> On 6/2/2026 2:21 AM, Lorenzo Bianconi wrote:
> > airoha_metadata_dst_free() runs metadata_dst_free() which frees the
> > metadata_dst with kfree() immediately, bypassing the RCU grace period.
> > In the RX path, skb_dst_set_noref() sets a non-refcounted pointer from
> > the skb to the metadata_dst. This function requires RCU read-side
> > protection and the dst must remain valid until all RCU readers complete.
> > Since metadata_dst_free() calls kfree() directly, an use-after-free can
> > occur if any skb still holds a noref pointer to the dst when the driver
> > tears it down.
> > Replace metadata_dst_free() with dst_release() which properly goes
> > through the refcount path: when the refcount drops to zero, it schedules
> > the actual free via call_rcu_hurry(), ensuring all RCU readers have
> > completed before the memory is freed.
> > 
> > Fixes: af3cf757d5c9 ("net: airoha: Move DSA tag in DMA descriptor")
> > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
> > ---
> >  drivers/net/ethernet/airoha/airoha_eth.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/net/ethernet/airoha/airoha_eth.c b/drivers/net/ethernet/airoha/airoha_eth.c
> > index cecd66251dba..eab6a98d62b9 100644
> > --- a/drivers/net/ethernet/airoha/airoha_eth.c
> > +++ b/drivers/net/ethernet/airoha/airoha_eth.c
> > @@ -2936,7 +2936,7 @@ static void airoha_metadata_dst_free(struct airoha_gdm_port *port)
> >  		if (!port->dsa_meta[i])
> >  			continue;
> >  
> > -		metadata_dst_free(port->dsa_meta[i]);
> > +		dst_release(&port->dsa_meta[i]->dst);
> >  	}
> >  }
> >  
> > 
> 
> the port->dsa_meta is allocated using metadata_dst_alloc().. how is it
> safe to use dst_release here? Seems like we should be calling dst_alloc
> instead of metadata_dst_alloc in order to use dst_release??

We need to allocate the metadata_dst using metadata_dst_alloc() since
md_dst->u.port_info.port_id is consumed in dsa_switch_rcv() to get the
switch conduit port.
I guess it is fine to free metadata_dst running dst_release() since dst_init()
sets DST_METADATA flag and so dst_destroy() runs metadata_dst_free() after the
RCU grace period.

> 
> metadata_dst_alloc does call __metadata_dst_init which calls dst_init..
> 
> I guess the start of the metadata_dst structure is also the same address
> as the internal dst_entry struct...
> 
> But dst_destroy does a whole lot more than metadata_dst_release so I
> don't feel confident in this actually being a drop-in replacement... It
> calls netdev_put, it calls the dst->ops->destroy, it releases child
> refs.. Or for metadata dst entries is that all basically a no-op??

__metadata_dst_init() calls dst_init() with dev = NULL so netdev_put() is a
no-op. Same for dst->ops is dst_blackhole_ops and and dst_blackhole_ops has no
destroy callback.

> 
> I feel like I'm missing something here.. The driver also calls
> metadata_dst_free in the remove path and that wasn't changed by this
> patch either.

can you please explain what you mean here? we do not run metadata_dst_free()
anymore.

Regards,
Lorenzo

> 
> Generally it seems like we should be using the same API to allocate as
> to release the object... This is confusing. What am I missing?