On Sat, 31 Jan 2026 06:12:12 +0000, Kohei Enju wrote:
> On Sat, 31 Jan 2026 13:52:17 +0800, Haoxiang Li wrote:
>
> > In i40e_xsk_pool_enable(), add an error handling path to
> > prevent potential memory leaks.
> >
> > Fixes: 1742b3d52869 ("xsk: i40e: ice: ixgbe: mlx5: Pass buffer pool to
> > driver instead of umem")
> > Cc: [email protected]
> > Signed-off-by: Haoxiang Li <[email protected]>
> > ---
> > Changes in v2:
> > - Add a Fixes tag. Thanks, Paul!
> > - Replace unmap with i40e_xsk_pool_disable() to prevent
> > a limbo state of queues. Thanks, Maciej!
> > ---
> > drivers/net/ethernet/intel/i40e/i40e_xsk.c | 11 +++++++----
> > 1 file changed, 7 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
> > b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
> > index 9f47388eaba5..a72a309540c3 100644
> > --- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
> > +++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
> > @@ -108,23 +108,26 @@ static int i40e_xsk_pool_enable(struct i40e_vsi *vsi,
> > if (if_running) {
> > err = i40e_queue_pair_disable(vsi, qid);
> > if (err)
> > - return err;
> > + goto err_out;
> >
> > err = i40e_realloc_rx_xdp_bi(vsi->rx_rings[qid], true);
> > if (err)
> > - return err;
> > + goto err_out;
> >
> > err = i40e_queue_pair_enable(vsi, qid);
> > if (err)
> > - return err;
> > + goto err_out;
> >
> > /* Kick start the NAPI context so that receiving will start */
> > err = i40e_xsk_wakeup(vsi->netdev, qid, XDP_WAKEUP_RX);
> > if (err)
> > - return err;
> > + goto err_out;
> > }
> >
> > return 0;
> > +
> > +err_out:
> > + i40e_xsk_pool_disable(vsi, qid);
>
> I think return err; is missing...
>
> Also, since i40e_xsk_pool_disable is not declared before this line,
> compilation fails due to a 'Call to undeclared function
> i40e_xsk_pool_disable' error. Adding declaration or moving
> i40e_xsk_pool_enable() after i40e_xsk_pool_disable() is needed.
After skimming the code, I think i40e_xsk_pool_disable() can fail in
this error path, right?
Let's assume a scenario that i40e_queue_pair_disable() succeeds but
i40e_realloc_rx_xdp_bi() fails in i40e_xsk_pool_enable().
At this point, i40e_enter_busy_conf() called by
i40e_queue_pair_disable() has set __I40E_CONFIG_BUSY state.
Then the err_out: path is executed and i40e_xsk_pool_disable() tries to
call i40e_queue_pair_disable() when if_running.
In this case, i40e_enter_busy_conf() fails because __I40E_CONFIG_BUSY is
set, and finally xsk_pool_dma_unmap() isn't called.
Do you have any thoughts on how to handle this case?
Thanks.
>
> > }
> >
> > /**
> > --
> > 2.25.1
