linux-rdma.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Pradeep Satyanarayana <pradeeps-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
To: "Davis, Arlin R" <arlin.r.davis-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Cc: linux-rdma <linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>
Subject: dat_ep_disconnect() with ABRUPT
Date: Thu, 18 Nov 2010 09:40:28 -0800	[thread overview]
Message-ID: <4CE5650C.1090706@linux.vnet.ibm.com> (raw)

Hi Arlin,

We are seeing some issues with dat_ep_disconnect() with ABRUPT flag. In fact it appears that the ABRUPT flag seems
to behave like the GRACEFUL flag. One difference between the DAT1.2 and DAT2.0 appears to be the following:

In dapls_ib_disconnect()


        /* ABRUPT close, wait for callback and DISCONNECTED state */
        if (close_flags == DAT_CLOSE_ABRUPT_FLAG) {
                dapl_os_lock(&ep_ptr->header.lock);
                while (ep_ptr->param.ep_state != DAT_EP_STATE_DISCONNECTED) {
                        dapl_os_unlock(&ep_ptr->header.lock);
                        dapl_os_sleep_usec(10000);
                        dapl_os_lock(&ep_ptr->header.lock);
                }
                dapl_os_unlock(&ep_ptr->header.lock);
        }

this loop exists in DAT2.0 and has been removed in DAT1.2. I am not sure why this is leading to different behaviors
in DAT1.2 and DAT2.0.

One thought is that both DAT1.2 and DAT2.0 have a missing check for ABRUPT flag in dapl_ep_disconnect().

    if ( ep_ptr->param.ep_state == DAT_EP_STATE_ACTIVE_CONNECTION_PENDING ||
         ep_ptr->param.ep_state == DAT_EP_STATE_COMPLETION_PENDING )
    {
        /*
         * Beginning or waiting on a connection: abort and reset the
         * state
         */
        ep_ptr->param.ep_state  = DAT_EP_STATE_DISCONNECTED;

        dapl_os_unlock ( &ep_ptr->header.lock );
        /* disconnect and make sure we get no callbacks */
        (void) dapls_ib_disconnect (ep_ptr, DAT_CLOSE_ABRUPT_FLAG);

        /* clean up connection state */
        dapl_sp_remove_ep (ep_ptr);

        evd_ptr = (DAPL_EVD *) ep_ptr->param.connect_evd_handle;
        dapls_evd_post_connection_event (evd_ptr,
                                        DAT_CONNECTION_EVENT_DISCONNECTED,
                                        (DAT_HANDLE) ep_ptr,
                                        0,
                                        0);
        dat_status = DAT_SUCCESS;
        goto bail;
    }

The if condition above should also have an additional check for disconnect_flags == ABRUPT. If the EP is in a
CONNECTED state and the remote end crashes and this node calls dat_ep_disconnect() with ABRUPT, the if condition
is not true and it is treated as though it was a GRACEFUL disconnect. If you agree with the assessment I can build
a patch.

Thanks
Pradeep

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

             reply	other threads:[~2010-11-18 17:40 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-11-18 17:40 Pradeep Satyanarayana [this message]
     [not found] ` <4CE5650C.1090706-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
2010-11-18 19:51   ` dat_ep_disconnect() with ABRUPT Pradeep Satyanarayana

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4CE5650C.1090706@linux.vnet.ibm.com \
    --to=pradeeps-23vcf4htsmix0ybbhkvfkdbpr1lh4cv8@public.gmane.org \
    --cc=arlin.r.davis-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org \
    --cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).