MySQL Bugs: #68196: vio_timeout doesn't work in Solaris, and this will cause vio

Bug #68196	vio_timeout doesn't work in Solaris, and this will cause vio_read block.
Submitted:	28 Jan 2013 2:36	Modified:	21 Jul 2014 20:04
Reporter:	Nan Xiao	Email Updates:
Status:	Verified	Impact on me:	None
Category:	MySQL Server: C API (client library)	Severity:	S2 (Serious)
Version:	5.5.29	OS:	Solaris
Assigned to:		CPU Architecture:	Any
Tags:	block, vio_read, vio_timeout

Description:
Hi, all;

	Now, the code of vio_timeout:
	{
	#if defined(SO_SNDTIMEO) && defined(SO_RCVTIMEO)
		......
		r= setsockopt(vio->sd, SOL_SOCKET, which ? SO_SNDTIMEO : SO_RCVTIMEO,
                IF_WIN((const char*), (const void*))&wait_timeout,
                sizeof(wait_timeout));
                ......
	#else
	#endif
	}
	
	Because Solaris OS doesn't support SO_SNDTIMEO and SO_RCVTIMEO socket options, the vio_timeout can't take effect. And this may cause vio_read block for a long time or ever.
	So does it need extra processing during vio_timeout for Solaris OS? Thanks very much!

Best Regards
Nan Xiao

How to repeat:
This bug occurs in Solaris OS.

When network does not support timeout, client - server protocol uses thread timeouts, which work beautifully, for example, on Solaris.

In short, this has been taken care of.

Hi, Sinisa:

    Thanks very much for your reply!
    How can I use thread timeouts? Does it need special compile options when building client and Connector/ODBC? Thanks in advance!

Best Regards
Nan Xiao

Hi Nan,

after analyzing the sources we concluded that send and receive timeouts are not supported neither in the socket nor in the protocol in Solarys.

setting mysql_options(mysql, MYSQL_OPT_READ_TIMEOUT, readtimeout) comes down to calling setsockopt() and as it was said before such calls have no effect.

However, there is not workaround, which is to use a library that implements the missing functionality.
More details about this library can be found here:

https://www.varnish-cache.org/lists/pipermail/varnish-dev/2011-March/006741.html

Hi, Michael Widenius:

    Thanks very much for your reply, and I will try it.

Best Regards
Nan Xiao

I believe this should be "Verified" since everyone agrees that the bug exists.