Re: [RFC PATCH v2 1/4] liveupdate: Rename handler compatible to name

[tarunsahu@google.com]

Mike Rapoport <rppt@kernel.org> writes:

> On Wed, Oct 21, 2026 at 07:35:28PM +0000, tarunsahu@google.com wrote:
>> Pasha Tatashin <pasha.tatashin@soleen.com> writes:
>> > On Wed, Oct 21, 2026 at 2:16 PM <tarunsahu@google.com> wrote:
>> >> Pasha Tatashin <pasha.tatashin@soleen.com> writes:
>> >>
>> >> > On Wed, Oct 21, 2026 at 11:25 AM Tarun Sahu <tarunsahu@google.com> wrote:
>> >> >>
>> >> >> --- a/include/linux/kho/abi/luo.h
>> >> >> @@ -84,18 +84,16 @@ struct luo_ser {
>> >> >>         u64 flbs_pa;
>> >> >>  } __packed;
>> >> >>
>> >> >> -#define LIVEUPDATE_HNDL_COMPAT_LENGTH  48
>> >> >> -
>> >> >>  /**
>> >> >>   * struct luo_file_ser - Represents the serialized preserves files.
>> >> >> - * @compatible:  File handler compatible string.
>> >> >> + * @name:        File handler name string.
>> >> >
>> >> > We are renaming from cpmaptible to name; I am not sure that is
>> >> > necessary. We can still use compatible for the same purpose as name.
>> >> > Unless it really bothers people, but for me 'compatible' works just
>> >> > the same.
>> >>
>> >> Yes it is necessary, For example if we have two handler for a single
>> >> subsystem. We should refer it by handle this or handler that, not like
>> >> handler with compatible string this and handler with compatible string
>> >> that.
>> >
>> > I still feel like compatability makes more sense to me. Take an example:
>> >
>> > We have two file handlers for memfd (i.e., FH_OLD and FH_NEW): one has
>> > compatibility = "memfd", and the other has compatibility =
>> > "memfd_next_generation". The user chooses one of them to preserve the
>> > memfd data, and the compatibility string is saved alongside the FD.
>> > When the next kernel scans through the registered file handlers, it
>> > must find exactly one that is compatible with the preserved data.
>> 
>> I will open this for vote.
>
> I'd rather won't ;-)
>
> "compatible" does not contradict more fine grained bitfields, and still has meaning.
> Just like in FDT world from where we inherited it, it means "these ops are
> somewhat compatible with that serialized format, will check the details and
> will know for sure".
>
> And IMO we should keep it.
>  
> And please, trim your responses. I had to scroll 463 lines to see there's
> nothing there.

Okay.

I will send the v3 soon with feature flags only. Also I will drop this
and its next patch, as It will be part of other compatiblity series with
exporting the schema to userspace so that userspace can make better
decision in selecting the handler (As per the response from the other thread).

~Tarun

>
> -- 
> Sincerely yours,
> Mike.