Re: panic after DRM & nouveau MSI changes

2020-02-13 Thread John D. Baker
On Thu, 13 Feb 2020, Jarom??r Dole?~Mek wrote:

> Le jeu. 13 f??vr. 2020 ?? 17:28, John D. Baker
>  a ??crit :
> > Instead of a blank screen, the display simply freezes in VGA text mode
> > when nouveaufb0 attaches.  See my latest addendum to kern/52440.
> 
> I see the problem happens on G84 and G92, but not G64.
> 
> Are all the cards with MSI problem by chance the Tesla family, i.e NV50?
> 
> Those are: GeForce 8, GeForce 9, GeForce 100, GeForce 200, GeForce 300

So far, the G92 is the only affected card which I can test easily.  I
have some others, but I'd need to arrange space to set up a system in
which to install them for testing.

It's not clear if it's the card, the host, or the combination of the
two that trigger the problem.  The machine with the G94 has no problem
with MSI.  I might be able to arrange swapping the G92 and G94 between
hosts to see if the problem moves with the card.

I also have a couple of AGP and at least one plain PCI NVidia graphics
card which at one time were supported by the nouveau kernel driver.
I haven't had those machines operational for some time, but perhaps
I can see how they behave these days as well.

(At present, the Toshiba Qosmio F15R is the only non-PCIe NVidia graphics
system that's readily accessible and I've posted its results to kern/52440
as well.  I haven't rebuilt i386-current since the changes as that build
host is currently engaged in other tasks.)

As it stands, the panic that was the prompt for this thread has been
fixed, and the issue is back to kern/52440.

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645

Re: panic after DRM & nouveau MSI changes

2020-02-13 Thread Jaromír Doleček
Le jeu. 13 févr. 2020 à 17:28, John D. Baker
 a écrit :
>
> On Wed, 12 Feb 2020, Jaromír DoleÄ~Mek wrote:
>
> > I've just committed a fix for the MSI interrupt allocation for
> > nouveau, can you try it on the system which had the trouble with blank
> > console?
>
> Instead of a blank screen, the display simply freezes in VGA text mode
> when nouveaufb0 attaches.  See my latest addendum to kern/52440.

I see the problem happens on G84 and G92, but not G64.

Are all the cards with MSI problem by chance the Tesla family, i.e NV50?

Those are: GeForce 8, GeForce 9, GeForce 100, GeForce 200, GeForce 300

Jaromir


Re: panic after DRM & nouveau MSI changes

2020-02-13 Thread John D. Baker
On Wed, 12 Feb 2020, Jarom??r Dole?~Mek wrote:

> I've just committed a fix for the MSI interrupt allocation for
> nouveau, can you try it on the system which had the trouble with blank
> console?

Instead of a blank screen, the display simply freezes in VGA text mode
when nouveaufb0 attaches.  See my latest addendum to kern/52440.

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645

Re: panic after DRM & nouveau MSI changes

2020-02-12 Thread John D. Baker
On Wed, 12 Feb 2020, Jarom??r Dole?~Mek wrote:

> I've just committed a fix for the MSI interrupt allocation for
> nouveau, can you try it on the system which had the trouble with blank
> console?
> 
> It needs rev. 1.7 of
> sys/external/bsd/drm2/dist/drm/nouveau/nvkm/subdev/pci/nouveau_nvkm_subdev_pci_base.c

I updated my sources as soon as I saw these commits.  A release build
is finishing as I write and I will update and test soon.

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645

Re: panic after DRM & nouveau MSI changes

2020-02-12 Thread Jaromír Doleček
Le mer. 12 févr. 2020 à 03:38, John D. Baker
 a écrit :
> I've checked this with plain GENERIC and one with the above file rolled
> back.  With plain generic, nouveau, intel, and radeon all work and have
> MSI off.  With the modified GENERIC, nouveau systems work (MSI off) and
> intel and radeon systems work (MSI on).

I've just committed a fix for the MSI interrupt allocation for
nouveau, can you try it on the system which had the trouble with blank
console?

It needs rev. 1.7 of
sys/external/bsd/drm2/dist/drm/nouveau/nvkm/subdev/pci/nouveau_nvkm_subdev_pci_base.c

Jaromir


Re: panic after DRM & nouveau MSI changes

2020-02-11 Thread John D. Baker
On Tue, 11 Feb 2020, Jarom??r Dole?~Mek wrote:

> If it's something simple let's fix it and enable MSI unconditionally,
> if it would be something difficult then it would be reasonable to add
> some conditionals to enable MSI only for non-nouveau.

At present, simply rolling back:

  sys/external/bsd/drm2/linux/linux_pci.c

to the revision where MSI is enabled (r1.8 in -current) allows intel-
and radeon-equipped machines to use MSI, while nouveau has it explicitly
disabled in the HEAD revision of:

  
sys/external/bsd/drm2/dist/drm/nouveau/nvkm/subdev/pci/nouveau_nvkm_subdev_pci_base.c

so nouveau machines use INTx.

I've checked this with plain GENERIC and one with the above file rolled
back.  With plain generic, nouveau, intel, and radeon all work and have
MSI off.  With the modified GENERIC, nouveau systems work (MSI off) and
intel and radeon systems work (MSI on).

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645

Re: panic after DRM & nouveau MSI changes

2020-02-11 Thread John D. Baker
On Tue, 11 Feb 2020, Jarom??r Dole?~Mek wrote:

> I've now got a nVidia card which triggers the problem on my rig x86
> system, will look into the nouveau-specific panic over next week or
> so.

I presume the problem you refer to is the panic after recent MSI
re-enabling?

What about the original problem in kern/52440 or the different
panic after using Jared's patch with the re-enabled MSI?

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645

Re: panic after DRM & nouveau MSI changes

2020-02-11 Thread John D. Baker
On Tue, 11 Feb 2020, Patrick Welche wrote:

> Just BTW, I remember a nouveau system that was unhappy with EFI but
> happy with legacy BIOS - different initialisation? worth a try?

One machine at my disposal predates the existence of EFI.  It is the
only one that had no problem when MSI was first enabled in 7.99.64.

The others have EFI, but I'm not sure what's being used when calling
for the boot menu to boot from USB.  The USB stick is a plain MBR
setup.  (The machines otherwise boot Linux Mint from hard disk.)

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645


Re: panic after DRM & nouveau MSI changes

2020-02-11 Thread Patrick Welche
On Fri, Feb 07, 2020 at 04:22:19PM -0600, John D. Baker wrote:
> I see that the MSI changes have been reverted.  I have not yet updated
> my sources and still have the MSI-enabled sources plus proposed nouveau
> patch applied in my kernels.
> 
> FWIW, quick tests with known-to-work radeon and intel graphics devices
> show that they are quite happy with MSI.  It's only NVIDIA/nouveau
> devices that are having trouble.
> 
> Is it possible then to enable MSI for DRM devices generally, but force
> nouveau to use INTx until the matter can be sorted?

Just BTW, I remember a nouveau system that was unhappy with EFI but
happy with legacy BIOS - different initialisation? worth a try?

Cheers,

Patrick


Re: panic after DRM & nouveau MSI changes

2020-02-11 Thread Jaromír Doleček
Le ven. 7 févr. 2020 à 23:22, John D. Baker  a écrit :
> I see that the MSI changes have been reverted.  I have not yet updated
> my sources and still have the MSI-enabled sources plus proposed nouveau
> patch applied in my kernels.
>
> FWIW, quick tests with known-to-work radeon and intel graphics devices
> show that they are quite happy with MSI.  It's only NVIDIA/nouveau
> devices that are having trouble.

I've now got a nVidia card which triggers the problem on my rig x86
system, will look into the nouveau-specific panic over next week or
so.

If it's something simple let's fix it and enable MSI unconditionally,
if it would be something difficult then it would be reasonable to add
some conditionals to enable MSI only for non-nouveau.

Jaromir


Re: panic after DRM & nouveau MSI changes

2020-02-07 Thread John D. Baker
I see that the MSI changes have been reverted.  I have not yet updated
my sources and still have the MSI-enabled sources plus proposed nouveau
patch applied in my kernels.

FWIW, quick tests with known-to-work radeon and intel graphics devices
show that they are quite happy with MSI.  It's only NVIDIA/nouveau
devices that are having trouble.

Is it possible then to enable MSI for DRM devices generally, but force
nouveau to use INTx until the matter can be sorted?

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645


Re: panic after DRM & nouveau MSI changes

2020-02-06 Thread Jared McNeill
I'll try to have a look at it over the next few days. Unfortunately this 
is an x86 specific problem and I don't have one to test so it could take 
some time to work out.


On Wed, 5 Feb 2020, John D. Baker wrote:


Recently MSI was re-enabled for DRM devices generally and nouveau devices
specifically.  The resulting kernel panics attaching the framebuffer
console on nouveau-equipped machines.

See the two most recent addenda to PR kern/52440 for details.

Thanks.

--
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645




panic after DRM & nouveau MSI changes

2020-02-05 Thread John D. Baker
Recently MSI was re-enabled for DRM devices generally and nouveau devices
specifically.  The resulting kernel panics attaching the framebuffer
console on nouveau-equipped machines.

See the two most recent addenda to PR kern/52440 for details.

Thanks.

-- 
|/"\ John D. Baker, KN5UKS   NetBSD Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSDFreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645