Nouveau - Jul 2021 - [PATCH] drm/nouveau: don't touch has_pr3 for likely-non-NVIDIA device

The call site of nouveau_dsm_pci_probe() uses single set of output
variables for all invocations. So, we must not write anything to them
until we think this is an NVIDIA device of interest. Otherwise, if we
are called with another device after the NVIDIA device, we'll clober the
result of the NVIDIA device.

In this case, if the other device doesn't have _PR3 resources, the
detection later would miss the presence of power resource support, and
the rest of the code will keep using Optimus DSM, breaking power
management for that machine. In particular, this fixes power management
of the NVIDIA card in Lenovo Legion 5-15ARH05... well, at least
partially. New error shows up, but this patch is correct in itself
anyway.

As a bonus, we'll also stop preventing _PR3 usage from the bridge for
unrelated devices, which is always nice, I guess.

As noted in commit ccfc2d5cdb024 ("drm/nouveau: Use generic helper to
check _PR3 presence"), care is taken to leave the _PR3 detection outside
of the optimus_func condition.

https://gitlab.freedesktop.org/drm/nouveau/-/issues/79

Fixes: ccfc2d5cdb024 ("drm/nouveau: Use generic helper to check _PR3
presence")
Signed-off-by: Ratchanan Srirattanamet <peathot at hotmail.com>
---

Hello,

This is my first time submitting a Linux patch. I've done a
number of PR/MR workflows on GitHub or GitLab, but never done any
email-oriented development. So, please excuse me if I'm doing something
incorrectly.

 drivers/gpu/drm/nouveau/nouveau_acpi.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nouveau_acpi.c
b/drivers/gpu/drm/nouveau/nouveau_acpi.c
index 7c15f6448428..c88bda3ac820 100644
--- a/drivers/gpu/drm/nouveau/nouveau_acpi.c
+++ b/drivers/gpu/drm/nouveau/nouveau_acpi.c
@@ -220,15 +220,6 @@ static void nouveau_dsm_pci_probe(struct pci_dev *pdev,
acpi_handle *dhandle_out
 	int optimus_funcs;
 	struct pci_dev *parent_pdev;
 
-	*has_pr3 = false;
-	parent_pdev = pci_upstream_bridge(pdev);
-	if (parent_pdev) {
-		if (parent_pdev->bridge_d3)
-			*has_pr3 = pci_pr3_present(parent_pdev);
-		else
-			pci_d3cold_disable(pdev);
-	}
-
 	dhandle = ACPI_HANDLE(&pdev->dev);
 	if (!dhandle)
 		return;
@@ -249,6 +240,15 @@ static void nouveau_dsm_pci_probe(struct pci_dev *pdev,
acpi_handle *dhandle_out
 	*has_opt = !!optimus_funcs;
 	*has_opt_flags = optimus_funcs & (1 << NOUVEAU_DSM_OPTIMUS_FLAGS);
 
+	*has_pr3 = false;
+	parent_pdev = pci_upstream_bridge(pdev);
+	if (parent_pdev) {
+		if (parent_pdev->bridge_d3)
+			*has_pr3 = pci_pr3_present(parent_pdev);
+		else
+			pci_d3cold_disable(pdev);
+	}
+
 	if (optimus_funcs) {
 		uint32_t result;
 		nouveau_optimus_dsm(dhandle, NOUVEAU_DSM_OPTIMUS_CAPS, 0,
-- 
2.25.1

hey, thanks for the patch. But I am a bit confused on why that patch
actually helps. It should only be called for nvidia GPUs, but are we
ending up checking it for AMD GPUs as well?

Mind posting the output of lspci -tvnn?

On Thu, Jul 22, 2021 at 5:10 AM Ratchanan Srirattanamet
<peathot at hotmail.com> wrote:
>
> The call site of nouveau_dsm_pci_probe() uses single set of output
> variables for all invocations. So, we must not write anything to them
> until we think this is an NVIDIA device of interest. Otherwise, if we
> are called with another device after the NVIDIA device, we'll clober
the
> result of the NVIDIA device.
>
> In this case, if the other device doesn't have _PR3 resources, the
> detection later would miss the presence of power resource support, and
> the rest of the code will keep using Optimus DSM, breaking power
> management for that machine. In particular, this fixes power management
> of the NVIDIA card in Lenovo Legion 5-15ARH05... well, at least
> partially. New error shows up, but this patch is correct in itself
> anyway.
>
> As a bonus, we'll also stop preventing _PR3 usage from the bridge for
> unrelated devices, which is always nice, I guess.
>
> As noted in commit ccfc2d5cdb024 ("drm/nouveau: Use generic helper to
> check _PR3 presence"), care is taken to leave the _PR3 detection
outside
> of the optimus_func condition.
>
> https://gitlab.freedesktop.org/drm/nouveau/-/issues/79
>
> Fixes: ccfc2d5cdb024 ("drm/nouveau: Use generic helper to check _PR3
> presence")
> Signed-off-by: Ratchanan Srirattanamet <peathot at hotmail.com>
> ---
>
> Hello,
>
> This is my first time submitting a Linux patch. I've done a
> number of PR/MR workflows on GitHub or GitLab, but never done any
> email-oriented development. So, please excuse me if I'm doing something
> incorrectly.
>
>  drivers/gpu/drm/nouveau/nouveau_acpi.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_acpi.c
b/drivers/gpu/drm/nouveau/nouveau_acpi.c
> index 7c15f6448428..c88bda3ac820 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_acpi.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_acpi.c
> @@ -220,15 +220,6 @@ static void nouveau_dsm_pci_probe(struct pci_dev
*pdev, acpi_handle *dhandle_out
>         int optimus_funcs;
>         struct pci_dev *parent_pdev;
>
> -       *has_pr3 = false;
> -       parent_pdev = pci_upstream_bridge(pdev);
> -       if (parent_pdev) {
> -               if (parent_pdev->bridge_d3)
> -                       *has_pr3 = pci_pr3_present(parent_pdev);
> -               else
> -                       pci_d3cold_disable(pdev);
> -       }
> -
>         dhandle = ACPI_HANDLE(&pdev->dev);
>         if (!dhandle)
>                 return;
> @@ -249,6 +240,15 @@ static void nouveau_dsm_pci_probe(struct pci_dev
*pdev, acpi_handle *dhandle_out
>         *has_opt = !!optimus_funcs;
>         *has_opt_flags = optimus_funcs & (1 <<
NOUVEAU_DSM_OPTIMUS_FLAGS);
>
> +       *has_pr3 = false;
> +       parent_pdev = pci_upstream_bridge(pdev);
> +       if (parent_pdev) {
> +               if (parent_pdev->bridge_d3)
> +                       *has_pr3 = pci_pr3_present(parent_pdev);
> +               else
> +                       pci_d3cold_disable(pdev);
> +       }
> +
>         if (optimus_funcs) {
>                 uint32_t result;
>                 nouveau_optimus_dsm(dhandle, NOUVEAU_DSM_OPTIMUS_CAPS, 0,
> --
> 2.25.1
>
> _______________________________________________
> Nouveau mailing list
> Nouveau at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/nouveau
>

On Thu, Jul 22, 2021 at 5:10 AM Ratchanan Srirattanamet
<peathot at hotmail.com> wrote:
>
> The call site of nouveau_dsm_pci_probe() uses single set of output
> variables for all invocations. So, we must not write anything to them
> until we think this is an NVIDIA device of interest. Otherwise, if we
> are called with another device after the NVIDIA device, we'll clober
the
> result of the NVIDIA device.
>
> In this case, if the other device doesn't have _PR3 resources, the
> detection later would miss the presence of power resource support, and
> the rest of the code will keep using Optimus DSM, breaking power
> management for that machine. In particular, this fixes power management
> of the NVIDIA card in Lenovo Legion 5-15ARH05... well, at least
> partially. New error shows up, but this patch is correct in itself
> anyway.
>
> As a bonus, we'll also stop preventing _PR3 usage from the bridge for
> unrelated devices, which is always nice, I guess.
>
> As noted in commit ccfc2d5cdb024 ("drm/nouveau: Use generic helper to
> check _PR3 presence"), care is taken to leave the _PR3 detection
outside
> of the optimus_func condition.
>
> https://gitlab.freedesktop.org/drm/nouveau/-/issues/79
>
> Fixes: ccfc2d5cdb024 ("drm/nouveau: Use generic helper to check _PR3
> presence")
> Signed-off-by: Ratchanan Srirattanamet <peathot at hotmail.com>
> ---
>
> Hello,
>
> This is my first time submitting a Linux patch. I've done a
> number of PR/MR workflows on GitHub or GitLab, but never done any
> email-oriented development. So, please excuse me if I'm doing something
> incorrectly.
>
>  drivers/gpu/drm/nouveau/nouveau_acpi.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_acpi.c
b/drivers/gpu/drm/nouveau/nouveau_acpi.c
> index 7c15f6448428..c88bda3ac820 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_acpi.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_acpi.c
> @@ -220,15 +220,6 @@ static void nouveau_dsm_pci_probe(struct pci_dev
*pdev, acpi_handle *dhandle_out
>         int optimus_funcs;
>         struct pci_dev *parent_pdev;
>
> -       *has_pr3 = false;
> -       parent_pdev = pci_upstream_bridge(pdev);
> -       if (parent_pdev) {
> -               if (parent_pdev->bridge_d3)
> -                       *has_pr3 = pci_pr3_present(parent_pdev);
> -               else
> -                       pci_d3cold_disable(pdev);
> -       }
> -
>         dhandle = ACPI_HANDLE(&pdev->dev);
>         if (!dhandle)
>                 return;
> @@ -249,6 +240,15 @@ static void nouveau_dsm_pci_probe(struct pci_dev
*pdev, acpi_handle *dhandle_out
>         *has_opt = !!optimus_funcs;
>         *has_opt_flags = optimus_funcs & (1 <<
NOUVEAU_DSM_OPTIMUS_FLAGS);
>
> +       *has_pr3 = false;
> +       parent_pdev = pci_upstream_bridge(pdev);
> +       if (parent_pdev) {
> +               if (parent_pdev->bridge_d3)
> +                       *has_pr3 = pci_pr3_present(parent_pdev);
> +               else
> +                       pci_d3cold_disable(pdev);
> +       }
> +
given that we call this code for all GPUs, I _think_ it is better to
check for GPU Vendors or check if the GPU we touch is the secondary
one. We also have systems with two Nvidia GPUs. But I don't know how
the detection worked there and if that's fine in the end... I might
have to investigate this on all laptops with various hybrid GPU types,
but... mhh. Maybe checking for "this is a secondary GPU" is good
enough.
>         if (optimus_funcs) {
>                 uint32_t result;
>                 nouveau_optimus_dsm(dhandle, NOUVEAU_DSM_OPTIMUS_CAPS, 0,
> --
> 2.25.1
>
> _______________________________________________
> Nouveau mailing list
> Nouveau at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/nouveau
>