Re: [PATCH net-next RFC v4 04/15] devlink: Add reload actions stats to dev get

[Jiri Pirko]

Fri, Sep 18, 2020 at 06:13:59PM CEST, moshe@xxxxxxxxxx wrote:
>
>On 9/15/2020 11:33 PM, Moshe Shemesh wrote:
>> External email: Use caution opening links or attachments
>> 
>> 
>> On 9/15/2020 4:34 PM, Jiri Pirko wrote:
>> > Tue, Sep 15, 2020 at 02:31:38PM CEST, moshe@xxxxxxxxxx wrote:
>> > > On 9/15/2020 10:44 AM, Jiri Pirko wrote:
>> > > > Tue, Sep 15, 2020 at 08:45:19AM CEST, idosch@xxxxxxxxxx wrote:
>> > > > > On Mon, Sep 14, 2020 at 03:45:00PM +0200, Jiri Pirko wrote:
>> > > > > > Mon, Sep 14, 2020 at 08:07:51AM CEST, moshe@xxxxxxxxxxxx wrote:
>> > > > > > > Expose devlink reload actions stats to the user through devlink dev
>> > > > > > > get command.
>> > > > > > > 
>> > > > > > > Examples:
>> > > > > > > $ devlink dev show
>> > > > > > > pci/0000:82:00.0:
>> > > > > > >    reload_action_stats:
>> > > > > > >      driver_reinit 2
>> > > > > > >      fw_activate 1
>> > > > > > >      driver_reinit_no_reset 0
>> > > > > > >      fw_activate_no_reset 0
>> > > > > > > pci/0000:82:00.1:
>> > > > > > >    reload_action_stats:
>> > > > > > >      driver_reinit 1
>> > > > > > >      fw_activate 1
>> > > > > > >      driver_reinit_no_reset 0
>> > > > > > >      fw_activate_no_reset 0
>> > > > > > I would rather have something like:
>> > > > > >      stats:
>> > > > > >        reload_action:
>> > > > > >          driver_reinit 1
>> > > > > >          fw_activate 1
>> > > > > >          driver_reinit_no_reset 0
>> > > > > >          fw_activate_no_reset 0
>> > > > > > 
>> > > > > > Then we can easily extend and add other stats in the tree.
>> > > 
>> > > Sure, I will add it.
>> > Could you please checkout the metrics patchset and figure out how to
>> > merge that with your usecase?
>> > 
>> 
>> I will check, I will discuss with Ido how it will fit.
>> 
>
>I have discussed it with Ido, it doesn't fit to merge with metrics:
>
>1. These counters are maintained by devlink unlike metrics which are read by
>the driver from HW.

Okay.

>
>2. The metrics counters push string name, while here I use enum.
>
>However, I did add another level as you suggested here for option to future
>stats that may fit.
>
>> > > > > > Also, I wonder if these stats could be somehow merged
>> > > > > > with Ido's metrics
>> > > > > > work:
>> > > > > > https://github.com/idosch/linux/commits/submit/devlink_metric_rfc_v1
>> > > > > > 
>> > > > > > Ido, would it make sense?
>> > > > > I guess. My original idea for devlink-metric was to expose
>> > > > > design-specific metrics to user space where the entity
>> > > > > registering the
>> > > > > metrics is the device driver. In this case the entity
>> > > > > would be devlink
>> > > > > itself and it would be auto-registered for each device.
>> > > > Yeah, the usecase is different, but it is still stats, right.
>> > > > 
>> > > > 
>> > > > > > > $ devlink dev show -jp
>> > > > > > > {
>> > > > > > >      "dev": {
>> > > > > > >          "pci/0000:82:00.0": {
>> > > > > > >              "reload_action_stats": [ {
>> > > > > > >                      "driver_reinit": 2
>> > > > > > >                  },{
>> > > > > > >                      "fw_activate": 1
>> > > > > > >                  },{
>> > > > > > >                      "driver_reinit_no_reset": 0
>> > > > > > >                  },{
>> > > > > > >                      "fw_activate_no_reset": 0
>> > > > > > >                  } ]
>> > > > > > >          },
>> > > > > > >          "pci/0000:82:00.1": {
>> > > > > > >              "reload_action_stats": [ {
>> > > > > > >                      "driver_reinit": 1
>> > > > > > >                  },{
>> > > > > > >                      "fw_activate": 1
>> > > > > > >                  },{
>> > > > > > >                      "driver_reinit_no_reset": 0
>> > > > > > >                  },{
>> > > > > > >                      "fw_activate_no_reset": 0
>> > > > > > >                  } ]
>> > > > > > >          }
>> > > > > > >      }
>> > > > > > > }
>> > > > > > > 
>> > > > > > [..]