[PATCH 0/1] net: ethtool: add SmartNIC reset support

[PATCH 0/1] net: ethtool: add SmartNIC reset support

[Scott Branden]

Ethtool provides support for resetting other internal portions of the
NIC already.  Seems appropriate to use one of the bits for resetting
the application processor (AP) for SmartNICs.
If we are not able to use ethtool to reset such portions of the NIC we
will need to create some other API that allows us to do so.  Please
suggest alternatives.
See patch for proposed bit define using ethtool.

Please comment on the best way to add reset support for Next-gen
NIC devices that contain an application processor (AP).

Scott Branden (1):
  net: ethtool: add support for reset of AP inside NIC interface.

 include/uapi/linux/ethtool.h | 1 +
 1 file changed, 1 insertion(+)

-- 
2.5.0

[PATCH 1/1] net: ethtool: add support for reset of AP inside NIC interface.

[Scott Branden]

Add ETH_RESET_AP to reset the application processor inside the NIC
interface.

Signed-off-by: Scott Branden <scott.branden@broadcom.com>
---
 include/uapi/linux/ethtool.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/include/uapi/linux/ethtool.h b/include/uapi/linux/ethtool.h
index 5bd1b1d..f927ab7 100644
--- a/include/uapi/linux/ethtool.h
+++ b/include/uapi/linux/ethtool.h
@@ -1685,6 +1685,7 @@ enum ethtool_reset_flags {
 	ETH_RESET_PHY		= 1 << 6,	/* Transceiver/PHY */
 	ETH_RESET_RAM		= 1 << 7,	/* RAM shared between
 						 * multiple components */
+	ETH_RESET_AP		= 1 << 8,	/* Application processor */
 
 	ETH_RESET_DEDICATED	= 0x0000ffff,	/* All components dedicated to
 						 * this interface */
-- 
2.5.0

Re: [PATCH 0/1] net: ethtool: add SmartNIC reset support

[Andrew Lunn]

On Wed, Oct 18, 2017 at 09:01:35AM -0700, Scott Branden wrote:
> Ethtool provides support for resetting other internal portions of the
> NIC already.  Seems appropriate to use one of the bits for resetting
> the application processor (AP) for SmartNICs.

Hi Scott

Do you also have a management processor on the NIC?

Or is the Application Processor just the Marketing Departments name
for the management processor?

	   Andrew

Re: [PATCH 0/1] net: ethtool: add SmartNIC reset support

[Scott Branden]

Hi Andrew,


On 17-10-18 09:28 AM, Andrew Lunn wrote:
> On Wed, Oct 18, 2017 at 09:01:35AM -0700, Scott Branden wrote:
>> Ethtool provides support for resetting other internal portions of the
>> NIC already.  Seems appropriate to use one of the bits for resetting
>> the application processor (AP) for SmartNICs.
> Hi Scott
>
> Do you also have a management processor on the NIC?
>
> Or is the Application Processor just the Marketing Departments name
> for the management processor?
Yes, there is also a management processor.
In our next gen SmartNIC the "Application processor" may actually
have 8 very powerful cores.
But, if they need to be reset, we reset them all.
>
> 	   Andrew

Re: [PATCH 0/1] net: ethtool: add SmartNIC reset support

[Andrew Lunn]

> Yes, there is also a management processor.

O.K.

Maybe it would be nice to add some more text to the commit message to
make this clear. Define what an application processor is, and how it
differs from a management processor. But othersize:

Reviewed-by: Andrew Lunn <andrew@lunn.ch>

I assume you have another kernel patch to actually make use of this?
It is normal to post the user of a new API in the same series as the
API.

    Andrew

Re: [PATCH 0/1] net: ethtool: add SmartNIC reset support

[Scott Branden]

Hi Andrew,


On 17-10-18 12:16 PM, Andrew Lunn wrote:
>> Yes, there is also a management processor.
> O.K.
>
> Maybe it would be nice to add some more text to the commit message to
> make this clear. Define what an application processor is, and how it
> differs from a management processor. But othersize:
OK -will add more description to differentiate management processor vs. 
application processor(s).
>
> Reviewed-by: Andrew Lunn <andrew@lunn.ch>
>
> I assume you have another kernel patch to actually make use of this?
> It is normal to post the user of a new API in the same series as the
> API.
I actually wanted to get agreement that the bit define could be added to 
ethtool before implementing it in driver.
If this direction approved we'll implement in driver and submit with 
this patch series.
>
>      Andrew
Thanks,
  Scott

Re: [PATCH 0/1] net: ethtool: add SmartNIC reset support

[Andy Gospodarek]

On Wed, Oct 18, 2017 at 12:31:28PM -0700, Scott Branden wrote:
> Hi Andrew,
> 
> 
> On 17-10-18 12:16 PM, Andrew Lunn wrote:
> > > Yes, there is also a management processor.
> > O.K.
> > 
> > Maybe it would be nice to add some more text to the commit message to
> > make this clear. Define what an application processor is, and how it
> > differs from a management processor. But othersize:
> OK -will add more description to differentiate management processor vs.
> application processor(s).
> > 
> > Reviewed-by: Andrew Lunn <andrew@lunn.ch>
> > 
> > I assume you have another kernel patch to actually make use of this?
> > It is normal to post the user of a new API in the same series as the
> > API.
> I actually wanted to get agreement that the bit define could be added to
> ethtool before implementing it in driver.
> If this direction approved we'll implement in driver and submit with this
> patch series.

I just noticed that you did not also post this to netdev@vger.kernel.org.  I
suspect you are more likely to get review and acceptance if that list is cc'd.

I'm not positive that Linus will take networking patches off this list.

> > 
> >      Andrew
> Thanks,
>  Scott

Re: [PATCH 0/1] net: ethtool: add SmartNIC reset support

[Scott Branden]

+netdev@vger.kernel.org


On 17-10-18 02:30 PM, Andy Gospodarek wrote:
> On Wed, Oct 18, 2017 at 12:31:28PM -0700, Scott Branden wrote:
>> Hi Andrew,
>>
>>
>> On 17-10-18 12:16 PM, Andrew Lunn wrote:
>>>> Yes, there is also a management processor.
>>> O.K.
>>>
>>> Maybe it would be nice to add some more text to the commit message to
>>> make this clear. Define what an application processor is, and how it
>>> differs from a management processor. But othersize:
>> OK -will add more description to differentiate management processor vs.
>> application processor(s).
>>> Reviewed-by: Andrew Lunn <andrew@lunn.ch>
>>>
>>> I assume you have another kernel patch to actually make use of this?
>>> It is normal to post the user of a new API in the same series as the
>>> API.
>> I actually wanted to get agreement that the bit define could be added to
>> ethtool before implementing it in driver.
>> If this direction approved we'll implement in driver and submit with this
>> patch series.
> I just noticed that you did not also post this to netdev@vger.kernel.org.  I
> suspect you are more likely to get review and acceptance if that list is cc'd.
>
> I'm not positive that Linus will take networking patches off this list.
>
>>>       Andrew
>> Thanks,
>>   Scott

Re: [PATCH 1/1] net: ethtool: add support for reset of AP inside NIC interface.

[Scott Branden]

+netdev@vger.kernel.org


On 17-10-18 09:01 AM, Scott Branden wrote:
> Add ETH_RESET_AP to reset the application processor inside the NIC
> interface.
>
> Signed-off-by: Scott Branden <scott.branden@broadcom.com>
> ---
>   include/uapi/linux/ethtool.h | 1 +
>   1 file changed, 1 insertion(+)
>
> diff --git a/include/uapi/linux/ethtool.h b/include/uapi/linux/ethtool.h
> index 5bd1b1d..f927ab7 100644
> --- a/include/uapi/linux/ethtool.h
> +++ b/include/uapi/linux/ethtool.h
> @@ -1685,6 +1685,7 @@ enum ethtool_reset_flags {
>   	ETH_RESET_PHY		= 1 << 6,	/* Transceiver/PHY */
>   	ETH_RESET_RAM		= 1 << 7,	/* RAM shared between
>   						 * multiple components */
> +	ETH_RESET_AP		= 1 << 8,	/* Application processor */
>   
>   	ETH_RESET_DEDICATED	= 0x0000ffff,	/* All components dedicated to
>   						 * this interface */