[Stgt-devel] Open connections

[FUJITA Tomonori]

On Thu, 27 Mar 2008 21:56:11 +0100 (CET)
"Niels de Carpentier" <stgt at decarpentier.com> wrote:

> >> On Wed, 19 Mar 2008 03:53:36 +0100 (CET)
> >> "Niels de Carpentier" <stgt at decarpentier.com> wrote:
> >>
> >>> The issue where a connection isn't properly closed because
> >>> it_nexus_destroy returns EBUSY.
> >>>
> >>> This probably isn't very important anymore, as the reconnect should now
> >>> close the old connection properly.
> >>
> >> We hit that issue since tgt doesn't implement session reinstatement
> >> properly, I think.
> >>
> >> You still hit that issue with the patch?
> >>
> > Yes, that issue is still there. (And it actually happens before the
> > session reinstatement).
> >
> 
> Some debug information about the close issue:
> 
> tgtd: iscsi_tcp_event_handler(160) connection closed
> tgtd: conn_close(88) connection closed
> tgtd: conn_close(111) Forcing release of tx task 34
> tgtd: conn_close(111) Forcing release of tx task 2f
> tgtd: conn_close(111) Forcing release of tx task 4e
> tgtd: conn_close(111) Forcing release of tx task 66
> tgtd: conn_close(111) Forcing release of tx task 39
> tgtd: conn_close(111) Forcing release of tx task 6a
> tgtd: conn_close(111) Forcing release of tx task 28
> tgtd: conn_close(111) Forcing release of tx task 40
> tgtd: conn_close(111) Forcing release of tx task 3d
> tgtd: conn_close(111) Forcing release of tx task 37
> tgtd: conn_close(111) Forcing release of tx task 2a
> tgtd: conn_close(111) Forcing release of tx task 4c
> tgtd: conn_close(111) Forcing release of tx task 6c
> tgtd: conn_close(111) Forcing release of tx task 4a
> tgtd: conn_close(111) Forcing release of tx task 49
> tgtd: conn_close(111) Forcing release of tx task 48
> tgtd: conn_close(111) Forcing release of tx task 41
> tgtd: conn_close(111) Forcing release of tx task 54
> tgtd: conn_close(111) Forcing release of tx task 38
> tgtd: conn_close(111) Forcing release of tx task 4f
> tgtd: conn_close(111) Forcing release of tx task 15
> tgtd: conn_close(111) Forcing release of tx task 20
> tgtd: it_nexus_destroy(257) 1 1
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 0 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 1 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 2 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 3 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 7 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 8 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 9 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 11 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 12 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 13 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 14 not empty
> tgtd: it_nexus_destroy(265) cmd_hash_list entry 15 not empty
> tgtd: iscsi_tcp_event_handler(167) connection closed
> 
> So it looks like the cmd_hash_list is not actually cleared on a connection
> close.

Thanks,

> Is there anything more I can do to troubleshoot this issue?

Is there any way to reproduce this easily?