On Fri, Sep 4, 2020 at 3:17 PM Till Maas <till(a)redhat.com> wrote:
Hi,
Am Fr., 4. Sept. 2020 um 15:47 Uhr schrieb Jarod Wilson <jarod(a)redhat.com>:
...
Just wondering, would it be possible to use port on the public part,
which might be just the output of /proc/net/bonding/$interface - so
Primary Port:
Currently Active Port:
Port Interface:
Port queue ID:
So this is one of those fun areas... No. The output of
/proc/net/bonding/<bond> is part of the user-facing api, essentially,
so the wording there will NOT be changing from saying Slave, per the
Linux kernel's "do not break userspace" imperative. People have
scripts that parse it, so a change there is forbidden. (This is
explicitly covered in the upstream netdev discussion).
> but hadn't looked into until now: master -> aggregate and
slave ->
> link, since bonding is explicitly described as being a link
> aggregation driver. There's use of 'link' in the bonding driver
> already, but it looks fairly simple to adjust (it's used for link
> state tracking, up/down/etc)...
sounds like good terms, too. Just wondering, who needs to be convinced
on the kernel side to move this forward?
It's currently stalled on me providing a sane way for -stable kernel
maintainers to do backports, and posting a viable patchset to do the
rename, that people agree to and we get merged. I should be able to
revisit this the upcoming week and hopefully send out an initial
patchset. Locally, I've already done some work to replace the current
use of "link" with "link_state", so that I can try out an
aggregator/link rename.
--
Jarod Wilson
jarod(a)redhat.com