On 4/22/19 7:41 AM, Ard Biesheuvel wrote: > On Sun, 21 Apr 2019 at 10:33, Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx> wrote: >> >> Hello all, >> >> I ran into an issue booting a graphical desktop (GNOME) on an ARM >> based system with nvidia graphics and the nouveau driver (which is the >> only one available on arm64). Bisect identified the following commit >> as the culprit >> >> commit 9bf7933fc3f306bc4ce74ad734f690a71670178a >> Author: Roman Penyaev <rpenyaev@xxxxxxx> >> Date: Mon Mar 25 20:09:24 2019 +0100 >> >> io_uring: offload write to async worker in case of -EAGAIN >> >> and reverting that on top of v5.1-rc5 gives me a working system again. >> >> I don't see any errors in the kernel log, but the graphical desktop >> never comes up. I can still ssh into the system and everything else >> appears to work normally. >> >> I'm a bit out of my depth there, so could you please help me diagnose >> this further? Any conditions I should check in the 'else' path? >> > > Actually, I am struggling to understand why this particular commit > would make a difference on a Debian Stretch based userland, since I > don't suppose the new syscalls are being used there yet ... Yeah, it really makes no sense at all, as that code path is not being exercised on your system at all. -- Jens Axboe
