RE: [PATCH v2 net-next 11/13] net: enetc: optimize the allocation of tx_bdr

[Wei Fang]

> -----Original Message-----
> From: Frank Li <frank.li@xxxxxxx>
> Sent: 2024年10月16日 0:58
> To: Wei Fang <wei.fang@xxxxxxx>
> Cc: davem@xxxxxxxxxxxxx; edumazet@xxxxxxxxxx; kuba@xxxxxxxxxx;
> pabeni@xxxxxxxxxx; robh@xxxxxxxxxx; krzk+dt@xxxxxxxxxx;
> conor+dt@xxxxxxxxxx; Vladimir Oltean <vladimir.oltean@xxxxxxx>; Claudiu
> Manoil <claudiu.manoil@xxxxxxx>; Clark Wang <xiaoning.wang@xxxxxxx>;
> christophe.leroy@xxxxxxxxxx; linux@xxxxxxxxxxxxxxx; bhelgaas@xxxxxxxxxx;
> horms@xxxxxxxxxx; imx@xxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx;
> devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> linux-pci@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH v2 net-next 11/13] net: enetc: optimize the allocation of
> tx_bdr
> 
> On Tue, Oct 15, 2024 at 08:58:39PM +0800, Wei Fang wrote:
> > From: Clark Wang <xiaoning.wang@xxxxxxx>
> >
> > There is a situation where num_tx_rings cannot be divided by bdr_int_num.
> > For example, num_tx_rings is 8 and bdr_int_num is 3. According to the
> > previous logic, this results in two tx_bdr corresponding memories not
> > being allocated, so when sending packets to tx ring 6 or 7, wild
> > pointers will be accessed. Of course, this issue doesn't exist on
> > LS1028A, because its num_tx_rings is 8, and bdr_int_num is either 1 or
> > 2. However, there is a risk for the upcoming i.MX95. Therefore, it is
> > necessary to ensure that each tx_bdr can be allocated to the corresponding
> memory.
> >
> > Signed-off-by: Clark Wang <xiaoning.wang@xxxxxxx>
> > Signed-off-by: Wei Fang <wei.fang@xxxxxxx>
> > ---
> > v2 changes:
> > This patch is separated from v1 patch 9 ("net: enetc: optimize the
> > allocation of tx_bdr"). Only the optimized part is kept.
> > ---
> >  drivers/net/ethernet/freescale/enetc/enetc.c | 10 +++++++---
> >  1 file changed, 7 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/freescale/enetc/enetc.c
> > b/drivers/net/ethernet/freescale/enetc/enetc.c
> > index d36af3f8ba31..72ddf8b16271 100644
> > --- a/drivers/net/ethernet/freescale/enetc/enetc.c
> > +++ b/drivers/net/ethernet/freescale/enetc/enetc.c
> > @@ -3049,10 +3049,10 @@ static void enetc_int_vector_destroy(struct
> > enetc_ndev_priv *priv, int i)  int enetc_alloc_msix(struct
> > enetc_ndev_priv *priv)  {
> >  	struct pci_dev *pdev = priv->si->pdev;
> > +	int v_tx_rings, v_remainder;
> >  	int num_stack_tx_queues;
> >  	int first_xdp_tx_ring;
> >  	int i, n, err, nvec;
> > -	int v_tx_rings;
> 
> Nit: Needn't move v_tx_rings.

Just to keep the reverse xmas tree style, of course I could add a new line to
define v_remainder, but these two variables are related, so I think it is more
appropriate to define them together.

> 
> Reviewed-by: Frank Li <Frank.Li@xxxxxxx>
> 
> >
> >  	nvec = ENETC_BDR_INT_BASE_IDX + priv->bdr_int_num;
> >  	/* allocate MSIX for both messaging and Rx/Tx interrupts */ @@
> > -3066,10 +3066,14 @@ int enetc_alloc_msix(struct enetc_ndev_priv
> > *priv)
> >
> >  	/* # of tx rings per int vector */
> >  	v_tx_rings = priv->num_tx_rings / priv->bdr_int_num;
> > +	v_remainder = priv->num_tx_rings % priv->bdr_int_num;
> >
> > -	for (i = 0; i < priv->bdr_int_num; i++)
> > -		if (enetc_int_vector_init(priv, i, v_tx_rings))
> > +	for (i = 0; i < priv->bdr_int_num; i++) {
> > +		int num_tx_rings = i < v_remainder ? v_tx_rings + 1 : v_tx_rings;
> > +
> > +		if (enetc_int_vector_init(priv, i, num_tx_rings))
> >  			goto fail;
> > +	}
> >
> >  	num_stack_tx_queues = enetc_num_stack_tx_queues(priv);
> >
> > --
> > 2.34.1
> >