On Thu, 28 Nov 2024 at 15:46, Johan Hovold <johan@xxxxxxxxxx> wrote: > > On Thu, Nov 28, 2024 at 12:05:09PM +0100, Ard Biesheuvel wrote: > > > If you're happy to experiment more, you could try and register a > > notification for EFI_EVENT_GROUP_BEFORE_EXIT_BOOT_SERVICES using > > CreateEventEx(), and see if it gets called when ExitBootServices() is > > called. That would at least help narrow it down. > > Thanks for the suggestion. > > I see the notify function being called when I signal it as well as on > each ExitBootServices(). > Interesting. That means the EDK2 fork is fairly recent. FYI https://github.com/tianocore/edk2/pull/6481 > With an efi_printk() in the callback ExitBootServices() fails as > expected, but with an empty function the kernel seems to start every > time. > > Interestingly, ExitBootServices() now succeeds also if I add back the > CloseEvent() call. In fact, it works also if I never signal the event > (i.e. if I just create and close the event). > Is it still invoked by the firmware if you closed the event before EBS()? > The patch below should suffice as a workaround I can carry until the > firmware has been fixed. > Ok. I'd prefer to get this fixed on the firmware side as well.
