[U-Boot] [RFC PATCH 00/50] net: net subsystem ops cleanup

Joe Hershberger joe.hershberger at gmail.com
Wed Nov 14 01:06:15 CET 2012


Hi Tomas,

On Sat, Nov 3, 2012 at 6:23 PM, Tomas Hlavacek <tmshlvck at gmail.com> wrote:
> Dear Wolfgang,
>
> On Sat, Nov 3, 2012 at 4:09 PM, Wolfgang Denk <wd at denx.de> wrote:
>> Dear Tomas Hlavacek,
>>
>> In message <1351876722-5183-1-git-send-email-tmshlvck at gmail.com> you wrote:
>>> This patchset is a first stage of preparation of the net subsystem
>>> for the driver model.
>>>
>>> The idea of this patchset is:
>>> 1) Remove ops .init, .send, .recv and .halt from the eth_device struct.
>>>    Add a sparate structure eth_ops which is ready for inclusion
>>>    to DM core.
>>> 2) Replace dynamic init of ops function pointers by static struct.
>>> 3) Do minor style cleanup.
>>>
>>> Tomas Hlavacek (50):
>>>   net: dm: Pull out ops from struct eth_device
>>>   net: 4xx_enet: Pull out init of struct eth_ops
>>>   net: altera_tse: Pull out init of struct eth_ops
>>>   net: dm9000x: Pull out init of struct eth_ops
>>>   net: armada100_fec: Pull out init of struct eth_ops
>>
>> Hm... looking at this patch series, I wonder if it is really
>> bisectable?  Can I really apply any number of these patches (the first
>> N, with N < 50) and expect the code to build and to work?
>
> It should be, because the first patch adds new struct eth_ops and
> changes all accesses to its' members in one step. Patches 2 .. 50
> remove dynamic ops settings and add static initialization to each
> affected driver - one patch per driver. I would rather try that by
> compiling U-Boot with only 1/50 applied and after some random N, say
> 30/50 to be absolutely sure. Let me get back later when I have my
> MAKEALL results.

Have you completed this bisectability test yet?  How about run
testing?  What boards did you test this on?

Thanks,
-Joe


More information about the U-Boot mailing list