Hi, These messages indicate real errors in zero copy bcopy operations. They mean that when sdp_sendmsg() returned to user space, the SDP driver thought that there were in-flight IB send operations pointing to user pages. Bad things can happen in this case.
I have never seen these in simple testing by hand. The patch I sent you yesterday for review adds this test to the error recovery path. If you hit ^C (for example) during a transfer, you may see these messages but no crash. That patch is not in the current code yet because I need to understand why cleanup is not coordinated with completions and thus generating these messages. Were the regressions running with the test patch or did you see these in normal operation? If normal operation, please let me know how to reproduce. Thanks, JIm Jim Mott Mellanox Technologies Ltd. mail: [EMAIL PROTECTED] Phone: 512-294-5481 -----Original Message----- From: Dotan Barak [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 21, 2007 10:01 AM To: Jim Mott; openib-general Subject: the SDP module prints allot of error messages to the /var/log/messages In our nightly regression i noticed that the /var/log/messages is filled with the following error messages: Nov 21 17:28:30 sw186 kernel: sdp_sock(42203:19000): Could not reap -32 in-flight sends Nov 21 17:28:30 sw186 kernel: sdp_sock(42202:19010): Could not reap -2 in-flight sends Nov 21 17:28:30 sw186 kernel: sdp_sock(42203:19000): Could not reap -29 in-flight sends Nov 21 17:28:30 sw186 kernel: sdp_sock(42204:19005): Could not reap -14 in-flight sends Nov 21 17:28:30 sw186 kernel: sdp_sock(42202:19010): Could not reap -2 in-flight sends Nov 21 17:28:30 sw186 kernel: sdp_sock(42202:19010): Could not reap -7 in-flight sends Nov 21 17:28:30 sw186 kernel: sdp_sock(42203:19000): Could not reap -28 in-flight sends Nov 21 17:28:30 sw186 kernel: sdp_sock(42203:19000): Could not reap -4 in-flight sends Nov 21 17:28:30 sw186 kernel: sdp_sock(42203:19000): Could not reap -32 in-flight sends Nov 21 17:28:31 sw186 kernel: sdp_sock(42204:19005): Could not reap -14 in-flight sends Are those error messages are really necessary? thanks Dotan _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
