platform/adaptation/renesas_rcar/renesas_kernel.git
13 years agoMerge branches 'amso1100', 'cma', 'cxgb3', 'cxgb4', 'ehca', 'iboe', 'ipoib', 'misc...
Roland Dreier [Tue, 26 Oct 2010 23:09:11 +0000 (16:09 -0700)]
Merge branches 'amso1100', 'cma', 'cxgb3', 'cxgb4', 'ehca', 'iboe', 'ipoib', 'misc', 'mlx4', 'nes', 'qib' and 'srp' into for-next

13 years agoIB/qib: clean up properly if pci_set_consistent_dma_mask() fails
Jason Gunthorpe [Tue, 26 Oct 2010 04:19:06 +0000 (21:19 -0700)]
IB/qib: clean up properly if pci_set_consistent_dma_mask() fails

Clean up properly if pci_set_consistent_dma_mask() fails.

Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/qib: Allow driver to load if PCIe AER fails
Ralph Campbell [Fri, 22 Oct 2010 22:29:54 +0000 (15:29 -0700)]
IB/qib: Allow driver to load if PCIe AER fails

Some PCIe root complex chip sets don't support advanced error reporting.
Allow the driver to load OK if pci_enable_pcie_error_reporting() fails.

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/qib: Fix uninitialized pointer if CONFIG_PCI_MSI not set
Ralph Campbell [Fri, 22 Oct 2010 22:29:46 +0000 (15:29 -0700)]
IB/qib: Fix uninitialized pointer if CONFIG_PCI_MSI not set

If CONFIG_PCI_MSI is not set, and a QLE7140 is present, the pointer
"dd" is uninitialized.

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/qib: Fix extra log level in qib_early_err()
Jason Gunthorpe [Fri, 22 Oct 2010 20:41:24 +0000 (20:41 +0000)]
IB/qib: Fix extra log level in qib_early_err()

Noticed this odd looking thing in dmesg:

    ib_qib 0000:02:00.0: <3>ib_qib: Unable to enable pcie error reporting: -5

which is due to a bad use of dev_info.

Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Acked-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Remove unnecessary KERN_<level> use
Joe Perches [Tue, 26 Oct 2010 02:44:22 +0000 (19:44 -0700)]
RDMA/cxgb4: Remove unnecessary KERN_<level> use

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb3: Remove unnecessary KERN_<level> use
Joe Perches [Tue, 26 Oct 2010 02:44:22 +0000 (19:44 -0700)]
RDMA/cxgb3: Remove unnecessary KERN_<level> use

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/core: Add link layer type information to sysfs
Eli Cohen [Fri, 22 Oct 2010 12:32:05 +0000 (14:32 +0200)]
IB/core: Add link layer type information to sysfs

Since an IB transport port may use either IB or Ethernet as its link layer,
add the file /sys/class/infiniband/<device>/ports/<port_num>/link_layer to
show the link layer for the port.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/mlx4: Add VLAN support for IBoE
Eli Cohen [Thu, 26 Aug 2010 14:19:22 +0000 (17:19 +0300)]
IB/mlx4: Add VLAN support for IBoE

This patch allows IBoE traffic to be encapsulated in 802.1Q tagged
VLAN frames.  The VLAN tag is encoded in the GID and derived from it
by a simple computation.

The netdev notifier callback is modified to catch VLAN device
addition/removal and the port's GID table is updated to reflect the
change, so that for each netdevice there is an entry in the GID table.
When the port's GID table is exhausted, GID entries will not be added.
Only children of the main interfaces can add to the GID table; if a
VLAN interface is added on another VLAN interface (e.g. "vconfig add
eth2.6 8"), then that interfaces will not add an entry to the GID
table.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/core: Add VLAN support for IBoE
Eli Cohen [Thu, 26 Aug 2010 14:18:59 +0000 (17:18 +0300)]
IB/core: Add VLAN support for IBoE

Add 802.1q VLAN support to IBoE. The VLAN tag is encoded within the
GID derived from a link local address in the following way:

    GID[11] GID[12] contain the VLAN ID when the GID contains a VLAN.

The 3 bits user priority field of the packets are identical to the 3
bits of the SL.

In case of rdma_cm apps, the TOS field is used to generate the SL
field by doing a shift right of 5 bits effectively taking to 3 MS bits
of the TOS field.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/mlx4: Add support for IBoE
Eli Cohen [Mon, 25 Oct 2010 04:08:52 +0000 (21:08 -0700)]
IB/mlx4: Add support for IBoE

Add support for IBoE to mlx4_ib.  The bulk of the code is handling the
new address vector fields; mlx4 needs the MAC address of a remote node
to include it in a WQE (for datagrams) or in the QP context (for
connected QPs).  Address resolution is done by assuming all unicast
GIDs are either link-local IPv6 addresses.

Multicast group attach/detach needs to update the NIC's multicast
filters; but since attaching a QP to a multicast group can be done
before the QP is bound to a port, for IBoE we need to keep track of
all multicast groups that a QP is attached too before it transitions
from INIT to RTR (since it does not have a port in the INIT state).

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
[ Many things cleaned up and otherwise monkeyed with; hope I didn't
  introduce too many bugs.  - Roland ]

Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agomlx4_en: Change multicast promiscuous mode to support IBoE
Eli Cohen [Thu, 21 Oct 2010 19:39:08 +0000 (12:39 -0700)]
mlx4_en: Change multicast promiscuous mode to support IBoE

Change multicast promiscuous mode to pass packets through the multicast group distribution table
before sending packets that miss to the default multicast QP.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agomlx4_core: Update data structures and constants for IBoE
Eli Cohen [Thu, 21 Oct 2010 04:57:02 +0000 (21:57 -0700)]
mlx4_core: Update data structures and constants for IBoE

Add fields to hardware data structures and add new constants required for IBoE
support.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agomlx4_core: Allow protocol drivers to find corresponding interfaces
Eli Cohen [Thu, 26 Aug 2010 14:18:43 +0000 (14:18 +0000)]
mlx4_core: Allow protocol drivers to find corresponding interfaces

Add a mechanism for mlx4 protocol drivers to get a pointer to other
drivers's device objects.  For this, an exported function,
mlx4_get_protocol_dev() is added, which allows a driver to get some
other driver's device based on the protocol that the driver
implements.  Two protocols are added: MLX4_PROTOCOL_IB and
MLX4_PROTOCOL_EN.

This will be used in mlx4 IBoE support so that mlx4_ib can find the
corresponding mlx4_en netdev.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
[ Clean up and rename a few things.  - Roland ]

Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/uverbs: Return link layer type to userspace for query port operation
Eli Cohen [Mon, 18 Oct 2010 21:45:20 +0000 (14:45 -0700)]
IB/uverbs: Return link layer type to userspace for query port operation

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/srp: Sync buffer before posting send
David Dillow [Mon, 18 Oct 2010 12:54:49 +0000 (08:54 -0400)]
IB/srp: Sync buffer before posting send

srp_send_tsk_mgmt() was missing the proper DMA sync calls before posting
the buffer to the device.

Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/srp: Use list_first_entry()
Bart Van Assche [Mon, 30 Aug 2010 19:27:40 +0000 (19:27 +0000)]
IB/srp: Use list_first_entry()

Use the list_first_entry() macro in ib_srp instead of open-coding the equivalent,
which makes the source code slightly more descriptive.  The list_first_entry()
macro itself was introduced in kernel 2.6.22.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/srp: Reduce number of BUSY conditions
Bart Van Assche [Mon, 30 Aug 2010 19:27:36 +0000 (19:27 +0000)]
IB/srp: Reduce number of BUSY conditions

As proposed by the SRP (draft) standard, ib_srp reserves one ring
element for SRP_TSK_MGMT requests. This patch makes sure that the SCSI
mid-layer never tries to queue more than (SRP request limit) - 1 SCSI
commands to ib_srp. This improves performance for targets whose request
limit is less than or equal to SRP_NORMAL_REQ_SQ_SIZE by reducing the
number of BUSY responses reported by ib_srp to the SCSI mid-layer.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/srp: Eliminate two forward declarations
David Dillow [Fri, 8 Oct 2010 18:48:14 +0000 (14:48 -0400)]
IB/srp: Eliminate two forward declarations

Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/mlx4: Signal node desc changes to SM by using FW to generate trap 144
Jack Morgenstein [Mon, 4 Oct 2010 12:11:34 +0000 (12:11 +0000)]
IB/mlx4: Signal node desc changes to SM by using FW to generate trap 144

The Node Description cannot be changed via MADs (it is read-only).
Until now, it was changed in the driver via sysfs, and the new Node
Description was simply inserted by the driver into MAD responses
(replacing the description returned by FW).

System startup scripts use the sysfs interface to change the node
description at driver startup to show the hostname, etc. However, this
has a race condition: the SM could discover the original FW node
description rather than the system-specific description if it queried the
port before the startup scripts finish running.

For mlx4, we fix this with a new FW command (SET_NODE) that allows
passing the new node description to FW.  When this command is invoked,
FW sends a trap 144 to the SM.  When it gets this trap, the SM can
query the node to obtain the new node description -- thus eliminating
the effects of the race.

This patch simply calls SET_NODE command when a new node description
is entered via sysfs (thus causing trap 144 to be issued by the FW).
We ignore all failures of the SET_NODE command (including those caused
by using a device FW that predates the SET_NODE command), since in
that case things work just as before.

Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB: Replace EXTRA_CFLAGS with ccflags-y
matt mooney [Fri, 24 Sep 2010 19:17:18 +0000 (19:17 +0000)]
IB: Replace EXTRA_CFLAGS with ccflags-y

Signed-off-by: matt mooney <mfm@muteddisk.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/ucma: Allow tuning the max listen backlog
Steve Wise [Fri, 13 Aug 2010 20:56:34 +0000 (20:56 +0000)]
RDMA/ucma: Allow tuning the max listen backlog

For iWARP connections, the connect request is carried in a TCP payload
on an already established TCP connection.  So if the ucma's backlog is
full, the connection request is transmitted and acked at the TCP level
by the time the connect request gets dropped in the ucma.  The end
result is the connection gets rejected by the iWARP provider.
Further, a 32 node 256NP OpenMPI job will generate > 128 connect
requests on some ranks.

This patch increases the default max backlog to 1024, and adds a
sysctl variable so the backlog can be adjusted at run time.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIPoIB: Set dev_id field of net_device
Eli Cohen [Mon, 20 Sep 2010 07:05:06 +0000 (07:05 +0000)]
IPoIB: Set dev_id field of net_device

Use the net device's dev_id field to encode the port number of the pci
device.  This can be used to to associate a net device with the pci
device's port. The encoding is: dev_id = port - 1.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/srp: Implement SRP_CRED_REQ and SRP_AER_REQ
David Dillow [Fri, 8 Oct 2010 18:40:47 +0000 (14:40 -0400)]
IB/srp: Implement SRP_CRED_REQ and SRP_AER_REQ

This patch adds support for SRP_CRED_REQ to avoid a lockup by targets
that use that mechanism to return credits to the initiator. This
prevents a lockup observed in the field where we would never add the
credits from the SRP_CRED_REQ to our current count, and would therefore
never send another command to the target.

Minimal support for SRP_AER_REQ is also added, as these messages can
also be used to convey additional credits to the initiator.

Based upon extensive debugging and code by Bart Van Assche and a bug
report by Chris Worley.

Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/srp: Preparation for transmit ring response allocation
Bart Van Assche [Mon, 30 Aug 2010 19:27:20 +0000 (19:27 +0000)]
IB/srp: Preparation for transmit ring response allocation

The transmit ring in ib_srp (srp_target.tx_ring) is currently only used
for allocating requests sent by the initiator to the target. This patch
prepares using that ring for allocation of both requests and responses.
Also, this patch differentiates the uses of SRP_SQ_SIZE, increases the
size of the IB send completion queue by one element and reserves one
transmit ring slot for SRP_TSK_MGMT requests.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/qib: Process RDMA WRITE ONLY with IMMEDIATE properly
Jason Gunthorpe [Fri, 22 Oct 2010 22:00:48 +0000 (22:00 +0000)]
IB/qib: Process RDMA WRITE ONLY with IMMEDIATE properly

See table 35 in IBA - the header order for RDMA_WRITE_ONLY_WITH_IMMEDIATE
and SEND_LAST_WITH_IMMEDIATE is different: the RDMA_WRITE_ONLY has
a RETH header before the immediate data, so we need a different code path
to extract the immediate data.

I tested this with a userspace app that does RDMA_WRITE with immediate
on a QLE7140.

Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb3: When a user QP is marked in error, also mark the CQs in error
Steve Wise [Thu, 21 Oct 2010 12:37:06 +0000 (12:37 +0000)]
RDMA/cxgb3: When a user QP is marked in error, also mark the CQs in error

The flushing of work requests for user QPs is implemented entirely in
the user mode library.  The only kernel interaction is to mark the
user QP object indicating it is in error when the QP exits RTS.  When
the user QP operations are called by the application (eg: post_send,
post_recv), the QP in error bit is checked and if set, the library
flushes the QP.  If, however, the application is not doing IO, but
rather just polling the CQ, it will never get flushed work requests.
This breaks some classes of applications.

This patch adds logic to mark user CQs in error when a QP that is bound
to the CQ is marked in error.  The library poll code can then notice
the CQ is in error and flush all the in error QPs bound to that CQ.

Design:

 - add 1 extra CQE entry to the CQ memory that will be used to indicate
   in error status.
 - return the desired CQ memory size that should be mapped by the library
 - bump the ABI since the create_cq uverbs response changes.
 - detect older libraries and reduce the mmap size accordingly.
   (The ABI bump doesn't break old libraries, since they didn't check
   the ABI field anyway)

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Use cxgb4 service for packet gl to skb
Steve Wise [Mon, 18 Oct 2010 15:16:45 +0000 (15:16 +0000)]
RDMA/cxgb4: Use cxgb4 service for packet gl to skb

Remove the local service t4_pktgl_to_skb() and use cxgb4_pktgl_to_skb()
exported by cxgb4.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Export T4 TCP MIB
Steve Wise [Mon, 18 Oct 2010 15:16:40 +0000 (15:16 +0000)]
RDMA/cxgb4: Export T4 TCP MIB

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/pack: IBoE UD packet packing support
Eli Cohen [Thu, 26 Aug 2010 14:17:56 +0000 (14:17 +0000)]
IB/pack: IBoE UD packet packing support

Add support for packing IBoE packet headers.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
[ Clean up and fix ib_ud_header_init() a bit.  - Roland ]

Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cm: Add RDMA CM support for IBoE devices
Eli Cohen [Wed, 13 Oct 2010 19:26:51 +0000 (21:26 +0200)]
RDMA/cm: Add RDMA CM support for IBoE devices

Add support for IBoE device binding and IP --> GID resolution.  Path
resolving and multicast joining are implemented within cma.c by
filling in the responses and running callbacks in the CMA work queue.

IP --> GID resolution always yields IPv6 link local addresses; remote
GIDs are derived from the destination MAC address of the remote port.
Multicast GIDs are always mapped to multicast MACs as is done in IPv6.
(IPv4 multicast is enabled by translating IPv4 multicast addresses to
IPv6 multicast as described in
<http://www.mail-archive.com/ipng@sunroof.eng.sun.com/msg02134.html>.)

Some helper functions are added to ib_addr.h.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/mad: IBoE supports only QP1 (no QP0)
Eli Cohen [Tue, 28 Sep 2010 00:51:11 +0000 (17:51 -0700)]
IB/mad: IBoE supports only QP1 (no QP0)

Since IBoE is using Ethernet as its link layer, there is no central
management entity so there is need for QP0.  QP1 is still needed since
it handles communications between CM agents.  This patch will skip QP0
and create only QP1 for IBoE ports.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIPoIB: Skip IBoE ports
Eli Cohen [Tue, 28 Sep 2010 00:51:11 +0000 (17:51 -0700)]
IPoIB: Skip IBoE ports

IPoIB is IP-over-Infiniband link layer. In the case of IBoE, the link
layer is Ethernet and IP can work directly over Ethernet, so disable
IPoIB for non-IB_LINK_LAYER_INFINIBAND ports.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/iwcm: Fix hang in uninterruptible wait on cm_id destroy
Animesh K Trivedi [Tue, 28 Sep 2010 14:44:02 +0000 (14:44 +0000)]
RDMA/iwcm: Fix hang in uninterruptible wait on cm_id destroy

A process can get stuck in an uninterruptible wait in the
kernel while destroying a cm_id when iw_cm_connect() fails:

For example, When creation of a PD fails but the user continues with
an attempt to connect to the server without checking the return value,
in iw_cm_connect() a NULL qp is found so the call fails.  However the
IWCM_F_CONNECT_WAIT bit is not cleared.  destroy_cm_id() then waits
forever for IWCM_F_CONNECT_WAIT to be cleared.

The same problem exists on the passive side with the accept call.

Fix this by clearing the bit and waking up any waiters in the
appropriate spots.

Signed-off-by: Animesh Trivedi <atr@zurich.ibm.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Use simple_read_from_buffer() for debugfs handlers
Steve Wise [Wed, 29 Sep 2010 18:21:33 +0000 (18:21 +0000)]
RDMA/cxgb4: Use simple_read_from_buffer() for debugfs handlers

We can replace our equivalent open-coded version.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Add default_llseek to debugfs files
Steve Wise [Wed, 29 Sep 2010 14:11:12 +0000 (14:11 +0000)]
RDMA/cxgb4: Add default_llseek to debugfs files

Incorporate BKL removal changes.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/mlx4: Limit size of fast registration WRs
Eli Cohen [Thu, 7 Oct 2010 14:24:16 +0000 (16:24 +0200)]
IB/mlx4: Limit size of fast registration WRs

Fix the limit on the size of max fast registration WRs that can be
posted to match hardware capabilities.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoMAINTAINERS: Hand off SCSI RDMA Protocol (SRP) initiator to Dave Dillow
Roland Dreier [Thu, 7 Oct 2010 16:54:53 +0000 (09:54 -0700)]
MAINTAINERS: Hand off SCSI RDMA Protocol (SRP) initiator to Dave Dillow

Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/qib: Remove unnecessary casts of private_data
Joe Perches [Sun, 5 Sep 2010 01:52:43 +0000 (01:52 +0000)]
IB/qib: Remove unnecessary casts of private_data

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/nes: Turn carrier off on ifdown
Maciej Sosnowski [Wed, 1 Sep 2010 13:35:35 +0000 (13:35 +0000)]
RDMA/nes: Turn carrier off on ifdown

This lets the bonding driver to detect when an interface goes down.

Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Chien Tung <chien.tin.tung@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/nes: Report correct port state if interface is down
Chien Tung [Mon, 23 Aug 2010 14:32:36 +0000 (14:32 +0000)]
RDMA/nes: Report correct port state if interface is down

With commit cd6860eb ("RDMA/nes: Fix hangs on ifdown") we no longer
remove nes interfaces on ifdown.  On nes_query_port(), add an
additional check of the netdev queue and report IB_PORT_DOWN if the
queue is not running.

Signed-off-by: Chien Tung <chien.tin.tung@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/ehca: Fix driver on relocatable kernel
Sonny Rao [Fri, 20 Aug 2010 04:10:19 +0000 (04:10 +0000)]
IB/ehca: Fix driver on relocatable kernel

the eHCA driver registers a MR for all of kernel memory, but makes the
assumption that valid memory exists at KERNELBASE.  This assumption
may not be true in the case of a relocatable kernel, so use KERNELBASE
+ PHYSICAL_START to get the true beginning of usable kernel memory.

cc: Joachim Fenkes <fenkes@de.ibm.com>
cc: Christoph Raisch <raisch@de.ibm.com>
cc: Hoan-Ham Hguyen <hnguyen@de.ibm.com>
Signed-off-by: Sonny Rao <sonnyrao@us.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/umad: Make user_mad semaphore a real one
Thomas Gleixner [Tue, 7 Sep 2010 14:33:31 +0000 (14:33 +0000)]
IB/umad: Make user_mad semaphore a real one

Get rid of init_MUTEX[_LOCKED]() and use sema_init() instead.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/amso1100: Remove KERN_<level> from pr_<level> use
Joe Perches [Sun, 12 Sep 2010 05:10:55 +0000 (22:10 -0700)]
RDMA/amso1100: Remove KERN_<level> from pr_<level> use

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/nes: Remove unneeded variable
Dan Carpenter [Wed, 15 Sep 2010 16:32:39 +0000 (18:32 +0200)]
RDMA/nes: Remove unneeded variable

Just a small cleanup.  The "passive_state" variable isn't used any
more after commit dae58728dc ("RDMA/nes: Fix double CLOSE event
indication crash")

Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: Faisal Latif <faisal.latif@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agomlx4_core: Allow log_mtts_per_seg up to 7
Eli Cohen [Mon, 20 Sep 2010 06:42:38 +0000 (08:42 +0200)]
mlx4_core: Allow log_mtts_per_seg up to 7

As there are now machines with a lot more physical memory, we need to
be able to register more memory. This patch lifts the upper limit of
log_mtts_per_seg from 5 to 7, increasing the amount of memory that can
be registered.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIPoIB: Set pkt_type correctly for multicast packets (fix IGMP breakage)
Christoph Lameter [Fri, 27 Aug 2010 13:29:38 +0000 (08:29 -0500)]
IPoIB: Set pkt_type correctly for multicast packets (fix IGMP breakage)

IGMP processing is broken because the IPOIB does not set the
skb->pkt_type the right way for multicast traffic.  All incoming
packets are set to PACKET_HOST which means that igmp_recv() will
ignore the IGMP broadcasts/multicasts.

This in turn means that the IGMP timers are firing and are sending
information about multicast subscriptions unnecessarily.  In a large
private network this can cause traffic spikes.

Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Fastreg NSMR fixes
Steve Wise [Fri, 17 Sep 2010 20:40:15 +0000 (15:40 -0500)]
RDMA/cxgb4: Fastreg NSMR fixes

- Remove dsgl support - doesn't work in T4.
- Wrap the immediate PBL as needed when building it in the wr.
- Adjust max pbl depth allowed based on ulptx alignment requirements.
- Bump the slots per SQ to 5 to allow up to 128MB fast registers.
- Advertise fastreg support by default.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Don't set completion flag for read requests
Steve Wise [Fri, 17 Sep 2010 20:40:09 +0000 (15:40 -0500)]
RDMA/cxgb4: Don't set completion flag for read requests

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Set the default TCP send window to 128KB
Steve Wise [Fri, 10 Sep 2010 16:15:41 +0000 (11:15 -0500)]
RDMA/cxgb4: Set the default TCP send window to 128KB

This helps with large IO throughput.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Use a mutex for QP and EP state transitions
Steve Wise [Fri, 10 Sep 2010 16:15:36 +0000 (11:15 -0500)]
RDMA/cxgb4: Use a mutex for QP and EP state transitions

Move the connection setup/teardown paths to the workq thread removing
spin lock/irq disable requirements for these paths.  This allows calls
down to the LLD for EP and QP state transition actions to be atomic
with respect to processing CPL messages coming up from the HW.
Namely, calls to rdma_init() and rdma_fini() can now be called with
the mutex held avoiding many race conditions with the abort path.

The QP spinlock is still used but only to manipulate the qp state.  This
allows the fastpaths, poll, post_send, and pos_recv, to run in the
irq context.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Support on-chip SQs
Steve Wise [Mon, 13 Sep 2010 16:23:57 +0000 (11:23 -0500)]
RDMA/cxgb4: Support on-chip SQs

T4 support on-chip SQs to reduce latency.  This patch adds support for
this in iw_cxgb4:

 - Manage ocqp memory like other adapter mem resources.
 - Allocate user mode SQs from ocqp mem if available.
 - Map ocqp mem to user process using write combining.
 - Map PCIE_MA_SYNC reg to user process.

Bump uverbs ABI.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Centralize the wait logic
Steve Wise [Fri, 10 Sep 2010 16:15:25 +0000 (11:15 -0500)]
RDMA/cxgb4: Centralize the wait logic

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: debugfs files for dumping active stags
Steve Wise [Fri, 10 Sep 2010 16:15:20 +0000 (11:15 -0500)]
RDMA/cxgb4: debugfs files for dumping active stags

Add "stags" debugfs file.  This is useful for examining the TPTE and
PBL entries in adapter memory.  It allows scripts to dump just the
active entries.

Also clean up the "qps" file handlers and shared common code.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Log HW lack-of-resource errors
Steve Wise [Fri, 10 Sep 2010 16:15:14 +0000 (11:15 -0500)]
RDMA/cxgb4: Log HW lack-of-resource errors

This helps debug cases where HW resources are depleted.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Handle CPL_RDMA_TERMINATE messages
Steve Wise [Fri, 10 Sep 2010 16:15:09 +0000 (11:15 -0500)]
RDMA/cxgb4: Handle CPL_RDMA_TERMINATE messages

T4 FW sends up CPL_RDMA_TERMINATE to indicate a peer TERM.  This
triggers the QP moving to TERMINATE state.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Ignore TERMINATE CQEs
Steve Wise [Fri, 10 Sep 2010 16:15:04 +0000 (11:15 -0500)]
RDMA/cxgb4: Ignore TERMINATE CQEs

T4 incorrectly inserts TERM CQEs into the CQ.  Silently ignore them.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Ignore positive return values from cxgb4_*_send() functions
Steve Wise [Fri, 10 Sep 2010 16:14:58 +0000 (11:14 -0500)]
RDMA/cxgb4: Ignore positive return values from cxgb4_*_send() functions

The cxgb4_*_send() functions return NET_XMIT_ values, which are
positive integers or negative errno values.  So don't treat positive
return values as an error.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Zero out ISGL padding
Steve Wise [Fri, 10 Sep 2010 16:14:53 +0000 (11:14 -0500)]
RDMA/cxgb4: Zero out ISGL padding

The HW design requires zeroing any pad in SGLs.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Don't use null ep ptr
Steve Wise [Fri, 10 Sep 2010 16:14:48 +0000 (11:14 -0500)]
RDMA/cxgb4: Don't use null ep ptr

In c4iw_modify_qp() error path, only use qhp->ep if ep is not already set.
Otherwise qhp->ep can be NULL and we crash.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/nes: Fix cast-to-pointer warnings on 32-bit
Roland Dreier [Tue, 28 Sep 2010 00:51:33 +0000 (17:51 -0700)]
RDMA/nes: Fix cast-to-pointer warnings on 32-bit

Fix:

  drivers/infiniband/hw/nes/nes_verbs.c: In function 'nes_alloc_fast_reg_page_list':
  drivers/infiniband/hw/nes/nes_verbs.c:477: warning: cast to pointer from integer of different size
  drivers/infiniband/hw/nes/nes_verbs.c: In function 'nes_post_send':
  drivers/infiniband/hw/nes/nes_verbs.c:3486: warning: cast to pointer from integer of different size
  drivers/infiniband/hw/nes/nes_verbs.c:3486: warning: cast to pointer from integer of different size

by printing u64 quantities by casting to unsigned long and long and
using %llx, rather than casting to void* and using %p.

Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoMAINTAINERS: Fix broken link to http://www.openib.org/
Roland Dreier [Tue, 28 Sep 2010 00:51:24 +0000 (17:51 -0700)]
MAINTAINERS: Fix broken link to www.openib.org/

openib.org is now openfabrics.org.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoIB/core: Add link layer property to ports
Eli Cohen [Tue, 28 Sep 2010 00:51:10 +0000 (17:51 -0700)]
IB/core: Add link layer property to ports

This patch allows ports to have different link layers:
IB_LINK_LAYER_INFINIBAND or IB_LINK_LAYER_ETHERNET.  This is required
for adding IBoE (InfiniBand-over-Ethernet, aka RoCE) support.  For
devices that do not provide an implementation for querying the link
layer property of a port, we return a default value based on the
transport: RMA_TRANSPORT_IB nodes will return IB_LINK_LAYER_INFINIBAND
and RDMA_TRANSPORT_IWARP nodes will return IB_LINK_LAYER_ETHERNET.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoRDMA/cxgb4: Fix warnings about casts to/from pointers of different sizes
Roland Dreier [Tue, 28 Sep 2010 00:51:04 +0000 (17:51 -0700)]
RDMA/cxgb4: Fix warnings about casts to/from pointers of different sizes

Fix:

  drivers/infiniband/hw/cxgb4/qp.c: In function â€˜create_qp’:
  drivers/infiniband/hw/cxgb4/qp.c:147: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/qp.c: In function â€˜rdma_fini’:
  drivers/infiniband/hw/cxgb4/qp.c:988: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/qp.c: In function â€˜rdma_init’:
  drivers/infiniband/hw/cxgb4/qp.c:1063: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/mem.c: In function â€˜write_adapter_mem’:
  drivers/infiniband/hw/cxgb4/mem.c:74: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/cq.c: In function â€˜destroy_cq’:
  drivers/infiniband/hw/cxgb4/cq.c:58: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/cq.c: In function â€˜create_cq’:
  drivers/infiniband/hw/cxgb4/cq.c:135: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/cm.c: In function â€˜fw6_msg’:
  drivers/infiniband/hw/cxgb4/cm.c:2326: warning: cast to pointer from integer of different size

by casting pointers to unsigned long instead of u64.

Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoFix pktcdvd ioctl dev_minor range check
Dan Rosenberg [Mon, 27 Sep 2010 16:30:28 +0000 (12:30 -0400)]
Fix pktcdvd ioctl dev_minor range check

The PKT_CTRL_CMD_STATUS device ioctl retrieves a pointer to a
pktcdvd_device from the global pkt_devs array.  The index into this
array is provided directly by the user and is a signed integer, so the
comparison to ensure that it falls within the bounds of this array will
fail when provided with a negative index.

This can be used to read arbitrary kernel memory or cause a crash due to
an invalid pointer dereference.  This can be exploited by users with
permission to open /dev/pktcdvd/control (on many distributions, this is
readable by group "cdrom").

Signed-off-by: Dan Rosenberg <dan.j.rosenberg@gmail.com>
[ Rather than add a cast, just make the function take the right type -Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMN10300: Default config choice GDBSTUB_TTYSM0 should be GDBSTUB_ON_TTYSM0
David Howells [Mon, 27 Sep 2010 12:12:33 +0000 (13:12 +0100)]
MN10300: Default config choice GDBSTUB_TTYSM0 should be GDBSTUB_ON_TTYSM0

The configuration choice for the port on which the GDB stub listens has
a default of GDBSTUB_TTYSM0, but this should be GDBSTUB_ON_TTYSM0 to
match the option.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Mon, 27 Sep 2010 19:33:54 +0000 (12:33 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  RDMA/cxgb3: Turn off RX coalescing for iWARP connections

13 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Mon, 27 Sep 2010 19:32:36 +0000 (12:32 -0700)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm: (28 commits)
  ARM: 6411/1: vexpress: set RAM latencies to 1 cycle for PL310 on ct-ca9x4 tile
  ARM: 6409/1: davinci: map sram using MT_MEMORY_NONCACHED instead of MT_DEVICE
  ARM: 6408/1: omap: Map only available sram memory
  ARM: 6407/1: mmu: Setup MT_MEMORY and MT_MEMORY_NONCACHED L1 entries
  ARM: pxa: remove pr_<level> uses of KERN_<level>
  ARM: pxa168fb: clear enable bit when not active
  ARM: pxa: fix cpu_is_pxa*() not expanding to zero when not configured
  ARM: pxa168: fix corrected reset vector
  ARM: pxa: Use PIO for PI2C communication on Palm27x
  ARM: pxa: Fix Vpac270 gpio_power for MMC
  ARM: 6401/1: plug a race in the alignment trap handler
  ARM: 6406/1: at91sam9g45: fix i2c bus speed
  leds: leds-ns2: fix locking
  ARM: dove: fix __io() definition to use bus based offset
  dmaengine: fix interrupt clearing for mv_xor
  ARM: kirkwood: Unbreak PCIe I/O port
  ARM: Fix build error when using KCONFIG_CONFIG
  ARM: 6383/1: Implement phys_mem_access_prot() to avoid attributes aliasing
  ARM: 6400/1: at91: fix arch_gettimeoffset fallout
  ARM: 6398/1: add proc info for ARM11MPCore/Cortex-A9 from ARM
  ...

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh...
Linus Torvalds [Mon, 27 Sep 2010 19:32:00 +0000 (12:32 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/ericvh/v9fs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs:
  net/9p: fix memory handling/allocation in rdma_request()

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp
Linus Torvalds [Mon, 27 Sep 2010 19:31:12 +0000 (12:31 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/bp/bp

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp:
  amd64_edac: Fix driver module removal

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Mon, 27 Sep 2010 19:29:39 +0000 (12:29 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jmorris/security-testing-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/security-testing-2.6:
  TOMOYO: Don't abuse sys_getpid(), sys_getppid()

13 years agoMerge branch 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ickle...
Linus Torvalds [Mon, 27 Sep 2010 19:28:19 +0000 (12:28 -0700)]
Merge branch 'drm-intel-fixes' of git://git./linux/kernel/git/ickle/drm-intel

* 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ickle/drm-intel:
  drm/i915/sdvo: Handle unsupported GET_SUPPORTED_ENHANCEMENTS gracefully
  drm/i915/sdvo: Cleanup connector on error path
  drm/i915: Fix 945GM regression in e259befd

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc
Linus Torvalds [Mon, 27 Sep 2010 19:27:00 +0000 (12:27 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/cjb/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc:
  mmc: sdhci-s3c: fix NULL ptr access in sdhci_s3c_remove
  mmc: sdhci-s3c: fix incorrect spinlock usage after merge
  mmc: MAINTAINERS: add myself as MMC maintainer

13 years agoMerge branch 'urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/brodo/pcmcia-2.6
Linus Torvalds [Mon, 27 Sep 2010 19:26:33 +0000 (12:26 -0700)]
Merge branch 'urgent' of git://git./linux/kernel/git/brodo/pcmcia-2.6

* 'urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/brodo/pcmcia-2.6:
  pcmcia: pd6729: Fix error path
  pcmcia: preserve configuration information if request_io fails partly

13 years agoMerge git://git.infradead.org/iommu-2.6
Linus Torvalds [Mon, 27 Sep 2010 19:25:10 +0000 (12:25 -0700)]
Merge git://git.infradead.org/iommu-2.6

* git://git.infradead.org/iommu-2.6:
  intel-iommu: Use symbolic values instead of magic numbers in Lenovo w/a
  intel-iommu: Abort IOMMU setup for igfx if BIOS gave no shadow GTT space

13 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 27 Sep 2010 19:22:21 +0000 (12:22 -0700)]
Merge branch 'x86-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86/amd-iommu: Fix rounding-bug in __unmap_single
  x86/amd-iommu: Work around S3 BIOS bug
  x86/amd-iommu: Set iommu configuration flags in enable-loop
  x86, setup: Fix earlyprintk=serial,0x3f8,115200
  x86, setup: Fix earlyprintk=serial,ttyS0,115200

13 years agoMerge branch 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 27 Sep 2010 19:21:48 +0000 (12:21 -0700)]
Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  perf, x86: Catch spurious interrupts after disabling counters
  tracing/x86: Don't use mcount in kvmclock.c
  tracing/x86: Don't use mcount in pvclock.c

13 years agomn10300: check __get_user/__put_user results...
Al Viro [Sun, 26 Sep 2010 18:29:12 +0000 (19:29 +0100)]
mn10300: check __get_user/__put_user results...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agomn10300: get rid of set_fs(USER_DS) in sigframe setup
Al Viro [Sun, 26 Sep 2010 18:29:02 +0000 (19:29 +0100)]
mn10300: get rid of set_fs(USER_DS) in sigframe setup

It really has no business being there; short of a serious kernel bug
we should already have USER_DS at that point.  It shouldn't have been
done on x86 either...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agomn10300: ->restart_block.fn needs to be reset on sigreturn
Al Viro [Sun, 26 Sep 2010 18:28:52 +0000 (19:28 +0100)]
mn10300: ->restart_block.fn needs to be reset on sigreturn

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agomn10300: prevent double syscall restarts
Al Viro [Sun, 26 Sep 2010 18:28:42 +0000 (19:28 +0100)]
mn10300: prevent double syscall restarts

set ->orig_d0 to -1, same as what sigreturn does

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agomn10300: avoid SIGSEGV delivery loop
Al Viro [Sun, 26 Sep 2010 18:28:32 +0000 (19:28 +0100)]
mn10300: avoid SIGSEGV delivery loop

force_sigsegv() is there for purpose...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoalpha: __get_user/__put_user results need to be checked...
Al Viro [Sun, 26 Sep 2010 18:28:22 +0000 (19:28 +0100)]
alpha: __get_user/__put_user results need to be checked...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoalpha: switch osf_sigprocmask() to use of sigprocmask()
Al Viro [Sun, 26 Sep 2010 18:28:12 +0000 (19:28 +0100)]
alpha: switch osf_sigprocmask() to use of sigprocmask()

get rid of a useless wrapper, while we are at it

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoRDMA/cxgb3: Turn off RX coalescing for iWARP connections
Steve Wise [Sun, 19 Sep 2010 00:38:21 +0000 (19:38 -0500)]
RDMA/cxgb3: Turn off RX coalescing for iWARP connections

The HW by default has RX coalescing on.  For iWARP connections, this
causes a 100ms delay in connection establishement due to the ingress
MPA Start message being stalled in HW.  So explicitly turn RX
coalescing off when setting up iWARP connections.

This was causing very bad performance for NP64 gather operations using
Open MPI, due to the way it sets up connections on larger jobs.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Cc: <stable@kernel.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
13 years agoARM: 6411/1: vexpress: set RAM latencies to 1 cycle for PL310 on ct-ca9x4 tile
Will Deacon [Mon, 27 Sep 2010 13:55:15 +0000 (14:55 +0100)]
ARM: 6411/1: vexpress: set RAM latencies to 1 cycle for PL310 on ct-ca9x4 tile

The PL310 on the ct-ca9x4 tile for the Versatile Express does not need
to add additional latency when accessing its cache RAMs. Unfortunately,
the boot monitor sets this up for an 8-cycle delay on reads and writes,
resulting in greatly reduced memory performance when the L2 cache is
enabled.

This patch sets the L2 RAM latencies to the correct value of 1 cycle
on the ct-ca9x4 tile before enabling the L2 cache.

Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
13 years agonet/9p: fix memory handling/allocation in rdma_request()
Davidlohr Bueso [Mon, 13 Sep 2010 15:53:18 +0000 (15:53 +0000)]
net/9p: fix memory handling/allocation in rdma_request()

Return -ENOMEM when erroring on kmalloc and fix memory leaks when returning on error.

Signed-off-by: Davidlohr Bueso <dave@gnu.org>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
13 years agoamd64_edac: Fix driver module removal
Borislav Petkov [Sun, 26 Sep 2010 10:42:23 +0000 (12:42 +0200)]
amd64_edac: Fix driver module removal

f4347553b30ec66530bfe63c84530afea3803396 removed the edac polling
mechanism in favor of using a notifier chain for conveying MCE
information to edac. However, the module removal path didn't test
whether the driver had setup the polling function workqueue at all and
the rmmod process was hanging in the kernel at try_to_del_timer_sync()
in the cancel_delayed_work() path, trying to cancel an uninitialized
work struct.

Fix that by adding a balancing check to the workqueue removal path.

Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
13 years agoTOMOYO: Don't abuse sys_getpid(), sys_getppid()
Ben Hutchings [Sun, 26 Sep 2010 04:55:13 +0000 (05:55 +0100)]
TOMOYO: Don't abuse sys_getpid(), sys_getppid()

System call entry functions sys_*() are never to be called from
general kernel code.  The fact that they aren't declared in header
files should have been a clue.  These functions also don't exist on
Alpha since it has sys_getxpid() instead.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Acked-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: James Morris <jmorris@namei.org>
13 years agommc: sdhci-s3c: fix NULL ptr access in sdhci_s3c_remove
Marek Szyprowski [Thu, 23 Sep 2010 14:22:05 +0000 (16:22 +0200)]
mmc: sdhci-s3c: fix NULL ptr access in sdhci_s3c_remove

If not all clocks have been defined in platform data, the driver will
cause a null pointer dereference when it is removed. This patch fixes
this issue.

Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
13 years agommc: sdhci-s3c: fix incorrect spinlock usage after merge
Marek Szyprowski [Mon, 20 Sep 2010 13:03:42 +0000 (15:03 +0200)]
mmc: sdhci-s3c: fix incorrect spinlock usage after merge

In the commit f522886e202a34a2191dd5d471b3c4d46410a9a0 a merge conflict
in the sdhci-s3c driver been fixed. However the fix used incorrect
spinlock operation - it caused a race with sdhci interrupt service. The
correct way to solve it is to use spin_lock_irqsave/irqrestore() calls.

Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
13 years agommc: MAINTAINERS: add myself as MMC maintainer
Chris Ball [Fri, 10 Sep 2010 16:05:24 +0000 (12:05 -0400)]
mmc: MAINTAINERS: add myself as MMC maintainer

Signed-off-by: Chris Ball <cjb@laptop.org>
Cc: Pierre Ossman <pierre-list@ossman.eu>
Cc: <linux-mmc@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
13 years agopcmcia: pd6729: Fix error path
Rahul Ruikar [Sun, 26 Sep 2010 11:30:29 +0000 (17:00 +0530)]
pcmcia: pd6729: Fix error path

In error return path
call pci_disable_device() which was enabled earlier.

Signed-off-by: Rahul Ruikar <rahul.ruikar@gmail.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
13 years agoalpha: fix usp value in multithreaded coredumps
Al Viro [Sat, 25 Sep 2010 20:07:51 +0000 (21:07 +0100)]
alpha: fix usp value in multithreaded coredumps

rdusp() gives us the right value only for the current thread...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoalpha: fix hae_cache race in RESTORE_ALL
Al Viro [Sat, 25 Sep 2010 20:07:14 +0000 (21:07 +0100)]
alpha: fix hae_cache race in RESTORE_ALL

We want interrupts disabled on all paths leading to RESTORE_ALL;
otherwise, we are risking an IRQ coming between the updates of
alpha_mv->hae_cache and *alpha_mv->hae_register and set_hae()
within the IRQ getting badly confused.

RESTORE_ALL used to play with disabling IRQ itself, but that got
removed back in 2002, without making sure we had them disabled
on all paths.  It's cheaper to make sure we have them disabled than
to revert to original variant...

Remove the detritus left from that commit back in 2002; we used to
need a reload of $0 and $1 since swpipl would change those, but
doing that had become pointless when we stopped doing swpipl in
there...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Sat, 25 Sep 2010 16:51:54 +0000 (09:51 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  block: prevent merges of discard and write requests

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Sat, 25 Sep 2010 16:51:31 +0000 (09:51 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ALSA: sound/pci/rme9652: prevent reading uninitialized stack memory
  ALSA: hda - Fix auto-parse of SPDIF input of Realtek codecs
  ASoC: Fix multi-componentism
  ASoC: Fix soc-cache buffer overflow bug
  ALSA: oxygen: fix analog capture on Claro halo cards
  ALSA: hda - Add Dell Latitude E6400 model quirk
  ASoC: fix clkdev API usage in sh/migor.c

13 years agoAvoid pgoff overflow in remap_file_pages
Larry Woodman [Fri, 24 Sep 2010 16:04:48 +0000 (12:04 -0400)]
Avoid pgoff overflow in remap_file_pages

Thomas Pollet noticed that the remap_file_pages() system call in
fremap.c has a potential overflow in the first part of the if statement
below, which could cause it to process bogus input parameters.
Specifically the pgoff + size parameters could be wrap thereby
preventing the system call from failing when it should.

Reported-by: Thomas Pollet <thomas.pollet@gmail.com>
Signed-off-by: Larry Woodman <lwoodman@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'fix/hda' into for-linus
Takashi Iwai [Sat, 25 Sep 2010 15:57:53 +0000 (17:57 +0200)]
Merge branch 'fix/hda' into for-linus

13 years agoMerge branch 'fix/asoc' into for-linus
Takashi Iwai [Sat, 25 Sep 2010 15:57:49 +0000 (17:57 +0200)]
Merge branch 'fix/asoc' into for-linus