Openssh 3.7.1p1 and 3.7p1 were complied with PAM support. When we try to connect in(to the openssh 3.7.1p1/3.7p1 server) from F-Secure ssh clients the PAM modules are totally getting bypassed. Is there a way to fix this? However there are no problems connecting in from Openssh clients(PAM works fine) The options that were used here were similar to the options used to compile openssh 3.6p1. No problems are encountered when connecting to a 3.6p1 server either from openssh client or a f-secure ssh client.
Read the comment next to UsePAM in sshd_config.
PasswordAuthentication is set to no and UsePAM is set to yes on the sshd_config file Running sshd in debug mode while trying to connect in , shows PAM modules being invoked while coming in from openssh clients but not from f-secure.
Are your F-Secure clients configured to use keyboard-interactive authentication?
F-Secure SSH client for me (on OpenVMS) works fine with UsePAM=yes and PasswordAuthentication=no for the ssh client: SYS$ ssh2 "jmccormick@rowan" Keyboard-interactive: Password: Authentication successful. [jmccormick@rowan jmccormick]$ My F-Secure install by default seems to be using keyboard-interactive as I'm not explicitly enabling it anywhere.
Yes the clients are configured to use keyboard-interactive. The same client connects fine to a 3.6p1 server(no problems with PAM) but has problems talking with 3.7p1 or 3.7.1p1.
The same problem has been noticed on Secure CRT and Putty clients as well. The only client that seems to work so far is the openssh client.
you will have to provide more evidence. A debug trace from the server perhaps? Are you using 3.7.1p2?
Created attachment 463 [details] Debug output from the server and verbose o/p from the client side(both f-secure and openssh) Yes we upgraded to 3.7.1p2 and the problem still persists. Setting UsePAM to yes and PasswordAuthentication to no the f-secure client is not able to login to the machine at all.
You are not even trying challenge response authentication. Try connecting using ssh protocol 2 or looking for a f-secure option "tisauthentication" or similar to enable challenge-response for protocol 1. This does work (it has been tested by a number of developers) - the problem is at the client.
Mass change of RESOLVED bugs to CLOSED