- 29 Aug, 2017 9 commits
-
-
Artemy Kovalyov authored
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. Signed-off-by:
Artemy Kovalyov <artemyko@mellanox.com> Reviewed-by:
Yossi Itigin <yosefe@mellanox.com> Signed-off-by:
Leon Romanovsky <leon@kernel.org> Signed-off-by:
Doug Ledford <dledford@redhat.com>
-
Artemy Kovalyov authored
Provide driver specific values for XRQ capabilities. Signed-off-by:
-
Artemy Kovalyov authored
Make XRQ capabilities available via ibv_query_device() verb. Signed-off-by:
-
Artemy Kovalyov authored
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. Signed-off-by:
-
Artemy Kovalyov authored
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. Signed-off-by:
-
Artemy Kovalyov authored
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. Signed-off-by:
-
Artemy Kovalyov authored
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. Signed-off-by:
-
Artemy Kovalyov authored
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 Signed-off-by: -
Artemy Kovalyov authored
* add offload_type field to mlx5_ifc_qpc_bits * update mlx5_ifc_xrqc_bits layout Signed-off-by:
-
- 28 Aug, 2017 31 commits
-
-
Andrew Boyer authored
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: 8700e3e7 ("Soft RoCE driver") Signed-off-by:
Andrew Boyer <andrew.boyer@dell.com> Signed-off-by:
-
Andrew Boyer authored
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: 8700e3e7 ("Soft RoCE driver") Signed-off-by:
-
Andrew Boyer authored
This fixes another path in rxe_requester() that might overlook stale SKBs, preventing cleanup. Fixes: 12171971 ("rxe: fix broken receive queue draining") Signed-off-by:
-
Andrew Boyer authored
Fixes: 4ed6ad1e ("IB/rxe: Cache dst in QP instead of getting it...") Signed-off-by:
Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by:
-
Andrew Boyer authored
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: 4ed6ad1e ("IB/rxe: Cache dst in QP instead of getting it...") Signed-off-by:
-
Andrew Boyer authored
Otherwise the reference count goes negative as IPv6 packets complete. Fixes: 4ed6ad1e ("IB/rxe: Cache dst in QP instead of getting it...") Signed-off-by:
-
Andrew Boyer authored
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: 4ed6ad1e ("IB/rxe: Cache dst in QP instead of getting it...") Signed-off-by:
-
Andrew Boyer authored
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: 8700e3e7 ("Soft RoCE driver") Signed-off-by:
-
Andrew Boyer authored
Fixes: 8700e3e7 ("Soft RoCE driver") Signed-off-by:
Moni Shoua <monis@mellanox.com> Reviewed-by:
-
Andrew Boyer authored
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: 8700e3e7 ("Soft RoCE driver") Signed-off-by:
-
Andrew Boyer authored
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: fda85ce9 ("IB/rxe: Fix kernel panic from skb destructor") Signed-off-by:
-
Mike Marciniszyn authored
A destroy of an MR prior to destroying the QP can cause the following diagnostic if the QP is referencing the MR being de-registered: hfi1 0000:05:00.0: hfi1_0: rvt_dereg_mr timeout mr ffff8808562108 00 pd ffff880859b20b00 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:
-
Mike Marciniszyn authored
Continue porting copy/paste code into rdmavt from qib. Reviewed-by:
-
Mike Marciniszyn authored
Continue moving copy/paste code into rdmavt. Reviewed-by:
-
Mike Marciniszyn authored
Change hfi1_error_port_qps() to use the new rvt_qp_iter() in its QP scanning. Reviewed-by:
Kaike Wan <kaike.wan@intel.com> Signed-off-by:
-
Mike Marciniszyn authored
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:
-
Kaike Wan authored
The qp_stats print will soon be moving to rdmavt, so use the proper accessor to get the ring size rather than a driver supplied constant. Fixes: Commit ff8d836e ("IB/hfi1: Add receiving queue info to qp_stats") Reviewed-by:
-
Kamenee Arumugam authored
Replace 'strcpy' with 'strncpy' to restrict the number of bytes copied to the buffer. Reviewed-by:
Michael J. Ruhl <michael.j.ruhl@intel.com> Signed-off-by:
Kamenee Arumugam <kamenee.arumugam@intel.com> Signed-off-by:
-
Michael J. Ruhl authored
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:
-
Jan Sokolowski authored
Currently, QSFP information is not queried in cases where loopback was set up and QSFP module is present. Acquire QSFP information in case of loopback. Reviewed-by:
Jan Sokolowski <jan.sokolowski@intel.com> Signed-off-by:
-
Bhumika Goyal authored
Make some structures const as they are only used during a copy operation. Signed-off-by:
Bhumika Goyal <bhumirks@gmail.com> Signed-off-by:
-
Arvind Yadav authored
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. Signed-off-by:
Arvind Yadav <arvind.yadav.cs@gmail.com> Signed-off-by:
-
Himanshu Jha authored
call to memset to assign 0 value immediately after allocating memory with kzalloc is unnecesaary as kzalloc allocates the memory filled with 0 value. Signed-off-by:
Himanshu Jha <himanshujha199640@gmail.com> Reviewed-by:
-
Dan Carpenter authored
usnic_uiom_get_dev_list() can return ERR_PTR(-ENOMEM) so we should check for that. Fixes: e3cf00d0 ("IB/usnic: Add Cisco VIC low-level hardware driver") Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by:
-
Mike Marciniszyn authored
These fields allow for debugging send engine processing. Reviewed-by:
-
Kaike Wan authored
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:
-
Kamenee Arumugam authored
Fix a tab alignment issue present in pr_err_ratelimited error message. Reviewed-by:
-
Harish Chegondi authored
Clean up user_sdma.c by moving the structure and MACRO definitions into the header file user_sdma.h Reviewed-by:
Harish Chegondi <harish.chegondi@intel.com> Signed-off-by:
-
Harish Chegondi authored
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:
-
Harish Chegondi authored
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:
-
Harish Chegondi authored
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:
-
