Re: "rfkill: add a reason to the HW rfkill state" breaks userspace

[Benjamin Berg <bberg@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1248 bytes --]

Hi,

On Wed, 2021-04-14 at 12:29 +0200, Hans de Goede wrote:
> On 4/14/21 11:52 AM, Benjamin Berg wrote:
> > [SNIP]
> > 
> > That said, the g-s-d event handler is checking the size of the read
> > against the struct size. This is obviously wrong, it should only
> > check
> > that the read was successful (or check for >= V1 size).
> 
> Right this is what I was referring to a g-s-d compiled against the
> new headers with a struct size of 9 will read 8 bytes on an old
> kernel
> and that will fail the len check, so it won't work.
> 
> But this will actually never happen as I just noticed that g-s-d
> uses a private rfkill.h copy with the old 8 bytes struct definition.
> 
> So with the buffered-io disabled everything should work fine,
> see my other email in this thread.

OK, so at least that part of the fix should be backported to older
release branches.

> We should probably still fix / clean the code a bit though, as
> you are working on in:
>   
> https://gitlab.gnome.org/GNOME/gnome-settings-daemon/-/merge_requests/234

Yeah, the code there is somewhat messy overall. It could probably be
improved more than the MR does, but hopefully at least the biggest
inconsistencies are gone now.

Benjamin

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]