Skip to contentSkip to conversation

drm/i915: Another version of the vga stop_machine monstrosity

PATCH New [email protected] 5 participants 18 messages Sep 30, 2013 – Oct 7, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover drm/i915: Another version of the vga stop_machine monstrosity

Here's a slightly improved version of the VGA stop_machine() patch.

The main new thing is switching ctg/elk over to port IO. That appears to have cured the hangs from Chris's lid_notify riddled ctg machine.

What I'd really like is some kind of tested-by note from Alex Williamson to make sure I didn't accidentally break something important. Well, assuming people are willing to overlook the uglyness to slip this in at all.

#1 lore raw

PATCH v2 1/1 drm/i915: Fix VGA handling using stop_machine() or mmio

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

[1] http://lists.x.org/archives/xorg-devel/2013-September/037763.html

v2: Use SNB_GMCH_TRL on SNB+
Use port IO instead of MMIO on CTG/ELK Add WaEnableVGAAccessThroughIOPort comment Fix the max number of devices on the bus limit

Cc Alex Williamson <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Dave Airlie <[email protected]>
---
 drivers/gpu/drm/i915/i915_dma.c      |  39 +-----
 drivers/gpu/drm/i915/i915_suspend.c  |   4 +-
 drivers/gpu/drm/i915/intel_display.c | 248 +++++++++++++++++++++++++++++------
 3 files changed, 217 insertions(+), 74 deletions(-)
drivers/gpu/drm/i915/i915_dma.c +6 -33
index d35de1b..59a2048 100644
@@ -1225,19 +1225,6 @@ intel_teardown_mchbar(struct drm_device *dev)
 		release_resource(&dev_priv->mch_res);
 }
 
-/* true = enable decode, false = disable decoder */
-static unsigned int i915_vga_set_decode(void *cookie, bool state)
-{
-	struct drm_device *dev = cookie;
-
-	intel_modeset_vga_set_state(dev, state);
-	if (state)
-		return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
-		       VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
-	else
-		return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
-}
-
 static void i915_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_state state)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
@@ -1283,25 +1270,11 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	if (ret)
 		DRM_INFO("failed to find VBIOS tables\n");
 
-	/* 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.
-	 */
-	if (!HAS_PCH_SPLIT(dev)) {
-		ret = vga_client_register(dev->pdev, dev, NULL,
-					  i915_vga_set_decode);
-		if (ret && ret != -ENODEV)
-			goto out;
-	}
-
 	intel_register_dsm_handler();
 
 	ret = vga_switcheroo_register_client(dev->pdev, &i915_switcheroo_ops, false);
 	if (ret)
-		goto cleanup_vga_client;
+		goto out;
 
 	/* Initialise stolen first so that we may reserve preallocated
 	 * objects for the BIOS to KMS transition.
@@ -1316,7 +1289,7 @@ static int i915_load_modeset_init(struct drm_device *dev)
 
 	intel_init_power_well(dev);
 
-	/* Keep VGA alive until i915_disable_vga_mem() */
+	/* Keep VGA alive until intel_modeset_vga_set_state() */
 	intel_display_power_get(dev, POWER_DOMAIN_VGA);
 
 	/* Important: The output setup functions called by modeset_init need
@@ -1359,10 +1332,13 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	intel_fbdev_initial_config(dev);
 
 	/*
+	 * Disable VGA IO and memory, and
+	 * tell the arbiter to ignore us.
+	 *
 	 * Must do this after fbcon init so that
 	 * vgacon_save_screen() works during the handover.
 	 */
-	i915_disable_vga_mem(dev);
+	intel_modeset_vga_set_state(dev, false);
 	intel_display_power_put(dev, POWER_DOMAIN_VGA);
 
 	/* Only enable hotplug handling once the fbdev is fully set up. */
@@ -1386,8 +1362,6 @@ cleanup_gem_stolen:
 	i915_gem_cleanup_stolen(dev);
 cleanup_vga_switcheroo:
 	vga_switcheroo_unregister_client(dev->pdev);
-cleanup_vga_client:
-	vga_client_register(dev->pdev, NULL, NULL, NULL);
 out:
 	return ret;
 }
@@ -1758,7 +1732,6 @@ int i915_driver_unload(struct drm_device *dev)
 		}
 
 		vga_switcheroo_unregister_client(dev->pdev);
-		vga_client_register(dev->pdev, NULL, NULL, NULL);
 	}
 
 	/* Free error state after interrupts are fully disabled. */
drivers/gpu/drm/i915/i915_suspend.c +3 -1
index 3538370..cc0d459 100644
@@ -331,8 +331,10 @@ static void i915_restore_display(struct drm_device *dev)
 
 	if (!drm_core_check_feature(dev, DRIVER_MODESET))
 		i915_restore_vga(dev);
-	else
+	else {
 		i915_redisable_vga(dev);
+		intel_modeset_vga_set_state(dev, false);
+	}
 }
 
 int i915_save_state(struct drm_device *dev)
drivers/gpu/drm/i915/intel_display.c +208 -40

291 lines not on this page · raw message

Signature · 6 lines
1.8.1.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tested-by on v2 1/1
#2 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 8 lines · On Mon, Sep 30, 2013 at 05:08:31PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

I'd like to emphasize that X disables DRI if it detects 2 vga cards, effectively breaking all machines with a discrete GPU. Even if one of those is not being used.

drivers/gpu/drm/i915/intel_display.c static void intel_init_quirks(struct drm_device *dev) · line 10252

Quoted 6 lines
drivers/gpu/drm/i915/intel_display.c +6 -0
+/*
+ * 21 devices with 8 functions per device max on the same bus.
+ * We don't need locking for these due to stop_machine().
+ */
+static u16 vga_cmd[21*8];
+static u16 vga_ctl[21*8];

Should we just allocate storage for when we need it? We are now adding several hundred bytes to our module, which is bound to cause us to use an extra page, and they can be passed around through the stop_machine closure rather than static.

But anyway, it does what is says on the tin and makes my dGPU testing box usable again, without breaking any other machine that I've tested on so far,

-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#3 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 14 lines · Chris Wilson
Quoted 8 lines · On Mon, Sep 30, 2013 at 05:08:31PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

I'd like to emphasize that X disables DRI if it detects 2 vga cards, effectively breaking all machines with a discrete GPU. Even if one of those is not being used.

Quoted 11 lines

drivers/gpu/drm/i915/intel_display.c static void intel_init_quirks(struct drm_device *dev) · line 10252

Quoted 6 lines
drivers/gpu/drm/i915/intel_display.c +6 -0
+/*
+ * 21 devices with 8 functions per device max on the same bus.
+ * We don't need locking for these due to stop_machine().
+ */
+static u16 vga_cmd[21*8];
+static u16 vga_ctl[21*8];

Should we just allocate storage for when we need it? We are now adding several hundred bytes to our module, which is bound to cause us to use an extra page, and they can be passed around through the stop_machine closure rather than static.

I guess we could do that. Although I do wonder a bit if we'd race with hotplug. Not sure there's a way to hotplug stuff onto the root bus...

Quoted 9 lines

But anyway, it does what is says on the tin and makes my dGPU testing box usable again, without breaking any other machine that I've tested on so far,

-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
Signature · 2 lines
Ville Syrjälä
Intel OTC
#4 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 27 lines · Ville Syrjälä
Quoted 14 lines · Chris Wilson
Quoted 8 lines · On Mon, Sep 30, 2013 at 05:08:31PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

I'd like to emphasize that X disables DRI if it detects 2 vga cards, effectively breaking all machines with a discrete GPU. Even if one of those is not being used.

Quoted 11 lines

drivers/gpu/drm/i915/intel_display.c static void intel_init_quirks(struct drm_device *dev) · line 10252

Quoted 6 lines
drivers/gpu/drm/i915/intel_display.c +6 -0
+/*
+ * 21 devices with 8 functions per device max on the same bus.
+ * We don't need locking for these due to stop_machine().
+ */
+static u16 vga_cmd[21*8];
+static u16 vga_ctl[21*8];

Should we just allocate storage for when we need it? We are now adding several hundred bytes to our module, which is bound to cause us to use an extra page, and they can be passed around through the stop_machine closure rather than static.

Quoted 2 lines

I guess we could do that. Although I do wonder a bit if we'd race with hotplug. Not sure there's a way to hotplug stuff onto the root bus...

Within stop-machine? I surely hope not!
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#5 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 28 lines · Chris Wilson
Quoted 27 lines · Ville Syrjälä
Quoted 25 lines · Chris Wilson
Quoted 8 lines · On Mon, Sep 30, 2013 at 05:08:31PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

I'd like to emphasize that X disables DRI if it detects 2 vga cards, effectively breaking all machines with a discrete GPU. Even if one of those is not being used.

drivers/gpu/drm/i915/intel_display.c static void intel_init_quirks(struct drm_device *dev) · line 10252

Quoted 6 lines
drivers/gpu/drm/i915/intel_display.c +6 -0
+/*
+ * 21 devices with 8 functions per device max on the same bus.
+ * We don't need locking for these due to stop_machine().
+ */
+static u16 vga_cmd[21*8];
+static u16 vga_ctl[21*8];

Should we just allocate storage for when we need it? We are now adding several hundred bytes to our module, which is bound to cause us to use an extra page, and they can be passed around through the stop_machine closure rather than static.

Quoted 4 lines
Quoted 2 lines

I guess we could do that. Although I do wonder a bit if we'd race with hotplug. Not sure there's a way to hotplug stuff onto the root bus...

Within stop-machine? I surely hope not!

Can we allocate there? I was thinking of allocating before stop_machine().

Signature · 2 lines
Ville Syrjälä
Intel OTC
#6 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 29 lines · Ville Syrjälä
Quoted 28 lines · Chris Wilson
Quoted 27 lines · Ville Syrjälä
Quoted 25 lines · Chris Wilson
Quoted 8 lines · On Mon, Sep 30, 2013 at 05:08:31PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

I'd like to emphasize that X disables DRI if it detects 2 vga cards, effectively breaking all machines with a discrete GPU. Even if one of those is not being used.

drivers/gpu/drm/i915/intel_display.c static void intel_init_quirks(struct drm_device *dev) · line 10252

Quoted 6 lines
drivers/gpu/drm/i915/intel_display.c +6 -0
+/*
+ * 21 devices with 8 functions per device max on the same bus.
+ * We don't need locking for these due to stop_machine().
+ */
+static u16 vga_cmd[21*8];
+static u16 vga_ctl[21*8];

Should we just allocate storage for when we need it? We are now adding several hundred bytes to our module, which is bound to cause us to use an extra page, and they can be passed around through the stop_machine closure rather than static.

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

I guess we could do that. Although I do wonder a bit if we'd race with hotplug. Not sure there's a way to hotplug stuff onto the root bus...

Within stop-machine? I surely hope not!

Can we allocate there? I was thinking of allocating before stop_machine().

Using GFP_ATOMIC should be permissible, I think. But yeah, you can allocate space upfront for struct vga_op.
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#7 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

In reply to Chris Wilson
Quoted 10 lines · Chris Wilson
Quoted 8 lines · On Mon, Sep 30, 2013 at 05:08:31PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

Quoted 3 lines

I'd like to emphasize that X disables DRI if it detects 2 vga cards, effectively breaking all machines with a discrete GPU. Even if one of those is not being used.

Why does it do this? It seems like DRI would make little or no use of VGA space. Having more than one VGA card seems like a pretty common condition when integrated graphics are available. We also seem to have quite some interest in assigning one or more of the cards to a virtual machine, so I worry we're headed the wrong way if X starts deciding not to release the VGA arbiter lock. On a modern desktop what touches VGA space with a high enough frequency that we care about it's performance? Thanks,

Alex

Quoted 12 lines

drivers/gpu/drm/i915/intel_display.c static void intel_init_quirks(struct drm_device *dev) · line 10252

Quoted 6 lines
drivers/gpu/drm/i915/intel_display.c +6 -0
+/*
+ * 21 devices with 8 functions per device max on the same bus.
+ * We don't need locking for these due to stop_machine().
+ */
+static u16 vga_cmd[21*8];
+static u16 vga_ctl[21*8];

Should we just allocate storage for when we need it? We are now adding several hundred bytes to our module, which is bound to cause us to use an extra page, and they can be passed around through the stop_machine closure rather than static.

Quoted 6 lines

But anyway, it does what is says on the tin and makes my dGPU testing box usable again, without breaking any other machine that I've tested on so far,

-Chris

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

maintainer
#8 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 15 lines · Alex Williamson
Quoted 10 lines · Chris Wilson
Quoted 8 lines · On Mon, Sep 30, 2013 at 05:08:31PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

Quoted 3 lines

I'd like to emphasize that X disables DRI if it detects 2 vga cards, effectively breaking all machines with a discrete GPU. Even if one of those is not being used.

Quoted 9 lines

Why does it do this? It seems like DRI would make little or no use of VGA space. Having more than one VGA card seems like a pretty common condition when integrated graphics are available. We also seem to have quite some interest in assigning one or more of the cards to a virtual machine, so I worry we're headed the wrong way if X starts deciding not to release the VGA arbiter lock. On a modern desktop what touches VGA space with a high enough frequency that we care about it's performance? Thanks,

Because we don't know what DRI clients can do, and a lot of this was constructed when DRI1 still occured. The idea being that sane GPUs would remove themselves from arbitration by just switching off their VGA spaces, who was to know Intel hw guys would fuck this up, probably should have guessed though.

So the problem we have is that VGA ARB's current interface causes interactions with the current X server that probably aren't totally required, so by trying to actually VGA arbitrate properly we are going to regress userspace so we now need to design a solution that avoids the regression but possibly allows us to move forward.

I'm think I'm going to make an executive decision to merge Ville's latest patch to avoid the regression, the other option being just to revert everything back to the status quo by reverting all the patches from the past few months, if people are happier with that then maybe we should just do that for now and try and design our way out of it properly by reengineering userspace and plan to avoid the regression next time.

Dave.

#9 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 7 lines · Dave Airlie

I'm think I'm going to make an executive decision to merge Ville's latest patch to avoid the regression, the other option being just to revert everything back to the status quo by reverting all the patches from the past few months, if people are happier with that then maybe we should just do that for now and try and design our way out of it properly by reengineering userspace and plan to avoid the regression next time.

Ack on merging Ville's hack. Longer-term we could beautify the pig a bit by moving it into vgaarb.c and avoiding the stop_machine if no one else has a vga lock. And I think in next we could try to limit the vga redisable hack in the lid notifier a bit on more modern platforms to avoid the dreaded stop_machine at runtime and relegate it to just resume/driver load.

Really long term (once broken X servers have died) we can try to rip this hack out again. Or maybe intel hw engineers get their act together beforehand ;-)

So can you directly pick it up since I've just flushed out my -fixes queue?

Thanks, Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#10 lore raw

PATCH v3 1/1 drm/i915: Fix VGA handling using stop_machine() or mmio

In reply to Chris Wilson

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

[1] http://lists.x.org/archives/xorg-devel/2013-September/037763.html

v2: Use SNB_GMCH_TRL on SNB+
Use port IO instead of MMIO on CTG/ELK
Add WaEnableVGAAccessThroughIOPort comment
Fix the max number of devices on the bus limit
v3: Allocate the temp space dynamically
Print some errors if we fail to execute the vga "op" due to alloc failure

Cc Alex Williamson <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Dave Airlie <[email protected]>
---
 drivers/gpu/drm/i915/i915_dma.c      |  39 +----
 drivers/gpu/drm/i915/i915_suspend.c  |   4 +-
 drivers/gpu/drm/i915/intel_display.c | 318 ++++++++++++++++++++++++++++++-----
 3 files changed, 287 insertions(+), 74 deletions(-)
drivers/gpu/drm/i915/i915_dma.c +6 -33
index d35de1b..59a2048 100644
@@ -1225,19 +1225,6 @@ intel_teardown_mchbar(struct drm_device *dev)
 		release_resource(&dev_priv->mch_res);
 }
 
-/* true = enable decode, false = disable decoder */
-static unsigned int i915_vga_set_decode(void *cookie, bool state)
-{
-	struct drm_device *dev = cookie;
-
-	intel_modeset_vga_set_state(dev, state);
-	if (state)
-		return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
-		       VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
-	else
-		return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
-}
-
 static void i915_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_state state)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
@@ -1283,25 +1270,11 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	if (ret)
 		DRM_INFO("failed to find VBIOS tables\n");
 
-	/* 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.
-	 */
-	if (!HAS_PCH_SPLIT(dev)) {
-		ret = vga_client_register(dev->pdev, dev, NULL,
-					  i915_vga_set_decode);
-		if (ret && ret != -ENODEV)
-			goto out;
-	}
-
 	intel_register_dsm_handler();
 
 	ret = vga_switcheroo_register_client(dev->pdev, &i915_switcheroo_ops, false);
 	if (ret)
-		goto cleanup_vga_client;
+		goto out;
 
 	/* Initialise stolen first so that we may reserve preallocated
 	 * objects for the BIOS to KMS transition.
@@ -1316,7 +1289,7 @@ static int i915_load_modeset_init(struct drm_device *dev)
 
 	intel_init_power_well(dev);
 
-	/* Keep VGA alive until i915_disable_vga_mem() */
+	/* Keep VGA alive until intel_modeset_vga_set_state() */
 	intel_display_power_get(dev, POWER_DOMAIN_VGA);
 
 	/* Important: The output setup functions called by modeset_init need
@@ -1359,10 +1332,13 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	intel_fbdev_initial_config(dev);
 
 	/*
+	 * Disable VGA IO and memory, and
+	 * tell the arbiter to ignore us.
+	 *
 	 * Must do this after fbcon init so that
 	 * vgacon_save_screen() works during the handover.
 	 */
-	i915_disable_vga_mem(dev);
+	intel_modeset_vga_set_state(dev, false);
 	intel_display_power_put(dev, POWER_DOMAIN_VGA);
 
 	/* Only enable hotplug handling once the fbdev is fully set up. */
@@ -1386,8 +1362,6 @@ cleanup_gem_stolen:
 	i915_gem_cleanup_stolen(dev);
 cleanup_vga_switcheroo:
 	vga_switcheroo_unregister_client(dev->pdev);
-cleanup_vga_client:
-	vga_client_register(dev->pdev, NULL, NULL, NULL);
 out:
 	return ret;
 }
@@ -1758,7 +1732,6 @@ int i915_driver_unload(struct drm_device *dev)
 		}
 
 		vga_switcheroo_unregister_client(dev->pdev);
-		vga_client_register(dev->pdev, NULL, NULL, NULL);
 	}
 
 	/* Free error state after interrupts are fully disabled. */
drivers/gpu/drm/i915/i915_suspend.c +3 -1
index 3538370..cc0d459 100644
@@ -331,8 +331,10 @@ static void i915_restore_display(struct drm_device *dev)
 
 	if (!drm_core_check_feature(dev, DRIVER_MODESET))
 		i915_restore_vga(dev);
-	else
+	else {
 		i915_redisable_vga(dev);
+		intel_modeset_vga_set_state(dev, false);
+	}
 }
 
 int i915_save_state(struct drm_device *dev)
drivers/gpu/drm/i915/intel_display.c +278 -40

361 lines not on this page · raw message

Signature · 6 lines
1.8.1.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#11 lore raw

Re: [PATCH v3] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 28 lines · On Wed, Oct 02, 2013 at 04:42:55PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

Quoted 8 lines

[1] http://lists.x.org/archives/xorg-devel/2013-September/037763.html

v2: Use SNB_GMCH_TRL on SNB+
Use port IO instead of MMIO on CTG/ELK
Add WaEnableVGAAccessThroughIOPort comment
Fix the max number of devices on the bus limit
v3: Allocate the temp space dynamically
Print some errors if we fail to execute the vga "op" due to alloc failure

Passes the dGPU test, the SNB laptop, but is doa for CTG. -Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#12 lore raw

Re: [PATCH v3] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 31 lines · Chris Wilson
Quoted 30 lines · On Wed, Oct 02, 2013 at 04:42:55PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

[1] http://lists.x.org/archives/xorg-devel/2013-September/037763.html

Quoted 8 lines
Quoted 6 lines

v2: Use SNB_GMCH_TRL on SNB+
Use port IO instead of MMIO on CTG/ELK
Add WaEnableVGAAccessThroughIOPort comment
Fix the max number of devices on the bus limit
v3: Allocate the temp space dynamically
Print some errors if we fail to execute the vga "op" due to alloc failure

Passes the dGPU test, the SNB laptop, but is doa for CTG.

Crap. And v2 still works all right there?

Signature · 2 lines
Ville Syrjälä
Intel OTC
#13 lore raw

Re: [PATCH v3] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 39 lines · Ville Syrjälä
Quoted 38 lines · Chris Wilson
Quoted 36 lines · On Wed, Oct 02, 2013 at 04:42:55PM +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

[1] http://lists.x.org/archives/xorg-devel/2013-September/037763.html

v2: Use SNB_GMCH_TRL on SNB+
Use port IO instead of MMIO on CTG/ELK
Add WaEnableVGAAccessThroughIOPort comment
Fix the max number of devices on the bus limit
v3: Allocate the temp space dynamically
Print some errors if we fail to execute the vga "op" due to alloc failure

Quoted 3 lines
Quoted 1 line

Passes the dGPU test, the SNB laptop, but is doa for CTG.

Crap. And v2 still works all right there?

No, appears I mislead you. The last one to successfully work was the composite v1 patch set. Sorry.
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#14 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

In reply to [email protected]
Quoted 23 lines · On Mon, 2013-09-30 at 17:08 +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

Quoted 4 lines

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

What?! Why would we not simply wait for the arbiter lock?

Quoted 1 line

[1] http://lists.x.org/archives/xorg-devel/2013-September/037763.html

Eek, does this X patch effectively kill VGA Arbiter interaction unless an instance of X has multiple adapters needing VGA?

"Secondly, we then only wrap the operations with vgaarb get/put for the drivers that require VGA access. If we only have a single driver requring VGA access, we just wrap Enter/LeaveVT and lock the VGA arbiter for the entire duration that the Xserver is active."

Maybe I'm reading the code wrong, but it looks like if xf86VGAarbiterInit() is not called multiple times then X just locks VGA and never unlocks it. If so, this is broken! How is this supposed to work with multiple X sessions? How can it ever work if we want to assign a VGA device to a virtual machine? Chris, please tell me I'm misreading the intent of this patch. Thanks,

Alex

Quoted 414 lines

v2: Use SNB_GMCH_TRL on SNB+
Use port IO instead of MMIO on CTG/ELK Add WaEnableVGAAccessThroughIOPort comment Fix the max number of devices on the bus limit

Cc Alex Williamson <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Dave Airlie <[email protected]>
---
 drivers/gpu/drm/i915/i915_dma.c      |  39 +-----
 drivers/gpu/drm/i915/i915_suspend.c  |   4 +-
 drivers/gpu/drm/i915/intel_display.c | 248 +++++++++++++++++++++++++++++------
 3 files changed, 217 insertions(+), 74 deletions(-)
drivers/gpu/drm/i915/i915_dma.c +6 -33
index d35de1b..59a2048 100644
@@ -1225,19 +1225,6 @@ intel_teardown_mchbar(struct drm_device *dev)
 		release_resource(&dev_priv->mch_res);
 }
 
-/* true = enable decode, false = disable decoder */
-static unsigned int i915_vga_set_decode(void *cookie, bool state)
-{
-	struct drm_device *dev = cookie;
-
-	intel_modeset_vga_set_state(dev, state);
-	if (state)
-		return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
-		       VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
-	else
-		return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
-}
-
 static void i915_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_state state)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
@@ -1283,25 +1270,11 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	if (ret)
 		DRM_INFO("failed to find VBIOS tables\n");
 
-	/* 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.
-	 */
-	if (!HAS_PCH_SPLIT(dev)) {
-		ret = vga_client_register(dev->pdev, dev, NULL,
-					  i915_vga_set_decode);
-		if (ret && ret != -ENODEV)
-			goto out;
-	}
-
 	intel_register_dsm_handler();
 
 	ret = vga_switcheroo_register_client(dev->pdev, &i915_switcheroo_ops, false);
 	if (ret)
-		goto cleanup_vga_client;
+		goto out;
 
 	/* Initialise stolen first so that we may reserve preallocated
 	 * objects for the BIOS to KMS transition.
@@ -1316,7 +1289,7 @@ static int i915_load_modeset_init(struct drm_device *dev)
 
 	intel_init_power_well(dev);
 
-	/* Keep VGA alive until i915_disable_vga_mem() */
+	/* Keep VGA alive until intel_modeset_vga_set_state() */
 	intel_display_power_get(dev, POWER_DOMAIN_VGA);
 
 	/* Important: The output setup functions called by modeset_init need
@@ -1359,10 +1332,13 @@ static int i915_load_modeset_init(struct drm_device *dev)
 	intel_fbdev_initial_config(dev);
 
 	/*
+	 * Disable VGA IO and memory, and
+	 * tell the arbiter to ignore us.
+	 *
 	 * Must do this after fbcon init so that
 	 * vgacon_save_screen() works during the handover.
 	 */
-	i915_disable_vga_mem(dev);
+	intel_modeset_vga_set_state(dev, false);
 	intel_display_power_put(dev, POWER_DOMAIN_VGA);
 
 	/* Only enable hotplug handling once the fbdev is fully set up. */
@@ -1386,8 +1362,6 @@ cleanup_gem_stolen:
 	i915_gem_cleanup_stolen(dev);
 cleanup_vga_switcheroo:
 	vga_switcheroo_unregister_client(dev->pdev);
-cleanup_vga_client:
-	vga_client_register(dev->pdev, NULL, NULL, NULL);
 out:
 	return ret;
 }
@@ -1758,7 +1732,6 @@ int i915_driver_unload(struct drm_device *dev)
 		}
 
 		vga_switcheroo_unregister_client(dev->pdev);
-		vga_client_register(dev->pdev, NULL, NULL, NULL);
 	}
 
 	/* Free error state after interrupts are fully disabled. */
drivers/gpu/drm/i915/i915_suspend.c +3 -1
index 3538370..cc0d459 100644
@@ -331,8 +331,10 @@ static void i915_restore_display(struct drm_device *dev)
 
 	if (!drm_core_check_feature(dev, DRIVER_MODESET))
 		i915_restore_vga(dev);
-	else
+	else {
 		i915_redisable_vga(dev);
+		intel_modeset_vga_set_state(dev, false);
+	}
 }
 
 int i915_save_state(struct drm_device *dev)
drivers/gpu/drm/i915/intel_display.c +201 -40
index 76870f0..25955be 100644
@@ -40,6 +40,7 @@
 #include <drm/drm_dp_helper.h>
 #include <drm/drm_crtc_helper.h>
 #include <linux/dma_remapping.h>
+#include <linux/stop_machine.h>
 
 static void intel_increase_pllclock(struct drm_crtc *crtc);
 static void intel_crtc_update_cursor(struct drm_crtc *crtc, bool on);
@@ -10232,51 +10233,195 @@ static void intel_init_quirks(struct drm_device *dev)
 	}
 }
 
-/* Disable the VGA plane that we never use */
-static void i915_disable_vga(struct drm_device *dev)
+enum vga_op {
+	VGA_OP_MEM_DISABLE,
+	VGA_OP_MEM_ENABLE,
+	VGA_OP_SCREEN_OFF,
+};
+
+struct i915_vga_op {
+	enum vga_op op;
+	struct drm_device *dev;
+};
+
+/*
+ * 21 devices with 8 functions per device max on the same bus.
+ * We don't need locking for these due to stop_machine().
+ */
+static u16 vga_cmd[21*8];
+static u16 vga_ctl[21*8];
+
+/*
+ * Disable IO decode and VGA bridge forwarding
+ * for everyone else on the same bus.
+ */
+static void i915_vga_bus_disable(struct drm_device *dev)
 {
-	struct drm_i915_private *dev_priv = dev->dev_private;
-	u8 sr1;
-	u32 vga_reg = i915_vgacntrl_reg(dev);
+	struct pci_dev *pdev, *self = dev->pdev;
+	struct pci_bus *bus;
+	int i;
 
-	vga_get_uninterruptible(dev->pdev, VGA_RSRC_LEGACY_IO);
-	outb(SR01, VGA_SR_INDEX);
-	sr1 = inb(VGA_SR_DATA);
-	outb(sr1 | 1<<5, VGA_SR_DATA);
-	vga_put(dev->pdev, VGA_RSRC_LEGACY_IO);
-	udelay(300);
+	i = 0;
+	list_for_each_entry(bus, &self->bus->children, node) {
+		pci_read_config_word(bus->self, PCI_BRIDGE_CONTROL, &vga_ctl[i]);
+		pci_write_config_word(bus->self, PCI_BRIDGE_CONTROL, vga_ctl[i] & ~PCI_BRIDGE_CTL_VGA);
 
-	I915_WRITE(vga_reg, VGA_DISP_DISABLE);
-	POSTING_READ(vga_reg);
+		if (WARN_ON(++i >= ARRAY_SIZE(vga_ctl)))
+			break;
+	}
+
+	i = 0;
+	list_for_each_entry(pdev, &self->bus->devices, bus_list) {
+		if (pdev == self)
+			continue;
+
+		pci_read_config_word(pdev, PCI_COMMAND, &vga_cmd[i]);
+		pci_write_config_word(pdev, PCI_COMMAND, vga_cmd[i] & ~PCI_COMMAND_IO);
+
+		if (WARN_ON(++i >= ARRAY_SIZE(vga_cmd)))
+			break;
+	}
 }
 
-static void i915_enable_vga_mem(struct drm_device *dev)
+/*
+ * Restore IO decode and VGA bridge forwarding
+ * for everyone else on the same bus.
+ */
+static void i915_vga_bus_restore(struct drm_device *dev)
 {
-	/* Enable VGA memory on Intel HD */
-	if (HAS_PCH_SPLIT(dev)) {
-		vga_get_uninterruptible(dev->pdev, VGA_RSRC_LEGACY_IO);
-		outb(inb(VGA_MSR_READ) | VGA_MSR_MEM_EN, VGA_MSR_WRITE);
-		vga_set_legacy_decoding(dev->pdev, VGA_RSRC_LEGACY_IO |
-						   VGA_RSRC_LEGACY_MEM |
-						   VGA_RSRC_NORMAL_IO |
-						   VGA_RSRC_NORMAL_MEM);
-		vga_put(dev->pdev, VGA_RSRC_LEGACY_IO);
+	struct pci_dev *pdev, *self = dev->pdev;
+	struct pci_bus *bus;
+	int i;
+
+	i = 0;
+	list_for_each_entry(pdev, &self->bus->devices, bus_list) {
+		if (pdev == self)
+			continue;
+
+		pci_write_config_word(pdev, PCI_COMMAND, vga_cmd[i]);
+
+		if (WARN_ON(++i >= ARRAY_SIZE(vga_cmd)))
+			break;
+	}
+
+	i = 0;
+	list_for_each_entry(bus, &self->bus->children, node) {
+		pci_write_config_word(bus->self, PCI_BRIDGE_CONTROL, vga_ctl[i]);
+
+		if (WARN_ON(++i >= ARRAY_SIZE(vga_ctl)))
+			break;
 	}
 }
 
-void i915_disable_vga_mem(struct drm_device *dev)
+/*
+ * Hide our VGA IO access from the rest of the system
+ * using stop_machine().
+ *
+ * Note that we assume that the IGD is on the root bus.
+ */
+static int i915_vga_stop_machine_cb(void *data)
 {
-	/* Disable VGA memory on Intel HD */
-	if (HAS_PCH_SPLIT(dev)) {
-		vga_get_uninterruptible(dev->pdev, VGA_RSRC_LEGACY_IO);
-		outb(inb(VGA_MSR_READ) & ~VGA_MSR_MEM_EN, VGA_MSR_WRITE);
-		vga_set_legacy_decoding(dev->pdev, VGA_RSRC_LEGACY_IO |
-						   VGA_RSRC_NORMAL_IO |
-						   VGA_RSRC_NORMAL_MEM);
-		vga_put(dev->pdev, VGA_RSRC_LEGACY_IO);
+	struct i915_vga_op *op = data;
+	struct drm_device *dev = op->dev;
+	u16 cmd;
+	u8 tmp;
+
+	i915_vga_bus_disable(dev);
+
+	pci_read_config_word(dev->pdev, PCI_COMMAND, &cmd);
+	pci_write_config_word(dev->pdev, PCI_COMMAND, cmd | PCI_COMMAND_IO);
+
+	switch (op->op) {
+	case VGA_OP_MEM_DISABLE:
+		tmp = inb(VGA_MSR_READ);
+		tmp &= ~VGA_MSR_MEM_EN;
+		outb(tmp, VGA_MSR_WRITE);
+		break;
+	case VGA_OP_MEM_ENABLE:
+		tmp = inb(VGA_MSR_READ);
+		tmp |= VGA_MSR_MEM_EN;
+		outb(tmp, VGA_MSR_WRITE);
+		break;
+	case VGA_OP_SCREEN_OFF:
+		outb(SR01, VGA_SR_INDEX);
+		tmp = inb(VGA_SR_DATA);
+		tmp |= 1 << 5;
+		outb(tmp, VGA_SR_DATA);
+		break;
+	}
+
+	pci_write_config_word(dev->pdev, PCI_COMMAND, cmd);
+
+	i915_vga_bus_restore(dev);
+
+	return 0;
+}
+
+/* MMIO based VGA register access, pre-Gen5 only */
+static void i915_vga_execute_mmio(struct drm_device *dev, enum vga_op op)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u8 tmp;
+
+	switch (op) {
+	case VGA_OP_MEM_DISABLE:
+		tmp = I915_READ8(VGA_MSR_READ);
+		tmp &= ~VGA_MSR_MEM_EN;
+		I915_WRITE8(VGA_MSR_WRITE, tmp);
+		break;
+	case VGA_OP_MEM_ENABLE:
+		tmp = I915_READ8(VGA_MSR_READ);
+		tmp |= VGA_MSR_MEM_EN;
+		I915_WRITE8(VGA_MSR_WRITE, tmp);
+		break;
+	case VGA_OP_SCREEN_OFF:
+		I915_WRITE8(VGA_SR_INDEX, SR01);
+		tmp = I915_READ8(VGA_SR_DATA);
+		tmp |= 1 << 5;
+		I915_WRITE8(VGA_SR_DATA, tmp);
+		break;
 	}
 }
 
+static void i915_vga_execute(struct drm_device *dev, enum vga_op op)
+{
+	/* WaEnableVGAAccessThroughIOPort:ctg,elg,ilk,snb,ivb,vlv,hsw */
+	if (IS_G4X(dev) || INTEL_INFO(dev)->gen >= 5) {
+		struct i915_vga_op vga_op = {
+			.op = op,
+			.dev = dev,
+		};
+
+		stop_machine(i915_vga_stop_machine_cb, &vga_op, NULL);
+	} else {
+		i915_vga_execute_mmio(dev, op);
+	}
+}
+
+/* Disable the VGA plane that we never use */
+static void i915_disable_vga(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u32 vga_reg = i915_vgacntrl_reg(dev);
+	u16 gmch_ctrl;
+
+	/* already disabled? */
+	if (I915_READ(vga_reg) & VGA_DISP_DISABLE)
+		return;
+
+	pci_read_config_word(dev_priv->bridge_dev, INTEL_INFO(dev)->gen >= 6 ?
+			     SNB_GMCH_CTRL : INTEL_GMCH_CTRL, &gmch_ctrl);
+	if (gmch_ctrl & INTEL_GMCH_VGA_DISABLE)
+		DRM_ERROR("VGA plane enabled while VGA disabled via GMCH control\n");
+
+	i915_vga_execute(dev, VGA_OP_SCREEN_OFF);
+
+	udelay(300);
+
+	I915_WRITE(vga_reg, VGA_DISP_DISABLE);
+	POSTING_READ(vga_reg);
+}
+
 void intel_modeset_init_hw(struct drm_device *dev)
 {
 	intel_prepare_ddi(dev);
@@ -10553,7 +10698,6 @@ void i915_redisable_vga(struct drm_device *dev)
 	if (I915_READ(vga_reg) != VGA_DISP_DISABLE) {
 		DRM_DEBUG_KMS("Something enabled VGA plane, disabling it\n");
 		i915_disable_vga(dev);
-		i915_disable_vga_mem(dev);
 	}
 }
 
@@ -10755,7 +10899,7 @@ void intel_modeset_cleanup(struct drm_device *dev)
 
 	intel_disable_fbc(dev);
 
-	i915_enable_vga_mem(dev);
+	intel_modeset_vga_set_state(dev, true);
 
 	intel_disable_gt_powersave(dev);
 
@@ -10798,12 +10942,36 @@ int intel_modeset_vga_set_state(struct drm_device *dev, bool state)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u16 gmch_ctrl;
 
-	pci_read_config_word(dev_priv->bridge_dev, INTEL_GMCH_CTRL, &gmch_ctrl);
-	if (state)
-		gmch_ctrl &= ~INTEL_GMCH_VGA_DISABLE;
-	else
-		gmch_ctrl |= INTEL_GMCH_VGA_DISABLE;
-	pci_write_config_word(dev_priv->bridge_dev, INTEL_GMCH_CTRL, gmch_ctrl);
+	/* Is VGA totally disabled for the IGD? */
+	pci_read_config_word(dev_priv->bridge_dev, INTEL_INFO(dev)->gen >= 6 ?
+			     SNB_GMCH_CTRL : INTEL_GMCH_CTRL, &gmch_ctrl);
+	if (gmch_ctrl & INTEL_GMCH_VGA_DISABLE)
+		return 0;
+
+	if (state) {
+		i915_vga_execute(dev, VGA_OP_MEM_ENABLE);
+
+		/*
+		 * Leave PCI_COMMAND_IO alone for now. vgaarb
+		 * should re-enable it if and when needed.
+		 */
+
+		vga_set_legacy_decoding(dev->pdev, VGA_RSRC_LEGACY_IO |
+						   VGA_RSRC_LEGACY_MEM |
+						   VGA_RSRC_NORMAL_IO |
+						   VGA_RSRC_NORMAL_MEM);
+	} else {
+		u16 cmd;
+
+		i915_vga_execute(dev, VGA_OP_MEM_DISABLE);
+

drivers/gpu/drm/i915/intel_display.c int intel_modeset_vga_set_state(struct drm_device *dev, bool state) · line 10977

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +7 -0
+		pci_read_config_word(dev->pdev, PCI_COMMAND, &cmd);
+		cmd &= ~PCI_COMMAND_IO;
+		pci_write_config_word(dev->pdev, PCI_COMMAND, cmd);
+
+		vga_set_legacy_decoding(dev->pdev, VGA_RSRC_NORMAL_MEM);
+	}
+
 	return 0;
 }
 

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#15 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 24 lines · Alex Williamson
Quoted 23 lines · On Mon, 2013-09-30 at 17:08 +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

Quoted 6 lines
Quoted 4 lines

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

What?! Why would we not simply wait for the arbiter lock?

Well, there are the X problems which I really don't want to attempt solving.

Also the arbiter looks a lot like deadlock heaven to me.

What if the other guy doesn't release the arbiter lock in a timely fashion? It could be some userspace process that's stopped inside gdb or something.

What if we're doing the restore thing in intel_lid_notify() and we've already locked the modeset locks and are now waiting for the arbiter lock, but the other guy who is holding the arbiter lock is doing a modeset ioctl at the same time and gets stuck waiting for a modeset lock?

I guess we might be able to solve those problems by killing the userspace client after a while. But I'd rather just hide and go code up something more productive ;)

Signature · 2 lines
Ville Syrjälä
Intel OTC
#16 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 47 lines · Ville Syrjälä
Quoted 24 lines · Alex Williamson
Quoted 23 lines · On Mon, 2013-09-30 at 17:08 +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

Quoted 6 lines
Quoted 4 lines

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

What?! Why would we not simply wait for the arbiter lock?

Well, there are the X problems which I really don't want to attempt solving.

Also the arbiter looks a lot like deadlock heaven to me.

What if the other guy doesn't release the arbiter lock in a timely fashion? It could be some userspace process that's stopped inside gdb or something.

What if we're doing the restore thing in intel_lid_notify() and we've already locked the modeset locks and are now waiting for the arbiter lock, but the other guy who is holding the arbiter lock is doing a modeset ioctl at the same time and gets stuck waiting for a modeset lock?

Quoted 3 lines

I guess we might be able to solve those problems by killing the userspace client after a while. But I'd rather just hide and go code up something more productive ;)

So in summary, ignore the infrastructure intended to solve this problem because it's hard and may have corner cases and instead stop the entire machine so we can be sure our access is exclusive. If this is the solution then VGA arbiter has failed us. Thanks,

Alex

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#17 lore raw

Re: [PATCH v2 1/1] drm/i915: Fix VGA handling using stop_machine() or mmio

Quoted 48 lines · Alex Williamson
Quoted 47 lines · Ville Syrjälä
Quoted 24 lines · Alex Williamson
Quoted 23 lines · On Mon, 2013-09-30 at 17:08 +0300, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We have several problems with out VGA handling:
- We try to use the GMCH control VGA disable bit even though it may be locked
- If we manage to disable VGA throuh GMCH control, we're no longer able to correctly disable the VGA plane
- Taking part in the VGA arbitration is too expensive for X [1]

So let's treat the GMCH control VGA disable bit as read-only and leave it for the BIOS to set, as it was intended. To disable VGA we will use the VGA misc register, and to disable VGA IO we will disable IO space completely via the PCI command register.

But we still need VGA register access during resume (and possibly during lid event on insane BIOSen) to disable the VGA plane. Also we need to re-disable VGA memory decode via the VGA misc register on resume.

Luckily up to gen4, VGA registers can be accessed through MMIO. Unfortunately from gen5 onwards only the legacy VGA IO port range works. So on gen5+ we still need IO space to be enabled during those few special moments when we need to access VGA registers.

Quoted 6 lines
Quoted 4 lines

We still want to opt out of VGA arbitration on gen5+, so we have keep IO space disabled most of the time. And when we do need to poke at VGA registers, we enable IO space briefly while no one is looking. To guarantee that no one is looking we will use stop_machine().

What?! Why would we not simply wait for the arbiter lock?

Well, there are the X problems which I really don't want to attempt solving.

Also the arbiter looks a lot like deadlock heaven to me.

What if the other guy doesn't release the arbiter lock in a timely fashion? It could be some userspace process that's stopped inside gdb or something.

What if we're doing the restore thing in intel_lid_notify() and we've already locked the modeset locks and are now waiting for the arbiter lock, but the other guy who is holding the arbiter lock is doing a modeset ioctl at the same time and gets stuck waiting for a modeset lock?

Quoted 8 lines
Quoted 3 lines

I guess we might be able to solve those problems by killing the userspace client after a while. But I'd rather just hide and go code up something more productive ;)

So in summary, ignore the infrastructure intended to solve this problem because it's hard and may have corner cases and instead stop the entire machine so we can be sure our access is exclusive. If this is the solution then VGA arbiter has failed us. Thanks,

Well, if you want to fix the vga arbiter (assuming it's possible) feel free to send patches. I can even give a vague promise of reviewing them. But I guess we need to fix this ASAP or face the wrath of the angry mob who lost their dri capabilities.

The other option is to move the stop_machine infrastructure inside vgaarb and then we can say we're using vgaarb as *deity* intended ;)

Signature · 2 lines
Ville Syrjälä
Intel OTC