On 10/23/2015 12:54 PM, Aaron Durbin wrote:
Do people realize these binaries sit in cbfs? Are we going to compile random c files into object files then objcopy them? Then add to cbfs? 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.
Nobody said "encode" that in a C struct. The proposal, AFAIU is to have dynamic code that serializes/deserializes from struct to SPD. Last time someone tried to compile a C struct into a binary SPD, it didn't turn out too well (and it's probably why we don't do it).
Not that I agree with the proposal, but we already have deserialization code in the tree. It's not a stretch to write the inverse operation.
On Fri, Oct 23, 2015 at 12:44 PM, ron minnich firstname.lastname@example.org wrote:
Aaron is my hero :-)
On Fri, Oct 23, 2015 at 11:35 AM Aaron Durbin email@example.com wrote:
This one's for Ron.
On Fri, Oct 23, 2015 at 10:32 AM, ron minnich firstname.lastname@example.org wrote:
Build the tool in go. It's trivial. If you have an idea how it ought to work I can set it up in the playground in a few minutes.
On Fri, Oct 23, 2015 at 8:24 AM Patrick Georgi email@example.com wrote: > Hi, > > Some mainboards come with soldered-on memory without SPD ROM. For > these, we carry the SPD data in coreboot. > > Currently, they're stored in a hexdump format that is then converted > to binary at build time. The various mechanisms of doing so fail on > some platform or another: > - "echo -e -n $stuff" isn't well-liked by some shells (emits "e -n > $stuff") > - "printf '\x42'" isn't well-liked by some shells (or /usr/bin/printf > tools) that don't support hexadecimal formats > - "printf '\0377'" isn't well-liked by some non-conforming, but > existing > shells > - xxd -rg1 $file > $file.bin requires xxd, which comes with vim and > may just not exist > > I see essentially two ways out of this, before we can start reducing > duplication across the tree in that area: > We could build our own tool to parse hex files and dump binary, or we > could ship SPD data as binary from the start (and only have to > concatenate them). > > The second option has the appeal of being much simpler (and there > isn't really a "preferred form" for editing SPD data that I'm aware of > - is there?), but looks icky at a glance because it's binary (but it's > really just as impenetrable as the equivalent hexdump). > > So what do these files contain? Parameters (as in: facts) about the > hardware's size, layout, and timing, and a bunch of vendor/model > identifier strings. > > > So, is there a third option that I'm missing? Other opinions? > Patrick >