Re: [PATCH net-next V2 2/3] xen-netback: avoid allocating variable size array on stack
From: Ian Campbell <hidden>
Date: 2013-05-02 15:55:48
On Thu, 2013-05-02 at 13:04 +0100, Jan Beulich wrote:
quoted
quoted
quoted
On 02.05.13 at 12:43, Wei Liu [off-list ref] wrote:@@ -934,11 +934,14 @@ static int netbk_count_requests(struct xenvif *vif, RING_IDX cons = vif->tx.req_cons; int slots = 0; int drop_err = 0; + int more_data; if (!(first->flags & XEN_NETTXF_more_data)) return 0; do { + struct xen_netif_tx_request dropped_tx = { 0 }; +No need for an initializer here.quoted
if (slots >= work_to_do) { netdev_err(vif->dev, "Asked for %d slots but exceeds this limit\n",@@ -972,6 +975,9 @@ static int netbk_count_requests(struct xenvif *vif, drop_err = -E2BIG; } + if (drop_err) + txp = &dropped_tx; + memcpy(txp, RING_GET_REQUEST(&vif->tx, cons + slots), sizeof(*txp));@@ -1001,7 +1007,13 @@ static int netbk_count_requests(struct xenvif *vif, netbk_fatal_tx_err(vif); return -EINVAL; } - } while ((txp++)->flags & XEN_NETTXF_more_data); + + more_data = txp->flags & XEN_NETTXF_more_data; + + if (!drop_err) + txp++;And no need for the conditional here afaict.
I think it is needed, in the case where you've assigned txp = &dropped_tx you don't want to increment txp. Or perhaps it just gets reassigned back to &dropped_tx at the top of the next loop, so the increment doesn't matter. Subtle! I'm happy with whichever way Wei prefers, but it is probably worthy of a comment
Janquoted
+ + } while (more_data); if (drop_err) { netbk_tx_err(vif, first, cons + slots);