Re: [dpdk-dev] [PATCH v6 0/5] Extended xstats API in ethdev library to allow grouping of stats

[Thomas Monjalon <thomas@monjalon.net>]

24/04/2017 14:32, Olivier Matz:
> Hi,
> 
> On Thu, 20 Apr 2017 22:31:35 +0200, Thomas Monjalon <thomas@monjalon.net> 
wrote:
> > 13/04/2017 16:59, Kuba Kozak:
> > > Extended xstats API in ethdev library to allow grouping of stats
> > > logically
> > > so they can be retrieved per logical grouping  managed by the
> > > application.
> > > Changed existing functions rte_eth_xstats_get_names and
> > > rte_eth_xstats_get
> > > to use a new list of arguments: array of ids and array of values.
> > > ABI versioning mechanism was used to support backward compatibility.
> > > Introduced two new functions rte_eth_xstats_get_all and
> > > rte_eth_xstats_get_names_all which keeps functionality of the previous
> > > ones (respectively rte_eth_xstats_get and rte_eth_xstats_get_names)
> > > but use new API inside. Both functions marked as deprecated.
> > > Introduced new function: rte_eth_xstats_get_id_by_name to retrieve
> > > xstats ids by its names.
> > > Extended functionality of proc_info application:
> > > --xstats-name NAME: to display single xstat value by NAME
> > > Updated test-pmd application to use new API.
> > 
> > Applied, thanks
> 
> I'm adapting my application to the upcoming dpdk 17.05. I see
> several problems with this patchset:
> 
> - the API of rte_eth_xstats_get() and rte_eth_xstats_get_names()
>   has been modified, and from what I see it was not announced.
>   It looks that ABI is preserved however.
> 
> - the new functions rte_eth_xstats_get_all() and
>   rte_eth_xstats_get_names_all() are marked as deprecated, which
>   looks strange for new functions.
> 
> About the new api:
> 
> int rte_eth_xstats_get(uint8_t port_id, uint64_t *ids, uint64_t *values,
>        unsigned int n);
> 
> int rte_eth_xstats_get_names(uint8_t port_id,
>        struct rte_eth_xstat_name *xstats_names, unsigned int size,
>        uint64_t *ids);
> 
> - the argument "id" is not at the same place
> 
> - why having "size" in one function and "n" in the second (it was
>   renamed in the patch)?
> 
> - the argument "id" should be const
> 
> - a table of uint64_t is returned in place of the struct rte_eth_xstat
>   table: if no ids are given, the driver cannot return partial or
>   disordered stats anymore. See
>   513c78ae3fd6 ("ethdev: fix extended statistics name index")
> 
> 
> So, I wonder if it wouldn't be more simple to keep the old
> API intact (it would avoid unannounced breakage). The new feature
> can be implemented in an additional API:
> 
>  rte_eth_xstats_get_by_id(uint8_t port_id, const uint64_t *ids,
>        uint64_t *values, unsigned int size)
>  rte_eth_xstats_get_names_by_id(uint8_t port_id, const uint64_t *ids,
>        struct rte_eth_xstat_name *xstats_names, unsigned int size)
> 
> Or:
> 
>  rte_eth_xstats_get_by_id(uint8_t port_id, const uint64_t *ids,
>        struct rte_eth_xstat *values, unsigned int size)
>  rte_eth_xstats_get_names_by_id(uint8_t port_id, const uint64_t *ids,
>        struct rte_eth_xstat_name *xstats_names, unsigned int size)
> 
>  (which would allow to deprecate the old API, but I'm not sure
>   we need to)
> 
> 
> Can we fix that for 17.05?

Bottom line is that I have skipped the complete review of this patchset
because it was not properly split, preventing a good review.
Lesson learned: I won't accept anymore a patchset which is not split
enough to allow a proper overview.

Back to the issues, please try to fix it quickly or we should revert
it for 17.05-rc3.