There has been a nasty problem upon removing an SRP target when the SRP target machine crashed accidently without giving back any IB events. In that case, the admin cannot make use of deleting remote ports for the purpose of tearing down SRP targets as well as SCSI host. One of the reasons was a completion on target->done, the other was the invocation order of srp_disconnect_target() and scsi_remove_host(). Consequence of the latter was unfortunately hanging forever on device_del(), until the target machine comes up again after having rebooted.
That symptom is simply reproducible via sysfs. First of all, trigger an immediate reboot via /proc/sysrq-trigger on the SRP target. target# echo b > /proc/sysrq-trigger After doing that, the Infiniband connection will be completely gone for several minutes at least. Then on the initiator's side, delete a remote port by writing 1 to /sys/class/srp_remote_ports/port-*\:1/delete, e.g.: initiator# echo 1 > /sys/class/srp_remote_ports/port-6\:1/delete , where the SRP remote port to be deleted has its number 6. Then you will see stale SCSI targets remaining despite of rport delete, which is not expected though. That was resulted from device_del() hanging forever on destroying SCSI LLD. The solution consists of two modifications. The first one was already committed to jejb/for-next. See the commit 55d93898 "IB/srp: send disconnect request without waiting for CM timewait exit" by Vu Pham <[email protected]>. That will prevent from waiting for completion. The next one, which this commit is saying about, is changing the invocation order in srp_remove_target(). Call srp_disconnect_target() before scsi_remove_host(). This change will prevent device_del() from hanging indefinitely. This patch is based on the srp-ha-v3.7 tree by Bart Van Assche. See also <https://github.com/advance38/linux/tree/ib-srp-remove-target-v3.7>. If necessary, I could rebase it on the stable tree. Signed-off-by: Dongsu Park <[email protected]> Cc: Sebastian Riemer <[email protected]> Cc: Bart Van Assche <[email protected]> Cc: David Dillow <[email protected]> Cc: Roland Dreier <[email protected]> Cc: Sean Hefty <[email protected]> Cc: Hal Rosenstock <[email protected]> --- drivers/infiniband/ulp/srp/ib_srp.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 307430e..ca4bf40 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -553,10 +553,11 @@ static void srp_remove_target(struct srp_target_port *target) if (scsi_host_added) { srp_del_scsi_host_attr(shost); srp_remove_host(shost); + srp_disconnect_target(target); scsi_remove_host(shost); - } + } else + srp_disconnect_target(target); - srp_disconnect_target(target); ib_destroy_cm_id(target->cm_id); cancel_work_sync(&target->tl_err_work); srp_free_target_ib(target); -- 1.7.10.4 -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
