1f0/0x2c4)
[   19.090642] [<c04965d4>] (cpts_register) from [<c04931dc>] (cpsw_ndo_open+0x780/0x81c)
[   19.098948] [<c04931dc>] (cpsw_ndo_open) from [<c0599c2c>] (__dev_open+0xb4/0x118)

Signed-off-by: Benedikt Spranger <b.spranger@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Mon, 17 Mar 2014 23:54:49 +0000 (16:54 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull final final block IO fixes from Jens Axboe:
 "Yes, the last round was final. This one is final final.

  The mtip32xx fix could have waited, but it's so simple and gets rid of
  two warning spewages on load.  The two block flush fixes are critical
  for blk-mq, and are the primary reason for this late pull request"

* 'for-linus' of git://git.kernel.dk/linux-block:
  mtip32xx: fix bad use of smp_processor_id()
  block: change flush sequence list addition back to front add
  block: fix q->flush_rq NULL pointer crash on dm-mpath flush

10 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Mon, 17 Mar 2014 23:53:18 +0000 (16:53 -0700)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS updates from Ralf Baechle:
 "A fair number of fixes all across arch/mips.  Nothing really stands
  out though APRP, the FPU code and syscall tracing code received
  multiple patches those all were small"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: mark O32+FP64 experimental for now
  MIPS: ftrace: Fix icache flush range error
  MIPS: Fix syscall tracing interface
  MIPS: asm: syscall: Fix copying system call arguments
  MIPS: Octeon: Fix fall through on bar type OCTEON_DMA_BAR_TYPE_SMALL
  MIPS: FPU: Fix conflict of register usage
  MIPS: Replace CONFIG_MIPS64 and CONFIG_MIPS32_R2
  MIPS: math-emu: Fix prefx detection and COP1X function field definition
  MIPS: APRP: Choose the correct VPE loader by fixing the linking
  MIPS: APRP: Unregister rtlx interrupt hook at module exit
  MIPS: APRP: Fix the linking of rtlx interrupt hook
  MIPS: bcm47xx: Include missing errno.h for ENXIO
  MIPS: Alchemy: Fix unchecked kstrtoul return value
  MIPS: Fix randconfig build error.

10 years agoATHEROS-ALX: Use dma_set_mask_and_coherent and fix a bug
Peter Senna Tschudin [Sat, 15 Mar 2014 23:30:52 +0000 (00:30 +0100)]
ATHEROS-ALX: Use dma_set_mask_and_coherent and fix a bug

1. For the 64 bits dma mask use dma_set_mask_and_coherent instead of
   dma_set_mask and dma_set_coherent_mask.

2. For the 32 bits dma mask dma_set_coherent_mask is only called if
   dma_set_mask fails, which is unusual. Assuming this as a bug, fixes
   it by replacing calls to dma_set_mask and dma_set_coherent_mask by a
   call to dma_set_mask_and_coherent.

Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Tested-by: Jonas Hahnfeld <hahnjo@hahnjo.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agosparc64:tsb.c:use array size macro rather than number
Doug Wilson [Fri, 7 Mar 2014 10:59:03 +0000 (16:29 +0530)]
sparc64:tsb.c:use array size macro rather than number

    This is a small patch which uses ARRAY_SIZE macro
    rather than a number to make code readability better.

Signed-off-by: Doug Wilson <doug.lkml@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agosparc64: don't treat 64-bit syscall return codes as 32-bit
Dave Kleikamp [Fri, 14 Mar 2014 15:42:01 +0000 (10:42 -0500)]
sparc64: don't treat 64-bit syscall return codes as 32-bit

When checking a system call return code for an error,
linux_sparc_syscall was sign-extending the lower 32-bit value and
comparing it to -ERESTART_RESTARTBLOCK. lseek can return valid return
codes whose lower 32-bits alone would indicate a failure (such as 4G-1).
Use the whole 64-bit value to check for errors. Only the 32-bit path
should sign extend the lower 32-bit value.

Signed-off-by: Dave Kleikamp <dave.kleikamp@oracle.com>
Acked-by: Bob Picco <bob.picco@oracle.com>
Acked-by: Allen Pais <allen.pais@oracle.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: sparclinux@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMIPS: mark O32+FP64 experimental for now
Paul Burton [Fri, 14 Feb 2014 17:55:18 +0000 (17:55 +0000)]
MIPS: mark O32+FP64 experimental for now

Commit 597ce1723e0f "MIPS: Support for 64-bit FP with O32 binaries"
introduced support for setting Status.FR=1 for O32 binaries with the
EF_MIPS_FP64 ELF header flag set. Whilst this flag is currently
supported by binutils it does introduce an ABI break within userland.
Objects built with EF_MIPS_FP64 cannot be safely linked with those built
without it since code in either object may assume behaviour specific to
a value of FR.

More recently there has been discussion around avoiding further
fragmentation of the O32 ABI whilst still allowing the use of FR=1 and
features such as MSA which depend upon it. Details of the plan to allow
this are still being worked on, and whilst the kernel will need the
ability to handle FR=1 with O32 tasks it is unclear what else it may
need to provide to a userland which seeks to avoid another ABI break. In
order to prevent the proliferation of userland which may rely upon the
current EF_MIPS_FP64 behaviour this patch marks the kernel support for
it experimental & disables it by default. Under current proposals it is
likely that this support can simply be enabled again later, but possibly
after the introduction of further interfaces with userland and support
for the MIPS R5 UFR feature.

Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Matthew Fortune <matthew.fortune@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: Paul Burton <paul.burton@imgtec.com>
Patchwork: https://patchwork.linux-mips.org/patch/6549/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Mon, 17 Mar 2014 15:00:50 +0000 (08:00 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/hid

Pull HID fixes from Jiri Kosina:

 - fix for ordering of device removal actions in hidraw, by Fernando
   Luis Vázquez Cao

 - fix for uninitialized workqueue usage in hid-sony, by Frank Praznik

 - device ID addition for new variant of Logitech G27, from Simon Wood

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
  HID: hid-lg4ff: Support new version of G27
  HID: hidraw: fix warning destroying hidraw device files after parent
  HID: sony: Fix work queue issues.