Re: [PATCH net-next 1/4] net: mana: Fix potential deadlocks in mana napi ops

[Saurabh Singh Sengar]

On Wed, Jun 11, 2025 at 04:03:52AM -0700, Saurabh Singh Sengar wrote:
> On Wed, Jun 11, 2025 at 01:46:13AM -0700, Erni Sri Satya Vennela wrote:
> > When net_shaper_ops are enabled for MANA, netdev_ops_lock
> > becomes active.
> > 
> > The netvsc sets up MANA VF via following call chain:
> > 
> > netvsc_vf_setup()
> >         dev_change_flags()
> > 		...
> >          __dev_open() OR __dev_close()
> > 
> > dev_change_flags() holds the netdev mutex via netdev_lock_ops.
> > 
> > During this process, mana_create_txq() and mana_create_rxq()
> > invoke netif_napi_add_tx(), netif_napi_add_weight(), and napi_enable(),
> > all of which attempt to acquire the same lock,
> > leading to a potential deadlock.
> 
> commit message could be better oriented.
> 
> > 
> > Similarly, mana_destroy_txq() and mana_destroy_rxq() call
> > netif_napi_disable() and netif_napi_del(), which also contend
> > for the same lock.
> > 
> > Switch to the _locked variants of these APIs to avoid deadlocks
> > when the netdev_ops_lock is held.
> > 
> > Fixes: d4c22ec680c8 ("net: hold netdev instance lock during ndo_open/ndo_stop")
> > Signed-off-by: Erni Sri Satya Vennela <ernis@xxxxxxxxxxxxxxxxxxx>
> > Reviewed-by: Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>
> > Reviewed-by: Shradha Gupta <shradhagupta@xxxxxxxxxxxxxxxxxxx>
> > ---
> >  drivers/net/ethernet/microsoft/mana/mana_en.c | 39 ++++++++++++++-----
> >  1 file changed, 30 insertions(+), 9 deletions(-)
> > 
> > diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c
> > index ccd2885c939e..3c879d8a39e3 100644
> > --- a/drivers/net/ethernet/microsoft/mana/mana_en.c
> > +++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
> > @@ -1911,8 +1911,13 @@ static void mana_destroy_txq(struct mana_port_context *apc)
> >  		napi = &apc->tx_qp[i].tx_cq.napi;
> >  		if (apc->tx_qp[i].txq.napi_initialized) {
> >  			napi_synchronize(napi);
> > -			napi_disable(napi);
> > -			netif_napi_del(napi);
> > +			if (netdev_need_ops_lock(napi->dev)) {
> > +				napi_disable_locked(napi);
> > +				netif_napi_del_locked(napi);
> > +			} else {
> > +				napi_disable(napi);
> > +				netif_napi_del(napi);
> > +			}
> 
> Instead of using if-else, we can used netdev_lock_ops(), followed by *_locked api-s.
> Same for rest of the patch.
> 

I later realized that what we actually need is:

  if (!netdev_need_ops_lock(napi->dev))
	netdev_lock(dev);

not

  if (netdev_need_ops_lock(napi->dev))
        netdev_lock(dev);

Hence, netdev_lock_ops() is not appropriate. Instead, netdev_lock_ops_to_full()
seems to be a better choice.

> Reviewed-by: Saurabh Sengar <ssengar@xxxxxxxxxxxxxxxxxxx>
> 
> - Saurabh
>