Re: f17 and ATI RAGE128

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, 2012-05-29 at 10:00 -0400, Adam Jackson wrote:
> On Sat, 2012-05-26 at 20:22 -0700, John Reiser wrote:
> > > Peter, how in general do you want to proceed where we have documented
> > > cases of graphics mode setting failure in grub2? Do you want individual
> > > bug reports for each affected bit of hardware? Downstream, upstream?
> > > What kind of info should be included? Thanks!
> > 
> > What's not good enough about using [the equivalent of] vesafb,
> > so as to to avoid problems with mode-setting?  Or, use vesafb
> > unless the PCI vendor:device:version is on a known-good list?
> 
> Only all the things wrong with VBE, which I don't especially feel like
> repeating yet again.
> 
> Graphical bootloader is not my favorite idea.

I did propose disabling it again before release, but a bit too late; we
didn't really have enough solid data to justify the change. I'm hoping
it won't turn out _too_ terribly.

I did spend some time looking into what Ubuntu does (since they've been
using grub2 for some time and have a wide base of consumer hardware
among their users, I was thinking this may give an indication as to what
their experience with graphical grub has taught them). They default to
graphical grub, but with quite a few caveats:

1) Unless you're multi-booting, they set the default timeout to 0, so
you never see grub by default (you have to edit the timeout or hold down
a key on boot).
2) They have an explicit blacklist of hardware that's forced to text
mode - the list is in the 'grub-gfxpayload-lists' package.
3) They have a patch to grub that explicitly blacklists a particular
mode that is known to be problematic on a specific system:
ubuntu_blacklist_1440x900x32.patch , the result of
https://bugs.launchpad.net/ubuntu/+source/grub2/+bug/701111 . This is
obviously an insane road to go down.
4) They have a somewhat large patch to grub which implements a 'failsafe
mode', rather similar to the one pjones recently posted for UEFI mode to
devel@ under the title "[PATCH] Add check_completed_boot command on EFI
systems." Briefly, after a successful boot, a setting is flipped
somewhere which grub can check. At each boot, grub checks it. If it's
not flipped - indicating the last attempt to boot the system was
unsuccessful - grub starts up in the 'failsafe' mode, which among other
things, uses text mode, not graphical. This patch has not gone upstream,
and Ubuntu are still using grub2 1.99 - not any of the 2.0 betas
(probably precisely because of this and several other large,
non-upstreamed patches) - so we can't easily adopt the patch for Fedora
or get it upstreamed. The patch that creates this 'failsafe' mode is
'ubuntu_failed_boot_menu.patch' in the grub2 package; it should be
considered to also include the new files grub-common.init and
grub-common.pm-sleep that are created in the Ubuntu package (seriously,
Debian/Ubuntu, grow a sane fucking patch system already). The use of
text mode during failsafe is actually part of the patch
ubuntu_gfxpayload_filter.patch .

The good news from the Ubuntu investigation I did is that their
blacklisting taken in full doesn't actually cover a huge amount of
hardware. The ubuntu_blacklist_1440x900x32.patch covers one system
(albeit a fairly popular one) - the Thinkpad T400. The
grub-gfxpayload-lists blacklist appears to cover only a single Radeon
adapter and all VMware VMs. Obviously, I would be more worried if Ubuntu
had a huge blacklist, as that would indicate there was lots of known-bad
hardware out there we were about to run into trouble with.

On the principle of things, I agree broadly with ajax that what's going
on here seems suboptimal, with two different projects trying to
duplicate the same work (modesetting, and handling hardware that isn't
good at modesetting). I believe there are some legitimate reasons why it
may make sense to do it in the bootloader, though. Maybe kernel/X and
grub need to sit down and thrash out who's going to do the modesetting?
If that's not happened already? This obviously has implications for the
whole 'smooth boot' idea.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net

-- 
test mailing list
test@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe:
https://admin.fedoraproject.org/mailman/listinfo/test



[Index of Archives]     [Fedora Desktop]     [Fedora SELinux]     [Photo Sharing]     [Yosemite Forum]     [KDE Users]

  Powered by Linux