[PATCH 0/2] PCI: designware: let dw_pcie_link_up() beware of LTSSM training bit

[PATCH 0/2] PCI: designware: let dw_pcie_link_up() beware of LTSSM training bit

[Jisheng Zhang]

patch1 is a trivial clean up: move the parameters for wait for link
into the core pcie-designware.c

Since link may be UP but still in link training, if so, we can't think
the link is up and operating correctly. So patch2 teaches
dw_pcie_link_up() beware of the PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING bit.


Jisheng Zhang (2):
  PCI: designware: mv parameters for wait for link into
    pcie-designware.c
  PCI: designware: let dw_pcie_link_up() beware of LTSSM training bit

 drivers/pci/host/pcie-designware.c | 11 +++++++++--
 drivers/pci/host/pcie-designware.h |  5 -----
 2 files changed, 9 insertions(+), 7 deletions(-)

-- 
2.8.1

[PATCH 1/2] PCI: designware: mv parameters for wait for link into pcie-designware.c

[Jisheng Zhang]

These parameters are only used in the core pcie-designware.c, let's
move them to the core driver.

Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
---
 drivers/pci/host/pcie-designware.c | 5 +++++
 drivers/pci/host/pcie-designware.h | 5 -----
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/pci/host/pcie-designware.c b/drivers/pci/host/pcie-designware.c
index aafd766..9df879a 100644
--- a/drivers/pci/host/pcie-designware.c
+++ b/drivers/pci/host/pcie-designware.c
@@ -75,6 +75,11 @@
 #define PCIE_PHY_DEBUG_R1		(PLR_OFFSET + 0x2c)
 #define PCIE_PHY_DEBUG_R1_LINK_UP	0x00000010
 
+/* Parameters for the waiting for link up routine */
+#define LINK_WAIT_MAX_RETRIES		10
+#define LINK_WAIT_USLEEP_MIN		90000
+#define LINK_WAIT_USLEEP_MAX		100000
+
 static struct pci_ops dw_pcie_ops;
 
 int dw_pcie_cfg_read(void __iomem *addr, int size, u32 *val)
diff --git a/drivers/pci/host/pcie-designware.h b/drivers/pci/host/pcie-designware.h
index f437f9b..384e79b 100644
--- a/drivers/pci/host/pcie-designware.h
+++ b/drivers/pci/host/pcie-designware.h
@@ -22,11 +22,6 @@
 #define MAX_MSI_IRQS			32
 #define MAX_MSI_CTRLS			(MAX_MSI_IRQS / 32)
 
-/* Parameters for the waiting for link up routine */
-#define LINK_WAIT_MAX_RETRIES		10
-#define LINK_WAIT_USLEEP_MIN		90000
-#define LINK_WAIT_USLEEP_MAX		100000
-
 struct pcie_port {
 	struct device		*dev;
 	u8			root_bus_nr;
-- 
2.8.1

[PATCH 2/2] PCI: designware: let dw_pcie_link_up() beware of LTSSM training bit

[Jisheng Zhang]

The link may be UP but still in link training. In this case, we can't
think the link is up and operating correctly. So we need to teach
dw_pcie_link_up() beware of the PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING bit.

This patch also rewrite PCIE_PHY_DEBUG_R1_LINK_UP definition so that
it's consistent with other MACROS.

Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
---
 drivers/pci/host/pcie-designware.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/host/pcie-designware.c b/drivers/pci/host/pcie-designware.c
index 9df879a..29e10dd 100644
--- a/drivers/pci/host/pcie-designware.c
+++ b/drivers/pci/host/pcie-designware.c
@@ -73,7 +73,8 @@
 /* PCIe Port Logic registers */
 #define PLR_OFFSET			0x700
 #define PCIE_PHY_DEBUG_R1		(PLR_OFFSET + 0x2c)
-#define PCIE_PHY_DEBUG_R1_LINK_UP	0x00000010
+#define PCIE_PHY_DEBUG_R1_LINK_UP	(0x1 << 4)
+#define PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING	(0x1 << 29)
 
 /* Parameters for the waiting for link up routine */
 #define LINK_WAIT_MAX_RETRIES		10
@@ -417,7 +418,8 @@ int dw_pcie_link_up(struct pcie_port *pp)
 		return pp->ops->link_up(pp);
 
 	val = readl(pp->dbi_base + PCIE_PHY_DEBUG_R1);
-	return val & PCIE_PHY_DEBUG_R1_LINK_UP;
+	return ((val & PCIE_PHY_DEBUG_R1_LINK_UP) &&
+		(!(val & PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING)));
 }
 
 static int dw_pcie_msi_map(struct irq_domain *domain, unsigned int irq,
-- 
2.8.1

[PATCH 2/2] PCI: designware: let dw_pcie_link_up() beware of LTSSM training bit

[Joao Pinto]

Hi,

On 7/6/2016 11:59 AM, Jisheng Zhang wrote:
> The link may be UP but still in link training. In this case, we can't
> think the link is up and operating correctly. So we need to teach
> dw_pcie_link_up() beware of the PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING bit.
> 
> This patch also rewrite PCIE_PHY_DEBUG_R1_LINK_UP definition so that
> it's consistent with other MACROS.
> 
> Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
> ---
>  drivers/pci/host/pcie-designware.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/host/pcie-designware.c b/drivers/pci/host/pcie-designware.c
> index 9df879a..29e10dd 100644
> --- a/drivers/pci/host/pcie-designware.c
> +++ b/drivers/pci/host/pcie-designware.c
> @@ -73,7 +73,8 @@
>  /* PCIe Port Logic registers */
>  #define PLR_OFFSET			0x700
>  #define PCIE_PHY_DEBUG_R1		(PLR_OFFSET + 0x2c)
> -#define PCIE_PHY_DEBUG_R1_LINK_UP	0x00000010
> +#define PCIE_PHY_DEBUG_R1_LINK_UP	(0x1 << 4)
> +#define PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING	(0x1 << 29)

According to the databook bit 29 is inside a range that is dedicated to M-PCIe.
Have you checked bit 29 state by experience?

>  
>  /* Parameters for the waiting for link up routine */
>  #define LINK_WAIT_MAX_RETRIES		10
> @@ -417,7 +418,8 @@ int dw_pcie_link_up(struct pcie_port *pp)
>  		return pp->ops->link_up(pp);
>  
>  	val = readl(pp->dbi_base + PCIE_PHY_DEBUG_R1);
> -	return val & PCIE_PHY_DEBUG_R1_LINK_UP;
> +	return ((val & PCIE_PHY_DEBUG_R1_LINK_UP) &&
> +		(!(val & PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING)));
>  }
>  
>  static int dw_pcie_msi_map(struct irq_domain *domain, unsigned int irq,
> 

Thanks,
Joao

[PATCH 2/2] PCI: designware: let dw_pcie_link_up() beware of LTSSM training bit

[Jisheng Zhang]

Dear Joao,

On Fri, 15 Jul 2016 16:10:24 +0100 Joao Pinto wrote:

> Hi,
> 
> On 7/6/2016 11:59 AM, Jisheng Zhang wrote:
> > The link may be UP but still in link training. In this case, we can't
> > think the link is up and operating correctly. So we need to teach
> > dw_pcie_link_up() beware of the PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING bit.
> > 
> > This patch also rewrite PCIE_PHY_DEBUG_R1_LINK_UP definition so that
> > it's consistent with other MACROS.
> > 
> > Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
> > ---
> >  drivers/pci/host/pcie-designware.c | 6 ++++--
> >  1 file changed, 4 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/pci/host/pcie-designware.c b/drivers/pci/host/pcie-designware.c
> > index 9df879a..29e10dd 100644
> > --- a/drivers/pci/host/pcie-designware.c
> > +++ b/drivers/pci/host/pcie-designware.c
> > @@ -73,7 +73,8 @@
> >  /* PCIe Port Logic registers */
> >  #define PLR_OFFSET			0x700
> >  #define PCIE_PHY_DEBUG_R1		(PLR_OFFSET + 0x2c)
> > -#define PCIE_PHY_DEBUG_R1_LINK_UP	0x00000010
> > +#define PCIE_PHY_DEBUG_R1_LINK_UP	(0x1 << 4)
> > +#define PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING	(0x1 << 29)  
> 
> According to the databook bit 29 is inside a range that is dedicated to M-PCIe.
> Have you checked bit 29 state by experience?

bit 29 here is bit 61 of cxpl_debug_info(64bit) in databook, the debug info is
composited by debug 0 and debug 1 registers.

I checked databook 4.3 and 4.21, bit 29 (bit 61 in databook) isn't dedicated to
M-PCIe. I think you may misread the bit 29 here as the bit29 in databook?
Databook indeed mentioned that bit[31:28] is reserved for M-PCIe

And after more code checking, I think this is not only marvell have this
case(link is up but still in link training), but pci-imx6.c also has this
case, we could check imx6_pcie_link_up() for reference.

Thanks,
Jisheng

> 
> >  
> >  /* Parameters for the waiting for link up routine */
> >  #define LINK_WAIT_MAX_RETRIES		10
> > @@ -417,7 +418,8 @@ int dw_pcie_link_up(struct pcie_port *pp)
> >  		return pp->ops->link_up(pp);
> >  
> >  	val = readl(pp->dbi_base + PCIE_PHY_DEBUG_R1);
> > -	return val & PCIE_PHY_DEBUG_R1_LINK_UP;
> > +	return ((val & PCIE_PHY_DEBUG_R1_LINK_UP) &&
> > +		(!(val & PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING)));
> >  }
> >  
> >  static int dw_pcie_msi_map(struct irq_domain *domain, unsigned int irq,
> >   
> 
> Thanks,
> Joao

[PATCH 2/2] PCI: designware: let dw_pcie_link_up() beware of LTSSM training bit

[Joao Pinto]

Hi Jisheng,

On 7/18/2016 3:38 AM, Jisheng Zhang wrote:
> Dear Joao,
> 
> On Fri, 15 Jul 2016 16:10:24 +0100 Joao Pinto wrote:
> 
>> Hi,
>>
>> On 7/6/2016 11:59 AM, Jisheng Zhang wrote:
>>> The link may be UP but still in link training. In this case, we can't
>>> think the link is up and operating correctly. So we need to teach
>>> dw_pcie_link_up() beware of the PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING bit.
>>>
>>> This patch also rewrite PCIE_PHY_DEBUG_R1_LINK_UP definition so that
>>> it's consistent with other MACROS.
>>>
>>> Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
>>> ---
>>>  drivers/pci/host/pcie-designware.c | 6 ++++--
>>>  1 file changed, 4 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/pci/host/pcie-designware.c b/drivers/pci/host/pcie-designware.c
>>> index 9df879a..29e10dd 100644
>>> --- a/drivers/pci/host/pcie-designware.c
>>> +++ b/drivers/pci/host/pcie-designware.c
>>> @@ -73,7 +73,8 @@
>>>  /* PCIe Port Logic registers */
>>>  #define PLR_OFFSET			0x700
>>>  #define PCIE_PHY_DEBUG_R1		(PLR_OFFSET + 0x2c)
>>> -#define PCIE_PHY_DEBUG_R1_LINK_UP	0x00000010
>>> +#define PCIE_PHY_DEBUG_R1_LINK_UP	(0x1 << 4)
>>> +#define PCIE_PHY_DEBUG_R1_LINK_IN_TRAINING	(0x1 << 29)  
>>
>> According to the databook bit 29 is inside a range that is dedicated to M-PCIe.
>> Have you checked bit 29 state by experience?
> 
> bit 29 here is bit 61 of cxpl_debug_info(64bit) in databook, the debug info is
> composited by debug 0 and debug 1 registers.

You are absolutely correct... I misread the databook in this subject.
In the latest Core versions this is also valid.

Acked-By: Joao Pinto <jpinto@synopsys.com>