Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'follow-up-on-deduplicate-cookie-logic'
Willem de Bruijn says:
====================
follow-up on deduplicate cookie logic
1/3: I came across a leftover from cookie deduplication, due to UDP
having two code paths: lockless fast path and locked cork path.
3/3: Even though the leftover was in the fast path, this prompted me
to complete coverage to the cork path.
2/3: That uncovered a subtle API inconsistency in how dontfrag is
configured. It should not be possible to switch DF mid datagram.
====================
Link: https://patch.msgid.link/20250307033620.411611-1-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
+32
-15
+1
include/linux/ipv6.h
+1
include/linux/ipv6.h
+5
-6
net/ipv6/ip6_output.c
+5
-6
net/ipv6/ip6_output.c
···
1386
1386
}
1387
1387
v6_cork->hop_limit = ipc6->hlimit;
1388
1388
v6_cork->tclass = ipc6->tclass;
1389
+
v6_cork->dontfrag = ipc6->dontfrag;
1389
1390
if (rt->dst.flags & DST_XFRM_TUNNEL)
1390
1391
mtu = READ_ONCE(np->pmtudisc) >= IPV6_PMTUDISC_PROBE ?
1391
1392
READ_ONCE(rt->dst.dev->mtu) : dst_mtu(&rt->dst);
···
1422
1421
int getfrag(void *from, char *to, int offset,
1423
1422
int len, int odd, struct sk_buff *skb),
1424
1423
void *from, size_t length, int transhdrlen,
1425
-
unsigned int flags, struct ipcm6_cookie *ipc6)
1424
+
unsigned int flags)
1426
1425
{
1427
1426
struct sk_buff *skb, *skb_prev = NULL;
1428
1427
struct inet_cork *cork = &cork_full->base;
···
1476
1475
if (headersize + transhdrlen > mtu)
1477
1476
goto emsgsize;
1478
1477
1479
-
if (cork->length + length > mtu - headersize && ipc6->dontfrag &&
1478
+
if (cork->length + length > mtu - headersize && v6_cork->dontfrag &&
1480
1479
(sk->sk_protocol == IPPROTO_UDP ||
1481
1480
sk->sk_protocol == IPPROTO_ICMPV6 ||
1482
1481
sk->sk_protocol == IPPROTO_RAW)) {
···
1856
1855
1857
1856
return __ip6_append_data(sk, &sk->sk_write_queue, &inet->cork,
1858
1857
&np->cork, sk_page_frag(sk), getfrag,
1859
-
from, length, transhdrlen, flags, ipc6);
1858
+
from, length, transhdrlen, flags);
1860
1859
}
1861
1860
EXPORT_SYMBOL_GPL(ip6_append_data);
1862
1861
···
2055
2054
ip6_cork_release(cork, &v6_cork);
2056
2055
return ERR_PTR(err);
2057
2056
}
2058
-
if (ipc6->dontfrag < 0)
2059
-
ipc6->dontfrag = inet6_test_bit(DONTFRAG, sk);
2060
2057
2061
2058
err = __ip6_append_data(sk, &queue, cork, &v6_cork,
2062
2059
¤t->task_frag, getfrag, from,
2063
2060
length + exthdrlen, transhdrlen + exthdrlen,
2064
-
flags, ipc6);
2061
+
flags);
2065
2062
if (err) {
2066
2063
__ip6_flush_pending_frames(sk, &queue, cork, &v6_cork);
2067
2064
return ERR_PTR(err);
+7
-4
tools/testing/selftests/net/cmsg_ip.sh
+7
-4
tools/testing/selftests/net/cmsg_ip.sh
···
50
50
# IPV6_DONTFRAG
51
51
for ovr in setsock cmsg both diff; do
52
52
for df in 0 1; do
53
-
for p in u i r; do
53
+
for p in u U i r; do
54
54
[ $p == "u" ] && prot=UDP
55
+
[ $p == "U" ] && prot=UDP
55
56
[ $p == "i" ] && prot=ICMP
56
57
[ $p == "r" ] && prot=RAW
57
58
···
82
81
ip $IPVER -netns $NS route add table 300 prohibit any
83
82
84
83
for ovr in setsock cmsg both diff; do
85
-
for p in u i r; do
84
+
for p in u U i r; do
86
85
[ $p == "u" ] && prot=UDP
86
+
[ $p == "U" ] && prot=UDP
87
87
[ $p == "i" ] && prot=ICMP
88
88
[ $p == "r" ] && prot=RAW
89
89
···
136
134
local -r LIM=4
137
135
138
136
for ovr in setsock cmsg both diff; do
139
-
for p in u i r; do
137
+
for p in u U i r; do
140
138
[ $p == "u" ] && prot=UDP
139
+
[ $p == "U" ] && prot=UDP
141
140
[ $p == "i" ] && prot=ICMP
142
141
[ $p == "r" ] && prot=RAW
143
142
···
169
166
test_ttl_hoplimit -6 $TGT6 hlim
170
167
171
168
# IPV6 exthdr
172
-
for p in u i r; do
169
+
for p in u U i r; do
173
170
# Very basic "does it crash" test
174
171
for h in h d r; do
175
172
$NSEXE ./cmsg_sender -p $p -6 -H $h $TGT6 1234
+19
-5
tools/testing/selftests/net/cmsg_sender.c
+19
-5
tools/testing/selftests/net/cmsg_sender.c
···
33
33
ERN_RECVERR,
34
34
ERN_CMSG_RD,
35
35
ERN_CMSG_RCV,
36
+
ERN_SEND_MORE,
36
37
};
37
38
38
39
struct option_cmsg_u32 {
···
47
46
const char *service;
48
47
unsigned int size;
49
48
unsigned int num_pkt;
49
+
bool msg_more;
50
50
struct {
51
51
unsigned int mark;
52
52
unsigned int dontfrag;
···
96
94
"\t\t-S send() size\n"
97
95
"\t\t-4/-6 Force IPv4 / IPv6 only\n"
98
96
"\t\t-p prot Socket protocol\n"
99
-
"\t\t (u = UDP (default); i = ICMP; r = RAW)\n"
97
+
"\t\t (u = UDP (default); i = ICMP; r = RAW;\n"
98
+
"\t\t U = UDP with MSG_MORE)\n"
100
99
"\n"
101
100
"\t\t-m val Set SO_MARK with given value\n"
102
101
"\t\t-M val Set SO_MARK via setsockopt\n"
···
112
109
"\t\t-l val Set TTL/HOPLIMIT via cmsg\n"
113
110
"\t\t-L val Set TTL/HOPLIMIT via setsockopt\n"
114
111
"\t\t-H type Add an IPv6 header option\n"
115
-
"\t\t (h = HOP; d = DST; r = RTDST)"
116
-
"");
112
+
"\t\t (h = HOP; d = DST; r = RTDST)\n"
113
+
"\n");
117
114
exit(ERN_HELP);
118
115
}
119
116
···
136
133
opt.sock.family = AF_INET6;
137
134
break;
138
135
case 'p':
139
-
if (*optarg == 'u' || *optarg == 'U') {
136
+
if (*optarg == 'u') {
140
137
opt.sock.proto = IPPROTO_UDP;
138
+
} else if (*optarg == 'U') {
139
+
opt.sock.proto = IPPROTO_UDP;
140
+
opt.msg_more = true;
141
141
} else if (*optarg == 'i' || *optarg == 'I') {
142
142
opt.sock.proto = IPPROTO_ICMP;
143
143
} else if (*optarg == 'r') {
···
537
531
cs_write_cmsg(fd, &msg, cbuf, sizeof(cbuf));
538
532
539
533
for (i = 0; i < opt.num_pkt; i++) {
540
-
err = sendmsg(fd, &msg, 0);
534
+
err = sendmsg(fd, &msg, opt.msg_more ? MSG_MORE : 0);
541
535
if (err < 0) {
542
536
if (!opt.silent_send)
543
537
fprintf(stderr, "send failed: %s\n", strerror(errno));
···
547
541
fprintf(stderr, "short send\n");
548
542
err = ERN_SEND_SHORT;
549
543
goto err_out;
544
+
}
545
+
if (opt.msg_more) {
546
+
err = write(fd, NULL, 0);
547
+
if (err < 0) {
548
+
fprintf(stderr, "send more: %s\n", strerror(errno));
549
+
err = ERN_SEND_MORE;
550
+
goto err_out;
551
+
}
550
552
}
551
553
}
552
554
err = ERN_SUCCESS;