kernel/linux.git/drivers/infiniband/ulp, branch v6.1.87

RDMA/rtrs-clt: Check strnlen return len in sysfs mpath_policy_store()

2024-03-26T22:20:55+00:00

[ Upstream commit 7a7b7f575a25aa68ee934ee8107294487efcb3fe ] strnlen() may return 0 (e.g. for "\0\n" string), it's better to check the result of strnlen() before using 'len - 1' expression for the 'buf' array index. Detected using the static analysis tool - Svace. Fixes: dc3b66a0ce70 ("RDMA/rtrs-clt: Add a minimum latency multipath policy") Signed-off-by: Alexey Kodanev Link: https://lore.kernel.org/r/20240221113204.147478-1-aleksei.kodanev@bell-sw.com Acked-by: Jack Wang Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin

RDMA/srpt: Do not register event handler until srpt device is fully setup

2024-03-26T22:20:54+00:00

[ Upstream commit c21a8870c98611e8f892511825c9607f1e2cd456 ] Upon rare occasions, KASAN reports a use-after-free Write in srpt_refresh_port(). This seems to be because an event handler is registered before the srpt device is fully setup and a race condition upon error may leave a partially setup event handler in place. Instead, only register the event handler after srpt device initialization is complete. Fixes: a42d985bd5b2 ("ib_srpt: Initial SRP Target merge for v3.3-rc1") Signed-off-by: William Kucharski Link: https://lore.kernel.org/r/20240202091549.991784-2-william.kucharski@oracle.com Reviewed-by: Bart Van Assche Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin

RDMA/srpt: fix function pointer cast warnings

2024-03-01T12:26:35+00:00

[ Upstream commit eb5c7465c3240151cd42a55c7ace9da0026308a1 ] clang-16 notices that srpt_qp_event() gets called through an incompatible pointer here: drivers/infiniband/ulp/srpt/ib_srpt.c:1815:5: error: cast from 'void (*)(struct ib_event *, struct srpt_rdma_ch *)' to 'void (*)(struct ib_event *, void *)' converts to incompatible function type [-Werror,-Wcast-function-type-strict] 1815 | = (void(*)(struct ib_event *, void*))srpt_qp_event; Change srpt_qp_event() to use the correct prototype and adjust the argument inside of it. Fixes: a42d985bd5b2 ("ib_srpt: Initial SRP Target merge for v3.3-rc1") Signed-off-by: Arnd Bergmann Link: https://lore.kernel.org/r/20240213100728.458348-1-arnd@kernel.org Reviewed-by: Bart Van Assche Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin

RDMA/srpt: Support specifying the srpt_service_guid parameter

2024-03-01T12:26:35+00:00

[ Upstream commit fdfa083549de5d50ebf7f6811f33757781e838c0 ] Make loading ib_srpt with this parameter set work. The current behavior is that setting that parameter while loading the ib_srpt kernel module triggers the following kernel crash: BUG: kernel NULL pointer dereference, address: 0000000000000000 Call Trace: parse_one+0x18c/0x1d0 parse_args+0xe1/0x230 load_module+0x8de/0xa60 init_module_from_file+0x8b/0xd0 idempotent_init_module+0x181/0x240 __x64_sys_finit_module+0x5a/0xb0 do_syscall_64+0x5f/0xe0 entry_SYSCALL_64_after_hwframe+0x6e/0x76 Cc: LiHonggang Reported-by: LiHonggang Fixes: a42d985bd5b2 ("ib_srpt: Initial SRP Target merge for v3.3-rc1") Signed-off-by: Bart Van Assche Link: https://lore.kernel.org/r/20240205004207.17031-1-bvanassche@acm.org Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin

IB/ipoib: Fix mcast list locking

2024-02-05T20:12:56+00:00

[ Upstream commit 4f973e211b3b1c6d36f7c6a19239d258856749f9 ] Releasing the `priv->lock` while iterating the `priv->multicast_list` in `ipoib_mcast_join_task()` opens a window for `ipoib_mcast_dev_flush()` to remove the items while in the middle of iteration. If the mcast is removed while the lock was dropped, the for loop spins forever resulting in a hard lockup (as was reported on RHEL 4.18.0-372.75.1.el8_6 kernel): Task A (kworker/u72:2 below) | Task B (kworker/u72:0 below) -----------------------------------+----------------------------------- ipoib_mcast_join_task(work) | ipoib_ib_dev_flush_light(work) spin_lock_irq(&priv->lock) | __ipoib_ib_dev_flush(priv, ...) list_for_each_entry(mcast, | ipoib_mcast_dev_flush(dev = priv->dev) &priv->multicast_list, list) | ipoib_mcast_join(dev, mcast) | spin_unlock_irq(&priv->lock) | | spin_lock_irqsave(&priv->lock, flags) | list_for_each_entry_safe(mcast, tmcast, | &priv->multicast_list, list) | list_del(&mcast->list); | list_add_tail(&mcast->list, &remove_list) | spin_unlock_irqrestore(&priv->lock, flags) spin_lock_irq(&priv->lock) | | ipoib_mcast_remove_list(&remove_list) (Here, `mcast` is no longer on the | list_for_each_entry_safe(mcast, tmcast, `priv->multicast_list` and we keep | remove_list, list) spinning on the `remove_list` of | >>> wait_for_completion(&mcast->done) the other thread which is blocked | and the list is still valid on | it's stack.) Fix this by keeping the lock held and changing to GFP_ATOMIC to prevent eventual sleeps. Unfortunately we could not reproduce the lockup and confirm this fix but based on the code review I think this fix should address such lockups. crash> bc 31 PID: 747 TASK: ff1c6a1a007e8000 CPU: 31 COMMAND: "kworker/u72:2" -- [exception RIP: ipoib_mcast_join_task+0x1b1] RIP: ffffffffc0944ac1 RSP: ff646f199a8c7e00 RFLAGS: 00000002 RAX: 0000000000000000 RBX: ff1c6a1a04dc82f8 RCX: 0000000000000000 work (&priv->mcast_task{,.work}) RDX: ff1c6a192d60ac68 RSI: 0000000000000286 RDI: ff1c6a1a04dc8000 &mcast->list RBP: ff646f199a8c7e90 R8: ff1c699980019420 R9: ff1c6a1920c9a000 R10: ff646f199a8c7e00 R11: ff1c6a191a7d9800 R12: ff1c6a192d60ac00 mcast R13: ff1c6a1d82200000 R14: ff1c6a1a04dc8000 R15: ff1c6a1a04dc82d8 dev priv (&priv->lock) &priv->multicast_list (aka head) ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018 Signed-off-by: Sasha Levin

RDMA/IPoIB: Fix error code return in ipoib_mcast_join

2024-02-05T20:12:54+00:00

[ Upstream commit 753fff78f430704548f45eda52d6d55371a52c0f ] Return the error code in case of ib_sa_join_multicast fail. Signed-off-by: Jack Wang Link: https://lore.kernel.org/r/20231121130316.126364-2-jinpu.wang@ionos.com Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin

IB/iser: Prevent invalidating wrong MR

2024-01-25T23:27:38+00:00

[ Upstream commit 2f1888281e67205bd80d3e8f54dbd519a9653f26 ] The iser_reg_resources structure has two pointers to MR but only one mr_valid field. The implementation assumes that we use only *sig_mr when pi_enable is true. Otherwise, we use only *mr. However, it is only sometimes correct. Read commands without protection information occur even when pi_enble is true. For example, the following SCSI commands have a Data-In buffer but never have protection information: READ CAPACITY (16), INQUIRY, MODE SENSE(6), MAINTENANCE IN. So, we use *sig_mr for some SCSI commands and *mr for the other SCSI commands. In most cases, it works fine because the remote invalidation is applied. However, there are two cases when the remote invalidation is not applicable. 1. Small write commands when all data is sent as an immediate. 2. The target does not support the remote invalidation feature. The lazy invalidation is used if the remote invalidation is impossible. Since, at the lazy invalidation, we always invalidate the MR we want to use, the wrong MR may be invalidated. To fix the issue, we need a field per MR that indicates the MR needs invalidation. Since the ib_mr structure already has such a field, let's use ib_mr.need_inval instead of iser_reg_resources.mr_valid. Fixes: b76a439982f8 ("IB/iser: Use IB_WR_REG_MR_INTEGRITY for PI handover") Link: https://lore.kernel.org/r/20231219072311.40989-1-sergeygo@nvidia.com Acked-by: Max Gurtovoy Signed-off-by: Sergey Gorenko Signed-off-by: Jason Gunthorpe Signed-off-by: Sasha Levin

RDMA/rtrs-clt: Remove the warnings for req in_use check

2023-12-13T17:39:14+00:00

[ Upstream commit 0c8bb6eb70ca41031f663b4481aac9ac78b53bc6 ] As we chain the WR during write request: memory registration, rdma write, local invalidate, if only the last WR fail to send due to send queue overrun, the server can send back the reply, while client mark the req->in_use to false in case of error in rtrs_clt_req when error out from rtrs_post_rdma_write_sg. Fixes: 6a98d71daea1 ("RDMA/rtrs: client: main functionality") Signed-off-by: Jack Wang Reviewed-by: Md Haris Iqbal Signed-off-by: Grzegorz Prajsner Link: https://lore.kernel.org/r/20231120154146.920486-8-haris.iqbal@ionos.com Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin

RDMA/rtrs-clt: Fix the max_send_wr setting

2023-12-13T17:39:14+00:00

[ Upstream commit 6d09f6f7d7584e099633282ea915988914f86529 ] For each write request, we need Request, Response Memory Registration, Local Invalidate. Fixes: 6a98d71daea1 ("RDMA/rtrs: client: main functionality") Signed-off-by: Jack Wang Reviewed-by: Md Haris Iqbal Signed-off-by: Grzegorz Prajsner Link: https://lore.kernel.org/r/20231120154146.920486-7-haris.iqbal@ionos.com Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin

RDMA/rtrs-srv: Destroy path files after making sure no IOs in-flight

2023-12-13T17:39:13+00:00

[ Upstream commit c4d32e77fc1006f99eeb78417efc3d81a384072a ] Destroying path files may lead to the freeing of rdma_stats. This creates the following race. An IO is in-flight, or has just passed the session state check in process_read/process_write. The close_work gets triggered and the function rtrs_srv_close_work() starts and does destroy path which frees the rdma_stats. After this the function process_read/process_write resumes and tries to update the stats through the function rtrs_srv_update_rdma_stats This commit solves the problem by moving the destroy path function to a later point. This point makes sure any inflights are completed. This is done by qp drain, and waiting for all in-flights through ops_id. Fixes: 9cb837480424 ("RDMA/rtrs: server: main functionality") Signed-off-by: Md Haris Iqbal Signed-off-by: Santosh Kumar Pradhan Signed-off-by: Grzegorz Prajsner Link: https://lore.kernel.org/r/20231120154146.920486-6-haris.iqbal@ionos.com Signed-off-by: Leon Romanovsky Signed-off-by: Sasha Levin