blktests failures with v7.4-rc1 kernel

blktests failures with v7.4-rc1 kernel

[Shin'ichiro Kawasaki]

Hi all,

I ran the latest blktests (git hash: a65cf6bdca19) with the v7.4-rc1 kernel. I
observed 4 failures listed below. Compared with the previous report for v7.3
kernel [1], 2 new failure was observed. The failure #4 nvme/059 needs further
work. Help for it will be appreciated.


[1] https://lore.kernel.org/linux-block/atgwg4i6C1zGi1HU@shinmob/


List of failures
================
#1: nvme/004,031  (rdma transport)
#2: nvme/037 hang  (new)(rdma transport)
#3: nvme/052
#4: nvme/059  (new)(fc transport)(kmemleak)


Failure description
===================

#1: nvme/004,031 (rdma transport)

    The test cases nvme/004 and nvme/031 fail occasionally for rdma transport
    due to the lockdep WARN related to the two locks nvmet_config_sem and
    &id_priv->handler_mutex. It is easier to recreate with rxe driver. It is
    rather rare, but the failure is observed with siw driver also.

    The trigger commit was

     2a8513091d2f ("nvme-rdma: parallelize I/O queue allocation and startup")

    in v7.3-rc1. I posted fix patch candidate [2].

    [2] https://lore.kernel.org/linux-nvme/20261108120154.148824-1-shinichiro.kawasaki@wdc.com/

#2: nvme/037 hang  (rdma transport)

    When the test case nvme/037 is repeated several times for rdma transport,
    it hangs [3]. Nilay provided a fix patch, which was already applied for
    v7.4-rcX [4].

    [3] https://lore.kernel.org/linux-nvme/auhpmq499RVE9YpC@shinmob/
    [4] https://lore.kernel.org/linux-nvme/20261105123642.2120548-2-nilay@linux.ibm.com/

#3: nvme/052

    When it is repeated several times, the test case nvme/052 fails [5]. Nilay
    provided a fix patch [6].

    [5] https://lore.kernel.org/all/ar30Q5H1fuyFtDwM@shinmob/
    [6] https://lore.kernel.org/linux-nvme/20261002070249.338225-1-nilay@linux.ibm.com/

#4: nvme/059  (fc transport)(kmemleak)

    When the test cases nvme/057, 058 and 059 are repeated several times for
    fc transport, a kmemleak was reported at the test case nvme/059:

    unreferenced object 0xffff88812a768000 (size 512):
      comm "kworker/1:3H", pid 92985, jiffies 4301970763
      hex dump (first 32 bytes):
        40 cd f8 4b 81 88 ff ff 00 00 00 00 00 00 00 00  @..K............
        00 00 00 00 ad 4e ad de ff ff ff ff 00 00 00 00  .....N..........
      backtrace (crc 3cd92a37):
        __kmalloc_cache_noprof+0x4c1/0x6d0
        fcloop_fcp_req+0xc4/0x650 [nvme_fcloop]
        nvme_fc_start_fcp_op.part.0+0x642/0x17a0 [nvme_fc]
        nvme_fc_queue_rq+0x392/0x6e0 [nvme_fc]
        blk_mq_dispatch_rq_list+0x3e0/0x2400
        __blk_mq_sched_dispatch_requests+0x20a/0x15d0
        blk_mq_sched_dispatch_requests+0xab/0x150
        blk_mq_run_work_fn+0x1dc/0x2e0
        process_one_work+0x944/0x1760
        worker_thread+0x5fd/0xfe0
        kthread+0x36d/0x470
        ret_from_fork+0x653/0x9d0
        ret_from_fork_asm+0x1a/0x30