Re: [Nouveau] [PATCH v2 3/9] PCI: drop `is_thunderbolt` attribute
Hi, On Sun, Feb 13, 2022 at 09:39:28AM +0100, Lukas Wunner wrote: > On Fri, Feb 11, 2022 at 12:23:51PM +0200, Mika Westerberg wrote: > > On Thu, Feb 10, 2022 at 04:43:23PM -0600, Mario Limonciello wrote: > > > @@ -2955,7 +2955,7 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge) > > > return true; > > > > > > /* Even the oldest 2010 Thunderbolt controller supports D3. */ > > > - if (bridge->is_thunderbolt) > > > + if (dev_is_removable(&bridge->dev)) > > > > For this, I'm not entirely sure this is what we want. The purpose of > > this check is to enable port power management of Apple systems with > > Intel Thunderbolt controller and therefore checking for "removable" here > > is kind of misleading IMHO. > [...] > > and then make a quirk in quirks.c that adds the software node property > > for the Apple systems? Or something along those lines. > > Honestly, that feels wrong to me. > > There are non-Apple products with Thunderbolt controllers, > e.g. Supermicro X10SAT was a Xeon board with Redwood Ridge > which was introduced in 2013. This was way before Microsoft > came up with the HotPlugSupportInD3 property. It was also way > before the 2015 BIOS cut-off date that we use to disable > power management on older boards. > > Still, we currently whitelist the Thunderbolt ports on that > board for D3 because we know it works. What if products like > this one use their own power management scheme and we'd cause > a power regression if we needlessly disable D3 for them now? All the non-Apple Thunderbolt products before "HotPlugSupportInD3" use ACPI "assisted" hotplug which means all the PM is done in the BIOS. Essentially it means the controller is only present if there is anything connected and in that case it is always in D0. Unplugging the device makes the controller to be hot-removed (ACPI hotplug) too and that's the only way early Thunderbolt used to save energy.
Re: [Nouveau] [PATCH v2 3/9] PCI: drop `is_thunderbolt` attribute
On Fri, Feb 11, 2022 at 12:23:51PM +0200, Mika Westerberg wrote: > On Thu, Feb 10, 2022 at 04:43:23PM -0600, Mario Limonciello wrote: > > @@ -2955,7 +2955,7 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge) > > return true; > > > > /* Even the oldest 2010 Thunderbolt controller supports D3. */ > > - if (bridge->is_thunderbolt) > > + if (dev_is_removable(&bridge->dev)) > > For this, I'm not entirely sure this is what we want. The purpose of > this check is to enable port power management of Apple systems with > Intel Thunderbolt controller and therefore checking for "removable" here > is kind of misleading IMHO. [...] > and then make a quirk in quirks.c that adds the software node property > for the Apple systems? Or something along those lines. Honestly, that feels wrong to me. There are non-Apple products with Thunderbolt controllers, e.g. Supermicro X10SAT was a Xeon board with Redwood Ridge which was introduced in 2013. This was way before Microsoft came up with the HotPlugSupportInD3 property. It was also way before the 2015 BIOS cut-off date that we use to disable power management on older boards. Still, we currently whitelist the Thunderbolt ports on that board for D3 because we know it works. What if products like this one use their own power management scheme and we'd cause a power regression if we needlessly disable D3 for them now? Thanks, Lukas
Re: [Nouveau] [PATCH v2 3/9] PCI: drop `is_thunderbolt` attribute
Hi Mario,
On Thu, Feb 10, 2022 at 04:43:23PM -0600, Mario Limonciello wrote:
> The `is_thunderbolt` attribute is currently a dumping ground for a
> variety of things.
>
> Instead use the driver core removable attribute to indicate the
> detail a device is attached to a thunderbolt or USB4 chain.
>
> Signed-off-by: Mario Limonciello
> ---
> drivers/pci/pci.c | 2 +-
> drivers/pci/probe.c | 20 +++-
> drivers/platform/x86/apple-gmux.c | 2 +-
> include/linux/pci.h | 5 ++---
> 4 files changed, 11 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 9ecce435fb3f..1264984d5e6d 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -2955,7 +2955,7 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge)
> return true;
>
> /* Even the oldest 2010 Thunderbolt controller supports D3. */
> - if (bridge->is_thunderbolt)
> + if (dev_is_removable(&bridge->dev))
For this, I'm not entirely sure this is what we want. The purpose of
this check is to enable port power management of Apple systems with
Intel Thunderbolt controller and therefore checking for "removable" here
is kind of misleading IMHO.
I wonder if we could instead remove the check completely here and rely
on the below:
if (platform_pci_bridge_d3(bridge))
return true;
and that would then look like:
static inline bool platform_pci_bridge_d3(struct pci_dev *dev)
{
if (pci_use_mid_pm())
return false;
if (acpi_pci_bridge_d3(dev))
return true;
if (device_property_read_bool(&dev->dev, "HotPlugSupportInD3"))
return true;
return false;
}
and then make a quirk in quirks.c that adds the software node property
for the Apple systems? Or something along those lines.
@Lukas, what do you think?
