From: anti.sullin@artecdesign.ee (Anti Sullin) Date: Mon, 01 Feb 2010 00:34:44 +0200 Subject: [patch v2 1/2] at91_udc.c read_fifo timing issue In-Reply-To: <20100129162155.709756286@gmail.com>> References: <20100129162029.998540038@gmail.com>> <20100129162155.709756286@gmail.com>> Message-ID: <4B660584.7020302@artecdesign.ee> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Harro Haan wrote: > This solves the read_fifo timing issue for example seen with CDC Ethernet. [...] Still this bug not fixed? I sent almost the same fix and a lot of debug info with description of the issue and the critical path of the timing to linux-arm-kernel on 2009-03-25 ( http://lists.arm.linux.org.uk/lurker/message/20090325.150843.f515c02f.en.html ) and we had an off-list discussion with David Brownell about another udc patch on 2009-06-19 that later turned out to be the same issue. This fix has solved the problems on our devices and we've been using it for almost a year now. About Harro's patch: * why are the marker1 and marker2 comments necessary to be included? * maybe you should just link to the mailing list and/or move the long comment to patch description instead adding a page-long comment to code; just leave a small comment to explain why the work-around is needed so that nobody would optimize it away? * Is the error check and warning message necessary after we've got rid of the problem? Won't it add problems - as I found out, the first csr read (that comes too early) may return bogus data, you should not use its result at all. -- Anti Sullin Embedded Software Engineer Artec Design LLC Teaduspargi 6/2, 12618, Tallinn, Estonia http://www.artecdesign.ee