[V2] PCI: rcar: Add the initialization of PCIe link in resume_noirq

Commit Message

Marek Vasut Feb. 17, 2019, 1:24 p.m. UTC

From: Kazufumi Ikeda <[e-mail]>

Reestablish the PCIe link very early in the resume process in case it
went down to prevent PCI accesses from hanging the bus. Such accesses
can happen early in the PCI resume process, in the resume_noirq, thus
the link must be reestablished in the resume_noirq callback of the
driver.

Signed-off-by: Kazufumi Ikeda <[e-mail]>
Signed-off-by: Gaku Inami <[e-mail]>
Signed-off-by: Marek Vasut <[e-mail]>
Cc: Geert Uytterhoeven <[e-mail]>
Cc: Phil Edworthy <[e-mail]>
Cc: Simon Horman <[e-mail]>
Cc: Wolfram Sang <[e-mail]>
Cc: [e-mail]
---
V2: - Use BIT() macro for (1 << n)
    - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
      add extra changes to this function anymore
    - Make resume_noirq return early and clean up parenthesis therein
---
 drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

Comments

Wolfram Sang Feb. 19, 2019, 5:24 p.m. UTC | #1

On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> From: Kazufumi Ikeda <[e-mail]>
> 
> Reestablish the PCIe link very early in the resume process in case it
> went down to prevent PCI accesses from hanging the bus. Such accesses
> can happen early in the PCI resume process, in the resume_noirq, thus
> the link must be reestablished in the resume_noirq callback of the
> driver.
> 
> Signed-off-by: Kazufumi Ikeda <[e-mail]>
> Signed-off-by: Gaku Inami <[e-mail]>
> Signed-off-by: Marek Vasut <[e-mail]>

I am not a deep PCI expert. Still, the patch looks good to me and I
double checked with the documentation.

Acked-by: Wolfram Sang <[e-mail]>

Minor nit would be 'CONFIG_PM' protection around the new code. But I
have no strong opinion on that.

Simon Horman Feb. 25, 2019, 9:58 a.m. UTC | #2

On Tue, Feb 19, 2019 at 06:24:20PM +0100, Wolfram Sang wrote:
> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> > From: Kazufumi Ikeda <[e-mail]>
> > 
> > Reestablish the PCIe link very early in the resume process in case it
> > went down to prevent PCI accesses from hanging the bus. Such accesses
> > can happen early in the PCI resume process, in the resume_noirq, thus
> > the link must be reestablished in the resume_noirq callback of the
> > driver.
> > 
> > Signed-off-by: Kazufumi Ikeda <[e-mail]>
> > Signed-off-by: Gaku Inami <[e-mail]>
> > Signed-off-by: Marek Vasut <[e-mail]>
> 
> I am not a deep PCI expert. Still, the patch looks good to me and I
> double checked with the documentation.
> 
> Acked-by: Wolfram Sang <[e-mail]>
> 
> Minor nit would be 'CONFIG_PM' protection around the new code. But I
> have no strong opinion on that.

I agree that would be a nice enhancement but I also do not feel strongly
about it.

Reviewed-by: Simon Horman <[e-mail]>

Bjorn Helgaas March 7, 2019, 8:50 p.m. UTC | #3

On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> From: Kazufumi Ikeda <[e-mail]>
> 
> Reestablish the PCIe link very early in the resume process in case it
> went down to prevent PCI accesses from hanging the bus. Such accesses
> can happen early in the PCI resume process, in the resume_noirq, thus
> the link must be reestablished in the resume_noirq callback of the
> driver.
> 
> Signed-off-by: Kazufumi Ikeda <[e-mail]>
> Signed-off-by: Gaku Inami <[e-mail]>
> Signed-off-by: Marek Vasut <[e-mail]>
> Cc: Geert Uytterhoeven <[e-mail]>
> Cc: Phil Edworthy <[e-mail]>
> Cc: Simon Horman <[e-mail]>
> Cc: Wolfram Sang <[e-mail]>
> Cc: [e-mail]
> ---
> V2: - Use BIT() macro for (1 << n)
>     - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
>       add extra changes to this function anymore
>     - Make resume_noirq return early and clean up parenthesis therein
> ---
>  drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
>  1 file changed, 20 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-rcar.c b/drivers/pci/controller/pcie-rcar.c
> index c8febb009454..b8f8fb3bc640 100644
> --- a/drivers/pci/controller/pcie-rcar.c
> +++ b/drivers/pci/controller/pcie-rcar.c
> @@ -46,6 +46,7 @@
>  
>  /* Transfer control */
>  #define PCIETCTLR		0x02000
> +#define  DL_DOWN		BIT(3)
>  #define  CFINIT			1

I saw discussion after the V1 patch about using BIT() and making
similar constants also use BIT() for consistency.  That makes sense to
me, and I think the best way would be:

  1) in *this* patch, use "#define DL_DOWN 8"
  2) in a followup patch, convert them all to BIT()

That way each revision of pcie-rcar.c is self-consistent.

>  #define PCIETSTR		0x02004
>  #define  DATA_LINK_ACTIVE	1
> @@ -1130,6 +1131,7 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>  	pcie = pci_host_bridge_priv(bridge);
>  
>  	pcie->dev = dev;
> +	platform_set_drvdata(pdev, pcie);
>  
>  	err = pci_parse_request_of_pci_ranges(dev, &pcie->resources, NULL);
>  	if (err)
> @@ -1221,10 +1223,28 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>  	return err;
>  }
>  
> +static int rcar_pcie_resume_noirq(struct device *dev)
> +{
> +	struct rcar_pcie *pcie = dev_get_drvdata(dev);
> +
> +	if (rcar_pci_read_reg(pcie, PMSR) &&
> +	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
> +		return 0;
> +
> +	/* Re-establish the PCIe link */
> +	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
> +	return rcar_pcie_wait_for_dl(pcie);
> +}
> +
> +static const struct dev_pm_ops rcar_pcie_pm_ops = {
> +	.resume_noirq = rcar_pcie_resume_noirq,
> +};

I think there's the beginning of a convention to use #ifdef
CONFIG_PM_SLEEP around the ops themselves [1].  Otherwise I think
we'll get a warning about unused code when CONFIG_PM_SLEEP is unset.

>  static struct platform_driver rcar_pcie_driver = {
>  	.driver = {
>  		.name = "rcar-pcie",
>  		.of_match_table = rcar_pcie_of_match,
> +		.pm = &rcar_pcie_pm_ops,
>  		.suppress_bind_attrs = true,
>  	},
>  	.probe = rcar_pcie_probe,
> -- 
> 2.19.2
> 

[1] https://lore.kernel.org/linux-pci/[e-mail]

Marek Vasut March 7, 2019, 10:49 p.m. UTC | #4

On 3/7/19 9:50 PM, Bjorn Helgaas wrote:
> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
>> From: Kazufumi Ikeda <[e-mail]>
>>
>> Reestablish the PCIe link very early in the resume process in case it
>> went down to prevent PCI accesses from hanging the bus. Such accesses
>> can happen early in the PCI resume process, in the resume_noirq, thus
>> the link must be reestablished in the resume_noirq callback of the
>> driver.
>>
>> Signed-off-by: Kazufumi Ikeda <[e-mail]>
>> Signed-off-by: Gaku Inami <[e-mail]>
>> Signed-off-by: Marek Vasut <[e-mail]>
>> Cc: Geert Uytterhoeven <[e-mail]>
>> Cc: Phil Edworthy <[e-mail]>
>> Cc: Simon Horman <[e-mail]>
>> Cc: Wolfram Sang <[e-mail]>
>> Cc: [e-mail]
>> ---
>> V2: - Use BIT() macro for (1 << n)
>>     - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
>>       add extra changes to this function anymore
>>     - Make resume_noirq return early and clean up parenthesis therein
>> ---
>>  drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
>>  1 file changed, 20 insertions(+)
>>
>> diff --git a/drivers/pci/controller/pcie-rcar.c b/drivers/pci/controller/pcie-rcar.c
>> index c8febb009454..b8f8fb3bc640 100644
>> --- a/drivers/pci/controller/pcie-rcar.c
>> +++ b/drivers/pci/controller/pcie-rcar.c
>> @@ -46,6 +46,7 @@
>>  
>>  /* Transfer control */
>>  #define PCIETCTLR		0x02000
>> +#define  DL_DOWN		BIT(3)
>>  #define  CFINIT			1
> 
> I saw discussion after the V1 patch about using BIT() and making
> similar constants also use BIT() for consistency.  That makes sense to
> me, and I think the best way would be:
> 
>   1) in *this* patch, use "#define DL_DOWN 8"
>   2) in a followup patch, convert them all to BIT()
> 
> That way each revision of pcie-rcar.c is self-consistent.

But the BIT() macros are already cleaned , see commit
0ee40820989b330e24926d82953ffb9e1c7a8425

    PCI: rcar: Clean up the macros

>>  #define PCIETSTR		0x02004
>>  #define  DATA_LINK_ACTIVE	1
>> @@ -1130,6 +1131,7 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>>  	pcie = pci_host_bridge_priv(bridge);
>>  
>>  	pcie->dev = dev;
>> +	platform_set_drvdata(pdev, pcie);
>>  
>>  	err = pci_parse_request_of_pci_ranges(dev, &pcie->resources, NULL);
>>  	if (err)
>> @@ -1221,10 +1223,28 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>>  	return err;
>>  }
>>  
>> +static int rcar_pcie_resume_noirq(struct device *dev)
>> +{
>> +	struct rcar_pcie *pcie = dev_get_drvdata(dev);
>> +
>> +	if (rcar_pci_read_reg(pcie, PMSR) &&
>> +	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
>> +		return 0;
>> +
>> +	/* Re-establish the PCIe link */
>> +	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
>> +	return rcar_pcie_wait_for_dl(pcie);
>> +}
>> +
>> +static const struct dev_pm_ops rcar_pcie_pm_ops = {
>> +	.resume_noirq = rcar_pcie_resume_noirq,
>> +};
> 
> I think there's the beginning of a convention to use #ifdef
> CONFIG_PM_SLEEP around the ops themselves [1].  Otherwise I think
> we'll get a warning about unused code when CONFIG_PM_SLEEP is unset.

Only if I used SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() , but I set the
resume_noirq directly.

>>  static struct platform_driver rcar_pcie_driver = {
>>  	.driver = {
>>  		.name = "rcar-pcie",
>>  		.of_match_table = rcar_pcie_of_match,
>> +		.pm = &rcar_pcie_pm_ops,
>>  		.suppress_bind_attrs = true,
>>  	},
>>  	.probe = rcar_pcie_probe,
>> -- 
>> 2.19.2
>>
> 
> [1] https://lore.kernel.org/linux-pci/[e-mail]
>

Bjorn Helgaas March 8, 2019, 5:17 p.m. UTC | #5

[+cc linux-pm, Rafael for SET_NOIRQ_SYSTEM_SLEEP_PM_OPS question at the end]

On Thu, Mar 07, 2019 at 11:49:34PM +0100, Marek Vasut wrote:
> On 3/7/19 9:50 PM, Bjorn Helgaas wrote:
> > On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> >> From: Kazufumi Ikeda <[e-mail]>
> >>
> >> Reestablish the PCIe link very early in the resume process in case it
> >> went down to prevent PCI accesses from hanging the bus. Such accesses
> >> can happen early in the PCI resume process, in the resume_noirq, thus
> >> the link must be reestablished in the resume_noirq callback of the
> >> driver.
> >>
> >> Signed-off-by: Kazufumi Ikeda <[e-mail]>
> >> Signed-off-by: Gaku Inami <[e-mail]>
> >> Signed-off-by: Marek Vasut <[e-mail]>
> >> Cc: Geert Uytterhoeven <[e-mail]>
> >> Cc: Phil Edworthy <[e-mail]>
> >> Cc: Simon Horman <[e-mail]>
> >> Cc: Wolfram Sang <[e-mail]>
> >> Cc: [e-mail]
> >> ---
> >> V2: - Use BIT() macro for (1 << n)
> >>     - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
> >>       add extra changes to this function anymore
> >>     - Make resume_noirq return early and clean up parenthesis therein
> >> ---
> >>  drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
> >>  1 file changed, 20 insertions(+)
> >>
> >> diff --git a/drivers/pci/controller/pcie-rcar.c b/drivers/pci/controller/pcie-rcar.c
> >> index c8febb009454..b8f8fb3bc640 100644
> >> --- a/drivers/pci/controller/pcie-rcar.c
> >> +++ b/drivers/pci/controller/pcie-rcar.c
> >> @@ -46,6 +46,7 @@
> >>  
> >>  /* Transfer control */
> >>  #define PCIETCTLR		0x02000
> >> +#define  DL_DOWN		BIT(3)
> >>  #define  CFINIT			1
> > 
> > I saw discussion after the V1 patch about using BIT() and making
> > similar constants also use BIT() for consistency.  That makes sense to
> > me, and I think the best way would be:
> > 
> >   1) in *this* patch, use "#define DL_DOWN 8"
> >   2) in a followup patch, convert them all to BIT()
> > 
> > That way each revision of pcie-rcar.c is self-consistent.
> 
> But the BIT() macros are already cleaned , see commit
> 0ee40820989b330e24926d82953ffb9e1c7a8425
> 
>     PCI: rcar: Clean up the macros

Hmmm.  Maybe I'm missing something, but it looks like 0ee40820989b
didn't touch CFINIT, DATA_LINK_ACTIVE, or MSIFE.  Arguably,
LINK_SPEED_2_5GTS and LINK_SPEED_5_0GTS could use BIT() also.

I guess I'm just old-school, but in my personal opinion, BIT() is more
trouble than it's worth.  I'd rather see a complete bitmask because I
can easily match it with the typical pictures in a spec, multi-bit
fields are easy (you don't have to mix BIT() and GENMASK()), it gives
a hint about the register width, it's easy to match with a hexdump,
etc, e.g.,

  #define  DL_DOWN        0x00000008
  #define  CFINIT         0x00000001

But I'm not suggesting that you get rid of BIT() in this driver.  I'm
fine with it as long as it's used consistently.

BTW, while we're looking at it, I think rcar_pci_read_reg() and
rcar_pci_write_reg() really should use "u32" instead of "unsigned
long", since they deal with hardware registers that are explicitly
32 bits wide.

> >>  #define PCIETSTR		0x02004
> >>  #define  DATA_LINK_ACTIVE	1
> >> @@ -1130,6 +1131,7 @@ static int rcar_pcie_probe(struct platform_device *pdev)
> >>  	pcie = pci_host_bridge_priv(bridge);
> >>  
> >>  	pcie->dev = dev;
> >> +	platform_set_drvdata(pdev, pcie);
> >>  
> >>  	err = pci_parse_request_of_pci_ranges(dev, &pcie->resources, NULL);
> >>  	if (err)
> >> @@ -1221,10 +1223,28 @@ static int rcar_pcie_probe(struct platform_device *pdev)
> >>  	return err;
> >>  }
> >>  
> >> +static int rcar_pcie_resume_noirq(struct device *dev)
> >> +{
> >> +	struct rcar_pcie *pcie = dev_get_drvdata(dev);
> >> +
> >> +	if (rcar_pci_read_reg(pcie, PMSR) &&
> >> +	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
> >> +		return 0;
> >> +
> >> +	/* Re-establish the PCIe link */
> >> +	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
> >> +	return rcar_pcie_wait_for_dl(pcie);
> >> +}
> >> +
> >> +static const struct dev_pm_ops rcar_pcie_pm_ops = {
> >> +	.resume_noirq = rcar_pcie_resume_noirq,
> >> +};
> > 
> > I think there's the beginning of a convention to use #ifdef
> > CONFIG_PM_SLEEP around the ops themselves [1].  Otherwise I think
> > we'll get a warning about unused code when CONFIG_PM_SLEEP is unset.
> 
> Only if I used SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() , but I set the
> resume_noirq directly.

Fair enough.  I guess in this case if CONFIG_PM_SLEEP is unset, you
set the pointer, which avoids the "unused function" warning, but we
just never use that function pointer.

My intent is to avoid needless differences when possible, so when I
review things like this I look at how other drivers do things.  It
looks like all the other controllers use
SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() or similar:

  git grep -A3 "static.*dev_pm_ops" drivers/pci/controller

In the rcar case you only need the resume_fn, not the suspend_fn, so
SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() does a little more than you need and
you'd have to pass NULL for suspend_fn.  I didn't check them all
(suspend_noirq, freeze_noirq, poweroff_noirq), but at least for
suspend_noirq, all users check for NULL before calling through the
.suspend_noirq() function pointer, so I think that should be safe.

This *does* raise the question of whether you should be setting
.thaw_noirq and .restore_noirq in addition to .resume_noirq, as
SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() would.  I'm not a PM person, but maybe
Rafael or others will chime in.

Bjorn

Marek Vasut March 8, 2019, 9:35 p.m. UTC | #6

On 3/8/19 6:17 PM, Bjorn Helgaas wrote:
> [+cc linux-pm, Rafael for SET_NOIRQ_SYSTEM_SLEEP_PM_OPS question at the end]
> 
> On Thu, Mar 07, 2019 at 11:49:34PM +0100, Marek Vasut wrote:
>> On 3/7/19 9:50 PM, Bjorn Helgaas wrote:
>>> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
>>>> From: Kazufumi Ikeda <[e-mail]>
>>>>
>>>> Reestablish the PCIe link very early in the resume process in case it
>>>> went down to prevent PCI accesses from hanging the bus. Such accesses
>>>> can happen early in the PCI resume process, in the resume_noirq, thus
>>>> the link must be reestablished in the resume_noirq callback of the
>>>> driver.
>>>>
>>>> Signed-off-by: Kazufumi Ikeda <[e-mail]>
>>>> Signed-off-by: Gaku Inami <[e-mail]>
>>>> Signed-off-by: Marek Vasut <[e-mail]>
>>>> Cc: Geert Uytterhoeven <[e-mail]>
>>>> Cc: Phil Edworthy <[e-mail]>
>>>> Cc: Simon Horman <[e-mail]>
>>>> Cc: Wolfram Sang <[e-mail]>
>>>> Cc: [e-mail]
>>>> ---
>>>> V2: - Use BIT() macro for (1 << n)
>>>>     - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
>>>>       add extra changes to this function anymore
>>>>     - Make resume_noirq return early and clean up parenthesis therein
>>>> ---
>>>>  drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
>>>>  1 file changed, 20 insertions(+)
>>>>
>>>> diff --git a/drivers/pci/controller/pcie-rcar.c b/drivers/pci/controller/pcie-rcar.c
>>>> index c8febb009454..b8f8fb3bc640 100644
>>>> --- a/drivers/pci/controller/pcie-rcar.c
>>>> +++ b/drivers/pci/controller/pcie-rcar.c
>>>> @@ -46,6 +46,7 @@
>>>>  
>>>>  /* Transfer control */
>>>>  #define PCIETCTLR		0x02000
>>>> +#define  DL_DOWN		BIT(3)
>>>>  #define  CFINIT			1
>>>
>>> I saw discussion after the V1 patch about using BIT() and making
>>> similar constants also use BIT() for consistency.  That makes sense to
>>> me, and I think the best way would be:
>>>
>>>   1) in *this* patch, use "#define DL_DOWN 8"
>>>   2) in a followup patch, convert them all to BIT()
>>>
>>> That way each revision of pcie-rcar.c is self-consistent.
>>
>> But the BIT() macros are already cleaned , see commit
>> 0ee40820989b330e24926d82953ffb9e1c7a8425
>>
>>     PCI: rcar: Clean up the macros
> 
> Hmmm.  Maybe I'm missing something, but it looks like 0ee40820989b
> didn't touch CFINIT, DATA_LINK_ACTIVE, or MSIFE.  Arguably,
> LINK_SPEED_2_5GTS and LINK_SPEED_5_0GTS could use BIT() also.
> 
> I guess I'm just old-school, but in my personal opinion, BIT() is more
> trouble than it's worth.  I'd rather see a complete bitmask because I
> can easily match it with the typical pictures in a spec, multi-bit
> fields are easy (you don't have to mix BIT() and GENMASK()), it gives
> a hint about the register width, it's easy to match with a hexdump,
> etc, e.g.,
> 
>   #define  DL_DOWN        0x00000008
>   #define  CFINIT         0x00000001
> 
> But I'm not suggesting that you get rid of BIT() in this driver.  I'm
> fine with it as long as it's used consistently.
> 
> BTW, while we're looking at it, I think rcar_pci_read_reg() and
> rcar_pci_write_reg() really should use "u32" instead of "unsigned
> long", since they deal with hardware registers that are explicitly
> 32 bits wide.

OK, I can send those as separate patches.

>>>>  #define PCIETSTR		0x02004
>>>>  #define  DATA_LINK_ACTIVE	1
>>>> @@ -1130,6 +1131,7 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>>>>  	pcie = pci_host_bridge_priv(bridge);
>>>>  
>>>>  	pcie->dev = dev;
>>>> +	platform_set_drvdata(pdev, pcie);
>>>>  
>>>>  	err = pci_parse_request_of_pci_ranges(dev, &pcie->resources, NULL);
>>>>  	if (err)
>>>> @@ -1221,10 +1223,28 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>>>>  	return err;
>>>>  }
>>>>  
>>>> +static int rcar_pcie_resume_noirq(struct device *dev)
>>>> +{
>>>> +	struct rcar_pcie *pcie = dev_get_drvdata(dev);
>>>> +
>>>> +	if (rcar_pci_read_reg(pcie, PMSR) &&
>>>> +	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
>>>> +		return 0;
>>>> +
>>>> +	/* Re-establish the PCIe link */
>>>> +	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
>>>> +	return rcar_pcie_wait_for_dl(pcie);
>>>> +}
>>>> +
>>>> +static const struct dev_pm_ops rcar_pcie_pm_ops = {
>>>> +	.resume_noirq = rcar_pcie_resume_noirq,
>>>> +};
>>>
>>> I think there's the beginning of a convention to use #ifdef
>>> CONFIG_PM_SLEEP around the ops themselves [1].  Otherwise I think
>>> we'll get a warning about unused code when CONFIG_PM_SLEEP is unset.
>>
>> Only if I used SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() , but I set the
>> resume_noirq directly.
> 
> Fair enough.  I guess in this case if CONFIG_PM_SLEEP is unset, you
> set the pointer, which avoids the "unused function" warning, but we
> just never use that function pointer.
> 
> My intent is to avoid needless differences when possible, so when I
> review things like this I look at how other drivers do things.  It
> looks like all the other controllers use
> SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() or similar:
> 
>   git grep -A3 "static.*dev_pm_ops" drivers/pci/controller
> 
> In the rcar case you only need the resume_fn, not the suspend_fn, so
> SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() does a little more than you need and
> you'd have to pass NULL for suspend_fn.  I didn't check them all
> (suspend_noirq, freeze_noirq, poweroff_noirq), but at least for
> suspend_noirq, all users check for NULL before calling through the
> .suspend_noirq() function pointer, so I think that should be safe.
> 
> This *does* raise the question of whether you should be setting
> .thaw_noirq and .restore_noirq in addition to .resume_noirq, as
> SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() would.  I'm not a PM person, but maybe
> Rafael or others will chime in.

OK, let's wait for the feedback ...

Rafael J. Wysocki March 11, 2019, 10:10 a.m. UTC | #7

On Friday, March 8, 2019 10:35:49 PM CET Marek Vasut wrote:
> On 3/8/19 6:17 PM, Bjorn Helgaas wrote:
> > [+cc linux-pm, Rafael for SET_NOIRQ_SYSTEM_SLEEP_PM_OPS question at the end]
> > 
> > On Thu, Mar 07, 2019 at 11:49:34PM +0100, Marek Vasut wrote:
> >> On 3/7/19 9:50 PM, Bjorn Helgaas wrote:
> >>> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> >>>> From: Kazufumi Ikeda <[e-mail]>
> >>>>
> >>>> Reestablish the PCIe link very early in the resume process in case it
> >>>> went down to prevent PCI accesses from hanging the bus. Such accesses
> >>>> can happen early in the PCI resume process, in the resume_noirq, thus
> >>>> the link must be reestablished in the resume_noirq callback of the
> >>>> driver.
> >>>>
> >>>> Signed-off-by: Kazufumi Ikeda <[e-mail]>
> >>>> Signed-off-by: Gaku Inami <[e-mail]>
> >>>> Signed-off-by: Marek Vasut <[e-mail]>
> >>>> Cc: Geert Uytterhoeven <[e-mail]>
> >>>> Cc: Phil Edworthy <[e-mail]>
> >>>> Cc: Simon Horman <[e-mail]>
> >>>> Cc: Wolfram Sang <[e-mail]>
> >>>> Cc: [e-mail]
> >>>> ---
> >>>> V2: - Use BIT() macro for (1 << n)
> >>>>     - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
> >>>>       add extra changes to this function anymore
> >>>>     - Make resume_noirq return early and clean up parenthesis therein
> >>>> ---
> >>>>  drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
> >>>>  1 file changed, 20 insertions(+)
> >>>>
> >>>> diff --git a/drivers/pci/controller/pcie-rcar.c b/drivers/pci/controller/pcie-rcar.c
> >>>> index c8febb009454..b8f8fb3bc640 100644
> >>>> --- a/drivers/pci/controller/pcie-rcar.c
> >>>> +++ b/drivers/pci/controller/pcie-rcar.c
> >>>> @@ -46,6 +46,7 @@
> >>>>  
> >>>>  /* Transfer control */
> >>>>  #define PCIETCTLR		0x02000
> >>>> +#define  DL_DOWN		BIT(3)
> >>>>  #define  CFINIT			1
> >>>
> >>> I saw discussion after the V1 patch about using BIT() and making
> >>> similar constants also use BIT() for consistency.  That makes sense to
> >>> me, and I think the best way would be:
> >>>
> >>>   1) in *this* patch, use "#define DL_DOWN 8"
> >>>   2) in a followup patch, convert them all to BIT()
> >>>
> >>> That way each revision of pcie-rcar.c is self-consistent.
> >>
> >> But the BIT() macros are already cleaned , see commit
> >> 0ee40820989b330e24926d82953ffb9e1c7a8425
> >>
> >>     PCI: rcar: Clean up the macros
> > 
> > Hmmm.  Maybe I'm missing something, but it looks like 0ee40820989b
> > didn't touch CFINIT, DATA_LINK_ACTIVE, or MSIFE.  Arguably,
> > LINK_SPEED_2_5GTS and LINK_SPEED_5_0GTS could use BIT() also.
> > 
> > I guess I'm just old-school, but in my personal opinion, BIT() is more
> > trouble than it's worth.  I'd rather see a complete bitmask because I
> > can easily match it with the typical pictures in a spec, multi-bit
> > fields are easy (you don't have to mix BIT() and GENMASK()), it gives
> > a hint about the register width, it's easy to match with a hexdump,
> > etc, e.g.,
> > 
> >   #define  DL_DOWN        0x00000008
> >   #define  CFINIT         0x00000001
> > 
> > But I'm not suggesting that you get rid of BIT() in this driver.  I'm
> > fine with it as long as it's used consistently.
> > 
> > BTW, while we're looking at it, I think rcar_pci_read_reg() and
> > rcar_pci_write_reg() really should use "u32" instead of "unsigned
> > long", since they deal with hardware registers that are explicitly
> > 32 bits wide.
> 
> OK, I can send those as separate patches.
> 
> >>>>  #define PCIETSTR		0x02004
> >>>>  #define  DATA_LINK_ACTIVE	1
> >>>> @@ -1130,6 +1131,7 @@ static int rcar_pcie_probe(struct platform_device *pdev)
> >>>>  	pcie = pci_host_bridge_priv(bridge);
> >>>>  
> >>>>  	pcie->dev = dev;
> >>>> +	platform_set_drvdata(pdev, pcie);
> >>>>  
> >>>>  	err = pci_parse_request_of_pci_ranges(dev, &pcie->resources, NULL);
> >>>>  	if (err)
> >>>> @@ -1221,10 +1223,28 @@ static int rcar_pcie_probe(struct platform_device *pdev)
> >>>>  	return err;
> >>>>  }
> >>>>  
> >>>> +static int rcar_pcie_resume_noirq(struct device *dev)
> >>>> +{
> >>>> +	struct rcar_pcie *pcie = dev_get_drvdata(dev);
> >>>> +
> >>>> +	if (rcar_pci_read_reg(pcie, PMSR) &&
> >>>> +	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
> >>>> +		return 0;
> >>>> +
> >>>> +	/* Re-establish the PCIe link */
> >>>> +	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
> >>>> +	return rcar_pcie_wait_for_dl(pcie);
> >>>> +}
> >>>> +
> >>>> +static const struct dev_pm_ops rcar_pcie_pm_ops = {
> >>>> +	.resume_noirq = rcar_pcie_resume_noirq,
> >>>> +};
> >>>
> >>> I think there's the beginning of a convention to use #ifdef
> >>> CONFIG_PM_SLEEP around the ops themselves [1].  Otherwise I think
> >>> we'll get a warning about unused code when CONFIG_PM_SLEEP is unset.
> >>
> >> Only if I used SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() , but I set the
> >> resume_noirq directly.
> > 
> > Fair enough.  I guess in this case if CONFIG_PM_SLEEP is unset, you
> > set the pointer, which avoids the "unused function" warning, but we
> > just never use that function pointer.
> > 
> > My intent is to avoid needless differences when possible, so when I
> > review things like this I look at how other drivers do things.  It
> > looks like all the other controllers use
> > SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() or similar:
> > 
> >   git grep -A3 "static.*dev_pm_ops" drivers/pci/controller
> > 
> > In the rcar case you only need the resume_fn, not the suspend_fn, so
> > SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() does a little more than you need and
> > you'd have to pass NULL for suspend_fn.  I didn't check them all
> > (suspend_noirq, freeze_noirq, poweroff_noirq), but at least for
> > suspend_noirq, all users check for NULL before calling through the
> > .suspend_noirq() function pointer, so I think that should be safe.

Yes, this should be safe from the NULL pointer checks perspective.

> > This *does* raise the question of whether you should be setting
> > .thaw_noirq and .restore_noirq in addition to .resume_noirq, as
> > SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() would.  I'm not a PM person, but maybe
> > Rafael or others will chime in.

That depends on what happens during hibernation/restore.

The driver may not need to do anything before the memory snapshot is taken or
before the restore kernel jumps back to the image kernel.  In that case you
don't need freeze/thaw/restore_noirq at all.

However, chances are that it needs to do something and in that case doing
the same thing as during system-wide suspend/resume generally works, even
though it may not be particularly energy-efficient.  That's why it very
ofther is done.

Lorenzo Pieralisi March 19, 2019, 4:18 p.m. UTC | #8

On Fri, Mar 08, 2019 at 11:17:38AM -0600, Bjorn Helgaas wrote:

[...]

> > >> +static int rcar_pcie_resume_noirq(struct device *dev)
> > >> +{
> > >> +	struct rcar_pcie *pcie = dev_get_drvdata(dev);
> > >> +
> > >> +	if (rcar_pci_read_reg(pcie, PMSR) &&
> > >> +	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
> > >> +		return 0;
> > >> +
> > >> +	/* Re-establish the PCIe link */
> > >> +	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
> > >> +	return rcar_pcie_wait_for_dl(pcie);
> > >> +}
> > >> +
> > >> +static const struct dev_pm_ops rcar_pcie_pm_ops = {
> > >> +	.resume_noirq = rcar_pcie_resume_noirq,
> > >> +};
> > > 
> > > I think there's the beginning of a convention to use #ifdef
> > > CONFIG_PM_SLEEP around the ops themselves [1].  Otherwise I think
> > > we'll get a warning about unused code when CONFIG_PM_SLEEP is unset.
> > 
> > Only if I used SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() , but I set the
> > resume_noirq directly.
> 
> Fair enough.  I guess in this case if CONFIG_PM_SLEEP is unset, you
> set the pointer, which avoids the "unused function" warning, but we
> just never use that function pointer.
> 
> My intent is to avoid needless differences when possible, so when I
> review things like this I look at how other drivers do things.  It
> looks like all the other controllers use
> SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() or similar:
> 
>   git grep -A3 "static.*dev_pm_ops" drivers/pci/controller
> 
> In the rcar case you only need the resume_fn, not the suspend_fn, so
> SET_NOIRQ_SYSTEM_SLEEP_PM_OPS() does a little more than you need and
> you'd have to pass NULL for suspend_fn.  I didn't check them all
> (suspend_noirq, freeze_noirq, poweroff_noirq), but at least for
> suspend_noirq, all users check for NULL before calling through the
> .suspend_noirq() function pointer, so I think that should be safe.

It is a matter of consistency across drivers, yes, but that's something
I can easily solve with a clean-up patch on top of Marek's one.

I would merge this patch as-is and convert all the drivers to a uniform
convention (which one I shall see).

On a side note, this patch looks like a fix to me (even though it is
not trivial for me to add a sensible Fixes: tag) and should be treated
as such, so it should go in one of the upcoming -rc* (and I float the
idea of sending it to stable kernels on which S2R is basically broken).

Lorenzo

Lorenzo Pieralisi March 22, 2019, 11:31 a.m. UTC | #9

On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> From: Kazufumi Ikeda <[e-mail]>
> 
> Reestablish the PCIe link very early in the resume process in case it
> went down to prevent PCI accesses from hanging the bus. Such accesses

Hi Marek, Kazufumi,

Apologies for the delay.

Just as a clarification, when you state "in case it went down" isn't
this supposed to happen for every suspend cycle ? Let me know and I
will add a comment to the patch commit log.

> can happen early in the PCI resume process, in the resume_noirq, thus
> the link must be reestablished in the resume_noirq callback of the
> driver.
> 
> Signed-off-by: Kazufumi Ikeda <[e-mail]>
> Signed-off-by: Gaku Inami <[e-mail]>
> Signed-off-by: Marek Vasut <[e-mail]>
> Cc: Geert Uytterhoeven <[e-mail]>
> Cc: Phil Edworthy <[e-mail]>
> Cc: Simon Horman <[e-mail]>
> Cc: Wolfram Sang <[e-mail]>
> Cc: [e-mail]

This looks like a fix (most likely fixing initial S2R support, please
help me chase the commit ID), should we consider it for stable kernels ?

Without it I understand S2R is actually broken on platforms with this
host bridge.

Thanks,
Lorenzo

> ---
> V2: - Use BIT() macro for (1 << n)
>     - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
>       add extra changes to this function anymore
>     - Make resume_noirq return early and clean up parenthesis therein
> ---
>  drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
>  1 file changed, 20 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-rcar.c b/drivers/pci/controller/pcie-rcar.c
> index c8febb009454..b8f8fb3bc640 100644
> --- a/drivers/pci/controller/pcie-rcar.c
> +++ b/drivers/pci/controller/pcie-rcar.c
> @@ -46,6 +46,7 @@
>  
>  /* Transfer control */
>  #define PCIETCTLR		0x02000
> +#define  DL_DOWN		BIT(3)
>  #define  CFINIT			1
>  #define PCIETSTR		0x02004
>  #define  DATA_LINK_ACTIVE	1
> @@ -1130,6 +1131,7 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>  	pcie = pci_host_bridge_priv(bridge);
>  
>  	pcie->dev = dev;
> +	platform_set_drvdata(pdev, pcie);
>  
>  	err = pci_parse_request_of_pci_ranges(dev, &pcie->resources, NULL);
>  	if (err)
> @@ -1221,10 +1223,28 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>  	return err;
>  }
>  
> +static int rcar_pcie_resume_noirq(struct device *dev)
> +{
> +	struct rcar_pcie *pcie = dev_get_drvdata(dev);
> +
> +	if (rcar_pci_read_reg(pcie, PMSR) &&
> +	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
> +		return 0;
> +
> +	/* Re-establish the PCIe link */
> +	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
> +	return rcar_pcie_wait_for_dl(pcie);
> +}
> +
> +static const struct dev_pm_ops rcar_pcie_pm_ops = {
> +	.resume_noirq = rcar_pcie_resume_noirq,
> +};
> +
>  static struct platform_driver rcar_pcie_driver = {
>  	.driver = {
>  		.name = "rcar-pcie",
>  		.of_match_table = rcar_pcie_of_match,
> +		.pm = &rcar_pcie_pm_ops,
>  		.suppress_bind_attrs = true,
>  	},
>  	.probe = rcar_pcie_probe,
> -- 
> 2.19.2
>

Marek Vasut March 22, 2019, 12:09 p.m. UTC | #10

On 3/22/19 12:31 PM, Lorenzo Pieralisi wrote:
> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
>> From: Kazufumi Ikeda <[e-mail]>
>>
>> Reestablish the PCIe link very early in the resume process in case it
>> went down to prevent PCI accesses from hanging the bus. Such accesses
> 
> Hi Marek, Kazufumi,

Hi,

> Apologies for the delay.
> 
> Just as a clarification, when you state "in case it went down" isn't
> this supposed to happen for every suspend cycle ? Let me know and I
> will add a comment to the patch commit log.

It does happen on every suspend/resume cycle and if you manually put a
remote endpoint into non-L0 state.

>> can happen early in the PCI resume process, in the resume_noirq, thus
>> the link must be reestablished in the resume_noirq callback of the
>> driver.
>>
>> Signed-off-by: Kazufumi Ikeda <[e-mail]>
>> Signed-off-by: Gaku Inami <[e-mail]>
>> Signed-off-by: Marek Vasut <[e-mail]>
>> Cc: Geert Uytterhoeven <[e-mail]>
>> Cc: Phil Edworthy <[e-mail]>
>> Cc: Simon Horman <[e-mail]>
>> Cc: Wolfram Sang <[e-mail]>
>> Cc: [e-mail]
> 
> This looks like a fix (most likely fixing initial S2R support, please
> help me chase the commit ID), should we consider it for stable kernels ?
> 
> Without it I understand S2R is actually broken on platforms with this
> host bridge.
I don't think this ever worked, so it's hard to find a Fixes: commit for
this.

Lorenzo Pieralisi March 22, 2019, 12:18 p.m. UTC | #11

On Fri, Mar 22, 2019 at 01:09:13PM +0100, Marek Vasut wrote:
> On 3/22/19 12:31 PM, Lorenzo Pieralisi wrote:
> > On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> >> From: Kazufumi Ikeda <[e-mail]>
> >>
> >> Reestablish the PCIe link very early in the resume process in case it
> >> went down to prevent PCI accesses from hanging the bus. Such accesses
> > 
> > Hi Marek, Kazufumi,
> 
> Hi,
> 
> > Apologies for the delay.
> > 
> > Just as a clarification, when you state "in case it went down" isn't
> > this supposed to happen for every suspend cycle ? Let me know and I
> > will add a comment to the patch commit log.
> 
> It does happen on every suspend/resume cycle and if you manually put a
> remote endpoint into non-L0 state.

Ok I will update the log then.

> >> can happen early in the PCI resume process, in the resume_noirq, thus
> >> the link must be reestablished in the resume_noirq callback of the
> >> driver.
> >>
> >> Signed-off-by: Kazufumi Ikeda <[e-mail]>
> >> Signed-off-by: Gaku Inami <[e-mail]>
> >> Signed-off-by: Marek Vasut <[e-mail]>
> >> Cc: Geert Uytterhoeven <[e-mail]>
> >> Cc: Phil Edworthy <[e-mail]>
> >> Cc: Simon Horman <[e-mail]>
> >> Cc: Wolfram Sang <[e-mail]>
> >> Cc: [e-mail]
> > 
> > This looks like a fix (most likely fixing initial S2R support, please
> > help me chase the commit ID), should we consider it for stable kernels ?
> > 
> > Without it I understand S2R is actually broken on platforms with this
> > host bridge.
> I don't think this ever worked, so it's hard to find a Fixes: commit for
> this.

If we want to send it to stable kernels we have to select which versions
we are covering. I think the only options for a Fixes: tag are either
the initial S2R support commit for the platforms this driver runs on
or the initial driver commit that harks back to v3.16 AFAICS.

I will queue it next week so there is some time to think about it.

Thanks,
Lorenzo

Geert Uytterhoeven March 22, 2019, 12:29 p.m. UTC | #12

Hi Lorenzo, Marek,

On Fri, Mar 22, 2019 at 1:18 PM Lorenzo Pieralisi
<[e-mail]> wrote:
> On Fri, Mar 22, 2019 at 01:09:13PM +0100, Marek Vasut wrote:
> > On 3/22/19 12:31 PM, Lorenzo Pieralisi wrote:
> > > On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> > >> From: Kazufumi Ikeda <[e-mail]>
> > >>
> > >> Reestablish the PCIe link very early in the resume process in case it
> > >> went down to prevent PCI accesses from hanging the bus. Such accesses
> > >> can happen early in the PCI resume process, in the resume_noirq, thus
> > >> the link must be reestablished in the resume_noirq callback of the
> > >> driver.
> > >
> > > This looks like a fix (most likely fixing initial S2R support, please
> > > help me chase the commit ID), should we consider it for stable kernels ?
> > >
> > > Without it I understand S2R is actually broken on platforms with this
> > > host bridge.
> > I don't think this ever worked, so it's hard to find a Fixes: commit for
> > this.
>
> If we want to send it to stable kernels we have to select which versions
> we are covering. I think the only options for a Fixes: tag are either
> the initial S2R support commit for the platforms this driver runs on
> or the initial driver commit that harks back to v3.16 AFAICS.

This only started to become an issue when support for arm64 platforms
was added, where PSCI may power down the SoC, right?

Hence:
Fixes: e015f88c368da1e6 ("PCI: rcar: Add support for R-Car H3 to pcie-rcar")

Gr{oetje,eeting}s,

                        Geert

Marek Vasut March 22, 2019, 12:33 p.m. UTC | #13

On 3/22/19 1:29 PM, Geert Uytterhoeven wrote:
> Hi Lorenzo, Marek,

Hi,

> On Fri, Mar 22, 2019 at 1:18 PM Lorenzo Pieralisi
> <[e-mail]> wrote:
>> On Fri, Mar 22, 2019 at 01:09:13PM +0100, Marek Vasut wrote:
>>> On 3/22/19 12:31 PM, Lorenzo Pieralisi wrote:
>>>> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
>>>>> From: Kazufumi Ikeda <[e-mail]>
>>>>>
>>>>> Reestablish the PCIe link very early in the resume process in case it
>>>>> went down to prevent PCI accesses from hanging the bus. Such accesses
>>>>> can happen early in the PCI resume process, in the resume_noirq, thus
>>>>> the link must be reestablished in the resume_noirq callback of the
>>>>> driver.
>>>>
>>>> This looks like a fix (most likely fixing initial S2R support, please
>>>> help me chase the commit ID), should we consider it for stable kernels ?
>>>>
>>>> Without it I understand S2R is actually broken on platforms with this
>>>> host bridge.
>>> I don't think this ever worked, so it's hard to find a Fixes: commit for
>>> this.
>>
>> If we want to send it to stable kernels we have to select which versions
>> we are covering. I think the only options for a Fixes: tag are either
>> the initial S2R support commit for the platforms this driver runs on
>> or the initial driver commit that harks back to v3.16 AFAICS.
> 
> This only started to become an issue when support for arm64 platforms
> was added, where PSCI may power down the SoC, right?

Wouldn't you also hit this on ARM32 LPAE ones ?

> Hence:
> Fixes: e015f88c368da1e6 ("PCI: rcar: Add support for R-Car H3 to pcie-rcar")
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
>

Geert Uytterhoeven March 22, 2019, 12:42 p.m. UTC | #14

Hi Marek,

On Fri, Mar 22, 2019 at 1:33 PM Marek Vasut <[e-mail]> wrote:
> On 3/22/19 1:29 PM, Geert Uytterhoeven wrote:
> > On Fri, Mar 22, 2019 at 1:18 PM Lorenzo Pieralisi
> > <[e-mail]> wrote:
> >> On Fri, Mar 22, 2019 at 01:09:13PM +0100, Marek Vasut wrote:
> >>> On 3/22/19 12:31 PM, Lorenzo Pieralisi wrote:
> >>>> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> >>>>> From: Kazufumi Ikeda <[e-mail]>
> >>>>>
> >>>>> Reestablish the PCIe link very early in the resume process in case it
> >>>>> went down to prevent PCI accesses from hanging the bus. Such accesses
> >>>>> can happen early in the PCI resume process, in the resume_noirq, thus
> >>>>> the link must be reestablished in the resume_noirq callback of the
> >>>>> driver.
> >>>>
> >>>> This looks like a fix (most likely fixing initial S2R support, please
> >>>> help me chase the commit ID), should we consider it for stable kernels ?
> >>>>
> >>>> Without it I understand S2R is actually broken on platforms with this
> >>>> host bridge.
> >>> I don't think this ever worked, so it's hard to find a Fixes: commit for
> >>> this.
> >>
> >> If we want to send it to stable kernels we have to select which versions
> >> we are covering. I think the only options for a Fixes: tag are either
> >> the initial S2R support commit for the platforms this driver runs on
> >> or the initial driver commit that harks back to v3.16 AFAICS.
> >
> > This only started to become an issue when support for arm64 platforms
> > was added, where PSCI may power down the SoC, right?
>
> Wouldn't you also hit this on ARM32 LPAE ones ?

This is about the PCI link issue on system resume, right?

> > Hence:
> > Fixes: e015f88c368da1e6 ("PCI: rcar: Add support for R-Car H3 to pcie-rcar")

Gr{oetje,eeting}s,

                        Geert

Marek Vasut March 22, 2019, 12:49 p.m. UTC | #15

On 3/22/19 1:42 PM, Geert Uytterhoeven wrote:
> Hi Marek,

Hi,

> On Fri, Mar 22, 2019 at 1:33 PM Marek Vasut <[e-mail]> wrote:
>> On 3/22/19 1:29 PM, Geert Uytterhoeven wrote:
>>> On Fri, Mar 22, 2019 at 1:18 PM Lorenzo Pieralisi
>>> <[e-mail]> wrote:
>>>> On Fri, Mar 22, 2019 at 01:09:13PM +0100, Marek Vasut wrote:
>>>>> On 3/22/19 12:31 PM, Lorenzo Pieralisi wrote:
>>>>>> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
>>>>>>> From: Kazufumi Ikeda <[e-mail]>
>>>>>>>
>>>>>>> Reestablish the PCIe link very early in the resume process in case it
>>>>>>> went down to prevent PCI accesses from hanging the bus. Such accesses
>>>>>>> can happen early in the PCI resume process, in the resume_noirq, thus
>>>>>>> the link must be reestablished in the resume_noirq callback of the
>>>>>>> driver.
>>>>>>
>>>>>> This looks like a fix (most likely fixing initial S2R support, please
>>>>>> help me chase the commit ID), should we consider it for stable kernels ?
>>>>>>
>>>>>> Without it I understand S2R is actually broken on platforms with this
>>>>>> host bridge.
>>>>> I don't think this ever worked, so it's hard to find a Fixes: commit for
>>>>> this.
>>>>
>>>> If we want to send it to stable kernels we have to select which versions
>>>> we are covering. I think the only options for a Fixes: tag are either
>>>> the initial S2R support commit for the platforms this driver runs on
>>>> or the initial driver commit that harks back to v3.16 AFAICS.
>>>
>>> This only started to become an issue when support for arm64 platforms
>>> was added, where PSCI may power down the SoC, right?
>>
>> Wouldn't you also hit this on ARM32 LPAE ones ?
> 
> This is about the PCI link issue on system resume, right?

Forget what I said, yes, I mixed this with the MSI 64bit patch.

>>> Hence:
>>> Fixes: e015f88c368da1e6 ("PCI: rcar: Add support for R-Car H3 to pcie-rcar")
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
>

Lorenzo Pieralisi March 25, 2019, 4:54 p.m. UTC | #16

On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
> From: Kazufumi Ikeda <[e-mail]>
> 
> Reestablish the PCIe link very early in the resume process in case it
> went down to prevent PCI accesses from hanging the bus. Such accesses
> can happen early in the PCI resume process, in the resume_noirq, thus
> the link must be reestablished in the resume_noirq callback of the
> driver.
> 
> Signed-off-by: Kazufumi Ikeda <[e-mail]>
> Signed-off-by: Gaku Inami <[e-mail]>
> Signed-off-by: Marek Vasut <[e-mail]>
> Cc: Geert Uytterhoeven <[e-mail]>
> Cc: Phil Edworthy <[e-mail]>
> Cc: Simon Horman <[e-mail]>
> Cc: Wolfram Sang <[e-mail]>
> Cc: [e-mail]
> ---
> V2: - Use BIT() macro for (1 << n)
>     - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
>       add extra changes to this function anymore
>     - Make resume_noirq return early and clean up parenthesis therein
> ---
>  drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
>  1 file changed, 20 insertions(+)

I applied it to pci/controller-fixes but the kbot barfed at it so we
need to rework it.

Thanks,
Lorenzo

> diff --git a/drivers/pci/controller/pcie-rcar.c b/drivers/pci/controller/pcie-rcar.c
> index c8febb009454..b8f8fb3bc640 100644
> --- a/drivers/pci/controller/pcie-rcar.c
> +++ b/drivers/pci/controller/pcie-rcar.c
> @@ -46,6 +46,7 @@
>  
>  /* Transfer control */
>  #define PCIETCTLR		0x02000
> +#define  DL_DOWN		BIT(3)
>  #define  CFINIT			1
>  #define PCIETSTR		0x02004
>  #define  DATA_LINK_ACTIVE	1
> @@ -1130,6 +1131,7 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>  	pcie = pci_host_bridge_priv(bridge);
>  
>  	pcie->dev = dev;
> +	platform_set_drvdata(pdev, pcie);
>  
>  	err = pci_parse_request_of_pci_ranges(dev, &pcie->resources, NULL);
>  	if (err)
> @@ -1221,10 +1223,28 @@ static int rcar_pcie_probe(struct platform_device *pdev)
>  	return err;
>  }
>  
> +static int rcar_pcie_resume_noirq(struct device *dev)
> +{
> +	struct rcar_pcie *pcie = dev_get_drvdata(dev);
> +
> +	if (rcar_pci_read_reg(pcie, PMSR) &&
> +	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
> +		return 0;
> +
> +	/* Re-establish the PCIe link */
> +	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
> +	return rcar_pcie_wait_for_dl(pcie);
> +}
> +
> +static const struct dev_pm_ops rcar_pcie_pm_ops = {
> +	.resume_noirq = rcar_pcie_resume_noirq,
> +};
> +
>  static struct platform_driver rcar_pcie_driver = {
>  	.driver = {
>  		.name = "rcar-pcie",
>  		.of_match_table = rcar_pcie_of_match,
> +		.pm = &rcar_pcie_pm_ops,
>  		.suppress_bind_attrs = true,
>  	},
>  	.probe = rcar_pcie_probe,
> -- 
> 2.19.2
>

Marek Vasut March 25, 2019, 7:44 p.m. UTC | #17

On 3/25/19 5:54 PM, Lorenzo Pieralisi wrote:
> On Sun, Feb 17, 2019 at 02:24:41PM +0100, [e-mail] wrote:
>> From: Kazufumi Ikeda <[e-mail]>
>>
>> Reestablish the PCIe link very early in the resume process in case it
>> went down to prevent PCI accesses from hanging the bus. Such accesses
>> can happen early in the PCI resume process, in the resume_noirq, thus
>> the link must be reestablished in the resume_noirq callback of the
>> driver.
>>
>> Signed-off-by: Kazufumi Ikeda <[e-mail]>
>> Signed-off-by: Gaku Inami <[e-mail]>
>> Signed-off-by: Marek Vasut <[e-mail]>
>> Cc: Geert Uytterhoeven <[e-mail]>
>> Cc: Phil Edworthy <[e-mail]>
>> Cc: Simon Horman <[e-mail]>
>> Cc: Wolfram Sang <[e-mail]>
>> Cc: [e-mail]
>> ---
>> V2: - Use BIT() macro for (1 << n)
>>     - Since polling in rcar_pcie_wait_for_dl() uses udelay(), do not
>>       add extra changes to this function anymore
>>     - Make resume_noirq return early and clean up parenthesis therein
>> ---
>>  drivers/pci/controller/pcie-rcar.c | 20 ++++++++++++++++++++
>>  1 file changed, 20 insertions(+)
> 
> I applied it to pci/controller-fixes but the kbot barfed at it so we
> need to rework it.

Right, PMSR definition added, updated commit message retained, patch
resubmitted as V3.

Patch

diff --git a/drivers/pci/controller/pcie-rcar.c b/drivers/pci/controller/pcie-rcar.c
index c8febb009454..b8f8fb3bc640 100644
--- a/drivers/pci/controller/pcie-rcar.c
+++ b/drivers/pci/controller/pcie-rcar.c
@@ -46,6 +46,7 @@ 
 
 /* Transfer control */
 #define PCIETCTLR		0x02000
+#define  DL_DOWN		BIT(3)
 #define  CFINIT			1
 #define PCIETSTR		0x02004
 #define  DATA_LINK_ACTIVE	1
@@ -1130,6 +1131,7 @@  static int rcar_pcie_probe(struct platform_device *pdev)
 	pcie = pci_host_bridge_priv(bridge);
 
 	pcie->dev = dev;
+	platform_set_drvdata(pdev, pcie);
 
 	err = pci_parse_request_of_pci_ranges(dev, &pcie->resources, NULL);
 	if (err)
@@ -1221,10 +1223,28 @@  static int rcar_pcie_probe(struct platform_device *pdev)
 	return err;
 }
 
+static int rcar_pcie_resume_noirq(struct device *dev)
+{
+	struct rcar_pcie *pcie = dev_get_drvdata(dev);
+
+	if (rcar_pci_read_reg(pcie, PMSR) &&
+	    !(rcar_pci_read_reg(pcie, PCIETCTLR) & DL_DOWN))
+		return 0;
+
+	/* Re-establish the PCIe link */
+	rcar_pci_write_reg(pcie, CFINIT, PCIETCTLR);
+	return rcar_pcie_wait_for_dl(pcie);
+}
+
+static const struct dev_pm_ops rcar_pcie_pm_ops = {
+	.resume_noirq = rcar_pcie_resume_noirq,
+};
+
 static struct platform_driver rcar_pcie_driver = {
 	.driver = {
 		.name = "rcar-pcie",
 		.of_match_table = rcar_pcie_of_match,
+		.pm = &rcar_pcie_pm_ops,
 		.suppress_bind_attrs = true,
 	},
 	.probe = rcar_pcie_probe,