Re: [PATCH net-next 2/3] net: dsa: Allow default tag protocol to be overridden from DT

[Tobias Waldekranz <tobias@waldekranz.com>]

On Fri, Mar 26, 2021 at 14:57, Vladimir Oltean <olteanv@gmail.com> wrote:
> Hi Tobias,
>
> On Fri, Mar 26, 2021 at 11:56:47AM +0100, Tobias Waldekranz wrote:
>>  	} else {
>> -		dst->tag_ops = dsa_tag_driver_get(tag_protocol);
>> -		if (IS_ERR(dst->tag_ops)) {
>> -			if (PTR_ERR(dst->tag_ops) == -ENOPROTOOPT)
>> -				return -EPROBE_DEFER;
>> -			dev_warn(ds->dev, "No tagger for this switch\n");
>> -			dp->master = NULL;
>> -			return PTR_ERR(dst->tag_ops);
>> -		}
>> +		dst->tag_ops = tag_ops;
>>  	}
>
> This will conflict with George's bug fix for 'net', am I right?
> https://patchwork.kernel.org/project/netdevbpf/patch/20210322202650.45776-1-george.mccollister@gmail.com/

Yes; this version also fixes George's problem I think, as we do not
assign dst->tag_ops until we know it is good, but it will not merge
cleanly.

> Would you mind resending after David merges 'net' into 'net-next'?

Sure thing. Should I then call that v2 or a resend of v1? The patches
will not be identical, so v2 I guess?

> This process usually looks like commit d489ded1a369 ("Merge
> git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net"). However,
> during this kernel development cycle, I have seen no merge of 'net' into
> 'net-next' since commit 05a59d79793d ("Merge
> git://git.kernel.org:/pub/scm/linux/kernel/git/netdev/net"), but that
> comes directly from Linus Torvalds' v5.12-rc2.
>
> Nonetheless, at some point (and sooner rather than later, I think),
> David or Jakub should merge the two trees. I would prefer to do it this
> way because the merge is going to be a bit messy otherwise, and I might
> want to cherry-pick these patches to some trees and it would be nice if
> the history was linear.
>
> Thanks!