Re: [Nouveau] [PATCH] PCI: stop spamming info in quirk_nvidia_hda
On Thu, Mar 16, 2023 at 8:57 PM Bjorn Helgaas wrote:
>
> On Thu, Mar 16, 2023 at 03:31:22PM +0100, Karol Herbst wrote:
> > Users kept complaining about those messages and it's a little spammy on
> > prime systems so turn it into a debug print.
>
> What is a "prime system"?
>
Laptops with a iGPU + Nvidia Setup. That tech is usually marketed as
"Nvidia Optimus" and people in the open source world made "prime" out
of that, mostly in the context of "prime offloading".
> I'm a little surprised that users would really care about the message.
> But I do see comments like these:
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1836308/comments/15
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2002206
> that suggest the message happens frequently, maybe if we're resuming
> the controller after runtime suspend?
>
Yes, that happens every time the discrete Nvidia GPU gets runtime resumed.
> Maybe this should be a pci_info_once() sort of thing? I think there's
> some value in knowing that we're changing the BIOS configuration
> outside the purview of a driver, since I assume BIOS had some reason
> for hiding the HDA controller.
>
fair point. Most of the bugs happen on the first runtime resume
already, so if that one is good, it's unlikely the system will hit a
bug later on.
> > Cc: Bjorn Helgaas
> > Cc: Lukas Wunner
> > Cc: linux-...@vger.kernel.org
> > Cc: nouveau@lists.freedesktop.org
> > Fixes: b516ea586d71 ("PCI: Enable NVIDIA HDA controllers")
> > Signed-off-by: Karol Herbst
> > ---
> > drivers/pci/quirks.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> > index 44cab813bf951..b10c77bbe4716 100644
> > --- a/drivers/pci/quirks.c
> > +++ b/drivers/pci/quirks.c
> > @@ -5549,7 +5549,7 @@ static void quirk_nvidia_hda(struct pci_dev *gpu)
> > if (val & BIT(25))
> > return;
> >
> > - pci_info(gpu, "Enabling HDA controller\n");
> > + pci_dbg(gpu, "Enabling HDA controller\n");
> > pci_write_config_dword(gpu, 0x488, val | BIT(25));
> >
> > /* The GPU becomes a multi-function device when the HDA is enabled */
> > --
> > 2.39.2
> >
>
Re: [Nouveau] [PATCH] PCI: stop spamming info in quirk_nvidia_hda
On Thu, Mar 16, 2023 at 03:31:22PM +0100, Karol Herbst wrote:
> Users kept complaining about those messages and it's a little spammy on
> prime systems so turn it into a debug print.
What is a "prime system"?
I'm a little surprised that users would really care about the message.
But I do see comments like these:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1836308/comments/15
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2002206
that suggest the message happens frequently, maybe if we're resuming
the controller after runtime suspend?
Maybe this should be a pci_info_once() sort of thing? I think there's
some value in knowing that we're changing the BIOS configuration
outside the purview of a driver, since I assume BIOS had some reason
for hiding the HDA controller.
> Cc: Bjorn Helgaas
> Cc: Lukas Wunner
> Cc: linux-...@vger.kernel.org
> Cc: nouveau@lists.freedesktop.org
> Fixes: b516ea586d71 ("PCI: Enable NVIDIA HDA controllers")
> Signed-off-by: Karol Herbst
> ---
> drivers/pci/quirks.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> index 44cab813bf951..b10c77bbe4716 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -5549,7 +5549,7 @@ static void quirk_nvidia_hda(struct pci_dev *gpu)
> if (val & BIT(25))
> return;
>
> - pci_info(gpu, "Enabling HDA controller\n");
> + pci_dbg(gpu, "Enabling HDA controller\n");
> pci_write_config_dword(gpu, 0x488, val | BIT(25));
>
> /* The GPU becomes a multi-function device when the HDA is enabled */
> --
> 2.39.2
>
Re: [Nouveau] [PATCH] PCI: stop spamming info in quirk_nvidia_hda
On Thu, Mar 16, 2023 at 03:31:22PM +0100, Karol Herbst wrote:
> Users kept complaining about those messages and it's a little spammy on
> prime systems so turn it into a debug print.
>
> Cc: Bjorn Helgaas
> Cc: Lukas Wunner
> Cc: linux-...@vger.kernel.org
> Cc: nouveau@lists.freedesktop.org
> Fixes: b516ea586d71 ("PCI: Enable NVIDIA HDA controllers")
> Signed-off-by: Karol Herbst
Could you provide Link: tags to bugzillas or mailing list messages
of those complaints so that the sentence sounds a little less like
hand-waving?
The point of the message is that if users see adverse effects as a
result of exposing the hidden HDA controller, they get a hint in
dmesg as to the cause. Hopefully the existence of such adverse
effects can be ruled out by now.
Thanks,
Lukas
> ---
> drivers/pci/quirks.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> index 44cab813bf951..b10c77bbe4716 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -5549,7 +5549,7 @@ static void quirk_nvidia_hda(struct pci_dev *gpu)
> if (val & BIT(25))
> return;
>
> - pci_info(gpu, "Enabling HDA controller\n");
> + pci_dbg(gpu, "Enabling HDA controller\n");
> pci_write_config_dword(gpu, 0x488, val | BIT(25));
>
> /* The GPU becomes a multi-function device when the HDA is enabled */
> --
> 2.39.2
[Nouveau] [PATCH] PCI: stop spamming info in quirk_nvidia_hda
Users kept complaining about those messages and it's a little spammy on
prime systems so turn it into a debug print.
Cc: Bjorn Helgaas
Cc: Lukas Wunner
Cc: linux-...@vger.kernel.org
Cc: nouveau@lists.freedesktop.org
Fixes: b516ea586d71 ("PCI: Enable NVIDIA HDA controllers")
Signed-off-by: Karol Herbst
---
drivers/pci/quirks.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
index 44cab813bf951..b10c77bbe4716 100644
--- a/drivers/pci/quirks.c
+++ b/drivers/pci/quirks.c
@@ -5549,7 +5549,7 @@ static void quirk_nvidia_hda(struct pci_dev *gpu)
if (val & BIT(25))
return;
- pci_info(gpu, "Enabling HDA controller\n");
+ pci_dbg(gpu, "Enabling HDA controller\n");
pci_write_config_dword(gpu, 0x488, val | BIT(25));
/* The GPU becomes a multi-function device when the HDA is enabled */
--
2.39.2
