Re: [PATCH net] net: page_pool: don't try to stash the napi id

[Jakub Kicinski <kuba@kernel.org>]

On Mon, 27 Jan 2025 14:31:10 +0100 Toke Høiland-Jørgensen wrote:
> > +
> > +/* page_pool_destroy or page_pool_disable_direct_recycling must be
> > called before
> > + * netif_napi_del if pool->p.napi is set.
> > + */

FWIW the comment is better placed on the warning, that's where people
will look when they hit it ;)

> >  void page_pool_destroy(struct page_pool *pool);
> >  void page_pool_use_xdp_mem(struct page_pool *pool, void (*disconnect)(void *),
> >                            const struct xdp_mem_info *mem);
> >
> > diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> > index 5c4b788b811b..dc82767b2516 100644
> > --- a/net/core/page_pool.c
> > +++ b/net/core/page_pool.c
> > @@ -1161,6 +1161,8 @@ void page_pool_destroy(struct page_pool *pool)
> >         if (!page_pool_put(pool))
> >                 return;
> >
> > +       DEBUG_NET_WARN_ON(pool->p.napi && !napi_is_valid(pool->p.napi));

IDK what "napi_is_valid()" is. I think like this:

diff --git a/net/core/page_pool.c b/net/core/page_pool.c
index a3de752c5178..837ed36472db 100644
--- a/net/core/page_pool.c
+++ b/net/core/page_pool.c
@@ -1145,6 +1145,7 @@ void page_pool_disable_direct_recycling(struct page_pool *pool)
         * pool and NAPI are unlinked when NAPI is disabled.
         */
        WARN_ON(!test_bit(NAPI_STATE_SCHED, &pool->p.napi->state));
+       WARN_ON(!test_bit(NAPI_STATE_LISTED, &pool->p.napi->state));
        WARN_ON(READ_ONCE(pool->p.napi->list_owner) != -1);
 
        WRITE_ONCE(pool->p.napi, NULL);

Because page_pool_disable_direct_recycling() must also be called while
NAPI is listed. Technically we should also sync rcu if the driver calls
this directly, because NAPI may be reused :(

> >         page_pool_disable_direct_recycling(pool);
> >         page_pool_free_frag(pool);  
> 
> Yeah, good idea; care to send a proper patch? :)

...for net-next ? :)