- 29 Jun, 2017 2 commits
-
-
Vishwanathapura, Niranjana authored
Stats can be read from atomic context, hence make stats_lock as a spinlock. Fix the following trace with debug kernel. BUG: sleeping function called from invalid context at kernel/locking/mutex.c:238 in_atomic(): 1, irqs_disabled(): 0, pid: 6487, name: sadc Call Trace: dump_stack+0x63/0x90 ___might_sleep+0xda/0x130 __might_sleep+0x4a/0x90 mutex_lock+0x20/0x50 opa_vnic_get_stats64+0x56/0x140 [opa_vnic] dev_get_stats+0x74/0x130 dev_seq_printf_stats+0x37/0x120 dev_seq_show+0x14/0x30 seq_read+0x26d/0x3d0 Reviewed-by:
Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by:
Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by:
Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com> Signed-off-by:
Doug Ledford <dledford@redhat.com>
-
Vishwanathapura, Niranjana authored
Pass GFP_ATOMIC flag to ib_create_send_mad() while sending trap as it can be triggered from the atomic context. Fix the following trace with debug kernel. BUG: sleeping function called from invalid context at mm/slab.h:432 in_atomic(): 1, irqs_disabled(): 0, pid: 1771, name: NetworkManager Call Trace: dump_stack+0x63/0x90 ___might_sleep+0xda/0x130 __might_sleep+0x4a/0x90 __kmalloc+0x19e/0x220 ? ib_create_send_mad+0xea/0x390 [ib_core] ib_create_send_mad+0xea/0x390 [ib_core] opa_vnic_vema_send_trap+0x17b/0x460 [opa_vnic] opa_vnic_vema_report_event+0x57/0x80 [opa_vnic] opa_vnic_mac_send_event+0xaa/0xf0 [opa_vnic] opa_vnic_set_rx_mode+0x17/0x30 [opa_vnic] __dev_set_rx_mode+0x52/0x90 dev_set_rx_mode+0x26/0x40 __dev_open+0xe8/0x140 Reviewed-by:
-
- 14 Jun, 2017 14 commits
-
-
Raju Rangoju authored
Fix memory leaks of iw_cxgb4 module in the exit path Signed-off-by:
Raju Rangoju <rajur@chelsio.com> Reviewed-by:
Steve Wise <swise@opengridcomputing.com> Signed-off-by:
-
Feras Daoud authored
The flow of creating a new child goes through ipoib_vlan_add which allocates a new interface and checks the rtnl_lock. If the lock is taken, restart_syscall will be called to restart the system call again. In this case we are not releasing the already allocated interface, causing a leak. Fixes: 9baa0b03 ("IB/ipoib: Add rtnl_link_ops support") Signed-off-by:
Feras Daoud <ferasda@mellanox.com> Signed-off-by:
Leon Romanovsky <leon@kernel.org> Signed-off-by:
-
Alex Vesker authored
There is no need to re-enable napi since we set the initialized flag before calling ipoib_ib_dev_stop which will disable napi, disabling napi twice is harmless in case it was already disabled. One more reason for this fix is that when using IPoIB new device driver napi is not added to priv, this can lead to kernel panic when rn_ops ndo_open fails. [ 289.755840] invalid opcode: 0000 [#1] SMP [ 289.757111] task: ffff880036964440 ti: ffff880178ee8000 task.ti: ffff880178ee8000 [ 289.757111] RIP: 0010:[<ffffffffa05368d6>] [<ffffffffa05368d6>] napi_enable.part.24+0x4/0x6 [ib_ipoib] [ 289.757111] RSP: 0018:ffff880178eeb6d8 EFLAGS: 00010246 [ 289.757111] RAX: 0000000000000000 RBX: ffff880177a80010 RCX: 000000007fffffff [ 289.757111] RDX: ffffffff81d5f118 RSI: 0000000000000000 RDI: ffff880177a80010 [ 289.757111] RBP: ffff880178eeb6d8 R08: 0000000000000082 R09: 0000000000000283 [ 289.757111] R10: 0000000000000000 R11: 0000000000000000 R12: ffff880175a00000 [ 289.757111] R13: ffff880177a80080 R14: 0000000000000000 R15: 0000000000000001 [ 289.757111] FS: 00007fe2ee346880(0000) GS:ffff88017fc00000(0000) knlGS:0000000000000000 [ 289.757111] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 289.757111] CR2: 00007fffca979020 CR3: 00000001792e4000 CR4: 00000000000006f0 [ 289.757111] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 289.757111] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 [ 289.757111] Stack: [ 289.796027] ffff880178eeb6f0 ffffffffa05251f5 ffff880177a80000 ffff880178eeb718 [ 289.796027] ffffffffa0528505 ffff880175a00000 ffff880177a80000 0000000000000000 [ 289.796027] ffff880178eeb748 ffffffffa051f0ab ffff880175a00000 ffffffffa0537d60 [ 289.796027] Call Trace: [ 289.796027] [<ffffffffa05251f5>] napi_enable+0x25/0x30 [ib_ipoib] [ 289.796027] [<ffffffffa0528505>] ipoib_ib_dev_open+0x175/0x190 [ib_ipoib] [ 289.796027] [<ffffffffa051f0ab>] ipoib_open+0x4b/0x160 [ib_ipoib] [ 289.796027] [<ffffffff814fe33f>] _dev_open+0xbf/0x130 [ 289.796027] [<ffffffff814fe62d>] __dev_change_flags+0x9d/0x170 [ 289.796027] [<ffffffff814fe729>] dev_change_flags+0x29/0x60 [ 289.796027] [<ffffffff8150caf7>] do_setlink+0x397/0xa40 Fixes: cd565b4b ('IB/IPoIB: Support acceleration options callbacks') Signed-off-by:
Alex Vesker <valex@mellanox.com> Signed-off-by:
-
Alex Vesker authored
This patch mekas init_default and uninit_default symmetric with a call to delete napi. Additionally, the uninit_default gained delete napi call in case of init_default fails. Fixes: 515ed4f3 ('IB/IPoIB: Separate control and data related initializations') Signed-off-by:
-
Alex Vesker authored
Limit calls to free_rdma_netdev() for capable devices only. Fixes: cd565b4b ('IB/IPoIB: Support acceleration options callbacks') Signed-off-by:
-
Alex Vesker authored
There is a need to free priv explicitly and not just to release the device, child priv is freed explicitly on remove flow and this patch also includes priv free on error flow in P_key creation and also in add_port. Fixes: cd565b4b ('IB/IPoIB: Support acceleration options callbacks') Signed-off-by:
-
Jia-Ju Bai authored
The driver may sleep under a spin lock, and the function call path is: post_one_send (acquire the lock by spin_lock_irqsave) init_send_wqe copy_from_user --> may sleep There is no flow that makes "qp->is_user" true, and copy_from_user may cause bug when a non-user pointer is used. So the lines of copy_from_user and check of "qp->is_user" are removed. Signed-off-by:Jia-Ju Bai <baijiaju1990@163.com> Reviewed-by:
Leon Romanovsky <leonro@mellanox.com> Acked-by:
Moni Shoua <monis@mellanox.com> Signed-off-by:
-
Ram Amrani authored
Add 64KB PAGE_SIZE support to user-space CQ, SQ and RQ queues. De-facto it means that code was added to translate 64KB pages to smaller 4KB pages that the FW can handle. Otherwise, the FW would wrap (or jump to the next page) when reaching 4KB while the user space library will continue on the same large page. Note that MR code remains as is since the FW supports larger pages for MRs. Signed-off-by:
Ram Amrani <Ram.Amrani@cavium.com> Signed-off-by:
Michal Kalderon <Michal.Kalderon@cavium.com> Signed-off-by:
-
Michal Kalderon authored
Initialize byte_len in work completion of RDMA_READ and RDMA_SEND. Exposed by uDAPL application. Signed-off-by:
-
Selvin Xavier authored
Some issues observed with FMR implementation while running stress traffic. So removing the FMR verbs support for now. Signed-off-by:
Selvin Xavier <selvin.xavier@broadcom.com> Acked-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
-
Devesh Sharma authored
This patch adds code to ring RQ Doorbell aggressively so that the adapter can DMA RQ buffers sooner, instead of DMA all WQEs in the post_recv WR list together at the end of the post_recv verb. Also use spinlock to serialize RQ posting Signed-off-by:
Kalesh AP <kalesh-anakkur.purayil@broadcom.com> Signed-off-by:
Devesh Sharma <devesh.sharma@broadcom.com> Signed-off-by:
-
Somnath Kotur authored
HW stalls out after 0x800000 WQEs are posted for UD QPs. To workaround this problem, driver will send a modify_qp cmd to the HW at around the halfway mark(0x400000) so that FW can accordingly modify the QP context in the HW to prevent this stall. This workaround needs to be done for UD, QP1 and Raw Ethertype packets. Added a counter to keep track of WQEs posted during post_send. Signed-off-by:
Somnath Kotur <somnath.kotur@broadcom.com> Signed-off-by:
-
Selvin Xavier authored
If the host buffers are freed before destroying MR in HW, HW could try accessing these buffers. This could cause a host crash. Fixing the code to avoid this condition. Signed-off-by:
-
Eddie Wai authored
This patch implements the following HW workarounds 1. The SQ depth needs to be augmented by 128 + 1 to avoid running into an Out of order CQE issue 2. Workaround to handle the problem where the HW fast path engine continues to access DMA memory in retranmission mode even after the WQE has already been completed. If the HW reports this condition, driver detects it and posts a Fence WQE. The driver stops reporting the completions to stack until it receives completion for Fence WQE. Signed-off-by:
Eddie Wai <eddie.wai@broadcom.com> Signed-off-by:
Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com> Signed-off-by:
-
- 13 Jun, 2017 1 commit
-
-
Devesh Sharma authored
Fixing a concurrency issue with creq handling. Each caller was given a globally managed crsq element, which was accessed outside a lock. This could result in corruption, if lot of applications are simultaneously issuing Control Path commands. Now, each caller will provide its own response buffer and the responses will be copied under a lock. Also, Fixing the queue full condition check for the CMDQ. As a part of these changes, the control path code is refactored to remove the code replication in the response status checking. Signed-off-by:
-
- 07 Jun, 2017 1 commit
-
-
Roland Dreier authored
Commit eea40b8f ("infiniband: call ipv6 route lookup via the stub interface") introduced a regression in address resolution when connecting to IPv6 destination addresses. The old code called ip6_route_output(), while the new code calls ipv6_stub->ipv6_dst_lookup(). The two are almost the same, except that ipv6_dst_lookup() also calls ip6_route_get_saddr() if the source address is in6addr_any. This means that the test of ipv6_addr_any(&fl6.saddr) now never succeeds, and so we never copy the source address out. This ends up causing rdma_resolve_addr() to fail, because without a resolved source address, cma_acquire_dev() will fail to find an RDMA device to use. For me, this causes connecting to an NVMe over Fabrics target via RoCE / IPv6 to fail. Fix this by copying out fl6.saddr if ipv6_addr_any() is true for the original source address passed into addr6_resolve(). We can drop our call to ipv6_dev_get_saddr() because ipv6_dst_lookup() already does that work. Fixes: eea40b8f ("infiniband: call ipv6 route lookup via the stub interface") Cc: <stable@vger.kernel.org> # 3.12+ Signed-off-by:
Roland Dreier <roland@purestorage.com> Acked-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
-
- 01 Jun, 2017 22 commits
-
-
Majd Dibbiny authored
Commit 9fdca4da (IB/SA: Split struct sa_path_rec based on IB and ROCE specific fields) moved the service_id to be specific attribute for IB and OPA SA Path Record, and thus wasn't assigned for RoCE. This caused to the following kernel panic in the CMA request handler flow: [ 27.074594] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 [ 27.074731] IP: __radix_tree_lookup+0x1d/0xe0 ... [ 27.075356] Workqueue: ib_cm cm_work_handler [ib_cm] [ 27.075401] task: ffff88022e3b8000 task.stack: ffffc90001298000 [ 27.075449] RIP: 0010:__radix_tree_lookup+0x1d/0xe0 ... [ 27.075979] Call Trace: [ 27.076015] radix_tree_lookup+0xd/0x10 [ 27.076055] cma_ps_find+0x59/0x70 [rdma_cm] [ 27.076097] cma_id_from_event+0xd2/0x470 [rdma_cm] [ 27.076144] ? ib_init_ah_from_path+0x39a/0x590 [ib_core] [ 27.076193] cma_req_handler+0x25/0x480 [rdma_cm] [ 27.076237] cm_process_work+0x25/0x120 [ib_cm] [ 27.076280] ? cm_get_bth_pkey.isra.62+0x3c/0xa0 [ib_cm] [ 27.076350] cm_req_handler+0xb03/0xd40 [ib_cm] [ 27.076430] ? sched_clock_cpu+0x11/0xb0 [ 27.076478] cm_work_handler+0x194/0x1588 [ib_cm] [ 27.076525] process_one_work+0x160/0x410 [ 27.076565] worker_thread+0x137/0x4a0 [ 27.076614] kthread+0x112/0x150 [ 27.076684] ? max_active_store+0x60/0x60 [ 27.077642] ? kthread_park+0x90/0x90 [ 27.078530] ret_from_fork+0x2c/0x40 This patch moves it back to the common SA Path Record structure and removes the redundant setter and getter. Tested on Connect-IB and Connect-X4 in Infiniband and RoCE respectively. Fixes: 9fdca4da (IB/SA: Split struct sa_path_rec based on IB ands ROCE specific fields) Signed-off-by:
Majd Dibbiny <majd@mellanox.com> Reviewed-by:
Parav Pandit <parav@mellanox.com> Signed-off-by:
-
Leon Romanovsky authored
Add mmap_sem lock around VMA inspection in ib_umem_odp_get(). Fixes: 0008b84e ('IB/umem: Add support to huge ODP') Signed-off-by:
Artemy Kovalyov <artemyko@mellanox.com> Reviewed-by:
Eli Cohen <eli@mellanox.com> Signed-off-by:
-
Qing Huang authored
This change will optimize kernel memory deregistration operations. __ib_umem_release() used to call set_page_dirty_lock() against every writable page in its memory region. Its purpose is to keep data synced between CPU and DMA device when swapping happens after mem deregistration ops. Now we choose not to set page dirty bit if it's already set by kernel prior to calling __ib_umem_release(). This reduces memory deregistration time by half or even more when we ran application simulation test program. Signed-off-by:
Qing Huang <qing.huang@oracle.com> Signed-off-by:
-
Leon Romanovsky authored
Commit 57520751 ("IB/SA: Add OPA path record type") introduced new local function __ib_copy_path_rec_to_user, but didn't limit its scope. This produces the following sparse warning: drivers/infiniband/core/uverbs_marshall.c:99:6: warning: symbol '__ib_copy_path_rec_to_user' was not declared. Should it be static? In addition, it used sizeof ... notations instead of sizeof(...), which is correct in C, but a little bit misleading. Let's change it too. Fixes: 57520751 ("IB/SA: Add OPA path record type") Reviewed-by:
Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by:
-
Leon Romanovsky authored
RDMA netlink is part of ib_core, hence ibnl_chk_listeners(), ibnl_init() and ibnl_cleanup() don't need to be published in public header file. Let's remove EXPORT_SYMBOL from ibnl_chk_listeners() and move all these functions to private header file. CC: Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by:
-
Israel Rukshin authored
If srp_init_qp() fails at srp_create_ch_ib() then ch->send_cq may be NULL. Calling directly to ib_destroy_qp() is sufficient because no work requests were posted on the created qp. Fixes: 9294000d ("IB/srp: Drain the send queue before destroying a QP") Cc: <stable@vger.kernel.org> Signed-off-by:
Israel Rukshin <israelr@mellanox.com> Reviewed-by:
Max Gurtovoy <maxg@mellanox.com> Reviewed-by: Bart van Assche <bart.vanassche@sandisk.com>-- Signed-off-by:
-
Leon Romanovsky authored
ipoib_dev_uninit_default() call is used in ipoib_main.c file only and it generates the following warning from smatch tool: drivers/infiniband/ulp/ipoib/ipoib_main.c:1593:6: warning: symbol 'ipoib_dev_uninit_default' was not declared. Should it be static? so let's declare that function as static. Signed-off-by:
-
Honggang Li authored
ipoib_dev_init accesses the wrong private data for the IPoIB device. Commit cd565b4b (IB/IPoIB: Support acceleration options callbacks) changed ipoib_priv from being identical to netdev_priv to being an area inside of, but not the same pointer as, the netdev_priv pointer. As such, the struct we want is the ipoib_priv area, not the netdev_priv area, so use the right accessor, otherwise we kernel panic. [ 27.271938] IPv6: ADDRCONF(NETDEV_CHANGE): mlx5_ib0.8006: link becomes ready [ 28.156790] BUG: unable to handle kernel NULL pointer dereference at 000000000000067c [ 28.166309] IP: ib_query_port+0x30/0x180 [ib_core] ... [ 28.306282] RIP: 0010:ib_query_port+0x30/0x180 [ib_core] ... [ 28.393337] Call Trace: [ 28.397594] ipoib_get_link_ksettings+0x66/0xe0 [ib_ipoib] [ 28.405274] __ethtool_get_link_ksettings+0xa0/0x1c0 [ 28.412353] speed_show+0x74/0xa0 [ 28.417503] dev_attr_show+0x20/0x50 [ 28.422922] ? mutex_lock+0x12/0x40 [ 28.428179] sysfs_kf_seq_show+0xbf/0x1a0 [ 28.434002] kernfs_seq_show+0x21/0x30 [ 28.439470] seq_read+0x116/0x3b0 [ 28.444445] ? do_filp_open+0xa5/0x100 [ 28.449774] kernfs_fop_read+0xff/0x180 [ 28.455220] __vfs_read+0x37/0x150 [ 28.460167] ? security_file_permission+0x9d/0xc0 [ 28.466560] vfs_read+0x8c/0x130 [ 28.471318] SyS_read+0x55/0xc0 [ 28.475950] do_syscall_64+0x67/0x150 [ 28.481163] entry_SYSCALL64_slow_path+0x25/0x25 ... [ 28.584493] ---[ end trace 3549968a4bf0aa5d ]--- Fixes: cd565b4b (IB/IPoIB: Support acceleration options callbacks) Fixes: 0d7e2d21 (IB/ipoib: add get_link_ksettings in ethtool) Signed-off-by:
Honggang Li <honli@redhat.com> Reviewed-by:
-
Gustavo A. R. Silva authored
Add null check before dereferencing pointer sgid_attr.ndev inside function rdma_vlan_dev_vlan_id(). Addresses-Coverity-ID: 1373979 Signed-off-by:
Gustavo A. R. Silva <garsilva@embeddedor.com> Acked-by:
-
Max Gurtovoy authored
Cache the needed umr_fence and set the wqe ctrl segmennt accordingly. Signed-off-by:
Sagi Grimberg <sagi@grimberg.me> Signed-off-by:
-
Max Gurtovoy authored
HW can implement UMR wqe re-transmission in various ways. Thus, add HCA cap to distinguish the needed fence for UMR to make sure that the wqe wouldn't fail on mkey checks. Signed-off-by:
-
Jack Morgenstein authored
The cited patch added a type field to structures ib_ah and rdma_ah_attr. Function mlx4_ib_query_ah() builds an rdma_ah_attr structure from the data in an mlx4_ib_ah structure (which contains both an ib_ah structure and an address vector). For mlx4_ib_query_ah() to work properly, the type field in the contained ib_ah structure must be set correctly. In the outgoing MAD tunneling flow, procedure mlx4_ib_multiplex_mad() paravirtualizes a MAD received from a slave and sends the processed mad out over the wire. During this processing, it populates an mlx4_ib_ah structure and calls mlx4_ib_query_ah(). The cited commit overlooked setting the type field in the contained ib_ah structure before invoking mlx4_ib_query_ah(). As a result, the type field remained uninitialized, and the rdma_ah_attr structure was incorrectly built. This resulted in improperly built MADs being sent out over the wire. This patch properly initializes the type field in the contained ib_ah structure before calling mlx4_ib_query_ah(). The rdma_ah_attr structure is then generated correctly. Fixes: 44c58487 ("IB/core: Define 'ib' and 'roce' rdma_ah_attr types") Signed-off-by:
Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by:
-
Mike Marciniszyn authored
The handling of IB_RDMA_WRITE_ONLY_WITH_IMMEDIATE will leak a memory reference when a buffer cannot be allocated for returning the immediate data. The issue is that the rkey validation has already occurred and the RNR nak fails to release the reference that was fruitlessly gotten. The the peer will send the identical single packet request when its RNR timer pops. The fix is to release the held reference prior to the rnr nak exit. This is the only sequence the requires both rkey validation and the buffer allocation on the same packet. Cc: Stable <stable@vger.kernel.org> # 4.7+ Tested-by:
Tadeusz Struk <tadeusz.struk@intel.com> Reviewed-by:
-
Byczkowski, Jakub authored
Keep VL15 credits at 0 during LNI, before link-up. Store VL15 credits value during verify cap interrupt and set in after link-up. This addresses an issue where VL15 MAD packets could be sent by one side of the link before the other side is ready to receive them. Reviewed-by:
Dean Luick <dean.luick@intel.com> Reviewed-by:
Jakub Byczkowski <jakub.byczkowski@intel.com> Signed-off-by:
-
Steven L. Roberts authored
The Omni-Path adapter driver fails to load on the ppc64le platform due to invalid PCI setup. This patch makes the PCI configuration more robust and will fix 64 bit addressing for ppc64le. Signed-off-by:
Steven L Roberts <robers97@gmail.com> Acked-by:
-
Steven L. Roberts authored
This fixes a kernel panic when loading the hfi driver as a dynamic module. Signed-off-by:
-
Raju Rangoju authored
Take care of ipv6 checks while computing header length for deducing mtu size of ipv6 servers. Due to the incorrect header length computation for ipv6 servers, wrong mss is reported to the peer (client). Signed-off-by:
Ganesh Goudar <ganeshgr@chelsio.com> Acked-by:
-
Ganesh Goudar authored
use egrstatuspagesize to calculate t4_eq_status_entries. Fixes: bb58d079 ("cxgb4: Update IngPad and IngPack values") Reported-by:
Logan Gunthorpe <logang@deltatee.com> Signed-off-by:
Potnuri Bharat Teja <bharat@chelsio.com> Signed-off-by:
-
Raju Rangoju authored
The patch 761e19a5 (RDMA/iw_cxgb4: Handle return value of c4iw_ofld_send() in abort_arp_failure()) from May 6, 2016 leads to the following static checker warning: drivers/infiniband/hw/cxgb4/cm.c:575 abort_arp_failure() warn: passing freed memory 'skb' Also fixes skb leak when l2t resolution fails Fixes: 761e19a5 (RDMA/iw_cxgb4: Handle return value of c4iw_ofld_send() in abort_arp_failure()) Reported-by:
Dan Carpenter <dan.carpenter@oracle.com> Cc: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Tatyana Nikolova authored
Explicitly ACK the MPA Reply frame so the peer does not retransmit the frame. Signed-off-by:
Tatyana Nikolova <tatyana.e.nikolova@intel.com> Signed-off-by:
-
Tatyana Nikolova authored
Don't set control flag for 0-length FULPDU (Send) RTR indication in the enhanced MPA Request/Reply frames, because it isn't supported. Signed-off-by:
-
Gustavo A. R. Silva authored
Refactor code to avoid identical code for different branches. Addresses-Coverity-ID: 1357356 Reviewed-by:
Shiraz Saleem <shiraz.saleem@intel.com> Signed-off-by:
-
