- 13 Nov, 2017 32 commits
-
-
Selvin Xavier authored
Synchronize poll_cq and req_notify_cq verbs using cq_lock, instead of the lower level qplib->hwq.lock. Signed-off-by:
Selvin Xavier <selvin.xavier@broadcom.com> Signed-off-by:
Doug Ledford <dledford@redhat.com>
-
Selvin Xavier authored
Destroy_qp shall wait for any outstanding CQ notification to be flushed out before proceeding with QP destroy. Flushing the WQ before destroying the QP. Signed-off-by:
-
Selvin Xavier authored
Moves the driver QP state to error in case of response completion errors. Handles the scenarios which doesn't generate a terminal CQE. Signed-off-by:
-
Somnath Kotur authored
The code determines if the next ring entry is valid before proceeding further to read the rest of the entry. The CPU can re-order and read the rest of the entry first, possibly reading a stale entry, if DMA of a new entry happens right after reading it. Signed-off-by:
Somnath Kotur <somnath.kotur@broadcom.com> Signed-off-by:
-
Dennis Dalessandro authored
When the driver is loaded it sets the default CCTI value to be 0. When the FM starts and CCA is disabled the driver sets the max value to 65535 due the driver subtracting 1 from 0 and the fact that the CCTI value is a u16. Special case the subtraction to find the index for a 0 value. Reviewed-by:
Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by:
Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by:
-
Don Hiatt authored
Pass only the lower 16Bits of an Extended LIDs to rvt_cq_entry to avoid triggering a WARN_ON_ONCE during conversion there. These upper 16Bits are okay to drop as they are obtained elsewhere. Fixes: 62ede777 ("Add OPA extended LID support") Reviewed-by:
Ira Weiny <ira.weiny@intel.com> Signed-off-by:
Don Hiatt <don.hiatt@intel.com> Signed-off-by:
-
Don Hiatt authored
When deciding to convert an OPA AH to IB we were incorrectly including the IB multicast range. At this layer, all Extended LIDs will be larger than IB_LID_PERMISSIVE. Change comparison accordingly. Fixes: d541e455 ("IB/core: Convert ah_attr from OPA to IB when copying to user") Reviewed-by:
-
Jan Sokolowski authored
On host shutdown, driver sends 'SMA_Disabled' as a reason for link down. This is incorrect. Send 'reboot' as a linkdown reason. Signed-off-by:
Jan Sokolowski <jan.sokolowski@intel.com> Reviewed-by:
Jakub Byczkowski <jakub.byczkowski@intel.com> Signed-off-by:
-
Grzegorz Morys authored
It is invalid to change Link state from Init to Armed if IsSmConfigurationStarted bit is not set in Attribute modifier for Set subnet management method in case of PortInfo and PortStateInfo attribute. Set response MAD status field bits accordingly to react correctly in such situations and avoid changing Link state. Reviewed-by:
Michael J. Ruhl <michael.j.ruhl@intel.com> Signed-off-by:
Grzegorz Morys <grzegorz.morys@intel.com> Signed-off-by:
-
Niranjana Vishwanathapura authored
OPA VNIC does not use PIO contexts and instead only uses SDMA engines. Do not allocate PIO contexts for VNIC ports. Reviewed-by:
Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com> Signed-off-by:
-
Jan Sokolowski authored
A for loop condition of data_iovs in user_sdma_free_request is unnecessarily repeated before the loop as an if check. Remove the if enveloping the loop. Reviewed-by:
-
Mike Marciniszyn authored
The "2 * UINT_MAX" statement: if ((u64)(ts - cce->timestamp) > 2 * UINT_MAX) { is equivalent to: if ((u64)(ts - cce->timestamp) > UINT_MAX - 1) { This results in a premature timeout of the cong log entry. Fix by using unsigned 64 bit integers, removing casts, and using an algebraic equivalent test to avoid the "2 * UINT_MAX" issue. Also make use of kernel API to get nanoseconds instead of open coding. Reported-by:Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by:
-
Kamenee Arumugam authored
Wrapper functions were used to call the same function mmu_notifier_mem_invalidate for 2 callbacks in mmu_notifier. The commit 7def96f0 ("IB/hfi1: update to new mmu_notifier semantic") removed the invalidate_page callback. Therefore, the wrapper function is no longer needed. Reviewed-by:
Alex Estrin <alex.estrin@intel.com> Signed-off-by:
Kamenee Arumugam <kamenee.arumugam@intel.com> Signed-off-by:
-
Jakub Byczkowski authored
Timeout of 20 seconds is too long for active wait performed for 8051 command completion. It was required for scenarios when transition to polling was requested before offline.quiet state was reached. Currently wait for offline.quiet is properly implemented and timeout can be reduced to 1 second. Reviewed-by:
Dean Luick <dean.luick@intel.com> Reviewed-by:
Duane McCrory <duane.mccrory@intel.com> Signed-off-by:
-
Jan Sokolowski authored
HFI's are hard-wired to send Device Info frames with MgmtAllowed bit set to 0. This means in B2B setups, MgmtAllowed would never be allowed, which prevents remote opa management tools from working properly. Assume MgmtAllowed if a neighbor is also an HFI. Fixes: 98b9ee20 ("IB/hfi1: Cache neighbor secure data after link up") Reviewed-by:
Sebastian Sanchez <sebastian.sanchez@intel.com> Reviewed-by:
-
Bart Van Assche authored
The use_srq configfs attribute is created after it is read. Hence modify srpt_tpg_attrib_use_srq_store() such that this function switches dynamically between non-SRQ and SRQ mode. Reported-by:
Bart Van Assche <bart.vanassche@wdc.com> Signed-off-by:
-
Bart Van Assche authored
Introduce the helper function srpt_set_enabled(). Protect sport->enabled changes with sdev->mutex. Makes configfs writes into 'enabled' wait until all channel resources have been freed. Wait until channel release has finished during kernel module unload. Signed-off-by:
-
Bart Van Assche authored
Except for changing a BUG_ON() call into a WARN_ON_ONCE() call, this patch does not change any functionality. Signed-off-by:
-
Bart Van Assche authored
The only functional change in this patch is in the srpt_add_one() error path: if allocating the ring buffer for the SRQ fails, fall back to non-SRQ mode instead of disabling SRP target functionality. Signed-off-by:
-
Mike Marciniszyn authored
IB and iWARP specs both spell out that posting a receive work request to a queue pair in the RESET state is an invalid operation and required to fail. Postpone posting receive work requests until after the transition to the INIT state. Fixes: commit dea26209 ("IB/srpt: Change default behavior from using SRQ to using RC") Signed-off-by:
-
Steve Wise authored
iw_cxgb4 has many BUG_ON()s that were left over from various enhancemnets made over the years. Almost all of them should just be removed. Some, however indicate a ULP usage error and can be handled w/o bringing down the system. If the condition cannot happen with correctly implemented cxgb4 sw/fw, then remove the BUG_ON. If the condition indicates a misbehaving ULP (like CQ overflows), add proper recovery logic. Signed-off-by:
Steve Wise <swise@opengridcomputing.com> Reviewed-by:
-
Sriharsha Basavapatna authored
While adding a new gid, the driver currently does not return the context back to the stack. A subsequent del_gid() (e.g, when ip address is changed) doesn't find the right context in the driver and it ends up dropping that request. This results in the HW caching a stale gid entry and traffic fails because of that. Fix by returning the proper context in bnxt_re_add_gid(). Signed-off-by:
Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com> Signed-off-by:
-
Sriharsha Basavapatna authored
In bnxt_qplib_process_qp_event(), for qp error events we look up the qp-handle and pass it for further processing. But we don't check if the handle is NULL. This could lead to a crash in the called functions when that qp-handle is dereferenced, if the qp is destroyed in the meantime. Fix this by checking for a valid qp-handle in that function. Signed-off-by:
-
Noa Osherovich authored
Struct mlx5_ib_striding_rq_caps was not aligned to 64 bit as it should have been. Add a 32 bit reserved field. Fixes: b4f34597 ('IB/mlx5: Expose multi-packet RQ capabilities') Signed-off-by:
Noa Osherovich <noaos@mellanox.com> Reviewed-by:
Yishai Hadas <yishaih@mellanox.com> Signed-off-by:
Leon Romanovsky <leon@kernel.org> Signed-off-by:
-
Parav Pandit authored
Since there is nothing done with non zero return value, such check is avoided. Signed-off-by:
Parav Pandit <parav@mellanox.com> Reviewed-by:
Daniel Jurgens <danielj@mellanox.com> Signed-off-by:
-
Mark Bloch authored
Maximal message should be used as a limit to the max message payload allowed, without the headers. The ConnectX-3 check is done against this value includes the headers. When the payload is 4K this will cause the NIC to drop packets. Increase maximal message to 8K as workaround, this shouldn't change current behaviour because we continue to set the MTU to 4k. To reproduce; set MTU to 4296 on the corresponding interface, for example: ifconfig eth0 mtu 4296 (both server and client) On server: ib_send_bw -c UD -d mlx4_0 -s 4096 -n 1000000 -i1 -m 4096 On client: ib_send_bw -d mlx4_0 -c UD <server_ip> -s 4096 -n 1000000 -i 1 -m 4096 Fixes: 6e0d733d ("IB/mlx4: Allow 4K messages for UD QPs") Signed-off-by:
Mark Bloch <markb@mellanox.com> Reviewed-by:
Majd Dibbiny <majd@mellanox.com> Signed-off-by:
-
Guy Levi authored
Taking advantage of the optimization which was introduced in previous commit ("IB/mlx4: Use optimal numbers of MTT entries") to optimize the MTT usage for QP and CQ. Signed-off-by:Guy Levi <guyle@mellanox.com> Signed-off-by:
-
Guy Levi authored
Optimize the device performance by assigning multiple physical pages, which are contiguous, to a single MTT. As a result, the number of MTTs is reduced and in turn save cache misses of MTTs. Signed-off-by:
-
Majd Dibbiny authored
When working over a RoCE network, the UDP source port should be set only for statically connected QPs (RC, UC and XRC). Fixes: 2811ba51 ("IB/mlx5: Add RoCE fields to Address Vector") Signed-off-by:
-
Majd Dibbiny authored
The UMR's QP is created by calling mlx5_ib_create_qp directly, and therefore the send CQ and the recv CQ on the ibqp weren't assigned. Assign them right after calling the mlx5_ib_create_qp to assure that any access to those pointers will work as expected and won't crash the system as might happen as part of reset flow. Fixes: e126ba97 ("mlx5: Add driver for Mellanox Connect-IB adapters") Signed-off-by:
-
Leon Romanovsky authored
Smatch tool reports the following error: drivers/infiniband/hw/cxgb4/qp.c:1886 c4iw_create_qp() error: we previously assumed 'ucontext' could be null (see line 1804) Cc: Steve Wise <swise@opengridcomputing.com> Signed-off-by:
-
Leon Romanovsky authored
The Broadcom driver produces the following compilation warning drivers/infiniband/hw/bnxt_re/ib_verbs.c: In function ‘bnxt_re_create_ah’: drivers/infiniband/hw/bnxt_re/ib_verbs.c:668:6: warning: variable ‘vlan_tag’ set but not used [-Wunused-but-set-variable] u16 vlan_tag; Let's remove it till vlan_tag will be implemented properly. Cc: Selvin Xavier <selvin.xavier@broadcom.com> Signed-off-by:
Leon Romanovsky <leonro@mellanox.com> Signed-off-by:
-
- 10 Nov, 2017 8 commits
-
-
Noa Osherovich authored
Add the PCI write end padding flag to device_cap_flags enum and set it during mlx5_ib_query_device so it will be reported to user-space. During WQ/QP creation, set that capability for WQ/QP if user requested it and HW supports it. PCI write end padding modification is not supported for now. There's no such flag for a QP but for a WQ, create and modify use the same flag. Return an error if PCI write end padding flag is set during modify_wq. Signed-off-by:
-
Noa Osherovich authored
There are root complexes that are able to optimize their performance when incoming data is multiple full cache lines. PCI write end padding is the device's ability to pad the ending of incoming packets (scatter) to full cache line such that the last upstream write generated by an incoming packet will be a full cache line. Add a relevant entry to ib_device_cap_flags to report such capability of an RDMA device. Add the QP and WQ create flags: * A QP/WQ created with a scatter end padding flag will cause HW to pad the last upstream write generated by a packet to cache line. User should consider several factors before activating this feature: - In case of high CPU memory load (which may cause PCI back pressure in turn), if a large percent of the writes are partial cache line, this feature should be checked as an optional solution. - This feature might reduce performance if most packets are between one and two cache lines and PCIe throughput has reached its maximum capacity. E.g. 65B packet from the network port will lead to 128B write on PCIe, which may cause traffic on PCIe to reach high throughput. Signed-off-by:
-
Thomas Bogendoerfer authored
Following crash happens, if crc algorithm couldn't be allocated: [ 1087.989072] rdma_rxe: loaded [ 1097.855397] PCLMULQDQ-NI instructions are not detected. [ 1097.901220] rdma_rxe: failed to allocate crc algorithmi err:-2 [ 1097.901248] BUG: unable to handle kernel [ 1097.901249] NULL pointer dereference [ 1097.901250] at 0000000000000046 [...] Reason is that rxe->tfm is assigned the error return, which will then be used for crypto_free_shash() in rxe_cleanup. Fix by using a temporary variable and assigning it rxe->tfm after allocation succeeded. Fixes: cee2688e ("IB/rxe: Offload CRC calculation when possible") Signed-off-by:
Thomas Bogendoerfer <tbogendoerfer@suse.de> Reviewed-by:
Moni Shoua <monis@mellanox.com> Signed-off-by:
-
Parav Pandit authored
Below kernel crash is observed when Pkey security enforcement fails on received MADs. This issue is reported in [1]. ib_free_recv_mad() accesses the rmpp_list, whose initialization is needed before accessing it. When security enformcent fails on received MADs, MAD processing avoided due to security checks failed. OpenSM[3770]: SM port is down kernel: BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 kernel: IP: ib_free_recv_mad+0x44/0xa0 [ib_core] kernel: PGD 0 kernel: P4D 0 kernel: kernel: Oops: 0002 [#1] SMP kernel: CPU: 0 PID: 2833 Comm: kworker/0:1H Tainted: P IO 4.13.4-1-pve #1 kernel: Hardware name: Dell XS23-TY3 /9CMP63, BIOS 1.71 09/17/2013 kernel: Workqueue: ib-comp-wq ib_cq_poll_work [ib_core] kernel: task: ffffa069c6541600 task.stack: ffffb9a729054000 kernel: RIP: 0010:ib_free_recv_mad+0x44/0xa0 [ib_core] kernel: RSP: 0018:ffffb9a729057d38 EFLAGS: 00010286 kernel: RAX: ffffa069cb138a48 RBX: ffffa069cb138a10 RCX: 0000000000000000 kernel: RDX: ffffb9a729057d38 RSI: 0000000000000000 RDI: ffffa069cb138a20 kernel: RBP: ffffb9a729057d60 R08: ffffa072d2d49800 R09: ffffa069cb138ae0 kernel: R10: ffffa069cb138ae0 R11: ffffa072b3994e00 R12: ffffb9a729057d38 kernel: R13: ffffa069d1c90000 R14: 0000000000000000 R15: ffffa069d1c90880 kernel: FS: 0000000000000000(0000) GS:ffffa069dba00000(0000) knlGS:0000000000000000 kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 kernel: CR2: 0000000000000008 CR3: 00000011f51f2000 CR4: 00000000000006f0 kernel: Call Trace: kernel: ib_mad_recv_done+0x5cc/0xb50 [ib_core] kernel: __ib_process_cq+0x5c/0xb0 [ib_core] kernel: ib_cq_poll_work+0x20/0x60 [ib_core] kernel: process_one_work+0x1e9/0x410 kernel: worker_thread+0x4b/0x410 kernel: kthread+0x109/0x140 kernel: ? process_one_work+0x410/0x410 kernel: ? kthread_create_on_node+0x70/0x70 kernel: ? SyS_exit_group+0x14/0x20 kernel: ret_from_fork+0x25/0x30 kernel: RIP: ib_free_recv_mad+0x44/0xa0 [ib_core] RSP: ffffb9a729057d38 kernel: CR2: 0000000000000008 [1] : https://www.spinics.net/lists/linux-rdma/msg56190.html Fixes: 47a2b338 ("IB/core: Enforce security on management datagrams") Cc: stable@vger.kernel.org # 4.13+ Signed-off-by:
Chris Blake <chrisrblake93@gmail.com> Reviewed-by:
Hal Rosenstock <hal@mellanox.com> Signed-off-by:
-
Leon Romanovsky authored
Chelsio cxgb4 HW is big-endian, hence there is need to properly annotate r2 and stag fields as __be32 and not __u32 to fix the following sparse warnings. drivers/infiniband/hw/cxgb4/qp.c:614:16: warning: incorrect type in assignment (different base types) expected unsigned int [unsigned] [usertype] r2 got restricted __be32 [usertype] <noident> drivers/infiniband/hw/cxgb4/qp.c:615:18: warning: incorrect type in assignment (different base types) expected unsigned int [unsigned] [usertype] stag got restricted __be32 [usertype] <noident> Cc: Steve Wise <swise@opengridcomputing.com> Signed-off-by: -
Guy Levi authored
In the modify QP handler the base_qpn_udp field in the RSS QPC is overwrite later by irrelevant value assignment. Hence, ingress packets which gets to the RSS QP will be steered then to a garbage QPN. The patch fixes this by skipping the above assignment when a RSS QP is modified, also, the RSS context's attributes assignments are relocated just before the context is posted to avoid future issues like this. Additionally, this patch takes the opportunity to change the code to be disciplined to the device's manual and assigns the RSS QP context just at RESET to INIT transition. Fixes:3078f5f1 ("IB/mlx4: Add support for RSS QP") Signed-off-by:
-
Guy Levi authored
The mlx4's RSS patches submission missed a report of RSS capabilities which should be reported by the vendor channel in query_device. Signed-off-by:
-
Leon Romanovsky authored
The RDMA/umem uses generic RB-trees macros to generate various ib_umem access functions. The generation is performed with INTERVAL_TREE_DEFINE macro, which allows one of two modes: declare all functions as static or declare none of the function to be static. The second mode of operation produces the following sparse errors: drivers/infiniband/core/umem_rbtree.c:69:1: warning: symbol 'rbt_ib_umem_iter_first' was not declared. Should it be static? drivers/infiniband/core/umem_rbtree.c:69:1: warning: symbol 'rbt_ib_umem_iter_next' was not declared. Should it be static? Code relocation together with declaration of such functions to be "static" solves the issue. Because there is no need to have separate file for two functions, let's consolidate umem_rtree.c and umem_odp.c into one file. Signed-off-by:
-
