On Thu, 2021-06-10 at 14:45 +0200, Bastien Nocera wrote: > On Mon, 2021-05-03 at 15:12 +0200, Benjamin Berg wrote: > > From: Benjamin Berg <bberg@xxxxxxxxxx> > > > > The kernel will always send exactly one event, but the size of the > > passed struct will depend on the length of the submitted read() and > > the > > kernel version. i.e. the interface can be extended and we need to > > expect > > for a read to be longer than expected if we ask for it. > > > > Fix this by only requesting the needed length and explicitly check > > the > > length against the V1 version of the structure to make the code a bit > > more future proof in case the internal copy of the struct is updated > > to > > contain new fields. > > This fixes a bug in GNOME where to enable Bluetooth, we removed a soft > rfkill block on the Bluetooth interface. > > Without this, the bluetooth rfkill gets unblocked, but bluetoothd > doesn't see it as unblocked so never powers it on, causing the UI to > appear broken, as we expect Bluetooth devices to be either blocked > through rfkill, or powered on. > > The equivalent gnome-settings-daemon fix (which deals with rfkill) was > reviewed by Hans de Goede: > https://gitlab.gnome.org/GNOME/gnome-settings-daemon/-/merge_requests/234 > > Benjamin, it might be worth resending this with a better commit message > explaining exactly what it fixes and referencing the gnome-bluetooth > bug: > https://gitlab.gnome.org/GNOME/gnome-bluetooth/-/issues/38 > It's also been pushed to Fedora rawhide and Fedora 34: https://bodhi.fedoraproject.org/updates/FEDORA-2021-2cd83da751
