On Fri, Jun 23, 2023 at 2:24 PM Christian Brauner <brauner@xxxxxxxxxx> wrote: > > On Fri, Jun 23, 2023 at 11:23:57AM +0200, Alexander Mikhalitsyn wrote: > > On Fri, Jun 23, 2023 at 10:34 AM Alexander Mikhalitsyn > > <alexander@xxxxxxxxxxxxx> wrote: > > > > > > On Fri, Jun 23, 2023 at 12:28 AM Von Dentz, Luiz > > > <luiz.von.dentz@xxxxxxxxx> wrote: > > > > > > > > Hi Stephen, Alexander, > > > > > > Dear friends, > > > > > > > > > > > Looks like we have a similar problem as to: > > > > > > > > https://lore.kernel.org/lkml/20230524081933.44dc8bea@xxxxxxxxxx/ > > > > > > > > That said for unix socket it was decided to not allow it to be build as module, which is something I don't think we are willing to do for bluetooth, so we have to find a way to get around pidfd_prepare, which seems to be called due to the use of scm_recv (it is also used by netlink btw). > > > > > > Ugh, yep. That's bad and we can't workaround it like we did for unix sockets. > > > > > > Originally, Christian had objections against exporting the > > > pidfd_prepare function [1] > > > > > > [1] https://lore.kernel.org/all/20230523-flechten-ortsschild-e5724ecc4ed0@brauner/ > > > > > > +cc Christian > > > > Idea. (-: We can introduce a new function like "scm_recv_light" that > > Seems fine to me. Thanks, Christian! Let's wait for Jakub's comment then :) Then I'll prepare and send a patch. Kind regards, Alex
