From Bugzilla Helper: User-Agent: Mozilla/4.76 [en] (X11; U; Linux 2.2.16-3-usb i686) The following error messages are consistantly appearing in /var/log/messages: Mar 2 10:32:46 empress sshd[4309]: error: channel 0: internal error: we do not read, but chan_read_failed for istate 8 Here is a longer snippet: Mar 2 10:32:46 empress sshd[4309]: Accepted publickey for dprice from 10.0.0.1 port 1468 ssh2 Mar 2 10:32:46 empress sshd[4309]: error: channel 0: internal error: we do not read, but chan_read_failed for istate 8 Mar 2 10:32:46 empress PAM_pwdb[4309]: (sshd) session closed for user dprice Mar 2 10:32:48 empress sshd[4315]: Accepted publickey for dprice from 10.0.0.1 port 1469 ssh2 Mar 2 10:32:49 empress sshd[4315]: error: channel 0: internal error: we do not read, but chan_read_failed for istate 8 Mar 2 10:32:49 empress PAM_pwdb[4315]: (sshd) session closed for user dprice Mar 2 10:32:51 empress sshd[4321]: Accepted publickey for dprice from 10.0.0.1 port 1470 ssh2 Mar 2 10:32:51 empress sshd[4321]: error: channel 0: internal error: we do not read, but chan_read_failed for istate 8 Mar 2 10:32:51 empress PAM_pwdb[4321]: (sshd) session closed for user dprice This is non-fatal, but ocassionally I am getting failed connections as part of a test script (I have been unable to get a specific error message - the test script is merely checking whether the output of "ssh -n localhost 'echo hi'" = "hi"). I don't know if these two issues are related. Reproducible: Always Steps to Reproduce: 1. ssh -n localhost 'echo hi' 2. 3. Actual Results: Mostly works, but the error message above appears in /var/log/message. Intermittant and possibly unrelated failures. Expected Results: No error message, no failures.
Just managed to get the failure case on the command line. In the failure case, the "ssh -n empress.2-wit.com 'echo hi'" fails without outputting any text whatsoever and the error message I pointed out _does not_ appear in the log, interestingly enough. I went back through /var/log/messages and confirmed the lack of error message for one of the test script failure cases. This has been happening consistantly during evening test runs, and the two cases I recently managed to reproduce seem to happen after I haven't attempted opening any ssh sessions in awhile - something they have in common with the evening test runs. I recently downgraded my kernel to 2.2.16 with the 2.4.0-test1 USB support backpatch applied since your 2.2.17 kernel's USB support seems to be crashing my kernel. The timing seems to be the same. I am using OpenSSH 2.3.0p1-4.
/var/log/messages from last night also confirms that the error case lacked the error message. Successful runs still seem to consistantly produce the error message on session close. I can say that at least one ssh run that used '-n' but assumed the domain name of the host worked fine after no ssh sessions had been opened in hours. It was a significantly longer run than my test case (several hours), and was still running 1.5 hours later when the next (shorter) ssh run failed. Maybe the length of the run is related? The short test case is the one that has failed as part of the long one three nights in a row. Of course, I don't parse the output of the sucessful command directly, only it's log files, so I might not notice a similar discrepancy. Then again, the error message was present when the long session close and not for the (failed) short session. I will try my test command once an hour or so for awhile and let you know what I discover.
This should no longer happen if you upgrade to OpenSSH 2.5.2p2 in Errata.
This hasn't happened since I upgraded as requested and this bug could probably be closed.