Re: [PATCH 2/3] PCI: dwc: qcom: Set suspend_poweroff flag for SC7280

[Bjorn Helgaas <helgaas@kernel.org>]

[+cc Prasad, Andy, Rob, Krzysztof, Rajat, Saheed, Rama, Stephen,
Dmitry, Kalle for connection to https://lore.kernel.org/lkml/CAE-0n53ho2DX2rqQMvvKAuDCfsWW62TceTaNPzv5Mn_NQ-U6dA@mail.gmail.com/T/]

Subject line convention for this file is "PCI: qcom:" (not "PCI: dwc:
qcom:").

Find this from "git log --oneline drivers/pci/controller/dwc/pcie-qcom.c".

On Tue, May 17, 2022 at 08:41:34PM +0530, Manivannan Sadhasivam wrote:
> On Mon, May 16, 2022 at 03:19:50PM -0500, Bjorn Helgaas wrote:
> > On Fri, May 13, 2022 at 04:30:26PM +0530, Manivannan Sadhasivam wrote:
> > > For aggressive power saving on SC7280 SoCs, the power for the
> > > PCI devices will be taken off during system suspend. Hence,
> > > notify the same to the PCI device drivers using
> > > "suspend_poweroff" flag so that the drivers can prepare the PCI
> > > devices to handle the poweroff and recover them during resume.
> > 
> > No doubt "power ... will be taken off during system suspend" is
> > true, but this isn't very informative.  Is this a property of
> > SC7280?  A choice made by the SC7280 driver?  Why is this not
> > applicable to other systems?
> 
> The SC7280's RPMh firmware is cutting off the PCIe power domain
> during system suspend. And as I explained in previous patch, the RC
> driver itself may put the devices in D3cold conditionally on this
> platform. The reason is to save power as this chipset is being used
> in Chromebooks.

It looks like this should be squashed into the patch you mentioned:
https://lore.kernel.org/lkml/CAE-0n53ho2DX2rqQMvvKAuDCfsWW62TceTaNPzv5Mn_NQ-U6dA@mail.gmail.com/T/

If Prasad's patch is applied without this, devices will be powered
off, but nvme will not be prepared for it.  Apparently something would
be broken in that case?

Also, I think this patch should be reordered so the nvme driver is
prepared for suspend_poweroff before the qcom driver starts setting
it.  Otherwise there's a window where qcom sets suspend_poweroff and
powers off devices, but nvme doesn't know about it, and I assume
something will be broken in that case?

Please mention RPMh in the commit log, along with the specific
connection with system suspend, i.e., what OS action enables RPMh to
cut power.

> > > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> > > ---
> > >  drivers/pci/controller/dwc/pcie-qcom.c | 6 ++++++
> > >  1 file changed, 6 insertions(+)
> > > 
> > > diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c
> > > index 6ab90891801d..4b0ad2827f8f 100644
> > > --- a/drivers/pci/controller/dwc/pcie-qcom.c
> > > +++ b/drivers/pci/controller/dwc/pcie-qcom.c
> > > @@ -199,6 +199,7 @@ struct qcom_pcie_cfg {
> > >  	unsigned int has_ddrss_sf_tbu_clk:1;
> > >  	unsigned int has_aggre0_clk:1;
> > >  	unsigned int has_aggre1_clk:1;
> > > +	unsigned int suspend_poweroff:1;
> > >  };
> > >  
> > >  struct qcom_pcie {
> > > @@ -1220,6 +1221,10 @@ static int qcom_pcie_init_2_7_0(struct qcom_pcie *pcie)
> > >  	if (pcie->cfg->pipe_clk_need_muxing)
> > >  		clk_set_parent(res->pipe_clk_src, res->ref_clk_src);
> > >  
> > > +	/* Indicate PCI device drivers that the power will be taken off during system suspend */
> > > +	if (pcie->cfg->suspend_poweroff)
> > > +		pci->pp.bridge->suspend_poweroff = true;
> > > +
> > >  	ret = clk_bulk_prepare_enable(res->num_clks, res->clks);
> > >  	if (ret < 0)
> > >  		goto err_disable_regulators;
> > > @@ -1548,6 +1553,7 @@ static const struct qcom_pcie_cfg sc7280_cfg = {
> > >  	.ops = &ops_1_9_0,
> > >  	.has_tbu_clk = true,
> > >  	.pipe_clk_need_muxing = true,
> > > +	.suspend_poweroff = true,
> > >  };
> > >  
> > >  static const struct dw_pcie_ops dw_pcie_ops = {
> > > -- 
> > > 2.25.1
> > > 
> 
> -- 
> மணிவண்ணன் சதாசிவம்