follow-up on deduplicate cookie logic
Subsystems (4)
cover follow-up on deduplicate cookie logic
From: Willem de Bruijn <[email protected]>
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.
Willem de Bruijn (3):
ipv6: remove leftover ip6 cookie initializer
ipv6: save dontfrag in cork
selftests/net: expand cmsg_ip with MSG_MORE
include/linux/ipv6.h | 1 + net/ipv6/ip6_output.c | 11 +++++------ tools/testing/selftests/net/cmsg_ip.sh | 11 +++++++---- tools/testing/selftests/net/cmsg_sender.c | 24 ++++++++++++++++++----- 4 files changed, 32 insertions(+), 15 deletions(-)
Signature · 1 line
2.49.0.rc0.332.g42c0ae87b1-goog
PATCH 1/3 ipv6: remove leftover ip6 cookie initializer
From: Willem de Bruijn <[email protected]>
As of the blamed commit ipc6.dontfrag is always initialized at the start of udpv6_sendmsg, by ipcm6_init_sk, to either 0 or 1.
Later checks against -1 are no longer needed and the branches are now dead code.
The blamed commit had removed those branches. But I had overlooked this one case.
UDP has both a lockless fast path and a slower path for corked requests. This branch remained in the fast path.
--- net/ipv6/ip6_output.c | 2 -- 1 file changed, 2 deletions(-)
net/ipv6/ip6_output.c +0 -2
@@ -2054,8 +2054,6 @@ struct sk_buff *ip6_make_skb(struct sock *sk,ip6_cork_release (cork, &v6_cork);return ERR_PTR (err); } -if (ipc6->dontfrag <0 ) - ipc6->dontfrag =inet6_test_bit (DONTFRAG , sk); err = __ip6_append_data(sk, &queue, cork, &v6_cork, ¤t->task_frag, getfrag, from,
Signature · 1 line
2.49.0.rc0.332.g42c0ae87b1-goog
Re: [PATCH net-next 1/3] ipv6: remove leftover ip6 cookie initializer
Quoted 9 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
As of the blamed commit ipc6.dontfrag is always initialized at the start of
udpv6_sendmsg, byipcm6_init_sk, to either 0 or 1.Later checks against -1 are no longer needed and the branches are now dead code.
Quoted 8 lines
The blamed commit had removed those branches. But I had overlooked this one case.
UDP has both a lockless fast path and a slower path for corked requests. This branch remained in the fast path.
Fixes 096208592b09 ("ipv6: replace ipcm6_init calls with ipcm6_init_sk")Signed-off-by Willem de Bruijn <[email protected]>
PATCH 2/3 ipv6: save dontfrag in cork
In reply to Willem de BruijnFrom: Willem de Bruijn <[email protected]>
When spanning datagram construction over multiple send calls using MSG_MORE, per datagram settings are configured on the first send.
That is when ip(6)_setup_cork stores these settings for subsequent use in __ip(6)_append_data and others.
The only flag that escaped this was dontfrag. As a result, a datagram could be constructed with df=0 on the first sendmsg, but df=1 on a next. Which is what cmsg_ip.sh does in an upcoming MSG_MORE test in the "diff" scenario.
Changing datagram conditions in the middle of constructing an skb makes this already complex code path even more convoluted. It is here unintentional. Bring this flag in line with expected sockopt/cmsg behavior.
And stop passing ipc6 to __ip6_append_data, to avoid such issues in the future. This is already the case for __ip_append_data.
inet6_cork had a 6 byte hole, so the 1B flag has no impact.
--- include/linux/ipv6.h | 1 + net/ipv6/ip6_output.c | 9 +++++---- 2 files changed, 6 insertions(+), 4 deletions(-)
include/linux/ipv6.h +1 -0
@@ -207,6 +207,7 @@ struct inet6_cork {struct ipv6_txoptions *opt;u8 hop_limit;u8 tclass; + u8 dontfrag:1; };/* struct ipv6_pinfo - ipv6 private area */
net/ipv6/ip6_output.c +5 -4
@@ -1386,6 +1386,7 @@ static int ip6_setup_cork(struct sock *sk, struct inet_cork_full *cork, } v6_cork->hop_limit = ipc6->hlimit; v6_cork->tclass = ipc6->tclass; + v6_cork->dontfrag = ipc6->dontfrag;if (rt->dst.flags &DST_XFRM_TUNNEL ) mtu =READ_ONCE (np->pmtudisc) >=IPV6_PMTUDISC_PROBE ?READ_ONCE (rt->dst.dev->mtu) :dst_mtu (&rt->dst);
@@ -1421,7 +1422,7 @@ static int __ip6_append_data(struct sock *sk,int getfrag (void *from,char *to,int offset,int len,int odd,struct sk_buff *skb), void *from, size_t length, int transhdrlen, - unsigned int flags, struct ipcm6_cookie *ipc6) + unsigned int flags) { struct sk_buff *skb, *skb_prev = NULL;struct inet_cork *cork = &cork_full->base;
@@ -1475,7 +1476,7 @@ static int __ip6_append_data(struct sock *sk,if (headersize + transhdrlen > mtu) goto emsgsize; -if (cork->length + length > mtu - headersize && ipc6->dontfrag && +if (cork->length + length > mtu - headersize && v6_cork->dontfrag && (sk->sk_protocol ==IPPROTO_UDP || sk->sk_protocol ==IPPROTO_ICMPV6 || sk->sk_protocol ==IPPROTO_RAW )) {
@@ -1855,7 +1856,7 @@ int ip6_append_data(struct sock *sk,return __ip6_append_data (sk, &sk->sk_write_queue, &inet->cork, &np->cork,sk_page_frag (sk), getfrag, - from, length, transhdrlen, flags, ipc6); + from, length, transhdrlen, flags); }EXPORT_SYMBOL_GPL (ip6_append_data);
@@ -2058,7 +2059,7 @@ struct sk_buff *ip6_make_skb(struct sock *sk, err =__ip6_append_data (sk, &queue, cork, &v6_cork, ¤t->task_frag, getfrag, from, length + exthdrlen, transhdrlen + exthdrlen, - flags, ipc6); + flags);if (err) {__ip6_flush_pending_frames (sk, &queue, cork, &v6_cork);return ERR_PTR (err);
Signature · 1 line
2.49.0.rc0.332.g42c0ae87b1-goog
Re: [PATCH net-next 2/3] ipv6: save dontfrag in cork
Quoted 19 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
When spanning datagram construction over multiple send calls using
MSG_MORE, per datagram settings are configured on the first send.That is when ip(6)_setup_cork stores these settings for subsequent use in __ip(6)_append_data and others.
The only flag that escaped this was dontfrag. As a result, a datagram could be constructed with df=0 on the first sendmsg, but df=1 on a next. Which is what
cmsg_ip.shdoes in an upcomingMSG_MOREtest in the "diff" scenario.Changing datagram conditions in the middle of constructing an skb makes this already complex code path even more convoluted. It is here unintentional. Bring this flag in line with expected sockopt/cmsg behavior.
Quoted 6 lines
And stop passing ipc6 to
__ip6_append_data, to avoid such issues in the future. This is already the case for__ip_append_data.
inet6_corkhad a 6 byte hole, so the 1B flag has no impact.Signed-off-by Willem de Bruijn <[email protected]>
PATCH 3/3 selftests/net: expand cmsg_ip with MSG_MORE
In reply to Willem de BruijnFrom: Willem de Bruijn <[email protected]>
UDP send with MSG_MORE takes a slightly different path than the lockless fast path.
For completeness, add coverage to this case too.
Pass MSG_MORE on the initial sendmsg, then follow up with a zero byte write to unplug the cork.
Unrelated: also add two missing endlines in usage().
--- tools/testing/selftests/net/cmsg_ip.sh | 11 +++++++---- tools/testing/selftests/net/cmsg_sender.c | 24 ++++++++++++++++++----- 2 files changed, 26 insertions(+), 9 deletions(-)
tools/testing/selftests/net/cmsg_ip.sh +7 -4
@@ -50,8 +50,9 @@ check_result() { # IPV6_DONTFRAG for ovr in setsock cmsg both diff; do for df in 0 1; do - for p in u i r; do + for p in u U i r; do [ $p == "u" ] && prot=UDP + [ $p == "U" ] && prot=UDP [ $p == "i" ] && prot=ICMP [ $p == "r" ] && prot=RAW
@@ -81,8 +82,9 @@ test_dscp() { ip $IPVER -netns $NS route add table 300 prohibit any for ovr in setsock cmsg both diff; do - for p in u i r; do + for p in u U i r; do [ $p == "u" ] && prot=UDP + [ $p == "U" ] && prot=UDP [ $p == "i" ] && prot=ICMP [ $p == "r" ] && prot=RAW
@@ -134,8 +136,9 @@ test_ttl_hoplimit() { local -r LIM=4 for ovr in setsock cmsg both diff; do - for p in u i r; do + for p in u U i r; do [ $p == "u" ] && prot=UDP + [ $p == "U" ] && prot=UDP [ $p == "i" ] && prot=ICMP [ $p == "r" ] && prot=RAW
@@ -166,7 +169,7 @@ test_ttl_hoplimit -4 $TGT4 ttl test_ttl_hoplimit -6 $TGT6 hlim # IPV6 exthdr -for p in u i r; do +for p in u U i r; do # Very basic "does it crash" test for h in h d r; do $NSEXE ./cmsg_sender -p $p -6 -H $h $TGT6 1234
tools/testing/selftests/net/cmsg_sender.c +19 -5
@@ -33,6 +33,7 @@ enum {ERN_RECVERR ,ERN_CMSG_RD ,ERN_CMSG_RCV , + ERN_SEND_MORE, }; struct option_cmsg_u32 {
@@ -46,6 +47,7 @@ struct options {const char *service;unsigned int size;unsigned int num_pkt; +bool msg_more;struct {unsigned int mark;unsigned int dontfrag;
@@ -94,7 +96,8 @@ static void __attribute__((noreturn)) cs_usage(const char *bin)"\t\t-S send() size\n" "\t\t-4/-6 Force IPv4 / IPv6 only\n" "\t\t-p prot Socket protocol\n" -"\t\t (u = UDP (default); i = ICMP; r = RAW)\n" +"\t\t (u = UDP (default); i = ICMP; r = RAW;\n" +"\t\t U = UDP with MSG_MORE)\n" "\n" "\t\t-m val Set SO_MARK with given value\n" "\t\t-M val Set SO_MARK via setsockopt\n"
@@ -109,8 +112,8 @@ static void __attribute__((noreturn)) cs_usage(const char *bin)"\t\t-l val Set TTL/HOPLIMIT via cmsg\n" "\t\t-L val Set TTL/HOPLIMIT via setsockopt\n" "\t\t-H type Add an IPv6 header option\n" -"\t\t (h = HOP; d = DST; r = RTDST)" - ""); +"\t\t (h = HOP; d = DST; r = RTDST)\n" + "\n");exit (ERN_HELP ); }
@@ -133,8 +136,11 @@ static void cs_parse_args(int argc, char *argv[]) opt.sock.family =AF_INET6 ;break ;case 'p' : -if (*optarg =='u' || *optarg =='U' ) { +if (*optarg =='u' ) { opt.sock.proto =IPPROTO_UDP ; + }else if (*optarg =='U' ) { + opt.sock.proto =IPPROTO_UDP ; + opt.msg_more = true; }else if (*optarg =='i' || *optarg =='I' ) { opt.sock.proto =IPPROTO_ICMP ; }else if (*optarg =='r' ) {
@@ -531,7 +537,7 @@ int main(int argc, char *argv[])cs_write_cmsg (fd, &msg, cbuf,sizeof (cbuf));for (i =0 ; i < opt.num_pkt; i++) { - err =sendmsg (fd, &msg,0 ); + err =sendmsg (fd, &msg, opt.msg_more ?MSG_MORE :0 );if (err <0 ) {if (!opt.silent_send)fprintf (stderr,"send failed: %s\n" ,strerror (errno));
@@ -542,6 +548,14 @@ int main(int argc, char *argv[]) err =ERN_SEND_SHORT ; goto err_out; } +if (opt.msg_more) { + err =write (fd,NULL ,0 ); +if (err <0 ) { +fprintf (stderr,"send more: %s\n" ,strerror (errno)); + err =ERN_SEND_MORE ; + goto err_out; + } + } } err =ERN_SUCCESS ;
Signature · 1 line
2.49.0.rc0.332.g42c0ae87b1-goog
Re: [PATCH net-next 3/3] selftests/net: expand cmsg_ip with MSG_MORE
Quoted 6 lines · Willem de Bruijn
From: Willem de Bruijn <[email protected]>
UDP send with
MSG_MOREtakes a slightly different path than the lockless fast path.
Quoted 8 lines
For completeness, add coverage to this case too.
Pass
MSG_MOREon the initial sendmsg, then follow up with a zero byte write to unplug the cork.Unrelated: also add two missing endlines in
usage().Signed-off-by Willem de Bruijn <[email protected]>
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Quoted 9 lines · On Thu, 6 Mar 2025 22:34:07 -0500 you wrote:
From: Willem de Bruijn <[email protected]>
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.
[...]
Here is the summary with links:
- [net-next,1/3] ipv6: remove leftover ip6 cookie initializer
https://git.kernel.org/netdev/net-next/c/54580ccdd8a9 - [net-next,2/3] ipv6: save dontfrag in cork https://git.kernel.org/netdev/net-next/c/a18dfa9925b9 - [net-next,3/3] selftests/net: expand cmsg_ip with MSG_MORE https://git.kernel.org/netdev/net-next/c/0922cb68edfd
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html