fallback option is broken, second call to fallback crashes the SNIProxy with sniproxy SIGABRT, first call goes fine.

[balyanrobin]

Hi,

Thanks for providing a great software to the community. I just downloaded the latest Sniproxy code and installed it. I noticed when the request goes to fallback domain it crashes in the second call. First call goes fine. Do you have any Idea?

Thanks i advance for your efforts and time.

Robin

[balyanrobin]

When request does not include sin-extension/hostname, i see below errors in error.log first time and second time for the same call SniProxy crashes -

2015-04-02 07:53:05 Request from [::ffff:xx.xxx.xx.xx]:53536 did not include a hostname
2015-04-02 07:53:05 Invalid hostname (null) in client request

My configuration is below,
listen 443 {
protocol tls
table https_hosts
fallback localhost:8080
access_log {
# Same options as error_log
filename /logs/sni_https.log
priority debug
}
}

[dlundquist]

Could you try it with fallback 127.0.0.1:8080 instead of fallback localhost:8080?

[dlundquist]

I've added a test in 93af73a and have been been able to reproduce this behavior.

[balyanrobin]

Ok great, Do you still want me to test 127.0.0.1 or you got the handle on the issue?

[dlundquist]

@balyanrobin I can't reproduce your crash. Using an IP literal rather than a hostname will allow the fallback server to to be used even if there is a problem resolving DNS queries. I've added a warning in the 0.4.0 release to discourage this configuration, although I could see certain applications where this could be useful so its only a warning. In the case of localhost, using ::1 or 127.0.0.1 is a good idea.

[balyanrobin]

for me it is crashing even if i use IP literal also like : fallback 127.0.0.1:8080

[balyanrobin]

below line in connection.c under resolve_server_address is destroying the fallback memory, and hurting the next connection falling back on the same fallback server. Could you please review it once

free(server_address);