On 18-Oct-17 6:57 PM, Bjorn Helgaas wrote:
> On Tue, Oct 17, 2017 at 10:27:49PM +0200, Thierry Reding wrote:
>> On Tue, Oct 17, 2017 at 12:34:28PM -0500, Bjorn Helgaas wrote:
>>> On Wed, Sep 27, 2017 at 05:28:35PM +0530, Manikanta Maddireddy wrote:
>>>> UPHY programming is performed by BPMP, PHY enable calls are
>>>> not required for Tegra186 PCIe. Power partition ungate is
>>>> done by BPMP powergate driver.
>>>>
>>>> Signed-off-by: Manikanta Maddireddy <mmaddireddy@xxxxxxxxxx>
>>>> Reviewed-by: Mikko Perttunen <mperttunen@xxxxxxxxxx>
>>>> Tested-by: Mikko Perttunen <mperttunen@xxxxxxxxxx>
>>>> ---
>>>> V2: Add soc->program_uphy check for phy_exit call
>>>> drivers/pci/host/pci-tegra.c | 132 +++++++++++++++++++++++++++++++++++--------
>>>> 1 file changed, 108 insertions(+), 24 deletions(-)
>>>
>>>> @@ -1012,10 +1016,12 @@ static int tegra_pcie_enable_controller(struct tegra_pcie *pcie)
>>>> afi_writel(pcie, value, AFI_FUSE);
>>>> }
>>>>
>>>> - err = tegra_pcie_phy_power_on(pcie);
>>>> - if (err < 0) {
>>>> - dev_err(dev, "failed to power on PHY(s): %d\n", err);
>>>> - return err;
>>>> + if (soc->program_uphy) {
>>>> + err = tegra_pcie_phy_power_on(pcie);
>>>> + if (err < 0) {
>>>> + dev_err(dev, "failed to power on PHY(s): %d\n", err);
>>>> + return err;
>>>> + }
>>>
>>> This looks good: it's obvious that the previously-supported devices
>>> continue to work the same way because you set ".program_uphy = true"
>>> for them. (It would be even more obvious if you changed the sense so
>>> that only the new device had to add this initializaer, but in general
>>> I prefer the positive logic as you have here, and I did verify that
>>> you added the initializer for all tegra_pcie_soc variants.)
>>>
>>>> @@ -1048,19 +1054,23 @@ static int tegra_pcie_enable_controller(struct tegra_pcie *pcie)
>>>> static void tegra_pcie_power_off(struct tegra_pcie *pcie)
>>>> {
>>>> struct device *dev = pcie->dev;
>>>> + const struct tegra_pcie_soc *soc = pcie->soc;
>>>> int err;
>>>>
>>>> /* TODO: disable and unprepare clocks? */
>>>>
>>>> - err = tegra_pcie_phy_power_off(pcie);
>>>> - if (err < 0)
>>>> - dev_err(dev, "failed to power off PHY(s): %d\n", err);
>>>> + if (soc->program_uphy) {
>>>> + err = tegra_pcie_phy_power_off(pcie);
>>>> + if (err < 0)
>>>> + dev_err(dev, "failed to power off PHY(s): %d\n", err);
>>>> + }
>>>>
>>>> reset_control_assert(pcie->pcie_xrst);
>>>> reset_control_assert(pcie->afi_rst);
>>>> reset_control_assert(pcie->pex_rst);
>>>>
>>>> - tegra_powergate_power_off(TEGRA_POWERGATE_PCIE);
>>>> + if (!dev->pm_domain)
>>>> + tegra_powergate_power_off(TEGRA_POWERGATE_PCIE);
>>>
>>> But this one isn't obvious because nothing in your patch obviously
>>> affects dev->pm_domain, so I can't tell whether this is safe. It's
>>> worth a changelog comment to help justify this.
>>
>> The last sentence in the commit message is what this is about. Maybe it
>> should be more verbose:
>>
>> Since the BPMP controls the powergates on Tegra186, there is no
>> need to manually power on and off the PCIe power partition. The
>> BPMP generic power domain driver takes care of it instead.
>
> If you know Tegra intimately, maybe that clarifies it. I don't, so my
> problem is that there's nothing in the patch that helps me verify
> this. I infer that maybe there's something different in the DT that
> means dev->pm_domain will be set for Tegra186, but not for other
> variants?
>
> There should be something that helps the reader connect the dots.
>
> Bjorn
>
Hi Bjorn,
Till tegra210 pmc driver provides direct powergate power ON/OFF functions.
In tegra186.dtsi power-domains property is added in pcie@10003000 DT node. reference: https://patchwork.ozlabs.org/patch/819113/
powergate-bpmp.c is the pm domain provider for pcie powergate.
Platform driver will look for pm domain provider of Tegra pcie host driver and sets dev->pm_domains.
powergate-bpmp.c driver registers powergate power ON/OFF functions to generic PD power_on/powerr_off callback functions.
Generic power domain will take care of calling power_on before Tegra PCIe probe.
So in this patch I used dev->pm_domain to skip pmc driver calls.
