Re: [RFC PATCH 05/20] x86/virt/tdx: Export tdx module attributes via sysfs

["Huang, Kai" <kai.huang@intel.com>]

On Tue, 2025-06-10 at 09:37 +0800, Chao Gao wrote:
> On Tue, Jun 03, 2025 at 07:49:17AM +0800, Huang, Kai wrote:
> > 
> > > 
> > > Note changes to tdx_global_metadata.{hc} are auto-generated by following
> > > the instructions detailed in [1], after modifying "version" to "versions"
> > > in the TDX_STRUCT of tdx.py to accurately reflect that it is a collection
> > > of versions.
> > > 
> > 
> > [...]
> > 
> > > +static ssize_t version_show(struct device *dev, struct device_attribute *attr,
> > > +			    char *buf)
> > > +{
> > > +	const struct tdx_sys_info_versions *v = &tdx_sysinfo.versions;
> > > +
> > > +	return sysfs_emit(buf, "%u.%u.%u\n", v->major_version,
> > > +					     v->minor_version,
> > > +					     v->update_version);
> > > +}
> > > +
> > > +static DEVICE_ATTR_RO(version);
> > > +
> > 
> > Then for this attribute, I think it is better to name it 'versions' as well?
> 
> Using 'versions' for sysfs might be confusing, as it could imply multiple TDX
> modules. It makes more sense to me that each module has __a version__ in the
> x.y.z format.
> 
> And the convention for sysfs file names is to use 'version'. E.g.,
> 
> # find . -type f -exec grep 'version_show' {} + |wc -l
> 185
> # find . -type f -exec grep 'versions_show' {} + |wc -l
> 0
> 
> Concatenating major_version/minor_version is kinda common inside the kernel,
> but 'versions' is not typically used as a sysfs name.

Sure.

But then should we just use 'version' in the names of the structure and the
variable generated via the script?

It doesn't make a lot sense to me to have this inconsistency.