> 
> Hi Ciara,
> thanks for the patch, it looks good to me.
> I only have a minor comment:
> I'd like the requested values to depend only on the current database
> state.  With the current patch when a value is invalid (not pow2 or bigger
> than 4096) we keep the previous one.
> Could you change dpdk_process_queue_size() to return a default value
> (which can be passed as an argument) when the value from the database is
> absent or not valid?
> I guess dpdk_process_queue_size() could return this value directly, instead
> of returning it through a pointer.

Hi Daniele,

Thanks for the review. Can you please clarify your request.
What do you suggest we assign the return value of process_queues() to? If 
requested_size is to reflect the DB value then I assume not that.
The validity checks seem pointless in process_queue_size if we are doing to set 
the requested value regardless. If requested_size reflects the DB value  I see 
two options:
1. Do the pow2 and size checks in reconfigure, before assigning dev->xq_sizes, 
and only assign if valid.
2. Similar to n_rxq, try set up the queue and if it fails, fall back on a known 
good (previous) value. This removes the pow2 etc checks.

Let me know your opinion, or another option if you have it.

Thanks,
Ciara

> 
> 
> 2016-09-22 7:25 GMT-07:00 Ilya Maximets <i.maxim...@samsung.com>:
> I didn't test that, but it looks good to me.
> 
> Acked-by: Ilya Maximets <i.maxim...@samsung.com>
> _______________________________________________
> dev mailing list
> dev@openvswitch.org
> http://openvswitch.org/mailman/listinfo/dev

_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Reply via email to