Date: Thu, 05 Nov 2020 08:28:46 -0700 From: Ian Lepore <ian@freebsd.org> To: Kurt Jaeger <pi@freebsd.org>, John F Carr <jfc@mit.edu> Cc: "freebsd-arm@freebsd.org" <arm@freebsd.org> Subject: Re: if_mvneta rx/tx byte counts wrong Message-ID: <9b2c7b09be7524fc736c741487b1cfb039ce6eda.camel@freebsd.org> In-Reply-To: <20201104123909.GF53210@home.opsec.eu> References: <57104A54-AC82-4102-A637-116BF8D2E17D@gmail.com> <858E0BCD-C4E9-4997-A982-3C19DF21D161@exchange.mit.edu> <20201104123909.GF53210@home.opsec.eu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 2020-11-04 at 13:39 +0100, Kurt Jaeger wrote: > Hi! > > > If you view in hex, > > > > $ printf '%x %x %x\n' 443193380410133 3813930959736 274877907008 > > 1931500019315 37800000378 4000000040 > > > > you see a 32 bit value duplicated into both halves of a 64 bit > > value. > > Ah. I think Juniper switches have the same problem -- I heard > that this is a timing/atomicity problem of writing 64bit values > on arm32. > I don't think that's likely to be true. We'd see the problem in a lot more places, and the problem would be intermittent, which doesn't appear to be the case here. I would more tend to suspect the arm 32-bit implementation of bus_space_read/write_8(). -- Ian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9b2c7b09be7524fc736c741487b1cfb039ce6eda.camel>