Hi This is a race issue, I can't reproduce it stably yet. I have not tested the latest kernel. but in fact, I've synced some nvme-tcp patches from lastest upstream, But in fact, the issue still exists. I review nvme-tcp code and found that this issues may exist in nvme_tcp_poll processing. zhang.guanghui@xxxxxxxx From: Xu Chunguang Date: 2025-02-11 15:20 To: Maurizio Lombardi CC: Max Gurtovoy; zhang.guanghui; sagi; kbusch; sashal; linux-kernel; linux-nvme; linux-block Subject: Re: nvme-tcp: fix a possible UAF when failing to send request Hi: Does you have tested the latest kernel, can it reproduce the same issue？ Maurizio Lombardi <mlombard@xxxxxxxxxxxxxxx> 于 2025年2月11日周二 00:40写道： On Mon Feb 10, 2025 at 11:24 AM CET, Max Gurtovoy wrote: > > It seems to me that the HOST_PATH_ERROR handling can be improved in > nvme-tcp. > > In nvme-rdma we use nvme_host_path_error(rq) and nvme_cleanup_cmd(rq) in > case we fail to submit a command.. > > can you try to replacing nvme_tcp_end_request(blk_mq_rq_from_pdu(req), > NVME_SC_HOST_PATH_ERROR); call with the similar logic we use in > nvme-rdma for host path error handling ? Yes, I could try to prepare a patch. In any case, I think the main issue here is that nvme_tcp_poll() should be prevented from racing against io_work... and I also think there is a possible race condition if nvme_tcp_poll() races against the controller resetting code. Maurizio
