On 13/03/14 11:02, Ian Campbell wrote:

On Thu, 2014-03-13 at 10:56 +0000, David Vrabel wrote:

quoted

On 13/03/14 10:33, Ian Campbell wrote:

quoted

On Thu, 2014-03-06 at 21:48 +0000, Zoltan Kiss wrote:

quoted

@@ -135,13 +146,31 @@ struct xenvif {
 	pending_ring_idx_t pending_cons;
 	u16 pending_ring[MAX_PENDING_REQS];
 	struct pending_tx_info pending_tx_info[MAX_PENDING_REQS];
+	grant_handle_t grant_tx_handle[MAX_PENDING_REQS];
 
 	/* Coalescing tx requests before copying makes number of grant
 	 * copy ops greater or equal to number of slots required. In
 	 * worst case a tx request consumes 2 gnttab_copy.
 	 */
 	struct gnttab_copy tx_copy_ops[2*MAX_PENDING_REQS];
-
+	struct gnttab_map_grant_ref tx_map_ops[MAX_PENDING_REQS];
+	struct gnttab_unmap_grant_ref tx_unmap_ops[MAX_PENDING_REQS];

I wonder if we should break some of these arrays into separate
allocations? Wasn't there a problem with sizeof(struct xenvif) at one
point?

alloc_netdev() falls back to vmalloc() if the kmalloc failed so there's
no need to split these structures.

Is vmalloc space in abundant supply? For some reason I thought it was
limited (maybe that's a 32-bit only limitation?)

It is limited in 32-bit, but 64-bit has stupid amounts.

/proc/meminfo:

VmallocTotal:   34359738367 kB

David