-stable review patch. If anyone has any objections, please let us know.
---------------------
From: Ben Hutchings <[email protected]>
upstream commit: 718cff1eec595ce6ab0635b8160a51ee37d9268d
sfc could call netif_napi_add() multiple times for the same
napi_struct, corrupting the list of napi_structs for the associated
device and leading to a busy-loop on device removal. Move the call to
netif_napi_add() and add a call to netif_napi_del() in the obvious
places.
[bhutchings: backport to 2.6.29]
Signed-off-by: Ben Hutchings <[email protected]>
Signed-off-by: Chris Wright <[email protected]>
---
We didn't spot this earlier because only netpoll would look at the list,
but this changed with the addition of GRO in 2.6.29. I have no excuse
for not catching it during the .29 release cycle though.
Ben.
drivers/net/sfc/efx.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
--- a/drivers/net/sfc/efx.c
+++ b/drivers/net/sfc/efx.c
@@ -424,10 +424,6 @@ static void efx_start_channel(struct efx
EFX_LOG(channel->efx, "starting chan %d\n", channel->channel);
- if (!(channel->efx->net_dev->flags & IFF_UP))
- netif_napi_add(channel->napi_dev, &channel->napi_str,
- efx_poll, napi_weight);
-
/* The interrupt handler for this channel may set work_pending
* as soon as we enable it. Make sure it's cleared before
* then. Similarly, make sure it sees the enabled flag set. */
@@ -1273,6 +1269,8 @@ static int efx_init_napi(struct efx_nic
efx_for_each_channel(channel, efx) {
channel->napi_dev = efx->net_dev;
+ netif_napi_add(channel->napi_dev, &channel->napi_str,
+ efx_poll, napi_weight);
rc = efx_lro_init(&channel->lro_mgr, efx);
if (rc)
goto err;
@@ -1289,6 +1287,8 @@ static void efx_fini_napi(struct efx_nic
efx_for_each_channel(channel, efx) {
efx_lro_fini(&channel->lro_mgr);
+ if (channel->napi_dev)
+ netif_napi_del(&channel->napi_str);
channel->napi_dev = NULL;
}
}