Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'add-support-for-so_priority-cmsg'
Anna Emese Nyiri says:
====================
Add support for SO_PRIORITY cmsg
Introduce a new helper function, `sk_set_prio_allowed`,
to centralize the logic for validating priority settings.
Add support for the `SO_PRIORITY` control message,
enabling user-space applications to set socket priority
via control messages (cmsg).
====================
Link: https://patch.msgid.link/20241213084457.45120-1-annaemesenyiri@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
+228
-16
+2
arch/alpha/include/uapi/asm/socket.h
+2
arch/alpha/include/uapi/asm/socket.h
+2
arch/mips/include/uapi/asm/socket.h
+2
arch/mips/include/uapi/asm/socket.h
+2
arch/parisc/include/uapi/asm/socket.h
+2
arch/parisc/include/uapi/asm/socket.h
+2
arch/sparc/include/uapi/asm/socket.h
+2
arch/sparc/include/uapi/asm/socket.h
+1
-1
include/net/inet_sock.h
+1
-1
include/net/inet_sock.h
+1
-1
include/net/ip.h
+1
-1
include/net/ip.h
···
81
81
__u8 protocol;
82
82
__u8 ttl;
83
83
__s16 tos;
84
-
char priority;
85
84
__u16 gso_size;
86
85
};
87
86
···
95
96
ipcm_init(ipcm);
96
97
97
98
ipcm->sockc.mark = READ_ONCE(inet->sk.sk_mark);
99
+
ipcm->sockc.priority = READ_ONCE(inet->sk.sk_priority);
98
100
ipcm->sockc.tsflags = READ_ONCE(inet->sk.sk_tsflags);
99
101
ipcm->oif = READ_ONCE(inet->sk.sk_bound_dev_if);
100
102
ipcm->addr = inet->inet_saddr;
+6
-2
include/net/sock.h
+6
-2
include/net/sock.h
···
953
953
SOCK_XDP, /* XDP is attached */
954
954
SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */
955
955
SOCK_RCVMARK, /* Receive SO_MARK ancillary data with packet */
956
+
SOCK_RCVPRIORITY, /* Receive SO_PRIORITY ancillary data with packet */
956
957
};
957
958
958
959
#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
···
1815
1814
u32 mark;
1816
1815
u32 tsflags;
1817
1816
u32 ts_opt_id;
1817
+
u32 priority;
1818
1818
};
1819
1819
1820
1820
static inline void sockcm_init(struct sockcm_cookie *sockc,
1821
1821
const struct sock *sk)
1822
1822
{
1823
1823
*sockc = (struct sockcm_cookie) {
1824
-
.tsflags = READ_ONCE(sk->sk_tsflags)
1824
+
.tsflags = READ_ONCE(sk->sk_tsflags),
1825
+
.priority = READ_ONCE(sk->sk_priority),
1825
1826
};
1826
1827
}
1827
1828
···
2661
2658
{
2662
2659
#define FLAGS_RECV_CMSGS ((1UL << SOCK_RXQ_OVFL) | \
2663
2660
(1UL << SOCK_RCVTSTAMP) | \
2664
-
(1UL << SOCK_RCVMARK))
2661
+
(1UL << SOCK_RCVMARK) |\
2662
+
(1UL << SOCK_RCVPRIORITY))
2665
2663
#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2666
2664
SOF_TIMESTAMPING_RAW_HARDWARE)
2667
2665
+2
include/uapi/asm-generic/socket.h
+2
include/uapi/asm-generic/socket.h
+1
-1
net/can/raw.c
+1
-1
net/can/raw.c
+23
-3
net/core/sock.c
+23
-3
net/core/sock.c
···
454
454
return 0;
455
455
}
456
456
457
+
static bool sk_set_prio_allowed(const struct sock *sk, int val)
458
+
{
459
+
return ((val >= TC_PRIO_BESTEFFORT && val <= TC_PRIO_INTERACTIVE) ||
460
+
sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) ||
461
+
sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN));
462
+
}
463
+
457
464
static bool sock_needs_netstamp(const struct sock *sk)
458
465
{
459
466
switch (sk->sk_family) {
···
1200
1193
/* handle options which do not require locking the socket. */
1201
1194
switch (optname) {
1202
1195
case SO_PRIORITY:
1203
-
if ((val >= 0 && val <= 6) ||
1204
-
sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) ||
1205
-
sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1196
+
if (sk_set_prio_allowed(sk, val)) {
1206
1197
sock_set_priority(sk, val);
1207
1198
return 0;
1208
1199
}
···
1517
1512
break;
1518
1513
case SO_RCVMARK:
1519
1514
sock_valbool_flag(sk, SOCK_RCVMARK, valbool);
1515
+
break;
1516
+
1517
+
case SO_RCVPRIORITY:
1518
+
sock_valbool_flag(sk, SOCK_RCVPRIORITY, valbool);
1520
1519
break;
1521
1520
1522
1521
case SO_RXQ_OVFL:
···
1949
1940
1950
1941
case SO_RCVMARK:
1951
1942
v.val = sock_flag(sk, SOCK_RCVMARK);
1943
+
break;
1944
+
1945
+
case SO_RCVPRIORITY:
1946
+
v.val = sock_flag(sk, SOCK_RCVPRIORITY);
1952
1947
break;
1953
1948
1954
1949
case SO_RXQ_OVFL:
···
2954
2941
/* SCM_RIGHTS and SCM_CREDENTIALS are semantically in SOL_UNIX. */
2955
2942
case SCM_RIGHTS:
2956
2943
case SCM_CREDENTIALS:
2944
+
break;
2945
+
case SO_PRIORITY:
2946
+
if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2947
+
return -EINVAL;
2948
+
if (!sk_set_prio_allowed(sk, *(u32 *)CMSG_DATA(cmsg)))
2949
+
return -EPERM;
2950
+
sockc->priority = *(u32 *)CMSG_DATA(cmsg);
2957
2951
break;
2958
2952
default:
2959
2953
return -EINVAL;
+2
-2
net/ipv4/ip_output.c
+2
-2
net/ipv4/ip_output.c
···
1333
1333
cork->ttl = ipc->ttl;
1334
1334
cork->tos = ipc->tos;
1335
1335
cork->mark = ipc->sockc.mark;
1336
-
cork->priority = ipc->priority;
1336
+
cork->priority = ipc->sockc.priority;
1337
1337
cork->transmit_time = ipc->sockc.transmit_time;
1338
1338
cork->tx_flags = 0;
1339
1339
sock_tx_timestamp(sk, &ipc->sockc, &cork->tx_flags);
···
1470
1470
ip_options_build(skb, opt, cork->addr, rt);
1471
1471
}
1472
1472
1473
-
skb->priority = (cork->tos != -1) ? cork->priority: READ_ONCE(sk->sk_priority);
1473
+
skb->priority = cork->priority;
1474
1474
skb->mark = cork->mark;
1475
1475
if (sk_is_tcp(sk))
1476
1476
skb_set_delivery_time(skb, cork->transmit_time, SKB_CLOCK_MONOTONIC);
+1
-1
net/ipv4/ip_sockglue.c
+1
-1
net/ipv4/ip_sockglue.c
+1
-1
net/ipv4/raw.c
+1
-1
net/ipv4/raw.c
···
358
358
skb_reserve(skb, hlen);
359
359
360
360
skb->protocol = htons(ETH_P_IP);
361
-
skb->priority = READ_ONCE(sk->sk_priority);
361
+
skb->priority = sockc->priority;
362
362
skb->mark = sockc->mark;
363
363
skb_set_delivery_type_by_clockid(skb, sockc->transmit_time, sk->sk_clockid);
364
364
skb_dst_set(skb, &rt->dst);
+2
-1
net/ipv6/ip6_output.c
+2
-1
net/ipv6/ip6_output.c
···
1401
1401
cork->base.gso_size = ipc6->gso_size;
1402
1402
cork->base.tx_flags = 0;
1403
1403
cork->base.mark = ipc6->sockc.mark;
1404
+
cork->base.priority = ipc6->sockc.priority;
1404
1405
sock_tx_timestamp(sk, &ipc6->sockc, &cork->base.tx_flags);
1405
1406
if (ipc6->sockc.tsflags & SOCKCM_FLAG_TS_OPT_ID) {
1406
1407
cork->base.flags |= IPCORK_TS_OPT_ID;
···
1943
1942
hdr->saddr = fl6->saddr;
1944
1943
hdr->daddr = *final_dst;
1945
1944
1946
-
skb->priority = READ_ONCE(sk->sk_priority);
1945
+
skb->priority = cork->base.priority;
1947
1946
skb->mark = cork->base.mark;
1948
1947
if (sk_is_tcp(sk))
1949
1948
skb_set_delivery_time(skb, cork->base.transmit_time, SKB_CLOCK_MONOTONIC);
+1
net/ipv6/ping.c
+1
net/ipv6/ping.c
+2
-1
net/ipv6/raw.c
+2
-1
net/ipv6/raw.c
···
619
619
skb_reserve(skb, hlen);
620
620
621
621
skb->protocol = htons(ETH_P_IPV6);
622
-
skb->priority = READ_ONCE(sk->sk_priority);
622
+
skb->priority = sockc->priority;
623
623
skb->mark = sockc->mark;
624
624
skb_set_delivery_type_by_clockid(skb, sockc->transmit_time, sk->sk_clockid);
625
625
···
780
780
ipcm6_init(&ipc6);
781
781
ipc6.sockc.tsflags = READ_ONCE(sk->sk_tsflags);
782
782
ipc6.sockc.mark = fl6.flowi6_mark;
783
+
ipc6.sockc.priority = READ_ONCE(sk->sk_priority);
783
784
784
785
if (sin6) {
785
786
if (addr_len < SIN6_LEN_RFC2133)
+1
net/ipv6/udp.c
+1
net/ipv6/udp.c
+1
-1
net/packet/af_packet.c
+1
-1
net/packet/af_packet.c
+11
net/socket.c
+11
net/socket.c
···
1008
1008
}
1009
1009
}
1010
1010
1011
+
static void sock_recv_priority(struct msghdr *msg, struct sock *sk,
1012
+
struct sk_buff *skb)
1013
+
{
1014
+
if (sock_flag(sk, SOCK_RCVPRIORITY) && skb) {
1015
+
__u32 priority = skb->priority;
1016
+
1017
+
put_cmsg(msg, SOL_SOCKET, SO_PRIORITY, sizeof(__u32), &priority);
1018
+
}
1019
+
}
1020
+
1011
1021
void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1012
1022
struct sk_buff *skb)
1013
1023
{
1014
1024
sock_recv_timestamp(msg, sk, skb);
1015
1025
sock_recv_drops(msg, sk, skb);
1016
1026
sock_recv_mark(msg, sk, skb);
1027
+
sock_recv_priority(msg, sk, skb);
1017
1028
}
1018
1029
EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1019
1030
+2
tools/include/uapi/asm-generic/socket.h
+2
tools/include/uapi/asm-generic/socket.h
+1
tools/testing/selftests/net/Makefile
+1
tools/testing/selftests/net/Makefile
+10
-1
tools/testing/selftests/net/cmsg_sender.c
+10
-1
tools/testing/selftests/net/cmsg_sender.c
···
59
59
unsigned int proto;
60
60
} sock;
61
61
struct option_cmsg_u32 mark;
62
+
struct option_cmsg_u32 priority;
62
63
struct {
63
64
bool ena;
64
65
unsigned int delay;
···
98
97
"\n"
99
98
"\t\t-m val Set SO_MARK with given value\n"
100
99
"\t\t-M val Set SO_MARK via setsockopt\n"
100
+
"\t\t-P val Set SO_PRIORITY via setsockopt\n"
101
+
"\t\t-Q val Set SO_PRIORITY via cmsg\n"
101
102
"\t\t-d val Set SO_TXTIME with given delay (usec)\n"
102
103
"\t\t-t Enable time stamp reporting\n"
103
104
"\t\t-f val Set don't fragment via cmsg\n"
···
118
115
{
119
116
int o;
120
117
121
-
while ((o = getopt(argc, argv, "46sS:p:P:m:M:n:d:tf:F:c:C:l:L:H:")) != -1) {
118
+
while ((o = getopt(argc, argv, "46sS:p:P:m:M:n:d:tf:F:c:C:l:L:H:Q:")) != -1) {
122
119
switch (o) {
123
120
case 's':
124
121
opt.silent_send = true;
···
150
147
case 'm':
151
148
opt.mark.ena = true;
152
149
opt.mark.val = atoi(optarg);
150
+
break;
151
+
case 'Q':
152
+
opt.priority.ena = true;
153
+
opt.priority.val = atoi(optarg);
153
154
break;
154
155
case 'M':
155
156
opt.sockopt.mark = atoi(optarg);
···
259
252
260
253
ca_write_cmsg_u32(cbuf, cbuf_sz, &cmsg_len,
261
254
SOL_SOCKET, SO_MARK, &opt.mark);
255
+
ca_write_cmsg_u32(cbuf, cbuf_sz, &cmsg_len,
256
+
SOL_SOCKET, SO_PRIORITY, &opt.priority);
262
257
ca_write_cmsg_u32(cbuf, cbuf_sz, &cmsg_len,
263
258
SOL_IPV6, IPV6_DONTFRAG, &opt.v6.dontfrag);
264
259
ca_write_cmsg_u32(cbuf, cbuf_sz, &cmsg_len,
+151
tools/testing/selftests/net/cmsg_so_priority.sh
+151
tools/testing/selftests/net/cmsg_so_priority.sh
···
1
+
#!/bin/bash
2
+
# SPDX-License-Identifier: GPL-2.0
3
+
4
+
source lib.sh
5
+
6
+
readonly KSFT_SKIP=4
7
+
8
+
IP4=192.0.2.1/24
9
+
TGT4=192.0.2.2
10
+
TGT4_RAW=192.0.2.3
11
+
IP6=2001:db8::1/64
12
+
TGT6=2001:db8::2
13
+
TGT6_RAW=2001:db8::3
14
+
PORT=1234
15
+
TOTAL_TESTS=0
16
+
FAILED_TESTS=0
17
+
18
+
if ! command -v jq &> /dev/null; then
19
+
echo "SKIP cmsg_so_priroity.sh test: jq is not installed." >&2
20
+
exit "$KSFT_SKIP"
21
+
fi
22
+
23
+
check_result() {
24
+
((TOTAL_TESTS++))
25
+
if [ "$1" -ne 0 ]; then
26
+
((FAILED_TESTS++))
27
+
fi
28
+
}
29
+
30
+
cleanup()
31
+
{
32
+
cleanup_ns $NS
33
+
}
34
+
35
+
trap cleanup EXIT
36
+
37
+
setup_ns NS
38
+
39
+
create_filter() {
40
+
local handle=$1
41
+
local vlan_prio=$2
42
+
local ip_type=$3
43
+
local proto=$4
44
+
local dst_ip=$5
45
+
local ip_proto
46
+
47
+
if [[ "$proto" == "u" ]]; then
48
+
ip_proto="udp"
49
+
elif [[ "$ip_type" == "ipv4" && "$proto" == "i" ]]; then
50
+
ip_proto="icmp"
51
+
elif [[ "$ip_type" == "ipv6" && "$proto" == "i" ]]; then
52
+
ip_proto="icmpv6"
53
+
fi
54
+
55
+
tc -n $NS filter add dev dummy1 \
56
+
egress pref 1 handle "$handle" proto 802.1q \
57
+
flower vlan_prio "$vlan_prio" vlan_ethtype "$ip_type" \
58
+
dst_ip "$dst_ip" ${ip_proto:+ip_proto $ip_proto} \
59
+
action pass
60
+
}
61
+
62
+
ip -n $NS link set dev lo up
63
+
ip -n $NS link add name dummy1 up type dummy
64
+
65
+
ip -n $NS link add link dummy1 name dummy1.10 up type vlan id 10 \
66
+
egress-qos-map 0:0 1:1 2:2 3:3 4:4 5:5 6:6 7:7
67
+
68
+
ip -n $NS address add $IP4 dev dummy1.10
69
+
ip -n $NS address add $IP6 dev dummy1.10 nodad
70
+
71
+
ip netns exec $NS sysctl -wq net.ipv4.ping_group_range='0 2147483647'
72
+
73
+
ip -n $NS neigh add $TGT4 lladdr 00:11:22:33:44:55 nud permanent \
74
+
dev dummy1.10
75
+
ip -n $NS neigh add $TGT6 lladdr 00:11:22:33:44:55 nud permanent \
76
+
dev dummy1.10
77
+
ip -n $NS neigh add $TGT4_RAW lladdr 00:11:22:33:44:66 nud permanent \
78
+
dev dummy1.10
79
+
ip -n $NS neigh add $TGT6_RAW lladdr 00:11:22:33:44:66 nud permanent \
80
+
dev dummy1.10
81
+
82
+
tc -n $NS qdisc add dev dummy1 clsact
83
+
84
+
FILTER_COUNTER=10
85
+
86
+
for i in 4 6; do
87
+
for proto in u i r; do
88
+
echo "Test IPV$i, prot: $proto"
89
+
for priority in {0..7}; do
90
+
if [[ $i == 4 && $proto == "r" ]]; then
91
+
TGT=$TGT4_RAW
92
+
elif [[ $i == 6 && $proto == "r" ]]; then
93
+
TGT=$TGT6_RAW
94
+
elif [ $i == 4 ]; then
95
+
TGT=$TGT4
96
+
else
97
+
TGT=$TGT6
98
+
fi
99
+
100
+
handle="${FILTER_COUNTER}${priority}"
101
+
102
+
create_filter $handle $priority ipv$i $proto $TGT
103
+
104
+
pkts=$(tc -n $NS -j -s filter show dev dummy1 egress \
105
+
| jq ".[] | select(.options.handle == ${handle}) | \
106
+
.options.actions[0].stats.packets")
107
+
108
+
if [[ $pkts == 0 ]]; then
109
+
check_result 0
110
+
else
111
+
echo "prio $priority: expected 0, got $pkts"
112
+
check_result 1
113
+
fi
114
+
115
+
ip netns exec $NS ./cmsg_sender -$i -Q $priority \
116
+
-p $proto $TGT $PORT
117
+
118
+
pkts=$(tc -n $NS -j -s filter show dev dummy1 egress \
119
+
| jq ".[] | select(.options.handle == ${handle}) | \
120
+
.options.actions[0].stats.packets")
121
+
if [[ $pkts == 1 ]]; then
122
+
check_result 0
123
+
else
124
+
echo "prio $priority -Q: expected 1, got $pkts"
125
+
check_result 1
126
+
fi
127
+
128
+
ip netns exec $NS ./cmsg_sender -$i -P $priority \
129
+
-p $proto $TGT $PORT
130
+
131
+
pkts=$(tc -n $NS -j -s filter show dev dummy1 egress \
132
+
| jq ".[] | select(.options.handle == ${handle}) | \
133
+
.options.actions[0].stats.packets")
134
+
if [[ $pkts == 2 ]]; then
135
+
check_result 0
136
+
else
137
+
echo "prio $priority -P: expected 2, got $pkts"
138
+
check_result 1
139
+
fi
140
+
done
141
+
FILTER_COUNTER=$((FILTER_COUNTER + 10))
142
+
done
143
+
done
144
+
145
+
if [ $FAILED_TESTS -ne 0 ]; then
146
+
echo "FAIL - $FAILED_TESTS/$TOTAL_TESTS tests failed"
147
+
exit 1
148
+
else
149
+
echo "OK - All $TOTAL_TESTS tests passed"
150
+
exit 0
151
+
fi