Re: [Qemu-devel] [ping: PATCH v2 0/3] e1000: allow model/device_id selection on command line

[Gabriel L. Somlo]

ping ?

On Wed, May 21, 2014 at 02:27:41PM -0400, Gabriel L. Somlo wrote:
> This started out as a single patch (now patch 2/3):
> 
>     Allow selection of different card models from the qemu
>     command line, to better accomodate a wider range of guests.
> 
> New in v2:
> 
>   - moved check for 8257x out of the way of QOM, as suggested by
>     Michael (patch 1/3)

http://patchwork.ozlabs.org/patch/351274/

>   - resolved "Signed-off-by" misunderstanding and miscellaneous style
>     issues (patch 2/3)

http://patchwork.ozlabs.org/patch/351273/

>   - modified e1000 test to check for all supported models, as suggested
>     by Andreas (patch 3/3). I used eepro100-test.c as an example for
>     this change.

http://patchwork.ozlabs.org/patch/351275/

> On Wed, May 21, 2014 at 12:04:57PM +0300, Michael S. Tsirkin wrote:
> > On Wed, May 21, 2014 at 10:40:54AM +0200, Andreas F?rber wrote:
> > >    static inline uint16_t maybe?
> > 
> > inline in c file is an optimization hint for a compiler, so
> > really useless without a benchmark.
> > It's useful in headers since some compilers warn about unused
> > non-inline static functions.
> 
> I went with "static inline uint16_t" in both 1/3 and 2/3, since each
> only get called once, and I'm only using a function instead of a
> preprocessor macro for aesthetic reasons. I'm not religious about it
> though, so if anyone still objects I can change it to whatever we all
> agree on :)
> 
> > > > +    d->phy_reg[PHY_ID2] = e1000_phy_id2_init(dev_id);
> > 
> > I would just pass E1000State to e1000_phy_id2_init.
> 
> So I'm still passing "uint16_t device_id" instead of E1000State,
> because e1000_phy_id2_init() replaces "enum { PHY_ID2_INIT = ...}",
> which goes before the E1000State typedef in the source, and I wanted
> to have the new code show up in the same spot in the patch as the code
> it replaces. That, and we wouldn't save anything either way, I'd still
> have to de-ref E1000State to get at the device_id, either in
> e1000_reset() or in e1000_phy_id2_init(). Again, not religious about
> it, so I can change it if you really want me to :)
> 
> Thanks again,
>   Gabriel
> 
> Gabriel L. Somlo (3):
>   e1000: avoid relying on device id (and soon, QOM) on data path
>   e1000: allow command-line selection of card model
>   tests: e1000: test additional device IDs
> 
>  hw/net/e1000.c     | 127 
> ++++++++++++++++++++++++++++++++++++++++++++++-------
>  tests/e1000-test.c |  33 +++++++++++---
>  2 files changed, 136 insertions(+), 24 deletions(-)
> 
> -- 
> 1.9.0
>