Re: i915 module does not find 82865G if configured as secondary

[Tempura San <tempura.san@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 5026 bytes --]

This would be great, but...

I have just tested the patches and it really messes up the system.

On the first boot I got a shell and was able to to a dmesg:
...
[   46.775935] [drm] Initialized drm 1.1.0 20060810
[   46.926398] i915 0000:00:02.0: enabling device (0000 -> 0003)
[   46.926412] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[   46.926422] i915 0000:00:02.0: setting latency timer to 64
[   46.932858] [drm] set up 31M of stolen space
[   46.933087] i915 0000:00:02.0: BAR 6: can't allocate resource (bogus
alignment) [0x0-0x0] flags 0x0
[   46.933093] [drm] failed to find VBIOS tables
[   46.978294] [drm] initialized overlay support
[   47.308831] Console: switching to colour frame buffer device 160x64
[   47.308849] fb0: inteldrmfb frame buffer device
[   47.308853] registered panic notifier
[   47.308868] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on
minor 0
[   47.323940] vgaarb: this pci device is not a vga device
[   47.355806] render error detected, EIR: 0x00000010
[   47.355815] [drm:i915_handle_error] *ERROR* EIR stuck: 0x00000010,
masking
[   47.355835] render error detected, EIR: 0x00000010
[   48.112726] vgaarb: this pci device is not a vga device
[   48.568008] tap0: no IPv6 routers present
[   50.673130] lp0: using parport0 (interrupt-driven).
[   50.842153] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer
elapsed... GPU hung
[   50.842166] render error detected, EIR: 0x00000000
[   50.842200] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request
returns -5 (awaiting 1 at 0)
[   51.100509] ppdev: user-space parallel port driver
[   51.436065] EXT3-fs error (device sda2): ext3_lookup: deleted inode
referenced: 157299
[   51.436076] Aborting journal on device sda2.
[   51.436296] Remounting filesystem read-only
...

On the second boot the computer hung altogether.

Xorg seemed to initialise but ended up with this (last) line in the log
file:
(EE) intel(0): Detected a hung GPU, disabling acceleration.

I have attached the log for completeness.

So I believe for the older kernel there would be more to patch than
this... I have attached the diff of the changes I have done so far -
should be the two patches Chris had listed.

Compared against the sources from:
ii  linux-source-2.6.32                    
2.6.32-38                               Linux kernel source for version
2.6.32 with Debian patches

Were the changes I did correct? Just in case I messed things up myself...

Yours,
Tempura.

On Mon, 10 Oct 2011 21:06:27 +0200, Julien Cristau wrote:
> On Sun, Oct  9, 2011 at 19:36:27 +0100, Chris Wilson wrote:
>
>> On Sun, 9 Oct 2011 14:44:30 +0200, Daniel Vetter <daniel@ffwll.ch> wrote:
>>> On Sun, Oct 09, 2011 at 01:07:25PM +0200, Tempura San wrote:
>>>> Here is the output of lspci -nn:
>>>>
>>>> 00:00.0 Host bridge [0600]: Intel Corporation 82865G/PE/P DRAM
>>>> Controller/Host-Hub Interface [8086:2570] (rev 02)
>>>> 00:02.0 Display controller [0380]: Intel Corporation 82865G Integrated
>>>> Graphics Controller [8086:2572] (rev 02)
>>> The issue seems to be that the igd isn't a VGA pci class device anymore
>>> when used as secondary. The below (untested) patch should allow to still
>>> bind the i915 driver. Please test how far that gets us.
>> Note the old kernel, and probable lack of:
>>
>> commit 934f992c763ae1e5eefcce8af769c16444085df7
>> Author: Chris Wilson <chris@chris-wilson.co.uk>
>> Date:   Thu Jan 20 13:09:12 2011 +0000
>>
>>     drm/i915: Recognise non-VGA display devices
>>     
>>     Starting with SandyBridge (though possible with earlier hacked BIOSes),
>>     the BIOS may initialise the IGFX as secondary to a discrete GPU. Prior,
>>     it would simply disable the integrated GPU. So we adjust our PCI class
>>     mask to match any DISPLAY_CLASS device.
>>     
>>     In such a configuration, the IGFX is not a primary VGA controller and
>>     so should not take part in VGA arbitration, and the error return from
>>     vga_client_register() is expected.
>>     
>>     Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>>     Cc: stable@kernel.org
>>
>> and
>>
>> commit 5fe49d86f9d01044abf687a8cd21edef636d58aa
>> Author: Chris Wilson <chris@chris-wilson.co.uk>
>> Date:   Tue Feb 1 19:43:02 2011 +0000
>>
>>     drm/i915: Only bind to function 0 of the PCI device
>>     
>>     Early chipsets (gen2/3) used function 1 as a placeholder for multi-head.
>>     We used to ignore these since they were not assigned to
>>     PCI_CLASS_DISPLAY_VGA. However with 934f992c7 we attempt to bind to all
>>     Intel PCI_CLASS_DISPLAY devices (and functions) to work in multi-gpu
>>     systems. This fails hard on gen2/3.
>>     
>>     Reported-by: Ferenc W??gner <wferi@niif.hu>
>>     Tested-by: Ferenc W??gner <wferi@niif.hu>
>>     Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=28012
>>     Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>>     Cc: stable@kernel.org
> I'll try to get these included into a future squeeze kernel update.
> Thanks.
>
> Cheers,
> Julien

[-- Attachment #2: i915-for-squeeze.diff --]
[-- Type: text/x-patch, Size: 2525 bytes --]

diff -uNr linux-source-2.6.32/drivers/gpu/drm/i915/i915_dma.c linux-source-2.6.32-i915/drivers/gpu/drm/i915/i915_dma.c
--- linux-source-2.6.32/drivers/gpu/drm/i915/i915_dma.c	2011-10-02 01:46:54.000000000 +0200
+++ linux-source-2.6.32-i915/drivers/gpu/drm/i915/i915_dma.c	2011-10-09 21:55:14.000000000 +0200
@@ -1267,9 +1267,15 @@
 	if (ret)
 		DRM_INFO("failed to find VBIOS tables\n");
 
-	/* if we have > 1 VGA cards, then disable the radeon VGA resources */
-	ret = vga_client_register(dev->pdev, dev, NULL, i915_vga_set_decode);
-	if (ret)
+	/* If we have > 1 VGA cards, then we need to arbitrate access
+	 * to the common VGA resources.
+	 *
+	 * If we are a secondary display controller (!PCI_DISPLAY_CLASS_VGA),
+	 * then we do not take part in VGA arbitration and the
+	 * vga_client_register() fails with -ENODEV.
+	 */
+ 	ret = vga_client_register(dev->pdev, dev, NULL, i915_vga_set_decode);
+	if (ret && ret != -ENODEV)
 		goto destroy_ringbuffer;
 
 	/* IIR "flip pending" bit means done if this bit is set */
diff -uNr linux-source-2.6.32/drivers/gpu/drm/i915/i915_drv.c linux-source-2.6.32-i915/drivers/gpu/drm/i915/i915_drv.c
--- linux-source-2.6.32/drivers/gpu/drm/i915/i915_drv.c	2011-10-02 01:46:49.000000000 +0200
+++ linux-source-2.6.32-i915/drivers/gpu/drm/i915/i915_drv.c	2011-10-09 21:59:58.000000000 +0200
@@ -52,7 +52,7 @@
 
 #define INTEL_VGA_DEVICE(id, info) {		\
 	.class = PCI_CLASS_DISPLAY_VGA << 8,	\
-	.class_mask = 0xffff00,			\
+	.class_mask = 0xff0000,			\
 	.vendor = 0x8086,			\
 	.device = id,				\
 	.subvendor = PCI_ANY_ID,		\
@@ -393,6 +393,14 @@
 static int __devinit
 i915_pci_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 {
+	/* Only bind to function 0 of the device. Early generations
+	 * used function 1 as a placeholder for multi-head. This causes
+	 * us confusion instead, especially on the systems where both
+	 * functions have the same PCI-ID!
+	 */
+	if (PCI_FUNC(pdev->devfn))
+		return -ENODEV;
+
 	return drm_get_dev(pdev, ent, &driver);
 }
diff -uNr linux-source-2.6.32/drivers/gpu/vga/vgaarb.c linux-source-2.6.32-i915/drivers/gpu/vga/vgaarb.c
--- linux-source-2.6.32/drivers/gpu/vga/vgaarb.c	2011-10-02 01:46:46.000000000 +0200
+++ linux-source-2.6.32-i915/drivers/gpu/vga/vgaarb.c	2011-10-09 21:56:59.000000000 +0200
@@ -614,7 +614,7 @@
 			void (*irq_set_state)(void *cookie, bool state),
 			unsigned int (*set_vga_decode)(void *cookie, bool decode))
 {
-	int ret = -1;
+	int ret = -ENODEV;
 	struct vga_device *vgadev;
 	unsigned long flags;
 

[-- Attachment #3: Xorg.0.log.old --]
[-- Type: application/x-trash, Size: 18778 bytes --]

[-- Attachment #4: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx