nvme_tcp: Remove qpair from group->needs_poll when removing it from poll_group
spdk_sock_close() may not complete synchronously. Then the following scenario is possible. ctrlr_disconnected_qpair() calls spdk_sock_close(). Then any request may complete and call _pdu_write_done(). qpair is still associated with a poll_group and is inserted into the group->needs_poll list. After ctrlr_disconnected_qpair() completes, disconnected_qpair_cb() is called. disconnected_qpair_cb() calls spdk_nvme_ctrlr_free_io_qpair(). spdk_nvme_ctrlr_free_io_qpair() calls spdk_nvme_poll_group_remove(). spdk_nvme_poll_group() removes qpair only from the group->disconnected_qpairs list. Even after qpair->poll_group is nullified, qpair is still in the group->needs_poll list. Then spdk_nvme_poll_group_process_completions() polls all qpairs in the group->needs_poll list and hits the assert. Fixes the issue #2390 Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com> Change-Id: I6ede8bd3b7b1a57a34ac61436159975ab6253fbe Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11882 Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com> Community-CI: Mellanox Build Bot Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com> Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
This commit is contained in:
parent
22e8fc83ed
commit
999f0362ab
@ -2294,14 +2294,21 @@ nvme_tcp_poll_group_remove(struct spdk_nvme_transport_poll_group *tgroup,
|
||||
struct spdk_nvme_qpair *qpair)
|
||||
{
|
||||
struct nvme_tcp_qpair *tqpair;
|
||||
struct nvme_tcp_poll_group *group;
|
||||
|
||||
assert(qpair->poll_group_tailq_head == &tgroup->disconnected_qpairs);
|
||||
|
||||
tqpair = nvme_tcp_qpair(qpair);
|
||||
group = nvme_tcp_poll_group(tgroup);
|
||||
|
||||
assert(tqpair->shared_stats == true);
|
||||
tqpair->stats = &g_dummy_stats;
|
||||
|
||||
if (tqpair->needs_poll) {
|
||||
TAILQ_REMOVE(&group->needs_poll, tqpair, link);
|
||||
tqpair->needs_poll = false;
|
||||
}
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user