Date: Sat, 1 Aug 2015 14:46:38 -0500 From: Alan Cox <alan.l.cox@gmail.com> To: Zbigniew Bodek <zbb@semihalf.com> Cc: K Macy <kmacy@freebsd.org>, imp@freebsd.org, freebsd-current <freebsd-current@freebsd.org>, "freebsd-arm@freebsd.org" <freebsd-arm@freebsd.org> Subject: Re: Buf ring cleanups Message-ID: <CAJUyCcMOJOPnz6NV0eoP9ojdzG1OmLOb6zgCfmhJLY-S6KKMQA@mail.gmail.com> In-Reply-To: <CAG7dG%2ByT%2BhUBzUN3Zdg2MSaFEd77YXhbXxQh2FmS058UyxuLCA@mail.gmail.com> References: <CAG7dG%2ByT%2BhUBzUN3Zdg2MSaFEd77YXhbXxQh2FmS058UyxuLCA@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Jul 30, 2015 at 6:34 AM, Zbigniew Bodek <zbb@semihalf.com> wrote: > Hello, > > I'm writing to ensure what to do with that patch: > https://reviews.freebsd.org/D1945 > > It was created as a result of discussion related to this review: > https://reviews.freebsd.org/D1833 > The patch (D1945) is still waiting to be committed. We really need fix > for ARM in buf_ring so if someone is sure that the patch is OK then > please commit. > > Thanks in advance and best regards > bde@, kib@, and I have been slogging through kern/, net/, and sys/ reviewing the use of memory ordering primitives and correcting problems. We should get to buf_ring.h soon. However, we also need to pause and update the atomic(9) man page. It defines the semantics of acquire and release in a way that is inconsistent with the architectures that natively support the release consistency model, e.g., arm64 and ia64, as well as the C11 and C++11 standards. Lines 330 through 339 of the patch in D1945 is one example of where the incorrect definition in the current man page has an effect. Alan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJUyCcMOJOPnz6NV0eoP9ojdzG1OmLOb6zgCfmhJLY-S6KKMQA>