- 08 Jan, 2018 10 commits
-
-
Bart Van Assche authored
Introduce a function for converting a GUID into an ASCII string. This patch does not change any functionality. Signed-off-by:
Bart Van Assche <bart.vanassche@wdc.com> Signed-off-by:
Doug Ledford <dledford@redhat.com>
-
Bart Van Assche authored
Disabling an SRP target port causes the state of all QPs associated with a port to be changed into IB_QPS_ERR. Avoid that this causes one error message per I/O context to be reported. Signed-off-by:
-
Bart Van Assche authored
At least when running the ib_srpt driver on top of the rdma_rxe driver it is easy to trigger a zero-length write completion in the CH_DISCONNECTED state. Hence make the message that reports this less noisy. Signed-off-by:
-
Bart Van Assche authored
Make the ib_srpt driver use the IPv6 format for GIDs in log messages to improve consistency of this driver with other RDMA kernel drivers. Signed-off-by:
-
Bart Van Assche authored
Verify whether port numbers are in the expected range before using these as an array index. Complain if a port number is out of range. Signed-off-by:
-
Bart Van Assche authored
Since the SRQ event message is only useful for debugging purposes, reduce its severity from "informational" to "debug". Signed-off-by:
-
Bart Van Assche authored
Rename rsp_size into max_rsp_size and SRPT_RQ_SIZE into MAX_SRPT_RQ_SIZE. The new names better reflect the role of this member variable and constant. Since the prefix "srp_" is superfluous in the context of the function that creates an RDMA channel, rename srp_sq_size into sq_size. Signed-off-by:
-
Bart Van Assche authored
This patch avoids that the following command reports any warnings: scripts/kernel-doc -none drivers/infiniband/ulp/srpt/ib_srpt.h Signed-off-by:
-
Bart Van Assche authored
Avoid that warnings about missing parameter descriptions are reported when building with W=1. Signed-off-by:
-
Bart Van Assche authored
Fixes: commit a42d985b ("ib_srpt: Initial SRP Target merge for v3.3-rc1") Signed-off-by:
-
- 04 Jan, 2018 2 commits
-
-
Bart Van Assche authored
Make sure that the initiator port GUID is stored in ch->ini_guid. Note: when initiating a connection sgid and dgid members in struct sa_path_rec represent the source and destination GIDs. When accepting a connection however sgid represents the destination GID and dgid the source GID. Fixes: commit 2bce1a6d ("IB/srpt: Accept GUIDs as port names") Signed-off-by:
Jason Gunthorpe <jgg@mellanox.com>
-
Bart Van Assche authored
With the SRP protocol all RDMA operations are initiated by the target. Since no RDMA operations are initiated by the initiator, do not grant the initiator permission to submit RDMA reads or writes to the target. Signed-off-by:
-
- 02 Jan, 2018 3 commits
-
-
Leon Romanovsky authored
Holding locks is mandatory when calling __ib_device_get_by_index, otherwise there are races during the list iteration with device removal. Since the locks are static to device.c, __ib_device_get_by_index can never be called correctly by any user out side the file. Make the function static and provide a safe function that gets the correct locks and returns a kref'd pointer. Fix all callers. Fixes: e5c9469e ("RDMA/netlink: Add nldev device doit implementation") Fixes: c3f66f7b ("RDMA/netlink: Implement nldev port doit callback") Fixes: 7d02f605 ("RDMA/netlink: Add nldev port dumpit implementation") Reviewed-by:
Mark Bloch <markb@mellanox.com> Signed-off-by:
Leon Romanovsky <leonro@mellanox.com> Signed-off-by:
-
Erez Shitrit authored
When using enhanced mode for IPoIB, two threads may execute xmit in parallel to two different TX queues while the target is the same. In this case, both of them will add the same neighbor to the path's neigh link list and we might see the following message: list_add double add: new=ffff88024767a348, prev=ffff88024767a348... WARNING: lib/list_debug.c:31__list_add_valid+0x4e/0x70 ipoib_start_xmit+0x477/0x680 [ib_ipoib] dev_hard_start_xmit+0xb9/0x3e0 sch_direct_xmit+0xf9/0x250 __qdisc_run+0x176/0x5d0 __dev_queue_xmit+0x1f5/0xb10 __dev_queue_xmit+0x55/0xb10 Analysis: Two SKB are scheduled to be transmitted from two cores. In ipoib_start_xmit, both gets NULL when calling ipoib_neigh_get. Two calls to neigh_add_path are made. One thread takes the spin-lock and calls ipoib_neigh_alloc which creates the neigh structure, then (after the __path_find) the neigh is added to the path's neigh link list. When the second thread enters the critical section it also calls ipoib_neigh_alloc but in this case it gets the already allocated ipoib_neigh structure, which is already linked to the path's neigh link list and adds it again to the list. Which beside of triggering the list, it creates a loop in the linked list. This loop leads to endless loop inside path_rec_completion. Solution: Check list_empty(&neigh->list) before adding to the list. Add a similar fix in "ipoib_multicast.c::ipoib_mcast_send" Fixes: b63b70d8 ('IPoIB: Use a private hash table for path lookup in xmit path') 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:
-
Leon Romanovsky authored
ibmr.device is being set only after ib_alloc_mr() is successfully complete. Therefore, in case imlx4_mr_enable() returns with error, the error flow unwinder calls to mlx4_free_priv_pages(), which uses ibmr.device. Such usage causes to NULL dereference oops and to fix it, the IB device should be set in the mr struct earlier stage (e.g. prior to calling mlx4_free_priv_pages()). Fixes: 1b2cd0fc ("IB/mlx4: Support the new memory registration API") Signed-off-by:
Nitzan Carmi <nitzanc@mellanox.com> Signed-off-by:
-
- 27 Dec, 2017 4 commits
-
-
Nitzan Carmi authored
ibmr.device is being set only after ib_alloc_mr() is (successfully) complete. Therefore, in case mlx5_core_create_mkey() return with error, the error flow calls mlx5_free_priv_descs() which uses ibmr.device (which doesn't exist yet), causing a NULL dereference oops. To fix this, the IB device should be set in the mr struct earlier stage (e.g. prior to calling mlx5_core_create_mkey()). Fixes: 8a187ee5 ("IB/mlx5: Support the new memory registration API") Signed-off-by:
Max Gurtovoy <maxg@mellanox.com> Signed-off-by:
-
Moni Shoua authored
The XRC target QP create flow sets up qp_sec only if there is an IB link with LSM security enabled. However, several other related uAPI entry points blindly follow the qp_sec NULL pointer, resulting in a possible oops. Check for NULL before using qp_sec. Cc: <stable@vger.kernel.org> # v4.12 Fixes: d291f1a6 ("IB/core: Enforce PKey security on QPs") Reviewed-by:
Daniel Jurgens <danielj@mellanox.com> Signed-off-by:
Moni Shoua <monis@mellanox.com> Signed-off-by:
-
Moni Shoua authored
If the input command length is larger than the kernel supports an error should be returned in case the unsupported bytes are not cleared, instead of the other way aroudn. This matches what all other callers of ib_is_udata_cleared do and will avoid user ABI problems in the future. Cc: <stable@vger.kernel.org> # v4.10 Fixes: 189aba99 ("IB/uverbs: Extend modify_qp and support packet pacing") Reviewed-by:
Yishai Hadas <yishaih@mellanox.com> Signed-off-by:
-
Majd Dibbiny authored
User-space applications can do mmap and munmap directly at any time. Since the VMA list is not protected with a mutex, concurrent accesses to the VMA list from the mmap and munmap can cause data corruption. Add a mutex around the list. Cc: <stable@vger.kernel.org> # v4.7 Fixes: 7c2344c3 ("IB/mlx5: Implements disassociate_ucontext API") Reviewed-by:
Majd Dibbiny <majd@mellanox.com> Signed-off-by:
-
- 22 Dec, 2017 1 commit
-
-
Michael J. Ruhl authored
During driver init, various registers are saved to allow restoration after an FLR or gen3 bump. Some of these registers are not available in some circumstances (i.e. Virtual machines). This bug makes the driver unusable when the PCI device is passed into a VM, it fails during probe. Delete unnecessary register read/write, and only access register if the capability exists. Cc: <stable@vger.kernel.org> # 4.14.x Fixes: a618b7e4 ("IB/hfi1: Move saving PCI values to a separate 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:
-
- 21 Dec, 2017 8 commits
-
-
Alex Vesker authored
The locking order of vlan_rwsem (LOCK A) and then rtnl (LOCK B), contradicts other flows such as ipoib_open possibly causing a deadlock. To prevent this deadlock heavy flush is called with RTNL locked and only then tries to acquire vlan_rwsem. This deadlock is possible only when there are child interfaces. [ 140.941758] ====================================================== [ 140.946276] WARNING: possible circular locking dependency detected [ 140.950950] 4.15.0-rc1+ #9 Tainted: G O [ 140.954797] ------------------------------------------------------ [ 140.959424] kworker/u32:1/146 is trying to acquire lock: [ 140.963450] (rtnl_mutex){+.+.}, at: [<ffffffffc083516a>] __ipoib_ib_dev_flush+0x2da/0x4e0 [ib_ipoib] [ 140.970006] but task is already holding lock: [ 140.975141] (&priv->vlan_rwsem){++++}, at: [<ffffffffc0834ee1>] __ipoib_ib_dev_flush+0x51/0x4e0 [ib_ipoib] [ 140.982105] which lock already depends on the new lock. [ 140.990023] the existing dependency chain (in reverse order) is: [ 140.998650] -> #1 (&priv->vlan_rwsem){++++}: [ 141.005276] down_read+0x4d/0xb0 [ 141.009560] ipoib_open+0xad/0x120 [ib_ipoib] [ 141.014400] __dev_open+0xcb/0x140 [ 141.017919] __dev_change_flags+0x1a4/0x1e0 [ 141.022133] dev_change_flags+0x23/0x60 [ 141.025695] devinet_ioctl+0x704/0x7d0 [ 141.029156] sock_do_ioctl+0x20/0x50 [ 141.032526] sock_ioctl+0x221/0x300 [ 141.036079] do_vfs_ioctl+0xa6/0x6d0 [ 141.039656] SyS_ioctl+0x74/0x80 [ 141.042811] entry_SYSCALL_64_fastpath+0x1f/0x96 [ 141.046891] -> #0 (rtnl_mutex){+.+.}: [ 141.051701] lock_acquire+0xd4/0x220 [ 141.055212] __mutex_lock+0x88/0x970 [ 141.058631] __ipoib_ib_dev_flush+0x2da/0x4e0 [ib_ipoib] [ 141.063160] __ipoib_ib_dev_flush+0x71/0x4e0 [ib_ipoib] [ 141.067648] process_one_work+0x1f5/0x610 [ 141.071429] worker_thread+0x4a/0x3f0 [ 141.074890] kthread+0x141/0x180 [ 141.078085] ret_from_fork+0x24/0x30 [ 141.081559] other info that might help us debug this: [ 141.088967] Possible unsafe locking scenario: [ 141.094280] CPU0 CPU1 [ 141.097953] ---- ---- [ 141.101640] lock(&priv->vlan_rwsem); [ 141.104771] lock(rtnl_mutex); [ 141.109207] lock(&priv->vlan_rwsem); [ 141.114032] lock(rtnl_mutex); [ 141.116800] *** DEADLOCK *** Fixes: b4b678b0 ("IB/ipoib: Grab rtnl lock on heavy flush when calling ndo_open/stop") Signed-off-by: -
Majd Dibbiny authored
Congestion counters are counted and queried per physical function. When working in LAG mode, CNP packets can be sent or received on both of the functions, thus congestion counters should be aggregated from the two physical functions. Fixes: e1f24a79 ("IB/mlx5: Support congestion related counters") Signed-off-by:
Aviv Heller <avivh@mellanox.com> Signed-off-by:
-
Bryan Tan authored
The use of wait queues in vmw_pvrdma for handling concurrent access to a resource leaves a race condition which can cause a use after free bug. Fix this by using the pattern from other drivers, complete() protected by dec_and_test to ensure complete() is called only once. Fixes: 29c8d9eb ("IB: Add vmw_pvrdma driver") Signed-off-by:
Bryan Tan <bryantan@vmware.com> Signed-off-by:
-
Bryan Tan authored
refcount_dec generates a warning when the operation causes the refcount to hit zero. Avoid this by using refcount_dec_and_test. Fixes: 8b10ba78 ("RDMA/vmw_pvrdma: Add shared receive queue support") Reviewed-by:
Adit Ranadive <aditr@vmware.com> Reviewed-by:
Aditya Sarwade <asarwade@vmware.com> Reviewed-by:
Jorgen Hansen <jhansen@vmware.com> Signed-off-by:
-
Bryan Tan authored
The QP cleanup did not previously call ib_umem_release, resulting in a user-triggerable kernel resource leak. Fixes: 29c8d9eb ("IB: Add vmw_pvrdma driver") Reviewed-by:
-
Steve Wise authored
If a wr chain was posted and needed to be flushed, only the first wr in the chain was completed with FLUSHED status. The rest were never completed. This caused isert to hang on shutdown due to the missing completions which left iscsi IO commands referenced, stalling the shutdown. Fixes: 4fe7c296 ("iw_cxgb4: refactor sq/rq drain logic") Cc: stable@vger.kernel.org Signed-off-by:
Steve Wise <swise@opengridcomputing.com> Signed-off-by:
-
Steve Wise authored
The flush/drain logic was not retaining the original wr opcode in its completion. This can cause problems if the application uses the completion opcode to make decisions. Use bit 10 of the CQE header word to indicate the CQE is a special drain completion, and save the original WR opcode in the cqe header opcode field. Fixes: 4fe7c296 ("iw_cxgb4: refactor sq/rq drain logic") Cc: stable@vger.kernel.org Signed-off-by:
-
Steve Wise authored
If the RECV CQE is in error, ignore the MSN check. This was causing recvs that were flushed into the sw cq to be completed with the wrong status (BAD_MSN instead of FLUSHED). Cc: stable@vger.kernel.org Signed-off-by:
-
- 13 Dec, 2017 1 commit
-
-
Yuval Shaia authored
memalloc_noio_save modifies the behavior of MM, we must restore it after we are done. Fixes: d83187dd ("IB/IPoIB: Convert IPoIB to memalloc_noio_* calls") Signed-off-by:
Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by:
-
- 11 Dec, 2017 1 commit
-
-
Steve Wise authored
Only insert our special drain CQEs to support ib_drain_sq/rq() after the wq is flushed. Otherwise, existing but not yet polled CQEs can be returned out of order to the user application. This can happen when the QP has exited RTS but not yet flushed the QP, which can happen during a normal close (vs abortive close). In addition never count the drain CQEs when determining how many CQEs need to be synthesized during the flush operation. This latter issue should never happen if the QP is properly flushed before inserting the drain CQE, but I wanted to avoid corrupting the CQ state. So we handle it and log a warning once. Fixes: 4fe7c296 ("iw_cxgb4: refactor sq/rq drain logic") Signed-off-by:
-
- 07 Dec, 2017 5 commits
-
-
Steve Wise authored
In __flush_qp(), the CQ ARMED bit was being cleared regardless of whether any notification is actually needed. This resulted in the iser termination logic getting stuck in ib_drain_sq() because the CQ was not marked ARMED and thus the drain CQE notification wasn't triggered. This new bug was exposed when this commit was merged: commit cbb40fad ("iw_cxgb4: only call the cq comp_handler when the cq is armed") Signed-off-by:
-
Leon Romanovsky authored
The RDMA netlink core code checks validity of messages by ensuring that type and operand are in range. It works well for almost all clients except NLDEV, which has cb_table less than number of operands. Request to access such operand will trigger the following kernel panic. This patch updates all places where cb_table is declared for the consistency, but only NLDEV is actually need it. general protection fault: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN Modules linked in: CPU: 0 PID: 522 Comm: syz-executor6 Not tainted 4.13.0+ #4 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.7.5-0-ge51488c-20140602_164612-nilsson.home.kraxel.org 04/01/2014 task: ffff8800657799c0 task.stack: ffff8800695d000 RIP: 0010:rdma_nl_rcv_msg+0x13a/0x4c0 RSP: 0018:ffff8800695d7838 EFLAGS: 00010207 RAX: dffffc0000000000 RBX: 1ffff1000d2baf0b RCX: 00000000704ff4d7 RDX: 0000000000000000 RSI: ffffffff81ddb03c RDI: 00000003827fa6bc RBP: ffff8800695d7900 R08: ffffffff82ec0578 R09: 0000000000000000 R10: ffff8800695d7900 R11: 0000000000000001 R12: 000000000000001c R13: ffff880069d31e00 R14: 00000000ffffffff R15: ffff880069d357c0 FS: 00007fee6acb8700(0000) GS:ffff88006ca00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000201a9000 CR3: 0000000059766000 CR4: 00000000000006b0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: ? rdma_nl_multicast+0x80/0x80 rdma_nl_rcv+0x36b/0x4d0 ? ibnl_put_attr+0xc0/0xc0 netlink_unicast+0x4bd/0x6d0 ? netlink_sendskb+0x50/0x50 ? drop_futex_key_refs.isra.4+0x68/0xb0 netlink_sendmsg+0x9ab/0xbd0 ? nlmsg_notify+0x140/0x140 ? wake_up_q+0xa1/0xf0 ? drop_futex_key_refs.isra.4+0x68/0xb0 sock_sendmsg+0x88/0xd0 sock_write_iter+0x228/0x3c0 ? sock_sendmsg+0xd0/0xd0 ? do_futex+0x3e5/0xb20 ? iov_iter_init+0xaf/0x1d0 __vfs_write+0x46e/0x640 ? sched_clock_cpu+0x1b/0x190 ? __vfs_read+0x620/0x620 ? __fget+0x23a/0x390 ? rw_verify_area+0xca/0x290 vfs_write+0x192/0x490 SyS_write+0xde/0x1c0 ? SyS_read+0x1c0/0x1c0 ? trace_hardirqs_on_thunk+0x1a/0x1c entry_SYSCALL_64_fastpath+0x18/0xad RIP: 0033:0x7fee6a74a219 RSP: 002b:00007fee6acb7d58 EFLAGS: 00000212 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 0000000000638000 RCX: 00007fee6a74a219 RDX: 0000000000000078 RSI: 0000000020141000 RDI: 0000000000000006 RBP: 0000000000000046 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000212 R12: ffff8800695d7f98 R13: 0000000020141000 R14: 0000000000000006 R15: 00000000ffffffff Code: d6 48 b8 00 00 00 00 00 fc ff df 66 41 81 e4 ff 03 44 8d 72 ff 4a 8d 3c b5 c0 a6 7f 82 44 89 b5 4c ff ff ff 48 89 f9 48 c1 e9 03 <0f> b6 0c 01 48 89 f8 83 e0 07 83 c0 03 38 c8 7c 08 84 c9 0f 85 RIP: rdma_nl_rcv_msg+0x13a/0x4c0 RSP: ffff8800695d7838 ---[ end trace ba085d123959c8ec ]--- Kernel panic - not syncing: Fatal exception Cc: syzkaller <syzkaller@googlegroups.com> Fixes: b4c598a6 ("RDMA/netlink: Implement nldev device dumpit calback") Reviewed-by:
-
Guy Levi authored
Mistakenly the driver didn't allow RSS hash fields combinations which involve both IPv4 and IPv6 protocols. This bug caused to failures for user's use cases for RSS. Consequently, this patch fixes this bug and allows any combination that the HW can support. Additionally, the patch fixes the driver to return an error in case the user provides an unsupported mask for RSS hash fields. Fixes: 3078f5f1 ("IB/mlx4: Add support for RSS QP") Signed-off-by:
Guy Levi <guyle@mellanox.com> Reviewed-by:
-
Daniel Jurgens authored
Per the infiniband spec an SMI MAD can have any PKey. Checking the pkey on SMI MADs is not necessary, and it seems that some older adapters using the mthca driver don't follow the convention of using the default PKey, resulting in false denials, or errors querying the PKey cache. SMI MAD security is still enforced, only agents allowed to manage the subnet are able to receive or send SMI MADs. Reported-by:
Chris Blake <chrisrblake93@gmail.com> Cc: <stable@vger.kernel.org> # v4.12 Fixes: 47a2b338 ("IB/core: Enforce security on management datagrams") Signed-off-by:
Parav Pandit <parav@mellanox.com> Signed-off-by:
-
Daniel Jurgens authored
The alternate port number is used as an array index in the IB security implementation, invalid values can result in a kernel panic. Cc: <stable@vger.kernel.org> # v4.12 Fixes: d291f1a6 ("IB/core: Enforce PKey security on QPs") Signed-off-by:
-
- 01 Dec, 2017 5 commits
-
-
Daniel Jurgens authored
For now the only LSM security enforcement mechanism available is specific to InfiniBand. Bypass enforcement for non-IB link types. This fixes a regression where modify_qp fails for iWARP because querying the PKEY returns -EINVAL. Cc: Paul Moore <paul@paul-moore.com> Cc: Don Dutile <ddutile@redhat.com> Cc: stable@vger.kernel.org Reported-by:
Potnuri Bharat Teja <bharat@chelsio.com> Fixes: d291f1a6("IB/core: Enforce PKey security on QPs") Fixes: 47a2b338("IB/core: Enforce security on management datagrams") Signed-off-by:
-
Wei Hu\(Xavier\) authored
In general, dma_alloc_coherent() returns a CPU virtual address and a DMA address, and we have no guarantee that the underlying memory even has an associated struct page at all. This patch gets rid of the page operation after dma_alloc_coherent, and records the VA returned form dma_alloc_coherent in the struct of hem in hns RoCE driver. Fixes: 9a443537("IB/hns: Add driver files for hns RoCE driver") Signed-off-by:
Wei Hu (Xavier) <xavier.huwei@huawei.com> Signed-off-by:
Shaobo Xu <xushaobo2@huawei.com> Signed-off-by:
Lijun Ou <oulijun@huawei.com> Signed-off-by:
Yixian Liu <liuyixian@huawei.com> Signed-off-by:
Xiping Zhang (Francis) <zhangxiping3@huawei.com> Signed-off-by:
-
Wei Hu\(Xavier\) authored
In general dma_alloc_coherent() returns a CPU virtual address and a DMA address, and we have no guarantee that the virtual address is either in the linear map or vmalloc. It could be in some other special place. We have no guarantee that the underlying memory even has an associated struct page at all. In current code, there are incorrect usage as below: dma_alloc_coherent + virt_to_page + vmap. There will probably introduce coherency problem. This patch fixes it to get rid of virt_to_page and vmap calls at Leon's suggestion. The related link: https://lkml.org/lkml/2017/11/7/34 Fixes: 9a443537("IB/hns: Add driver files for hns RoCE driver") Signed-off-by:
-
Wei Hu\(Xavier\) authored
If the smmu is enabled, the length of sg obtained from __iommu_map_sg_attrs is not 4kB. When the IOVA is set with the sg dma address, the IOVA will not be page continuous. so, the current code has MTPT configuration error that probably cause dma operation failure. In order to fix this issue, the IOVA should be calculated based on the sg length. Fixes: 3958cc56("RDMA/hns: Configure the MTPT in hip08") Signed-off-by:
-
Dmitry Monakhov authored
Once infiniband is compiled as a core component its subsystem must be enabled before device initialization. Otherwise there is a NULL pointer dereference during mlx4_core init, calltrace: ->device_add if (dev->class) { deref dev->class->p =>NULLPTR #Config CONFIG_NET_DEVLINK=y CONFIG_MAY_USE_DEVLINK=y CONFIG_MLX4_EN=y Signed-off-by:Dmitry Monakhov <dmonakhov@openvz.org> Reviewed-by:
-
