[v3,0/2] IOPL related fixes

  Patches primary targeted to fix OVS unit test failures with
DPDK 18.11 due to following error:
    'IOPL call failed - cannot use virtio PMD'.

We can avoid test failures in OVS by filtering the output like this:
  https://patchwork.ozlabs.org/project/openvswitch/list/?series=77706
but it still looks very inconvenient for me to have this
message in the output of every command for the DPDK linked app.

Version 3:
    * Splitted in two patches.      [David Marchand]
    * Fixed typo in commit message. [David Marchand]

Version 2:
    * Fixed possible fd leak on BSD.

Ilya Maximets (2):
  eal/bsd: fix possible IOPL fd leak
  net/virtio: avoid annoying IOPL call related errors

 drivers/net/virtio/virtio_ethdev.c | 11 ++++++-----
 lib/librte_eal/bsdapp/eal/eal.c    |  6 ++++--
 2 files changed, 10 insertions(+), 7 deletions(-)
  

Hi,

On 11/23/18 4:39 PM, Ilya Maximets wrote:
> Patches primary targeted to fix OVS unit test failures with
> DPDK 18.11 due to following error:
>      'IOPL call failed - cannot use virtio PMD'.

You mention v18.11, do you mean this is a regression?

> 
> We can avoid test failures in OVS by filtering the output like this:
>    https://patchwork.ozlabs.org/project/openvswitch/list/?series=77706
> but it still looks very inconvenient for me to have this
> message in the output of every command for the DPDK linked app.
> 
> Version 3:
>      * Splitted in two patches.      [David Marchand]
>      * Fixed typo in commit message. [David Marchand]
> 
> Version 2:
>      * Fixed possible fd leak on BSD.
> 
> Ilya Maximets (2):
>    eal/bsd: fix possible IOPL fd leak
>    net/virtio: avoid annoying IOPL call related errors
> 
>   drivers/net/virtio/virtio_ethdev.c | 11 ++++++-----
>   lib/librte_eal/bsdapp/eal/eal.c    |  6 ++++--
>   2 files changed, 10 insertions(+), 7 deletions(-)
>
  

On 23.11.2018 18:41, Maxime Coquelin wrote:
> Hi,
> 
> On 11/23/18 4:39 PM, Ilya Maximets wrote:
>> Patches primary targeted to fix OVS unit test failures with
>> DPDK 18.11 due to following error:
>>      'IOPL call failed - cannot use virtio PMD'.
> 
> You mention v18.11, do you mean this is a regression?

Kind of. But not really a bug. It's just a message that shows up
every time you starting the app as a non-root user.

The message itself was introduced long time ago, but it wasn't
printed for unclear reasons. It's probably some change in logging
subsystem uncovered it.

I just checked and found that message appears starting from v18.02.
But OVS stuck with 17.11 LTS where this message exists, but not
printed for some reason. That's why this wasn't an issue previously.

P.S. OVS tests just checks the output for error messages.
     Tests themselves works fine. They are not actually using DPDK.

Best regards, Ilya Maximets.

> 
>>
>> We can avoid test failures in OVS by filtering the output like this:
>>    https://patchwork.ozlabs.org/project/openvswitch/list/?series=77706
>> but it still looks very inconvenient for me to have this
>> message in the output of every command for the DPDK linked app.
>>
>> Version 3:
>>      * Splitted in two patches.      [David Marchand]
>>      * Fixed typo in commit message. [David Marchand]
>>
>> Version 2:
>>      * Fixed possible fd leak on BSD.
>>
>> Ilya Maximets (2):
>>    eal/bsd: fix possible IOPL fd leak
>>    net/virtio: avoid annoying IOPL call related errors
>>
>>   drivers/net/virtio/virtio_ethdev.c | 11 ++++++-----
>>   lib/librte_eal/bsdapp/eal/eal.c    |  6 ++++--
>>   2 files changed, 10 insertions(+), 7 deletions(-)
>>
> 
>
  

On 11/23/18 4:53 PM, Ilya Maximets wrote:
> On 23.11.2018 18:41, Maxime Coquelin wrote:
>> Hi,
>>
>> On 11/23/18 4:39 PM, Ilya Maximets wrote:
>>> Patches primary targeted to fix OVS unit test failures with
>>> DPDK 18.11 due to following error:
>>>       'IOPL call failed - cannot use virtio PMD'.
>>
>> You mention v18.11, do you mean this is a regression?
> 
> Kind of. But not really a bug. It's just a message that shows up
> every time you starting the app as a non-root user.
> 
> The message itself was introduced long time ago, but it wasn't
> printed for unclear reasons. It's probably some change in logging
> subsystem uncovered it.
> 
> I just checked and found that message appears starting from v18.02.
> But OVS stuck with 17.11 LTS where this message exists, but not
> printed for some reason. That's why this wasn't an issue previously.

Thanks for the clarification.
In that case, we may want it in v18.11 LTS, but I think it can wait or
v18.11.1.

Once in the next-virtio tree, I'll send it to stable@dpdk.org.

Regards,
Maxime

> P.S. OVS tests just checks the output for error messages.
>       Tests themselves works fine. They are not actually using DPDK.
> 
> Best regards, Ilya Maximets.
> 
>>
>>>
>>> We can avoid test failures in OVS by filtering the output like this:
>>>     https://patchwork.ozlabs.org/project/openvswitch/list/?series=77706
>>> but it still looks very inconvenient for me to have this
>>> message in the output of every command for the DPDK linked app.
>>>
>>> Version 3:
>>>       * Splitted in two patches.      [David Marchand]
>>>       * Fixed typo in commit message. [David Marchand]
>>>
>>> Version 2:
>>>       * Fixed possible fd leak on BSD.
>>>
>>> Ilya Maximets (2):
>>>     eal/bsd: fix possible IOPL fd leak
>>>     net/virtio: avoid annoying IOPL call related errors
>>>
>>>    drivers/net/virtio/virtio_ethdev.c | 11 ++++++-----
>>>    lib/librte_eal/bsdapp/eal/eal.c    |  6 ++++--
>>>    2 files changed, 10 insertions(+), 7 deletions(-)
>>>
>>
>>
  

On 11/23/2018 04:25 PM, Maxime Coquelin wrote:
> 
> 
> On 11/23/18 4:53 PM, Ilya Maximets wrote:
>> On 23.11.2018 18:41, Maxime Coquelin wrote:
>>> Hi,
>>>
>>> On 11/23/18 4:39 PM, Ilya Maximets wrote:
>>>> Patches primary targeted to fix OVS unit test failures with
>>>> DPDK 18.11 due to following error:
>>>>       'IOPL call failed - cannot use virtio PMD'.
>>>
>>> You mention v18.11, do you mean this is a regression?
>>
>> Kind of. But not really a bug. It's just a message that shows up
>> every time you starting the app as a non-root user.
>>
>> The message itself was introduced long time ago, but it wasn't
>> printed for unclear reasons. It's probably some change in logging
>> subsystem uncovered it.
>>
>> I just checked and found that message appears starting from v18.02.
>> But OVS stuck with 17.11 LTS where this message exists, but not
>> printed for some reason. That's why this wasn't an issue previously.
> 
> Thanks for the clarification.
> In that case, we may want it in v18.11 LTS, but I think it can wait or
> v18.11.1.
> 

It would be better if it can be in 18.11. Upstream OVS will point at a
release tarball, and at present an 18.11.1 release is not due until
mid-January. If it's too late for 18.11, alternatively we could do an
earlier 18.11.1 release.

> Once in the next-virtio tree, I'll send it to stable@dpdk.org.
> 
> Regards,
> Maxime
> 
>> P.S. OVS tests just checks the output for error messages.
>>       Tests themselves works fine. They are not actually using DPDK.
>>
>> Best regards, Ilya Maximets.
>>
>>>
>>>>
>>>> We can avoid test failures in OVS by filtering the output like this:
>>>>     https://patchwork.ozlabs.org/project/openvswitch/list/?series=77706
>>>> but it still looks very inconvenient for me to have this
>>>> message in the output of every command for the DPDK linked app.
>>>>
>>>> Version 3:
>>>>       * Splitted in two patches.      [David Marchand]
>>>>       * Fixed typo in commit message. [David Marchand]
>>>>
>>>> Version 2:
>>>>       * Fixed possible fd leak on BSD.
>>>>
>>>> Ilya Maximets (2):
>>>>     eal/bsd: fix possible IOPL fd leak
>>>>     net/virtio: avoid annoying IOPL call related errors
>>>>
>>>>    drivers/net/virtio/virtio_ethdev.c | 11 ++++++-----
>>>>    lib/librte_eal/bsdapp/eal/eal.c    |  6 ++++--
>>>>    2 files changed, 10 insertions(+), 7 deletions(-)
>>>>
>>>
>>>
  

23/11/2018 21:07, Kevin Traynor:
> On 11/23/2018 04:25 PM, Maxime Coquelin wrote:
> > 
> > 
> > On 11/23/18 4:53 PM, Ilya Maximets wrote:
> >> On 23.11.2018 18:41, Maxime Coquelin wrote:
> >>> Hi,
> >>>
> >>> On 11/23/18 4:39 PM, Ilya Maximets wrote:
> >>>> Patches primary targeted to fix OVS unit test failures with
> >>>> DPDK 18.11 due to following error:
> >>>>       'IOPL call failed - cannot use virtio PMD'.
> >>>
> >>> You mention v18.11, do you mean this is a regression?
> >>
> >> Kind of. But not really a bug. It's just a message that shows up
> >> every time you starting the app as a non-root user.
> >>
> >> The message itself was introduced long time ago, but it wasn't
> >> printed for unclear reasons. It's probably some change in logging
> >> subsystem uncovered it.
> >>
> >> I just checked and found that message appears starting from v18.02.
> >> But OVS stuck with 17.11 LTS where this message exists, but not
> >> printed for some reason. That's why this wasn't an issue previously.
> > 
> > Thanks for the clarification.
> > In that case, we may want it in v18.11 LTS, but I think it can wait or
> > v18.11.1.
> 
> It would be better if it can be in 18.11. Upstream OVS will point at a
> release tarball, and at present an 18.11.1 release is not due until
> mid-January. If it's too late for 18.11, alternatively we could do an
> earlier 18.11.1 release.

I won't take any risk at this stage.
The memory of the headaches we had in the past with iopl is too much strong.
The real change of this patch is to register the PMD even without successful iopl.
I would like to be sure it won't cause any regression.
  

23/11/2018 23:29, Thomas Monjalon:
> 23/11/2018 21:07, Kevin Traynor:
> > On 11/23/2018 04:25 PM, Maxime Coquelin wrote:
> > > On 11/23/18 4:53 PM, Ilya Maximets wrote:
> > >> On 23.11.2018 18:41, Maxime Coquelin wrote:
> > >>> On 11/23/18 4:39 PM, Ilya Maximets wrote:
> > >>>> Patches primary targeted to fix OVS unit test failures with
> > >>>> DPDK 18.11 due to following error:
> > >>>>       'IOPL call failed - cannot use virtio PMD'.
> > >>>
> > >>> You mention v18.11, do you mean this is a regression?
> > >>
> > >> Kind of. But not really a bug. It's just a message that shows up
> > >> every time you starting the app as a non-root user.
> > >>
> > >> The message itself was introduced long time ago, but it wasn't
> > >> printed for unclear reasons. It's probably some change in logging
> > >> subsystem uncovered it.
> > >>
> > >> I just checked and found that message appears starting from v18.02.
> > >> But OVS stuck with 17.11 LTS where this message exists, but not
> > >> printed for some reason. That's why this wasn't an issue previously.
> > > 
> > > Thanks for the clarification.
> > > In that case, we may want it in v18.11 LTS, but I think it can wait or
> > > v18.11.1.
> > 
> > It would be better if it can be in 18.11. Upstream OVS will point at a
> > release tarball, and at present an 18.11.1 release is not due until
> > mid-January. If it's too late for 18.11, alternatively we could do an
> > earlier 18.11.1 release.
> 
> I won't take any risk at this stage.
> The memory of the headaches we had in the past with iopl is too much strong.
> The real change of this patch is to register the PMD even without successful iopl.
> I would like to be sure it won't cause any regression.

After more thinking, I decide to take the risk.
Take it as a gift to OVS :-)

Series applied
  

On 11/25/2018 10:42 AM, Thomas Monjalon wrote:
> 23/11/2018 23:29, Thomas Monjalon:
>> 23/11/2018 21:07, Kevin Traynor:
>>> On 11/23/2018 04:25 PM, Maxime Coquelin wrote:
>>>> On 11/23/18 4:53 PM, Ilya Maximets wrote:
>>>>> On 23.11.2018 18:41, Maxime Coquelin wrote:
>>>>>> On 11/23/18 4:39 PM, Ilya Maximets wrote:
>>>>>>> Patches primary targeted to fix OVS unit test failures with
>>>>>>> DPDK 18.11 due to following error:
>>>>>>>       'IOPL call failed - cannot use virtio PMD'.
>>>>>>
>>>>>> You mention v18.11, do you mean this is a regression?
>>>>>
>>>>> Kind of. But not really a bug. It's just a message that shows up
>>>>> every time you starting the app as a non-root user.
>>>>>
>>>>> The message itself was introduced long time ago, but it wasn't
>>>>> printed for unclear reasons. It's probably some change in logging
>>>>> subsystem uncovered it.
>>>>>
>>>>> I just checked and found that message appears starting from v18.02.
>>>>> But OVS stuck with 17.11 LTS where this message exists, but not
>>>>> printed for some reason. That's why this wasn't an issue previously.
>>>>
>>>> Thanks for the clarification.
>>>> In that case, we may want it in v18.11 LTS, but I think it can wait or
>>>> v18.11.1.
>>>
>>> It would be better if it can be in 18.11. Upstream OVS will point at a
>>> release tarball, and at present an 18.11.1 release is not due until
>>> mid-January. If it's too late for 18.11, alternatively we could do an
>>> earlier 18.11.1 release.
>>
>> I won't take any risk at this stage.
>> The memory of the headaches we had in the past with iopl is too much strong.
>> The real change of this patch is to register the PMD even without successful iopl.
>> I would like to be sure it won't cause any regression.
> 
> After more thinking, I decide to take the risk.
> Take it as a gift to OVS :-)
> 

ok, thanks :-) If there's any subsequent changes needed, I will get them
on an 18.11 stable branch straight away.

> Series applied
> 
>
  

On Sun, Nov 25, 2018 at 11:42 AM Thomas Monjalon <thomas@monjalon.net>
wrote:

> 23/11/2018 23:29, Thomas Monjalon:
> > 23/11/2018 21:07, Kevin Traynor:
> > > On 11/23/2018 04:25 PM, Maxime Coquelin wrote:
> > > > On 11/23/18 4:53 PM, Ilya Maximets wrote:
> > > >> On 23.11.2018 18:41, Maxime Coquelin wrote:
> > > >>> On 11/23/18 4:39 PM, Ilya Maximets wrote:
> > > >>>> Patches primary targeted to fix OVS unit test failures with
> > > >>>> DPDK 18.11 due to following error:
> > > >>>>       'IOPL call failed - cannot use virtio PMD'.
> > > >>>
> > > >>> You mention v18.11, do you mean this is a regression?
> > > >>
> > > >> Kind of. But not really a bug. It's just a message that shows up
> > > >> every time you starting the app as a non-root user.
> > > >>
> > > >> The message itself was introduced long time ago, but it wasn't
> > > >> printed for unclear reasons. It's probably some change in logging
> > > >> subsystem uncovered it.
> > > >>
> > > >> I just checked and found that message appears starting from v18.02.
> > > >> But OVS stuck with 17.11 LTS where this message exists, but not
> > > >> printed for some reason. That's why this wasn't an issue previously.
> > > >
> > > > Thanks for the clarification.
> > > > In that case, we may want it in v18.11 LTS, but I think it can wait
> or
> > > > v18.11.1.
> > >
> > > It would be better if it can be in 18.11. Upstream OVS will point at a
> > > release tarball, and at present an 18.11.1 release is not due until
> > > mid-January. If it's too late for 18.11, alternatively we could do an
> > > earlier 18.11.1 release.
> >
> > I won't take any risk at this stage.
> > The memory of the headaches we had in the past with iopl is too much
> strong.
> > The real change of this patch is to register the PMD even without
> successful iopl.
> > I would like to be sure it won't cause any regression.
>
> After more thinking, I decide to take the risk.
> Take it as a gift to OVS :-)
>

Just for the record, this log is visible by default since 18.02 with commit:
0062818856c3 ("net/virtio: implement dynamic logging")

So probably worth backporting in 18.08 at least.