On Sat, Jun 07, 2025 at 03:47:09PM +0800, Jeremy Kerr wrote:
> Hi Patrick,
> 
> [+CC Andrew, for openbmc kernel reasons]
> 
> > So, it seems like there's something more subtle happening here - or I
> > have misunderstood something about the fix (I'm unsure of the
> > reference to xa_for_each_start; for_each_netdev_dump only calls xa_start?).
> 
> Ah! Are you on the openbmc 6.6 backport perhaps?
> 
> It look the xa_for_each_start()-implementation of netdev_for_each_dump()
> would not be compatible with a direct backport of 2d20773aec14 ("mctp: no
> longer rely on net->dev_index_head[]").
> 
> This was the update for the for_each_netdev_dump() macro:
> 
> commit f22b4b55edb507a2b30981e133b66b642be4d13f
> Author: Jakub Kicinski <kuba@kernel.org>
> Date:   Thu Jun 13 14:33:16 2024 -0700
> 
>     net: make for_each_netdev_dump() a little more bug-proof
>     
>     I find the behavior of xa_for_each_start() slightly counter-intuitive.
>     It doesn't end the iteration by making the index point after the last
>     element. IOW calling xa_for_each_start() again after it "finished"
>     will run the body of the loop for the last valid element, instead
>     of doing nothing.
> 
> ... which sounds like what's happening here.

Ah, yep.  That's exactly what is going on here.  I guess this change
isn't needed for master and it looks like you've already requested a
revert for 6.6?

-- 
Patrick Williams