Re: [PATCH] net: fix memory leaks in flush_backlog() with RPS
> On May 1, 2020, at 11:32 PM, Eric Dumazet wrote: > > kfree_skb() is supposed to call skb_dst_drop() (look in > skb_release_head_state()) > > If you think about it, we would have hundreds of similar bugs if this was not > the case. Thanks for quick response. Funny thing is that once I applied this patch, the leaks went away. It could be the fuzzers do not always reproduce the leaks or it could be that call_rcu() in skb_dst_drop() takes a long time waiting for grace periods which may confuse kmemleak because skb has already gone.
Re: [PATCH] net: fix memory leaks in flush_backlog() with RPS
On 5/1/20 8:15 PM, Qian Cai wrote: > netif_receive_skb_list_internal() could call enqueue_to_backlog() to put > some skb to softnet_data.input_pkt_queue and then in > ip_route_input_slow(), it allocates a dst_entry to be used in > skb_dst_set(). Later, > > cleanup_net > default_device_exit_batch > unregister_netdevice_many > rollback_registered_many > flush_all_backlogs > > will call flush_backlog() for all CPUs which would call kfree_skb() for > each skb on the input_pkt_queue without calling skb_dst_drop() first. > > unreferenced object 0x97008e4c4040 (size 176): > comm "softirq", pid 0, jiffies 4295173845 (age 32012.550s) > hex dump (first 32 bytes): >00 d0 a5 74 04 97 ff ff 40 72 1a 96 ff ff ff ff ...t@r.. >c1 a3 c5 95 ff ff ff ff 00 00 00 00 00 00 00 00 > backtrace: >[<30483fae>] kmem_cache_alloc+0x184/0x430 >[<7ae17545>] dst_alloc+0x8e/0x128 >[<1efe9a1f>] rt_dst_alloc+0x6f/0x1e0 >rt_dst_alloc at net/ipv4/route.c:1628 >[] ip_route_input_rcu+0xdfe/0x1640 >ip_route_input_slow at net/ipv4/route.c:2218 >(inlined by) ip_route_input_rcu at net/ipv4/route.c:2348 >[<9f30cbc0>] ip_route_input_noref+0xab/0x1a0 >[<4f53bd04>] arp_process+0x83a/0xf50 >arp_process at net/ipv4/arp.c:813 (discriminator 1) >[<61fd547d>] arp_rcv+0x276/0x330 >[<07dbfa7a>] __netif_receive_skb_list_core+0x4d2/0x500 >[<62d5f6d2>] netif_receive_skb_list_internal+0x4cb/0x7d0 >[<2baa2b74>] gro_normal_list+0x55/0xc0 >[<93d04885>] napi_complete_done+0xea/0x350 >[<467dd088>] tg3_poll_msix+0x174/0x310 [tg3] >[<498af7d9>] net_rx_action+0x278/0x890 >[<1e81d7e6>] __do_softirq+0xd9/0x589 >[<087ee354>] irq_exit+0xa2/0xc0 >[<1c4db0cd>] do_IRQ+0x87/0x180 > > Signed-off-by: Qian Cai > --- > net/core/dev.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/net/core/dev.c b/net/core/dev.c > index 522288177bbd..b898cd3036da 100644 > --- a/net/core/dev.c > +++ b/net/core/dev.c > @@ -5496,6 +5496,7 @@ static void flush_backlog(struct work_struct *work) > skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) { > if (skb->dev->reg_state == NETREG_UNREGISTERING) { > __skb_unlink(skb, &sd->input_pkt_queue); > + skb_dst_drop(skb); > kfree_skb(skb); > input_queue_head_incr(sd); > } > kfree_skb() is supposed to call skb_dst_drop() (look in skb_release_head_state()) If you think about it, we would have hundreds of similar bugs if this was not the case.
[PATCH] net: fix memory leaks in flush_backlog() with RPS
netif_receive_skb_list_internal() could call enqueue_to_backlog() to put some skb to softnet_data.input_pkt_queue and then in ip_route_input_slow(), it allocates a dst_entry to be used in skb_dst_set(). Later, cleanup_net default_device_exit_batch unregister_netdevice_many rollback_registered_many flush_all_backlogs will call flush_backlog() for all CPUs which would call kfree_skb() for each skb on the input_pkt_queue without calling skb_dst_drop() first. unreferenced object 0x97008e4c4040 (size 176): comm "softirq", pid 0, jiffies 4295173845 (age 32012.550s) hex dump (first 32 bytes): 00 d0 a5 74 04 97 ff ff 40 72 1a 96 ff ff ff ff ...t@r.. c1 a3 c5 95 ff ff ff ff 00 00 00 00 00 00 00 00 backtrace: [<30483fae>] kmem_cache_alloc+0x184/0x430 [<7ae17545>] dst_alloc+0x8e/0x128 [<1efe9a1f>] rt_dst_alloc+0x6f/0x1e0 rt_dst_alloc at net/ipv4/route.c:1628 [] ip_route_input_rcu+0xdfe/0x1640 ip_route_input_slow at net/ipv4/route.c:2218 (inlined by) ip_route_input_rcu at net/ipv4/route.c:2348 [<9f30cbc0>] ip_route_input_noref+0xab/0x1a0 [<4f53bd04>] arp_process+0x83a/0xf50 arp_process at net/ipv4/arp.c:813 (discriminator 1) [<61fd547d>] arp_rcv+0x276/0x330 [<07dbfa7a>] __netif_receive_skb_list_core+0x4d2/0x500 [<62d5f6d2>] netif_receive_skb_list_internal+0x4cb/0x7d0 [<2baa2b74>] gro_normal_list+0x55/0xc0 [<93d04885>] napi_complete_done+0xea/0x350 [<467dd088>] tg3_poll_msix+0x174/0x310 [tg3] [<498af7d9>] net_rx_action+0x278/0x890 [<1e81d7e6>] __do_softirq+0xd9/0x589 [<087ee354>] irq_exit+0xa2/0xc0 [<1c4db0cd>] do_IRQ+0x87/0x180 Signed-off-by: Qian Cai --- net/core/dev.c | 1 + 1 file changed, 1 insertion(+) diff --git a/net/core/dev.c b/net/core/dev.c index 522288177bbd..b898cd3036da 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -5496,6 +5496,7 @@ static void flush_backlog(struct work_struct *work) skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) { if (skb->dev->reg_state == NETREG_UNREGISTERING) { __skb_unlink(skb, &sd->input_pkt_queue); + skb_dst_drop(skb); kfree_skb(skb); input_queue_head_incr(sd); } -- 2.21.0 (Apple Git-122.2)