Date: Mon, 30 Jun 2014 19:36:17 -0400 (EDT) From: Rick Macklem <rmacklem@uoguelph.ca> To: Johan Kooijman <mail@johankooijman.com> Cc: FreeBSD Net <freebsd-net@freebsd.org>, freebsd-stable@freebsd.org Subject: Re: ixgbe queue hang fixed in 9.3? Message-ID: <283522359.5786930.1404171377970.JavaMail.root@uoguelph.ca> In-Reply-To: <CAHvs-HW844pk1i_q=2GeGrsf5_5Ha2hhUVn7x4W%2B11ULf%2BFWow@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Johan Kooijman wrote: > Hey all, > > A while ago I started the thread "9.2 ixgbe tx queue hang" on > freebsd-net. > A lot more people had the same as I had. Could anybody tell me if > this > issue is resolved in 9.3 RC? I haven't been able to find anything > about it. > Well, it depends on your definition of "resolved". r264630 (in head) is in 9.3. With this patch, I believe that the hangs will not occur. However, ixgbe (for the 82599 chips, there shouldn't be an issue for the 82598 chips) will do a lot of m_defrag() calls, which does result in cpu overheads. To get rid of the m_defrag() calls, you need to disable TSO. I think a better fix for this is to allow the network device drivers specify a maximum number of mbufs in the list for a TSO segment and have tcp_output() chop the TSO segments up at this limit. (I have an untested patch to do this and it might make it in head someday, but will probably never be MFCible.) So, if the overhead caused by lots of m_defrag() calls isn't a problem for you, it is resolved in 9.3. rick > -- > Met vriendelijke groeten / With kind regards, > Johan Kooijman > _______________________________________________ > freebsd-stable@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-stable > To unsubscribe, send any mail to > "freebsd-stable-unsubscribe@freebsd.org" >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?283522359.5786930.1404171377970.JavaMail.root>