Re: [PATCH v1] i2c: eg20t: use generic power management

[Bjorn Helgaas]

On Wed, Aug 05, 2020 at 09:51:54PM +0530, Vaibhav Gupta wrote:
> On Wed, Aug 05, 2020 at 10:28:32AM -0500, Bjorn Helgaas wrote:
> > On Wed, Aug 05, 2020 at 10:23:31AM -0500, Bjorn Helgaas wrote:
> > > On Mon, Jul 20, 2020 at 07:30:32PM +0530, Vaibhav Gupta wrote:
> > > > Drivers using legacy PM have to manage PCI states and device's PM states
> > > > themselves. They also need to take care of configuration registers.
> > > > 
> > > > With improved and powerful support of generic PM, PCI Core takes care of
> > > > above mentioned, device-independent, jobs.
> > > > 
> > > > This driver makes use of PCI helper functions like
> > > > pci_save/restore_state(), pci_enable/disable_device(),
> > > > pci_enable_wake() and pci_set_power_state() to do required operations. In
> > > > generic mode, they are no longer needed.
> > > > 
> > > > Change function parameter in both .suspend() and .resume() to
> > > > "struct device*" type. Use to_pci_dev() and dev_get_drvdata() to get
> > > > "struct pci_dev*" variable and drv data.
> > > > 
> > > > Compile-tested only.
> > > > 
> > > > Signed-off-by: Vaibhav Gupta <vaibhavgupta40@xxxxxxxxx>
> > > > ---
> > > >  drivers/i2c/busses/i2c-eg20t.c | 39 ++++++++--------------------------
> > > >  1 file changed, 9 insertions(+), 30 deletions(-)
> > > > 
> > > > diff --git a/drivers/i2c/busses/i2c-eg20t.c b/drivers/i2c/busses/i2c-eg20t.c
> > > > index 73f139690e4e..c0ddc4cc2ce7 100644
> > > > --- a/drivers/i2c/busses/i2c-eg20t.c
> > > > +++ b/drivers/i2c/busses/i2c-eg20t.c
> > > > @@ -846,11 +846,10 @@ static void pch_i2c_remove(struct pci_dev *pdev)
> > > >  	kfree(adap_info);
> > > >  }
> > > >  
> > > > -#ifdef CONFIG_PM
> > > > -static int pch_i2c_suspend(struct pci_dev *pdev, pm_message_t state)
> > > > +static int __maybe_unused pch_i2c_suspend(struct device *dev)
> > > >  {
> > > > -	int ret;
> > > >  	int i;
> > > > +	struct pci_dev *pdev = to_pci_dev(dev);
> > > >  	struct adapter_info *adap_info = pci_get_drvdata(pdev);
> > > 
> > > Why don't you use "adap_info = dev_get_drvdata(dev)" as you did below,
> > > so you don't need to_pci_dev()?
> > >
> Actually, line 870, pch_pci_dbg() again needs "struct pci_dev*" type
> pointer.  Either way I had to use to_pci_dev(), so defined "pdev"
> which made less number of required changes in the code.

OK.  pch_pci_dbg() only needs the pdev in order to *get* the struct
device *, so it's all sort of in circles.  But it's fine to do as you
did here.

> > > >  	void __iomem *p = adap_info->pch_data[0].pch_base_address;
> > > >  
> > > > @@ -872,34 +871,17 @@ static int pch_i2c_suspend(struct pci_dev *pdev, pm_message_t state)
> > > >  		ioread32(p + PCH_I2CSR), ioread32(p + PCH_I2CBUFSTA),
> > > >  		ioread32(p + PCH_I2CESRSTA));
> > > >  
> > > > -	ret = pci_save_state(pdev);
> > > > -
> > > > -	if (ret) {
> > > > -		pch_pci_err(pdev, "pci_save_state\n");
> > > > -		return ret;
> > > > -	}
> > > > -
> > > > -	pci_enable_wake(pdev, PCI_D3hot, 0);
> > > > -	pci_disable_device(pdev);
> > > > -	pci_set_power_state(pdev, pci_choose_state(pdev, state));
> > > > +	device_wakeup_disable(dev);
> > > >  
> > > >  	return 0;
> > > >  }
> > > >  
> > > > -static int pch_i2c_resume(struct pci_dev *pdev)
> > > > +static int __maybe_unused pch_i2c_resume(struct device *dev)
> > > >  {
> > > >  	int i;
> > > > -	struct adapter_info *adap_info = pci_get_drvdata(pdev);
> > > > -
> > > > -	pci_set_power_state(pdev, PCI_D0);
> > > > -	pci_restore_state(pdev);
> > > > +	struct adapter_info *adap_info = dev_get_drvdata(dev);
> > > >  
> > > > -	if (pci_enable_device(pdev) < 0) {
> > > > -		pch_pci_err(pdev, "pch_i2c_resume:pci_enable_device FAILED\n");
> > > > -		return -EIO;
> > > > -	}
> > > > -
> > > > -	pci_enable_wake(pdev, PCI_D3hot, 0);
> > > > +	device_wakeup_disable(dev);
> > > 
> > > It *looks* wrong to disable wakeup in both suspend and resume.  I
> > > think the usual pattern is to enable wakeup in suspend and disable it
> > > in resume.
> > > 
> > > But it looks like it's been that way since the driver was added by
> > > e9bc8fa5df1c ("i2c-eg20t: add driver for Intel EG20T").
> > > 
> > > If the device doesn't support wakeup, I would not expect the driver to
> > > mention wakeup at all.
> > > 
> > > In any case, I think it's the right thing for *this* patch to preserve
> > > the previous wakeup behavior.  Maybe we want a follow-up patch to just
> > > remove both device_wakeup_disable() calls?
> > > 
> We have seen this issue earlier in other drivers too. I remember you even
> identified and listed them.
> The PCI core calls, pci_enable_wake(pci_dev, PCI_D0, false). And if the driver
> does not want to enable-wake on suspend, we discussed and concluded that the
> calls should be dropped.
> I am sending v2, to include dropping wakeup call.

Personally, I think I would do this in two patches, and in the reverse
order than what I first suggested:

  1) Drop pci_enable_wake() calls
  2) Convert to generic PM

Doing them in that order means patch 2 will be slightly simpler, and
if there's any issue with removing the wakeup stuff, we can debug it
in the context of the original PCI PM code we've been using for years
without muddying the water with the additional generic PM changes.

> > > >  	for (i = 0; i < adap_info->ch_num; i++)
> > > >  		pch_i2c_init(&adap_info->pch_data[i]);
> > > > @@ -908,18 +890,15 @@ static int pch_i2c_resume(struct pci_dev *pdev)
> > > >  
> > > >  	return 0;
> > > >  }
> > > > -#else
> > > > -#define pch_i2c_suspend NULL
> > > > -#define pch_i2c_resume NULL
> > > > -#endif
> > > > +
> > > > +static SIMPLE_DEV_PM_OPS(pch_i2c_pm_ops, pch_i2c_suspend, pch_i2c_resume);
> > > >  
> > > >  static struct pci_driver pch_pcidriver = {
> > > >  	.name = KBUILD_MODNAME,
> > > >  	.id_table = pch_pcidev_id,
> > > >  	.probe = pch_i2c_probe,
> > > >  	.remove = pch_i2c_remove,
> > > > -	.suspend = pch_i2c_suspend,
> > > > -	.resume = pch_i2c_resume
> > > > +	.driver.pm = &pch_i2c_pm_ops,
> > > >  };
> > > >  
> > > >  module_pci_driver(pch_pcidriver);
> > > > -- 
> > > > 2.27.0
> > > >