Date: Fri, 21 Nov 2014 00:29:37 +0100 From: Marko Zec <zec@fer.hr> To: Craig Rodrigues <rodrigc@freebsd.org> Cc: FreeBSD Net <freebsd-net@freebsd.org>, "Bjoern A.Zeeb" <bz@freebsd.org>, Robert Watson <rwatson@freebsd.org> Subject: Re: VIMAGE UDP memory leak fix Message-ID: <20141121002937.4f82daea@x23> In-Reply-To: <CAG=rPVehky00X4MuQQ-_Oe5ezWg52ZZrPASAh9GBy7baYv78CA@mail.gmail.com> References: <CAG=rPVehky00X4MuQQ-_Oe5ezWg52ZZrPASAh9GBy7baYv78CA@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 20 Nov 2014 10:02:46 -0800 Craig Rodrigues <rodrigc@freebsd.org> wrote: > Hi, > > Can folks take a look at this? > > https://reviews.freebsd.org/D1201 All UMA zones used in the network stack have been marked as UMA_ZONE_NOFREE for ages, probably for a reason, so perhaps it might not hurt to provide more insight why and how it suddenly became safe to remove that flag? One possible alternative is to de-virtualize V_udbinfo, V_udpcb_zone etc. which I have suggested a few times in the past but those proposals have been rejected based on expectations that one day our network stack may benefit from more parallelism of decoupled UMA zones for each VNET, though I'm not aware of further developments in that direction. The primary (and in many cases only) reason I have virtualized network stack zones was to simplify tracking of inter-VNET leaks. As bugs that caused such leaks seem to have been cleaned up some 7 years ago or so, I see no technical reason to maintain separate UMA zones for each VNET, especially not those which are size-limited to maxsockets global limit anyhow. Marko
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20141121002937.4f82daea>