[PATCH] PCI:MSI Return -ENOSPC when requested vectors is not enough

[PATCH] PCI:MSI Return -ENOSPC when requested vectors is not enough

[Dennis Chen]

The __pci_enable_msi_range() should return -ENOSPC instead of -EINVAL
when the device doesn't have enough vectors as required, just as the 
MSI-X vector allocator does in __pci_enable_msix_range(). Otherwise, 
some drivers depending on that return value will probably fallback to
the legacy interrupt directly, for example, in commit 17a51f12cfbd2814
("ahci: only try to use multi-MSI mode if there is more than 1 port"), the
ahci driver will fallback to single MSI mode only when the return value
is -ENOSPC in case of required vectors is not enough, else the driver will
use legacy interrupt which has been observed on a x86 box with 6-port SATA
controller.

With this patch, when a MSI-capable device doesn't have enough MSI
vectors as requested, it will fallback to single MSI mode while not
legacy interrupt.

Signed-off-by: Dennis Chen <dennis.chen@arm.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Tom Long Nguyen <tom.l.nguyen@intel.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Steve Capper <steve.capper@arm.com>
Cc: linux-ide@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
---
 drivers/pci/msi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
index ad70507..da37113 100644
--- a/drivers/pci/msi.c
+++ b/drivers/pci/msi.c
@@ -1084,7 +1084,7 @@ static int __pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec,
 	if (nvec < 0)
 		return nvec;
 	if (nvec < minvec)
-		return -EINVAL;
+		return -ENOSPC;
 
 	if (nvec > maxvec)
 		nvec = maxvec;
-- 
2.7.4


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] PCI:MSI Return -ENOSPC when requested vectors is not enough

[Christoph Hellwig]

Hi Dennis,

I've fixed ahci to treat all errors the same in the meantime, please
try latest Linux tree.  That being said I don't like the different
error returns from __pci_enable_msi_range (and __pci_enable_msix_range),
but they have been there for a while.

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] PCI:MSI Return -ENOSPC when requested vectors is not enough

[Dennis Chen]

On Thu, Dec 01, 2016 at 09:52:43AM +0100, Christoph Hellwig wrote:
> Hi Dennis,
> 
> I've fixed ahci to treat all errors the same in the meantime, please
> try latest Linux tree.  That being said I don't like the different
> error returns from __pci_enable_msi_range (and __pci_enable_msix_range),
> but they have been there for a while.

Ah, I've noticed that you have the fix recently which is somehow to weaken
the necessary of the change. But, that also being said that I don't like we insist
at least the inconsistent either just because something has been there *for a while*.
Both below comments from cpi_alloc_irq_vectors_affinity() and the logic itself
leads us to think that the correct return value is -NOSPC:
/**
 *...
 *Return the number of vectors allocated,
 * (which might be smaller than @max_vecs) if successful, or a negative
 * error code on error. If less than @min_vecs interrupt vectors are
 * available for @dev the function will fail with -ENOSPC.
 * ...
*/

People maybe argue that almost has no device drivers depending on the different
return value, then why we still need to do that?

Thanks,
Dennis 


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] PCI:MSI Return -ENOSPC when requested vectors is not enough

[Bjorn Helgaas]

On Thu, Dec 01, 2016 at 10:15:04AM +0800, Dennis Chen wrote:
> The __pci_enable_msi_range() should return -ENOSPC instead of -EINVAL
> when the device doesn't have enough vectors as required, just as the 
> MSI-X vector allocator does in __pci_enable_msix_range(). Otherwise, 
> some drivers depending on that return value will probably fallback to
> the legacy interrupt directly, for example, in commit 17a51f12cfbd2814
> ("ahci: only try to use multi-MSI mode if there is more than 1 port"), the
> ahci driver will fallback to single MSI mode only when the return value
> is -ENOSPC in case of required vectors is not enough, else the driver will
> use legacy interrupt which has been observed on a x86 box with 6-port SATA
> controller.

Unless Christoph objects, I'll apply this, but I don't understand the
situation with 17a51f12cfbd.  That commit doesn't check for EINVAL or
ENOSPC so I don't know what the connection with this patch is.

I know Christoph said he changed something in ahci to treat all errors
the same, but I don't know where that is, either.

If there's a revision of Linus' tree that is broken, please give the
details so I can at least describe which versions are broken, when it
got fixed by Christoph, and figure out whether we need stable
backports or anything.

Bjorn

> With this patch, when a MSI-capable device doesn't have enough MSI
> vectors as requested, it will fallback to single MSI mode while not
> legacy interrupt.
> 
> Signed-off-by: Dennis Chen <dennis.chen@arm.com>
> Cc: Tejun Heo <tj@kernel.org>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Tom Long Nguyen <tom.l.nguyen@intel.com>
> Cc: Bjorn Helgaas <bhelgaas@google.com>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Marc Zyngier <marc.zyngier@arm.com>
> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Steve Capper <steve.capper@arm.com>
> Cc: linux-ide@vger.kernel.org
> Cc: linux-arm-kernel@lists.infradead.org
> ---
>  drivers/pci/msi.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
> index ad70507..da37113 100644
> --- a/drivers/pci/msi.c
> +++ b/drivers/pci/msi.c
> @@ -1084,7 +1084,7 @@ static int __pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec,
>  	if (nvec < 0)
>  		return nvec;
>  	if (nvec < minvec)
> -		return -EINVAL;
> +		return -ENOSPC;
>  
>  	if (nvec > maxvec)
>  		nvec = maxvec;
> -- 
> 2.7.4
> 
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] PCI:MSI Return -ENOSPC when requested vectors is not enough

[Christoph Hellwig]

On Wed, Jan 11, 2017 at 12:18:53PM -0600, Bjorn Helgaas wrote:
> Unless Christoph objects, I'll apply this, but I don't understand the
> situation with 17a51f12cfbd.  That commit doesn't check for EINVAL or
> ENOSPC so I don't know what the connection with this patch is.

I don't think that commit is the culprit.

> I know Christoph said he changed something in ahci to treat all errors
> the same, but I don't know where that is, either.

"ahci: always fall back to single-MSI mode"

Re: [PATCH] PCI:MSI Return -ENOSPC when requested vectors is not enough

[Bjorn Helgaas]

On Thu, Dec 01, 2016 at 10:15:04AM +0800, Dennis Chen wrote:
> The __pci_enable_msi_range() should return -ENOSPC instead of -EINVAL
> when the device doesn't have enough vectors as required, just as the 
> MSI-X vector allocator does in __pci_enable_msix_range(). Otherwise, 
> some drivers depending on that return value will probably fallback to
> the legacy interrupt directly, for example, in commit 17a51f12cfbd2814
> ("ahci: only try to use multi-MSI mode if there is more than 1 port"), the
> ahci driver will fallback to single MSI mode only when the return value
> is -ENOSPC in case of required vectors is not enough, else the driver will
> use legacy interrupt which has been observed on a x86 box with 6-port SATA
> controller.
> 
> With this patch, when a MSI-capable device doesn't have enough MSI
> vectors as requested, it will fallback to single MSI mode while not
> legacy interrupt.
> 
> Signed-off-by: Dennis Chen <dennis.chen@arm.com>
> Cc: Tejun Heo <tj@kernel.org>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Tom Long Nguyen <tom.l.nguyen@intel.com>
> Cc: Bjorn Helgaas <bhelgaas@google.com>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Marc Zyngier <marc.zyngier@arm.com>
> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Steve Capper <steve.capper@arm.com>
> Cc: linux-ide@vger.kernel.org
> Cc: linux-arm-kernel@lists.infradead.org

Applied to pci/msi for v4.11, thanks, Dennis!

> ---
>  drivers/pci/msi.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
> index ad70507..da37113 100644
> --- a/drivers/pci/msi.c
> +++ b/drivers/pci/msi.c
> @@ -1084,7 +1084,7 @@ static int __pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec,
>  	if (nvec < 0)
>  		return nvec;
>  	if (nvec < minvec)
> -		return -EINVAL;
> +		return -ENOSPC;
>  
>  	if (nvec > maxvec)
>  		nvec = maxvec;
> -- 
> 2.7.4
> 
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel