Re: [PATCH 1/2] net: macb: remove gem_recv and reorder probe

[Steffen Trumtrar <s.trumtrar@pengutronix.de>]

On Wed, Mar 13, 2013 at 01:17:03PM +0100, Jean-Christophe PLAGNIOL-VILLARD wrote:
> On 11:15 Wed 13 Mar     , Sascha Hauer wrote:
> > On Wed, Mar 13, 2013 at 10:03:36AM +0100, Jean-Christophe PLAGNIOL-VILLARD wrote:
> > > On 09:48 Wed 13 Mar     , Steffen Trumtrar wrote:
> > > > The function gem_recv implements a buffer "ring" that stops at filling level 10.
> > > > That means, when the driver is used as gem, it stops receiving after exactly
> > > > 10 packets. Instead of implementing macb_recv twice, use it also for the gem
> > > > part. If there needs to be an extra recv function for the gigabit case, it can
> > > > be added than.
> > > > Also, first set the type of device (macb or gem) and then use functions that
> > > > use this info.
> > > > 
> > > > Tested on a Zynq7000.
> > > NACK
> > > 
> > > on gem we can receive the packet in one buffer the gem_recv implement this
> > > the macb can only receive splited buffer and then you have to reconstruct the
> > > packet
> > 
> > The gem received function was never used...
> > 
> > > >  static int macb_recv(struct eth_device *edev)
> > > >  {
> > > >  	struct macb_device *macb = edev->priv;
> > > > @@ -619,11 +588,6 @@ static int macb_probe(struct device_d *dev)
> > > >  
> > > >  	macb->phy_flags = pdata->phy_flags;
> > > >  
> > > > -	macb_init_rx_buffer_size(macb, PKTSIZE);
> > > > -	macb->rx_buffer = dma_alloc_coherent(macb->rx_buffer_size * macb->rx_ring_size);
> > > > -	macb->rx_ring = dma_alloc_coherent(RX_RING_BYTES(macb));
> > > > -	macb->tx_ring = dma_alloc_coherent(TX_RING_BYTES);
> > > > -
> > > >  	macb->regs = dev_request_mem_region(dev, 0);
> > > >  
> > > >  	/*
> > > > @@ -638,12 +602,15 @@ static int macb_probe(struct device_d *dev)
> > > >  
> > > >  	clk_enable(macb->pclk);
> > > >  
> > > > -	if (macb_is_gem(macb))
> > 
> > ... because macb_is_gem() is used here ...
> > 
> > > > -		edev->recv = gem_recv;
> > > > -	else
> > > > -		edev->recv = macb_recv;
> > > >  	macb->is_gem = read_is_gem(macb);
> > 
> > ... but the variable is initialized here. Up to this point macb_is_gem()
> > will return 0.
> so this is a bug when the driver was merge to the mailine but here I does use
> the gem_recv.
> 

What about the gem_recv function itself? It only increases rx_tail, but never
resets it. Therefore the buffer is full after 10 packets. Do you already have a
patch for that?

Regards,
Steffen

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox