Artemy Kovalyov [Thu, 17 Aug 2017 12:52:10 +0000 (15:52 +0300)]
net/mlx5: Add XRQ support
Add support to new XRQ(eXtended shared Receive Queue)
hardware object. It supports SRQ semantics with addition
of extended receive buffers topologies and offloads.
Currently supports tag matching topology and rendezvouz offload.
Artemy Kovalyov [Thu, 17 Aug 2017 12:52:07 +0000 (15:52 +0300)]
IB/uverbs: Add new SRQ type IB_SRQT_TM
Add new SRQ type capable of new tag matching feature.
When SRQ receives a message it will search through the matching list
for the corresponding posted receive buffer. The process of searching
the matching list is called tag matching.
In case the tag matching results in a match, the received message will
be placed in the address specified by the receive buffer. In case no
match was found the message will be placed in a generic buffer until the
corresponding receive buffer will be posted. These messages are called
unexpected and their set is called an unexpected list.
Artemy Kovalyov [Thu, 17 Aug 2017 12:52:06 +0000 (15:52 +0300)]
IB/uverbs: Add XRQ creation parameter to UAPI
Add tm_list_size parameter to struct ib_uverbs_create_xsrq.
If SRQ type is tag-matching this field defines maximum size
of tag matching list. Otherwise, it is expected to be zero.
Artemy Kovalyov [Thu, 17 Aug 2017 12:52:05 +0000 (15:52 +0300)]
IB/core: Add new SRQ type IB_SRQT_TM
This patch adds new SRQ type - IB_SRQT_TM. The new SRQ type supports tag
matching and rendezvous offloads for MPI applications.
When SRQ receives a message it will search through the matching list
for the corresponding posted receive buffer. The process of searching
the matching list is called tag matching.
In case the tag matching results in a match, the received message will
be placed in the address specified by the receive buffer. In case no
match was found the message will be placed in a generic buffer until the
corresponding receive buffer will be posted. These messages are called
unexpected and their set is called an unexpected list.
Artemy Kovalyov [Thu, 17 Aug 2017 12:52:04 +0000 (15:52 +0300)]
IB/core: Separate CQ handle in SRQ context
Before this change CQ attached to SRQ was part of XRC specific extension.
Moving CQ handle out makes it available to other types extending SRQ
functionality.
Artemy Kovalyov [Thu, 17 Aug 2017 12:52:03 +0000 (15:52 +0300)]
IB/core: Add XRQ capabilities
This patch adds following TM XRQ capabilities:
* max_rndv_hdr_size - Max size of rendezvous request message
* max_num_tags - Max number of entries in tag matching list
* max_ops - Max number of outstanding list operations
* max_sge - Max number of SGE in tag matching entry
* flags - the following flags are currently defined:
- IB_TM_CAP_RC - Support tag matching on RC transport
Andrew Boyer [Mon, 28 Aug 2017 20:11:59 +0000 (16:11 -0400)]
IB/rxe: Handle NETDEV_CHANGE events
Without this fix, ports configured on top of ixgbe miss link up
notifications. ibv_query_port() will continue to return IBV_PORT_DOWN even
though the port is up and working.
Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Andrew Boyer <andrew.boyer@dell.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Andrew Boyer [Mon, 28 Aug 2017 20:11:58 +0000 (16:11 -0400)]
IB/rxe: Avoid ICRC errors by copying into the skb first
The current process is to first calculate the CRC and then copy the client
data into the packet. This leaves a window in which the packet contents and
CRC can get out of sync, if the client changes the data after the CRC is
calculated but before the data is copied.
By copying the data into the packet and then calculating the CRC directly
from the packet contents we eliminate the window.
This can be seen with qperf's ud_bi_bw test. This seems like very
strange/reckless client behavior, but whether the client has mangled its
data or not RXE should be able to transfer it reliably.
Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Andrew Boyer <andrew.boyer@dell.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Andrew Boyer [Mon, 28 Aug 2017 20:11:55 +0000 (16:11 -0400)]
IB/rxe: Fix up rxe_qp_cleanup()
Replace sk_dst_get()/dst_release() in rxe_qp_cleanup() with sk_dst_reset().
sk_dst_get() takes a new reference on dst, so the dst_release() doesn't
actually release the original reference, which was the design intent.
Fixes: 4ed6ad1eb30e ("IB/rxe: Cache dst in QP instead of getting it...") Signed-off-by: Andrew Boyer <andrew.boyer@dell.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Andrew Boyer [Mon, 28 Aug 2017 20:11:53 +0000 (16:11 -0400)]
IB/rxe: Fix destination cache for IPv6
To successfully match an IPv6 path, the path cookie must match. Store it
in the QP so that the IPv6 path can be reused.
Replace open-coded version of dst_check() with the actual call, fixing the
logic. The open-coded version skips the check call if dst->obsolete is 0
(DST_OBSOLETE_NONE), proceeding to replace the route. DST_OBSOLETE_NONE
means that the route may continue to be used, though.
Fixes: 4ed6ad1eb30e ("IB/rxe: Cache dst in QP instead of getting it...") Signed-off-by: Andrew Boyer <andrew.boyer@dell.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Andrew Boyer [Mon, 28 Aug 2017 20:11:52 +0000 (16:11 -0400)]
IB/rxe: Fix up the responder's find_resources() function
The resource array is sized by max_dest_rd_atomic, not max_rd_atomic.
Iterating over max_rd_atomic entries of qp->resp.resources[] will cause
incorrect behavior when the two attributes are different (or even
crash if max_rd_atomic is larger).
Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Andrew Boyer <andrew.boyer@dell.com> Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Andrew Boyer [Mon, 28 Aug 2017 20:11:50 +0000 (16:11 -0400)]
IB/rxe: Disable completion upcalls when a CQ is destroyed
This prevents the stack from accessing userspace objects while they
are being torn down.
One possible sequence of events:
- Userspace program exits
- ib_uverbs_cleanup_ucontext() runs, calling ib_destroy_qp(),
ib_destroy_cq(), etc. and releasing/freeing the UCQ
- The QP still has tasklets running, so it isn't destroyed yet
- The CQ is referenced by the QP, so the CQ isn't destroyed yet
- The UCQ is kfree()'d anyway
- A send work request completes
- rxe_send_complete() calls cq->ibcq.comp_handler()
- ib_uverbs_comp_handler() runs and crashes; the event queue is checked
for is_closed, but it has no way to check the ib_ucq_object before
accessing it
The reference counting on the CQ doesn't protect against this since the CQ
hasn't been destroyed yet.
There's no available interface to deregister the UCQ from the CQ, and it
didn't appear that attempting to add reference counting to the UCQ was
going to be a good way to go since this solution is much simpler.
Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Andrew Boyer <andrew.boyer@dell.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Andrew Boyer [Mon, 28 Aug 2017 20:11:49 +0000 (16:11 -0400)]
IB/rxe: Move refcounting earlier in rxe_send()
The network stack will call nskb's destructor, rxe_skb_tx_dtor(), if the
packet gets dropped by ip_local_out()/ip6_local_out(). Thus we need to add
the QP ref before output to avoid extra dereferences during network
congestion. This could lead to unwanted destruction of the QP.
Fix up the skb_out accounting, too.
Fixes: fda85ce91240 ("IB/rxe: Fix kernel panic from skb destructor") Signed-off-by: Andrew Boyer <andrew.boyer@dell.com> Acked-by: Moni Shoua <monis@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
The solution is to when the a non-zero refcount is encountered when
the MR is destroyed the QPs needs to be iterated looking for QPs in
the same PD as the MR. If rvt_qp_mr_clean() detects any such QP
references the rkey/lkey, the QP needs to be put into an error state
via a call to rvt_qp_error() which will trigger the clean up of any
stuck references.
This solution is as specified in IBTA 1.3 Volume 1 11.2.10.5.
[This is reproduced with the 0.4.9 version of qperf and the rc_bw test]
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mike Marciniszyn [Mon, 28 Aug 2017 18:23:45 +0000 (11:23 -0700)]
IB/rdmavt: Add QP iterator API for QPs
There are currently 3 spots in the qib and hfi1 driver that have
knowledge of the internal QP hash list that should only be in
scope to rdmavt QP code.
Add an iterator API for processing all QPs to hide the
nature of the RCU hashlist.
The API consists of:
- rvt_qp_iter_init()
* For iterating QPs one at a time for seq_file semantics
- rvt_qp_iter_next()
* For iterating QPs one at a time for seq_file semantics
- rvt_qp_iter()
* For iterating all QPs
The first two are used for things like seq_file prints.
The last is for code that just needs to iterate all QPs
in the system.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Michael J. Ruhl [Mon, 28 Aug 2017 18:23:27 +0000 (11:23 -0700)]
IB/hif1: Remove static tracing from SDMA hot path
The hfi1_cdbg() macro can be instantiated in the hot path even when it
is not in use. This shows up on perf profiles.
Rework the macros (for SDMA and MMU), to use the trace interface directly
to eliminate this performance hit.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Arvind Yadav [Mon, 28 Aug 2017 04:29:28 +0000 (09:59 +0530)]
IB/hfi1: constify vm_operations_struct
vm_operations_struct are not supposed to change at runtime.
vm_area_struct structure working with const vm_operations_struct.
So mark the non-const vm_operations_struct structs as const.
Kaike Wan [Tue, 22 Aug 2017 01:27:41 +0000 (18:27 -0700)]
IB/hfi1: Add received request info to qp_stats
The rvt_ack_entry pointed to by s_tail_ack_queue provides important
info about the request that has just been processed or is being processed
on the responder side of a RC connection. This patch adds this info to
the qp_stats to assist debugging.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Kaike Wan <kaike.wan@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Harish Chegondi [Tue, 22 Aug 2017 01:27:23 +0000 (18:27 -0700)]
IB/hfi1: Move structure definitions from user_exp_rcv.c to user_exp_rcv.h
Clean up user_exp_rcv.c file by moving structure definitions into header
file user_exp_rcv.h. Since these structure definitions depend on the
structure definitions in mmu_rb.h, move #include "mmu_rb.h" above
the include "user_exp_rcv.h" or include of header files that include
user_exp_rcv.h
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Harish Chegondi <harish.chegondi@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Harish Chegondi [Tue, 22 Aug 2017 01:27:16 +0000 (18:27 -0700)]
IB/hfi1: Remove duplicate definitions of num_user_pages() function
num_user_pages() function has been defined in both user_exp_rcv.c file
and user_sdma.c file. Move the function definition to a header file so
there is only one definition in the source repo.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Harish Chegondi <harish.chegondi@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Harish Chegondi [Tue, 22 Aug 2017 01:27:09 +0000 (18:27 -0700)]
IB/hfi1: Fix the bail out code in pin_vector_pages() function
In pin_vector_pages() function, if there is any error while pinning the
pages or while adding a pinned buffer to the cache, the bail out code
needs to unpin any pinned pages that are not in the cache and adjust the
n_locked counter that counts the total pages pinned. The current bail
out code doesn't seem to be doing it right in two cases:
1. Before pinning required pages for a buffer, the SDMA pinned buffer
cache is searched to see if the virtual address range that needs to be
pinned is already pinned. If there isn't a hit in the cache, a new node
is created for the buffer and is added to the cache after the buffer is
pinned. If adding the new node to the cache fails, the n_locked count is
decremented properly but the pinned pages are not freed. This commit
fixes this issue.
2. If there is a hit in the SDMA cache, but the cached buffer doesn't
have enough pages to cover the entire address range that needs to be
pinned, the node for the cached buffer is extracted from the cache,
remaining pages needed are pinned and added to the node. The node is
finally added back into the cache. If there is an error pinning the
extra pages, the bail out code frees all the pages in the node but the
n_locked count is not being decremented by the no of pages in the node
that are freed. This commit fixes this issue.
This commit fixes the above two issues by creating a new function that
frees the pages in a node and decrements the n_locked count by the
number of pages freed.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Harish Chegondi <harish.chegondi@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Harish Chegondi [Tue, 22 Aug 2017 01:26:51 +0000 (18:26 -0700)]
IB/hfi1: Clean up hfi1_user_exp_rcv_setup function
Clean up hfi1_user_exp_rcv_setup function by moving page pinning and
unpinning related code to separate functions. In order to reduce the
number of parameters passed between functions, a new data structure
struct tid_user_buf is defined and used.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Harish Chegondi <harish.chegondi@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Michael J. Ruhl [Tue, 22 Aug 2017 01:26:45 +0000 (18:26 -0700)]
IB/hfi1: Improve local kmem_cache_alloc performance
Performance analysis shows that the cache callback function
sdma_kmem_cache_ctor contributes to 1/2 of the kmem_cache_allocs
time.
Since all of the fields in the allocated data structure are initialized
in the code path, remove the _ctor function.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mike Marciniszyn [Tue, 22 Aug 2017 01:26:20 +0000 (18:26 -0700)]
IB/{qib, hfi1}: Avoid flow control testing for RDMA write operation
Section 9.7.7.2.5 of the 1.3 IBTA spec clearly says that receive
credits should never apply to RDMA write.
qib and hfi1 were doing that. The following situation will result
in a QP hang:
- A prior SEND or RDMA_WRITE with immmediate consumed the last
credit for a QP using RC receive buffer credits
- The prior op is acked so there are no more acks
- The peer ULP fails to post receive for some reason
- An RDMA write sees that the credits are exhausted and waits
- The peer ULP posts receive buffers
- The ULP posts a send or RDMA write that will be hung
The fix is to avoid the credit test for the RDMA write operation.
Cc: <stable@vger.kernel.org> Reviewed-by: Kaike Wan <kaike.wan@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Bodong Wang [Thu, 17 Aug 2017 12:52:34 +0000 (15:52 +0300)]
IB/mlx5: Allow posting multi packet send WQEs if hardware supports
Set the field to allow posting multi packet send WQEs if hardware
supports this feature. This doesn't mean the send WQEs will be for
multi packet unless the send WQE was prepared according to multi
packet send WQE format.
User space shall use flag MLX5_IB_ALLOW_MPW to check if hardware
supports MPW and allows MPW in SQ context.
Signed-off-by: Bodong Wang <bodong@mellanox.com> Reviewed-by: Daniel Jurgens <danielj@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Yishai Hadas [Thu, 17 Aug 2017 12:52:33 +0000 (15:52 +0300)]
IB/mlx5: Add support for multi underlay QP
Set underlay QPN as part of flow rule when it's applicable.
There is one root flow table in the NIC RX namespace and all the
underlay QPs steer the traffic to this flow table.
In order to prevent QP to get traffic which is not target to its
underlay QP, we need to set the underlay QP number as part of
the steering matching.
Note:
When multicast traffic is sent the QPN filtering is done by the firmware
as some early step. Adding the QPN match on the flow table entry is
wrong as by that time the target QPN holds the multicast address (e.g.
FF(s)) and it won't match.
Ilya Lesokhin [Thu, 17 Aug 2017 12:52:29 +0000 (15:52 +0300)]
IB/mlx5: Enable UMR for MRs created with reg_create
This patch is the first step in decoupling UMR usage and
allocation from the MR cache. The only functional change
in this patch is to enables UMR for MRs created with
reg_create.
This change fixes a bug where ODP memory regions that
were not allocated from the MR cache did not have UMR
enabled.
Noa Osherovich [Thu, 17 Aug 2017 12:52:28 +0000 (15:52 +0300)]
IB/mlx5: Expose software parsing for Raw Ethernet QP
Software parsing (SWP) is a feature that can be used to instruct the
device to stop using its internal parser and to parse packets on the
transmit path according to offsets set for each packets.
Through this feature, the device allows the handling of checksum and
LSO by the hardware according to the location of IP and TCP/UDP
headers.
Enable SW parsing on Raw Ethernet send queue by default if firmware
supports it and report these capabilities to user space.
Signed-off-by: Noa Osherovich <noaos@mellanox.com> Reviewed-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Trivial fix to spelling mistake in DP_ERR error message
Signed-off-by: Colin Ian King <colin.king@canonical.com> Reviewed-by: Leon Romanovsky <leonro@mellanox.com> Reviewed-by: Ram Amrani <Ram.Amrani@cavium.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Selvin Xavier [Wed, 23 Aug 2017 08:08:07 +0000 (01:08 -0700)]
IB: Avoid ib_modify_port() failure for RoCE devices
IB CM calls ib_modify_port() irrespective of link layer. If the
failure is returned, the mad agent gets unregistered for those
devices. Recently, modify_port() hook was removed from some of the
low level drivers as it was always returning success. This breaks
rdma connection establishment over those devices.
For ethernet devices, Qkey violation and port capabilities are not
applicable. So returning success for RoCE when modify_port hook is
is not implemented.
Cc: Leon Romanovsky <leon@kernel.org> Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com> Reviewed-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Bryan Tan [Wed, 23 Aug 2017 06:19:00 +0000 (23:19 -0700)]
RDMA/vmw_pvrdma: Add RoCEv2 support
The driver version is bumped for compatibility purposes. Also, send correct
GID type during register to device. Added compatibility check macros for
the device.
Leon Romanovsky [Thu, 17 Aug 2017 12:50:54 +0000 (15:50 +0300)]
RDMA/mlx4: Properly annotate link layer variable
The rdma_port_get_link_layer() returns enum rdma_link_layer as
a return value, hence it is better to store the return value in
specially annotated variable and not in int.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Erez Shitrit [Thu, 17 Aug 2017 12:50:50 +0000 (15:50 +0300)]
IB/ipoib: Sync between remove_one to sysfs calls that use rtnl_lock
In order to avoid deadlock between sysfs functions (like create/delete
child) and remove_one (both of them are using the sysfs lock and
rtnl_lock) the driver will use a state mutex for sync.
That will fix traces as the following:
schedule+0x3e/0x90
kernfs_drain+0x75/0xf0
? wait_woken+0x90/0x90
__kernfs_remove+0x12e/0x1c0
kernfs_remove+0x25/0x40
sysfs_remove_dir+0x57/0x90
kobject_del+0x22/0x60
device_del+0x195/0x230
pm_runtime_set_memalloc_noio+0xac/0xf0
netdev_unregister_kobject+0x71/0x80
rollback_registered_many+0x205/0x2f0
rollback_registered+0x31/0x40
unregister_netdevice_queue+0x58/0xb0
unregister_netdev+0x20/0x30
ipoib_remove_one+0xb7/0x240 [ib_ipoib]
ib_unregister_device+0xbc/0x1b0 [ib_core]
ib_unregister_mad_agent+0x29/0x30 [ib_core]
mlx4_ib_remove+0x67/0x280 [mlx4_ib]
INFO: task echo:24082 blocked for more than 120 seconds.
Tainted: G OE 4.1.12-37.5.1.el6uek.x86_64 #2
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this
message.
Call Trace:
schedule+0x3e/0x90
schedule_preempt_disabled+0xe/0x10
__mutex_lock_slowpath+0x95/0x110
? _rcu_barrier+0x177/0x220
mutex_lock+0x23/0x40
rtnl_lock+0x15/0x20
netdev_run_todo+0x81/0x1f0
rtnl_unlock+0xe/0x10
ipoib_vlan_delete+0x12f/0x1c0 [ib_ipoib]
delete_child+0x69/0x80 [ib_ipoib]
dev_attr_store+0x20/0x30
sysfs_kf_write+0x41/0x50
Signed-off-by: Erez Shitrit <erezsh@mellanox.com> Reviewed-by: Alex Vesker <valex@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Guy Levi [Thu, 17 Aug 2017 12:50:49 +0000 (15:50 +0300)]
IB/mlx4: Check that reserved fields in mlx4_ib_create_qp_rss are zero
According to mlx4 convention, need to fail the command due to a non-zero
value in the user data which is expected to be zero.
Fixes: 3078f5f1bd8b ("IB/mlx4: Add support for RSS QP") Signed-off-by: Guy Levi <guyle@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Guy Levi [Thu, 17 Aug 2017 12:50:48 +0000 (15:50 +0300)]
IB/mlx4: Remove redundant attribute in mlx4_ib_create_qp_rss struct
rx_key_len is not in use and needs to be removed.
Fixes: 3078f5f1bd8b ("IB/mlx4: Add support for RSS QP") Signed-off-by: Guy Levi <guyle@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Guy Levi [Thu, 17 Aug 2017 12:50:47 +0000 (15:50 +0300)]
IB/mlx4: Fix struct mlx4_ib_create_wq alignment
The mlx4 ABI defines to have structures with alignment of 64B.
Fixes: 400b1ebcfe31 ("IB/mlx4: Add support for WQ related verbs") Signed-off-by: Guy Levi <guyle@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Guy Levi [Thu, 17 Aug 2017 12:50:46 +0000 (15:50 +0300)]
IB/mlx4: Fix RSS QP type in creation verb
The mlx4 was designed to support QP type of MLX4_IB_QPT_RAW_PACKET.
Fixes: 3078f5f1bd8b ("IB/mlx4: Add support for RSS QP") Signed-off-by: Guy Levi <guyle@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Leon Romanovsky [Thu, 17 Aug 2017 12:50:41 +0000 (15:50 +0300)]
RDMA/usnic: Fix remove address space warning
Sparse tool complains with the following error:
drivers/infiniband/hw/usnic/usnic_ib_main.c:445:16: warning: cast removes
address space of expression
Fix it by doing casting on correct field and convert function helper which
sets ifaddr to be void, because there are no users who are interested in
returned value.
Fixes: c7845bcafe4d ("IB/usnic: Add UDP support in u*verbs.c, u*main.c and u*util.h") Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Leon Romanovsky [Thu, 17 Aug 2017 12:50:37 +0000 (15:50 +0300)]
RDMA/core: Cleanup device capability enum
Cleanup patch prior exporting the ib_device_cap_flags
to the user space. In this patch, we are aligning the
indentation, removing IB_DEVICE_INIT_TYPE and IB_DEVICE_RESERVED
fields, because it is not used in the kernel.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Maor Gottlieb [Thu, 17 Aug 2017 12:50:35 +0000 (15:50 +0300)]
RDMA/mlx4: Fix create qp command alignment
Avoid extra padding by replacing the order of inl_recv_sz and reserved,
otherwise 'mlx4_ib_create_qp' structure might be larger than legacy user
input leading to copy of some garbage data from the user space buffer.
Fixes: ea30b966f7dd ('IB/mlx4: Add inline-receive support') Signed-off-by: Maor Gottlieb <maorg@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Parav Pandit [Thu, 17 Aug 2017 12:50:33 +0000 (15:50 +0300)]
IB/uverbs: Introduce and use helper functions to copy ah attributes
This patch introduces two helper functions to copy ah attributes
from uverbs to internal ib_ah_attr structure and the other way
during modify qp and query qp respectively.
Signed-off-by: Parav Pandit <parav@mellanox.com> Reviewed-by: Daniel Jurgens <danielj@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Leon Romanovsky [Thu, 17 Aug 2017 12:50:32 +0000 (15:50 +0300)]
IB/cma: Fix erroneous validation of supported default GID type
When rdma_cm is initializing a cma_device it checks if this device
supports the preferred default GID type. This check was done in a wrong way
and therefore sometimes rdma_cm is coming up with default GID type that is
not supported by the device.
Fix that by checking for supported GID type properly.
Fixes: 3c7f67d1880d ("IB/cma: Fix default RoCE type setting") Signed-off-by: Moni Shoua <monis@mellanox.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Reviewed-by: Selvin Xavier <selvin.xavier@broadcom.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Majd Dibbiny [Wed, 23 Aug 2017 05:35:42 +0000 (08:35 +0300)]
IB/mlx5: Always return success for RoCE modify port
CM layer calls ib_modify_port() regardless of the link layer.
For the Ethernet ports, qkey violation and Port capabilities
are meaningless. Therefore, always return success for ib_modify_port
calls on the Ethernet ports.
Cc: Selvin Xavier <selvin.xavier@broadcom.com> Signed-off-by: Majd Dibbiny <majd@mellanox.com> Reviewed-by: Moni Shoua <monis@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com>
Noa Osherovich [Wed, 23 Aug 2017 05:35:40 +0000 (08:35 +0300)]
IB/core: Avoid accessing non-allocated memory when inferring port type
Commit 44c58487d51a ("IB/core: Define 'ib' and 'roce' rdma_ah_attr types")
introduced the concept of type in ah_attr:
* During ib_register_device, each port is checked for its type which
is stored in ib_device's port_immutable array.
* During uverbs' modify_qp, the type is inferred using the port number
in ib_uverbs_qp_dest struct (address vector) by accessing the
relevant port_immutable array and the type is passed on to
providers.
IB spec (version 1.3) enforces a valid port value only in Reset to
Init. During Init to RTR, the address vector must be valid but port
number is not mentioned as a field in the address vector, so its
value is not validated, which leads to accesses to a non-allocated
memory when inferring the port type.
Save the real port number in ib_qp during modify to Init (when the
comp_mask indicates that the port number is valid) and use this value
to infer the port type.
Avoid copying the address vector fields if the matching bit is not set
in the attr_mask. Address vector can't be modified before the port, so
no valid flow is affected.
Li Dongyang [Wed, 16 Aug 2017 13:31:22 +0000 (23:31 +1000)]
IB/mlx5: use kvmalloc_array for mlx5_ib_wq
We observed multiple times on our Lustre OSS servers that when
the system memory is fragmented, kmalloc() in create_kernel_qp()
could fail order 4/5 allocations while we still have many free pages.
Switch to kvmalloc_array() to allow the operation to contine.
Signed-off-by: Li Dongyang <dongyang.li@anu.edu.au> Acked-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Shiraz Saleem [Wed, 9 Aug 2017 01:38:45 +0000 (20:38 -0500)]
i40iw: Improve CQP timeout logic
The current timeout logic for Control Queue-Pair (CQP) OPs
does not take into account whether CQP makes progress but
rather blindly waits for a large timeout value, 100000 jiffies
for the completion event. Improve this by setting the timeout
based on whether the CQP is making progress or not. If the CQP
is hung, the timeout will happen sooner, in 5000 jiffies. Each
time the CQP progress is detetcted, the timeout extends by 5000
jiffies.
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com> Signed-off-by: Christopher N Bednarz <christopher.n.bednarz@intel.com> Signed-off-by: Henry Orosco <henry.orosco@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>