Hello Eli,
If i undersand your code correctly, you post_receive 16 skb's at
once. when ib_post_receive returns error, all these 16 skb's would fail,
not just this skb failure, am I right? If all failure, then we should free
these 16 skb's not just one. Also the same issue in ipoib-cm code.
Thanks
Shirley
Eli Cohen
<[EMAIL PROTECTED]
.co.il> To
Shirley Ma/Beaverton/[EMAIL PROTECTED]
02/05/08 12:38 AM cc
openfabrics
<[EMAIL PROTECTED]
Please respond to rg>, Roland Dreier
[EMAIL PROTECTED] <[EMAIL PROTECTED]>,
co.il [EMAIL PROTECTED]
Subject
Re: [ofa-general] [PATCH 0/5]:
Improve small UDP messages
On Tue, 2008-02-05 at 00:29 -0800, Shirley Ma wrote:
> Hello Eli,
>
> Did you see my another email regarding the memory leak issue in
> post_receive()?
No, I saw you suggested using dynamic allocations for the list of
receive WQEs but nothing re memory leak. Can you be more specific?
>
> +static int ipoib_ib_post_receive(struct net_device *dev, int id)
> +{
> + struct ipoib_dev_priv *priv = netdev_priv(dev);
> + struct ib_recv_wr *bad_wr;
> + int ret = 0;
> + int i = priv->rx_outst;
> +
> + priv->sglist_draft[i].addr = priv->rx_ring[id].mapping;
> + priv->rx_wr_draft[i].wr_id = id | IPOIB_OP_RECV;
> + if (++priv->rx_outst == UD_POST_RCV_COUNT) {
> + ret = ib_post_recv(priv->qp, priv->rx_wr_draft,
> &bad_wr);
> +
> + if (unlikely(ret)) {
> + ipoib_warn(priv, "receive failed for buf %d (%
> d)\n", id, ret);
> + while (bad_wr) {
> + id = bad_wr->wr_id & ~IPOIB_OP_RECV;
> + ib_dma_unmap_single(priv->ca,
> priv->rx_ring[id].mapping,
> + IPOIB_BUF_SIZE,
> DMA_FROM_DEVICE);
> +
> dev_kfree_skb_any(priv->rx_ring[id].skb);
> + priv->rx_ring[id].skb = NULL;
> + }
>
> 16 skbs need to be freed here when errors, not just this one.
>
> + }
> + priv->rx_outst = 0;
> + }
>
> return ret;
> }
>
> Thanks
> Shirley
>
>
<<inline: graycol.gif>>
<<inline: pic01959.gif>>
<<inline: ecblank.gif>>
_______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
