net: mv643xx_eth: fix packet corruption with TSO and tiny unaligned packets.
authorNicolas Schichan <nschichan@freebox.fr>
Tue, 26 Jan 2016 15:12:35 +0000 (16:12 +0100)
committerDavid S. Miller <davem@davemloft.net>
Fri, 29 Jan 2016 00:08:01 +0000 (16:08 -0800)
The code in txq_put_data() would use txq->tx_curr_desc to index the
tso_hdrs/tso_hdrs_dma buffers, for less than 8 bytes unaligned
fragments, which is already moved to the next descriptor at the
beginning of the function.

If that fragment was the last of the the skb, the next skb would use
that same space to place the ip headers, overwritting that small
fragment data.

Fixes: 91986fd3d335 (net: mv643xx_eth: Ensure proper data alignment in TSO TX path)
Signed-off-by: Nicolas Schichan <nschichan@freebox.fr>
Reviewed-by: Philipp Kirchhofer <philipp@familie-kirchhofer.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/marvell/mv643xx_eth.c

index a0c03834a2f7b3d7494b864db6e96ee590b36c81..55831188bc32431e935550a7da5671fa1d24d89a 100644 (file)
@@ -762,10 +762,10 @@ txq_put_data_tso(struct net_device *dev, struct tx_queue *txq,
 
        if (length <= 8 && (uintptr_t)data & 0x7) {
                /* Copy unaligned small data fragment to TSO header data area */
-               memcpy(txq->tso_hdrs + txq->tx_curr_desc * TSO_HEADER_SIZE,
+               memcpy(txq->tso_hdrs + tx_index * TSO_HEADER_SIZE,
                       data, length);
                desc->buf_ptr = txq->tso_hdrs_dma
-                       + txq->tx_curr_desc * TSO_HEADER_SIZE;
+                       + tx_index * TSO_HEADER_SIZE;
        } else {
                /* Alignment is okay, map buffer and hand off to hardware */
                txq->tx_desc_mapping[tx_index] = DESC_DMA_MAP_SINGLE;