[dpdk-dev] kni: fix pci_enable_msix_range error handling
Checks
Commit Message
pci_enable_msix_range's return value has a different meaning than what
was previously returned by pci_enable_msix.
"It returns a negative errno if an error occurs. If it succeeds, it
returns the actual number of interrupts allocated and indicates the
successful configuration of MSI-X capability structure with new
allocated MSI-X interrupts."
The following commit introduced pci_enable_msix_range:
dpdk: 9fb3cd2c ("kni: fix ethtool build with kernel 4.11")
http://dpdk.org/dev/patchwork/patch/24076/
Signed-off-by: Andy Roulin <aroulin@cumulusnetworks.com>
---
lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
Comments
On 12/18/2017 6:16 AM, Andy Roulin wrote:
> pci_enable_msix_range's return value has a different meaning than what
> was previously returned by pci_enable_msix.
>
> "It returns a negative errno if an error occurs. If it succeeds, it
> returns the actual number of interrupts allocated and indicates the
> successful configuration of MSI-X capability structure with new
> allocated MSI-X interrupts."
>
> The following commit introduced pci_enable_msix_range:
> dpdk: 9fb3cd2c ("kni: fix ethtool build with kernel 4.11")
> http://dpdk.org/dev/patchwork/patch/24076/
Hi Andy,
This is updating driver within KNI, and these drivers are used for ethtool
support and most of the driver is already not used at all.
I doubt the updated code is functional at all.
Did you encounter any problem?
These drivers are already outdated, and I am definitely not for actively
maintaining them which is not required for ethtool support.
>
> Signed-off-by: Andy Roulin <aroulin@cumulusnetworks.com>
> ---
> lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c b/lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c
> index e0f427a..32ef3b9 100644
> --- a/lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c
> +++ b/lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c
> @@ -1043,14 +1043,16 @@ static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
> #ifdef HAVE_PCI_ENABLE_MSIX
> err = pci_enable_msix(pdev,
> adapter->msix_entries, numvecs);
> + if (err == 0)
> + break;
> #else
> err = pci_enable_msix_range(pdev,
> adapter->msix_entries,
> numvecs,
> numvecs);
> -#endif
> - if (err == 0)
> + if (err < 0)
> break;
> +#endif
> }
> /* MSI-X failed, so fall through and try MSI */
> dev_warn(pci_dev_to_dev(pdev), "Failed to initialize MSI-X interrupts. "
>
Hi Ferruh,
You can drop it then. I encountered the same porting bug in another
project and came across dpdk's code and patchwork comments which helped me
debugging. I just thought I would contribute the fix back. If it's not
functional then just drop it :)
Andy
On Mon, Feb 5, 2018 at 11:05 AM Ferruh Yigit <ferruh.yigit@intel.com> wrote:
> On 12/18/2017 6:16 AM, Andy Roulin wrote:
> > pci_enable_msix_range's return value has a different meaning than what
> > was previously returned by pci_enable_msix.
> >
> > "It returns a negative errno if an error occurs. If it succeeds, it
> > returns the actual number of interrupts allocated and indicates the
> > successful configuration of MSI-X capability structure with new
> > allocated MSI-X interrupts."
> >
> > The following commit introduced pci_enable_msix_range:
> > dpdk: 9fb3cd2c ("kni: fix ethtool build with kernel 4.11")
> > http://dpdk.org/dev/patchwork/patch/24076/
>
> Hi Andy,
>
> This is updating driver within KNI, and these drivers are used for ethtool
> support and most of the driver is already not used at all.
>
> I doubt the updated code is functional at all.
>
> Did you encounter any problem?
>
> These drivers are already outdated, and I am definitely not for actively
> maintaining them which is not required for ethtool support.
>
> >
> > Signed-off-by: Andy Roulin <aroulin@cumulusnetworks.com>
> > ---
> > lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c | 6 ++++--
> > 1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c
> b/lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c
> > index e0f427a..32ef3b9 100644
> > --- a/lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c
> > +++ b/lib/librte_eal/linuxapp/kni/ethtool/igb/igb_main.c
> > @@ -1043,14 +1043,16 @@ static void igb_set_interrupt_capability(struct
> igb_adapter *adapter, bool msix)
> > #ifdef HAVE_PCI_ENABLE_MSIX
> > err = pci_enable_msix(pdev,
> > adapter->msix_entries,
> numvecs);
> > + if (err == 0)
> > + break;
> > #else
> > err = pci_enable_msix_range(pdev,
> > adapter->msix_entries,
> > numvecs,
> > numvecs);
> > -#endif
> > - if (err == 0)
> > + if (err < 0)
> > break;
> > +#endif
> > }
> > /* MSI-X failed, so fall through and try MSI */
> > dev_warn(pci_dev_to_dev(pdev), "Failed to initialize MSI-X
> interrupts. "
> >
>
>
@@ -1043,14 +1043,16 @@ static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
#ifdef HAVE_PCI_ENABLE_MSIX
err = pci_enable_msix(pdev,
adapter->msix_entries, numvecs);
+ if (err == 0)
+ break;
#else
err = pci_enable_msix_range(pdev,
adapter->msix_entries,
numvecs,
numvecs);
-#endif
- if (err == 0)
+ if (err < 0)
break;
+#endif
}
/* MSI-X failed, so fall through and try MSI */
dev_warn(pci_dev_to_dev(pdev), "Failed to initialize MSI-X interrupts. "