Re: [Intel-wired-lan] [net-next PATCH 3/3] net: Add support for XPS with QoS via traffic classes

[John Fastabend <john.fastabend@gmail.com>]

On 16-10-28 07:58 AM, Alexander Duyck wrote:
> On Thu, Oct 27, 2016 at 7:38 PM, Tom Herbert <tom@herbertland.com> wrote:
>> On Thu, Oct 27, 2016 at 8:40 AM, Alexander Duyck
>> <alexander.h.duyck@intel.com> wrote:
>>> This patch adds support for setting and using XPS when QoS via traffic
>>> classes is enabled.  With this change we will factor in the priority and
>>> traffic class mapping of the packet and use that information to correctly
>>> select the queue.
>>>
>>> This allows us to define a set of queues for a given traffic class via
>>> mqprio and then configure the XPS mapping for those queues so that the
>>> traffic flows can avoid head-of-line blocking between the individual CPUs
>>> if so desired.
>>>
>> Does this change the sys API for XPS? Is it up the user to know which
>> are priority queues in sys?
> 
> The idea was to keep the change transparent.  So for now the only
> change in relation to XPS from the XPS point of view is that the map
> for a given queue is invalidated when either the dev->num_tcs changes
> or if the queue is moved into a dev->tx_to_txq mapping.  Otherwise the
> interface should behave exactly the same as before.
> 
> One thing I could look at doing is adding a read-only sysfs value that
> the user could use to identify which traffic class a given queue
> belongs to.  Then at least that way they would be able to dump both
> the XPS map and the tc to determine how the traffic will flow through
> the device.
> 

I could see some value in a sysfs read-only tc-queue mapping might be
especially for devices that are negotiating these things using firmware.

.John