Re: [PATCH v9 3/7] iommu/uapi: Introduce enum type for PASID data format

[Jacob Pan <jacob.jun.pan@linux.intel.com>]

Hi Joerg,

On Thu, 24 Sep 2020 10:40:16 +0200, Joerg Roedel <joro@8bytes.org> wrote:

> > On Fri, 18 Sep 2020 11:44:50 +0200, Joerg Roedel <joro@8bytes.org>
> > wrote: 
> > > On Fri, Sep 11, 2020 at 02:57:52PM -0700, Jacob Pan wrote:  
> > > > There can be multiple vendor-specific PASID data formats used in
> > > > UAPI structures. This patch adds enum type with a last entry which
> > > > makes range checking much easier.    
> > > 
> > > But it also makes it much easier to screw up the numbers (which are
> > > ABI) by inserting a new value into the middle. I prefer defines here,
> > > or alternativly BUILD_BUG_ON() checks for the numbers.
> > >   
> > I am not following, the purpose of IOMMU_PASID_FORMAT_LAST *is* for
> > preparing the future insertion of new value into the middle.
> > The checking against IOMMU_PASID_FORMAT_LAST is to protect ABI
> > compatibility by making sure that out of range format are rejected in
> > all versions of the ABI.  
> 
> But with the enum you could have:
> 
> enum {
> 	VTD_FOO,
> 	SMMU_FOO,
> 	LAST,
> };
> 
> which makes VTD_FOO==0 and SMMU_FOO==1, and when in the next version
> someone adds:
> 
> enum {
> 	VTD_FOO,
> 	VTD_BAR,
> 	SMMU_FOO,
> 	LAST,
> };
> 
> then SMMU_FOO will become 2 and break ABI. So I'd like to have this
> checked somewhere.
Got your point, will change to defines.

Thanks,

Jacob