On 23.10.2015 21:54, Aaron Durbin wrote: > On Fri, Oct 23, 2015 at 2:43 PM, Carl-Daniel Hailfinger > <c-d.hailfinger.devel.2...@gmx.net> wrote: >> On 23.10.2015 21:15, Martin Roth wrote: >>> I like what I understand Nico's proposal to be: Store the SPD data as >>> a c struct with all of the different JEDEC fields broken out. it >>> would relatively trivial to compile this into an SPD binary, or it can >>> be used in whatever other fashion is desired. A tool to convert from >>> a binary SPD to the structure file would be desirable to help the >>> transition, but it's out of the build path. Thanks for the support. One remark: I would really prefer serializing during runtime over compiling a C struct to a binary. The latter could cause much bit-field headache given the SPD format.
>>> >>> I believe this satisfies all the requirements: >>> - It's easy to review differences. >>> - It can be be built with no new tools. >>> - The fields are broken out, so you can actually tell what you're doing. >> >> Now that would be a nice way to combine the benefits of diffable source >> and no-tool builds. >> >> Ron, is that solution is acceptable to you? >> >> Side note: There is a tool called decode-dimms which can be fed with >> binary SPD dumps. It decodes everything in the SPD and could serve as a >> nice way to verify the output of Ron's magic SPD tool. > > Do people realize these binaries sit in cbfs? Yes, they do _currently_. We don't have to keep it that way. > Are we going to compile > random c files into object files then objcopy them? Then add to cbfs? I wouldn't like that either. > Also, the SPD format is quite silly as it is w.r.t. values crossing > multiple fields in a byte, etc. There's not really a good way to > encode that in a C struct. We don't have to encode the binary SPD format in the C struct. We can just use a C struct to store the SPD information. Like we used C structs before we knew bit fields (wasn't that a good time?). If we encounter a not so beautiful, unchangeable interface that needs the information as binary SPD, well, we could decide to be gracious and serialize the in- formation. Nico -- coreboot mailing list: coreboot@coreboot.org http://www.coreboot.org/mailman/listinfo/coreboot
