Re: 7.2.y took the on-air address fix without its companion

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Sat, Oct 31, 2026 at 01:16:21PM +0100, Radek Podgorny wrote:
> On Fri, Oct 30, 2026 at 11:32:40PM +0100, Greg Kroah-Hartman wrote:
> > I have no context here, sorry, what is wrong?  If that's a bugfix, why
> > is it not cc: stable?
> 
> Sorry, my last mail was too short.
> 
> It fixes a regression in stable. I sent two patches as a series:
> 
>   3889a5fcd091 Bluetooth: forget a peer's RPA once it advertises its
>                identity address
>   4914c4998961 Bluetooth: put the peer's on-air address on air when we
>                cannot resolve
> 
> The second depends on the first, but AUTOSEL took only the second into
> 7.2.8 (as a552b4bc058b). It also reached 7.3 ahead of the first.
> 
> So when a bonded peer turns privacy off and advertises its identity
> address, the host keeps connecting to the old RPA, which the peer no
> longer answers. That lasts until the adapter is powered off. 7.2.6
> connected to the identity address and worked.
> 
> There is no cc: stable because I left the tags off on purpose. When I
> posted the first patch, the kernel still hid the stale address, so the
> patch fixed nothing live on its own. Taking the second patch alone
> exposed the bug. In effect the first patch is:
> 
>   Fixes: 4914c4998961 ("Bluetooth: put the peer's on-air address on air when we cannot resolve")
> 
> Please queue 3889a5fcd091 for 7.3.y and 7.2.y. It cherry-picks cleanly
> onto 7.3.1 and 7.2.14. 6.18 and older don't have 4914c4998961 and don't
> need it. I have run the two patches together since September.

Thanks, that helps, now queued up.

greg k-h