GitHub/LineageOS/android_kernel_motorola_exynos9610.git
9 years agoIB/srp: Dont allocate a page vector when using fast_reg
Sagi Grimberg [Tue, 13 Oct 2015 16:11:41 +0000 (19:11 +0300)]
IB/srp: Dont allocate a page vector when using fast_reg

The new fast registration API does not reuqire a page vector
so we can't avoid allocating it.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Tested-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/srp: Remove srp_finish_mapping
Sagi Grimberg [Tue, 13 Oct 2015 16:11:40 +0000 (19:11 +0300)]
IB/srp: Remove srp_finish_mapping

No callers left, remove it.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Tested-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/srp: Convert to new registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:39 +0000 (19:11 +0300)]
IB/srp: Convert to new registration API

Instead of constructing a page list, call ib_map_mr_sg
and post a new ib_reg_wr. srp_map_finish_fr now returns
the number of sg elements registered.

Remove srp_finish_mapping since no one is calling it.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Tested-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/srp: Split srp_map_sg
Sagi Grimberg [Tue, 13 Oct 2015 16:11:38 +0000 (19:11 +0300)]
IB/srp: Split srp_map_sg

This is a preparation patch for the new registration API
conversion. It splits srp_map_sg per registration strategy
(srp_map_sg[fmr|fr|dma]. On its own it adds some code duplication,
but it makes the API switch easier to comprehend.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Tested-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDS/IW: Convert to new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:37 +0000 (19:11 +0300)]
RDS/IW: Convert to new memory registration API

Get rid of fast_reg page list and its construction.
Instead, just pass the RDS sg list to ib_map_mr_sg
and post the new ib_reg_wr.

This is done both for server IW RDMA_READ registration
and the client remote key registration.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agosvcrdma: Port to new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:36 +0000 (19:11 +0300)]
svcrdma: Port to new memory registration API

Instead of maintaining a fastreg page list, keep an sg table
and convert an array of pages to a sg list. Then call ib_map_mr_sg
and construct ib_reg_wr.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Selvin Xavier <selvin.xavier@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoxprtrdma: Port to new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:35 +0000 (19:11 +0300)]
xprtrdma: Port to new memory registration API

Instead of maintaining a fastreg page list, keep an sg table
and convert an array of pages to a sg list. Then call ib_map_mr_sg
and construct ib_reg_wr.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Selvin Xavier <selvin.xavier@avagotech.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoiser-target: Port to new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:34 +0000 (19:11 +0300)]
iser-target: Port to new memory registration API

Remove fastreg page list allocation as the page vector
is now private to the provider. Instead of constructing
the page list and fast_req work request, call ib_map_mr_sg
and construct ib_reg_wr.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/iser: Port to new fast registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:33 +0000 (19:11 +0300)]
IB/iser: Port to new fast registration API

Remove fastreg page list allocation as the page vector
is now private to the provider. Instead of constructing
the page list and fast_req work request, call ib_map_mr_sg
and construct ib_reg_wr.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/nes: Support the new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:32 +0000 (19:11 +0300)]
RDMA/nes: Support the new memory registration API

Support the new memory registration API by allocating a
private page list array in nes_mr and populate it when
nes_map_mr_sg is invoked. Also, support IB_WR_REG_MR
by duplicating IB_WR_FAST_REG_MR handling and take the
needed information from different places:
- page_size, iova, length (ib_mr)
- page array (nes_mr)
- key, access flags (ib_reg_wr)

The IB_WR_FAST_REG_MR handlers will be removed later when
all the ULPs will be converted.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/qib: Support the new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:31 +0000 (19:11 +0300)]
IB/qib: Support the new memory registration API

Support the new memory registration API by allocating a
private page list array in qib_mr and populate it when
qib_map_mr_sg is invoked. Also, support IB_WR_REG_MR
by duplicating qib_fastreg_mr just take the needed information
from different places:
- page_size, iova, length (ib_mr)
- page array (qib_mr)
- key, access flags (ib_reg_wr)

The IB_WR_FAST_REG_MR handlers will be removed later when
all the ULPs will be converted.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoiw_cxgb4: Support the new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:30 +0000 (19:11 +0300)]
iw_cxgb4: Support the new memory registration API

Support the new memory registration API by allocating a
private page list array in c4iw_mr and populate it when
c4iw_map_mr_sg is invoked. Also, support IB_WR_REG_MR
by duplicating build_fastreg just take the needed information
from different places:
- page_size, iova, length (ib_mr)
- page array (c4iw_mr)
- key, access flags (ib_reg_wr)

The IB_WR_FAST_REG_MR handlers will be removed later when
all the ULPs will be converted.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/cxgb3: Support the new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:29 +0000 (19:11 +0300)]
RDMA/cxgb3: Support the new memory registration API

Support the new memory registration API by allocating a
private page list array in iwch_mr and populate it when
iwch_map_mr_sg is invoked. Also, support IB_WR_REG_MR
by duplicating build_fastreg just take the needed information
from different places:
- page_size, iova, length (ib_mr)
- page array (iwch_mr)
- key, access flags (ib_reg_wr)

The IB_WR_FAST_REG_MR handlers will be removed later when
all the ULPs will be converted.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/ocrdma: Support the new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:28 +0000 (19:11 +0300)]
RDMA/ocrdma: Support the new memory registration API

Support the new memory registration API by allocating a
private page list array in ocrdma_mr and populate it when
ocrdma_map_mr_sg is invoked. Also, support IB_WR_REG_MR
by duplicating IB_WR_FAST_REG_MR, but take the needed
information from different places:
- page_size, iova, length, access flags (ib_mr)
- page array (ocrdma_mr)
- key (ib_reg_wr)

The IB_WR_FAST_REG_MR handlers will be removed later when
all the ULPs will be converted.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/mlx4: Support the new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:27 +0000 (19:11 +0300)]
IB/mlx4: Support the new memory registration API

Support the new memory registration API by allocating a
private page list array in mlx4_ib_mr and populate it when
mlx4_ib_map_mr_sg is invoked. Also, support IB_WR_REG_MR
by setting the exact WQE as IB_WR_FAST_REG_MR, just take the
needed information from different places:
- page_size, iova, length, access flags (ib_mr)
- page array (mlx4_ib_mr)
- key (ib_reg_wr)

The IB_WR_FAST_REG_MR handlers will be removed later when
all the ULPs will be converted.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Tested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/mlx5: Support the new memory registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:26 +0000 (19:11 +0300)]
IB/mlx5: Support the new memory registration API

Support the new memory registration API by allocating a
private page list array in mlx5_ib_mr and populate it when
mlx5_ib_map_mr_sg is invoked. Also, support IB_WR_REG_MR
by setting the exact WQE as IB_WR_FAST_REG_MR, just take the
needed information from different places:
- page_size, iova, length, access flags (ib_mr)
- page array (mlx5_ib_mr)
- key (ib_reg_wr)

The IB_WR_FAST_REG_MR handlers will be removed later when
all the ULPs will be converted.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/mlx5: Remove dead fmr code
Sagi Grimberg [Tue, 13 Oct 2015 16:11:25 +0000 (19:11 +0300)]
IB/mlx5: Remove dead fmr code

Just function declarations - no need for those
laying arround. If for some reason someone will want
FMR support in mlx5, it should be easy enough to restore
a few structs.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Introduce new fast registration API
Sagi Grimberg [Tue, 13 Oct 2015 16:11:24 +0000 (19:11 +0300)]
IB/core: Introduce new fast registration API

The new fast registration  verb ib_map_mr_sg receives a scatterlist
and converts it to a page list under the verbs API thus hiding
the specific HW mapping details away from the consumer.

The provider drivers are provided with a generic helper ib_sg_to_pages
that converts a scatterlist into a vector of page addresses. The
drivers can still perform any HW specific page address setting
by passing a set_page function pointer which will be invoked for
each page address. This allows drivers to avoid keeping a shadow
page vectors and convert them to HW specific translations by doing
extra copies.

This API will allow ULPs to remove the duplicated code of constructing
a page vector from a given sg list.

The send work request ib_reg_wr also shrinks as it will contain only
mr, key and access flags in addition.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Tested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoMerge branch 'wr-cleanup' into k.o/for-4.4
Doug Ledford [Thu, 29 Oct 2015 02:23:34 +0000 (22:23 -0400)]
Merge branch 'wr-cleanup' into k.o/for-4.4

9 years agoMerge branch 'wr-cleanup' of git://git.infradead.org/users/hch/rdma into wr-cleanup
Doug Ledford [Thu, 29 Oct 2015 02:21:09 +0000 (22:21 -0400)]
Merge branch 'wr-cleanup' of git://git.infradead.org/users/hch/rdma into wr-cleanup

Signed-off-by: Doug Ledford <dledford@redhat.com>
Conflicts:
drivers/infiniband/ulp/isert/ib_isert.c - Commit 4366b19ca5eb
(iser-target: Change the recv buffers posting logic) changed the
logic in isert_put_datain() and had to be hand merged

9 years agoIB/ucma: Take the network namespace from the process
Guy Shapiro [Thu, 22 Oct 2015 12:20:11 +0000 (15:20 +0300)]
IB/ucma: Take the network namespace from the process

Add support for network namespaces from user space. This is done by passing
the network namespace of the process instead of init_net.

Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Yotam Kenneth <yotamke@mellanox.com>
Signed-off-by: Shachar Raindel <raindel@mellanox.com>
Signed-off-by: Guy Shapiro <guysh@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/cma: Add support for network namespaces
Guy Shapiro [Thu, 22 Oct 2015 12:20:10 +0000 (15:20 +0300)]
IB/cma: Add support for network namespaces

Add support for network namespaces in the ib_cma module. This is
accomplished by:

1. Adding network namespace parameter for rdma_create_id. This parameter is
   used to populate the network namespace field in rdma_id_private.
   rdma_create_id keeps a reference on the network namespace.
2. Using the network namespace from the rdma_id instead of init_net inside
   of ib_cma, when listening on an ID and when looking for an ID for an
   incoming request.
3. Decrementing the reference count for the appropriate network namespace
   when calling rdma_destroy_id.

In order to preserve the current behavior init_net is passed when calling
from other modules.

Signed-off-by: Guy Shapiro <guysh@mellanox.com>
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Yotam Kenneth <yotamke@mellanox.com>
Signed-off-by: Shachar Raindel <raindel@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/cma: Separate port allocation to network namespaces
Haggai Eran [Thu, 22 Oct 2015 12:20:09 +0000 (15:20 +0300)]
IB/cma: Separate port allocation to network namespaces

Keep a struct for each network namespace containing the IDRs for the RDMA
CM port spaces. The struct is created dynamically using the generic_net
mechanism.

This patch is internal infrastructure work for the following patches. In
this patch, init_net is statically used as the network namespace for
the new port-space API.

Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Yotam Kenneth <yotamke@mellanox.com>
Signed-off-by: Shachar Raindel <raindel@mellanox.com>
Signed-off-by: Guy Shapiro <guysh@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/addr: Pass network namespace as a parameter
Guy Shapiro [Thu, 22 Oct 2015 12:20:08 +0000 (15:20 +0300)]
IB/addr: Pass network namespace as a parameter

Add network namespace support to the ib_addr module. For that, all the
address resolution and matching should be done using the appropriate
namespace instead of init_net.

This is achieved by:

1. Adding an explicit network namespace argument to exported function that
   require a namespace.
2. Saving the namespace in the rdma_addr_client structure.
3. Using it when calling networking functions.

In order to preserve the behavior of calling modules, &init_net is
passed as the parameter in calls from other modules. This is modified as
namespace support is added on more levels.

Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Yotam Kenneth <yotamke@mellanox.com>
Signed-off-by: Shachar Raindel <raindel@mellanox.com>
Signed-off-by: Guy Shapiro <guysh@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/iser: Enable SG clustering
Sagi Grimberg [Tue, 13 Oct 2015 16:12:59 +0000 (19:12 +0300)]
IB/iser: Enable SG clustering

iser is perfectly capable supporting SG clustering as it translates
the SG list to a page vector. Enabling SG clustering can dramatically
reduce the number of SG elements, which doesn't make much of a difference
at this point, but with arbitrary SG list support, reducing the
number of SG elements can benefit greatly as as it would reduce
the length of the HW descriptors array.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/iser: set block queue_virt_boundary
Sagi Grimberg [Tue, 13 Oct 2015 16:12:58 +0000 (19:12 +0300)]
IB/iser: set block queue_virt_boundary

The block layer can reliably guarantee that SG lists won't
contain gaps (page unaligned) if a driver set the queue
virt_boundary.

With this setting the block layer will:
- refuse merges if bios are not aligned to the virtual boundary
- split bios/requests that are not aligned to the virtual boundary
- or, bounce buffer SG_IOs that are not aligned to the virtual boundary

Since iser is working in 4K page size, set the virt_boundary to
4K pages. With this setting, we can now safely remove the bounce
buffering logic in iser.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoiser-target: Remove an unused variable
Bart Van Assche [Thu, 22 Oct 2015 18:14:18 +0000 (11:14 -0700)]
iser-target: Remove an unused variable

Detected this by compiling with W=1.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/iser: Remove an unused variable
Bart Van Assche [Thu, 22 Oct 2015 17:59:18 +0000 (10:59 -0700)]
IB/iser: Remove an unused variable

Detected this by compiling with W=1.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Remove smac and vlan id from path record
Matan Barak [Thu, 15 Oct 2015 15:38:54 +0000 (18:38 +0300)]
IB/core: Remove smac and vlan id from path record

The GID cache accompanies every GID with attributes.
The GID attributes link the GID with its netdevice, which could be
resolved to smac and vlan id easily. Since we've added the netdevice
(ifindex and net) to the path record, storing the L2 attributes is
duplicated data and hence these attributes are removed.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Remove smac and vlan id from qp_attr and ah_attr
Matan Barak [Thu, 15 Oct 2015 15:38:53 +0000 (18:38 +0300)]
IB/core: Remove smac and vlan id from qp_attr and ah_attr

Smac and vlan id could be resolved from the GID attribute, and thus
these attributes aren't needed anymore. Removing them.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/cm: Remove the usage of smac and vid of qp_attr and cm_av
Matan Barak [Thu, 15 Oct 2015 15:38:52 +0000 (18:38 +0300)]
IB/cm: Remove the usage of smac and vid of qp_attr and cm_av

The cm and cma don't need to explicitly handle vlan and smac,
as they are resolved from the GID index now. Removing this
portion of code.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Use GID table in AH creation and dmac resolution
Matan Barak [Thu, 15 Oct 2015 15:38:51 +0000 (18:38 +0300)]
IB/core: Use GID table in AH creation and dmac resolution

Previously, vlan id and source MAC were used from QP attributes. Since
the net device is now stored in the GID attributes, they could be used
instead of getting this information from the QP attributes.

IB_QP_SMAC, IB_QP_ALT_SMAC, IB_QP_VID and IB_QP_ALT_VID were removed
because there is no known libibverbs that uses them.

This commit also modifies the vendors (mlx4, ocrdma) drivers in order
to use the new approach.

ocrdma driver changes were done by Somnath Kotur <Somnath.Kotur@Avagotech.Com>

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/cache: Add ib_find_gid_by_filter cache API
Matan Barak [Thu, 15 Oct 2015 15:38:50 +0000 (18:38 +0300)]
IB/cache: Add ib_find_gid_by_filter cache API

GID cache API users might want to search for GIDs with specific
attributes rather than just specifying GID, net device and port.
This is used in a later patch, where we find the sgid index by
L2 Ethernet attributes.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/cma: cma_validate_port should verify the port and netdevice
Matan Barak [Thu, 15 Oct 2015 15:38:49 +0000 (18:38 +0300)]
IB/cma: cma_validate_port should verify the port and netdevice

Previously, cma_validate_port searched for GIDs in IB cache and then
tried to verify the found port. This could fail when there are
identical GIDs on both ports. In addition, netdevice should be taken
into account when searching the GID table.
Fixing cma_validate_port to search only the relevant port's cache
and netdevice.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/cm: cm_init_av_by_path should find a GID by its netdevice
Matan Barak [Thu, 15 Oct 2015 15:38:48 +0000 (18:38 +0300)]
IB/cm: cm_init_av_by_path should find a GID by its netdevice

Previously, the CM has searched the cache for any sgid_index whose
GID matches the path's GID. Since the path record stores the net
device, the CM should now search only for GIDs which originated from
this net device.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Add netdev to path record
Matan Barak [Thu, 15 Oct 2015 15:38:47 +0000 (18:38 +0300)]
IB/core: Add netdev to path record

In order to find the sgid_index, one could just query the IB cache
with the correct GID and netdevice. Therefore, instead of storing
the L2 attributes directly in the path, we only store the
ifindex and net and use them later to get the sgid_index.
The vlan_id and smac L2 attributes are removed in a later patch.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Expose and rename ib_find_cached_gid_by_port cache API
Matan Barak [Thu, 15 Oct 2015 15:38:46 +0000 (18:38 +0300)]
IB/core: Expose and rename ib_find_cached_gid_by_port cache API

Sometime consumers might want to search for a GID in a specific port.
For example, when a WC arrives and we want to search the GID
that matches that port - it's better to search only the relevant
port.
Exposing and renaming ib_cache_gid_find_by_port in order to match
the naming convention of the module.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Add netdev and gid attributes paramteres to cache
Matan Barak [Thu, 15 Oct 2015 15:38:45 +0000 (18:38 +0300)]
IB/core: Add netdev and gid attributes paramteres to cache

Adding an ability to query the IB cache by a netdev and get the
attributes of a GID. These parameters are necessary in order to
successfully resolve the required GID (when the netdevice is known)
and get the Ethernet L2 attributes from a GID.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Reviewed-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/mlx4: Add support for blocking multicast loopback QP creation user flag
Eran Ben Elisha [Thu, 15 Oct 2015 11:44:42 +0000 (14:44 +0300)]
IB/mlx4: Add support for blocking multicast loopback QP creation user flag

MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK is now supported downstream.

In addition, this flag was supported only for IB_QPT_UD, now, with the
new implementation it is supported for all QP types.

Support IB_USER_VERBS_EX_CMD_CREATE_QP in order to get the flag from
user space using the extension create qp command.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/mlx4: Add counter based implementation for QP multicast loopback block
Eran Ben Elisha [Thu, 15 Oct 2015 11:44:41 +0000 (14:44 +0300)]
IB/mlx4: Add counter based implementation for QP multicast loopback block

Current implementation for MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK is not
supported when link layer is Ethernet.

This patch will add counter based implementation for multicast loopback
prevention. HW can drop multicast loopback packets if sender QP counter
index is equal to receiver QP counter index. If qp flag
MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK is set and link layer is Ethernet,
create a new counter and attach it to the QP so it will continue
receiving multicast loopback traffic but it's own.

The decision if to create a new counter is being made at the qp
modification to RTR after the QP's port is set. When QP is destroyed or
moved back to reset state, delete the counter.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/mlx4: Add IB counters table
Eran Ben Elisha [Thu, 15 Oct 2015 11:44:40 +0000 (14:44 +0300)]
IB/mlx4: Add IB counters table

This is an infrastructure step for allocating and attaching more than
one counter to QPs on the same port. Allocate a counters table and
manage the insertion and removals of the counters in load and unload of
mlx4 IB.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agonet/mlx4_en: Implement mcast loopback prevention for ETH qps
Maor Gottlieb [Thu, 15 Oct 2015 11:44:39 +0000 (14:44 +0300)]
net/mlx4_en: Implement mcast loopback prevention for ETH qps

Set the mcast loopback prevention bit in the QPC for ETH MLX QPs (not
RSS QPs), when the firmware supports this feature. In addition, all rx
ring QPs need to be updated in order not to enforce loopback checks.
This prevents getting packets we sent both from the network stack and
the HCA. Loopback prevention is done by comparing the counter indices of
the sent and receiving QPs. If they're equal, packets aren't
loopback-ed.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agonet/mlx4_core: Add support for filtering multicast loopback
Maor Gottlieb [Thu, 15 Oct 2015 11:44:38 +0000 (14:44 +0300)]
net/mlx4_core: Add support for filtering multicast loopback

Update device capabilities regarding HW filtering multicast loopback support.

Add MLX4_UPDATE_QP_ETH_SRC_CHECK_MC_LB attribute to mlx4_update_qp to
enable changing QP context to support filtering incoming multicast
loopback traffic according the sender's counter index.

Set the corresponding bits in QP context to force the loopback source
checks if attribute is given and HW supports it.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Allow setting create flags in QP init attribute
Eran Ben Elisha [Thu, 15 Oct 2015 11:44:37 +0000 (14:44 +0300)]
IB/core: Allow setting create flags in QP init attribute

Allow setting IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK at create_flags in
ib_uverbs_create_qp_ex.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Extend ib_uverbs_create_qp
Eran Ben Elisha [Wed, 21 Oct 2015 14:00:42 +0000 (17:00 +0300)]
IB/core: Extend ib_uverbs_create_qp

ib_uverbs_ex_create_qp follows the extension verbs
mechanism. New features (for example, QP creation flags
field which is added in a downstream patch) could used
via user-space libraries without breaking the ABI.

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reviewed-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoiw_cxgb4: Adds support for T6 adapter
Hariprasad S [Wed, 23 Sep 2015 11:49:27 +0000 (17:19 +0530)]
iw_cxgb4: Adds support for T6 adapter

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agocxgb4: T6 adapter lld support for iw_cxgb4 driver
Hariprasad S [Wed, 23 Sep 2015 11:49:26 +0000 (17:19 +0530)]
cxgb4: T6 adapter lld support for iw_cxgb4 driver

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/ocrdma: Bump up ocrdma version number to 11.0.0.0
Selvin Xavier [Tue, 20 Oct 2015 08:48:00 +0000 (14:18 +0530)]
RDMA/ocrdma: Bump up ocrdma version number to 11.0.0.0

Updating the version number to 11.0.0.0

Signed-off-by: Selvin Xavier <selvin.xavier@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/ocrdma: Prevent CQ-Doorbell floods
Devesh Sharma [Tue, 20 Oct 2015 08:47:59 +0000 (14:17 +0530)]
RDMA/ocrdma: Prevent CQ-Doorbell floods

Changing CQ-Doorbell(DB) logic to prevent DB floods, it is supposed to be
pressed only if any hw CQE is polled. If cq-arm was requested
previously then don't bother about number of hw CQEs polled and
arm the CQ.

Signed-off-by: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Selvin Xavier <selvin.xavier@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/ocrdma: Check resource ids received in Async CQE
Naga Irrinki [Tue, 20 Oct 2015 08:47:58 +0000 (14:17 +0530)]
RDMA/ocrdma: Check resource ids received in Async CQE

Some versions of the FW sends wrong QP or CQ IDs in the
Async CQE. Adding a check to see whether qp or cq structures
associated with the CQE is valid.

Signed-off-by: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Selvin Xavier <selvin.xavier@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/ocrdma: Avoid a possible crash in ocrdma_rem_port_stats
Selvin Xavier [Tue, 20 Oct 2015 08:47:57 +0000 (14:17 +0530)]
RDMA/ocrdma: Avoid a possible crash in ocrdma_rem_port_stats

debugfs_remove should be called before freeing the driver
stats resources to avoid any crash during ocrdma_remove.

Signed-off-by: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Selvin Xavier <selvin.xavier@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/ocrdma: Cleanup unused device list and rcu variables
Selvin Xavier [Tue, 20 Oct 2015 08:47:56 +0000 (14:17 +0530)]
RDMA/ocrdma: Cleanup unused device list and rcu variables

ocrdma_dev_list is not used by the driver. So removing
the references of this variable. dev->rcu was introduced
for the ipv6 notifier for GID management. This is no longer
required as the GID management is outside the HW driver.

Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@avagotech.com>
Signed-off-by: Selvin Xavier <selvin.xavier@avagotech.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoiw_cxgb4: reverse the ord/ird in the ESTABLISHED upcall
Hariprasad S [Tue, 8 Sep 2015 04:27:00 +0000 (09:57 +0530)]
iw_cxgb4: reverse the ord/ird in the ESTABLISHED upcall

The ESTABLISHED event should have the peer's ord/ird so
swap the values in the event before the upcall.

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoiw_cxgb4: fix misuse of ep->ord for minimum ird calculation
Hariprasad S [Tue, 8 Sep 2015 04:26:59 +0000 (09:56 +0530)]
iw_cxgb4: fix misuse of ep->ord for minimum ird calculation

When calculating the minimum ird in c4iw_accept_cr(), we need to always
have a value of at least 1 if the RTR message is a 0B read.  The code
was
incorrectly using ep->ord for this logic which was incorrectly adjusting
the ird and causing incorrect ord/ird negotiation when using MPAv2 to
negotiate these values.

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoiw_cxgb4: pass the ord/ird in connect reply events
Hariprasad S [Tue, 8 Sep 2015 04:26:58 +0000 (09:56 +0530)]
iw_cxgb4: pass the ord/ird in connect reply events

This allows client ULPs to get the negotiated ord/ird which is useful
to avoid stalling the SQ due to exceeding the ORD.

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoiw_cxgb4: detect fatal errors while creating listening filters
Hariprasad S [Tue, 8 Sep 2015 04:26:57 +0000 (09:56 +0530)]
iw_cxgb4: detect fatal errors while creating listening filters

In c4iw_create_listen(), if we're using listen filters, then bail out
of the busy loop if the device becomes fatally dead

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: avoid 32-bit warning
Arnd Bergmann [Wed, 7 Oct 2015 12:29:51 +0000 (14:29 +0200)]
IB/core: avoid 32-bit warning

The INIT_UDATA() macro requires a pointer or unsigned long argument for
both input and output buffer, and all callers had a cast from when
the code was merged until a recent restructuring, so now we get

core/uverbs_cmd.c: In function 'ib_uverbs_create_cq':
core/uverbs_cmd.c:1481:66: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]

This makes the code behave as before by adding back the cast to
unsigned long.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 565197dd8fb1 ("IB/core: Extend ib_uverbs_create_cq")
Reviewed-by: Yann Droneaud <ydroneaud@opteya.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoRDMA/cxgb4: re-fix 32-bit build warning
Arnd Bergmann [Wed, 7 Oct 2015 12:10:04 +0000 (14:10 +0200)]
RDMA/cxgb4: re-fix 32-bit build warning

Casting a pointer to __be64 produces a warning on 32-bit architectures:

drivers/infiniband/hw/cxgb4/mem.c:147:20: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
    req->wr.wr_lo = (__force __be64)&wr_wait;

This was fixed at least twice for this driver in different places,
and accidentally reverted once more. This puts the correct version
back in place.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 6198dd8d7a6a7 ("iw_cxgb4: 32b platform fixes")
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoib_pack.h: Fix commentary IBA reference for CNP in IB opcode enum
Hal Rosenstock [Wed, 30 Sep 2015 19:04:29 +0000 (15:04 -0400)]
ib_pack.h: Fix commentary IBA reference for CNP in IB opcode enum

IBA spec is now 1.3 not 3.1 and vol 1 should be mentioned
as there is also vol 2.

Signed-off-by: Hal Rosenstock <hal@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/iser: fix a comment typo
Geliang Tang [Sun, 4 Oct 2015 09:00:05 +0000 (17:00 +0800)]
IB/iser: fix a comment typo

Just fix a typo in the code comment.

Signed-off-by: Geliang Tang <geliangtang@163.com>
Acked-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agousnic: correctly handle kzalloc return value
Insu Yun [Mon, 19 Oct 2015 16:57:10 +0000 (16:57 +0000)]
usnic: correctly handle kzalloc return value

Since kzalloc returns memory address, not error code,
it should be checked whether it is null or not.

Signed-off-by: Insu Yun <wuninsu@gmail.com>
Reviewed-by: Dave Goodell <dgoodell@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoMAINTAINERS: update usnic maintainer contacts
Dave Goodell \(dgoodell\) [Fri, 16 Oct 2015 03:01:29 +0000 (20:01 -0700)]
MAINTAINERS: update usnic maintainer contacts

Upinder hasn't worked for Cisco for a little while now, updating to more
active maintainers.

Signed-off-by: Dave Goodell <dgoodell@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agousnic: correctly check failed allocation
Insu Yun [Thu, 15 Oct 2015 21:15:15 +0000 (21:15 +0000)]
usnic: correctly check failed allocation

Since ib_alloc_device returns allocated memory address, not error,
it should be checked as IS_NULL, not IS_ERR_OR_NULL.

Signed-off-by: Insu Yun <wuninsu@gmail.com>
Reviewed-by: Dave Goodell <dgoodell@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoMerge branch 'k.o/for-4.3-v1' into k.o/for-4.4
Doug Ledford [Wed, 21 Oct 2015 20:40:21 +0000 (16:40 -0400)]
Merge branch 'k.o/for-4.3-v1' into k.o/for-4.4

Pick up the late fixes from the 4.3 cycle so we have them in our
next branch.

9 years agoIB/cm: Fix rb-tree duplicate free and use-after-free
Doron Tsur [Sun, 11 Oct 2015 12:58:17 +0000 (15:58 +0300)]
IB/cm: Fix rb-tree duplicate free and use-after-free

ib_send_cm_sidr_rep could sometimes erase the node from the sidr
(depending on errors in the process). Since ib_send_cm_sidr_rep is
called both from cm_sidr_req_handler and cm_destroy_id, cm_id_priv
could be either erased from the rb_tree twice or not erased at all.
Fixing that by making sure it's erased only once before freeing
cm_id_priv.

Fixes: a977049dacde ('[PATCH] IB: Add the kernel CM implementation')
Signed-off-by: Doron Tsur <doront@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/cma: Use inner P_Key to determine netdev
Haggai Eran [Tue, 20 Oct 2015 06:53:01 +0000 (09:53 +0300)]
IB/cma: Use inner P_Key to determine netdev

When discussing the patches to demux ids in rdma_cm instead of ib_cm, it
was decided that it is best to use the P_Key value in the packet headers.
However, the mlx5 and ipath drivers are currently unable to send correct
P_Key values in GMP headers. They always send using a single P_Key that is
set during the GSI QP initialization.

Change the rdma_cm code to look at the P_Key value that is part of the
packet payload as a workaround. Once the drivers are fixed this patch can
be reverted.

Fixes: 4c21b5bcef73 ("IB/cma: Add net_dev and private data checks to
RDMA CM")
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/ucma: check workqueue allocation before usage
Sasha Levin [Thu, 17 Sep 2015 20:04:19 +0000 (16:04 -0400)]
IB/ucma: check workqueue allocation before usage

Allocating a workqueue might fail, which wasn't checked so far and would
lead to NULL ptr derefs when an attempt to use it was made.

Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/cma: Potential NULL dereference in cma_id_from_event
Haggai Eran [Mon, 21 Sep 2015 13:02:02 +0000 (16:02 +0300)]
IB/cma: Potential NULL dereference in cma_id_from_event

If the lookup of a listening ID failed for an AF_IB request, the code
would try to call dev_put() on a NULL net_dev.

Fixes: be688195bd08 ("IB/cma: Fix net_dev reference leak with failed
requests")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Fix use after free of ifa
Matan Barak [Thu, 15 Oct 2015 12:01:03 +0000 (15:01 +0300)]
IB/core: Fix use after free of ifa

When using ifup/ifdown while executing enum_netdev_ipv4_ips,
ifa could become invalid and cause use after free error.
Fixing it by protecting with RCU lock.

Fixes: 03db3a2d81e6 ('IB/core: Add RoCE GID table management')
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Fix memory corruption in ib_cache_gid_set_default_gid
Doron Tsur [Thu, 15 Oct 2015 12:01:02 +0000 (15:01 +0300)]
IB/core: Fix memory corruption in ib_cache_gid_set_default_gid

When ib_cache_gid_set_default_gid is called from several threads,
updating the table could make find_gid fail, therefore a negative
index will be retruned and an invalid table entry will be used.
Locking find_gid as well fixes this problem.

Fixes: 03db3a2d81e6 ('IB/core: Add RoCE GID table management')
Signed-off-by: Doron Tsur <doront@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/ipoib: For sendonly join free the multicast group on leave
Christoph Lameter [Sun, 11 Oct 2015 23:49:42 +0000 (18:49 -0500)]
IB/ipoib: For sendonly join free the multicast group on leave

When we leave the multicast group on expiration of a neighbor we
do not free the mcast structure. This results in a memory leak
that causes ib_dealloc_pd to fail and print a WARN_ON message
and backtrace.

Fixes: bd99b2e05c4d (IB/ipoib: Expire sendonly multicast joins)
Signed-off-by: Christoph Lameter <cl@linux.com>
Tested-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB: remove xrc_remote_srq_num from struct ib_send_wr
Christoph Hellwig [Wed, 19 Aug 2015 12:58:43 +0000 (14:58 +0200)]
IB: remove xrc_remote_srq_num from struct ib_send_wr

The field is only initialized in mlx, but never used.

If we want to add proper XRC support it should be done with a new
struct ib_xrc_wr.

This shrinks the various WR structures by another 4 bytes.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Tested-by: Haggai Eran <haggaie@mellanox.com>
9 years agoIB: split struct ib_send_wr
Christoph Hellwig [Thu, 8 Oct 2015 08:16:33 +0000 (09:16 +0100)]
IB: split struct ib_send_wr

This patch split up struct ib_send_wr so that all non-trivial verbs
use their own structure which embedds struct ib_send_wr.  This dramaticly
shrinks the size of a WR for most common operations:

sizeof(struct ib_send_wr) (old): 96

sizeof(struct ib_send_wr): 48
sizeof(struct ib_rdma_wr): 64
sizeof(struct ib_atomic_wr): 96
sizeof(struct ib_ud_wr): 88
sizeof(struct ib_fast_reg_wr): 88
sizeof(struct ib_bind_mw_wr): 96
sizeof(struct ib_sig_handover_wr): 80

And with Sagi's pending MR rework the fast registration WR will also be
down to a reasonable size:

sizeof(struct ib_fastreg_wr): 64

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> [srp, srpt]
Reviewed-by: Chuck Lever <chuck.lever@oracle.com> [sunrpc]
Tested-by: Haggai Eran <haggaie@mellanox.com>
Tested-by: Sagi Grimberg <sagig@mellanox.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
9 years agoIB/cma: Accept connection without a valid netdev on RoCE
Haggai Eran [Tue, 6 Oct 2015 14:56:00 +0000 (17:56 +0300)]
IB/cma: Accept connection without a valid netdev on RoCE

The netdev checks recently added to RDMA CM expect a valid netdev to be
found for both InfiniBand and RoCE, but the code that find a netdev is
only implemented for InfiniBand.

Currently RoCE doesn't provide an API to find the netdev matching a
given set of parameters, so this patch just disables the netdev enforcement
for each incoming connections when the link layer is RoCE.

Fixes: 4c21b5bcef73 ("IB/cma: Add net_dev and private data checks to RDMA CM")
Reported-by: Kamal Heib <kamalh@mellanox.com>
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoxprtrdma: Don't require LOCAL_DMA_LKEY support for fastreg
Sagi Grimberg [Tue, 6 Oct 2015 16:52:37 +0000 (19:52 +0300)]
xprtrdma: Don't require LOCAL_DMA_LKEY support for fastreg

There is no need to require LOCAL_DMA_LKEY support as the
PD allocation makes sure that there is a local_dma_lkey. Also
correctly set a return value in error path.

This caused a NULL pointer dereference in mlx5 which removed
the support for LOCAL_DMA_LKEY.

Fixes: bb6c96d72879 ("xprtrdma: Replace global lkey with lkey local to PD")
Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agousnic: add missing clauses to BSD license
Jeff Squyres [Wed, 30 Sep 2015 20:34:00 +0000 (13:34 -0700)]
usnic: add missing clauses to BSD license

The usnic_verbs kernel module was clearly marked with the following in
its code:

  MODULE_LICENSE("Dual BSD/GPL");

However, we accidentally left a few clauses of the BSD text out of the
license header in all the source files.  This commit fixes that: all
the files are properly dual BSD/GPL-licensed.  Contributors that might
have been confused by this have been contacted to get their permission
and are Cc:ed here.

Cc: Benoit Taine <benoit.taine@lip6.fr>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Joerg Roedel <jroedel@suse.de>
Cc: Masanari Iida <standby24x7@gmail.com>
Cc: Matan Barak <matanb@mellanox.com>
Cc: Michael Wang <yun.wang@profitbricks.com>
Cc: Roland Dreier <roland@purestorage.com>
Cc: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Fabio Estevam <fabio.estevam@freescale.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/mlx4: Report checksum offload cap for RAW QP when query device
Bodong Wang [Tue, 22 Sep 2015 20:18:11 +0000 (23:18 +0300)]
IB/mlx4: Report checksum offload cap for RAW QP when query device

Signed-off-by: Bodong Wang <bodong@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoIB/core: Add support of checksum capability reporting for RC and RAW
Bodong Wang [Tue, 22 Sep 2015 20:18:10 +0000 (23:18 +0300)]
IB/core: Add support of checksum capability reporting for RC and RAW

Two enum members IB_DEVICE_RC_IP_CSUM and IB_DEVICE_RAW_IP_CSUM are
added to ib_device_cap_flags. Device should set these two flags if they support
insertion of UDP and TCP checksum on outgoing IPv4 messages and can verify the
validity of checksum for incoming IPv4 messages, for RC IPoIB and RAW over
Ethernet respectively. They are similar to IB_DEVICE_UD_IP_CSUM.

Signed-off-by: Bodong Wang <bodong@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoLinux 4.3-rc3
Linus Torvalds [Sun, 27 Sep 2015 11:50:08 +0000 (07:50 -0400)]
Linux 4.3-rc3

9 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 27 Sep 2015 10:51:42 +0000 (06:51 -0400)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Thomas Gleixner:
 "Two bugfixes from Andy addressing at least some of the subtle NMI
  related wreckage which has been reported by Sasha Levin"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/nmi/64: Fix a paravirt stack-clobbering bug in the NMI code
  x86/paravirt: Replace the paravirt nop with a bona fide empty function

9 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 27 Sep 2015 10:50:23 +0000 (06:50 -0400)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fix from Thomass Gleixner:
 "A bugfix for the atmel aic5 irq chip driver which caches the wrong
  data and thereby breaking resume"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/atmel-aic5: Use per chip mask caches in mask/unmask()

9 years agoMerge branch 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm
Linus Torvalds [Sun, 27 Sep 2015 10:48:48 +0000 (06:48 -0400)]
Merge branch 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm

Pull ARM fixes from Russell King:
 "Just two fixes: wire up the new system calls added during the last
  merge window, and fix another user access site"

* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
  ARM: alignment: fix alignment handling for uaccess changes
  ARM: wire up new syscalls

9 years agoMerge tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Linus Torvalds [Sun, 27 Sep 2015 10:45:18 +0000 (06:45 -0400)]
Merge tag 'armsoc-fixes' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "Our first real batch of fixes this release cycle.  Nothing really
  concerning, and diffstat is a bit inflated due to some DT contents
  moving around on STi platforms.

  There's a collection of them here:

   - A fixup for a build breakage that hits on arm64 allmodconfig in
     QCOM SCM firmware drivers
   - MMC fixes for OMAP that had quite a bit of breakage this merge
     window.
   - Misc build/warning fixes on PXA and OMAP
   - A couple of minor fixes for Beagleboard X15 which is now starting
     to see a few more users in the wild"

* tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (31 commits)
  ARM: sti: dt: adapt DT to fix probe/bind issues in DRM driver
  ARM: dts: fix omap2+ address translation for pbias
  firmware: qcom: scm: Add function stubs for ARM64
  ARM: dts: am57xx-beagle-x15: use palmas-usb for USB2
  ARM: omap2plus_defconfig: enable GPIO_PCA953X
  ARM: dts: omap5-uevm.dts: fix i2c5 pinctrl offsets
  ARM: OMAP2+: AM43XX: Enable autoidle for clks in am43xx_init_late
  ARM: dts: am57xx-beagle-x15: Update Phy supplies
  ARM: pxa: balloon3: Fix build error
  ARM: dts: Fixup model name for HP t410 dts
  ARM: dts: DRA7: fix a typo in ethernet
  ARM: omap2plus_defconfig: make PCF857x built-in
  ARM: dts: Use ti,pbias compatible string for pbias
  ARM: OMAP5: Cleanup options for SoC only build
  ARM: DRA7: Select missing options for SoC only build
  ARM: OMAP2+: board-generic: Remove stale of_irq macros
  ARM: OMAP4+: PM: erratum is used by OMAP5 and DRA7 as well
  ARM: dts: omap3-igep: Move eth IRQ pinmux to IGEPv2 common dtsi
  ARM: dts: am57xx-beagle-x15: Add wakeup irq for mcp79410
  ARM: dts: am335x-phycore-som: Fix mpu voltage
  ...

9 years agoMerge branch 'for-next' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sun, 27 Sep 2015 10:42:54 +0000 (06:42 -0400)]
Merge branch 'for-next' of git://git.samba.org/sfrench/cifs-2.6

Pull CIFS fixes from Steve French:
 "Four fixes from testing at the recent SMB3 Plugfest including two
  important authentication ones (one fixes authentication problems to
  some popular servers when clock times differ more than two hours
  between systems, the other fixes Kerberos authentication for SMB3)"

* 'for-next' of git://git.samba.org/sfrench/cifs-2.6:
  fix encryption error checks on mount
  [SMB3] Fix sec=krb5 on smb3 mounts
  cifs: use server timestamp for ntlmv2 authentication
  disabling oplocks/leases via module parm enable_oplocks broken for SMB3

9 years agoMerge tag 'pxa-fixes-v4.3' of https://github.com/rjarzmik/linux into fixes
Olof Johansson [Sun, 27 Sep 2015 05:23:26 +0000 (22:23 -0700)]
Merge tag 'pxa-fixes-v4.3' of https://github.com/rjarzmik/linux into fixes

ARM: pxa: fixes for v4.3

These fixes are mainly regression fixes triggered by irq changes,
common clock framework introduction and sound side-effect of
other platforms.

* tag 'pxa-fixes-v4.3' of https://github.com/rjarzmik/linux:
  ARM: pxa: balloon3: Fix build error
  ARM: pxa: ssp: Fix build error by removing originally incorrect DT binding
  ARM: pxa: fix DFI bus lockups on startup

Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agoMerge tag 'omap-for-v4.3/fixes-rc2' of git://git.kernel.org/pub/scm/linux/kernel...
Olof Johansson [Sun, 27 Sep 2015 05:22:31 +0000 (22:22 -0700)]
Merge tag 'omap-for-v4.3/fixes-rc2' of git://git./linux/kernel/git/tmlind/linux-omap into fixes

Fixes for omaps for v4.3-rc cycle:

- Two more patches to fix most of the MMC regressions with the
  PBIAS regulator changes. At least two MMC driver related issues
  still seems to remain for omap3 legacy booting and omap4 duovero.
  Note that the dts changes depend on a recent regulator fix, and
  are based on the regulator commit now in mainline kernel

- Enable autoidle for am43xx clocks to prevent clocks from staying
  always on

- Fix i2c5 pinctrl offsets for omap5-uevm

- Enable PCA953X as that's needed for HDMI to work on omap5

- Update phy supplies for beagle x15 beta board

- Use palmas-usb for on beagle x15 to start using the related
  driver that recently got merged

* tag 'omap-for-v4.3/fixes-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap:
  ARM: dts: fix omap2+ address translation for pbias
  ARM: dts: am57xx-beagle-x15: use palmas-usb for USB2
  ARM: omap2plus_defconfig: enable GPIO_PCA953X
  ARM: dts: omap5-uevm.dts: fix i2c5 pinctrl offsets
  ARM: OMAP2+: AM43XX: Enable autoidle for clks in am43xx_init_late
  ARM: dts: am57xx-beagle-x15: Update Phy supplies
  regulator: pbias: program pbias register offset in pbias driver
  ARM: omap2plus_defconfig: Enable MUSB DMA support
  ARM: DRA752: Add ID detect for ES2.0
  ARM: OMAP3: vc: fix 'or' always true warning
  ARM: OMAP2+: Fix booting if no timer parent clock is available
  ARM: OMAP2+: omap-device: fix race deferred probe of omap_hsmmc vs omap_device_late_init

Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Sun, 27 Sep 2015 01:05:23 +0000 (21:05 -0400)]
Merge git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fixes from Herbert Xu:
 "This fixes the following issues:

   - check the return value of platform_get_irq as signed int in xgene.

   - skip adf_dev_restore on virtual functions in qat.

   - fix double-free with backlogged requests in marvell_cesa"

* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  hwrng: xgene - fix handling platform_get_irq
  crypto: qat - VF should never trigger SBR on PH
  crypto: marvell - properly handle CRYPTO_TFM_REQ_MAY_BACKLOG-flagged requests

9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending
Linus Torvalds [Sun, 27 Sep 2015 01:02:42 +0000 (21:02 -0400)]
Merge git://git./linux/kernel/git/nab/target-pending

Pull SCSI target fixes from Nicholas Bellinger:
 "This includes a iser-target series from Jenny + Sagi @ Mellanox that
  addresses the few remaining active I/O shutdown bugs, along with a
  patch to support zero-copy for immediate data payloads that gives a
  nice performance improvement for small block WRITEs.

  Also included are some recent >= v4.2 regression bug-fixes.  The most
  notable is a RCU conversion regression for SPC-3 PR registrations, and
  recent removal of obsolete RFC-3720 markers that introduced a login
  regression bug with MSFT iSCSI initiators.

  Thanks to everyone who has been testing + reporting bugs for v4.x"

* git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  iscsi-target: Avoid OFMarker + IFMarker negotiation
  target: Make TCM_WRITE_PROTECT failure honor D_SENSE bit
  target: Fix target_sense_desc_format NULL pointer dereference
  target: Propigate backend read-only to core_tpg_add_lun
  target: Fix PR registration + APTPL RCU conversion regression
  iser-target: Skip data copy if all the command data comes as immediate
  iser-target: Change the recv buffers posting logic
  iser-target: Fix pending connections handling in target stack shutdown sequnce
  iser-target: Remove np_ prefix from isert_np members
  iser-target: Remove unused variables
  iser-target: Put the reference on commands waiting for unsol data
  iser-target: remove command with state ISTATE_REMOVE

9 years agoMerge tag 'usb-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 27 Sep 2015 01:00:28 +0000 (21:00 -0400)]
Merge tag 'usb-4.3-rc3' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some USB driver fixes for 4.3-rc3.

  There's the usual assortment of new device ids, combined with xhci and
  gadget driver fixes.  Full details in the shortlog.  All of these have
  been in linux-next with no reported problems"

* tag 'usb-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (34 commits)
  MAINTAINERS: remove amd5536udc USB gadget driver maintainer
  USB: whiteheat: fix potential null-deref at probe
  xhci: init command timeout timer earlier to avoid deleting it uninitialized
  xhci: change xhci 1.0 only restrictions to support xhci 1.1
  usb: xhci: exit early in xhci_setup_device() if we're halted or dying
  usb: xhci: stop everything on the first call to xhci_stop
  usb: xhci: Clear XHCI_STATE_DYING on start
  usb: xhci: lock mutex on xhci_stop
  xhci: Move xhci_pme_quirk() behind #ifdef CONFIG_PM
  xhci: give command abortion one more chance before killing xhci
  usb: Use the USB_SS_MULT() macro to get the burst multiplier.
  usb: dwc3: gadget: Fix BUG in RT config
  usb: musb: fix cppi channel teardown for isoch transfer
  usb: phy: isp1301: Export I2C module alias information
  usb: gadget: drop null test before destroy functions
  usb: gadget: dummy_hcd: in transfer(), return data sent, not limit
  usb: gadget: dummy_hcd: fix rescan logic for transfer
  usb: gadget: dummy_hcd: fix unneeded else-if condition
  usb: gadget: dummy_hcd: emulate sending zlp in packet logic
  usb: musb: dsps: fix polling in device-only mode
  ...

9 years agoMerge tag 'tty-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sun, 27 Sep 2015 00:58:38 +0000 (20:58 -0400)]
Merge tag 'tty-4.3-rc3' of git://git./linux/kernel/git/gregkh/tty

Pull serial driver fix from Greg KH:
 "Here is one serial driver fix for 4.3-rc3 that resolves a module
  loading issue due to splitting up of the 8250 driver into smaller
  pieces.  It's been in linux-next with no reported problems"

* tag 'tty-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty: serial: Add missing module license for 8250_base.ko

9 years agoMerge tag 'staging-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sun, 27 Sep 2015 00:56:50 +0000 (20:56 -0400)]
Merge tag 'staging-4.3-rc3' of git://git./linux/kernel/git/gregkh/staging

Pull staging driver fixes from Greg KH:
 "Here are some tiny staging driver and documentation fixes for 4.3-rc3.

  All of these resolve reported issues that people have found and have
  been in the linux-next tree for a while with no problems"

* tag 'staging-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  MAINTAINERS: Update email address for Martyn Welch
  staging: ion: fix corruption of ion_import_dma_buf
  staging: dgap: Remove myself from the MAINTAINERS file
  staging: most: Add dependency to HAS_IOMEM
  staging: unisys: remove reference of visorutil
  staging: unisys: visornic: handle error return from device registration
  staging: unisys: stop device registration before visorbus registration
  staging: unisys: visorbus: Unregister driver on error
  staging: unisys: visornic: Fix receive bytes statistics
  staging: unisys: unregister netdev when create debugfs fails
  staging: fbtft: replace master->setup() with spi_setup()
  staging: fbtft: fix 9-bit SPI support detection
  staging/lustre: change Lustre URLs and mailing list
  staging/android: Update ION TODO per LPC discussion
  Staging: most: MOST and MOSTCORE should depend on HAS_DMA
  staging: most: fix HDM_USB dependencies and build errors

9 years agoMerge tag 'driver-core-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 27 Sep 2015 00:54:53 +0000 (20:54 -0400)]
Merge tag 'driver-core-4.3-rc3' of git://git./linux/kernel/git/gregkh/driver-core

Pull driver core fix from Greg KH:
 "Here is one driver core fix for 4.3-rc3 that resolves a reported oops"

* tag 'driver-core-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  cpu/cacheinfo: Fix teardown path

9 years agoMerge tag 'char-misc-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sun, 27 Sep 2015 00:53:15 +0000 (20:53 -0400)]
Merge tag 'char-misc-4.3-rc3' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here's some tiny char and misc driver fixes that resolve some reported
  errors for 4.3-rc3.

  All of these have been in linux-next with no problems for a while"

* tag 'char-misc-4.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  extcon: Fix attached value returned by is_extcon_changed
  Drivers: hv: vmbus: fix init_vp_index() for reloading hv_netvsc
  mei: fix debugfs files leak on error path
  thunderbolt: Allow loading of module on recent Apple MacBooks with thunderbolt 2 controller

9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sat, 26 Sep 2015 10:01:33 +0000 (06:01 -0400)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) When we run a tap on netlink sockets, we have to copy mmap'd SKBs
    instead of cloning them.  From Daniel Borkmann.

 2) When converting classical BPF into eBPF, fix the setting of the
    source reg to BPF_REG_X.  From Tycho Andersen.

 3) Fix igmpv3/mldv2 report parsing in the bridge multicast code, from
    Linus Lussing.

 4) Fix dst refcounting for ipv6 tunnels, from Martin KaFai Lau.

 5) Set NLM_F_REPLACE flag properly when replacing ipv6 routes, from
    Roopa Prabhu.

 6) Add some new cxgb4 PCI device IDs, from Hariprasad Shenai.

 7) Fix headroom tests and SKB leaks in ipv6 fragmentation code, from
    Florian Westphal.

 8) Check DMA mapping errors in bna driver, from Ivan Vecera.

 9) Several 8139cp bug fixes (dev_kfree_skb_any in interrupt context,
    misclearing of interrupt status in TX timeout handler, etc.) from
    David Woodhouse.

10) In tipc, reset SKB header pointer after skb_linearize(), from Erik
    Hugne.

11) Fix autobind races et al. in netlink code, from Herbert Xu with
    help from Tejun Heo and others.

12) Missing SET_NETDEV_DEV in sunvnet driver, from Sowmini Varadhan.

13) Fix various races in timewait timer and reqsk_queue_hadh_req, from
    Eric Dumazet.

14) Fix array overruns in mac80211, from Johannes Berg and Dan
    Carpenter.

15) Fix data race in rhashtable_rehash_one(), from Dmitriy Vyukov.

16) Fix race between poll_one_napi and napi_disable, from Neil Horman.

17) Fix byte order in geneve tunnel port config, from John W Linville.

18) Fix handling of ARP replies over lightweight tunnels, from Jiri
    Benc.

19) We can loop when fib rule dumps cross multiple SKBs, fix from Wilson
    Kok and Roopa Prabhu.

20) Several reference count handling bug fixes in the PHY/MDIO layer
    from Russel King.

21) Fix lockdep splat in ppp_dev_uninit(), from Guillaume Nault.

22) Fix crash in icmp_route_lookup(), from David Ahern.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (116 commits)
  net: Fix panic in icmp_route_lookup
  net: update docbook comment for __mdiobus_register()
  ppp: fix lockdep splat in ppp_dev_uninit()
  net: via/Kconfig: GENERIC_PCI_IOMAP required if PCI not selected
  phy: marvell: add link partner advertised modes
  net: fix net_device refcounting
  phy: add phy_device_remove()
  phy: fixed-phy: properly validate phy in fixed_phy_update_state()
  net: fix phy refcounting in a bunch of drivers
  of_mdio: fix MDIO phy device refcounting
  phy: add proper phy struct device refcounting
  phy: fix mdiobus module safety
  net: dsa: fix of_mdio_find_bus() device refcount leak
  phy: fix of_mdio_find_bus() device refcount leak
  ip6_tunnel: Reduce log level in ip6_tnl_err() to debug
  ip6_gre: Reduce log level in ip6gre_err() to debug
  fib_rules: fix fib rule dumps across multiple skbs
  bnx2x: byte swap rss_key to comply to Toeplitz specs
  net: revert "net_sched: move tp->root allocation into fw_init()"
  lwtunnel: remove source and destination UDP port config option
  ...

9 years agonet: Fix panic in icmp_route_lookup
David Ahern [Thu, 24 Sep 2015 21:31:29 +0000 (15:31 -0600)]
net: Fix panic in icmp_route_lookup

Andrey reported a panic:

[ 7249.865507] BUG: unable to handle kernel pointer dereference at 000000b4
[ 7249.865559] IP: [<c16afeca>] icmp_route_lookup+0xaa/0x320
[ 7249.865598] *pdpt = 0000000030f7f001 *pde = 0000000000000000
[ 7249.865637] Oops: 0000 [#1]
...
[ 7249.866811] CPU: 0 PID: 0 Comm: swapper/0 Not tainted
4.3.0-999-generic #201509220155
[ 7249.866876] Hardware name: MSI MS-7250/MS-7250, BIOS 080014  08/02/2006
[ 7249.866916] task: c1a5ab00 ti: c1a52000 task.ti: c1a52000
[ 7249.866949] EIP: 0060:[<c16afeca>] EFLAGS: 00210246 CPU: 0
[ 7249.866981] EIP is at icmp_route_lookup+0xaa/0x320
[ 7249.867012] EAX: 00000000 EBX: f483ba48 ECX: 00000000 EDX: f2e18a00
[ 7249.867045] ESI: 000000c0 EDI: f483ba70 EBP: f483b9ec ESP: f483b974
[ 7249.867077]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[ 7249.867108] CR0: 8005003b CR2: 000000b4 CR3: 36ee07c0 CR4: 000006f0
[ 7249.867141] Stack:
[ 7249.867165]  320310ee 00000000 00000042 320310ee 00000000 c1aeca00
f3920240 f0c69180
[ 7249.867268]  f483ba04 f855058b a89b66cd f483ba44 f8962f4b 00000000
e659266c f483ba54
[ 7249.867361]  8004753c f483ba5c f8962f4b f2031140 000003c1 ffbd8fa0
c16b0e00 00000064
[ 7249.867448] Call Trace:
[ 7249.867494]  [<f855058b>] ? e1000_xmit_frame+0x87b/0xdc0 [e1000e]
[ 7249.867534]  [<f8962f4b>] ? tcp_in_window+0xeb/0xb10 [nf_conntrack]
[ 7249.867576]  [<f8962f4b>] ? tcp_in_window+0xeb/0xb10 [nf_conntrack]
[ 7249.867615]  [<c16b0e00>] ? icmp_send+0xa0/0x380
[ 7249.867648]  [<c16b102f>] icmp_send+0x2cf/0x380
[ 7249.867681]  [<f89c8126>] nf_send_unreach+0xa6/0xc0 [nf_reject_ipv4]
[ 7249.867714]  [<f89cd0da>] reject_tg+0x7a/0x9f [ipt_REJECT]
[ 7249.867746]  [<f88c29a7>] ipt_do_table+0x317/0x70c [ip_tables]
[ 7249.867780]  [<f895e0a6>] ? __nf_conntrack_find_get+0x166/0x3b0
[nf_conntrack]
[ 7249.867838]  [<f895eea8>] ? nf_conntrack_in+0x398/0x600 [nf_conntrack]
[ 7249.867889]  [<f84c0035>] iptable_filter_hook+0x35/0x80 [iptable_filter]
[ 7249.867933]  [<c16776a1>] nf_iterate+0x71/0x80
[ 7249.867970]  [<c1677715>] nf_hook_slow+0x65/0xc0
[ 7249.868002]  [<c1681811>] __ip_local_out_sk+0xc1/0xd0
[ 7249.868034]  [<c1680f30>] ? ip_forward_options+0x1a0/0x1a0
[ 7249.868066]  [<c1681836>] ip_local_out_sk+0x16/0x30
[ 7249.868097]  [<c1684054>] ip_send_skb+0x14/0x80
[ 7249.868129]  [<c16840f4>] ip_push_pending_frames+0x34/0x40
[ 7249.868163]  [<c16844a2>] ip_send_unicast_reply+0x282/0x310
[ 7249.868196]  [<c16a0863>] tcp_v4_send_reset+0x1b3/0x380
[ 7249.868227]  [<c16a1b63>] tcp_v4_rcv+0x323/0x990
[ 7249.868257]  [<c16776a1>] ? nf_iterate+0x71/0x80
[ 7249.868289]  [<c167dc2b>] ip_local_deliver_finish+0x8b/0x230
[ 7249.868322]  [<c167df4c>] ip_local_deliver+0x4c/0xa0
[ 7249.868353]  [<c167dba0>] ? ip_rcv_finish+0x390/0x390
[ 7249.868384]  [<c167d88c>] ip_rcv_finish+0x7c/0x390
[ 7249.868415]  [<c167e280>] ip_rcv+0x2e0/0x420
...

Prior to the VRF change the oif was not set in the flow struct, so the
VRF support should really have only added the vrf_master_ifindex lookup.

Fixes: 613d09b30f8b ("net: Use VRF device index for lookups on TX")
Cc: Andrey Melnikov <temnota.am@gmail.com>
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: update docbook comment for __mdiobus_register()
Russell King [Fri, 25 Sep 2015 10:56:56 +0000 (11:56 +0100)]
net: update docbook comment for __mdiobus_register()

Update the docbook comment for __mdiobus_register() to include the new
module owner argument.  This resolves a warning found by the 0-day
builder.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMAINTAINERS: remove amd5536udc USB gadget driver maintainer
Greg Kroah-Hartman [Sat, 26 Sep 2015 03:45:27 +0000 (20:45 -0700)]
MAINTAINERS: remove amd5536udc USB gadget driver maintainer

Thomas can no longer work on the driver, so he asked me to mark the
MAINTAINER entry as "Orphan" with the hope that someone else would
someday pick it up.

Cc: Thomas Dahlmann <dahlmann.thomas@arcor.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoIB/ipoib: increase the max mcast backlog queue
Doug Ledford [Sat, 26 Sep 2015 02:30:24 +0000 (22:30 -0400)]
IB/ipoib: increase the max mcast backlog queue

When performing sendonly joins, we queue the packets that trigger
the join until the join completes.  This may take on the order of
hundreds of milliseconds.  It is easy to have many more than three
packets come in during that time.  Expand the maximum queue depth
in order to try and prevent dropped packets during the time it
takes to join the multicast group.

Signed-off-by: Doug Ledford <dledford@redhat.com>
9 years agoMerge branch 'for-4.3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj...
Linus Torvalds [Fri, 25 Sep 2015 23:20:55 +0000 (16:20 -0700)]
Merge branch 'for-4.3-fixes' of git://git./linux/kernel/git/tj/cgroup

Pull another cgroup fix from Tejun Heo:
 "The cgroup writeback support got inadvertently enabled for traditional
  hierarchies revealing two regressions which are currently being worked
  on.  It shouldn't have been enabled on traditional hierarchies, so
  disable it on them.  This is enough to make the regressions go away
  for people who aren't experimenting with cgroup"

* 'for-4.3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  cgroup, writeback: don't enable cgroup writeback on traditional hierarchies

9 years agoppp: fix lockdep splat in ppp_dev_uninit()
Guillaume Nault [Thu, 24 Sep 2015 10:54:01 +0000 (12:54 +0200)]
ppp: fix lockdep splat in ppp_dev_uninit()

ppp_dev_uninit() locks all_ppp_mutex while under rtnl mutex protection.
ppp_create_interface() must then lock these mutexes in that same order
to avoid possible deadlock.

[  120.880011] ======================================================
[  120.880011] [ INFO: possible circular locking dependency detected ]
[  120.880011] 4.2.0 #1 Not tainted
[  120.880011] -------------------------------------------------------
[  120.880011] ppp-apitest/15827 is trying to acquire lock:
[  120.880011]  (&pn->all_ppp_mutex){+.+.+.}, at: [<ffffffffa0145f56>] ppp_dev_uninit+0x64/0xb0 [ppp_generic]
[  120.880011]
[  120.880011] but task is already holding lock:
[  120.880011]  (rtnl_mutex){+.+.+.}, at: [<ffffffff812e4255>] rtnl_lock+0x12/0x14
[  120.880011]
[  120.880011] which lock already depends on the new lock.
[  120.880011]
[  120.880011]
[  120.880011] the existing dependency chain (in reverse order) is:
[  120.880011]
[  120.880011] -> #1 (rtnl_mutex){+.+.+.}:
[  120.880011]        [<ffffffff81073a6f>] lock_acquire+0xcf/0x10e
[  120.880011]        [<ffffffff813ab18a>] mutex_lock_nested+0x56/0x341
[  120.880011]        [<ffffffff812e4255>] rtnl_lock+0x12/0x14
[  120.880011]        [<ffffffff812d9d94>] register_netdev+0x11/0x27
[  120.880011]        [<ffffffffa0147b17>] ppp_ioctl+0x289/0xc98 [ppp_generic]
[  120.880011]        [<ffffffff8113b367>] do_vfs_ioctl+0x4ea/0x532
[  120.880011]        [<ffffffff8113b3fd>] SyS_ioctl+0x4e/0x7d
[  120.880011]        [<ffffffff813ad7d7>] entry_SYSCALL_64_fastpath+0x12/0x6f
[  120.880011]
[  120.880011] -> #0 (&pn->all_ppp_mutex){+.+.+.}:
[  120.880011]        [<ffffffff8107334e>] __lock_acquire+0xb07/0xe76
[  120.880011]        [<ffffffff81073a6f>] lock_acquire+0xcf/0x10e
[  120.880011]        [<ffffffff813ab18a>] mutex_lock_nested+0x56/0x341
[  120.880011]        [<ffffffffa0145f56>] ppp_dev_uninit+0x64/0xb0 [ppp_generic]
[  120.880011]        [<ffffffff812d5263>] rollback_registered_many+0x19e/0x252
[  120.880011]        [<ffffffff812d5381>] rollback_registered+0x29/0x38
[  120.880011]        [<ffffffff812d53fa>] unregister_netdevice_queue+0x6a/0x77
[  120.880011]        [<ffffffffa0146a94>] ppp_release+0x42/0x79 [ppp_generic]
[  120.880011]        [<ffffffff8112d9f6>] __fput+0xec/0x192
[  120.880011]        [<ffffffff8112dacc>] ____fput+0x9/0xb
[  120.880011]        [<ffffffff8105447a>] task_work_run+0x66/0x80
[  120.880011]        [<ffffffff81001801>] prepare_exit_to_usermode+0x8c/0xa7
[  120.880011]        [<ffffffff81001900>] syscall_return_slowpath+0xe4/0x104
[  120.880011]        [<ffffffff813ad931>] int_ret_from_sys_call+0x25/0x9f
[  120.880011]
[  120.880011] other info that might help us debug this:
[  120.880011]
[  120.880011]  Possible unsafe locking scenario:
[  120.880011]
[  120.880011]        CPU0                    CPU1
[  120.880011]        ----                    ----
[  120.880011]   lock(rtnl_mutex);
[  120.880011]                                lock(&pn->all_ppp_mutex);
[  120.880011]                                lock(rtnl_mutex);
[  120.880011]   lock(&pn->all_ppp_mutex);
[  120.880011]
[  120.880011]  *** DEADLOCK ***

Fixes: 8cb775bc0a34 ("ppp: fix device unregistration upon netns deletion")
Reported-by: Sedat Dilek <sedat.dilek@gmail.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>