Re: [PATCH net-next 4/6] rtnetlink: Add support for SyncE recovered clock configuration

[Ido Schimmel <idosch@idosch.org>]

On Fri, Nov 05, 2021 at 12:17:19PM +0000, Machnikowski, Maciej wrote:
> 
> 
> > -----Original Message-----
> > From: Roopa Prabhu <roopa@nvidia.com>
> > Sent: Thursday, November 4, 2021 7:25 PM
> > To: Machnikowski, Maciej <maciej.machnikowski@intel.com>;
> > netdev@vger.kernel.org; intel-wired-lan@lists.osuosl.org
> > Cc: richardcochran@gmail.com; abyagowi@fb.com; Nguyen, Anthony L
> > <anthony.l.nguyen@intel.com>; davem@davemloft.net; kuba@kernel.org;
> > linux-kselftest@vger.kernel.org; idosch@idosch.org; mkubecek@suse.cz;
> > saeed@kernel.org; michael.chan@broadcom.com
> > Subject: Re: [PATCH net-next 4/6] rtnetlink: Add support for SyncE recovered
> > clock configuration
> > 
> > 
> > On 11/4/21 1:12 AM, Maciej Machnikowski wrote:
> > > Add support for RTNL messages for reading/configuring SyncE recovered
> > > clocks.
> > > The messages are:
> > > RTM_GETRCLKRANGE: Reads the allowed pin index range for the
> > recovered
> > > 		  clock outputs. This can be aligned to PHY outputs or
> > > 		  to EEC inputs, whichever is better for a given
> > > 		  application
> > >
> > > RTM_GETRCLKSTATE: Read the state of recovered pins that output
> > recovered
> > > 		  clock from a given port. The message will contain the
> > > 		  number of assigned clocks (IFLA_RCLK_STATE_COUNT) and
> > > 		  a N pin inexes in IFLA_RCLK_STATE_OUT_IDX
> > >
> > > RTM_SETRCLKSTATE: Sets the redirection of the recovered clock for
> > > 		  a given pin
> > >
> > > Signed-off-by: Maciej Machnikowski <maciej.machnikowski@intel.com>
> > > ---
> > 
> > 
> > Can't we just use a single RTM msg with nested attributes ?
> > 
> > With separate RTM msgtype for each syncE attribute we will end up
> > bloating the RTM msg namespace.
> > 
> > (these api's could also be in ethtool given its directly querying the
> > drivers)
> 
> I'm not a fan of merging those messages. The mergeable ones are
> GETRCLKRANGE and GETCLKSTATE, but the get range function may be
> result in a significantly longer call if the information about the underlying
> HW require any HW calls.
> They are currently grouped in 3 categories:
> - reading the boundaries in GetRclkRange (we can later add more to it, like
> 	configurable frequency limits)
> - Reading current configuration
> - setting the required configuration
> 
> I don't plan adding any additional RTM msg types for those (and that's
> the reason why I pulled them before EEC state which may have more
> messages in the future)
> 
> We also discussed ethtool way in the past RFCs, but this concept
> is applicable to any transport layer so I'd rather not limit it to ethernet
> devices (i.e. SONET, infiniband and others).

I'm still not convinced that this doesn't belong in ethtool. I find it
very weird to have a message called "Get Ethernet Equipment Clock State"
in rtnetlink and not in ethtool. I also believe it was a mistake to add
DCB to rtnetlink (for example, why PAUSE is configured via ethtool, but
PFC via rtnetlink?)