On Wed, Feb 1, 2017 at 3:09 AM, David Laight <david.lai...@aculab.com> wrote: > From Florian Fainelli >> Sent: 31 January 2017 20:33 >> On 01/31/2017 11:03 AM, Iyappan Subramanian wrote: >> > This patch adds, >> > - Transmit >> > - Transmit completion poll >> > - Receive poll >> > - NAPI handler >> > >> > and enables the driver. >> > >> > Signed-off-by: Iyappan Subramanian <isubraman...@apm.com> >> > Signed-off-by: Keyur Chudgar <kchud...@apm.com> >> > --- >> >> > + >> > + tx_ring = pdata->tx_ring; >> > + tail = tx_ring->tail; >> > + len = skb_headlen(skb); >> > + raw_desc = &tx_ring->raw_desc[tail]; >> > + >> > + /* Tx descriptor not available */ >> > + if (!GET_BITS(E, le64_to_cpu(raw_desc->m0)) || >> > + GET_BITS(PKT_SIZE, le64_to_cpu(raw_desc->m0))) >> > + return NETDEV_TX_BUSY; > > Aren't you supposed to detect 'ring full' and stop the code > giving you packets to transmit.
I'll add stop queue and wake queue. > >> > + >> > + /* Packet buffers should be 64B aligned */ > > Is that really a requirement of the hardware? > Almost all ethernet frames are 4n+2 aligned. Yes, it's a hardware requirement. > >> > + pkt_buf = dma_alloc_coherent(dev, XGENE_ENET_STD_MTU, &dma_addr, >> > + GFP_ATOMIC); >> > + if (unlikely(!pkt_buf)) >> > + goto out; >> >> Can't you obtain a DMA-API mapping for skb->data and pass it down to the >> hardware? This copy here is inefficient. >> >> > + >> > + memcpy(pkt_buf, skb->data, len); > > You really need to verify that the len <= XGENE_ENET_STD_MTU. This version of the driver, doesn't support jumbo frame. So, the check is not required. > > Isn't this code only transmitting the 'head' of the packet? > What about the fragments?? This driver doesn't enable SG yet. > ... > David >
