CAS becomes unresponsive when beacons are sent to port where search requests are expected

[Dirk Zimoch]

This problem happens if an IOC is started when EPICS_CA_ADDR_LIST includes a port number like "gateway:5066" and EPICS_CAS_BEACON_ADDR_LIST is not set. The IOC then sends beacons to port 5066, where the server (e.g. gateway) expects only search requests. CAS starts printing error messages:
CAS: CAS Request: ? on pc3853.psi.ch:38210: cmd=13 cid=0 typ=11 cnt=5064 psz=0 avail=81818260
bad request code=13 in DG
filename="../../../../src/cas/generic/st/casDGIntfOS.cc" line number=498
protocol from client was invalid unexpected problem with UDP input from "pc3853.psi.ch:38210"

The CAS does not terminate, but once confused like this it does not respond to any search requests any more.

Additional information:
The IOC should probably add 1 to the port number in EPICS_CA_ADDR_LIST before using it as the default value for EPICS_CAS_BEACON_ADDR_LIST.

Original Mantis Bug: mantis-302
    http://www.aps.anl.gov/epics/mantis/view_bug_page.php?f_id=302

[Jeff Hill (johill-lanl)]

I agree that it shouldnt do that.

I see the Mantis entry. At some point I will need to reproduce this and see what can be done, but at the moment I need to stay focused on other tasks (that are funded).

edited on: 2007-10-22 17:52

[Jeff Hill (johill-lanl)]

More details from Dirk,

I see strange crashes of the CA gateway which probably originate in the CAS code.

If I use an explicit port in EPICS_CA_ADDR_LIST like this

EPICS_CA_ADDR_LIST="gateway:5066"

and then start a softioc running a record with an INP link to a record behind the gateway (running on that port), the gateway starts printing errors and refuses to handle any more requests. It has to be restarted.

The gateway prints:

CAS: CAS Request: ? on pc3853.psi.ch:38210: cmd=13 cid=0 typ=11 cnt=5064 psz=0 avail=81818260

Oct 12 14:01:36 !!! Errlog message received (message is above) bad request code=13 in DG

Oct 12 14:01:36 !!! Errlog message received (message is above) filename="../../../../src/cas/generic/st/casDGIntfOS.cc" line number=498 protocol from client was invalid unexpected problem with UDP input from "pc3853.psi.ch:38210"

Oct 12 14:01:36 !!! Errlog message received (message is above)
CAS: CAS Request: ? on pc3853.psi.ch:38210: cmd=13 cid=0 typ=11 cnt=5064 psz=0 avail=81818260

Oct 12 14:01:36 !!! Errlog message received (message is above) bad request code=13 in DG

Oct 12 14:01:36 !!! Errlog message received (message is above) filename="../../../../src/cas/generic/st/casDGIntfOS.cc" line number=498 protocol from client was invalid unexpected problem with UDP input from "pc3853.psi.ch:38210"

Oct 12 14:01:36 !!! Errlog message received (message is above) ....

However, a simple caget works without problems, at least before the softioc crashes the gateway.

I tried 3.14.8 and 3.14.9 on the ioc side. The gateway runs with some pre-3.13.9
CVS snapshot. But I also tried it with the 3.14.9 libraries.

edited on: 2007-10-22 17:52

[Jeff Hill (johill-lanl)]

Fixed in R3.14.10

This patch fixes the issue.

@@ -831,6 +831,7 @@
             }
       status = ( this->*pHandler ) ();
       if ( status ) {
+ this->in.removeMsg ( this->in.bytesPresent() );
        break;
       }

[Andrew Johnson (anj)]

R3.14.10 released.