savage inproper init

[Alex Deucher]

On Dec 11, 2007 2:09 PM, Ralovich Kristóf <kristof.ralovich at gmail.com> wrote:
> On Dec 11, 2007 6:04 PM, Alex Deucher <alexdeucher at gmail.com> wrote:
> > On Dec 11, 2007 11:49 AM, Ralovich Kristóf <kristof.ralovich at gmail.com> wrote:
> > > Hi,
> > >
> > > I would like to report about hard lockups (ping and ssh stops too)
> > > caused by the savage driver. I am getting the same results no matter
> > > if I use the savage fbdev or not. Neither EXA / XAA changes the
> > > situation. Additionally if I use EXA, it garbles the bitmaps and the X
> > > screen around the cursor. The machine is an IBM T22. I can not find
> > > any further details in the system logs or X log.
> > >
> > > The symptom is if boot up, X wont run, it hangs before switching to
> > > the desired resolution. But if I run windows first then reboot and
> > > start linux, everything is going fine. GL and Xv works too.
> > >
> > > The other way avoiding the lockup2 was putting the AGPMode "2" line to
> > > the Device section, it was a little bit slower, but it was stable at
> > > least.
> >
> > Savage IX/MX cards are only AGP 2x cards.  Some savage agp 2x cards do
> > not like 1x mode (the driver default).  I never had a chance to figure
> > out which cards this affected as it was hit or miss.  We should
> > probably just default to whatever AGP mode the bios sets up.
> >
> > Alex
> >
>
> Alex,
>
> first of all, thanks for the fast reply. You were right, even win is
> using the card in agp 2x mode, so from now on I am using that too.
> With agp 2x the cold startup problem is vanished again. But now I have
> some messages in /var/log/kern.log. After X startup:
>
> Dec 11 18:26:55 t22 kernel: mtrr: base(0xf2000000) is not aligned on a
> size(0x5000000) boundary
> Dec 11 18:26:55 t22 last message repeated 2 times
> Dec 11 18:26:55 t22 kernel: agpgart: Found an AGP 1.0 compliant device
> at 0000:00:00.0.
> Dec 11 18:26:55 t22 kernel: agpgart: Putting AGP V2 device at
> 0000:00:00.0 into 2x mode
> Dec 11 18:26:55 t22 kernel: agpgart: Putting AGP V2 device at
> 0000:01:00.0 into 2x mode
>
> X is working properly, so is Xv! But if I start glxgears, there is a
> reproducible lockup again:
>
> Dec 11 18:28:01 t22 kernel: [drm:savage_bci_wait_event_shadow] *ERROR* failed!
>
> OR
>
> Dec 11 19:27:15 t22 kernel: [drm:savage_bci_wait_event_shadow] *ERROR* failed!
> Dec 11 19:27:15 t22 kernel: [drm]    status=0x00009631, e=0x9633
> Dec 11 19:27:28 t22 kernel: [drm:savage_bci_wait_fifo_shadow] *ERROR* failed!
> Dec 11 19:27:28 t22 kernel: [drm]    status=0x003c7620, threshold=0x00007620
>
> with the above lines in kern.log. I am still able to restart the
> machine using keys, so the kernel keeps running.
>
> Putting ShadowStatus "off" into xorg.conf wont solve the problem, but
> the kernel message is different:
>
> Dec 11 19:36:53 t22 kernel: [drm:savage_bci_wait_event_reg] *ERROR* failed!
> Dec 11 19:36:53 t22 kernel: [drm]    status=0x00000002, e=0x0004
>
> and causing a hard lockup again!

IIRC, you need one or more of the following options:
Option "DmaMode" "None"
Option "BusType" "PCI"

see the savage man page for more info.

>
> With BCIforXv "off" the result is always soft lockup:

BCIforXv has no affect on old savage cards like the MX/IX.  It only
affects savage4 and IGP cards.

Alex