Commit 29578624 authored by Olaf Kirch's avatar Olaf Kirch Committed by David S. Miller
Browse files

[NET]: Fix races in net_rx_action vs netpoll.

Keep netpoll/poll_napi from messing with the poll_list.
Only net_rx_action is allowed to manipulate the list.
Signed-off-by: default avatarOlaf Kirch <>
Signed-off-by: default avatarDavid S. Miller <>
parent 1fd05ba5
......@@ -261,6 +261,8 @@ enum netdev_state_t
/* Set by the netpoll NAPI code */
......@@ -1014,6 +1016,14 @@ static inline void netif_rx_complete(struct net_device *dev)
unsigned long flags;
/* Prevent race with netpoll - yes, this is a kludge.
* But at least it doesn't penalize the non-netpoll
* code path. */
if (test_bit(__LINK_STATE_POLL_LIST_FROZEN, &dev->state))
......@@ -124,6 +124,13 @@ static void poll_napi(struct netpoll *np)
if (test_bit(__LINK_STATE_RX_SCHED, &np->dev->state) &&
npinfo->poll_owner != smp_processor_id() &&
spin_trylock(&npinfo->poll_lock)) {
/* When calling dev->poll from poll_napi, we may end up in
* netif_rx_complete. However, only the CPU to which the
* device was queued is allowed to remove it from poll_list.
* Setting POLL_LIST_FROZEN tells netif_rx_complete
* to leave the NAPI state alone.
set_bit(__LINK_STATE_POLL_LIST_FROZEN, &np->dev->state);
npinfo->rx_flags |= NETPOLL_RX_DROP;
......@@ -131,6 +138,7 @@ static void poll_napi(struct netpoll *np)
npinfo->rx_flags &= ~NETPOLL_RX_DROP;
clear_bit(__LINK_STATE_POLL_LIST_FROZEN, &np->dev->state);
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment