From: Graeme Russ Date: Mon, 21 Sep 2009 19:44:43 +1000 Subject: [U-Boot] [PATCH 3/6] sbc8548: use I/O accessors In-Reply-To: <20090921084835.6F291832E864@gemini.denx.de> References: <1253493366-26427-1-git-send-email-paul.gortmaker@windriver.com> <4b92f833dfac8470519a71aee9833c0c67edadbf.1253492532.git.paul.gortmaker@windriver.com> <20090921084835.6F291832E864@gemini.denx.de> Message-ID: List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On Mon, Sep 21, 2009 at 6:48 PM, Wolfgang Denk wrote: > Dear Graeme Russ, > > In message you wrote: >> >> > - ecm->eedr = 0xffffffff; /* clear ecm errors */ >> > - ecm->eeer = 0xffffffff; /* enable ecm errors */ >> > + out_be32(&ecm->eedr, 0xffffffff); /* clear ecm errors */ >> > + out_be32(&ecm->eeer, 0xffffffff); /* enable ecm errors */ >> > return 0; >> >> Correct me if I'm wrong, but I thought the general rule was the other >> way (i.e. the way it alreay was). See for example: > > You are wrong. We must use I/O accessors. Thanks for the correction :) > >> http://git.denx.de/cgi-bin/gitweb.cgi?p=u-boot.git;a=commit;h=ed7a1b681de1e31d18d5b92e2767ae8df3241687 > > Ouch. This patch is clearly bloken. Sorry it escaped my attention. Hence my incorrect assumption ;) > > Using a C structure to describe the hardware is a good thing, but we > still must use I/O accessors to access the registers. > > This needs to be fixed in your code. Will do > >> This prevents, for example, accidentally writing words and longs to byte >> sized registers. > > Right. The I/O accessors still allow for strict type checking. Right, which still makes my patch better than the completely unchecked previous version of the SC520 MMCR accessors Regards, G