[dpdk-dev] [PATCH v8 03/14] eal/pci, ethdev: Remove assumption that port will not be detached

[Bruce Richardson]

On Wed, Feb 18, 2015 at 10:57:25AM +0100, Thomas Monjalon wrote:
> 2015-02-18 15:10, Tetsuya Mukawa:
> > On 2015/02/18 10:54, Tetsuya Mukawa wrote:
> > > On 2015/02/18 9:31, Thomas Monjalon wrote:
> > >> 2015-02-17 15:14, Tetsuya Mukawa:
> > >>> On 2015/02/17 9:36, Thomas Monjalon wrote:
> > >>>> 2015-02-16 13:14, Tetsuya Mukawa:
> > >>>> Is uint8_t sill a good size for hotpluggable virtual device ids?
> > >>> I am not sure it's enough, but uint8_t is widely used in "rte_ethdev.c"
> > >>> as port id.
> > >>> If someone reports it doesn't enough, I guess it will be the time to
> > >>> write a patch to change all uint_8 in one patch.
> > >> It's a big ABI breakage. So if we feel it's going to be required,
> > >> it's better to do it now in 2.0 release I think.
> > >>
> > >> Any opinion?
> > >>
> > > Hi Thomas,
> > >
> > > I agree with it.
> > > I will add an one more patch to change uint8_t to uint16_t.
> > >
> > > Thanks,
> > > Tetsuya
> > >
> > 
> > Hi Thomas,
> > 
> > Could I make sure.
> > After changing uint8_t to uint16_t in "rte_ethdev.[ch]", must I also
> > need to change other applications and libraries that call ethdev APIs?
> > If so, I would not finish it by 23rd.
> > 
> > I've counted how many lines call ethdev APIs that are related to port_id.
> > Could you please check an attached file?
> > It's over 1200 lines. Probably to fix  one of caller, I will need to
> > check how port_id is used, and fix more related lines. So probably
> > thousands lines may need to be fixed.
> > 
> > When is deadline for fixing this changing?
> > Also, if you have a good idea to fix it easier, could you please let me
> > know?
> 
> It was an open question.
> If everybody is fine with 255 ports maximum, let's keep it as is.
> 
I think we are probably ok for now (and forseeable future) with 255 max.

However, if we did change it, I agree that in 2.0 is a very good time to do so.
Since we are expanding the field, rather than shrinking it, I don't see why we
can't just make the change at the ethdev level (and in libs API) in 2.0 and then in
later releases (e.g. 2.1) update the apps and examples to match. That way the
ABI stays the same from 2.0 onwards, and we don't have a huge amount of churn
changing it everywhere late in the 2.0 release cycle.

/Bruce