Re: [PATCH v5 1/5] dt-bindings: PCI: brcmstb: Add brcm,enable-l1ss property

[Manivannan Sadhasivam <mani@kernel.org>]

On Tue, Aug 29, 2023 at 07:22:18AM -0500, Rob Herring wrote:
> On Fri, Aug 25, 2023 at 1:45 AM Manivannan Sadhasivam <mani@kernel.org> wrote:
> >
> > On Thu, Aug 24, 2023 at 10:55:02AM -0400, Jim Quinlan wrote:
> > > On Wed, Aug 23, 2023 at 2:17 PM Manivannan Sadhasivam <mani@kernel.org> wrote:
> > > >
> > > > On Wed, Aug 23, 2023 at 09:09:25AM -0400, Jim Quinlan wrote:
> > > > > On Wed, Aug 23, 2023 at 3:43 AM Manivannan Sadhasivam <mani@kernel.org> wrote:
> > > > > >
> > > > > > On Mon, May 08, 2023 at 06:01:21PM -0400, Jim Quinlan wrote:
> > > > > > > This commit adds the boolean "brcm,enable-l1ss" property:
> > > > > > >
> > > > > > >   The Broadcom STB/CM PCIe HW -- a core that is also used by RPi SOCs --
> > > > > > >   requires the driver probe() to deliberately place the HW one of three
> > > > > > >   CLKREQ# modes:
> > > > > > >
> > > > > > >   (a) CLKREQ# driven by the RC unconditionally
> > > > > > >   (b) CLKREQ# driven by the EP for ASPM L0s, L1
> > > > > > >   (c) Bidirectional CLKREQ#, as used for L1 Substates (L1SS).
> > > > > > >
> > > > > > >   The HW+driver can tell the difference between downstream devices that
> > > > > > >   need (a) and (b), but does not know when to configure (c).  All devices
> > > > > > >   should work fine when the driver chooses (a) or (b), but (c) may be
> > > > > > >   desired to realize the extra power savings that L1SS offers.  So we
> > > > > > >   introduce the boolean "brcm,enable-l1ss" property to inform the driver
> > > > > > >   that (c) is desired.  Setting this property only makes sense when the
> > > > > > >   downstream device is L1SS-capable and the OS is configured to activate
> > > > > > >   this mode (e.g. policy==powersupersave).
> > > > > > >
> > > > > > >   This property is already present in the Raspian version of Linux, but the
> > > > > > >   upstream driver implementation that follows adds more details and
> > > > > > >   discerns between (a) and (b).
> > > > > > >
> > > > > > > Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
> > > > > > > Reviewed-by: Rob Herring <robh@kernel.org>
> > > > > > > ---
> > > > > > >  Documentation/devicetree/bindings/pci/brcm,stb-pcie.yaml | 9 +++++++++
> > > > > > >  1 file changed, 9 insertions(+)
> > > > > > >
> > > > > > > diff --git a/Documentation/devicetree/bindings/pci/brcm,stb-pcie.yaml b/Documentation/devicetree/bindings/pci/brcm,stb-pcie.yaml
> > > > > > > index 7e15aae7d69e..8b61c2179608 100644
> > > > > > > --- a/Documentation/devicetree/bindings/pci/brcm,stb-pcie.yaml
> > > > > > > +++ b/Documentation/devicetree/bindings/pci/brcm,stb-pcie.yaml
> > > > > > > @@ -64,6 +64,15 @@ properties:
> > > > > > >
> > > > > > >    aspm-no-l0s: true
> > > > > > >
> > > > > > > +  brcm,enable-l1ss:
> > > > > > > +    description: Indicates that PCIe L1SS power savings
> > > > > > > +      are desired, the downstream device is L1SS-capable, and the
> > > > > > > +      OS has been configured to enable this mode.  For boards
> > > > > > > +      using a mini-card connector, this mode may not meet the
> > > > > > > +      TCRLon maximum time of 400ns, as specified in 3.2.5.2.2
> > > > > > > +      of the PCI Express Mini CEM 2.0 specification.
> > > > > >
> > > > > > As Lorenzo said, this property doesn't belong in DT. DT is supposed to specify
> > > > > > the hardware capability and not system/OS behavior.
> > > > >
> > > > > The "brcm,enable-l1ss" does NOT configure the OS behavior.
> > > > > It sets or not a mode bit to enable l1SS HW, whether or not the OS is
> > > > > configured for L1SS.
> > > > > It compensates for a problem in the PCIe core: the HW is not capable
> > > > > of dynamically
> > > > > switching between ASPM modes powersave and superpowersave.  I am actively
> > > > > advocating for our HW to change but that will take years.
> > > > >
> > > >
> > > > Okay, then I would say that the property name and commit message were a bit
> > > > misleading.
> > > >
> > > > I had briefly gone through the driver patch now. As per my understanding, you
> > > > have 2 modes in hw:
> > > >
> > > > 1. Clock PM - Refclk will be turned off by the host if CLKREQ# is deasserted by
> > > > the device (driving high) when the link is in L1.
> > > >
> > > > 2. L1SS - CLKREQ# will be used to decide L1SS entry and exit by the host.
> > >
> > > No, there are three, as enumerated in the commit message of
> > > "PCI: brcmstb: Configure HW CLKREQ# mode appropriate for downstream device"
> > >
> >
> > Yeah, another one is refclk always on.
> >
> > > >
> > > > Till now the driver only supported Clock PM through mode (1) but for supporting
> > > > L1SS you need to enable mode (2). And you are using this property to select mode
> > > > (2) when the L1SS supported devices are connected to the slot. Also, by
> > > > selecting this mode, you are loosing the benefit of mode (1) as both are not
> > > > compatible.
> > > >
> > > > My suggestion would be to just drop mode (1) and use mode (2) in the driver as
> > > > most of the recent devices should support L1SS (ofc there are exemptions).
> > > The disadvantage of this, as stated by the PCIe core HW designer, was
> > > that "doing so means
> > > we cannot enable the Cock Power Management capability since it may run afoul of
> > > the Tclron requirement."
> > >
> >
> > Ok.
> >
> > > I will attempt to press him on exactly what configurations and form
> > > factors would be
> > > vulnerable to this -- he was so convinced that it was a danger that he
> > > is against
> > > making L1SS mode the default.
> > >
> >
> > Hmm. After looking at this problem in detail, it looks to me that you can still
> > use DT but not with the property you proposed. Since these are hardware modes,
> > you can have a single DT property that specifies the mode that the driver can
> > use to configure the hw. It is similar to "phy-mode" property we have for the
> > network controllers.
> >
> > So you should have the property defined as below in binding:
> >
> > brcm,clkreq-mode:
> >   $ref: /schemas/types.yaml#/definitions/uint32
> >   enum: [ 0, 1, 2 ]
> 
> Is this really Broadcom specific?
> 

AFAIU, rest of the controllers do not have issues supporting Clock PM and L1SS
in hw. Neither does the spec define any incompatibility. So to me, this looks
like Broadcom specific.

- Mani

> Rob

-- 
மணிவண்ணன் சதாசிவம்