Re: [dpdk-dev] [RFC PATCH 2/2] ethdev: add capability to keep indirect actions on restart

[Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>]

On 10/7/21 11:16 AM, Dmitry Kozlyuk wrote:
>> -----Original Message-----
>> From: Ajit Khaparde <ajit.khaparde@broadcom.com>
>> Sent: 6 октября 2021 г. 20:13
>> To: Dmitry Kozlyuk <dkozlyuk@nvidia.com>
>> Cc: dpdk-dev <dev@dpdk.org>; Matan Azrad <matan@nvidia.com>; Ori Kam
>> <orika@nvidia.com>; NBU-Contact-Thomas Monjalon
>> <thomas@monjalon.net>; Ferruh Yigit <ferruh.yigit@intel.com>; Andrew
>> Rybchenko <andrew.rybchenko@oktetlabs.ru>
>> Subject: Re: [dpdk-dev] [RFC PATCH 2/2] ethdev: add capability to keep indirect
>> actions on restart
>>
>> On Wed, Sep 1, 2021 at 1:55 AM Dmitry Kozlyuk <dkozlyuk@nvidia.com> wrote:
>>>
>>> rte_flow_action_handle_create() did not mention what happens
>>> with an indirect action when a device is stopped, possibly reconfigured,
>>> and started again. It is natural for some indirect actions to be
>>> persistent, like counters and meters; keeping others just saves
>>> application time and complexity. However, not all PMDs can support it.
>>> It is proposed to add a device capability to indicate if indirect actions
>>> are kept across the above sequence or implicitly destroyed.
>>>
>>> It may happen that in the future a PMD acquires support for a type of
>>> indirect actions that it cannot keep across a restart. It is undesirable
>>> to stop advertising the capability so that applications that don't use
>>> actions of the problematic type can still take advantage of it.
>>> This is why PMDs are allowed to keep only a subset of indirect actions
>>> provided that the vendor mandatorily documents it.
>> Sorry - I am seeing this late.
>> This could become confusing.
>> May be it is better for the PMDs to specify which actions are persistent.
>> How about adding a bit for the possible actions of interest.
>> And then PMDs can set bits for actions which can be persistent across
>> stop, start and reconfigurations?
> 
> This approach was considered, but there is a risk of quickly running out of capability bits. Each action would consume one bit plus as many bits as there are special conditions for it in all the PMDs, because conditions are likely to be PMD-specific. And the application will anyway need to consider specific conditions to know which bit to test, so the meaning of the bits will be PMD-specific. On the other hand, PMDs are not expected to exercise this loophole unless absolutely needed.
> 

May be we should separate at least transfer and non-transfer
rules? Transfer rules are less configuration dependent.