Re: [PATCH v2 01/13] seqnum_ops: Introduce Sequence Number Ops

[Shuah Khan]

On 11/17/20 10:38 AM, Matthew Wilcox wrote:
> On Tue, Nov 17, 2020 at 09:34:24AM -0700, Shuah Khan wrote:
> > > seqnum_inc() should just return the new value -- seqnum_inc_return is
> > > too verbose.  And do we not need a seqnum_add()?
> >
> > I had the patch series with seqnum_inc() all ready to go and then
> > revisited the choice. My thinking is that matching the current atomic
> > api that has _inc() and inc_return() might be less confusing. That
>
> No, it's more confusing.  I know you're converting things from using
> atomic_t, but you really need to think about this in terms of "What
> makes sense for this API".  Unless you really want to have inc that
> returns void and inc_return that returns the new value, having only
> inc_return makes no sense.

I am fine with that. As I said I have a patch series saved with just
seqnum_inc() that increments and returns. I anticipated people would
have problems with seqnum_inc() that returns. :)

> > being said, I have no problems with making just _inc(). The reason
> > for 32 and 64 appended is based on comments that it including size
> > in the api makes it very clear.
>
> By putting 32 and 64 in the name of the API, I would contend you're making
> people think about something that they should not need to think about.

Are you recommending seqnum32_*() for 32bit and seqnum_*() for 64bit
which would make 64bit as a default? We have to make a distinction
for 32bit vs. 64-bit api.

> > No need for atomic_add() - inc_return() is sufficient for this use-case.
>
> I haven't looked at the various potential users of this API, but there
> are often cases where we account, eg, number of bytes transmitted.
>
> There are also cases where read-and-zero would be a useful operation
> to have.  I'm thinking about sampling statistics.

The idea is isolating sequence number use-case first and restrict this
api for that.

thanks,
-- Shuah