[PATCH] intel-agp: Avoid oops for G33 on 1MB stolen case

[PATCH] intel-agp: Avoid oops for G33 on 1MB stolen case

[Brandon Philips]

This is similar to f443675affe3f16dd428e46f0f7fd3f4d703eeab which was
reverted because it broke older X.org driver. This patch only fixes
the 1MB stolen case since it causes an oops.

Xorg will not work without the accompanying patch[1] but avoiding an
oops and making it possible to work with patched xorg driver is
reasonable.

[1] http://ifup.org/~philips/review/xf86-video-intel-G33-1mb.patch

Explanation of the oops:

> static void intel_i830_init_gtt_entries(void)
...
>         } else if (IS_G33) {
>         /* G33's GTT size defined in gmch_ctrl */
>                 switch (gmch_ctrl & G33_PGETBL_SIZE_MASK) {
>                 case G33_PGETBL_SIZE_1M:
>                         size = 1024;
>                         break;
...
>                 size += 4;

size = 1028

Then since we have the BIOS setting 1MB for the device in the GMCH
control we get to here:

>         } else {
>                 switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
>                 case I855_GMCH_GMS_STOLEN_1M:
>                         gtt_entries = MB(1) - KB(size);
>                         break;

MB(1) = 1 * 1024 * 1024
KB(1028) = 1028 * 1024

MB(1) - KB(1028) = -4096

>         gtt_entries /= KB(4);
>         intel_private.gtt_entries = gtt_entries;

We end up with -1 in gtt_entries.

This leads to intel_i915_configure reading/writing to areas outside of
mapped memory and the oops.

Bugzilla: https://bugzilla.novell.com/show_bug.cgi?id=391261

Signed-off-by: Brandon Philips <bphilips@suse.de>

---
 drivers/char/agp/intel-agp.c |    8 ++++++++
 1 file changed, 8 insertions(+)

Index: linux-2.6/drivers/char/agp/intel-agp.c
===================================================================
--- linux-2.6.orig/drivers/char/agp/intel-agp.c
+++ linux-2.6/drivers/char/agp/intel-agp.c
@@ -559,6 +559,14 @@ static void intel_i830_init_gtt_entries(
 	} else {
 		switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
 		case I855_GMCH_GMS_STOLEN_1M:
+			if (IS_G33) {
+				size = 0;
+				printk(KERN_WARNING PFX
+				       "Warning: G33 chipset with 1MB"
+					" allocated. Older X.org Intel drivers"
+					" will not work.\n");
+				WARN_ON(1);
+			}
 			gtt_entries = MB(1) - KB(size);
 			break;
 		case I855_GMCH_GMS_STOLEN_4M:

Re: [PATCH] intel-agp: Avoid oops for G33 on 1MB stolen case

[Andrew Morton]

On Tue, 14 Oct 2008 15:58:17 -0700
Brandon Philips <bphilips@suse.de> wrote:

> This is similar to f443675affe3f16dd428e46f0f7fd3f4d703eeab which was
> reverted because it broke older X.org driver. This patch only fixes
> the 1MB stolen case since it causes an oops.
> 
> Xorg will not work without the accompanying patch[1] but avoiding an
> oops and making it possible to work with patched xorg driver is
> reasonable.
> 
> [1] http://ifup.org/~philips/review/xf86-video-intel-G33-1mb.patch
> 
> Explanation of the oops:
> 
> > static void intel_i830_init_gtt_entries(void)
> ...
> >         } else if (IS_G33) {
> >         /* G33's GTT size defined in gmch_ctrl */
> >                 switch (gmch_ctrl & G33_PGETBL_SIZE_MASK) {
> >                 case G33_PGETBL_SIZE_1M:
> >                         size = 1024;
> >                         break;
> ...
> >                 size += 4;
> 
> size = 1028
> 
> Then since we have the BIOS setting 1MB for the device in the GMCH
> control we get to here:
> 
> >         } else {
> >                 switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
> >                 case I855_GMCH_GMS_STOLEN_1M:
> >                         gtt_entries = MB(1) - KB(size);
> >                         break;
> 
> MB(1) = 1 * 1024 * 1024
> KB(1028) = 1028 * 1024
> 
> MB(1) - KB(1028) = -4096
> 
> >         gtt_entries /= KB(4);
> >         intel_private.gtt_entries = gtt_entries;
> 
> We end up with -1 in gtt_entries.
> 
> This leads to intel_i915_configure reading/writing to areas outside of
> mapped memory and the oops.
> 
> Bugzilla: https://bugzilla.novell.com/show_bug.cgi?id=391261
> 
> Signed-off-by: Brandon Philips <bphilips@suse.de>
> 
> ---
>  drivers/char/agp/intel-agp.c |    8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> Index: linux-2.6/drivers/char/agp/intel-agp.c
> ===================================================================
> --- linux-2.6.orig/drivers/char/agp/intel-agp.c
> +++ linux-2.6/drivers/char/agp/intel-agp.c
> @@ -559,6 +559,14 @@ static void intel_i830_init_gtt_entries(
>  	} else {
>  		switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
>  		case I855_GMCH_GMS_STOLEN_1M:
> +			if (IS_G33) {
> +				size = 0;
> +				printk(KERN_WARNING PFX
> +				       "Warning: G33 chipset with 1MB"
> +					" allocated. Older X.org Intel drivers"
> +					" will not work.\n");
> +				WARN_ON(1);
> +			}
>  			gtt_entries = MB(1) - KB(size);
>  			break;
>  		case I855_GMCH_GMS_STOLEN_4M:

Is the bug which this patch addresses present in the 2.6.27 kernel?

Thanks.

Re: [PATCH] intel-agp: Avoid oops for G33 on 1MB stolen case

[Dave Airlie]

On Mon, 2008-10-20 at 14:21 -0700, Andrew Morton wrote:
> On Tue, 14 Oct 2008 15:58:17 -0700
> Brandon Philips <bphilips@suse.de> wrote:
> 
> > This is similar to f443675affe3f16dd428e46f0f7fd3f4d703eeab which was
> > reverted because it broke older X.org driver. This patch only fixes
> > the 1MB stolen case since it causes an oops.
> > 
> > Xorg will not work without the accompanying patch[1] but avoiding an
> > oops and making it possible to work with patched xorg driver is
> > reasonable.
> > 
> > [1] http://ifup.org/~philips/review/xf86-video-intel-G33-1mb.patch
> > 
> > Explanation of the oops:
> > 
> > > static void intel_i830_init_gtt_entries(void)
> > ...
> > >         } else if (IS_G33) {
> > >         /* G33's GTT size defined in gmch_ctrl */
> > >                 switch (gmch_ctrl & G33_PGETBL_SIZE_MASK) {
> > >                 case G33_PGETBL_SIZE_1M:
> > >                         size = 1024;
> > >                         break;
> > ...
> > >                 size += 4;
> > 
> > size = 1028
> > 
> > Then since we have the BIOS setting 1MB for the device in the GMCH
> > control we get to here:
> > 
> > >         } else {
> > >                 switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
> > >                 case I855_GMCH_GMS_STOLEN_1M:
> > >                         gtt_entries = MB(1) - KB(size);
> > >                         break;
> > 
> > MB(1) = 1 * 1024 * 1024
> > KB(1028) = 1028 * 1024
> > 
> > MB(1) - KB(1028) = -4096
> > 
> > >         gtt_entries /= KB(4);
> > >         intel_private.gtt_entries = gtt_entries;
> > 
> > We end up with -1 in gtt_entries.
> > 
> > This leads to intel_i915_configure reading/writing to areas outside of
> > mapped memory and the oops.
> > 
> > Bugzilla: https://bugzilla.novell.com/show_bug.cgi?id=391261
> > 
> > Signed-off-by: Brandon Philips <bphilips@suse.de>
> > 
> > ---
> >  drivers/char/agp/intel-agp.c |    8 ++++++++
> >  1 file changed, 8 insertions(+)
> > 
> > Index: linux-2.6/drivers/char/agp/intel-agp.c
> > ===================================================================
> > --- linux-2.6.orig/drivers/char/agp/intel-agp.c
> > +++ linux-2.6/drivers/char/agp/intel-agp.c
> > @@ -559,6 +559,14 @@ static void intel_i830_init_gtt_entries(
> >  	} else {
> >  		switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
> >  		case I855_GMCH_GMS_STOLEN_1M:
> > +			if (IS_G33) {
> > +				size = 0;
> > +				printk(KERN_WARNING PFX
> > +				       "Warning: G33 chipset with 1MB"
> > +					" allocated. Older X.org Intel drivers"
> > +					" will not work.\n");
> > +				WARN_ON(1);
> > +			}
> >  			gtt_entries = MB(1) - KB(size);
> >  			break;
> >  		case I855_GMCH_GMS_STOLEN_4M:
> 
> Is the bug which this patch addresses present in the 2.6.27 kernel?
> 

I've been a bit wary about this patch, but on re-review I suppose it
should be fine. We'll just get WARN_ONs in places we don't really want
them, and I'm sure Arjan will come complaining about them from
kerneloops.

However as I said at KS there is no nice answer to this, we are
hopelessly screwed.

Dave.

Re: [PATCH] intel-agp: Avoid oops for G33 on 1MB stolen case

[Arjan van de Ven]

On Tue, 21 Oct 2008 07:44:46 +1000
Dave Airlie <airlied@redhat.com> wrote:
> > +				printk(KERN_WARNING PFX
> > > +				       "Warning: G33 chipset
> > > with 1MB"
> > > +					" allocated. Older X.org
> > > Intel drivers"
> > > +					" will not work.\n");
> > > +				WARN_ON(1);
> > > +			}
> > >  			gtt_entries = MB(1) - KB(size);
> > >  			break;
> > >  		case I855_GMCH_GMS_STOLEN_4M:
> > 
> > Is the bug which this patch addresses present in the 2.6.27 kernel?
> > 
> 
> I've been a bit wary about this patch, but on re-review I suppose it
> should be fine. We'll just get WARN_ONs in places we don't really want
> them, and I'm sure Arjan will come complaining about them from
> kerneloops.
> 

the patch is not so nice for this;
the printk+WARN_ON() really should be using WARN() instead; that also
allows me to filter these guys out easily if needed.

-- 
Arjan van de Ven 	Intel Open Source Technology Centre
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org

Re: [PATCH] intel-agp: Avoid oops for G33 on 1MB stolen case

[Brandon Philips]

On 14:52 Mon 20 Oct 2008, Arjan van de Ven wrote:
> On Tue, 21 Oct 2008 07:44:46 +1000
> Dave Airlie <airlied@redhat.com> wrote:
> > > +				printk(KERN_WARNING PFX
> > > > +				       "Warning: G33 chipset
> > > > with 1MB"
> > > > +					" allocated. Older X.org
> > > > Intel drivers"
> > > > +					" will not work.\n");
> > > > +				WARN_ON(1);
> > > > +			}
> > > >  			gtt_entries = MB(1) - KB(size);
> > > >  			break;
> > > >  		case I855_GMCH_GMS_STOLEN_4M:
> > > 
> > > Is the bug which this patch addresses present in the 2.6.27 kernel?
> > > 
> > 
> > I've been a bit wary about this patch, but on re-review I suppose it
> > should be fine. We'll just get WARN_ONs in places we don't really want
> > them, and I'm sure Arjan will come complaining about them from
> > kerneloops.

I figured it was better to be noisy than to have people complaining
about X.org being broken.

Should it be something other than a WARN?

> the patch is not so nice for this;
> the printk+WARN_ON() really should be using WARN() instead; that also
> allows me to filter these guys out easily if needed.

Ok. I will send a new patch in a moment using WARN().

Thanks,

	Brandon

[PATCH v2] intel-agp: Avoid oops for G33 on 1MB stolen case

[Brandon Philips]

This is similar to f443675affe3f16dd428e46f0f7fd3f4d703eeab which was
reverted because it broke older X.org driver. This patch only fixes
the 1MB stolen case since it causes an oops.

Xorg will not work without the accompanying patch[1] but avoiding an
oops and making it possible to work with patched xorg driver is
reasonable.

[1] http://ifup.org/~philips/review/xf86-video-intel-G33-1mb.patch

Explanation of the oops:

> static void intel_i830_init_gtt_entries(void)
...
>         } else if (IS_G33) {
>         /* G33's GTT size defined in gmch_ctrl */
>                 switch (gmch_ctrl & G33_PGETBL_SIZE_MASK) {
>                 case G33_PGETBL_SIZE_1M:
>                         size = 1024;
>                         break;
...
>                 size += 4;

size = 1028

Then since we have the BIOS setting 1MB for the device in the GMCH
control we get to here:

>         } else {
>                 switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
>                 case I855_GMCH_GMS_STOLEN_1M:
>                         gtt_entries = MB(1) - KB(size);
>                         break;

MB(1) = 1 * 1024 * 1024
KB(1028) = 1028 * 1024

MB(1) - KB(1028) = -4096

>         gtt_entries /= KB(4);
>         intel_private.gtt_entries = gtt_entries;

We end up with -1 in gtt_entries.

This leads to intel_i915_configure reading/writing to areas outside of
mapped memory and the oops.

Bugzilla: https://bugzilla.novell.com/show_bug.cgi?id=391261

Signed-off-by: Brandon Philips <bphilips@suse.de>

---
 drivers/char/agp/intel-agp.c |    7 +++++++
 1 file changed, 7 insertions(+)

Index: linux-2.6/drivers/char/agp/intel-agp.c
===================================================================
--- linux-2.6.orig/drivers/char/agp/intel-agp.c
+++ linux-2.6/drivers/char/agp/intel-agp.c
@@ -561,6 +561,13 @@ static void intel_i830_init_gtt_entries(
 	} else {
 		switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
 		case I855_GMCH_GMS_STOLEN_1M:
+			if (IS_G33) {
+				size = 0;
+				WARN(1, KERN_WARNING
+				       "Warning: G33 chip with 1MB allocated."
+					" Older X.org Intel drivers will not"
+					" work.\n");
+			}
 			gtt_entries = MB(1) - KB(size);
 			break;
 		case I855_GMCH_GMS_STOLEN_4M:

Re: [PATCH] intel-agp: Avoid oops for G33 on 1MB stolen case

[Brandon Philips]

On 14:21 Mon 20 Oct 2008, Andrew Morton wrote:
> On Tue, 14 Oct 2008 15:58:17 -0700
> Brandon Philips <bphilips@suse.de> wrote:
> 
> > This is similar to f443675affe3f16dd428e46f0f7fd3f4d703eeab which was
> > reverted because it broke older X.org driver. This patch only fixes
> > the 1MB stolen case since it causes an oops.
> > 
> > Xorg will not work without the accompanying patch[1] but avoiding an
> > oops and making it possible to work with patched xorg driver is
> > reasonable.
> > 
> > [1] http://ifup.org/~philips/review/xf86-video-intel-G33-1mb.patch
> > 
> > Explanation of the oops:
> > 
> > > static void intel_i830_init_gtt_entries(void)
> > ...
> > >         } else if (IS_G33) {
> > >         /* G33's GTT size defined in gmch_ctrl */
> > >                 switch (gmch_ctrl & G33_PGETBL_SIZE_MASK) {
> > >                 case G33_PGETBL_SIZE_1M:
> > >                         size = 1024;
> > >                         break;
> > ...
> > >                 size += 4;
> > 
> > size = 1028
> > 
> > Then since we have the BIOS setting 1MB for the device in the GMCH
> > control we get to here:
> > 
> > >         } else {
> > >                 switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
> > >                 case I855_GMCH_GMS_STOLEN_1M:
> > >                         gtt_entries = MB(1) - KB(size);
> > >                         break;
> > 
> > MB(1) = 1 * 1024 * 1024
> > KB(1028) = 1028 * 1024
> > 
> > MB(1) - KB(1028) = -4096
> > 
> > >         gtt_entries /= KB(4);
> > >         intel_private.gtt_entries = gtt_entries;
> > 
> > We end up with -1 in gtt_entries.
> > 
> > This leads to intel_i915_configure reading/writing to areas outside of
> > mapped memory and the oops.
> > 
> > Bugzilla: https://bugzilla.novell.com/show_bug.cgi?id=391261
> > 
> > Signed-off-by: Brandon Philips <bphilips@suse.de>
> > 
> > ---
> >  drivers/char/agp/intel-agp.c |    8 ++++++++
> >  1 file changed, 8 insertions(+)
> > 
> > Index: linux-2.6/drivers/char/agp/intel-agp.c
> > ===================================================================
> > --- linux-2.6.orig/drivers/char/agp/intel-agp.c
> > +++ linux-2.6/drivers/char/agp/intel-agp.c
> > @@ -559,6 +559,14 @@ static void intel_i830_init_gtt_entries(
> >  	} else {
> >  		switch (gmch_ctrl & I855_GMCH_GMS_MASK) {
> >  		case I855_GMCH_GMS_STOLEN_1M:
> > +			if (IS_G33) {
> > +				size = 0;
> > +				printk(KERN_WARNING PFX
> > +				       "Warning: G33 chipset with 1MB"
> > +					" allocated. Older X.org Intel drivers"
> > +					" will not work.\n");
> > +				WARN_ON(1);
> > +			}
> >  			gtt_entries = MB(1) - KB(size);
> >  			break;
> >  		case I855_GMCH_GMS_STOLEN_4M:
> 
> Is the bug which this patch addresses present in the 2.6.27 kernel?

Yes, ever since 3a4986955c0d9806e96a9d738ef7c40cb0cdaba3

Thanks,

	Brandon