Re: [RFC PATCH net-next] net: sched: flower: refactor reoffload for concurrent access

[Vlad Buslov <vladbu@mellanox.com>]

On Thu 18 Apr 2019 at 21:02, Jakub Kicinski <jakub.kicinski@netronome.com> wrote:
> On Thu, 18 Apr 2019 17:58:26 +0000, Vlad Buslov wrote:
>> On Thu 18 Apr 2019 at 20:46, Jakub Kicinski <jakub.kicinski@netronome.com> wrote:
>> > On Thu, 18 Apr 2019 16:33:22 +0000, Vlad Buslov wrote:  
>> >> Considering this, I tried to improve my solution to remove possibility
>> >> of multiple adds of same filter and it seems to me that it would be
>> >> enough to move hw_filters list management in flower offloads functions:
>> >> add filter to list while holding rtnl lock in fl_hw_replace_filter() and
>> >> remove it from list while holding rtnl lock in fl_hw_destroy_filter().
>> >> What do you think?  
>> >
>> > Sounds good for now, but I presume the plan is to get rid of rtnl
>> > around the driver call.. at which point we would switch to a rwlock? :)  
>> 
>> Yes, but I would like the lock to be in cls hw offloads API
>> (tc_setup_cb_replace(), etc) and not in flower itself. That would also
>> solve deadlock issue and make code reusable for any further unlocked
>> classifiers implementations.
>
> And then the HW list goes along with it into the common code?
> That'd be quite nice.

The goal is to have a lock in tcf_block and use it synchronize cb_list
and all related counters (offloadcnt, etc). Now I also want to use it to
protect hw_filters list and prevent the double-add issue. Meanwhile rtnl
lock can do the job.