deliverable/linux.git
8 years agoIB: remove the struct ib_phys_buf definition
Christoph Hellwig [Wed, 23 Dec 2015 18:12:53 +0000 (19:12 +0100)] 
IB: remove the struct ib_phys_buf definition

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> [core]
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoehca: stop using struct ib_phys_buf
Christoph Hellwig [Wed, 23 Dec 2015 18:12:52 +0000 (19:12 +0100)] 
ehca: stop using struct ib_phys_buf

And simplify the calling convention for full-memory registrations.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> [core]
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoamso1100: fold c2_reg_phys_mr into c2_get_dma_mr
Christoph Hellwig [Wed, 23 Dec 2015 18:12:51 +0000 (19:12 +0100)] 
amso1100: fold c2_reg_phys_mr into c2_get_dma_mr

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> [core]
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agones: simplify nes_reg_phys_mr calling conventions
Christoph Hellwig [Wed, 23 Dec 2015 18:12:50 +0000 (19:12 +0100)] 
nes: simplify nes_reg_phys_mr calling conventions

Just pass and address/size pair instead of an ib_phys_buf array.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> [core]
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agocxgb3: simplify iwch_get_dma_wr
Christoph Hellwig [Wed, 23 Dec 2015 18:12:49 +0000 (19:12 +0100)] 
cxgb3: simplify iwch_get_dma_wr

Fold simplified versions of build_phys_page_list and
iwch_register_phys_mem into iwch_get_dma_wr now that no other callers
are left.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> [core]
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB: remove in-kernel support for memory windows
Christoph Hellwig [Wed, 23 Dec 2015 18:12:48 +0000 (19:12 +0100)] 
IB: remove in-kernel support for memory windows

Remove the unused ib_allow_mw and ib_bind_mw functions, remove the
unused IB_WR_BIND_MW and IB_WC_BIND_MW opcodes and move ib_dealloc_mw
into the uverbs module.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> [core]
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB: remove support for phys MRs
Christoph Hellwig [Wed, 23 Dec 2015 18:12:47 +0000 (19:12 +0100)] 
IB: remove support for phys MRs

We have stopped using phys MRs in the kernel a while ago, so let's
remove all the cruft used to implement them.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> [core]
Reviewed-By: Devesh Sharma<devesh.sharma@avagotech.com> [ocrdma]
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB: remove ib_query_mr
Christoph Hellwig [Wed, 23 Dec 2015 18:12:46 +0000 (19:12 +0100)] 
IB: remove ib_query_mr

This functionality has no users and was only supported by the staged out
EHCA driver.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> [core]
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB: start documenting device capabilities
Christoph Hellwig [Wed, 23 Dec 2015 18:12:45 +0000 (19:12 +0100)] 
IB: start documenting device capabilities

Just IB_DEVICE_LOCAL_DMA_LKEY and IB_DEVICE_MEM_MGT_EXTENSIONS for now
as I'm most familar with those.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-By: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/IPoIB: Move multicast specific code out of ipoib_main.c
Christoph Lameter [Mon, 21 Dec 2015 14:42:54 +0000 (08:42 -0600)] 
IB/IPoIB: Move multicast specific code out of ipoib_main.c

Code cleanup to move multicast specific code that checks for
a sendonly join to ipoib_multicast.c. This allows the removal
of the export of __ipoib_mcast_find().

Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/IPoIB: factor out common multicast list removal code
Christoph Lameter [Mon, 21 Dec 2015 14:42:53 +0000 (08:42 -0600)] 
IB/IPoIB: factor out common multicast list removal code

Code cleanup to remove multicast specific code from ipoib_main.c

The removal of a list of multicast groups occurs in three places.
Create a new function ipoib_mcast_remove_list(). Use this new
function in ipoib_main.c too.
That in turn allows the dropping of two functions that were
exported from ipoib_multicast.c for expiration of mc groups.

Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/mlx5: Support RoCE
Achiad Shochat [Wed, 23 Dec 2015 16:47:25 +0000 (18:47 +0200)] 
IB/mlx5: Support RoCE

Advertise RoCE support for IB/core layer and set the hardware to
work in RoCE mode.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/mlx5: Add RoCE fields to Address Vector
Achiad Shochat [Wed, 23 Dec 2015 16:47:24 +0000 (18:47 +0200)] 
IB/mlx5: Add RoCE fields to Address Vector

Set the address handle and QP address path fields according to the
link layer type (IB/Eth).

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/mlx5: Support IB device's callbacks for adding/deleting GIDs
Achiad Shochat [Wed, 23 Dec 2015 16:47:23 +0000 (18:47 +0200)] 
IB/mlx5: Support IB device's callbacks for adding/deleting GIDs

These callbacks write into the mlx5 RoCE address table.
Upon del_gid we write a zero'd GID.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/mlx5: Set network_hdr_type upon RoCE responder completion
Achiad Shochat [Wed, 23 Dec 2015 16:47:22 +0000 (18:47 +0200)] 
IB/mlx5: Set network_hdr_type upon RoCE responder completion

When handling a responder completion, if the link layer is Ethernet,
set the work completion network_hdr_type field according to CQE's
info and the IB_WC_WITH_NETWORK_HDR_TYPE flag.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/mlx5: Extend query_device/port to support RoCE
Achiad Shochat [Wed, 23 Dec 2015 16:47:21 +0000 (18:47 +0200)] 
IB/mlx5: Extend query_device/port to support RoCE

Using the vport access functions to retrieve the Ethernet
specific information and return this information in
ib_query_device and ib_query_port.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agonet/mlx5_core: Introduce access functions to query vport RoCE fields
Achiad Shochat [Wed, 23 Dec 2015 16:47:20 +0000 (18:47 +0200)] 
net/mlx5_core: Introduce access functions to query vport RoCE fields

Introduce access functions to query NIC vport system_image_guid,
node_guid and qkey_viol_cntr.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agonet/mlx5_core: Introduce access functions to enable/disable RoCE
Achiad Shochat [Wed, 23 Dec 2015 16:47:19 +0000 (18:47 +0200)] 
net/mlx5_core: Introduce access functions to enable/disable RoCE

A mlx5 Ethernet port must be explicitly enabled for RoCE.
When RoCE is not enabled on the port, the NIC will refuse to create
QPs attached to it and incoming RoCE packets will be considered by the
NIC as plain Ethernet packets.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agonet/mlx5_core: Break down the vport mac address query function
Achiad Shochat [Wed, 23 Dec 2015 16:47:18 +0000 (18:47 +0200)] 
net/mlx5_core: Break down the vport mac address query function

Introduce a new function called mlx5_query_nic_vport_context().
This function gets all the NIC vport attributes from the device.

The MAC address is just one of the NIC vport attributes, so
mlx5_query_nic_vport_mac_address() is now just a wrapper function
above mlx5_query_nic_vport_context().

More NIC vport attributes will be used in following commits.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/mlx5: Support IB device's callback for getting its netdev
Achiad Shochat [Wed, 23 Dec 2015 16:47:17 +0000 (18:47 +0200)] 
IB/mlx5: Support IB device's callback for getting its netdev

For Eth ports only:
Maintain a net device pointer in mlx5_ib_device and update it
upon NETDEV_REGISTER and NETDEV_UNREGISTER events if the
net-device and IB device have the same PCI parent device.
Implement the get_netdev callback to return this net device.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/mlx5: Support IB device's callback for getting the link layer
Achiad Shochat [Wed, 23 Dec 2015 16:47:16 +0000 (18:47 +0200)] 
IB/mlx5: Support IB device's callback for getting the link layer

Make the existing mlx5_ib_port_link_layer() signature match
the ib device callback signature (add port_num parameter).
Refactor it to use a sub function so that the link layer could
be queried also before the ibdev is created.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/usnic: delete unneeded IS_ERR test
Julia Lawall [Sat, 19 Dec 2015 20:48:59 +0000 (21:48 +0100)] 
IB/usnic: delete unneeded IS_ERR test

kzalloc doesn't return ERR_PTR, so there is no need to test for it.

The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression x,e;
@@

* x = kzalloc(...)
... when != x = e
* IS_ERR_OR_NULL(x)
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Reviewed-by: Dave Goodell <dgoodell@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/usnic: Handle 0 counts in resource allocation
Nelson Escobar [Wed, 9 Dec 2015 18:42:19 +0000 (10:42 -0800)] 
IB/usnic: Handle 0 counts in resource allocation

Signed-off-by: Dave Goodell <dgoodell@cisco.com>
Reviewed-by: Reese Faucette <rfaucett@cisco.com>
Reviewed-by: Xuyang Wang <xuywang@cisco.com>
Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/usnic: Fix resource leak in error case
Nelson Escobar [Wed, 9 Dec 2015 18:42:18 +0000 (10:42 -0800)] 
IB/usnic: Fix resource leak in error case

Signed-off-by: Dave Goodell <dgoodell@cisco.com>
Reviewed-by: Reese Faucette <rfaucett@cisco.com>
Reviewed-by: Xuyang Wang <xuywang@cisco.com>
Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/usnic: Support more QP state transitions
Nelson Escobar [Wed, 9 Dec 2015 18:42:17 +0000 (10:42 -0800)] 
IB/usnic: Support more QP state transitions

They were already implemented at a lower layer, but the upper level
routine placed arbitrary restrictions on which transitions were
permitted.  Simplify the state machine logic to live wholly in
usnic_ib_qp_grp_modify.

Signed-off-by: Dave Goodell <dgoodell@cisco.com>
Reviewed-by: Reese Faucette <rfaucett@cisco.com>
Reviewed-by: Xuyang Wang <xuywang@cisco.com>
Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/usnic: Fix message typo
Nelson Escobar [Wed, 9 Dec 2015 18:42:16 +0000 (10:42 -0800)] 
IB/usnic: Fix message typo

Signed-off-by: Dave Goodell <dgoodell@cisco.com>
Reviewed-by: Reese Faucette <rfaucett@cisco.com>
Reviewed-by: Xuyang Wang <xuywang@cisco.com>
Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/usnic: Fix incorrect cast in usnic_ib_fw_string_to_u64
Nelson Escobar [Wed, 9 Dec 2015 18:42:15 +0000 (10:42 -0800)] 
IB/usnic: Fix incorrect cast in usnic_ib_fw_string_to_u64

Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Reviewed-by: Dave Goodell <dgoodell@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/usnic: Improve a failure message
Nelson Escobar [Wed, 9 Dec 2015 18:42:14 +0000 (10:42 -0800)] 
IB/usnic: Improve a failure message

Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Reviewed-by: Dave Goodell <dgoodell@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/usnic: Remove unused prototype
Nelson Escobar [Wed, 9 Dec 2015 18:39:49 +0000 (10:39 -0800)] 
IB/usnic: Remove unused prototype

query_protocol() was added in commit 6b90a6d66b17 ("IB/Verbs:
Implement new callback query_protocol()") and then removed in
commit f9b22e355d38 ("IB/core: Convert core to use bitfield
for caps").

This left behind an unused prototype.

Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Reviewed-by: Dave Goodell <dgoodell@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/cma: Join and leave multicast groups with IGMP
Moni Shoua [Wed, 23 Dec 2015 12:56:57 +0000 (14:56 +0200)] 
IB/cma: Join and leave multicast groups with IGMP

Since RoCEv2 is a protocol over IP header it is required to send IGMP
join and leave requests to the network when joining and leaving
multicast groups.

Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Initialize UD header structure with IP and UDP headers
Moni Shoua [Wed, 23 Dec 2015 12:56:56 +0000 (14:56 +0200)] 
IB/core: Initialize UD header structure with IP and UDP headers

ib_ud_header_init() is used to format InfiniBand headers
in a buffer up to (but not with) BTH. For RoCE UDP ENCAP it is
required that this function would be able to build also IP and UDP
headers.

Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/cma: Add configfs for rdma_cm
Matan Barak [Wed, 23 Dec 2015 12:56:55 +0000 (14:56 +0200)] 
IB/cma: Add configfs for rdma_cm

Users would like to control the behaviour of rdma_cm.
For example, old applications which don't set the
required RoCE gid type could be executed on RoCE V2
network types. In order to support this configuration,
we implement a configfs for rdma_cm.

In order to use the configfs, one needs to mount it and
mkdir <IB device name> inside rdma_cm directory.

The patch adds support for a single configuration file,
default_roce_mode. The mode can either be "IB/RoCE v1" or
"RoCE v2".

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/rdma_cm: Add wrapper for cma reference count
Matan Barak [Wed, 23 Dec 2015 12:56:54 +0000 (14:56 +0200)] 
IB/rdma_cm: Add wrapper for cma reference count

Currently, cma users can't increase or decrease the cma reference
count. This is necassary when setting cma attributes (like the
default GID type) in order to avoid use-after-free errors.
Adding cma_ref_dev and cma_deref_dev APIs.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Validate route when we init ah
Matan Barak [Wed, 23 Dec 2015 12:56:53 +0000 (14:56 +0200)] 
IB/core: Validate route when we init ah

In order to make sure API users don't try to use SGIDs which don't
conform to the routing table, validate the route before searching
the RoCE GID table.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Move rdma_is_upper_dev_rcu to header file
Matan Barak [Wed, 23 Dec 2015 12:56:52 +0000 (14:56 +0200)] 
IB/core: Move rdma_is_upper_dev_rcu to header file

In order to validate the route, we need an easy way to check if a
net-device belongs to our RDMA device. Move this helper function
to a header file in order to make this check easier.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Add rdma_network_type to wc
Somnath Kotur [Wed, 23 Dec 2015 12:56:51 +0000 (14:56 +0200)] 
IB/core: Add rdma_network_type to wc

Providers should tell IB core the wc's network type.
This is used in order to search for the proper GID in the
GID table. When using HCAs that can't provide this info,
IB core tries to deep examine the packet and extract
the GID type by itself.

We choose sgid_index and type from all the matching entries in
RDMA-CM based on hint from the IP stack and we set hop_limit for
the IP packet based on above hint from IP stack.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Somnath Kotur <Somnath.Kotur@Avagotech.Com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Add ROCE_UDP_ENCAP (RoCE V2) type
Matan Barak [Wed, 23 Dec 2015 12:56:50 +0000 (14:56 +0200)] 
IB/core: Add ROCE_UDP_ENCAP (RoCE V2) type

Adding RoCE v2 GID type and port type. Vendors
which support this type will get their GID table
populated with RoCE v2 GIDs automatically.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Add gid attributes to sysfs
Matan Barak [Wed, 23 Dec 2015 12:56:49 +0000 (14:56 +0200)] 
IB/core: Add gid attributes to sysfs

This patch set adds attributes of net device and gid type to each GID
in the GID table. Users that use verbs directly need to specify
the GID index. Since the same GID could have different types or
associated net devices, users should have the ability to query the
associated GID attributes. Adding these attributes to sysfs.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/cm: Use the source GID index type
Matan Barak [Wed, 23 Dec 2015 12:56:48 +0000 (14:56 +0200)] 
IB/cm: Use the source GID index type

Previosuly, cm and cma modules supported only IB and RoCE v1 GID type.
In order to support multiple GID types, the gid_type is passed to
cm_init_av_by_path and stored in the path record.

The rdma cm client would use a default GID type that will be saved in
rdma_id_private.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Add gid_type to gid attribute
Matan Barak [Wed, 23 Dec 2015 12:56:47 +0000 (14:56 +0200)] 
IB/core: Add gid_type to gid attribute

In order to support multiple GID types, we need to store the gid_type
with each GID. This is also aligned with the RoCE v2 annex "RoCEv2 PORT
GID table entries shall have a "GID type" attribute that denotes the L3
Address type". The currently supported GID is IB_GID_TYPE_IB which is
also RoCE v1 GID type.

This implies that gid_type should be added to roce_gid_table meta-data.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: don't search the GID table twice
Matan Barak [Wed, 28 Oct 2015 14:52:41 +0000 (16:52 +0200)] 
IB/core: don't search the GID table twice

Previously, we've searched the GID table twice: first when we searched
the table for a GID matching the proposed new one, and second when we
didn't find a match, we searched again for an empty GID slot in the
table.  Instead, search the table once noting the first empty slot as
we search for our target GID.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Change per-entry lock in RoCE GID table to one lock
Matan Barak [Wed, 28 Oct 2015 14:52:40 +0000 (16:52 +0200)] 
IB/core: Change per-entry lock in RoCE GID table to one lock

Previously, IB GID cached used a lock per entry. This could result
in spending a lot of CPU cycles for locking and unlocking just
in order to find a GID. Changing this in favor of one lock per
a GID table.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Refactor GID cache's ib_dispatch_event
Matan Barak [Wed, 28 Oct 2015 14:52:39 +0000 (16:52 +0200)] 
IB/core: Refactor GID cache's ib_dispatch_event

Refactor ib_dispatch_event into a new function in order to avoid
duplicating code in the next patch.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoMerge branches '4.5/Or-cleanup' and '4.5/rdma-cq' into k.o/for-4.5
Doug Ledford [Tue, 22 Dec 2015 22:03:15 +0000 (17:03 -0500)] 
Merge branches '4.5/Or-cleanup' and '4.5/rdma-cq' into k.o/for-4.5

Signed-off-by: Doug Ledford <dledford@redhat.com>
Conflicts:
drivers/infiniband/ulp/iser/iser_verbs.c

8 years agoIB/core: Remove ib_query_device
Or Gerlitz [Fri, 18 Dec 2015 08:59:50 +0000 (10:59 +0200)] 
IB/core: Remove ib_query_device

The copy of the attributes present on the device is now used by all consumers
except for uverbs in case of serving user-space query, where dev->query_device
is called.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agostaging/o2iblnd: Avoid calling ib_query_device
Or Gerlitz [Fri, 18 Dec 2015 08:59:49 +0000 (10:59 +0200)] 
staging/o2iblnd: Avoid calling ib_query_device

Instead, use the cached copy of the attributes present on the device.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoxprtrdma: Avoid calling ib_query_device
Or Gerlitz [Fri, 18 Dec 2015 08:59:48 +0000 (10:59 +0200)] 
xprtrdma: Avoid calling ib_query_device

Instead, use the cached copy of the attributes present on the device.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agonet/rds: Avoid calling ib_query_device
Or Gerlitz [Fri, 18 Dec 2015 08:59:47 +0000 (10:59 +0200)] 
net/rds: Avoid calling ib_query_device

Instead, use the cached copy of the attributes present on the device.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/ulps: Avoid calling ib_query_device
Or Gerlitz [Fri, 18 Dec 2015 08:59:46 +0000 (10:59 +0200)] 
IB/ulps: Avoid calling ib_query_device

Instead, use the cached copy of the attributes present on the device.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Avoid calling ib_query_device
Or Gerlitz [Fri, 18 Dec 2015 08:59:45 +0000 (10:59 +0200)] 
IB/core: Avoid calling ib_query_device

Use the cached copy of the attributes present on the device, except for
the case of a query originating from user-space, where we have to invoke
the driver query_device entry, so they can fill in their udata.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoIB/core: Save the device attributes on the device structure
Ira Weiny [Fri, 18 Dec 2015 08:59:44 +0000 (10:59 +0200)] 
IB/core: Save the device attributes on the device structure

This way both the IB core and upper level drivers can access these cached
device attributes rather than querying or caching them on their own.

Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
8 years agoLinux 4.4-rc6
Linus Torvalds [Mon, 21 Dec 2015 00:06:09 +0000 (16:06 -0800)] 
Linux 4.4-rc6

8 years agoMerge tag 'rtc-4.4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux
Linus Torvalds [Sun, 20 Dec 2015 18:01:11 +0000 (10:01 -0800)] 
Merge tag 'rtc-4.4-3' of git://git./linux/kernel/git/abelloni/linux

Pull RTC fixes from Alexandre Belloni:
 "Late fixes for the RTC subsystem for 4.4:

  A fix for a nasty hardware bug in rk808 and an initialization
  reordering in da9063 to fix a possible crash"

* tag 'rtc-4.4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux:
  rtc: da9063: fix access ordering error during RTC interrupt at system power on
  rtc: rk808: Compensate for Rockchip calendar deviation on November 31st

8 years agortc: da9063: fix access ordering error during RTC interrupt at system power on
Steve Twiss [Tue, 8 Dec 2015 16:28:39 +0000 (16:28 +0000)] 
rtc: da9063: fix access ordering error during RTC interrupt at system power on

This fix alters the ordering of the IRQ and device registrations in the RTC
driver probe function. This change will apply to the RTC driver that supports
both DA9063 and DA9062 PMICs.

A problem could occur with the existing RTC driver if:

A system is started from a cold boot using the PMIC RTC IRQ to initiate a
power on operation. For instance, if an RTC alarm is used to start a
platform from power off.
The existing driver IRQ is requested before the device has been properly
registered.
i.e.
    ret = devm_request_threaded_irq()
comes before
    rtc->rtc_dev = devm_rtc_device_register();

In this case, the interrupt can be called before the device has been
registered and the handler can be called immediately. The IRQ handler
da9063_alarm_event() contains the function call

    rtc_update_irq(rtc->rtc_dev, 1, RTC_IRQF | RTC_AF);

which in turn tries to access the unavailable rtc->rtc_dev.

The fix is to reorder the functions inside the RTC probe. The IRQ is
requested after the RTC device resource has been registered so that
get_irq_byname is the last thing to happen.

Signed-off-by: Steve Twiss <stwiss.opensource@diasemi.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
8 years agortc: rk808: Compensate for Rockchip calendar deviation on November 31st
Julius Werner [Tue, 15 Dec 2015 23:02:49 +0000 (15:02 -0800)] 
rtc: rk808: Compensate for Rockchip calendar deviation on November 31st

In A.D. 1582 Pope Gregory XIII found that the existing Julian calendar
insufficiently represented reality, and changed the rules about
calculating leap years to account for this. Similarly, in A.D. 2013
Rockchip hardware engineers found that the new Gregorian calendar still
contained flaws, and that the month of November should be counted up to
31 days instead. Unfortunately it takes a long time for calendar changes
to gain widespread adoption, and just like more than 300 years went by
before the last Protestant nation implemented Greg's proposal, we will
have to wait a while until all religions and operating system kernels
acknowledge the inherent advantages of the Rockchip system. Until then
we need to translate dates read from (and written to) Rockchip hardware
back to the Gregorian format.

This patch works by defining Jan 1st, 2016 as the arbitrary anchor date
on which Rockchip and Gregorian calendars are in sync. From that we can
translate arbitrary later dates back and forth by counting the number
of November/December transitons since the anchor date to determine the
offset between the calendars. We choose this method (rather than trying
to regularly "correct" the date stored in hardware) since it's the only
way to ensure perfect time-keeping even if the system may be shut down
for an unknown number of years. The drawback is that other software
reading the same hardware (e.g. mainboard firmware) must use the same
translation convention (including the same anchor date) to be able to
read and write correct timestamps from/to the RTC.

Signed-off-by: Julius Werner <jwerner@chromium.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
8 years agoMerge tag 'tty-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sun, 20 Dec 2015 01:44:19 +0000 (17:44 -0800)] 
Merge tag 'tty-4.4-rc6' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial fixes from Greg KH:
 "Here are some tty/serial driver fixes for 4.4-rc6 that resolve some
  reported problems.  All of these have been in linux-next.  The details
  are in the shortlog"

* tag 'tty-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty: Fix GPF in flush_to_ldisc()
  serial: earlycon: Add missing spinlock initialization
  serial: sh-sci: Fix length of scatterlist
  n_tty: Fix poll() after buffer-limited eof push read
  serial: 8250_uniphier: fix dl_read and dl_write functions

8 years agoMerge tag 'usb-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 20 Dec 2015 01:33:58 +0000 (17:33 -0800)] 
Merge tag 'usb-4.4-rc6' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some USB and PHY fixes for 4.4-rc6.  All of them resolve some
  reported problems.  Full details in the shortlog"

* tag 'usb-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  USB: fix invalid memory access in hub_activate()
  USB: ipaq.c: fix a timeout loop
  phy: core: Get a refcount to phy in devm_of_phy_get_by_index()
  phy: cygnus: pcie: add missing of_node_put
  phy: miphy365x: add missing of_node_put
  phy: miphy28lp: add missing of_node_put
  phy: rockchip-usb: add missing of_node_put
  phy: berlin-sata: add missing of_node_put
  phy: mt65xx-usb3: add missing of_node_put
  phy: brcmstb-sata: add missing of_node_put
  phy: sun9i-usb: add USB dependency

8 years agoMerge tag 'md/4.4-rc5-fixes' of git://neil.brown.name/md
Linus Torvalds [Sun, 20 Dec 2015 00:46:46 +0000 (16:46 -0800)] 
Merge tag 'md/4.4-rc5-fixes' of git://neil.brown.name/md

Pull md fixes from Neil Brown:
 "Four fixes for md:

   - two recently introduced regressions fixed.
   - one older bug in RAID10 - tagged for -stable since 4.2
   - one minor sysfs api improvement"

* tag 'md/4.4-rc5-fixes' of git://neil.brown.name/md:
  Fix remove_and_add_spares removes drive added as spare in slot_store
  md: fix bug due to nested suspend
  MD: change journal disk role to disk 0
  md/raid10: fix data corruption and crash during resync

8 years agoMerge tag 'powerpc-4.4-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sun, 20 Dec 2015 00:40:48 +0000 (16:40 -0800)] 
Merge tag 'powerpc-4.4-5' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 - Partial revert of "powerpc: Individual System V IPC system calls"
 - pr_warn_once on unsupported OPAL_MSG type from Stewart
 - Fix deadlock in opal-irqchip introduced by "Fix double endian
   conversion" from Alistair

* tag 'powerpc-4.4-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/opal-irqchip: Fix deadlock introduced by "Fix double endian conversion"
  powerpc/powernv: pr_warn_once on unsupported OPAL_MSG type
  Partial revert of "powerpc: Individual System V IPC system calls"

8 years agoMerge tag 'spi-fix-v4.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi
Linus Torvalds [Sat, 19 Dec 2015 18:10:43 +0000 (10:10 -0800)] 
Merge tag 'spi-fix-v4.4-rc5' of git://git./linux/kernel/git/broonie/spi

Pull spi fixes from Mark Brown:
 "A couple of reference counting bugs here, one in spidev and one with
  holding an extra reference in the core that we never freed if we
  removed a device, plus a driver specific fix.  Both of the refcounting
  bugs are very old but they've only been found by observation so
  hopefully their impact has been low"

* tag 'spi-fix-v4.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: fix parent-device reference leak
  spi: spidev: Hold spi_lock over all defererences of spi in release()
  spi-fsl-dspi: Fix CTAR Register access

8 years agoMerge tag 'gpio-v4.4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux...
Linus Torvalds [Sat, 19 Dec 2015 18:05:00 +0000 (10:05 -0800)] 
Merge tag 'gpio-v4.4-3' of git://git./linux/kernel/git/linusw/linux-gpio

Pull GPIO fixes from Linus Walleij:
 "Some GPIO fixes for the v4.4 series.  Most prominent: I revert the
  error propagation from the .get() function until we can fix up all the
  drivers properly for v4.5.

   - Revert the error number propagation from the .get() vtable entry
     temporarily, until we make the proper fixes to all drivers.
   - Fix the clamping behaviour in the generic GPIO driver.
   - Driver fix for the ath79 driver"

* tag 'gpio-v4.4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio:
  gpio: revert get() to non-errorprogating behaviour
  gpio: generic: clamp values from bgpio_get_set()
  gpio: ath79: Fix the logic to clear offset bit of AR71XX_GPIO_REG_OE register

8 years agoMerge tag 'pinctrl-v4.4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Sat, 19 Dec 2015 18:01:03 +0000 (10:01 -0800)] 
Merge tag 'pinctrl-v4.4-3' of git://git./linux/kernel/git/linusw/linux-pinctrl

Pull pin control fixes from Linus Walleij:
 - Driver fixes for Freescale i.MX7D, Intel, Broadcom 2835
 - One MAINTAINERS entry

* tag 'pinctrl-v4.4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  MAINTAINERS: pinctrl: Add maintainers for pinctrl-single
  pinctrl: bcm2835: Fix initial value for direction_output
  pinctrl: intel: fix offset calculation issue of register PAD_OWN
  pinctrl: intel: fix bug of register offset calculation
  pinctrl: freescale: add ZERO_OFFSET_VALID flag for vf610 pinctrl

8 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sat, 19 Dec 2015 17:52:44 +0000 (09:52 -0800)] 
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "A set of 'usual' driver bugfixes for the I2C subsystem"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: rcar: disable runtime PM correctly in slave mode
  i2c: designware: Keep pm_runtime_enable/_disable calls in sync
  i2c: designware: fix IO timeout issue for AMD controller
  i2c: imx: init bus recovery info before adding i2c adapter
  i2c: do not use 0x in front of %pa
  i2c: davinci: Increase module clock frequency
  i2c: mv64xxx: The n clockdiv factor is 0 based on sunxi SoCs
  i2c: rk3x: populate correct variable for sda_falling_time

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 19 Dec 2015 17:51:11 +0000 (09:51 -0800)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Just a few assorted driver fixes"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: elants_i2c - fix wake-on-touch
  Input: elan_i2c - set input device's vendor and product IDs
  Input: sun4i-lradc-keys - fix typo in binding documentation
  Input: atmel_mxt_ts - add maxtouch to I2C table for module autoload
  Input: arizona-haptic - fix disabling of haptics device
  Input: aiptek - fix crash on detecting device without endpoints
  Input: atmel_mxt_ts - add generic platform data for Chromebooks
  Input: parkbd - clear unused function pointers
  Input: walkera0701 - clear unused function pointers
  Input: turbografx - clear unused function pointers
  Input: gamecon - clear unused function pointers
  Input: db9 - clear unused function pointers

8 years agoi2c: rcar: disable runtime PM correctly in slave mode
Wolfram Sang [Wed, 16 Dec 2015 19:05:18 +0000 (20:05 +0100)] 
i2c: rcar: disable runtime PM correctly in slave mode

When we also are I2C slave, we need to disable runtime PM because the
address detection mechanism needs to be active all the time. However, we
can reenable runtime PM once the slave instance was unregistered. So,
use pm_runtime_get_sync/put to achieve this, since it has proper
refcounting. pm_runtime_allow/forbid is like a global knob controllable
from userspace which is unsuitable here.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Cc: stable@kernel.org
8 years agoMerge tag 'pm+acpi-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Sat, 19 Dec 2015 05:01:35 +0000 (21:01 -0800)] 
Merge tag 'pm+acpi-4.4-rc6' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael Wysocki:
 "These fix a potential regression introduced during the 4.3 cycle
  (generic power domains framework), a nasty bug that has been present
  forever (power capping RAPL driver), a build issue (Tegra cpufreq
  driver) and a minor ugliness introduced recently (intel_pstate).

  Specifics:

   - Fix a potential regression in the generic power domains framework
     introduced during the 4.3 development cycle that may lead to
     spurious failures of system suspend in certain situations (Ulf
     Hansson).

   - Fix a problem in the power capping RAPL (Running Average Power
     Limits) driver that causes it to initialize successfully on some
     systems where it is not supposed to do that which is due to an
     incorrect check in an initialization routine (Prarit Bhargava).

   - Fix a build problem in the cpufreq Tegra driver that depends on the
     regulator framework, but that dependency is not reflected in
     Kconfig (Arnd Bergmann).

   - Fix a recent mistake in the intel_pstate driver where a numeric
     constant is used directly instead of a symbol defined specifically
     for the case in question (Prarit Bhargava)"

* tag 'pm+acpi-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  powercap / RAPL: fix BIOS lock check
  cpufreq: intel_pstate: Minor cleanup for FRAC_BITS
  cpufreq: tegra: add regulator dependency for T124
  PM / Domains: Allow runtime PM callbacks to be re-used during system PM

8 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 19 Dec 2015 04:35:35 +0000 (20:35 -0800)] 
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Three fixes this time, two in SES picked up by KASAN for various types
  of buffer overrun.  The first is a USB array which returns page 8
  whatever is asked for and causes us to overrun with incorrect data
  format assumptions and the second is an invalid iteration of page 10
  (the additional information page).

  The final fix is a reversion of a NULL deref fix which caused
  suspend/resume not to be called in pairs leading to incorrect device
  operation (Jens has queued a more proper fix for the problem in
  block)"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  ses: fix additional element traversal bug
  Revert "SCSI: Fix NULL pointer dereference in runtime PM"
  ses: Fix problems with simple enclosures

8 years agoInput: elants_i2c - fix wake-on-touch
James Chen [Fri, 18 Dec 2015 23:51:48 +0000 (15:51 -0800)] 
Input: elants_i2c - fix wake-on-touch

When sending "SLEEP" command to the controller it ceases scanning
completely and is unable to wake the system up from sleep, so if it is
configured as a wakeup source we should simply configure interrupt for
wakeup and rely on idle logic within the controller to reduce power
consumption while it is not used.

Signed-off-by: James Chen <james.chen@emc.com.tw>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
8 years agoMerge tag 'media/v4.4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Fri, 18 Dec 2015 23:41:35 +0000 (15:41 -0800)] 
Merge tag 'media/v4.4-3' of git://git./linux/kernel/git/mchehab/linux-media

Pull media fixes from Mauro Carvalho Chehab.

* tag 'media/v4.4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  [media] airspy: increase USB control message buffer size
  [media] hackrf: move RF gain ctrl enable behind module parameter
  [media] hackrf: fix possible null ptr on debug printing
  [media] Revert "[media] ivtv: avoid going past input/audio array"

8 years agoMerge branch 'for-linus-4.4' of git://git.kernel.org/pub/scm/linux/kernel/git/mason...
Linus Torvalds [Fri, 18 Dec 2015 23:35:08 +0000 (15:35 -0800)] 
Merge branch 'for-linus-4.4' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "A couple of small fixes"

* 'for-linus-4.4' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: check prepare_uptodate_page() error code earlier
  Btrfs: check for empty bitmap list in setup_cluster_bitmaps
  btrfs: fix misleading warning when space cache failed to load
  Btrfs: fix transaction handle leak in balance
  Btrfs: fix unprotected list move from unused_bgs to deleted_bgs list

8 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 18 Dec 2015 22:25:57 +0000 (14:25 -0800)] 
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "Three patches"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  include/linux/mmdebug.h: should include linux/bug.h
  mm/zswap: change incorrect strncmp use to strcmp
  proc: fix -ESRCH error when writing to /proc/$pid/coredump_filter

8 years agoinclude/linux/mmdebug.h: should include linux/bug.h
James Morse [Fri, 18 Dec 2015 22:22:07 +0000 (14:22 -0800)] 
include/linux/mmdebug.h: should include linux/bug.h

mmdebug.h uses BUILD_BUG_ON_INVALID(), assuming someone else included
linux/bug.h.  Include it ourselves.

This saves build-failures such as:

  arch/arm64/include/asm/pgtable.h: In function 'set_pte_at':
  arch/arm64/include/asm/pgtable.h:281:3: error: implicit declaration of function 'BUILD_BUG_ON_INVALID' [-Werror=implicit-function-declaration]
   VM_WARN_ONCE(!pte_young(pte),

Fixes: 02602a18c32d7 ("bug: completely remove code generated by disabled VM_BUG_ON()")
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agomm/zswap: change incorrect strncmp use to strcmp
Dan Streetman [Fri, 18 Dec 2015 22:22:04 +0000 (14:22 -0800)] 
mm/zswap: change incorrect strncmp use to strcmp

Change the use of strncmp in zswap_pool_find_get() to strcmp.

The use of strncmp is no longer correct, now that zswap_zpool_type is
not an array; sizeof() will return the size of a pointer, which isn't
the right length to compare.  We don't need to use strncmp anyway,
because the existing params and the passed in params are all guaranteed
to be null terminated, so strcmp should be used.

Signed-off-by: Dan Streetman <ddstreet@ieee.org>
Reported-by: Weijie Yang <weijie.yang@samsung.com>
Cc: Seth Jennings <sjennings@variantweb.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoproc: fix -ESRCH error when writing to /proc/$pid/coredump_filter
Colin Ian King [Fri, 18 Dec 2015 22:22:01 +0000 (14:22 -0800)] 
proc: fix -ESRCH error when writing to /proc/$pid/coredump_filter

Writing to /proc/$pid/coredump_filter always returns -ESRCH because commit
774636e19ed51 ("proc: convert to kstrto*()/kstrto*_from_user()") removed
the setting of ret after the get_proc_task call and incorrectly left it as
-ESRCH.  Instead, return 0 when successful.

Example breakage:

  echo 0 > /proc/self/coredump_filter
  bash: echo: write error: No such process

Fixes: 774636e19ed51 ("proc: convert to kstrto*()/kstrto*_from_user()")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: <stable@vger.kernel.org> [4.3+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoMerge tag 'hwmon-for-linus-v4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 18 Dec 2015 20:51:52 +0000 (12:51 -0800)] 
Merge tag 'hwmon-for-linus-v4.4-rc6' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fixes from Guenter Roeck:

 - Select CONFIG_BITREVERSE for sht15 driver to avoid build failure if
   it is not configured.

 - Force wait for conversion time for the first valid data in tmp102
   driver to avoid reporting erroneous data to the thermal subsystem.

* tag 'hwmon-for-linus-v4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (sht15) Select CONFIG_BITREVERSE
  hwmon: (tmp102) Force wait for conversion time for the first valid data

8 years agoMerge tag 'iommu-fixes-v4.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 18 Dec 2015 20:38:35 +0000 (12:38 -0800)] 
Merge tag 'iommu-fixes-v4.4-rc5' of git://git./linux/kernel/git/joro/iommu

Pull IOMMU fixes from Joerg Roedel:
 "Two similar fixes for the Intel and AMD IOMMU drivers to add proper
  access checks before calling handle_mm_fault"

* tag 'iommu-fixes-v4.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/vt-d: Do access checks before calling handle_mm_fault()
  iommu/amd: Do proper access checking before calling handle_mm_fault()

8 years agoMerge tag 'for-linus-4.4-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 18 Dec 2015 20:24:52 +0000 (12:24 -0800)] 
Merge tag 'for-linus-4.4-rc5-tag' of git://git./linux/kernel/git/xen/tip

Pull xen bug fixes from David Vrabel:
 - XSA-155 security fixes to backend drivers.
 - XSA-157 security fixes to pciback.

* tag 'for-linus-4.4-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen-pciback: fix up cleanup path when alloc fails
  xen/pciback: Don't allow MSI-X ops if PCI_COMMAND_MEMORY is not set.
  xen/pciback: For XEN_PCI_OP_disable_msi[|x] only disable if device has MSI(X) enabled.
  xen/pciback: Do not install an IRQ handler for MSI interrupts.
  xen/pciback: Return error on XEN_PCI_OP_enable_msix when device has MSI or MSI-X enabled
  xen/pciback: Return error on XEN_PCI_OP_enable_msi when device has MSI or MSI-X enabled
  xen/pciback: Save xen_pci_op commands before processing it
  xen-scsiback: safely copy requests
  xen-blkback: read from indirect descriptors only once
  xen-blkback: only read request operation from shared ring once
  xen-netback: use RING_COPY_REQUEST() throughout
  xen-netback: don't use last request to determine minimum Tx credit
  xen: Add RING_COPY_REQUEST()
  xen/x86/pvh: Use HVM's flush_tlb_others op
  xen: Resume PMU from non-atomic context
  xen/events/fifo: Consume unprocessed events when a CPU dies

8 years agoMerge tag 'arc-fixes-for-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 18 Dec 2015 20:19:01 +0000 (12:19 -0800)] 
Merge tag 'arc-fixes-for-4.4-rc6' of git://git./linux/kernel/git/vgupta/arc

Pull ARC architecture fixes from Vineet Gupta:
 "Fixes for:

 - perf interrupts on SMP: Not enabled (at boot) and disabled (at runtime)
 - stack unwinder regression (for modules, ignoring dwarf3)
 - nsim hosed for non default kernel link base builds"

* tag 'arc-fixes-for-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARC: smp: Rename platform hook @init_cpu_smp -> @init_per_cpu
  ARC: rename smp operation init_irq_cpu() to init_per_cpu()
  ARC: dw2 unwind: Ignore CIE version !=1 gracefully instead of bailing
  ARC: dw2 unwind: Reinstante unwinding out of modules
  ARC: [plat-sim] unbork non default CONFIG_LINUX_LINK_BASE
  ARC: intc: Document arc_request_percpu_irq() better
  ARCv2: perf: Ensure perf intr gets enabled on all cores
  ARC: intc: No need to clear IRQ_NOAUTOEN
  ARCv2: intc: Fix random perf irq disabling in SMP setup
  ARC: [axs10x] cap ethernet phy to 100 Mbit/sec

8 years agoMerge tag 'sound-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 18 Dec 2015 19:47:06 +0000 (11:47 -0800)] 
Merge tag 'sound-4.4-rc6' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "As usual in rc6, this update contains only a few HD-audio and
  USB-audio device-specific quirks: yet another Thinkpad noise fixes,
  Dell headphone mic fixes, and AudioQuest DragonFly fixes"

* tag 'sound-4.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda - Add a fixup for Thinkpad X1 Carbon 2nd
  ALSA: hda - Set codec to D3 at reboot/shutdown on Thinkpads
  ALSA: hda - Apply click noise workaround for Thinkpads generically
  ALSA: hda - Fix headphone mic input on a few Dell ALC293 machines
  ALSA: usb-audio: Add sample rate inquiry quirk for AudioQuest DragonFly
  ALSA: usb-audio: Add a more accurate volume quirk for AudioQuest DragonFly

8 years agoMerge tag 'for-linus-20151217' of git://git.infradead.org/linux-mtd
Linus Torvalds [Fri, 18 Dec 2015 19:19:16 +0000 (11:19 -0800)] 
Merge tag 'for-linus-20151217' of git://git.infradead.org/linux-mtd

Pull MTD fixes from Brian Norris:
 "I was holding out on this pull request for a bit, since there are a
  few other small issues being discussed that look like 4.4-rc
  regressions.  Hopefully I can get those stabilized soon, but these are
  ready at any rate:

   - A little bit of a last-minute change for the device tree "fixed
     partition" binding.  This is needed because we might want to reuse
     the 'partitions' subnode for other sorts of partitioning
     descriptions -- e.g., for describing which on-flash partition
     format(s) might be used on the system.

   - Also tone down a warning message, since it is probably going to
     show up on a lot of systems where it should just be ignored"

* tag 'for-linus-20151217' of git://git.infradead.org/linux-mtd:
  doc: dt: mtd: partitions: add compatible property to "partitions" node
  mtd: ofpart: don't complain about missing 'partitions' node too loudly

8 years agoUSB: fix invalid memory access in hub_activate()
Alan Stern [Wed, 16 Dec 2015 18:32:38 +0000 (13:32 -0500)] 
USB: fix invalid memory access in hub_activate()

Commit 8520f38099cc ("USB: change hub initialization sleeps to
delayed_work") changed the hub_activate() routine to make part of it
run in a workqueue.  However, the commit failed to take a reference to
the usb_hub structure or to lock the hub interface while doing so.  As
a result, if a hub is plugged in and quickly unplugged before the work
routine can run, the routine will try to access memory that has been
deallocated.  Or, if the hub is unplugged while the routine is
running, the memory may be deallocated while it is in active use.

This patch fixes the problem by taking a reference to the usb_hub at
the start of hub_activate() and releasing it at the end (when the work
is finished), and by locking the hub interface while the work routine
is running.  It also adds a check at the start of the routine to see
if the hub has already been disconnected, in which nothing should be
done.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Reported-by: Alexandru Cornea <alexandru.cornea@intel.com>
Tested-by: Alexandru Cornea <alexandru.cornea@intel.com>
Fixes: 8520f38099cc ("USB: change hub initialization sleeps to delayed_work")
CC: <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: ipaq.c: fix a timeout loop
Dan Carpenter [Wed, 16 Dec 2015 11:06:37 +0000 (14:06 +0300)] 
USB: ipaq.c: fix a timeout loop

The code expects the loop to end with "retries" set to zero but, because
it is a post-op, it will end set to -1.  I have fixed this by moving the
decrement inside the loop.

Fixes: 014aa2a3c32e ('USB: ipaq: minor ipaq_open() cleanup.')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years ago[media] airspy: increase USB control message buffer size
Antti Palosaari [Mon, 26 Oct 2015 20:58:14 +0000 (18:58 -0200)] 
[media] airspy: increase USB control message buffer size

Driver requested device firmware version string during probe using
only 24 byte long buffer. That buffer is too small for newer firmware
versions, which causes device firmware hang - device stops responding
to any commands after that. Increase buffer size to 128 which should
be enough for any current and future version strings.

Link: https://github.com/airspy/host/issues/27
Cc: <stable@vger.kernel.org> # 3.17+
Reported-by: Benjamin Vernoux <bvernoux@gmail.com>
Signed-off-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
8 years ago[media] hackrf: move RF gain ctrl enable behind module parameter
Antti Palosaari [Fri, 23 Oct 2015 22:01:31 +0000 (20:01 -0200)] 
[media] hackrf: move RF gain ctrl enable behind module parameter

Used Avago MGA-81563 RF amplifier could be destroyed pretty easily
with too strong signal or transmitting to bad antenna.
Add module parameter 'enable_rf_gain_ctrl' which allows enabling
RF gain control - otherwise, default without the module parameter,
RF gain control is set to 'grabbed' state which prevents setting
value to the control.

Signed-off-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
8 years ago[media] hackrf: fix possible null ptr on debug printing
Antti Palosaari [Wed, 21 Oct 2015 21:02:41 +0000 (19:02 -0200)] 
[media] hackrf: fix possible null ptr on debug printing

drivers/media/usb/hackrf/hackrf.c:1533 hackrf_probe()
error: we previously assumed 'dev' could be null (see line 1366)

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Antti Palosaari <crope@iki.fi>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
8 years ago[media] Revert "[media] ivtv: avoid going past input/audio array"
Mauro Carvalho Chehab [Wed, 11 Nov 2015 11:22:36 +0000 (09:22 -0200)] 
[media] Revert "[media] ivtv: avoid going past input/audio array"

This patch broke ivtv logic, as reported at
 https://bugzilla.redhat.com/show_bug.cgi?id=1278942

This reverts commit 09290cc885937cab3b2d60a6d48fe3d2d3e04061.

Cc: stable@vger.kernel.org # for v4.1 and upper
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
8 years agoMerge tag 'phy-for-4.4-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/kishon...
Greg Kroah-Hartman [Fri, 18 Dec 2015 17:24:50 +0000 (09:24 -0800)] 
Merge tag 'phy-for-4.4-rc' of git://git./linux/kernel/git/kishon/linux-phy into usb-linus

Kishon writes:

phy: for 4.4 -rc

*) Add missing of_node_put in a bunch of PHY drivers
*) Add get_device in devm_of_phy_get_by_index()
*) Fix randconfig build error in sun9i usb driver

8 years agoxen-pciback: fix up cleanup path when alloc fails
Doug Goldstein [Thu, 26 Nov 2015 20:32:39 +0000 (14:32 -0600)] 
xen-pciback: fix up cleanup path when alloc fails

When allocating a pciback device fails, clear the private
field. This could lead to an use-after free, however
the 'really_probe' takes care of setting
dev_set_drvdata(dev, NULL) in its failure path (which we would
exercise if the ->probe function failed), so we we
are OK. However lets be defensive as the code can change.

Going forward we should clean up the pci_set_drvdata(dev, NULL)
in the various code-base. That will be for another day.

Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reported-by: Jonathan Creekmore <jonathan.creekmore@gmail.com>
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agohwmon: (sht15) Select CONFIG_BITREVERSE
Arnd Bergmann [Fri, 18 Dec 2015 14:52:28 +0000 (15:52 +0100)] 
hwmon: (sht15) Select CONFIG_BITREVERSE

If CONFIG_BITREVERSE is not built-in, the sht15 driver fails to link:

drivers/built-in.o: In function `sht15_crc8':
drivers/hwmon/sht15.c:195: undefined reference to `byte_rev_table'

This adds a Kconfig 'select' statement, like all other users of
bitrev.h have it.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 33836ee98533 ("hwmon:change sht15_reverse()")
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
8 years agoxen/pciback: Don't allow MSI-X ops if PCI_COMMAND_MEMORY is not set.
Konrad Rzeszutek Wilk [Mon, 2 Nov 2015 23:13:27 +0000 (18:13 -0500)] 
xen/pciback: Don't allow MSI-X ops if PCI_COMMAND_MEMORY is not set.

commit f598282f51 ("PCI: Fix the NIU MSI-X problem in a better way")
teaches us that dealing with MSI-X can be troublesome.

Further checks in the MSI-X architecture shows that if the
PCI_COMMAND_MEMORY bit is turned of in the PCI_COMMAND we
may not be able to access the BAR (since they are memory regions).

Since the MSI-X tables are located in there.. that can lead
to us causing PCIe errors. Inhibit us performing any
operation on the MSI-X unless the MEMORY bit is set.

Note that Xen hypervisor with:
"x86/MSI-X: access MSI-X table only after having enabled MSI-X"
will return:
xen_pciback: 0000:0a:00.1: error -6 enabling MSI-X for guest 3!

When the generic MSI code tries to setup the PIRQ without
MEMORY bit set. Which means with later versions of Xen
(4.6) this patch is not neccessary.

This is part of XSA-157

CC: stable@vger.kernel.org
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen/pciback: For XEN_PCI_OP_disable_msi[|x] only disable if device has MSI(X) enabled.
Konrad Rzeszutek Wilk [Wed, 1 Apr 2015 14:49:47 +0000 (10:49 -0400)] 
xen/pciback: For XEN_PCI_OP_disable_msi[|x] only disable if device has MSI(X) enabled.

Otherwise just continue on, returning the same values as
previously (return of 0, and op->result has the PIRQ value).

This does not change the behavior of XEN_PCI_OP_disable_msi[|x].

The pci_disable_msi or pci_disable_msix have the checks for
msi_enabled or msix_enabled so they will error out immediately.

However the guest can still call these operations and cause
us to disable the 'ack_intr'. That means the backend IRQ handler
for the legacy interrupt will not respond to interrupts anymore.

This will lead to (if the device is causing an interrupt storm)
for the Linux generic code to disable the interrupt line.

Naturally this will only happen if the device in question
is plugged in on the motherboard on shared level interrupt GSI.

This is part of XSA-157

CC: stable@vger.kernel.org
Reviewed-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen/pciback: Do not install an IRQ handler for MSI interrupts.
Konrad Rzeszutek Wilk [Mon, 2 Nov 2015 22:24:08 +0000 (17:24 -0500)] 
xen/pciback: Do not install an IRQ handler for MSI interrupts.

Otherwise an guest can subvert the generic MSI code to trigger
an BUG_ON condition during MSI interrupt freeing:

 for (i = 0; i < entry->nvec_used; i++)
        BUG_ON(irq_has_action(entry->irq + i));

Xen PCI backed installs an IRQ handler (request_irq) for
the dev->irq whenever the guest writes PCI_COMMAND_MEMORY
(or PCI_COMMAND_IO) to the PCI_COMMAND register. This is
done in case the device has legacy interrupts the GSI line
is shared by the backend devices.

To subvert the backend the guest needs to make the backend
to change the dev->irq from the GSI to the MSI interrupt line,
make the backend allocate an interrupt handler, and then command
the backend to free the MSI interrupt and hit the BUG_ON.

Since the backend only calls 'request_irq' when the guest
writes to the PCI_COMMAND register the guest needs to call
XEN_PCI_OP_enable_msi before any other operation. This will
cause the generic MSI code to setup an MSI entry and
populate dev->irq with the new PIRQ value.

Then the guest can write to PCI_COMMAND PCI_COMMAND_MEMORY
and cause the backend to setup an IRQ handler for dev->irq
(which instead of the GSI value has the MSI pirq). See
'xen_pcibk_control_isr'.

Then the guest disables the MSI: XEN_PCI_OP_disable_msi
which ends up triggering the BUG_ON condition in 'free_msi_irqs'
as there is an IRQ handler for the entry->irq (dev->irq).

Note that this cannot be done using MSI-X as the generic
code does not over-write dev->irq with the MSI-X PIRQ values.

The patch inhibits setting up the IRQ handler if MSI or
MSI-X (for symmetry reasons) code had been called successfully.

P.S.
Xen PCIBack when it sets up the device for the guest consumption
ends up writting 0 to the PCI_COMMAND (see xen_pcibk_reset_device).
XSA-120 addendum patch removed that - however when upstreaming said
addendum we found that it caused issues with qemu upstream. That
has now been fixed in qemu upstream.

This is part of XSA-157

CC: stable@vger.kernel.org
Reviewed-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen/pciback: Return error on XEN_PCI_OP_enable_msix when device has MSI or MSI-X...
Konrad Rzeszutek Wilk [Mon, 2 Nov 2015 23:07:44 +0000 (18:07 -0500)] 
xen/pciback: Return error on XEN_PCI_OP_enable_msix when device has MSI or MSI-X enabled

The guest sequence of:

  a) XEN_PCI_OP_enable_msix
  b) XEN_PCI_OP_enable_msix

results in hitting an NULL pointer due to using freed pointers.

The device passed in the guest MUST have MSI-X capability.

The a) constructs and SysFS representation of MSI and MSI groups.
The b) adds a second set of them but adding in to SysFS fails (duplicate entry).
'populate_msi_sysfs' frees the newly allocated msi_irq_groups (note that
in a) pdev->msi_irq_groups is still set) and also free's ALL of the
MSI-X entries of the device (the ones allocated in step a) and b)).

The unwind code: 'free_msi_irqs' deletes all the entries and tries to
delete the pdev->msi_irq_groups (which hasn't been set to NULL).
However the pointers in the SysFS are already freed and we hit an
NULL pointer further on when 'strlen' is attempted on a freed pointer.

The patch adds a simple check in the XEN_PCI_OP_enable_msix to guard
against that. The check for msi_enabled is not stricly neccessary.

This is part of XSA-157

CC: stable@vger.kernel.org
Reviewed-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen/pciback: Return error on XEN_PCI_OP_enable_msi when device has MSI or MSI-X enabled
Konrad Rzeszutek Wilk [Fri, 3 Apr 2015 15:08:22 +0000 (11:08 -0400)] 
xen/pciback: Return error on XEN_PCI_OP_enable_msi when device has MSI or MSI-X enabled

The guest sequence of:

 a) XEN_PCI_OP_enable_msi
 b) XEN_PCI_OP_enable_msi
 c) XEN_PCI_OP_disable_msi

results in hitting an BUG_ON condition in the msi.c code.

The MSI code uses an dev->msi_list to which it adds MSI entries.
Under the above conditions an BUG_ON() can be hit. The device
passed in the guest MUST have MSI capability.

The a) adds the entry to the dev->msi_list and sets msi_enabled.
The b) adds a second entry but adding in to SysFS fails (duplicate entry)
and deletes all of the entries from msi_list and returns (with msi_enabled
is still set).  c) pci_disable_msi passes the msi_enabled checks and hits:

BUG_ON(list_empty(dev_to_msi_list(&dev->dev)));

and blows up.

The patch adds a simple check in the XEN_PCI_OP_enable_msi to guard
against that. The check for msix_enabled is not stricly neccessary.

This is part of XSA-157.

CC: stable@vger.kernel.org
Reviewed-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen/pciback: Save xen_pci_op commands before processing it
Konrad Rzeszutek Wilk [Mon, 16 Nov 2015 17:40:48 +0000 (12:40 -0500)] 
xen/pciback: Save xen_pci_op commands before processing it

Double fetch vulnerabilities that happen when a variable is
fetched twice from shared memory but a security check is only
performed the first time.

The xen_pcibk_do_op function performs a switch statements on the op->cmd
value which is stored in shared memory. Interestingly this can result
in a double fetch vulnerability depending on the performed compiler
optimization.

This patch fixes it by saving the xen_pci_op command before
processing it. We also use 'barrier' to make sure that the
compiler does not perform any optimization.

This is part of XSA155.

CC: stable@vger.kernel.org
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Jan Beulich <JBeulich@suse.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen-scsiback: safely copy requests
David Vrabel [Mon, 16 Nov 2015 18:02:32 +0000 (18:02 +0000)] 
xen-scsiback: safely copy requests

The copy of the ring request was lacking a following barrier(),
potentially allowing the compiler to optimize the copy away.

Use RING_COPY_REQUEST() to ensure the request is copied to local
memory.

This is part of XSA155.

CC: stable@vger.kernel.org
Reviewed-by: Juergen Gross <jgross@suse.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen-blkback: read from indirect descriptors only once
Roger Pau Monné [Tue, 3 Nov 2015 16:40:43 +0000 (16:40 +0000)] 
xen-blkback: read from indirect descriptors only once

Since indirect descriptors are in memory shared with the frontend, the
frontend could alter the first_sect and last_sect values after they have
been validated but before they are recorded in the request.  This may
result in I/O requests that overflow the foreign page, possibly
overwriting local pages when the I/O request is executed.

When parsing indirect descriptors, only read first_sect and last_sect
once.

This is part of XSA155.

CC: stable@vger.kernel.org
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen-blkback: only read request operation from shared ring once
Roger Pau Monné [Tue, 3 Nov 2015 16:34:09 +0000 (16:34 +0000)] 
xen-blkback: only read request operation from shared ring once

A compiler may load a switch statement value multiple times, which could
be bad when the value is in memory shared with the frontend.

When converting a non-native request to a native one, ensure that
src->operation is only loaded once by using READ_ONCE().

This is part of XSA155.

CC: stable@vger.kernel.org
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen-netback: use RING_COPY_REQUEST() throughout
David Vrabel [Fri, 30 Oct 2015 15:17:06 +0000 (15:17 +0000)] 
xen-netback: use RING_COPY_REQUEST() throughout

Instead of open-coding memcpy()s and directly accessing Tx and Rx
requests, use the new RING_COPY_REQUEST() that ensures the local copy
is correct.

This is more than is strictly necessary for guest Rx requests since
only the id and gref fields are used and it is harmless if the
frontend modifies these.

This is part of XSA155.

CC: stable@vger.kernel.org
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
8 years agoxen-netback: don't use last request to determine minimum Tx credit
David Vrabel [Fri, 30 Oct 2015 15:16:01 +0000 (15:16 +0000)] 
xen-netback: don't use last request to determine minimum Tx credit

The last from guest transmitted request gives no indication about the
minimum amount of credit that the guest might need to send a packet
since the last packet might have been a small one.

Instead allow for the worst case 128 KiB packet.

This is part of XSA155.

CC: stable@vger.kernel.org
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
This page took 0.051601 seconds and 5 git commands to generate.