RE: [EXT] Re: [PATCH net-next v1] octeon_ep: get max rx packet length from firmware

[Shinas Rasheed]

> -----Original Message-----
> From: Jesse Brandeburg <jesse.brandeburg@xxxxxxxxx>
> Sent: Wednesday, November 22, 2023 2:34 AM
> To: Shinas Rasheed <srasheed@xxxxxxxxxxx>; netdev@xxxxxxxxxxxxxxx;
> linux-kernel@xxxxxxxxxxxxxxx
> Cc: Haseeb Gani <hgani@xxxxxxxxxxx>; Vimlesh Kumar
> <vimleshk@xxxxxxxxxxx>; egallen@xxxxxxxxxx; mschmidt@xxxxxxxxxx;
> pabeni@xxxxxxxxxx; horms@xxxxxxxxxx; kuba@xxxxxxxxxx;
> davem@xxxxxxxxxxxxx; wizhao@xxxxxxxxxx; konguyen@xxxxxxxxxx;
> Veerasenareddy Burru <vburru@xxxxxxxxxxx>; Sathesh B Edara
> <sedara@xxxxxxxxxxx>; Eric Dumazet <edumazet@xxxxxxxxxx>
> Subject: [EXT] Re: [PATCH net-next v1] octeon_ep: get max rx packet length
> from firmware
> 
> External Email
> 
> ----------------------------------------------------------------------
> On 11/21/2023 11:12 AM, Shinas Rasheed wrote:
> > Fill max rx packet length value from firmware.
> 
> Hi Shinas, thanks for the patch.
> 
> Please provide why, and make sure you're talking to the linux kernel
> developer audience who don't know anything about your hardware. We're
> interested to know why this patch is useful to the kernel and why it
> might need to be applied.

Sure will update that in the changelog in the next version.
 
> > +/** Get max MTU from firmware.
> > + *
> > + * @param oct: non-null pointer to struct octep_device.
> > + * @param vfid: Index of virtual function.
> > + *
> > + * return value: mtu on success, -errno on failure.
> > + */
> 
> The above block is definitely not correctly formatted kdoc (if that's
> what you wanted), and you can probably get feedback about it from
> scripts/kernel-doc -v
> drivers/net/ethernet/marvell/octeon_ep/octep_ctrl_net.h
> 
> I see you have some correctly formatted doc in octep_tx.c

I'll see to it to properly format it.

> > diff --git a/drivers/net/ethernet/marvell/octeon_ep/octep_main.c
> b/drivers/net/ethernet/marvell/octeon_ep/octep_main.c
> > index 3cee69b3ac38..f9c539178114 100644
> > --- a/drivers/net/ethernet/marvell/octeon_ep/octep_main.c
> > +++ b/drivers/net/ethernet/marvell/octeon_ep/octep_main.c
> > @@ -1276,6 +1276,7 @@ static int octep_probe(struct pci_dev *pdev,
> const struct pci_device_id *ent)
> >  {
> >  	struct octep_device *octep_dev = NULL;
> >  	struct net_device *netdev;
> > +	int max_rx_pktlen;
> >  	int err;
> >
> >  	err = pci_enable_device(pdev);
> > @@ -1346,8 +1347,15 @@ static int octep_probe(struct pci_dev *pdev,
> const struct pci_device_id *ent)
> >
> >  	netdev->hw_features = NETIF_F_SG;
> >  	netdev->features |= netdev->hw_features;
> > +
> > +	max_rx_pktlen = octep_ctrl_net_get_mtu(octep_dev,
> OCTEP_CTRL_NET_INVALID_VFID);
> > +	if (max_rx_pktlen < 0) {
> > +		dev_err(&octep_dev->pdev->dev,
> > +			"Failed to get max receive packet size; err = %d\n",
> max_rx_pktlen);
> > +		goto register_dev_err;
> > +	}
> >  	netdev->min_mtu = OCTEP_MIN_MTU;
> > -	netdev->max_mtu = OCTEP_MAX_MTU;
> > +	netdev->max_mtu = max_rx_pktlen - (ETH_HLEN + ETH_FCS_LEN);
> >  	netdev->mtu = OCTEP_DEFAULT_MTU;
> 
> Not part of this patch, but was there a point to setting the mtu here
> without telling the netdev? most of the time it seems sufficient to just
> set max and min since the kernel default is already 1500 (which your
> internal define also duplicates)

I suppose that piece of code is redundant, but serves to atleast say that the default expected is 1500 for the device.