[PATCH] PCI/ASPM: Don't retrain link if ASPM not possible.

[PATCH] PCI/ASPM: Don't retrain link if ASPM not possible.

[David Daney]

From: David Daney <david.daney@cavium.com>

Some (defective) PCIe devices are not able to reliably do link
retraining.

Check to see if ASPM is possible between link partners before
configuring common clocking, and doing the resulting link retraining.
If ASPM is not possible, there is no reason to risk losing access to a
device due to an unnecessary link retraining.

Signed-off-by: David Daney <david.daney@cavium.com>
---
 drivers/pci/pcie/aspm.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c
index 0ec649d..d6667db 100644
--- a/drivers/pci/pcie/aspm.c
+++ b/drivers/pci/pcie/aspm.c
@@ -351,12 +351,26 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)
 		return;
 	}
 
+	/* Get upstream/downstream components' register state */
+	pcie_get_aspm_reg(parent, &upreg);
+	child = list_entry(linkbus->devices.next, struct pci_dev, bus_list);
+	pcie_get_aspm_reg(child, &dwreg);
+
+	/*
+	 * If ASPM not supported, don't mess with the clocks and link,
+	 * bail out now.
+	 */
+	if (!(upreg.support & dwreg.support))
+		return;
+
 	/* Configure common clock before checking latencies */
 	pcie_aspm_configure_common_clock(link);
 
-	/* Get upstream/downstream components' register state */
+	/*
+	 * Re-read upstream/downstream components' register state
+	 * after clock configuration
+	 */
 	pcie_get_aspm_reg(parent, &upreg);
-	child = list_entry(linkbus->devices.next, struct pci_dev, bus_list);
 	pcie_get_aspm_reg(child, &dwreg);
 
 	/*
-- 
1.7.11.7

Ping: [PATCH] PCI/ASPM: Don't retrain link if ASPM not possible.

[David Daney]

On 11/17/2016 02:25 PM, David Daney wrote:
> From: David Daney <david.daney@cavium.com>
>
> Some (defective) PCIe devices are not able to reliably do link
> retraining.
>
> Check to see if ASPM is possible between link partners before
> configuring common clocking, and doing the resulting link retraining.
> If ASPM is not possible, there is no reason to risk losing access to a
> device due to an unnecessary link retraining.
>

Hi Bjorn,

It has been a couple of weeks since I originally sent this, and I wanted 
to know if you have had a chance to look at it.  If possible, it would 
be nice to consider for the approaching merge window.

Thanks,
David Daney


> Signed-off-by: David Daney <david.daney@cavium.com>
> ---
>   drivers/pci/pcie/aspm.c | 18 ++++++++++++++++--
>   1 file changed, 16 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c
> index 0ec649d..d6667db 100644
> --- a/drivers/pci/pcie/aspm.c
> +++ b/drivers/pci/pcie/aspm.c
> @@ -351,12 +351,26 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)
>   		return;
>   	}
>
> +	/* Get upstream/downstream components' register state */
> +	pcie_get_aspm_reg(parent, &upreg);
> +	child = list_entry(linkbus->devices.next, struct pci_dev, bus_list);
> +	pcie_get_aspm_reg(child, &dwreg);
> +
> +	/*
> +	 * If ASPM not supported, don't mess with the clocks and link,
> +	 * bail out now.
> +	 */
> +	if (!(upreg.support & dwreg.support))
> +		return;
> +
>   	/* Configure common clock before checking latencies */
>   	pcie_aspm_configure_common_clock(link);
>
> -	/* Get upstream/downstream components' register state */
> +	/*
> +	 * Re-read upstream/downstream components' register state
> +	 * after clock configuration
> +	 */
>   	pcie_get_aspm_reg(parent, &upreg);
> -	child = list_entry(linkbus->devices.next, struct pci_dev, bus_list);
>   	pcie_get_aspm_reg(child, &dwreg);
>
>   	/*
>

Re: Ping: [PATCH] PCI/ASPM: Don't retrain link if ASPM not possible.

[Bjorn Helgaas]

Hi David,

On Wed, Nov 30, 2016 at 09:11:11AM -0800, David Daney wrote:
> On 11/17/2016 02:25 PM, David Daney wrote:
> >From: David Daney <david.daney@cavium.com>
> >
> >Some (defective) PCIe devices are not able to reliably do link
> >retraining.
> >
> >Check to see if ASPM is possible between link partners before
> >configuring common clocking, and doing the resulting link retraining.
> >If ASPM is not possible, there is no reason to risk losing access to a
> >device due to an unnecessary link retraining.
> >
> 
> Hi Bjorn,
> 
> It has been a couple of weeks since I originally sent this, and I
> wanted to know if you have had a chance to look at it.  If possible,
> it would be nice to consider for the approaching merge window.

Yes, definitely.  I haven't had a chance yet, but I hope to.  Right
now my top priority is getting the arm64 ECAM quirks in for v4.10.
Once that's settled I'll come back and pick up some of these fixes.

Bjorn

Re: [PATCH] PCI/ASPM: Don't retrain link if ASPM not possible.

[Bjorn Helgaas]

On Thu, Nov 17, 2016 at 02:25:01PM -0800, David Daney wrote:
> From: David Daney <david.daney@cavium.com>
> 
> Some (defective) PCIe devices are not able to reliably do link
> retraining.
> 
> Check to see if ASPM is possible between link partners before
> configuring common clocking, and doing the resulting link retraining.
> If ASPM is not possible, there is no reason to risk losing access to a
> device due to an unnecessary link retraining.
> 
> Signed-off-by: David Daney <david.daney@cavium.com>

Applied to pci/aspm for v4.10, thanks, David!

> ---
>  drivers/pci/pcie/aspm.c | 18 ++++++++++++++++--
>  1 file changed, 16 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c
> index 0ec649d..d6667db 100644
> --- a/drivers/pci/pcie/aspm.c
> +++ b/drivers/pci/pcie/aspm.c
> @@ -351,12 +351,26 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)
>  		return;
>  	}
>  
> +	/* Get upstream/downstream components' register state */
> +	pcie_get_aspm_reg(parent, &upreg);
> +	child = list_entry(linkbus->devices.next, struct pci_dev, bus_list);
> +	pcie_get_aspm_reg(child, &dwreg);
> +
> +	/*
> +	 * If ASPM not supported, don't mess with the clocks and link,
> +	 * bail out now.
> +	 */
> +	if (!(upreg.support & dwreg.support))
> +		return;
> +
>  	/* Configure common clock before checking latencies */
>  	pcie_aspm_configure_common_clock(link);
>  
> -	/* Get upstream/downstream components' register state */
> +	/*
> +	 * Re-read upstream/downstream components' register state
> +	 * after clock configuration
> +	 */
>  	pcie_get_aspm_reg(parent, &upreg);
> -	child = list_entry(linkbus->devices.next, struct pci_dev, bus_list);
>  	pcie_get_aspm_reg(child, &dwreg);
>  
>  	/*
> -- 
> 1.7.11.7
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-pci" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html