Hi,
> > for (modlist = modinfo->list; modlist != NULL; modlist =
> > modlist->next) {
> > + if (modlist->value->has_arch) {
> > + if (strcmp(modlist->value->name, module_name) == 0) {
> > + if (!module_arch ||
> > + strcmp(modlist->value->arch, module_arch) != 0) {
> > + return false;
> > + }
> > + }
> > + }
>
> I have a little hard time following the code paths, but IIUC, with this
> change, instead of "module.so" we would have multiple copies of something
> like "module-$arch.so" ?
Not yet with this series, but easily doable on top of this (see other
patch series sent today).
> Then we load them all, read their modinfo section
> and discard the ones with non-matching arch ?
No. There is a utility reading the modinfo section (patch #2), write
out the info as json (patch #2 has the schema), then qemu will read that
json file (patch #13) ...
> for arch specific modules. That would let us load only the modules we know
> are applicable for the system target arch and not need this post-load
> filtering from metadata.
... so it's pre-load filtering, not post-load.
take care,
Gerd
