Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Pull networking fixes from David Miller:
1) The per-network-namespace loopback device, and thus its namespace,
can have its teardown deferred for a long time if a kernel created
TCP socket closes and the namespace is exiting meanwhile. The kernel
keeps trying to finish the close sequence until it times out (which
takes quite some time).
Fix this by forcing the socket closed in this situation, from Dan
Streetman.
2) Fix regression where we're trying to invoke the update_pmtu method
on route types (in this case metadata tunnel routes) that don't
implement the dst_ops method. Fix from Nicolas Dichtel.
3) Fix long standing memory corruption issues in r8169 driver by
performing the chip statistics DMA programming more correctly. From
Francois Romieu.
4) Handle local broadcast sends over VRF routes properly, from David
Ahern.
5) Don't refire the DCCP CCID2 timer endlessly, otherwise the socket
can never be released. From Alexey Kodanev.
6) Set poll flags properly in VSOCK protocol layer, from Stefan
Hajnoczi.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
VSOCK: set POLLOUT | POLLWRNORM for TCP_CLOSING
dccp: don't restart ccid2_hc_tx_rto_expire() if sk in closed state
net: vrf: Add support for sends to local broadcast address
r8169: fix memory corruption on retrieval of hardware statistics.
net: don't call update_pmtu unconditionally
net: tcp: close sock if net namespace is exiting
+57
-28
+1
-2
drivers/infiniband/ulp/ipoib/ipoib_cm.c
+1
-2
drivers/infiniband/ulp/ipoib/ipoib_cm.c
···
1456
1456
struct ipoib_dev_priv *priv = ipoib_priv(dev);
1457
1457
int e = skb_queue_empty(&priv->cm.skb_queue);
1458
1458
1459
-
if (skb_dst(skb))
1460
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1459
+
skb_dst_update_pmtu(skb, mtu);
1461
1460
1462
1461
skb_queue_tail(&priv->cm.skb_queue, skb);
1463
1462
if (e)
+2
-7
drivers/net/ethernet/realtek/r8169.c
+2
-7
drivers/net/ethernet/realtek/r8169.c
···
2244
2244
void __iomem *ioaddr = tp->mmio_addr;
2245
2245
dma_addr_t paddr = tp->counters_phys_addr;
2246
2246
u32 cmd;
2247
-
bool ret;
2248
2247
2249
2248
RTL_W32(CounterAddrHigh, (u64)paddr >> 32);
2249
+
RTL_R32(CounterAddrHigh);
2250
2250
cmd = (u64)paddr & DMA_BIT_MASK(32);
2251
2251
RTL_W32(CounterAddrLow, cmd);
2252
2252
RTL_W32(CounterAddrLow, cmd | counter_cmd);
2253
2253
2254
-
ret = rtl_udelay_loop_wait_low(tp, &rtl_counters_cond, 10, 1000);
2255
-
2256
-
RTL_W32(CounterAddrLow, 0);
2257
-
RTL_W32(CounterAddrHigh, 0);
2258
-
2259
-
return ret;
2254
+
return rtl_udelay_loop_wait_low(tp, &rtl_counters_cond, 10, 1000);
2260
2255
}
2261
2256
2262
2257
static bool rtl8169_reset_counters(struct net_device *dev)
+2
-2
drivers/net/geneve.c
+2
-2
drivers/net/geneve.c
···
829
829
int mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr) -
830
830
GENEVE_BASE_HLEN - info->options_len - 14;
831
831
832
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
832
+
skb_dst_update_pmtu(skb, mtu);
833
833
}
834
834
835
835
sport = udp_flow_src_port(geneve->net, skb, 1, USHRT_MAX, true);
···
875
875
int mtu = dst_mtu(dst) - sizeof(struct ipv6hdr) -
876
876
GENEVE_BASE_HLEN - info->options_len - 14;
877
877
878
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
878
+
skb_dst_update_pmtu(skb, mtu);
879
879
}
880
880
881
881
sport = udp_flow_src_port(geneve->net, skb, 1, USHRT_MAX, true);
+3
-2
drivers/net/vrf.c
+3
-2
drivers/net/vrf.c
···
673
673
struct sock *sk,
674
674
struct sk_buff *skb)
675
675
{
676
-
/* don't divert multicast */
677
-
if (ipv4_is_multicast(ip_hdr(skb)->daddr))
676
+
/* don't divert multicast or local broadcast */
677
+
if (ipv4_is_multicast(ip_hdr(skb)->daddr) ||
678
+
ipv4_is_lbcast(ip_hdr(skb)->daddr))
678
679
return skb;
679
680
680
681
if (qdisc_tx_is_default(vrf_dev))
+2
-4
drivers/net/vxlan.c
+2
-4
drivers/net/vxlan.c
···
2158
2158
if (skb_dst(skb)) {
2159
2159
int mtu = dst_mtu(ndst) - VXLAN_HEADROOM;
2160
2160
2161
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL,
2162
-
skb, mtu);
2161
+
skb_dst_update_pmtu(skb, mtu);
2163
2162
}
2164
2163
2165
2164
tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
···
2199
2200
if (skb_dst(skb)) {
2200
2201
int mtu = dst_mtu(ndst) - VXLAN6_HEADROOM;
2201
2202
2202
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL,
2203
-
skb, mtu);
2203
+
skb_dst_update_pmtu(skb, mtu);
2204
2204
}
2205
2205
2206
2206
tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
+8
include/net/dst.h
+8
include/net/dst.h
···
521
521
}
522
522
#endif
523
523
524
+
static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu)
525
+
{
526
+
struct dst_entry *dst = skb_dst(skb);
527
+
528
+
if (dst && dst->ops->update_pmtu)
529
+
dst->ops->update_pmtu(dst, NULL, skb, mtu);
530
+
}
531
+
524
532
#endif /* _NET_DST_H */
+10
include/net/net_namespace.h
+10
include/net/net_namespace.h
···
223
223
return net1 == net2;
224
224
}
225
225
226
+
static inline int check_net(const struct net *net)
227
+
{
228
+
return atomic_read(&net->count) != 0;
229
+
}
230
+
226
231
void net_drop_ns(void *);
227
232
228
233
#else
···
248
243
249
244
static inline
250
245
int net_eq(const struct net *net1, const struct net *net2)
246
+
{
247
+
return 1;
248
+
}
249
+
250
+
static inline int check_net(const struct net *net)
251
251
{
252
252
return 1;
253
253
}
+3
net/dccp/ccids/ccid2.c
+3
net/dccp/ccids/ccid2.c
+1
-2
net/ipv4/ip_tunnel.c
+1
-2
net/ipv4/ip_tunnel.c
···
520
520
else
521
521
mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
522
522
523
-
if (skb_dst(skb))
524
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
523
+
skb_dst_update_pmtu(skb, mtu);
525
524
526
525
if (skb->protocol == htons(ETH_P_IP)) {
527
526
if (!skb_is_gso(skb) &&
+1
-1
net/ipv4/ip_vti.c
+1
-1
net/ipv4/ip_vti.c
···
200
200
201
201
mtu = dst_mtu(dst);
202
202
if (skb->len > mtu) {
203
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
203
+
skb_dst_update_pmtu(skb, mtu);
204
204
if (skb->protocol == htons(ETH_P_IP)) {
205
205
icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
206
206
htonl(mtu));
+3
net/ipv4/tcp.c
+3
net/ipv4/tcp.c
···
2298
2298
tcp_send_active_reset(sk, GFP_ATOMIC);
2299
2299
__NET_INC_STATS(sock_net(sk),
2300
2300
LINUX_MIB_TCPABORTONMEMORY);
2301
+
} else if (!check_net(sock_net(sk))) {
2302
+
/* Not possible to send reset; just close */
2303
+
tcp_set_state(sk, TCP_CLOSE);
2301
2304
}
2302
2305
}
2303
2306
+15
net/ipv4/tcp_timer.c
+15
net/ipv4/tcp_timer.c
···
48
48
* to prevent DoS attacks. It is called when a retransmission timeout
49
49
* or zero probe timeout occurs on orphaned socket.
50
50
*
51
+
* Also close if our net namespace is exiting; in that case there is no
52
+
* hope of ever communicating again since all netns interfaces are already
53
+
* down (or about to be down), and we need to release our dst references,
54
+
* which have been moved to the netns loopback interface, so the namespace
55
+
* can finish exiting. This condition is only possible if we are a kernel
56
+
* socket, as those do not hold references to the namespace.
57
+
*
51
58
* Criteria is still not confirmed experimentally and may change.
52
59
* We kill the socket, if:
53
60
* 1. If number of orphaned sockets exceeds an administratively configured
54
61
* limit.
55
62
* 2. If we have strong memory pressure.
63
+
* 3. If our net namespace is exiting.
56
64
*/
57
65
static int tcp_out_of_resources(struct sock *sk, bool do_reset)
58
66
{
···
89
81
__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONMEMORY);
90
82
return 1;
91
83
}
84
+
85
+
if (!check_net(sock_net(sk))) {
86
+
/* Not possible to send reset; just close */
87
+
tcp_done(sk);
88
+
return 1;
89
+
}
90
+
92
91
return 0;
93
92
}
94
93
+2
-4
net/ipv6/ip6_tunnel.c
+2
-4
net/ipv6/ip6_tunnel.c
···
642
642
if (rel_info > dst_mtu(skb_dst(skb2)))
643
643
goto out;
644
644
645
-
skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2,
646
-
rel_info);
645
+
skb_dst_update_pmtu(skb2, rel_info);
647
646
}
648
647
649
648
icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
···
1130
1131
mtu = 576;
1131
1132
}
1132
1133
1133
-
if (skb_dst(skb) && !t->parms.collect_md)
1134
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1134
+
skb_dst_update_pmtu(skb, mtu);
1135
1135
if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1136
1136
*pmtu = mtu;
1137
1137
err = -EMSGSIZE;
+1
-1
net/ipv6/ip6_vti.c
+1
-1
net/ipv6/ip6_vti.c
+2
-2
net/ipv6/sit.c
+2
-2
net/ipv6/sit.c
···
934
934
df = 0;
935
935
}
936
936
937
-
if (tunnel->parms.iph.daddr && skb_dst(skb))
938
-
skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
937
+
if (tunnel->parms.iph.daddr)
938
+
skb_dst_update_pmtu(skb, mtu);
939
939
940
940
if (skb->len > mtu && !skb_is_gso(skb)) {
941
941
icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
+1
-1
net/vmw_vsock/af_vsock.c
+1
-1
net/vmw_vsock/af_vsock.c
···
951
951
* POLLOUT|POLLWRNORM when peer is closed and nothing to read,
952
952
* but local send is not shutdown.
953
953
*/
954
-
if (sk->sk_state == TCP_CLOSE) {
954
+
if (sk->sk_state == TCP_CLOSE || sk->sk_state == TCP_CLOSING) {
955
955
if (!(sk->sk_shutdown & SEND_SHUTDOWN))
956
956
mask |= POLLOUT | POLLWRNORM;
957
957