When reading received messages with MSG_PEEK, we sometines have to read
the leading bytes of the stream several times, only to reach the bytes
we really want. This is clearly non-optimal.
What we would want is something similar to pread/preadv(), but working
even for tcp sockets. At the same time, we obviously don't want to add
any new arguments to the recv/recvmsg() calls.
In this commit, we allow the user to set iovec.iov_base in the first
vector entry to NULL. This tells the socket to skip the first entry,
hence making the iov_len field of that entry indicate the offset value.
This way, there is no need to add any new arguments.
This change is simple and non-intrusive, and should be safe addition to
the socket API. We have measured it to give a throughput improvement of
8-10 % for the protocol splicer 'passst', which is used in KubeVirt
containers.
Signed-off-by: Jon Maloy <jmaloy(a)redhat.com>
works with original msghdr
Signed-off-by: Jon Maloy <jmaloy(a)redhat.com>
---
net/ipv4/tcp.c | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 33f559f491c8..1d89337e89b6 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2428,6 +2428,7 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg,
size_t len,
struct tcp_sock *tp = tcp_sk(sk);
int copied = 0;
u32 peek_seq;
+ u32 peek_offset;
u32 *seq;
unsigned long used;
int err;
@@ -2435,7 +2436,6 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg,
size_t len,
long timeo;
struct sk_buff *skb, *last;
u32 urg_hole = 0;
-
err = -ENOTCONN;
if (sk->sk_state == TCP_LISTEN)
goto out;
@@ -2469,6 +2469,14 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg,
size_t len,
if (flags & MSG_PEEK) {
peek_seq = tp->copied_seq;
seq = &peek_seq;
+ if (msg->msg_iter.iov[0].iov_base == NULL) {
+ peek_offset = msg->msg_iter.iov[0].iov_len;
+ msg->msg_iter.iov = &msg->msg_iter.iov[1];
+ msg->msg_iter.nr_segs -= 1;
+ msg->msg_iter.count -= peek_offset;
+ len -= peek_offset;
+ *seq += peek_offset;
+ }
}
target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
--
2.39.0
(Added <passt-dev(a)passt.top> to distribution list. I found that my first approach was too simplistic, since it only moved the reading area forward in the receive buffer, but continue to fill in iov[0] with the indicated length. This commit exactly what we want: we indicate a NULL pointer in iov[0], but want the actually read bytes to end up in the remaining entries, and also the returned value to indicate the actually read length. I look forward to feedback to this, then I can hopefully post it to the netdev list next week. ///jon On 2023-06-22 22:12, Jon Maloy wrote:When reading received messages with MSG_PEEK, we sometines have to read the leading bytes of the stream several times, only to reach the bytes we really want. This is clearly non-optimal. What we would want is something similar to pread/preadv(), but working even for tcp sockets. At the same time, we obviously don't want to add any new arguments to the recv/recvmsg() calls. In this commit, we allow the user to set iovec.iov_base in the first vector entry to NULL. This tells the socket to skip the first entry, hence making the iov_len field of that entry indicate the offset value. This way, there is no need to add any new arguments. This change is simple and non-intrusive, and should be safe addition to the socket API. We have measured it to give a throughput improvement of 8-10 % for the protocol splicer 'passst', which is used in KubeVirt containers. Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> works with original msghdr Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> --- net/ipv4/tcp.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 33f559f491c8..1d89337e89b6 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2428,6 +2428,7 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, struct tcp_sock *tp = tcp_sk(sk); int copied = 0; u32 peek_seq; + u32 peek_offset; u32 *seq; unsigned long used; int err; @@ -2435,7 +2436,6 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, long timeo; struct sk_buff *skb, *last; u32 urg_hole = 0; - err = -ENOTCONN; if (sk->sk_state == TCP_LISTEN) goto out; @@ -2469,6 +2469,14 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, if (flags & MSG_PEEK) { peek_seq = tp->copied_seq; seq = &peek_seq; + if (msg->msg_iter.iov[0].iov_base == NULL) { + peek_offset = msg->msg_iter.iov[0].iov_len; + msg->msg_iter.iov = &msg->msg_iter.iov[1]; + msg->msg_iter.nr_segs -= 1; + msg->msg_iter.count -= peek_offset; + len -= peek_offset; + *seq += peek_offset; + } } target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
On Thu, 22 Jun 2023 22:12:27 -0400 Jon Maloy <jmaloy(a)redhat.com> wrote:When reading received messages with MSG_PEEK, we sometines have to read the leading bytes of the stream several times, only to reach the bytes we really want. This is clearly non-optimal. What we would want is something similar to pread/preadv(), but working even for tcp sockets. At the same time, we obviously don't want to add any new arguments to the recv/recvmsg() calls. In this commit, we allow the user to set iovec.iov_base in the first vector entry to NULL. This tells the socket to skip the first entry, hence making the iov_len field of that entry indicate the offset value. This way, there is no need to add any new arguments.Ah-ha! I'm glad you found an acceptable way to pass a NULL pointer there. :)This change is simple and non-intrusive, and should be safe addition to the socket API. We have measured it to give a throughput improvement of...it would be nice to also do a bit of profiling with perf(1) -- that's where I originally noticed we were wasting cycles on filling up tcp_buf_discard. Plus, sure, there's also some value in dropping a useless 16 MiB buffer. If you need examples/inspiration: the pasta (automated) demo shows that, skip at 9:20 in: https://passt.top/passt/about/#pasta_2 (the one on the left) and that's simply done like this: https://passt.top/passt/tree/test/demo/pasta#n1638-10 % for the protocol splicer 'passst', which is used in KubeVirtpasstcontainers.Those are virtual machines in containers -- but other than KubeVirt, passt is generally supported in libvirt (>= 9.2) with QEMU (>= 7.2)Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> works with original msghdr Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> --- net/ipv4/tcp.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 33f559f491c8..1d89337e89b6 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2428,6 +2428,7 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, struct tcp_sock *tp = tcp_sk(sk); int copied = 0; u32 peek_seq; + u32 peek_offset;Kernel networking code observes the reverse-Christmas tree notation (at least for new additions), this would need to go after *tp: struct tcp_sock *tp = tcp_sk(sk); u32 peek_offset;u32 *seq; unsigned long used; int err; @@ -2435,7 +2436,6 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, long timeo; struct sk_buff *skb, *last; u32 urg_hole = 0; -Probably not intended.err = -ENOTCONN; if (sk->sk_state == TCP_LISTEN) goto out; @@ -2469,6 +2469,14 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, if (flags & MSG_PEEK) { peek_seq = tp->copied_seq; seq = &peek_seq; + if (msg->msg_iter.iov[0].iov_base == NULL) { + peek_offset = msg->msg_iter.iov[0].iov_len; + msg->msg_iter.iov = &msg->msg_iter.iov[1]; + msg->msg_iter.nr_segs -= 1;Do you also need to make sure that nr_segs > 1 if iov[0].iov_base is NULL? I'm not sure if we need to check explicitly that msg_iter.iov[1] is valid here (but I haven't followed the whole path from the syscall handler).+ msg->msg_iter.count -= peek_offset; + len -= peek_offset; + *seq += peek_offset; + } } target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);-- Stefano
On 2023-06-23 07:02, Stefano Brivio wrote:On Thu, 22 Jun 2023 22:12:27 -0400 Jon Maloy <jmaloy(a)redhat.com> wrote:This looks like a script language, but I don't recognize it. How do I run it?When reading received messages with MSG_PEEK, we sometines have to read the leading bytes of the stream several times, only to reach the bytes we really want. This is clearly non-optimal. What we would want is something similar to pread/preadv(), but working even for tcp sockets. At the same time, we obviously don't want to add any new arguments to the recv/recvmsg() calls. In this commit, we allow the user to set iovec.iov_base in the first vector entry to NULL. This tells the socket to skip the first entry, hence making the iov_len field of that entry indicate the offset value. This way, there is no need to add any new arguments.Ah-ha! I'm glad you found an acceptable way to pass a NULL pointer there. :)This change is simple and non-intrusive, and should be safe addition to the socket API. We have measured it to give a throughput improvement of...it would be nice to also do a bit of profiling with perf(1) -- that's where I originally noticed we were wasting cycles on filling up tcp_buf_discard. Plus, sure, there's also some value in dropping a useless 16 MiB buffer. If you need examples/inspiration: the pasta (automated) demo shows that, skip at 9:20 in: https://passt.top/passt/about/#pasta_2 (the one on the left) and that's simply done like this: https://passt.top/passt/tree/test/demo/pasta#n163> 8-10 % for the protocol splicer 'passst', which is used in KubeVirtNow at 20+ %passtChecked. This is taken care of by the generic code. ///joncontainers.Those are virtual machines in containers -- but other than KubeVirt, passt is generally supported in libvirt (>= 9.2) with QEMU (>= 7.2)Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> works with original msghdr Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> --- net/ipv4/tcp.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 33f559f491c8..1d89337e89b6 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2428,6 +2428,7 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, struct tcp_sock *tp = tcp_sk(sk); int copied = 0; u32 peek_seq; + u32 peek_offset;Kernel networking code observes the reverse-Christmas tree notation (at least for new additions), this would need to go after *tp: struct tcp_sock *tp = tcp_sk(sk); u32 peek_offset;u32 *seq; unsigned long used; int err; @@ -2435,7 +2436,6 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, long timeo; struct sk_buff *skb, *last; u32 urg_hole = 0; -Probably not intended.err = -ENOTCONN; if (sk->sk_state == TCP_LISTEN) goto out; @@ -2469,6 +2469,14 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len, if (flags & MSG_PEEK) { peek_seq = tp->copied_seq; seq = &peek_seq; + if (msg->msg_iter.iov[0].iov_base == NULL) { + peek_offset = msg->msg_iter.iov[0].iov_len; + msg->msg_iter.iov = &msg->msg_iter.iov[1]; + msg->msg_iter.nr_segs -= 1;Do you also need to make sure that nr_segs > 1 if iov[0].iov_base is NULL? I'm not sure if we need to check explicitly that msg_iter.iov[1] is valid here (but I haven't followed the whole path from the syscall handler).> + msg->msg_iter.count -= peek_offset; > + len -= peek_offset; > + *seq += peek_offset; > + } > } > > target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
On Mon, Sep 04, 2023 at 01:14:17PM -0400, Jon Maloy wrote:On 2023-06-23 07:02, Stefano Brivio wrote:It's the hand rolled script language of the passt tests. The interpreter is in test/lib/test. Easiest way to run it is probably to "make check" in test/ of the passt tree. If you want to selectively run certain things, edit test/run. -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibsonOn Thu, 22 Jun 2023 22:12:27 -0400 Jon Maloy <jmaloy(a)redhat.com> wrote:This looks like a script language, but I don't recognize it. How do I run it?When reading received messages with MSG_PEEK, we sometines have to read the leading bytes of the stream several times, only to reach the bytes we really want. This is clearly non-optimal. What we would want is something similar to pread/preadv(), but working even for tcp sockets. At the same time, we obviously don't want to add any new arguments to the recv/recvmsg() calls. In this commit, we allow the user to set iovec.iov_base in the first vector entry to NULL. This tells the socket to skip the first entry, hence making the iov_len field of that entry indicate the offset value. This way, there is no need to add any new arguments.Ah-ha! I'm glad you found an acceptable way to pass a NULL pointer there. :)This change is simple and non-intrusive, and should be safe addition to the socket API. We have measured it to give a throughput improvement of...it would be nice to also do a bit of profiling with perf(1) -- that's where I originally noticed we were wasting cycles on filling up tcp_buf_discard. Plus, sure, there's also some value in dropping a useless 16 MiB buffer. If you need examples/inspiration: the pasta (automated) demo shows that, skip at 9:20 in: https://passt.top/passt/about/#pasta_2 (the one on the left) and that's simply done like this: https://passt.top/passt/tree/test/demo/pasta#n163
On Tue, 5 Sep 2023 12:16:11 +1000 David Gibson <david(a)gibson.dropbear.id.au> wrote:On Mon, Sep 04, 2023 at 01:14:17PM -0400, Jon Maloy wrote:...some more details at test/README.md. -- StefanoOn 2023-06-23 07:02, Stefano Brivio wrote:It's the hand rolled script language of the passt tests. The interpreter is in test/lib/test. Easiest way to run it is probably to "make check" in test/ of the passt tree. If you want to selectively run certain things, edit test/run.On Thu, 22 Jun 2023 22:12:27 -0400 Jon Maloy <jmaloy(a)redhat.com> wrote:This looks like a script language, but I don't recognize it. How do I run it?When reading received messages with MSG_PEEK, we sometines have to read the leading bytes of the stream several times, only to reach the bytes we really want. This is clearly non-optimal. What we would want is something similar to pread/preadv(), but working even for tcp sockets. At the same time, we obviously don't want to add any new arguments to the recv/recvmsg() calls. In this commit, we allow the user to set iovec.iov_base in the first vector entry to NULL. This tells the socket to skip the first entry, hence making the iov_len field of that entry indicate the offset value. This way, there is no need to add any new arguments.Ah-ha! I'm glad you found an acceptable way to pass a NULL pointer there. :)This change is simple and non-intrusive, and should be safe addition to the socket API. We have measured it to give a throughput improvement of...it would be nice to also do a bit of profiling with perf(1) -- that's where I originally noticed we were wasting cycles on filling up tcp_buf_discard. Plus, sure, there's also some value in dropping a useless 16 MiB buffer. If you need examples/inspiration: the pasta (automated) demo shows that, skip at 9:20 in: https://passt.top/passt/about/#pasta_2 (the one on the left) and that's simply done like this: https://passt.top/passt/tree/test/demo/pasta#n163