Re: [PATCH net-next RFC 01/13] devlink: Add reload level option to devlink reload command

[Moshe Shemesh <moshe@mellanox.com>]

On 7/30/2020 12:07 AM, Jakub Kicinski wrote:
> On Wed, 29 Jul 2020 17:54:08 +0300 Moshe Shemesh wrote:
>> On 7/28/2020 11:06 PM, Jakub Kicinski wrote:
>>> On Tue, 28 Jul 2020 12:18:30 -0700 Jacob Keller wrote:
>>>> On 7/28/2020 11:44 AM, Jakub Kicinski wrote:
>>>>>   From user perspective what's important is what the reset achieves (and
>>>>> perhaps how destructive it is). We can define the reset levels as:
>>>>>
>>>>> $ devlink dev reload pci/0000:82:00.0 net-ns-respawn
>>>>> $ devlink dev reload pci/0000:82:00.0 driver-param-init
>>>>> $ devlink dev reload pci/0000:82:00.0 fw-activate
>>>>>
>>>>> combining should be possible when user wants multiple things to happen:
>>>>>
>>>>> $ devlink dev reload pci/0000:82:00.0 fw-activate driver-param-init
>>>> Where today "driver-param-init" is the default behavior. But didn't we
>>>> just say that mlxsw also does the equivalent of fw-activate?
>>> Actually the default should probably be the combination of
>>> driver-param-init and net-ns-respawn.
>> What about the support of these combinations, one device needs to reset
>> fw to apply the param init, while another device can apply param-init
>> without fw reset, but has to reload the driver for fw-reset.
>>
>> So the support per driver will be a matrix of combinations ?
> Note that there is no driver reload in my examples, driver reload is
> likely not user's goal. Whatever the driver needs to reset to satisfy
> the goal is fair game IMO.


Actually, driver-param-init (cmode driverinit) implicit driver 
re-initialization.

> It's already the case that some drivers reset FW for param init and some
> don't and nobody is complaining.


Right, driver may need more than driver re-initialization for 
driver-param-init, but I think that driver re-initialization is the 
minimum for driver-param-init.

> We should treat constraints separate (in this set we have the live
> activation which is a constraint on the reload operation).
>
>>> My expectations would be that the driver must perform the lowest
>>> reset level possible that satisfies the requested functional change.
>>> IOW driver may do more, in fact it should be acceptable for the
>>> driver to always for a full HW reset (unless --live or other
>>> constraint is specified).
>> OK, but some combinations may still not be valid for specific driver
>> even if it tries lowest level possible.
> Can you give an example?


For example take the combination of fw-live-patch and param-init.

The fw-live-patch needs no re-initialization, while the param-init 
requires driver re-initialization.

So the only way to do that is to the one command after the other, not 
really combining.

Other combination, as fw-atcivate and param-init may not be valid for a 
specific driver as it doesn't support one of them and so can't even run 
one after the other.