Hi, ALL I am testing on nvmf, but I meet a bug with spdk, the spdk community say it is a bug about *soft-roce*. Here is the link I talk with SPDK community. https://github.com/spdk/spdk/issues/1184#issuecomment-582795537 And there is the stack of hang process.The bug come every time when I restart the network.service. PID: 44459 TASK: ffff9ff048322600 CPU: 0 COMMAND: "reactor_0" #0 [ffffc324ecae7bb0] __schedule at ffffffffbc89d270 #1 [ffffc324ecae7c38] schedule at ffffffffbc89d882 #2 [ffffc324ecae7c48] schedule_timeout at ffffffffbc8a1622 #3 [ffffc324ecae7cd8] wait_for_completion at ffffffffbc89eecf #4 [ffffc324ecae7d38] flush_workqueue at ffffffffbc0c2d65 #5 [ffffc324ecae7dc8] ucma_destroy_id at ffffffffc04bb7e2 [rdma_ucm] #6 [ffffc324ecae7e08] ucma_write at ffffffffc04b929b [rdma_ucm] #7 [ffffc324ecae7e38] __vfs_write at ffffffffbc27cc56 #8 [ffffc324ecae7eb8] vfs_write at ffffffffbc27cf3d #9 [ffffc324ecae7ef0] sys_write at ffffffffbc27d182 #10 [ffffc324ecae7f30] do_syscall_64 at ffffffffbc003774 #11 [ffffc324ecae7f50] entry_SYSCALL_64_after_hwframe at ffffffffbca00081 RIP: 00007fec5bf0169d RSP: 00007ffc251449d0 RFLAGS: 00000293 RAX: ffffffffffffffda RBX: 00000000024ef240 RCX: 00007fec5bf0169d RDX: 0000000000000018 RSI: 00007ffc251449f0 RDI: 000000000000000e RBP: 0000000000000000 R8: 0000000000000000 R9: 0000000000000010 R10: 0000000000000060 R11: 0000000000000293 R12: 00007ffc25144a40 R13: 00000000024efc30 R14: 00000000024efca0 R15: 0000000000000001 ORIG_RAX: 0000000000000001 CS: 0033 SS: 002b Is any one meet this before? What's the right way to handle the RDMA_CM_EVENT_DEVICE_REMOVAL event? The driver what I used is come from MLNX_OFED 4.7-3.2.9.0