[PATCH] PCI: Always call pci_enable_acs() regardless of pdev->acs_cap

[PATCH] PCI: Always call pci_enable_acs() regardless of pdev->acs_cap

[Rajat Jain]

Some devices may have have anomalies with the ACS cpability structure,
and they may be using quirks to support ACS functionality via other
registers. For such devices, it is important we always call
pci_enable_acs() to give the quirks a chance to enable ACS in other ways.

For Eg:
There seems a class of Intel devices quirked with *_intel_pch_acs_*
functions, that do not expose the standard ACS capability structure. But
these quirks help support ACS on these devices using other registers:
pci_quirk_enable_intel_pch_acs() -> doesn't use acs_cap to enable ACS

This has already been taken care of in the quirks, in the other direction
i.e. when checking if the ACS is enabled or not. So no need to do
anything there.

Reported-by: Boris V <borisvk@bstnet.org>
Fixes: 52fbf5bdeeef ("PCI: Cache ACS capability offset in device")
Signed-off-by: Rajat Jain <rajatja@google.com>
---
 drivers/pci/pci.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 6d4d5a2f923d..ab398226c55e 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -3516,8 +3516,13 @@ void pci_acs_init(struct pci_dev *dev)
 {
 	dev->acs_cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ACS);
 
-	if (dev->acs_cap)
-		pci_enable_acs(dev);
+	/*
+	 * Attempt to enable ACS regardless of capability because some rootports
+	 * (e.g. the ones quirked with *_intel_pch_acs_*) may not expose
+	 * standard rootport capability structure, but still may support ACS via
+	 * those quirks.
+	 */
+	pci_enable_acs(dev);
 }
 
 /**
-- 
2.29.1.341.ge80a0c044ae-goog

Re: [PATCH] PCI: Always call pci_enable_acs() regardless of pdev->acs_cap

[Rajat Jain]

-random email address (typo)

Re: [PATCH] PCI: Always call pci_enable_acs() regardless of pdev->acs_cap

[Alex Williamson]

On Wed, 28 Oct 2020 16:15:45 -0700
Rajat Jain <rajatja@google.com> wrote:

> Some devices may have have anomalies with the ACS cpability structure,
> and they may be using quirks to support ACS functionality via other
> registers. For such devices, it is important we always call
> pci_enable_acs() to give the quirks a chance to enable ACS in other ways.
> 
> For Eg:
> There seems a class of Intel devices quirked with *_intel_pch_acs_*
> functions, that do not expose the standard ACS capability structure. But
> these quirks help support ACS on these devices using other registers:
> pci_quirk_enable_intel_pch_acs() -> doesn't use acs_cap to enable ACS
> 
> This has already been taken care of in the quirks, in the other direction
> i.e. when checking if the ACS is enabled or not. So no need to do
> anything there.
> 
> Reported-by: Boris V <borisvk@bstnet.org>
> Fixes: 52fbf5bdeeef ("PCI: Cache ACS capability offset in device")
> Signed-off-by: Rajat Jain <rajatja@google.com>
> ---
>  drivers/pci/pci.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 6d4d5a2f923d..ab398226c55e 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -3516,8 +3516,13 @@ void pci_acs_init(struct pci_dev *dev)
>  {
>  	dev->acs_cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ACS);
>  
> -	if (dev->acs_cap)
> -		pci_enable_acs(dev);
> +	/*
> +	 * Attempt to enable ACS regardless of capability because some rootports
> +	 * (e.g. the ones quirked with *_intel_pch_acs_*) may not expose
> +	 * standard rootport capability structure, but still may support ACS via
> +	 * those quirks.
> +	 */
> +	pci_enable_acs(dev);
>  }
>  
>  /**

Much needed regression fix for v5.9:

Reviewed-by: Alex Williamson <alex.williamson@redhat.com>

Re: [PATCH] PCI: Always call pci_enable_acs() regardless of pdev->acs_cap

[Bjorn Helgaas]

[+cc Alex]

On Wed, Oct 28, 2020 at 04:15:45PM -0700, Rajat Jain wrote:
> Some devices may have have anomalies with the ACS cpability structure,
> and they may be using quirks to support ACS functionality via other
> registers. For such devices, it is important we always call
> pci_enable_acs() to give the quirks a chance to enable ACS in other ways.
> 
> For Eg:
> There seems a class of Intel devices quirked with *_intel_pch_acs_*
> functions, that do not expose the standard ACS capability structure. But
> these quirks help support ACS on these devices using other registers:
> pci_quirk_enable_intel_pch_acs() -> doesn't use acs_cap to enable ACS
> 
> This has already been taken care of in the quirks, in the other direction
> i.e. when checking if the ACS is enabled or not. So no need to do
> anything there.
> 
> Reported-by: Boris V <borisvk@bstnet.org>
> Fixes: 52fbf5bdeeef ("PCI: Cache ACS capability offset in device")
> Signed-off-by: Rajat Jain <rajatja@google.com>

Applied to for-linus for v5.10, thanks!

> ---
>  drivers/pci/pci.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 6d4d5a2f923d..ab398226c55e 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -3516,8 +3516,13 @@ void pci_acs_init(struct pci_dev *dev)
>  {
>  	dev->acs_cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ACS);
>  
> -	if (dev->acs_cap)
> -		pci_enable_acs(dev);
> +	/*
> +	 * Attempt to enable ACS regardless of capability because some rootports
> +	 * (e.g. the ones quirked with *_intel_pch_acs_*) may not expose
> +	 * standard rootport capability structure, but still may support ACS via
> +	 * those quirks.
> +	 */
> +	pci_enable_acs(dev);
>  }
>  
>  /**
> -- 
> 2.29.1.341.ge80a0c044ae-goog
>