From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Tyser Date: Sat, 24 Oct 2009 16:30:15 -0500 Subject: [U-Boot] [PATCH 1/5] 8xxx: Add 'ecc' command In-Reply-To: <20091024172110.D07EF185D4A5@gemini.denx.de> References: <1256258353-25589-1-git-send-email-ptyser@xes-inc.com> <20091024172110.D07EF185D4A5@gemini.denx.de> Message-ID: <4AE371E7.3000209@xes-inc.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Hi Wolfgang, > In message <1256258353-25589-1-git-send-email-ptyser@xes-inc.com> you wrote: >> Add a new 'ecc' command to interact with the 85xx and 86xx DDR ECC >> registers. The 'ecc' command can inject data/ECC errors to simulate >> errors and provides an 'info' subcommand which displays ECC error >> information such as failure address, read vs expected data/ECC, >> physical signal which failed, single-bit error count, and multiple bit >> error occurrence. An example of the 'ecc info' command follows: > > We already have similar commands for other architectures, see for > example cpu/mpc83xx/ecc.c > > I'm not sure if it's possible to use a common implementation, but I > would like to ask you to check if this is possible. 83xx, 85xx, and 86xx could all share an implementation I believe. I didn't integrate the 83xx in this patch because it seemed to have a different "goal" than the patch I submitted. The 83xx implementation supported a high degree of tweaking registers which I personally find unnecessary for general use. I think that if someone wants that level of control, they could just modify the registers directly since they have to have the 83xx user's manual handy anyway. The implementation I submitted has limited, common features and much better error reporting. The error reporting is the feature that would be used 98% of the time, not the tweaking of registers. I'd be happy to include the 83xx implementation in this patch, but I'd vote to strip out most of the current 83xx features - ie basically remove the 83xx ecc code and replace it with the 85/86xx implementation I submitted. Would 83xx people be OK with this? Or have any suggestions on what the combined implementation should look like? > In any case I ask that we use a common user interface for both > implementations. It makes no sense that the same command name behaves > differently on different boards (even from the same vendor). I see your point. As far as a common implementation, what did you have in mind? Are you referring to only consolidating the 83xx/85xx/86xx implementations? I'm fine with that, but don't think you could expand the "common interface" much past them as ECC reporting/injection features vary greatly from architecture to architecture. Best, Peter