GitHub/exynos8895/android_kernel_samsung_universal8895.git
12 years agoMerge branches 'core', 'cxgb4', 'ipath', 'iser', 'lockdep', 'mlx4', 'nes', 'ocrdma...
Roland Dreier [Mon, 21 May 2012 16:00:47 +0000 (09:00 -0700)]
Merge branches 'core', 'cxgb4', 'ipath', 'iser', 'lockdep', 'mlx4', 'nes', 'ocrdma', 'qib' and 'raw-qp' into for-linus

12 years agoRDMA/cxgb4: Include vmalloc.h for vmalloc and vfree
Vipul Pandya [Mon, 21 May 2012 12:01:13 +0000 (17:31 +0530)]
RDMA/cxgb4: Include vmalloc.h for vmalloc and vfree

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/mlx4: Fix mlx4_ib_add() error flow
Jack Morgenstein [Thu, 10 May 2012 20:28:09 +0000 (23:28 +0300)]
IB/mlx4: Fix mlx4_ib_add() error flow

We need to use a different loop index for mlx4_counter_alloc() and for
device_create_file() iterations: the mlx4_counter_alloc() loop index
is used in the error flow to free counters.

If the same loop index is used for device_create_file() and, say, the
device_create_file() loop fails on the first iteration, the allocated
counters will not be freed.

Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/core: Fix IB_SA_COMP_MASK macro
Jack Morgenstein [Thu, 10 May 2012 20:28:07 +0000 (23:28 +0300)]
IB/core: Fix IB_SA_COMP_MASK macro

It needs parentheses around the argument, so that it can be used with
complex arguments (e.g., "n+5").

Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/iser: Fix error flow in iser ep connection establishment
Or Gerlitz [Sun, 29 Apr 2012 14:04:21 +0000 (17:04 +0300)]
IB/iser: Fix error flow in iser ep connection establishment

The current error flow code was releasing the IB connection object and
calling iscsi_destroy_endpoint() directly without going through the
reference counting mechanism introduced in commit 39ff05d ("IB/iser:
Enhance disconnection logic for multi-pathing"). This resulted in a
double free of the iscsi endpoint object, which causes a kernel NULL
pointer dereference.  Fix that by plugging into the IB conn reference
counting correctly.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/mlx4: Increase the number of vectors (EQs) available for ULPs
Shlomo Pongratz [Sun, 29 Apr 2012 14:04:27 +0000 (17:04 +0300)]
IB/mlx4: Increase the number of vectors (EQs) available for ULPs

Enable IB ULPs to use a larger portion of the device EQs (which map to
IRQs).  The mlx4_ib driver follows the mlx4_core framework of the EQs
to be divided among the device ports.  In this scheme, for each IB
port, the number of allocated EQs follows the number of cores, subject
to other system constraints, such as number available MSI-X vectors.

Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Add query_qp support
Vipul Pandya [Fri, 18 May 2012 09:59:33 +0000 (15:29 +0530)]
RDMA/cxgb4: Add query_qp support

This allows querying the QP state before flushing.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Remove kfifo usage
Vipul Pandya [Fri, 18 May 2012 09:59:32 +0000 (15:29 +0530)]
RDMA/cxgb4: Remove kfifo usage

Using kfifos for ID management was limiting the number of QPs and
preventing NP384 MPI jobs.  So replace it with a simple bitmap
allocator.

Remove IDs from the IDR tables before deallocating them.  This bug was
causing the BUG_ON() in insert_handle() to fire because the ID was
getting reused before being removed from the IDR table.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Use vmalloc() for debugfs QP dump
Vipul Pandya [Fri, 18 May 2012 09:59:31 +0000 (15:29 +0530)]
RDMA/cxgb4: Use vmalloc() for debugfs QP dump

This allows dumping thousands of QPs.  Log active open failures of
interest.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: DB Drop Recovery for RDMA and LLD queues
Vipul Pandya [Fri, 18 May 2012 09:59:30 +0000 (15:29 +0530)]
RDMA/cxgb4: DB Drop Recovery for RDMA and LLD queues

Add module option db_fc_threshold which is the count of active QPs
that trigger automatic db flow control mode.  Automatically transition
to/from flow control mode when the active qp count crosses
db_fc_theshold.

Add more db debugfs stats

On DB DROP event from the LLD, recover all the iwarp queues.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Disable interrupts in c4iw_ev_dispatch()
Vipul Pandya [Fri, 18 May 2012 09:59:29 +0000 (15:29 +0530)]
RDMA/cxgb4: Disable interrupts in c4iw_ev_dispatch()

Use GFP_ATOMIC in _insert_handle() if ints are disabled.

Don't panic if we get an abort with no endpoint found.  Just log a
warning.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Add DB Overflow Avoidance
Vipul Pandya [Fri, 18 May 2012 09:59:28 +0000 (15:29 +0530)]
RDMA/cxgb4: Add DB Overflow Avoidance

Get FULL/EMPTY/DROP events from LLD.  On FULL event, disable normal
user mode DB rings.

Add modify_qp semantics to allow user processes to call into the
kernel to ring doobells without overflowing.

Add DB Full/Empty/Drop stats.

Mark queues when created indicating the doorbell state.

If we're in the middle of db overflow avoidance, then newly created
queues should start out in this mode.

Bump the C4IW_UVERBS_ABI_VERSION to 2 so the user mode library can
know if the driver supports the kernel mode db ringing.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Add debugfs RDMA memory stats
Vipul Pandya [Fri, 18 May 2012 09:59:27 +0000 (15:29 +0530)]
RDMA/cxgb4: Add debugfs RDMA memory stats

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agocxgb4: DB Drop Recovery for RDMA and LLD queues
Vipul Pandya [Fri, 18 May 2012 09:59:26 +0000 (15:29 +0530)]
cxgb4: DB Drop Recovery for RDMA and LLD queues

recover LLD EQs for DB drop interrupts.  This includes adding a new
db_lock, a spin lock disabling BH too, used by the recovery thread and
the ring_tx_db() paths to allow db drop recovery.

Clean up initial DB avoidance code.

Add read_eq_indices() - this allows the LLD to use the PCIe mw to
efficiently read hw eq contexts.

Add cxgb4_sync_txq_pidx() - called by iw_cxgb4 to sync up the sw/hw
pidx value.

Add flush_eq_cache() and cxgb4_flush_eq_cache().  This allows iw_cxgb4
to flush the sge eq context cache before beginning db drop recovery.

Add module parameter, dbfoifo_int_thresh, to allow tuning the db
interrupt threshold value.

Add dbfifo_int_thresh to cxgb4_lld_info so iw_cxgb4 knows the threshold.

Add module parameter, dbfoifo_drain_delay, to allow tuning the amount
of time delay between DB FULL and EMPTY upcalls to iw_cxgb4.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agocxgb4: Common platform specific changes for DB Drop Recovery
Vipul Pandya [Fri, 18 May 2012 09:59:25 +0000 (15:29 +0530)]
cxgb4: Common platform specific changes for DB Drop Recovery

Add platform-specific callback functions for interrupts.  This is
needed to do a single read-clear of the CAUSE register and then call
out to platform specific functions for DB threshold interrupts and DB
drop interrupts.

Add t4_mem_win_read_len() - mem-window reads for arbitrary lengths.
This is used to read the CIDX/PIDX values from EC contexts during DB
drop recovery.

Add t4_fwaddrspace_write() - sends addrspace write cmds to the fw.
Needed to flush the sge eq context cache.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agocxgb4: Detect DB FULL events and notify RDMA ULD
Vipul Pandya [Fri, 18 May 2012 09:59:24 +0000 (15:29 +0530)]
cxgb4: Detect DB FULL events and notify RDMA ULD

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Drop peer_abort when no endpoint found
Steve Wise [Mon, 30 Apr 2012 20:31:29 +0000 (15:31 -0500)]
RDMA/cxgb4: Drop peer_abort when no endpoint found

Log a warning and drop the abort message.  Otherwise we will do a
bogus wake_up() and crash.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Always wake up waiters in c4iw_peer_abort_intr()
Steve Wise [Fri, 27 Apr 2012 14:59:16 +0000 (09:59 -0500)]
RDMA/cxgb4: Always wake up waiters in c4iw_peer_abort_intr()

This fixes a race where an ingress abort fails to wake up the thread
blocked in rdma_init() causing the app to hang.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agomlx4_core: Change bitmap allocator to work in round-robin fashion
Jack Morgenstein [Thu, 10 May 2012 20:28:04 +0000 (23:28 +0300)]
mlx4_core: Change bitmap allocator to work in round-robin fashion

Under most circumstances, the bitmap allocator does not allocate the
same full 24-bit QP number immediately after a QP is destroyed.

This works by using the upper bits of a 24-bit QP number, beyond the
number of QPs that are actually available in the low level driver.
For example, say that the HCA is willing to allocate a maximum of 64K
qps.  We use the bits 23..16 as a "counter" which is incremented by 1
at each allocation so that even if the same physical QP is
re-allocated, it will not receive the same 24-bit QP number.

However, we have seen the following scenario:
1. Allocate, say, 255 QPs in succession.  This will cause a wrap of the "counter".
2. Destroy the first QP allocated, then allocate a new QP.  The new QP,
   because of the counter wraparound, will get the same FULL QP number as
   the QP just destroyed!

This is a problem because packets in transit can be erroneously
delivered to the new QP when they were meant for the old (destroyed)
QP, because the full QP number of the new QP is identical to the
destroyed QP.  (The "counter" mechanism is meant to prevent this by
having the full 24-bit QP numbers differ even if the physical QP on
the HCA is the same.  As we see above, however, this mechanism does
not always work).

The best fix for this problem is to allocate QPs in round-robin mode,
so that the physical QP numbers are not immediately re-used.

Found-by: Matthew Finlay <matt@mellanox.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/nes: Don't call event handler if pointer is NULL
Tatyana Nikolova [Fri, 11 May 2012 19:29:32 +0000 (14:29 -0500)]
RDMA/nes: Don't call event handler if pointer is NULL

Don't call the ibqp event_handler pointer in the case it wasn't initialized.

Signed-off-by: Tatyana Nikolova <Tatyana.E.Nikolova@intel.com>
Signed-off-by: Donald Wood <Donald.E.Wood@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/nes: Fix for the ORD value of the connecting peer
Tatyana Nikolova [Fri, 11 May 2012 15:46:16 +0000 (10:46 -0500)]
RDMA/nes: Fix for the ORD value of the connecting peer

Set ORD value of the connecting peer to be at least one in order to
accommodate an RDMA READ Request message.

Signed-off-by: Tatyana Nikolova <Tatyana.E.Nikolova@intel.com>
Signed-off-by: Donald Wood <Donald.E.Wood@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/qib: Add cache line awareness to qib_qp and qib_devdata structures
Mike Marciniszyn [Mon, 7 May 2012 18:02:31 +0000 (14:02 -0400)]
IB/qib: Add cache line awareness to qib_qp and qib_devdata structures

This patch reorganizes the QP and devdata files to be more cache line aware.

qib_qp fields in particular are split into read-mostly, send, and receive fields.

qib_devdata fields are split into read-mostly and read/write fields

Testing has show that bidirectional tests improve by as much as 100%
with this patch.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/qib: MADs with misset M_Keys should return failure
Jim Foraker [Wed, 2 May 2012 18:57:23 +0000 (14:57 -0400)]
IB/qib: MADs with misset M_Keys should return failure

If a MAD is sent directly to the local HCA rather than placed on a QP
and the MAD fails M_Key checks, there is no means to generate a
timeout for the client, which may hang.  Instead we report
IB_MAD_RESULT_FAILURE, which operates the same for on-the-wire
packets, but will generate a send failure back to the client.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Jim Foraker <foraker1@llnl.gov>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/qib: Fix M_Key lease timeout handling
Jim Foraker [Wed, 2 May 2012 18:39:11 +0000 (14:39 -0400)]
IB/qib: Fix M_Key lease timeout handling

If a port has an M_Key lease set, the M_Key protect bits set to 1, and
a SubnSet arrives with an invalid M_Key, an M_Key mismatch trap is
generated, the lease timer begins as expected, and eventually the
M_Key protect bits will be set back to 0 as per the spec.  However, if
any other SMP with an invalid M_Key arrives, the lease timer is expired
and the M_Key protect bits remain in force.

This is not according to to spec.  In particular, C14-17 says that
a lease timer that is underway is not affected by protection level
checks (ie, at protection level 1, a SubnGet with a bad M_Key may be
successful, but does not stop the timer), and C14-19 says that the timer
shall stop when a valid M_Key has been received.  C14-19 is the only
compliance statement that specifies a stopping condition for the timer.

This behavior is magnified if the port's Master SM LID attribute
points at itself.  In that case, the M_Key mismatch trap is sufficient to
expire the timer, and the mkey lease attribute is rendered useless.

Reviewed-by: Ram Vepa <ram.vepa@qlogic.com>
Signed-off-by: Jim Foraker <foraker1@llnl.gov>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/qib: Fix QLE734X link cycling
Mitko Haralanov [Mon, 7 May 2012 18:02:53 +0000 (14:02 -0400)]
IB/qib: Fix QLE734X link cycling

The SERDES was using the incorrect Frequency Loop Bandwidth setting
causing the link to cycle through the Physical link negotiation state
machine.  Fixing the Frequency Loop Bandwidth setting in the SERDES
helps the link come up faster and more reliably.

Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/qib: Display correct value for number of contexts
Mitko Haralanov [Mon, 7 May 2012 18:03:02 +0000 (14:03 -0400)]
IB/qib: Display correct value for number of contexts

A "fix" for a bug with the number of contexts on a single-port board
caused the calculation to be off by one, which causes problems with
the upper layers.  The same problem exists for number of free
contexts, which is also fixed here.

Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/qib: Correct ordering of reregister vs. port active events
Todd Rimmer [Wed, 2 May 2012 18:35:18 +0000 (14:35 -0400)]
IB/qib: Correct ordering of reregister vs. port active events

When a port first goes active with SMA Set(PortInfo) and reregister
bit set, the driver sends up the reregister event followed by a port
active event.

The problem is that in response to reregister event most apps try to
issue a SA query of some sort, but that fails because port is not
active.

The qib driver needs to a trivial change to correct this behavior.

This issue has been there for a while; however the recent serdes work
has probably made the delay between the reregister event and the
active event larger and hence opened the race far enough so that its
being seen more often.

The patch also changes the clientrereg local to a u8 and saves off the
rereg bit into it.  The code following the nested subn_get_portinfo()
now restores that bit per o14-12.2.1 with a logical OR from that copy.

Reviewed-by: Ram Vepa <ram.vepa@qlogic.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/qib: Optimize pio ack buffer allocation
Mike Marciniszyn [Mon, 7 May 2012 18:02:42 +0000 (14:02 -0400)]
IB/qib: Optimize pio ack buffer allocation

This patch optimizes pio buffer allocation in the kernel.

For qib, kernel pio buffers are used for sending acks.  The code to
allocate the buffer would always start at 0 until it found a buffer.

This means that an average of 64 comparisions were done on each
allocate, since the busy bit won't be cleared until the bits are
refreshed when buffers are exhausted.

This patch adds two new fields in the devdata struct, last_pio and
min_kernel_pio.  last_pio is the last buffer that was allocated.
min_kernel_pio is the lowest potential available buffer.

min_kernel_pio is modifed as contexts are allocated and deallocted.

Reviewed-by: Ramkrishna Vepa <ramkrishna.vepa@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/qib: Add prefetch for eager buffers
Mike Marciniszyn [Wed, 2 May 2012 15:21:53 +0000 (11:21 -0400)]
IB/qib: Add prefetch for eager buffers

Add a prefetch call when a packet has been stored.  The nature of the
prefetch is correctly determined by the alternatives mechanism.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoMAINTAINERS: Update qib and ipath entries from QLogic to Intel
Mike Marciniszyn [Thu, 10 May 2012 13:25:20 +0000 (09:25 -0400)]
MAINTAINERS: Update qib and ipath entries from QLogic to Intel

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/core: Fix mismatch between locked and pinned pages
Yishai Hadas [Thu, 10 May 2012 20:28:05 +0000 (23:28 +0300)]
IB/core: Fix mismatch between locked and pinned pages

Commit bc3e53f682d9 ("mm: distinguish between mlocked and pinned
pages") introduced a separate counter for pinned pages and used it in
the IB stack.  However, in ib_umem_get() the pinned counter is
incremented, but ib_umem_release() wrongly decrements the locked
counter.  Fix this.

Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/mlx4: Replace printk(KERN_yyy...) with pr_yyy(...)
Shlomo Pongratz [Sun, 29 Apr 2012 14:04:26 +0000 (17:04 +0300)]
IB/mlx4: Replace printk(KERN_yyy...) with pr_yyy(...)

Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
[ Replace one more printk_once() with pr_info_once().  - Roland ]

Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agomlx4_core: Add second capabilities flags field
Shlomo Pongratz [Sun, 29 Apr 2012 14:04:25 +0000 (17:04 +0300)]
mlx4_core: Add second capabilities flags field

This patch adds a 64-bit flags2 features member to struct mlx4_dev to
export further features of the hardware.  The original flags field
tracks features whose support bits are advertised by the firmware in
offsets 0x40 and 0x44 of the query device capabilities command.
flags2 will track features whose support bits are scattered at various
offsets.

RSS support is the first feature to be exported through flags2.  RSS
capabilities are located at offset 0x2e.  The size of the RSS
indirection table is also given in this offset.

Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/mlx4: Put priority bits in WQE of IBoE MLX QP
Oren Duer [Sun, 29 Apr 2012 14:04:24 +0000 (17:04 +0300)]
IB/mlx4: Put priority bits in WQE of IBoE MLX QP

Otherwise CM packets going over MLX QP1 get fixed scheduling priority 0.
We want CM packets to get the same scheduling priority, and therefore
map to the same SQ (Schedule Queue) and eventually TC (Traffic Class),
as the application requested for the actual QP used for the connection.

Signed-off-by: Oren Duer <oren@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/mlx4: Add raw packet QP support
Or Gerlitz [Tue, 17 Jan 2012 11:39:07 +0000 (13:39 +0200)]
IB/mlx4: Add raw packet QP support

Implement raw packet QPs for Ethernet ports using the MLX transport (as
done by the mlx4_en Ethernet netdevice driver).

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/core: Add raw packet QP type
Or Gerlitz [Thu, 1 Mar 2012 10:17:51 +0000 (12:17 +0200)]
IB/core: Add raw packet QP type

IB_QPT_RAW_PACKET allows applications to build a complete packet,
including L2 headers, when sending; on the receive side, the HW will
not strip any headers.

This QP type is designed for userspace direct access to Ethernet; for
example by applications that do TCP/IP themselves.  Only processes
with the NET_RAW capability are allowed to create raw packet QPs (the
name "raw packet QP" is supposed to suggest an analogy to AF_PACKET /
SOL_RAW sockets).

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Fix build with IPV6=n
Roland Dreier [Wed, 2 May 2012 22:42:33 +0000 (15:42 -0700)]
RDMA/ocrdma: Fix build with IPV6=n

When IPV6 is not enabled:

    ERROR: "register_inet6addr_notifier" [drivers/infiniband/hw/ocrdma/ocrdma.ko] undefined!
    ERROR: "unregister_inet6addr_notifier" [drivers/infiniband/hw/ocrdma/ocrdma.ko] undefined!

Fix this by wrapping the inet6 calls in #ifdef IPV6.  Also make the
ocrdma module depend on (IPV6 || IPV6=n) to forbid the case of modular
ipv6 but built-in ocrdma (which can't work, because ocrdma calls ipv6
functions).

Reported-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Tiny locking cleanup
Dan Carpenter [Wed, 2 May 2012 06:14:47 +0000 (09:14 +0300)]
RDMA/ocrdma: Tiny locking cleanup

We only need to disable the IRQs one time.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
[ Rename "wq_flags" to more conventional "flags."  - Roland ]

Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Fix check for NULL instead of IS_ERR
Dan Carpenter [Wed, 2 May 2012 06:13:57 +0000 (09:13 +0300)]
RDMA/ocrdma: Fix check for NULL instead of IS_ERR

The ocrdma_alloc_lkey() function never returns NULL pointers -- it
returns ERR_PTRs.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Don't sleep in atomic notifier handler
Sasha Levin [Sat, 28 Apr 2012 05:40:01 +0000 (07:40 +0200)]
RDMA/ocrdma: Don't sleep in atomic notifier handler

Events sent to ocrdma_inet6addr_event() are sent from an atomic context,
therefore we can't try to lock a mutex within the notifier callback.

We could just switch the mutex to a spinlock since all it does it
protect a list, but I've gone ahead and switched the list to use RCU
instead.  I couldn't fully test it since I don't have IB hardware, so
if it doesn't fully work for some reason let me know and I'll switch
it back to using a spinlock.

Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
[ Fixed locking in ocrdma_add().  - Roland ]

Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Remove write-only variables
Roland Dreier [Tue, 17 Apr 2012 08:18:28 +0000 (01:18 -0700)]
RDMA/ocrdma: Remove write-only variables

Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Set event's device member in ocrdma_dispatch_ibevent()
Roland Dreier [Mon, 16 Apr 2012 19:13:24 +0000 (12:13 -0700)]
RDMA/ocrdma: Set event's device member in ocrdma_dispatch_ibevent()

We need to set ib_evt.device, or else ib_dispatch_event() will crash
when we call it for unaffiliated events (and consumers may get
confused in their QP/CQ/SRQ event handler for affiliated events).

Also fix sparse warning:

    drivers/infiniband/hw/ocrdma/ocrdma_hw.c:678:36: warning: Using plain integer as NULL pointer

There's no need to clear ib_evt, since every member is initialized.

Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Make needlessly global functions/structs static
Roland Dreier [Mon, 16 Apr 2012 18:36:29 +0000 (11:36 -0700)]
RDMA/ocrdma: Make needlessly global functions/structs static

Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Fix warnings about uninitialized variables
Roland Dreier [Mon, 16 Apr 2012 18:32:17 +0000 (11:32 -0700)]
RDMA/ocrdma: Fix warnings about uninitialized variables

First, fix

    drivers/infiniband/hw/ocrdma/ocrdma_verbs.c: In function 'ocrdma_alloc_pd':
    drivers/infiniband/hw/ocrdma/ocrdma_verbs.c:371:17: warning: 'dpp_page_addr' may be used uninitialized in this function [-Wuninitialized]
    drivers/infiniband/hw/ocrdma/ocrdma_verbs.c:337:6: note: 'dpp_page_addr' was declared here

which seems that it may border on a bug (the call to ocrdma_del_mmap()
might conceivably do bad things if pd->dpp_enabled is not set and
dpp_page_addr ends up with just the wrong value).

Also take care of:

    drivers/infiniband/hw/ocrdma/ocrdma_hw.c: In function 'ocrdma_init_hw':
    drivers/infiniband/hw/ocrdma/ocrdma_hw.c:2587:5: warning: 'status' may be used uninitialized in this function [-Wuninitialized]
    drivers/infiniband/hw/ocrdma/ocrdma_hw.c:2549:17: note: 'status' was declared here

which is only real if num_eq == 0, which should be impossible.

Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/ocrdma: Add driver for Emulex OneConnect IBoE RDMA adapter
Parav Pandit [Tue, 20 Mar 2012 22:39:06 +0000 (04:09 +0530)]
RDMA/ocrdma: Add driver for Emulex OneConnect IBoE RDMA adapter

Signed-off-by: Parav Pandit <parav.pandit@emulex.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agobe2net: Add functionality to support RoCE driver
Parav Pandit [Mon, 26 Mar 2012 14:27:13 +0000 (14:27 +0000)]
be2net: Add functionality to support RoCE driver

- Increase MSI-X vectors by 5 for RoCE traffic.
- Add macro to check roce support on a device.
- Add device-specific doorbell and MSI-X vector fields shared with NIC
  functionality.
- Provide RoCE driver registration and deregistration functions.
- Add support functions which will be invoked on adapter add/remove
  and port up/down events.
- Traverse through the list of adapters to invoke callback functions.

Signed-off-by: Parav Pandit <parav.pandit@emulex.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agobe2net: Add function to issue mailbox cmd on MQ
Parav Pandit [Mon, 26 Mar 2012 14:27:12 +0000 (14:27 +0000)]
be2net: Add function to issue mailbox cmd on MQ

- Add generic function to issue mailbox cmd on MQ as export function.
- RoCE driver will use this before it setups its own MQ.

Signed-off-by: Parav Pandit <parav.pandit@emulex.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cma: Fix lockdep false positive recursive locking
Sean Hefty [Wed, 25 Apr 2012 17:42:35 +0000 (17:42 +0000)]
RDMA/cma: Fix lockdep false positive recursive locking

The following lockdep problem was reported by Or Gerlitz <ogerlitz@mellanox.com>:

    [ INFO: possible recursive locking detected ]
    3.3.0-32035-g1b2649e-dirty #4 Not tainted
    ---------------------------------------------
    kworker/5:1/418 is trying to acquire lock:
     (&id_priv->handler_mutex){+.+.+.}, at: [<ffffffffa0138a41>] rdma_destroy_i    d+0x33/0x1f0 [rdma_cm]

    but task is already holding lock:
     (&id_priv->handler_mutex){+.+.+.}, at: [<ffffffffa0135130>] cma_disable_ca    llback+0x24/0x45 [rdma_cm]

    other info that might help us debug this:
     Possible unsafe locking scenario:

           CPU0
           ----
      lock(&id_priv->handler_mutex);
      lock(&id_priv->handler_mutex);

     *** DEADLOCK ***

     May be due to missing lock nesting notation

    3 locks held by kworker/5:1/418:
     #0:  (ib_cm){.+.+.+}, at: [<ffffffff81042ac1>] process_one_work+0x210/0x4a    6
     #1:  ((&(&work->work)->work)){+.+.+.}, at: [<ffffffff81042ac1>] process_on    e_work+0x210/0x4a6
     #2:  (&id_priv->handler_mutex){+.+.+.}, at: [<ffffffffa0135130>] cma_disab    le_callback+0x24/0x45 [rdma_cm]

    stack backtrace:
    Pid: 418, comm: kworker/5:1 Not tainted 3.3.0-32035-g1b2649e-dirty #4
    Call Trace:
     [<ffffffff8102b0fb>] ? console_unlock+0x1f4/0x204
     [<ffffffff81068771>] __lock_acquire+0x16b5/0x174e
     [<ffffffff8106461f>] ? save_trace+0x3f/0xb3
     [<ffffffff810688fa>] lock_acquire+0xf0/0x116
     [<ffffffffa0138a41>] ? rdma_destroy_id+0x33/0x1f0 [rdma_cm]
     [<ffffffff81364351>] mutex_lock_nested+0x64/0x2ce
     [<ffffffffa0138a41>] ? rdma_destroy_id+0x33/0x1f0 [rdma_cm]
     [<ffffffff81065a78>] ? trace_hardirqs_on_caller+0x11e/0x155
     [<ffffffff81065abc>] ? trace_hardirqs_on+0xd/0xf
     [<ffffffffa0138a41>] rdma_destroy_id+0x33/0x1f0 [rdma_cm]
     [<ffffffffa0139c02>] cma_req_handler+0x418/0x644 [rdma_cm]
     [<ffffffffa012ee88>] cm_process_work+0x32/0x119 [ib_cm]
     [<ffffffffa0130299>] cm_req_handler+0x928/0x982 [ib_cm]
     [<ffffffffa01302f3>] ? cm_req_handler+0x982/0x982 [ib_cm]
     [<ffffffffa0130326>] cm_work_handler+0x33/0xfe5 [ib_cm]
     [<ffffffff81065a78>] ? trace_hardirqs_on_caller+0x11e/0x155
     [<ffffffffa01302f3>] ? cm_req_handler+0x982/0x982 [ib_cm]
     [<ffffffff81042b6e>] process_one_work+0x2bd/0x4a6
     [<ffffffff81042ac1>] ? process_one_work+0x210/0x4a6
     [<ffffffff813669f3>] ? _raw_spin_unlock_irq+0x2b/0x40
     [<ffffffff8104316e>] worker_thread+0x1d6/0x350
     [<ffffffff81042f98>] ? rescuer_thread+0x241/0x241
     [<ffffffff81046a32>] kthread+0x84/0x8c
     [<ffffffff8136e854>] kernel_thread_helper+0x4/0x10
     [<ffffffff81366d59>] ? retint_restore_args+0xe/0xe
     [<ffffffff810469ae>] ? __init_kthread_worker+0x56/0x56
     [<ffffffff8136e850>] ? gs_change+0xb/0xb

The actual locking is fine, since we're dealing with different locks,
but from the same lock class.  cma_disable_callback() acquires the
listening id mutex, whereas rdma_destroy_id() acquires the mutex for
the new connection id.  To fix this, delay the call to
rdma_destroy_id() until we've released the listening id mutex.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/uverbs: Lock SRQ / CQ / PD objects in a consistent order
Roland Dreier [Mon, 30 Apr 2012 19:51:50 +0000 (12:51 -0700)]
IB/uverbs: Lock SRQ / CQ / PD objects in a consistent order

Since XRC support was added, the uverbs code has locked SRQ, CQ and PD
objects needed during QP and SRQ creation in different orders
depending on the the code path.  This leads to the (at least
theoretical) possibility of deadlock, and triggers the lockdep splat
below.

Fix this by making sure we always lock the SRQ first, then CQs and
finally the PD.

    ======================================================
    [ INFO: possible circular locking dependency detected ]
    3.4.0-rc5+ #34 Not tainted
    -------------------------------------------------------
    ibv_srq_pingpon/2484 is trying to acquire lock:
     (SRQ-uobj){+++++.}, at: [<ffffffffa00af51b>] idr_read_uobj+0x2f/0x4d [ib_uverbs]

    but task is already holding lock:
     (CQ-uobj){+++++.}, at: [<ffffffffa00af51b>] idr_read_uobj+0x2f/0x4d [ib_uverbs]

    which lock already depends on the new lock.

    the existing dependency chain (in reverse order) is:

    -> #2 (CQ-uobj){+++++.}:
           [<ffffffff81070fd0>] lock_acquire+0xbf/0xfe
           [<ffffffff81384f28>] down_read+0x34/0x43
           [<ffffffffa00af51b>] idr_read_uobj+0x2f/0x4d [ib_uverbs]
           [<ffffffffa00af542>] idr_read_obj+0x9/0x19 [ib_uverbs]
           [<ffffffffa00b16c3>] ib_uverbs_create_qp+0x180/0x684 [ib_uverbs]
           [<ffffffffa00ae3dd>] ib_uverbs_write+0xb7/0xc2 [ib_uverbs]
           [<ffffffff810fe47f>] vfs_write+0xa7/0xee
           [<ffffffff810fe65f>] sys_write+0x45/0x69
           [<ffffffff8138cdf9>] system_call_fastpath+0x16/0x1b

    -> #1 (PD-uobj){++++++}:
           [<ffffffff81070fd0>] lock_acquire+0xbf/0xfe
           [<ffffffff81384f28>] down_read+0x34/0x43
           [<ffffffffa00af51b>] idr_read_uobj+0x2f/0x4d [ib_uverbs]
           [<ffffffffa00af542>] idr_read_obj+0x9/0x19 [ib_uverbs]
           [<ffffffffa00af8ad>] __uverbs_create_xsrq+0x96/0x386 [ib_uverbs]
           [<ffffffffa00b31b9>] ib_uverbs_detach_mcast+0x1cd/0x1e6 [ib_uverbs]
           [<ffffffffa00ae3dd>] ib_uverbs_write+0xb7/0xc2 [ib_uverbs]
           [<ffffffff810fe47f>] vfs_write+0xa7/0xee
           [<ffffffff810fe65f>] sys_write+0x45/0x69
           [<ffffffff8138cdf9>] system_call_fastpath+0x16/0x1b

    -> #0 (SRQ-uobj){+++++.}:
           [<ffffffff81070898>] __lock_acquire+0xa29/0xd06
           [<ffffffff81070fd0>] lock_acquire+0xbf/0xfe
           [<ffffffff81384f28>] down_read+0x34/0x43
           [<ffffffffa00af51b>] idr_read_uobj+0x2f/0x4d [ib_uverbs]
           [<ffffffffa00af542>] idr_read_obj+0x9/0x19 [ib_uverbs]
           [<ffffffffa00b1728>] ib_uverbs_create_qp+0x1e5/0x684 [ib_uverbs]
           [<ffffffffa00ae3dd>] ib_uverbs_write+0xb7/0xc2 [ib_uverbs]
           [<ffffffff810fe47f>] vfs_write+0xa7/0xee
           [<ffffffff810fe65f>] sys_write+0x45/0x69
           [<ffffffff8138cdf9>] system_call_fastpath+0x16/0x1b

    other info that might help us debug this:

    Chain exists of:
      SRQ-uobj --> PD-uobj --> CQ-uobj

     Possible unsafe locking scenario:

           CPU0                    CPU1
           ----                    ----
      lock(CQ-uobj);
                                   lock(PD-uobj);
                                   lock(CQ-uobj);
      lock(SRQ-uobj);

     *** DEADLOCK ***

    3 locks held by ibv_srq_pingpon/2484:
     #0:  (QP-uobj){+.+...}, at: [<ffffffffa00b162c>] ib_uverbs_create_qp+0xe9/0x684 [ib_uverbs]
     #1:  (PD-uobj){++++++}, at: [<ffffffffa00af51b>] idr_read_uobj+0x2f/0x4d [ib_uverbs]
     #2:  (CQ-uobj){+++++.}, at: [<ffffffffa00af51b>] idr_read_uobj+0x2f/0x4d [ib_uverbs]

    stack backtrace:
    Pid: 2484, comm: ibv_srq_pingpon Not tainted 3.4.0-rc5+ #34
    Call Trace:
     [<ffffffff8137eff0>] print_circular_bug+0x1f8/0x209
     [<ffffffff81070898>] __lock_acquire+0xa29/0xd06
     [<ffffffffa00af37c>] ? __idr_get_uobj+0x20/0x5e [ib_uverbs]
     [<ffffffffa00af51b>] ? idr_read_uobj+0x2f/0x4d [ib_uverbs]
     [<ffffffff81070fd0>] lock_acquire+0xbf/0xfe
     [<ffffffffa00af51b>] ? idr_read_uobj+0x2f/0x4d [ib_uverbs]
     [<ffffffff81070eee>] ? lock_release+0x166/0x189
     [<ffffffff81384f28>] down_read+0x34/0x43
     [<ffffffffa00af51b>] ? idr_read_uobj+0x2f/0x4d [ib_uverbs]
     [<ffffffffa00af51b>] idr_read_uobj+0x2f/0x4d [ib_uverbs]
     [<ffffffffa00af542>] idr_read_obj+0x9/0x19 [ib_uverbs]
     [<ffffffffa00b1728>] ib_uverbs_create_qp+0x1e5/0x684 [ib_uverbs]
     [<ffffffff81070fec>] ? lock_acquire+0xdb/0xfe
     [<ffffffff81070c09>] ? lock_release_non_nested+0x94/0x213
     [<ffffffff810d470f>] ? might_fault+0x40/0x90
     [<ffffffff810d470f>] ? might_fault+0x40/0x90
     [<ffffffffa00ae3dd>] ib_uverbs_write+0xb7/0xc2 [ib_uverbs]
     [<ffffffff810fe47f>] vfs_write+0xa7/0xee
     [<ffffffff810ff736>] ? fget_light+0x3b/0x99
     [<ffffffff810fe65f>] sys_write+0x45/0x69
     [<ffffffff8138cdf9>] system_call_fastpath+0x16/0x1b

Reported-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/uverbs: Make lockdep output more readable
Roland Dreier [Mon, 30 Apr 2012 17:27:26 +0000 (10:27 -0700)]
IB/uverbs: Make lockdep output more readable

Add names for our lockdep classes, so instead of having to decipher
lockdep output with mysterious names:

    Chain exists of:
      key#14 --> key#11 --> key#13

lockdep will give us something nicer:

    Chain exists of:
      SRQ-uobj --> PD-uobj --> CQ-uobj

Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/ipath: Replace open-coded ARRAY_SIZE with macro
Mike Marciniszyn [Thu, 19 Apr 2012 15:59:15 +0000 (11:59 -0400)]
IB/ipath: Replace open-coded ARRAY_SIZE with macro

Change sizeof(array)/sizeof(array[0]) to ARRAY_SIZE.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/ipath: Replace open-coded ARRAY_SIZE with macro
Jim Cromie [Wed, 11 Apr 2012 00:55:25 +0000 (18:55 -0600)]
IB/ipath: Replace open-coded ARRAY_SIZE with macro

Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Acked-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoRDMA/cxgb4: Use dst parameter in import_ep()
Steve Wise [Fri, 27 Apr 2012 15:24:33 +0000 (10:24 -0500)]
RDMA/cxgb4: Use dst parameter in import_ep()

Function import_ep() is incorrectly using ep->dst instead of the dst
ptr passed in.  This causes a crash when accepting new rdma connections
becase ep->dst is not initialized yet.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoIB/core: Use qp->usecnt to track multicast attach/detach
Or Gerlitz [Sun, 29 Apr 2012 14:04:22 +0000 (17:04 +0300)]
IB/core: Use qp->usecnt to track multicast attach/detach

Just as we don't allow PDs, CQs, etc. to be destroyed if there are QPs
that are attached to them, don't let a QP be destroyed if there are
multicast group(s) attached to it.  Use the existing usecnt field of
struct ib_qp which was added by commit 0e0ec7e ("RDMA/core: Export
ib_open_qp() to share XRC TGT QPs") to track this.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
12 years agoLinux 3.4-rc6
Linus Torvalds [Sun, 6 May 2012 22:07:32 +0000 (15:07 -0700)]
Linux 3.4-rc6

12 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 6 May 2012 19:19:38 +0000 (12:19 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes form Peter Anvin

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  intel_mid_powerbtn: mark irq as IRQF_NO_SUSPEND
  arch/x86/platform/geode/net5501.c: change active_low to 0 for LED driver
  x86, relocs: Remove an unused variable
  asm-generic: Use __BITS_PER_LONG in statfs.h
  x86/amd: Re-enable CPU topology extensions in case BIOS has disabled it

12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Sun, 6 May 2012 17:20:07 +0000 (10:20 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "The big ones here are a memory leak we introduced in rc1, and a
  scheduling while atomic if the transid on disk doesn't match the
  transid we expected.  This happens for corrupt blocks, or out of date
  disks.

  It also fixes up the ioctl definition for our ioctl to resolve logical
  inode numbers.  The __u32 was a merging error and doesn't match what
  we ship in the progs."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: avoid sleeping in verify_parent_transid while atomic
  Btrfs: fix crash in scrub repair code when device is missing
  btrfs: Fix mismatching struct members in ioctl.h
  Btrfs: fix page leak when allocing extent buffers
  Btrfs: Add properly locking around add_root_to_dirty_list

12 years agox86: fix broken TASK_SIZE for ia32_aout
Al Viro [Sun, 6 May 2012 16:20:00 +0000 (17:20 +0100)]
x86: fix broken TASK_SIZE for ia32_aout

Setting TIF_IA32 in load_aout_binary() used to be enough; these days
TASK_SIZE is controlled by TIF_ADDR32 and that one doesn't get set
there.  Switch to use of set_personality_ia32()...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoBtrfs: avoid sleeping in verify_parent_transid while atomic
Chris Mason [Sun, 6 May 2012 11:23:47 +0000 (07:23 -0400)]
Btrfs: avoid sleeping in verify_parent_transid while atomic

verify_parent_transid needs to lock the extent range to make
sure no IO is underway, and so it can safely clear the
uptodate bits if our checks fail.

But, a few callers are using it with spinlocks held.  Most
of the time, the generation numbers are going to match, and
we don't want to switch to a blocking lock just for the error
case.  This adds an atomic flag to verify_parent_transid,
and changes it to return EAGAIN if it needs to block to
properly verifiy things.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mattst88...
Linus Torvalds [Sat, 5 May 2012 23:34:38 +0000 (16:34 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mattst88/alpha

Pull alpha fixes from Matt Turner:
 "My alpha tree is back up (after taking quite some time to get my GPG
  key signed).  It contains just some simple fixes."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mattst88/alpha:
  alpha: silence 'const' warning in sys_marvel.c
  alpha: include module.h to fix modpost on Tsunami
  alpha: properly define get/set_rtc_time on Marvel/SMP
  alpha: VGA_HOSE depends on VGA_CONSOLE

12 years agoTTY: pdc_cons, fix regression in close
Jiri Slaby [Sat, 5 May 2012 20:49:10 +0000 (22:49 +0200)]
TTY: pdc_cons, fix regression in close

The test in pdc_console_tty_close '!tty->count' was always wrong
because tty->count is decremented after tty->ops->close is called and
thus can never be zero. Hence the 'then' branch was never executed and
the timer never deleted.

This did not matter until commit 5dd5bc40f3b6 ("TTY: pdc_cons, use
tty_port").  There we needed to set TTY in tty_port to NULL, but this
never happened due to the bug above.

So change the test to really trigger at the last close by changing the
condition to 'tty->count == 1'.

Well, the driver should not touch tty->count at all.  It should use
tty_port->count and count open count there itself.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Reported-and-tested-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Helge Deller <deller@gmx.de>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge tag 'sound-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
Linus Torvalds [Sat, 5 May 2012 17:07:06 +0000 (10:07 -0700)]
Merge tag 'sound-3.4' of git://git./linux/kernel/git/tiwai/sound

Pull sound sound fixes from Takashi Iwai:
 "As good as nothing exciting here; just a few trivial fixes for various
  ASoC stuff."

* tag 'sound-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ASoC: omap-pcm: Free dma buffers in case of error.
  ASoC: s3c2412-i2s: Fix dai registration
  ASoC: wm8350: Don't use locally allocated codec struct
  ASoC: tlv312aic23: unbreak resume
  ASoC: bf5xx-ssm2602: Set DAI format
  ASoC: core: check of_property_count_strings failure
  ASoC: dt: sgtl5000.txt: Add description for 'reg' field
  ASoC: wm_hubs: Make sure we don't disable differential line outputs

12 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux
Linus Torvalds [Sat, 5 May 2012 17:06:06 +0000 (10:06 -0700)]
Merge branch 'release' of git://git./linux/kernel/git/lenb/linux

Pull an ACPI patch from Len Brown:
 "It fixes a D3 issue new in 3.4-rc1."

By Lin Ming via Len Brown:
* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux:
  ACPI: Fix D3hot v D3cold confusion

12 years agoinit: don't try mounting device as nfs root unless type fully matches
Sasha Levin [Sat, 5 May 2012 15:06:35 +0000 (17:06 +0200)]
init: don't try mounting device as nfs root unless type fully matches

Currently, we'll try mounting any device who's major device number is
UNNAMED_MAJOR as NFS root.  This would happen for non-NFS devices as
well (such as 9p devices) but it wouldn't cause any issues since
mounting the device as NFS would fail quickly and the code proceeded to
doing the proper mount:

       [  101.522716] VFS: Unable to mount root fs via NFS, trying floppy.
       [  101.534499] VFS: Mounted root (9p filesystem) on device 0:18.

Commit 6829a048102a ("NFS: Retry mounting NFSROOT") introduced retries
when mounting NFS root, which means that now we don't immediately fail
and instead it takes an additional 90+ seconds until we stop retrying,
which has revealed the issue this patch fixes.

This meant that it would take an additional 90 seconds to boot when
we're not using a device type which gets detected in order before NFS.

This patch modifies the NFS type check to require device type to be
'Root_NFS' instead of requiring the device to have an UNNAMED_MAJOR
major.  This makes boot process cleaner since we now won't go through
the NFS mounting code at all when the device isn't an NFS root
("/dev/nfs").

Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge branch 'fix/asoc' into for-linus
Takashi Iwai [Sat, 5 May 2012 09:27:26 +0000 (11:27 +0200)]
Merge branch 'fix/asoc' into for-linus

12 years agoMerge branch 'for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/lrg/asoc...
Takashi Iwai [Sat, 5 May 2012 09:26:50 +0000 (11:26 +0200)]
Merge branch 'for-3.4' of git://git./linux/kernel/git/lrg/asoc into fix/asoc

12 years agoMerge tag 'asoc-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound...
Takashi Iwai [Sat, 5 May 2012 09:25:17 +0000 (11:25 +0200)]
Merge tag 'asoc-3.4' of git://git./linux/kernel/git/broonie/sound into for-linus

ASoC: Updates for 3.4

Nothing terribly exciting here, a bunch of small and simple fixes
scattered around the place.

12 years agoACPI: Fix D3hot v D3cold confusion
Lin Ming [Mon, 23 Apr 2012 01:03:49 +0000 (09:03 +0800)]
ACPI: Fix D3hot v D3cold confusion

Before this patch, ACPI_STATE_D3 incorrectly referenced D3hot
in some places, but D3cold in other places.

After this patch, ACPI_STATE_D3 always means ACPI_STATE_D3_COLD;
and all references to D3hot use ACPI_STATE_D3_HOT.

ACPI's _PR3 method is used to enter both D3hot and D3cold states.
What distinguishes D3hot from D3cold is the presence _PR3
(Power Resources for D3hot)  If these resources are all ON,
then the state is D3hot.  If _PR3 is not present,
or all _PR0 resources for the devices are OFF,
then the state is D3cold.

This patch applies after Linux-3.4-rc1.
A future syntax cleanup may remove ACPI_STATE_D3
to emphasize that it always means ACPI_STATE_D3_COLD.

Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Reviewed-by: Aaron Lu <aaron.lu@amd.com>
Signed-off-by: Len Brown <len.brown@intel.com>
12 years agohfsplus: Fix potential buffer overflows
Greg Kroah-Hartman [Fri, 4 May 2012 19:09:39 +0000 (12:09 -0700)]
hfsplus: Fix potential buffer overflows

Commit ec81aecb2966 ("hfs: fix a potential buffer overflow") fixed a few
potential buffer overflows in the hfs filesystem.  But as Timo Warns
pointed out, these changes also need to be made on the hfsplus
filesystem as well.

Reported-by: Timo Warns <warns@pre-sense.de>
Acked-by: WANG Cong <amwang@redhat.com>
Cc: Alexey Khoroshilov <khoroshilov@ispras.ru>
Cc: Miklos Szeredi <mszeredi@suse.cz>
Cc: Sage Weil <sage@newdream.net>
Cc: Eugene Teo <eteo@redhat.com>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Dave Anderson <anderson@redhat.com>
Cc: stable <stable@vger.kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 4 May 2012 22:35:09 +0000 (15:35 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Thomas Gleixner.

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  rtc: Fix possible null pointer dereference in rtc-mpc5121.c

12 years agoMerge git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Fri, 4 May 2012 22:34:21 +0000 (15:34 -0700)]
Merge git://git.samba.org/sfrench/cifs-2.6

Pull CIFS fixes from Steve French.

* git://git.samba.org/sfrench/cifs-2.6:
  fs/cifs: fix parsing of dfs referrals
  cifs: make sure we ignore the credentials= and cred= options
  [CIFS] Update cifs version to 1.78
  cifs - check S_AUTOMOUNT in revalidate
  cifs: add missing initialization of server->req_lock
  cifs: don't cap ra_pages at the same level as default_backing_dev_info
  CIFS: Fix indentation in cifs_show_options

12 years agoCPU frequency drivers MAINTAINERS update
Dave Jones [Fri, 4 May 2012 16:04:17 +0000 (12:04 -0400)]
CPU frequency drivers MAINTAINERS update

Remove myself as cpufreq maintainer.
x86 driver changes can go through the regular x86/ACPI trees.
ARM driver changes through the ARM trees.
cpufreq core changes are rare these days, and can just go to lkml/direct.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoseqlock: add 'raw_seqcount_begin()' function
Linus Torvalds [Fri, 4 May 2012 22:13:54 +0000 (15:13 -0700)]
seqlock: add 'raw_seqcount_begin()' function

The normal read_seqcount_begin() function will wait for any current
writers to exit their critical region by looping until the sequence
count is even.

That "wait for sequence count to stabilize" is the right thing to do if
the read-locker will just retry the whole operation on contention: no
point in doing a potentially expensive reader sequence if we know at the
beginning that we'll just end up re-doing it all.

HOWEVER.  Some users don't actually retry the operation, but instead
will abort and do the operation with proper locking.  So the sequence
count case may be the optimistic quick case, but in the presense of
writers you may want to do full locking in order to guarantee forward
progress.  The prime example of this would be the RCU name lookup.

And in that case, you may well be better off without the "retry early",
and are in a rush to instead get to the failure handling.  Thus this
"raw" interface that just returns the sequence number without testing it
- it just forces the low bit to zero so that read_seqcount_retry() will
always fail such a "active concurrent writer" scenario.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoFix __read_seqcount_begin() to use ACCESS_ONCE for sequence value read
Linus Torvalds [Fri, 4 May 2012 21:46:02 +0000 (14:46 -0700)]
Fix __read_seqcount_begin() to use ACCESS_ONCE for sequence value read

We really need to use a ACCESS_ONCE() on the sequence value read in
__read_seqcount_begin(), because otherwise the compiler might end up
reloading the value in between the test and the return of it.  As a
result, it might end up returning an odd value (which means that a write
is in progress).

If the reader is then fast enough that that odd value is still the
current one when the read_seqcount_retry() is done, we might end up with
a "successful" read sequence, even despite the concurrent write being
active.

In practice this probably never really happens - there just isn't
anything else going on around the read of the sequence count, and the
common case is that we end up having a read barrier immediately
afterwards.

So the code sequence in which gcc might decide to reaload from memory is
small, and there's no reason to believe it would ever actually do the
reload.  But if the compiler ever were to decide to do so, it would be
incredibly annoying to debug.  Let's just make sure.

Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agointel_mid_powerbtn: mark irq as IRQF_NO_SUSPEND
Yong Wang [Fri, 4 May 2012 21:02:44 +0000 (14:02 -0700)]
intel_mid_powerbtn: mark irq as IRQF_NO_SUSPEND

So that the power button still wakes up the platform.

Signed-off-by: Pierre Tardy <pierre.tardy@intel.com>
Link: http://lkml.kernel.org/r/20120504210244.F2EA5A018B@akpm.mtv.corp.google.com
Tested-by: Kangkai Yin <kangkai.yin@intel.com>
Tested-by: Yong Wang <yong.y.wang@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Cc: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
12 years agoarch/x86/platform/geode/net5501.c: change active_low to 0 for LED driver
Bjarke Istrup Pedersen [Fri, 4 May 2012 21:01:45 +0000 (14:01 -0700)]
arch/x86/platform/geode/net5501.c: change active_low to 0 for LED driver

It seems that there was an error with the active_low = 1 for the
LED, since it should be set to 0 (meaning that active is high,
since 0 is false, hence the confusion.

The wiki article about it confuses it, since it contradicts itself,
regarding what turns on the LED.

I have tested 3.4-rc2 on my net5501 with this patch, and it makes the LED
behave correctly, where "none" turns it off, and "default-on" turns it on,
when echoed onto the trigger "file" in /sys/class/leds.

Signed-off-by: Bjarke Istrup Pedersen <gurligebis@gentoo.org>
Link: http://lkml.kernel.org/r/20120504210146.62186A018B@akpm.mtv.corp.google.com
Cc: Philip Prindeville <philipp@redfish-solutions.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
12 years agoBtrfs: fix crash in scrub repair code when device is missing
Stefan Behrens [Fri, 4 May 2012 19:16:07 +0000 (15:16 -0400)]
Btrfs: fix crash in scrub repair code when device is missing

Fix that when scrub tries to repair an I/O or checksum error and one of
the devices containing the mirror is missing, it crashes in bio_add_page
because the bdev is a NULL pointer for missing devices.

Reported-by: Marco L. Crociani <marco.crociani@gmail.com>
Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agobtrfs: Fix mismatching struct members in ioctl.h
Alexander Block [Fri, 4 May 2012 19:16:06 +0000 (15:16 -0400)]
btrfs: Fix mismatching struct members in ioctl.h

Fix the size members of btrfs_ioctl_ino_path_args and
btrfs_ioctl_logical_ino_args. The user space btrfs-progs utilities used
__u64 and the kernel headers used __u32 before.

Signed-off-by: Alexander Block <ablock84@googlemail.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoBtrfs: fix page leak when allocing extent buffers
Josef Bacik [Fri, 4 May 2012 19:16:06 +0000 (15:16 -0400)]
Btrfs: fix page leak when allocing extent buffers

If we happen to alloc a extent buffer and then alloc a page and notice that
page is already attached to an extent buffer, we will only unlock it and
free our existing eb.  Any pages currently attached to that eb will be
properly freed, but we don't do the page_cache_release() on the page where
we noticed the other extent buffer which can cause us to leak pages and I
hope cause the weird issues we've been seeing in this area.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoBtrfs: Add properly locking around add_root_to_dirty_list
Chris Mason [Thu, 3 May 2012 16:08:48 +0000 (12:08 -0400)]
Btrfs: Add properly locking around add_root_to_dirty_list

add_root_to_dirty_list happens once at the very beginning of the
transaction, but it is still racey.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 4 May 2012 14:57:13 +0000 (07:57 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Some minor fixes from Intel and a radeon fix.

  I have the nouveau fix for the i2c regression queued for next week,
  its mostly a revert and seems to work on the system it was originally
  introduced for thanks to some i2c core changes."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/radeon: clarify and extend wb setup on APUs and NI+ asics
  drm/i915: enable dip before writing data on gen4
  fixing dmi match for hp t5745 and hp st5747 thin client
  drm/i915: Only enable IPS polling for gen5
  drm/i915: Do not read non-existent DPLL registers on PCH hardware

12 years agoMerge tag 'md-3.4-fixes' of git://neil.brown.name/md
Linus Torvalds [Fri, 4 May 2012 14:56:22 +0000 (07:56 -0700)]
Merge tag 'md-3.4-fixes' of git://neil.brown.name/md

Pull one small fix for md/bitmaps from NeilBrown:
 "This fixes a regression that was introduced in the merge window."

* tag 'md-3.4-fixes' of git://neil.brown.name/md:
  md/bitmap: fix calculation of 'chunks' - missing shift.

12 years agoMerge branch 'fix-unmapped-word-at-a-time'
Linus Torvalds [Fri, 4 May 2012 14:50:50 +0000 (07:50 -0700)]
Merge branch 'fix-unmapped-word-at-a-time'

Jana Saout confirmed that this fixes the page faults he saw.

His problem was triggered by ocfs2 and autofs symlink lookups, where the
symlink allocation was at the end of a page.  But the deeper reason
seems to be the use of Xen-PV, which is what then causes him to have all
these unmapped pages, which is what then makes it a problem when the
unaligned word-at-a-time code fetches data past the end of a page.

* fix-unmapped-word-at-a-time:
  vfs: make word-at-a-time accesses handle a non-existing page

12 years agoASoC: omap-pcm: Free dma buffers in case of error.
Oleg Matcovschi [Wed, 25 Apr 2012 02:02:02 +0000 (19:02 -0700)]
ASoC: omap-pcm: Free dma buffers in case of error.

Signed-off-by: Oleg Matcovschi <oleg.matcovschi@ti.com>
Acked-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Acked-by: Jarkko Nikula <jarkko.nikula@bitmer.com>
Acked-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Liam Girdwood <lrg@ti.com>
12 years agodrm/radeon: clarify and extend wb setup on APUs and NI+ asics
Alex Deucher [Thu, 3 May 2012 21:06:28 +0000 (17:06 -0400)]
drm/radeon: clarify and extend wb setup on APUs and NI+ asics

Use family rather than DCE check for clarity, also always use
wb on APUs, there will never be AGP variants.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agomd/bitmap: fix calculation of 'chunks' - missing shift.
NeilBrown [Fri, 4 May 2012 07:03:18 +0000 (17:03 +1000)]
md/bitmap: fix calculation of 'chunks' - missing shift.

commit 61a0d80c "md/bitmap: discard CHUNK_BLOCK_SHIFT macro"
replaced CHUNK_BLOCK_RATIO() by the same text that was
replacing CHUNK_BLOCK_SHIFT() - which is clearly wrong.

The result is that 'chunks' is often too small by 1,
which can sometimes result in a crash (not sure how).

So use the correct replacement, and get rid of CHUNK_BLOCK_RATIO
which is no longe used.

Reported-by: Karl Newman <siliconfiend@gmail.com>
Tested-by: Karl Newman <siliconfiend@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>
12 years agofs/cifs: fix parsing of dfs referrals
Stefan Metzmacher [Thu, 3 May 2012 22:19:28 +0000 (00:19 +0200)]
fs/cifs: fix parsing of dfs referrals

The problem was that the first referral was parsed more than once
and so the caller tried the same referrals multiple times.

The problem was introduced partly by commit
066ce6899484d9026acd6ba3a8dbbedb33d7ae1b,
where 'ref += le16_to_cpu(ref->Size);' got lost,
but that was also wrong...

Cc: <stable@vger.kernel.org>
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Tested-by: Björn Jacke <bj@sernet.de>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
12 years agoMerge tag 'mfd-for-linus-3.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 4 May 2012 00:21:05 +0000 (17:21 -0700)]
Merge tag 'mfd-for-linus-3.4-rc6' of git://git./linux/kernel/git/sameo/mfd-2.6

Pull second set of MFD fixes from Samuel Ortiz:
 "This time we only have a one liner fixing an omap-usb build error."

* tag 'mfd-for-linus-3.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6:
  mfd: Fix build breakage in omap-usb-host.c

12 years agoMerge branch 'efi-vars' from Matthew Garrett
Linus Torvalds [Fri, 4 May 2012 00:19:48 +0000 (17:19 -0700)]
Merge branch 'efi-vars' from Matthew Garrett

* efi-vars:
  efivars: Improve variable validation

12 years agoefivars: Improve variable validation
Matthew Garrett [Thu, 3 May 2012 20:50:46 +0000 (16:50 -0400)]
efivars: Improve variable validation

Ben Hutchings pointed out that the validation in efivars was inadequate -
most obviously, an entry with size 0 would server as a DoS against the
kernel. Improve this based on his suggestions.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge tag 'tag/upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarz...
Linus Torvalds [Fri, 4 May 2012 00:16:52 +0000 (17:16 -0700)]
Merge tag 'tag/upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

Pull libata fixes from Jeff Garzik:

1) Fix regression that could cause a misdiagnosis, which in turn could
   lead to an erroneous 3.0 Gbps -> 1.5 downshift, particularly when hotplug
   and suspend/resume is involved.

2) Fix a regression that led to ata%d controller ids being numbered one
   larger than in <= 3.4-rc3 (oh, the horror!).  Controller ids should now be
   as expected.

3) add some DT, PCI id's

4) ata/pata_arasan_cf: minor cpp fixing/cleaning

* tag 'tag/upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  ata: ahci_platform: Add synopsys ahci controller in DT's compatible list
  ata/pata_arasan_cf: Move arasan_cf_pm_ops out of #ifdef, #endif macros
  libata: init ata_print_id to 0
  ahci: Detect Marvell 88SE9172 SATA controller
  libata: skip old error history when counting probe trials

12 years agoMerge branch 'i2c-embedded/for-current' of git://git.pengutronix.de/git/wsa/linux
Linus Torvalds [Fri, 4 May 2012 00:15:47 +0000 (17:15 -0700)]
Merge branch 'i2c-embedded/for-current' of git://git.pengutronix.de/git/wsa/linux

Pull i2c embedded fixes from Wolfram Sang:
 "Here are some typical i2c driver bugfixes for 3.4.  Missed clock
  handling, improper timeout fixes, hardware wrokarounds...  All
  patches have been in linux-next for a few days, too."

* 'i2c-embedded/for-current' of git://git.pengutronix.de/git/wsa/linux:
  i2c: mxs: disable QUEUE when sending is done
  i2c: mxs: handle spurious interrupt
  i2c-eg20t: Modify MODULE_AUTHOR's email address
  i2c-eg20t: change timeout value 50msec to 1000msec
  i2c: tegra: Add delay before resetting the controller after NACK
  i2c: pnx: Disable clk in suspend

12 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 4 May 2012 00:14:55 +0000 (17:14 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Just some regression fixes from Ben along with a variable that gcc
  failed to spot is uninitialised."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  nouveau: initialise has_optimus variable.
  drm/nv10/gpio: fix thinko in mask for gpio lines 2-9
  nvc0/fb: shut up PMFB interrupt after the first occurrence
  drm/nouveau/hdmi: use correct hdmi regs for nvaa/nvac
  drm/nouveau/bios: fix regression on some nv4x board

12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 4 May 2012 00:10:39 +0000 (17:10 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Transfer padding was wrong for full-speed USB in ASIX driver, fix
    from Ingo van Lil.

 2) Propagate the negative packet offset fix into the PowerPC BPF JIT.
    From Jan Seiffert.

 3) dl2k driver's private ioctls were letting unprivileged tasks make
    MII writes and other ugly bits like that.  Fix from Jeff Mahoney.

 4) Fix TX VLAN and RX packet drops in ucc_geth, from Joakim Tjernlund.

 5) OOPS and network namespace fixes in IPVS from Hans Schillstrom and
    Julian Anastasov.

 6) Fix races and sleeping in locked context bugs in drop_monitor, from
    Neil Horman.

 7) Fix link status indication in smsc95xx driver, from Paolo Pisati.

 8) Fix bridge netfilter OOPS, from Peter Huang.

 9) L2TP sendmsg can return on error conditions with the socket lock
    held, oops.  Fix from Sasha Levin.

10) udp_diag should return meaningful values for socket memory usage,
    from Shan Wei.

11) Eric Dumazet is so awesome he gets his own section:

       Socket memory cgroup code (I never should have applied those
       patches, grumble...) made erroneous changes to
       sk_sockets_allocated_read_positive().  It was changed to
       use percpu_counter_sum_positive (which requires BH disabling)
       instead of percpu_counter_read_positive (which does not).
       Revert back to avoid crashes and lockdep warnings.

       Adjust the default tcp_adv_win_scale and tcp_rmem[2] values
       to fix throughput regressions.  This is necessary as a result
       of our more precise skb->truesize tracking.

       Fix SKB leak in netem packet scheduler.

12) New device IDs for various bluetooth devices, from Manoj Iyer,
    AceLan Kao, and Steven Harms.

13) Fix command completion race in ipw2200, from Stanislav Yakovlev.

14) Fix rtlwifi oops on unload, from Larry Finger.

15) Fix hard_mtu when adjusting hard_header_len in smsc95xx driver.
    From Stephane Fillod.

16) ehea driver registers it's IRQ before all the necessary state is
    setup, resulting in crashes.  Fix from Thadeu Lima de Souza
    Cascardo.

17) Fix PHY connection failures in davinci_emac driver, from Anatolij
    Gustschin.

18) Missing break; in switch statement in bluetooth's
    hci_cmd_complete_evt().  Fix from Szymon Janc.

19) Fix queue programming in iwlwifi, from Johannes Berg.

20) Interrupt throttling defaults not being actually programmed into the
    hardware, fix from Jeff Kirsher and Ying Cai.

21) TLAN driver SKB encoding in descriptor busted on 64-bit, fix from
    Benjamin Poirier.

22) Fix blind status block RX producer pointer deref in TG3 driver, from
    Matt Carlson.

23) Promisc and multicast are busted on ehea, fixes from Thadeu Lima de
    Souza Cascardo.

24) Fix crashes in 6lowpan, from Alexander Smirnov.

25) tcp_complete_cwr() needs to be careful to not rewind the CWND to
    ssthresh if ssthresh has the "infinite" value.  Fix from Yuchung
    Cheng.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (81 commits)
  sungem: Fix WakeOnLan
  tcp: change tcp_adv_win_scale and tcp_rmem[2]
  net: l2tp: unlock socket lock before returning from l2tp_ip_sendmsg
  drop_monitor: prevent init path from scheduling on the wrong cpu
  usbnet: fix failure handling in usbnet_probe
  usbnet: fix leak of transfer buffer of dev->interrupt
  ucc_geth: Add 16 bytes to max TX frame for VLANs
  net: ucc_geth, increase no. of HW RX descriptors
  netem: fix possible skb leak
  sky2: fix receive length error in mixed non-VLAN/VLAN traffic
  sky2: propogate rx hash when packet is copied
  net: fix two typos in skbuff.h
  cxgb3: Don't call cxgb_vlan_mode until q locks are initialized
  ixgbe: fix calling skb_put on nonlinear skb assertion bug
  ixgbe: Fix a memory leak in IEEE DCB
  igbvf: fix the bug when initializing the igbvf
  smsc75xx: enable mac to detect speed/duplex from phy
  smsc75xx: declare smsc75xx's MII as GMII capable
  smsc75xx: fix phy interrupt acknowledge
  smsc75xx: fix phy init reset loop
  ...

12 years agoMerge tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck...
Linus Torvalds [Fri, 4 May 2012 00:08:58 +0000 (17:08 -0700)]
Merge tag 'hwmon-for-linus' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fixes from Guenter Roeck:
 "Fix OOPS seen in coretemp driver if the CPU core ID is too large"

* tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (coretemp) Increase CPU core limit
  hwmon: (coretemp) fix oops on cpu unplug

12 years agovfs: make word-at-a-time accesses handle a non-existing page
Linus Torvalds [Thu, 3 May 2012 17:16:43 +0000 (10:16 -0700)]
vfs: make word-at-a-time accesses handle a non-existing page

It turns out that there are more cases than CONFIG_DEBUG_PAGEALLOC that
can have holes in the kernel address space: it seems to happen easily
with Xen, and it looks like the AMD gart64 code will also punch holes
dynamically.

Actually hitting that case is still very unlikely, so just do the
access, and take an exception and fix it up for the very unlikely case
of it being a page-crosser with no next page.

And hey, this abstraction might even help other architectures that have
other issues with unaligned word accesses than the possible missing next
page.  IOW, this could do the byte order magic too.

Peter Anvin fixed a thinko in the shifting for the exception case.

Reported-and-tested-by: Jana Saout <jana@saout.de>
Cc: Peter Anvin <hpa@zytor.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocifs: make sure we ignore the credentials= and cred= options
Jeff Layton [Wed, 2 May 2012 18:02:40 +0000 (14:02 -0400)]
cifs: make sure we ignore the credentials= and cred= options

Older mount.cifs programs passed this on to the kernel after parsing
the file. Make sure the kernel ignores that option.

Should fix:

    https://bugzilla.kernel.org/show_bug.cgi?id=43195

Cc: Sachin Prabhu <sprabhu@redhat.com>
Reported-by: Ronald <ronald645@gmail.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
12 years ago[CIFS] Update cifs version to 1.78
Steve French [Wed, 2 May 2012 16:58:19 +0000 (11:58 -0500)]
[CIFS] Update cifs version to 1.78

Signed-off-by: Steve French <sfrench@us.ibm.com>
12 years agocifs - check S_AUTOMOUNT in revalidate
Ian Kent [Wed, 2 May 2012 11:19:09 +0000 (07:19 -0400)]
cifs - check S_AUTOMOUNT in revalidate

When revalidating a dentry, if the inode wasn't known to be a dfs
entry when the dentry was instantiated, such as when created via
->readdir(), the DCACHE_NEED_AUTOMOUNT flag needs to be set on the
dentry in ->d_revalidate().

The false return from cifs_d_revalidate(), due to the inode now
being marked with the S_AUTOMOUNT flag, might not invalidate the
dentry if there is a concurrent unlazy path walk. This is because
the dentry reference count will be at least 2 in this case causing
d_invalidate() to return EBUSY. So the asumption that the dentry
will be discarded then correctly instantiated via ->lookup() might
not hold.

Signed-off-by: Ian Kent <raven@themaw.net>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Cc: Steve French <smfrench@gmail.com>
Cc: linux-cifs@vger.kernel.org
Signed-off-by: Steve French <sfrench@us.ibm.com>
12 years agoata: ahci_platform: Add synopsys ahci controller in DT's compatible list
Viresh Kumar [Sat, 21 Apr 2012 12:10:12 +0000 (17:40 +0530)]
ata: ahci_platform: Add synopsys ahci controller in DT's compatible list

SPEAr13xx series of SoCs contain Synopsys AHCI SATA Controller which shares
ahci_platform driver with other controller versions.

This patch updates DT compatible list for ahci_platform. It also updates and
renames binding documentation to more generic name.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Cc: Rob Herring <rob.herring@calxeda.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>