Experimental Physics and
| |||||||||||||||||
|
I am referring to my e-mail from 11 July 2005. The problem coudn't be solved yet. In the meantime, I got the latest version of dec21x40End.h/c from Wind River. It improved the situation, but didn't solve it to satisfaction. See below. We are experiencing problems here with VxWorks 5.5 and EPICS 3.13.10. In short, if a running IOC loaded with reasonable CA traffic is disconnected from the network for 5 seconds and then reconnected, the network connection will die forever (until reboot). No UDP, TCP, ICMP packets are processed by the IOC anymore. In more detail: Whenever this happens, the serial debug console outputs messages like: sendto failed: S_errno_ENOBUFS 0x195c160 (CA_online): ../online_notify.c: CA beacon error was "S_errno_ENOBUFS" Another observation is that I see several TCP and UDP sockets with 0.0.0.0 addresses, even before I let the system die. See below: TEST-VME-ANDY > inetstatShow Active Internet connections (including servers) PCB Proto Recv-Q Send-Q Local Address Foreign Address (state) -------- ----- ------ ------ ------------------ ------------------ ------- 1d392c0 TCP 0 0 129.129.129.245.5064 129.129.128.46.59589 ESTABLISHED 1d393c8 TCP 0 0 0.0.0.0.5064 0.0.0.0.0 LISTEN 1d39134 TCP 0 0 129.129.129.245.1034 129.129.128.46.6500 CLOSE_WAIT 1d38fa8 TCP 0 0 0.0.0.0.111 0.0.0.0.0 LISTEN 1d38d98 TCP 0 0 0.0.0.0.513 0.0.0.0.0 LISTEN 1d38b88 TCP 0 0 0.0.0.0.23 0.0.0.0.0 LISTEN 1d394d0 UDP 0 0 0.0.0.0.1028 0.0.0.0.0 1d3944c UDP 0 0 0.0.0.0.5064 0.0.0.0.0 1d3923c UDP 0 0 0.0.0.0.1027 0.0.0.0.0 1d38f24 UDP 0 0 0.0.0.0.1026 0.0.0.0.0 1d391b8 UDP 0 0 0.0.0.0.5065 0.0.0.0.0 1d38ea0 UDP 0 0 0.0.0.0.111 0.0.0.0.0 value = 1 = 0x1 TEST-VME-ANDY > Has this maybe to do with the NFS client bug in VxWorks? Finally: it doesn't seem to be an mbuf or cluster allocation problem. We are using higher values than VxWorks default. When I check with mbufShow, ifShow, inetstatShow, netStackDataPoolShow and netStackSysPoolShow, no unusual numbers are shown, except: I can see with netStackDataPoolShow, that the 128byte clusters decrease dramatically from initially 400 free clusters to 250 or with more network load even down to 0 after a disconnect/reconnect. However, they don't have to become 0 in order to kill the network stack. All help is appreciated. Andreas ---------------------------------- Andreas Isenegger SLS Controls Paul Scherrer Institute, WSLA/207 CH-5232 Villigen PSI, Switzerland Tel/Fax: ++41 (56) 310 5125/4413 ----------------------------------
| ||||||||||||||||
ANJ, 02 Sep 2010 |
·
Home
·
News
·
About
·
Base
·
Modules
·
Extensions
·
Distributions
·
Download
·
· Search · EPICS V4 · IRMIS · Talk · Bugs · Documents · Links · Licensing · |