[ofa-general] Re: mlx4 violating radix tree API locking rules?

[Michael S. Tsirkin]

> Quoting Roland Dreier <rdreier at cisco.com>:
> Subject: Re: mlx4 violating radix tree API locking rules?
> 
>  > I guess CQ spinlock implies rcu_read_lock - is that right?
>  > But I do not see any synchronize_rcu calls anywhere in mlx4.
>  > Should destroy QP and friends call synchronize_rcu after
>  > removing the QP from radix tree but before freeing the QP structure?
> 
> Well, I don't think we're really trying to use RCU to synchronize the
> radix tree.  It's the same locking scheme as in mthca, except without
> the home-grown sparse array stuff: we have a qp table lock that
> protects inserting and removing QPs, and then we use the CQ locks to
> avoid looking up a QP that is being removed.
> 
> However, I think you're right: we do violate the radix tree locking
> rules.  So maybe we need to fall back to our own homegrown array stuff
> as in mthca.

Why not just call synchronize_rcu instead?

-- 
MST