linux/include/net/gro_cells.h
Eric Dumazet f8e8f97c11 net: fix a race in gro_cell_poll()
Dmitry Kravkov reported packet drops for GRE packets since GRO support
was added.

There is a race in gro_cell_poll() because we call napi_complete()
without any synchronization with a concurrent gro_cells_receive()

Once bug was triggered, we queued packets but did not schedule NAPI
poll.

We can fix this issue using the spinlock protected the napi_skbs queue,
as we have to hold it to perform skb dequeue anyway.

As we open-code skb_dequeue(), we no longer need to mask IRQS, as both
producer and consumer run under BH context.

Bug added in commit c9e6bc644e (net: add gro_cells infrastructure)

Reported-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-12-11 12:49:53 -05:00

108 lines
2.4 KiB
C

#ifndef _NET_GRO_CELLS_H
#define _NET_GRO_CELLS_H
#include <linux/skbuff.h>
#include <linux/slab.h>
#include <linux/netdevice.h>
struct gro_cell {
struct sk_buff_head napi_skbs;
struct napi_struct napi;
} ____cacheline_aligned_in_smp;
struct gro_cells {
unsigned int gro_cells_mask;
struct gro_cell *cells;
};
static inline void gro_cells_receive(struct gro_cells *gcells, struct sk_buff *skb)
{
struct gro_cell *cell = gcells->cells;
struct net_device *dev = skb->dev;
if (!cell || skb_cloned(skb) || !(dev->features & NETIF_F_GRO)) {
netif_rx(skb);
return;
}
if (skb_rx_queue_recorded(skb))
cell += skb_get_rx_queue(skb) & gcells->gro_cells_mask;
if (skb_queue_len(&cell->napi_skbs) > netdev_max_backlog) {
atomic_long_inc(&dev->rx_dropped);
kfree_skb(skb);
return;
}
/* We run in BH context */
spin_lock(&cell->napi_skbs.lock);
__skb_queue_tail(&cell->napi_skbs, skb);
if (skb_queue_len(&cell->napi_skbs) == 1)
napi_schedule(&cell->napi);
spin_unlock(&cell->napi_skbs.lock);
}
/* called unser BH context */
static inline int gro_cell_poll(struct napi_struct *napi, int budget)
{
struct gro_cell *cell = container_of(napi, struct gro_cell, napi);
struct sk_buff *skb;
int work_done = 0;
spin_lock(&cell->napi_skbs.lock);
while (work_done < budget) {
skb = __skb_dequeue(&cell->napi_skbs);
if (!skb)
break;
spin_unlock(&cell->napi_skbs.lock);
napi_gro_receive(napi, skb);
work_done++;
spin_lock(&cell->napi_skbs.lock);
}
if (work_done < budget)
napi_complete(napi);
spin_unlock(&cell->napi_skbs.lock);
return work_done;
}
static inline int gro_cells_init(struct gro_cells *gcells, struct net_device *dev)
{
int i;
gcells->gro_cells_mask = roundup_pow_of_two(netif_get_num_default_rss_queues()) - 1;
gcells->cells = kcalloc(sizeof(struct gro_cell),
gcells->gro_cells_mask + 1,
GFP_KERNEL);
if (!gcells->cells)
return -ENOMEM;
for (i = 0; i <= gcells->gro_cells_mask; i++) {
struct gro_cell *cell = gcells->cells + i;
skb_queue_head_init(&cell->napi_skbs);
netif_napi_add(dev, &cell->napi, gro_cell_poll, 64);
napi_enable(&cell->napi);
}
return 0;
}
static inline void gro_cells_destroy(struct gro_cells *gcells)
{
struct gro_cell *cell = gcells->cells;
int i;
if (!cell)
return;
for (i = 0; i <= gcells->gro_cells_mask; i++,cell++) {
netif_napi_del(&cell->napi);
skb_queue_purge(&cell->napi_skbs);
}
kfree(gcells->cells);
gcells->cells = NULL;
}
#endif