The state of m_td0TD0.f1Owner should change after the buff_addr
has been filled otherwise the device grabs the buffer too early.

m_td0TD0.f1Owner is protected by memory barriers on both sides
of change.

iTDUsed is best incremented after MACvTransmit.

It appears that f1Owner actually polls to do the memory transfer.

A back port patch will be needed for v3.19

Signed-off-by: Malcolm Priestley <tvbox...@gmail.com>
Cc: <stable@vger.kernel.org> # v4.0+
---
 drivers/staging/vt6655/device_main.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/staging/vt6655/device_main.c 
b/drivers/staging/vt6655/device_main.c
index c5f8233..ed90b6f 100644
--- a/drivers/staging/vt6655/device_main.c
+++ b/drivers/staging/vt6655/device_main.c
@@ -1201,14 +1201,6 @@ static int vnt_tx_packet(struct vnt_private *priv, 
struct sk_buff *skb)
        if (dma_idx == TYPE_AC0DMA)
                head_td->pTDInfo->byFlags = TD_FLAGS_NETIF_SKB;
 
-       priv->iTDUsed[dma_idx]++;
-
-       /* Take ownership */
-       wmb();
-       head_td->m_td0TD0.f1Owner = OWNED_BY_NIC;
-
-       /* get Next */
-       wmb();
        priv->apCurrTD[dma_idx] = head_td->next;
 
        spin_unlock_irqrestore(&priv->lock, flags);
@@ -1229,11 +1221,18 @@ static int vnt_tx_packet(struct vnt_private *priv, 
struct sk_buff *skb)
 
        head_td->buff_addr = cpu_to_le32(head_td->pTDInfo->skb_dma);
 
+       /* Poll Transmit the adapter */
+       wmb();
+       head_td->m_td0TD0.f1Owner = OWNED_BY_NIC;
+       wmb(); /* second memory barrier */
+
        if (head_td->pTDInfo->byFlags & TD_FLAGS_NETIF_SKB)
                MACvTransmitAC0(priv->PortOffset);
        else
                MACvTransmit0(priv->PortOffset);
 
+       priv->iTDUsed[dma_idx]++;
+
        spin_unlock_irqrestore(&priv->lock, flags);
 
        return 0;
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to