intel-gfx.lists.freedesktop.org archive mirror
 help / color / mirror / Atom feed
From: Jani Nikula <jani.nikula@linux.intel.com>
To: Ville Syrjala <ville.syrjala@linux.intel.com>,
	intel-gfx@lists.freedesktop.org
Cc: intel-xe@lists.freedesktop.org
Subject: Re: [PATCH 08/19] drm/i915/vga: Stop trying to use GMCH_CTRL for VGA decode control
Date: Tue, 09 Dec 2025 12:39:31 +0200	[thread overview]
Message-ID: <d56eff8d61d6560dca0357923d1e6f406c854926@intel.com> (raw)
In-Reply-To: <20251208182637.334-9-ville.syrjala@linux.intel.com>

On Mon, 08 Dec 2025, Ville Syrjala <ville.syrjala@linux.intel.com> wrote:
> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
>
> intel_gmch_vga_set_state() is a complete lie on ILK+ because
> the GMCH_CTRL register is locked and can't actually be written.
> But we still need to remove the iGPU from the VGA arbitration
> on iGPU+dGPU systems, or else Xorg performace will tank due

*performance

> to the constant VGA arbiter accesess.

*accesses

>
> For VGA memory decode we can't turn off the PCI_COMMAND
> memory deocde as that would disable even normal MMIO.

*decode

> Instead we can disable just the VGA memory decode via
> the VGA MSR register. And we can do that just once
> when disablign the VGA plane. That way we don't have

*disabling

> to touch VGA registers anywhere else.
>
> We can also inform the arbiter that we're no longer decding

*decoding

> VGA memory. This will stop the arbitter from disabling all

*arbiter

> memory decode for the iGPU via PCI_COMMAND (and thus breaking
> everything) whenever some other GPU wants to own the VGA memory
> accesses.
>
> For IO we can disable all IO decode via the PCI_COMMAND
> register, except around the few VGA register accesses that
> we need to do in intel_vga_disable(). Unfortunately we can't
> disable IO decode permanently as it makes some laptops (eg.
> Dell Latitude E5400) hang during reboot/shutdown. One option
> would be to re-enable IO decode from the poweroff hooks, but
> that won't help the sysrq emergency reboot/shutdown since it
> won't call said hooks. So let's try to keep IO decode in its
> original setting unless we really need to disable it to
> exclude the GPU from VGA arbitration.
>
> I suppose we could keep frobbing GMCH_CTRL on pre-ILK, but
> it seems better to not do it since it has other side effects
> such as changing the class code of the PCI device.
>
> For discrete GPUs we'll rely on the bridge control instead.
>
> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

Acked-by: Jani Nikula <jani.nikula@intel.com>

> ---
>  drivers/gpu/drm/i915/display/intel_vga.c | 93 +++++++++++++++---------
>  1 file changed, 57 insertions(+), 36 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/display/intel_vga.c b/drivers/gpu/drm/i915/display/intel_vga.c
> index a2a1c33d053e..f2f7d396c556 100644
> --- a/drivers/gpu/drm/i915/display/intel_vga.c
> +++ b/drivers/gpu/drm/i915/display/intel_vga.c
> @@ -71,6 +71,19 @@ static bool intel_pci_set_io_decode(struct pci_dev *pdev, bool enable)
>  	return old & PCI_COMMAND_IO;
>  }
>  
> +static bool intel_pci_bridge_set_vga(struct pci_dev *pdev, bool enable)
> +{
> +	u16 old = 0, ctl;
> +
> +	pci_read_config_word(pdev->bus->self, PCI_BRIDGE_CONTROL, &old);
> +	ctl = old & ~PCI_BRIDGE_CTL_VGA;
> +	if (enable)
> +		ctl |= PCI_BRIDGE_CTL_VGA;
> +	pci_write_config_word(pdev->bus->self, PCI_BRIDGE_CONTROL, ctl);
> +
> +	return old & PCI_BRIDGE_CTL_VGA;
> +}
> +
>  static bool intel_vga_get(struct intel_display *display)
>  {
>  	struct pci_dev *pdev = to_pci_dev(display->drm->dev);
> @@ -108,6 +121,7 @@ static void intel_vga_put(struct intel_display *display, bool io_decode)
>  /* Disable the VGA plane that we never use */
>  void intel_vga_disable(struct intel_display *display)
>  {
> +	struct pci_dev *pdev = to_pci_dev(display->drm->dev);
>  	i915_reg_t vga_reg = intel_vga_cntrl_reg(display);
>  	bool io_decode;
>  	u8 msr, sr1;
> @@ -160,6 +174,12 @@ void intel_vga_disable(struct intel_display *display)
>  	outb(sr1 | VGA_SR01_SCREEN_OFF, VGA_SEQ_D);
>  
>  	msr = inb(VGA_MIS_R);
> +	/*
> +	 * Always disable VGA memory decode for iGPU so that
> +	 * intel_vga_set_decode() doesn't need to access VGA registers.
> +	 * VGA_MIS_ENB_MEM_ACCESS=0 is also the reset value.
> +	 */
> +	msr &= ~VGA_MIS_ENB_MEM_ACCESS;
>  	/*
>  	 * VGA_MIS_COLOR controls both GPU level and display engine level
>  	 * MDA vs. CGA decode logic. But when the register gets reset
> @@ -177,6 +197,14 @@ void intel_vga_disable(struct intel_display *display)
>  
>  	intel_vga_put(display, io_decode);
>  
> +	/*
> +	 * Inform the arbiter about VGA memory decode being disabled so
> +	 * that it doesn't disable all memory decode for the iGPU when
> +	 * targeting another GPU.
> +	 */
> +	if (!display->platform.dgfx)
> +		vga_set_legacy_decoding(pdev, VGA_RSRC_LEGACY_IO);
> +
>  	udelay(300);
>  
>  reset_vgacntr:
> @@ -184,45 +212,38 @@ void intel_vga_disable(struct intel_display *display)
>  	intel_de_posting_read(display, vga_reg);
>  }
>  
> -static int intel_gmch_vga_set_state(struct intel_display *display, bool enable_decode)
> -{
> -	struct pci_dev *pdev = to_pci_dev(display->drm->dev);
> -	u16 gmch_ctrl;
> -
> -	if (pci_bus_read_config_word(pdev->bus, PCI_DEVFN(0, 0),
> -				     intel_gmch_ctrl_reg(display), &gmch_ctrl)) {
> -		drm_err(display->drm, "failed to read control word\n");
> -		return -EIO;
> -	}
> -
> -	if (!!(gmch_ctrl & INTEL_GMCH_VGA_DISABLE) == !enable_decode)
> -		return 0;
> -
> -	if (enable_decode)
> -		gmch_ctrl &= ~INTEL_GMCH_VGA_DISABLE;
> -	else
> -		gmch_ctrl |= INTEL_GMCH_VGA_DISABLE;
> -
> -	if (pci_bus_write_config_word(pdev->bus, PCI_DEVFN(0, 0),
> -				      intel_gmch_ctrl_reg(display), gmch_ctrl)) {
> -		drm_err(display->drm, "failed to write control word\n");
> -		return -EIO;
> -	}
> -
> -	return 0;
> -}
> -
> -static unsigned int intel_gmch_vga_set_decode(struct pci_dev *pdev, bool enable_decode)
> +static unsigned int intel_vga_set_decode(struct pci_dev *pdev, bool enable_decode)
>  {
>  	struct intel_display *display = to_intel_display(pdev);
> +	unsigned int decodes = VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
>  
> -	intel_gmch_vga_set_state(display, enable_decode);
> +	drm_dbg_kms(display->drm, "%s VGA decode due to VGA arbitration\n",
> +		    str_enable_disable(enable_decode));
>  
> -	if (enable_decode)
> -		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;
> +	/*
> +	 * Can't use GMCH_CTRL INTEL_GMCH_VGA_DISABLE to disable VGA
> +	 * decode on ILK+ since the register is locked. Instead
> +	 * intel_disable_vga() will disable VGA memory decode for the
> +	 * iGPU, and here we just need to take care of the IO decode.
> +	 * For discrete GPUs we rely on the bridge VGA control.
> +	 *
> +	 * We can't disable IO decode already in intel_vga_disable()
> +	 * because at least some laptops (eg. CTG Dell Latitude E5400)
> +	 * will hang during reboot/shutfown with IO decode disabled.
> +	 */
> +	if (display->platform.dgfx) {
> +		if (!enable_decode)
> +			intel_pci_bridge_set_vga(pdev, false);
> +		else
> +			decodes |= VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM;
> +	} else {
> +		if (!enable_decode)
> +			intel_pci_set_io_decode(pdev, false);
> +		else
> +			decodes |= VGA_RSRC_LEGACY_IO;
> +	}
> +
> +	return decodes;
>  }
>  
>  void intel_vga_register(struct intel_display *display)
> @@ -239,7 +260,7 @@ void intel_vga_register(struct intel_display *display)
>  	 * then we do not take part in VGA arbitration and the
>  	 * vga_client_register() fails with -ENODEV.
>  	 */
> -	ret = vga_client_register(pdev, intel_gmch_vga_set_decode);
> +	ret = vga_client_register(pdev, intel_vga_set_decode);
>  	drm_WARN_ON(display->drm, ret && ret != -ENODEV);
>  }

-- 
Jani Nikula, Intel

  reply	other threads:[~2025-12-09 10:39 UTC|newest]

Thread overview: 51+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-12-08 18:26 [PATCH 00/19] drm/i915/vga: Try to sort out the VGA decode mess Ville Syrjala
2025-12-08 18:26 ` [PATCH 01/19] drm/i915/vga: Register vgaarb client later Ville Syrjala
2025-12-09 10:23   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 02/19] drm/i915/vga: Get rid of intel_vga_reset_io_mem() Ville Syrjala
2025-12-09 10:26   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 03/19] drm/i915/power: Remove i915_power_well_desc::has_vga Ville Syrjala
2025-12-09 10:27   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 04/19] drm/i915/vga: Extract intel_gmch_ctrl_reg() Ville Syrjala
2025-12-09 10:28   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 05/19] drm/i915/vga: Don't touch VGA registers if VGA decode is fully disabled Ville Syrjala
2025-12-09 10:29   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 06/19] drm/i915/vga: Clean up VGA registers even if VGA plane is disabled Ville Syrjala
2025-12-09 10:32   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 07/19] drm/i915/vga: Avoid VGA arbiter during intel_vga_disable() for iGPUs Ville Syrjala
2025-12-09 10:35   ` Jani Nikula
2025-12-09 12:17     ` Ville Syrjälä
2025-12-08 18:26 ` [PATCH 08/19] drm/i915/vga: Stop trying to use GMCH_CTRL for VGA decode control Ville Syrjala
2025-12-09 10:39   ` Jani Nikula [this message]
2025-12-08 18:26 ` [PATCH 09/19] drm/i915/vga: Assert that VGA register accesses are going to the right GPU Ville Syrjala
2025-12-09 10:40   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 10/19] drm/i915/de: Simplify intel_de_read8() Ville Syrjala
2025-12-09 10:47   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 11/19] drm/i915/de: Add intel_de_write8() Ville Syrjala
2025-12-09 10:49   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 12/19] drm/i915/vga: Introduce intel_vga_{read,write}() Ville Syrjala
2025-12-09 10:52   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 13/19] drm/i915/vga: Use MMIO for VGA registers on pre-g4x Ville Syrjala
2025-12-09 10:53   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 14/19] video/vga: Add VGA_IS0_R Ville Syrjala
2025-12-08 21:07   ` kernel test robot
2025-12-08 21:18   ` kernel test robot
2025-12-08 22:22   ` kernel test robot
2025-12-09  7:55   ` [PATCH v2 " Ville Syrjala
2025-12-09 10:55     ` Jani Nikula
2025-12-10 14:13   ` [PATCH " kernel test robot
2025-12-10 14:24   ` kernel test robot
2025-12-08 18:26 ` [PATCH 15/19] drm/i915/crt: Use IS0_R instead of VGA_MIS_W Ville Syrjala
2025-12-09 10:56   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 16/19] drm/i915/crt: Extract intel_crt_sense_above_threshold() Ville Syrjala
2025-12-09 10:57   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 17/19] drm/i915: Get rid of the INTEL_GMCH_CTRL alias Ville Syrjala
2025-12-09 10:58   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 18/19] drm/i915: Clean up PCI config space reg defines Ville Syrjala
2025-12-09 11:00   ` Jani Nikula
2025-12-09 11:01   ` Jani Nikula
2025-12-08 18:26 ` [PATCH 19/19] drm/i915: Document the GMCH_CTRL register a bit Ville Syrjala
2025-12-09 11:03   ` Jani Nikula
2025-12-08 19:11 ` ✗ Fi.CI.BUILD: failure for drm/i915/vga: Try to sort out the VGA decode mess Patchwork
2025-12-09 11:31 ` ✗ i915.CI.BAT: failure for drm/i915/vga: Try to sort out the VGA decode mess (rev2) Patchwork
2025-12-10 19:14 ` ✓ i915.CI.BAT: success for drm/i915/vga: Try to sort out the VGA decode mess (rev3) Patchwork
2025-12-11  3:23 ` ✓ i915.CI.Full: " Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=d56eff8d61d6560dca0357923d1e6f406c854926@intel.com \
    --to=jani.nikula@linux.intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=ville.syrjala@linux.intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).