- dev - passt

[PATCH] tcp_splice: A typo three years ago and SO_RCVLOWAT is gone
by Stefano Brivio 17 Feb '25

17 Feb '25

In commit e5eefe77435a ("tcp: Refactor to use events instead of states, split out spliced implementation"), this: if (!bitmap_isset(rcvlowat_set, conn - ts) && readlen > (long)c->tcp.pipe_size / 10) { (note the !) became: if (conn->flags & lowat_set_flag && readlen > (long)c->tcp.pipe_size / 10) { in the new tcp_splice_sock_handler(). We want to check, there, if we should set SO_RCVLOWAT, only if we haven't set it already. But, instead, we're checking if it's already set before we set it, so we'll never set it, of course. Fix the check and re-enable the functionality, which should give us improved CPU utilisation in non-interactive cases where we are not transferring at full pipe capacity. Fixes: e5eefe77435a ("tcp: Refactor to use events instead of states, split out spliced implementation") Signed-off-by: Stefano Brivio <sbrivio(a)redhat.com> --- tcp_splice.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcp_splice.c b/tcp_splice.c index 8a39a6f..5d845c9 100644 --- a/tcp_splice.c +++ b/tcp_splice.c @@ -556,7 +556,7 @@ eintr: if (readlen >= (long)c->tcp.pipe_size * 10 / 100) continue; - if (conn->flags & lowat_set_flag && + if (!(conn->flags & lowat_set_flag) && readlen > (long)c->tcp.pipe_size / 10) { int lowat = c->tcp.pipe_size / 4; -- 2.43.0

2 4

[PATCH v2 0/2] Reconstruct ICMP headers for failed UDP connect
by Jon Maloy 17 Feb '25

17 Feb '25

Reconstruct incoming ICMP headers for failed UDP connect and forward back to local peer. v2: - Added patch breaking out udp header creation from function tap_udp4_send(). - Updated the ICMP creation by using the new function. - Added logics to find correct flow, depending on origin. - All done after feedback from David Gibson. Jon Maloy (2): tap: break out building of udp header from tap_udp4_send function udp: create and send ICMPv4 to local peer when applicable tap.c | 36 +++++++++++++++------ tap.h | 7 ++++ udp.c | 86 ++++++++++++++++++++++++++++++++++++++++++-------- udp_internal.h | 2 +- udp_vu.c | 4 +-- 5 files changed, 110 insertions(+), 25 deletions(-) -- 2.48.1

2 4

[PATCH v3] repair, passt-repair: Build and warning fixes for musl
by Stefano Brivio 17 Feb '25

17 Feb '25

Checked against musl 1.2.5. Signed-off-by: Stefano Brivio <sbrivio(a)redhat.com> --- v3: Now 100% more trusted, with cppcheck absolutely positive about us initialising 'buf'. v2: Now 100% more tested: don't try to use the whole buffer as msg_controllen. passt-repair.c | 4 +++- repair.c | 13 +++++++++---- 2 files changed, 12 insertions(+), 5 deletions(-) diff --git a/passt-repair.c b/passt-repair.c index 1174ae3..e0c366e 100644 --- a/passt-repair.c +++ b/passt-repair.c @@ -63,6 +63,7 @@ int main(int argc, char **argv) struct cmsghdr *cmsg; struct msghdr msg; struct iovec iov; + size_t cmsg_len; int op; prctl(PR_SET_DUMPABLE, 0); @@ -138,8 +139,9 @@ loop: } } if (!n) { + cmsg_len = cmsg->cmsg_len; /* socklen_t is 'unsigned' on musl */ fprintf(stderr, "Invalid ancillary data length %zu from peer\n", - cmsg->cmsg_len); + cmsg_len); _exit(1); } diff --git a/repair.c b/repair.c index d288617..dac28a6 100644 --- a/repair.c +++ b/repair.c @@ -13,6 +13,7 @@ */ #include <errno.h> +#include <sys/socket.h> #include <sys/uio.h> #include "util.h" @@ -145,9 +146,9 @@ void repair_handler(struct ctx *c, uint32_t events) */ int repair_flush(struct ctx *c) { - struct iovec iov = { &repair_cmd, sizeof(repair_cmd) }; char buf[CMSG_SPACE(sizeof(int) * SCM_MAX_FD)] - __attribute__ ((aligned(__alignof__(struct cmsghdr)))); + __attribute__ ((aligned(__alignof__(struct cmsghdr)))) = { 0 }; + struct iovec iov = { &repair_cmd, sizeof(repair_cmd) }; struct cmsghdr *cmsg; struct msghdr msg; int8_t reply; @@ -155,8 +156,12 @@ int repair_flush(struct ctx *c) if (!repair_nfds) return 0; - msg = (struct msghdr){ NULL, 0, &iov, 1, - buf, CMSG_SPACE(sizeof(int) * repair_nfds), 0 }; + msg = (struct msghdr){ .msg_name = NULL, .msg_namelen = 0, + .msg_iov = &iov, .msg_iovlen = 1, + .msg_control = buf, + .msg_controllen = CMSG_SPACE(sizeof(int) * + repair_nfds), + .msg_flags = 0 }; cmsg = CMSG_FIRSTHDR(&msg); cmsg->cmsg_level = SOL_SOCKET; -- 2.43.0

2 1

[PATCH v27 0/2] State migration, perhaps final (#2)?
by Stefano Brivio 17 Feb '25

17 Feb '25

Compared to v25, test changes only: run all the tests we added. Compared to v26: actually commit all the tests. Stefano Brivio (2): migrate: Migrate TCP flows test: Add migration tests contrib/selinux/passt.te | 4 +- flow.c | 243 +++++++++ flow.h | 8 + migrate.c | 10 + passt.c | 6 +- repair.c | 1 - tcp.c | 919 +++++++++++++++++++++++++++++++++ tcp_conn.h | 103 ++++ test/lib/layout | 55 +- test/lib/setup | 138 ++++- test/lib/test | 48 ++ test/migrate/basic | 59 +++ test/migrate/basic_fin | 62 +++ test/migrate/bidirectional | 64 +++ test/migrate/bidirectional_fin | 64 +++ test/migrate/iperf3_bidir6 | 58 +++ test/migrate/iperf3_in4 | 50 ++ test/migrate/iperf3_in6 | 58 +++ test/migrate/iperf3_many_out6 | 60 +++ test/migrate/iperf3_out4 | 47 ++ test/migrate/iperf3_out6 | 58 +++ test/migrate/rampstream_in | 12 +- test/migrate/rampstream_out | 8 +- test/run | 42 +- 24 files changed, 2158 insertions(+), 19 deletions(-) create mode 100644 test/migrate/basic create mode 100644 test/migrate/basic_fin create mode 100644 test/migrate/bidirectional create mode 100644 test/migrate/bidirectional_fin create mode 100644 test/migrate/iperf3_bidir6 create mode 100644 test/migrate/iperf3_in4 create mode 100644 test/migrate/iperf3_in6 create mode 100644 test/migrate/iperf3_many_out6 create mode 100644 test/migrate/iperf3_out4 create mode 100644 test/migrate/iperf3_out6 -- 2.43.0

2 4

[PATCH v26 0/2] State migration, perhaps final?
by Stefano Brivio 16 Feb '25

16 Feb '25

Compared to v25, test changes only: run all the tests we added. Stefano Brivio (2): migrate: Migrate TCP flows test: Add migration tests contrib/selinux/passt.te | 4 +- flow.c | 243 ++++++++++ flow.h | 8 + migrate.c | 10 + passt.c | 6 +- repair.c | 1 - tcp.c | 919 ++++++++++++++++++++++++++++++++++++ tcp_conn.h | 103 ++++ test/lib/layout | 55 ++- test/lib/setup | 138 +++++- test/lib/test | 48 ++ test/migrate/basic | 59 +++ test/migrate/bidirectional | 64 +++ test/migrate/iperf3_bidir6 | 58 +++ test/migrate/iperf3_in4 | 50 ++ test/migrate/iperf3_in6 | 58 +++ test/migrate/iperf3_out4 | 47 ++ test/migrate/iperf3_out6 | 58 +++ test/migrate/rampstream_in | 12 +- test/migrate/rampstream_out | 8 +- test/run | 42 +- 21 files changed, 1972 insertions(+), 19 deletions(-) create mode 100644 test/migrate/basic create mode 100644 test/migrate/bidirectional create mode 100644 test/migrate/iperf3_bidir6 create mode 100644 test/migrate/iperf3_in4 create mode 100644 test/migrate/iperf3_in6 create mode 100644 test/migrate/iperf3_out4 create mode 100644 test/migrate/iperf3_out6 -- 2.43.0

1 2

[PATCH] tap: always set the no_frag flag in IPv4 headers.
by Jon Maloy 16 Feb '25

16 Feb '25

When studying the Linux source code and Wireshark dumps it seems like the no_frag flag in the IPv4 header is always set. Following discussions in the Internet on this subject indicates that modern routers never fragment packets, and that it isn't even supported in many cases. Adding to this that incoming messages forwarded on the tap interface never even pass through a router it seems safe to always set this flag. This makes the IPv4 headers of forwarded messages identical to those sent by the external sockets, something we must consider desirable. Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> --- tap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tap.c b/tap.c index d0673e5..44b0fc0 100644 --- a/tap.c +++ b/tap.c @@ -153,7 +153,7 @@ static void *tap_push_ip4h(struct iphdr *ip4h, struct in_addr src, ip4h->tos = 0; ip4h->tot_len = htons(l3len); ip4h->id = 0; - ip4h->frag_off = 0; + ip4h->frag_off = htons(IP_DF); ip4h->ttl = 255; ip4h->protocol = proto; ip4h->saddr = src.s_addr; -- 2.48.1

2 4

[PATCH v2] repair, passt-repair: Build and warning fixes for musl
by Stefano Brivio 15 Feb '25

15 Feb '25

Checked against musl 1.2.5. Signed-off-by: Stefano Brivio <sbrivio(a)redhat.com> --- v2: Now 100% more tested: don't try to use the whole buffer as msg_controllen. passt-repair.c | 4 +++- repair.c | 9 +++++++-- 2 files changed, 10 insertions(+), 3 deletions(-) diff --git a/passt-repair.c b/passt-repair.c index 1174ae3..e0c366e 100644 --- a/passt-repair.c +++ b/passt-repair.c @@ -63,6 +63,7 @@ int main(int argc, char **argv) struct cmsghdr *cmsg; struct msghdr msg; struct iovec iov; + size_t cmsg_len; int op; prctl(PR_SET_DUMPABLE, 0); @@ -138,8 +139,9 @@ loop: } } if (!n) { + cmsg_len = cmsg->cmsg_len; /* socklen_t is 'unsigned' on musl */ fprintf(stderr, "Invalid ancillary data length %zu from peer\n", - cmsg->cmsg_len); + cmsg_len); _exit(1); } diff --git a/repair.c b/repair.c index d288617..1d26fe6 100644 --- a/repair.c +++ b/repair.c @@ -13,6 +13,7 @@ */ #include <errno.h> +#include <sys/socket.h> #include <sys/uio.h> #include "util.h" @@ -155,8 +156,12 @@ int repair_flush(struct ctx *c) if (!repair_nfds) return 0; - msg = (struct msghdr){ NULL, 0, &iov, 1, - buf, CMSG_SPACE(sizeof(int) * repair_nfds), 0 }; + msg = (struct msghdr){ .msg_name = NULL, .msg_namelen = 0, + .msg_iov = &iov, .msg_iovlen = 1, + .msg_control = buf, + .msg_controllen = CMSG_SPACE(sizeof(int) * + repair_nfds), + .msg_flags = 0 }; cmsg = CMSG_FIRSTHDR(&msg); cmsg->cmsg_level = SOL_SOCKET; -- 2.43.0

1 0

[PATCH v25 0/2] State migration, somewhat polished
by Stefano Brivio 15 Feb '25

15 Feb '25

Notable changes compared to v24: - dropped thaw/freeze mechanism, as duplicate keep-alives gave me issues with migration of half-closed connections and in general they proved to be too rigid to quickly fix things as I needed to do. Kept all the fixes, though - fixed error handling on target, e.g. clearing flags, calling flow_cancel() as needed, etc. - added tests for half-closed sockets, fixed headers and descriptions - fix build with musl (1.25), and warnings on i686, armv6l, cppcheck and clang-tidy reports - drop "queue plugging" mechanism: it turns out that the kernel actually generates ACKs from the padding we send to our own receive queue, even in repair mode (!) All tests are happy. This looks quite solid and mergeable to me as it is. Stefano Brivio (2): migrate: Migrate TCP flows test: Add migration tests contrib/selinux/passt.te | 4 +- flow.c | 243 ++++++++++ flow.h | 8 + migrate.c | 10 + passt.c | 6 +- repair.c | 1 - tcp.c | 919 ++++++++++++++++++++++++++++++++++++ tcp_conn.h | 103 ++++ test/lib/layout | 55 ++- test/lib/setup | 138 +++++- test/lib/test | 48 ++ test/migrate/basic | 59 +++ test/migrate/bidirectional | 64 +++ test/migrate/iperf3_bidir6 | 58 +++ test/migrate/iperf3_in4 | 50 ++ test/migrate/iperf3_in6 | 58 +++ test/migrate/iperf3_out4 | 47 ++ test/migrate/iperf3_out6 | 58 +++ test/migrate/rampstream_in | 12 +- test/migrate/rampstream_out | 8 +- test/run | 34 +- 21 files changed, 1964 insertions(+), 19 deletions(-) create mode 100644 test/migrate/basic create mode 100644 test/migrate/bidirectional create mode 100644 test/migrate/iperf3_bidir6 create mode 100644 test/migrate/iperf3_in4 create mode 100644 test/migrate/iperf3_in6 create mode 100644 test/migrate/iperf3_out4 create mode 100644 test/migrate/iperf3_out6 -- 2.43.0

1 2

[PATCH] repair, passt-repair: Build and warning fixes for musl
by Stefano Brivio 15 Feb '25

15 Feb '25

Checked against musl 1.2.5. Signed-off-by: Stefano Brivio <sbrivio(a)redhat.com> --- passt-repair.c | 4 +++- repair.c | 8 ++++++-- 2 files changed, 9 insertions(+), 3 deletions(-) diff --git a/passt-repair.c b/passt-repair.c index 1174ae3..e0c366e 100644 --- a/passt-repair.c +++ b/passt-repair.c @@ -63,6 +63,7 @@ int main(int argc, char **argv) struct cmsghdr *cmsg; struct msghdr msg; struct iovec iov; + size_t cmsg_len; int op; prctl(PR_SET_DUMPABLE, 0); @@ -138,8 +139,9 @@ loop: } } if (!n) { + cmsg_len = cmsg->cmsg_len; /* socklen_t is 'unsigned' on musl */ fprintf(stderr, "Invalid ancillary data length %zu from peer\n", - cmsg->cmsg_len); + cmsg_len); _exit(1); } diff --git a/repair.c b/repair.c index d288617..64f18b7 100644 --- a/repair.c +++ b/repair.c @@ -13,6 +13,7 @@ */ #include <errno.h> +#include <sys/socket.h> #include <sys/uio.h> #include "util.h" @@ -155,8 +156,11 @@ int repair_flush(struct ctx *c) if (!repair_nfds) return 0; - msg = (struct msghdr){ NULL, 0, &iov, 1, - buf, CMSG_SPACE(sizeof(int) * repair_nfds), 0 }; + msg = (struct msghdr){ .msg_name = NULL, .msg_namelen = 0, + .msg_iov = &iov, .msg_iovlen = 1, + .msg_control = buf, + .msg_controllen = sizeof(buf), + .msg_flags = 0 }; cmsg = CMSG_FIRSTHDR(&msg); cmsg->cmsg_level = SOL_SOCKET; -- 2.43.0

1 0

[PATCH v24 0/5] State migration
by David Gibson 15 Feb '25

15 Feb '25

Rebased / integrated Stefano's patch for allowing us to carry on by sacrificing some flows in more cases. David Gibson (3): vhost_user: Clear ring address on GET_VRING_BASE flow: Flow table traversing macros flow, migrate: Flow migration skeleton Stefano Brivio (2): migrate: Migrate TCP flows test: Add migration tests contrib/selinux/passt.te | 4 +- flow.c | 335 ++++++++++++- flow.h | 13 + flow_table.h | 54 ++ migrate.c | 61 ++- migrate.h | 2 + passt.c | 6 +- repair.c | 1 - tcp.c | 950 ++++++++++++++++++++++++++++++++++++ tcp_conn.h | 99 ++++ test/lib/layout | 55 ++- test/lib/setup | 138 +++++- test/lib/test | 48 ++ test/migrate/basic | 59 +++ test/migrate/bidirectional | 64 +++ test/migrate/iperf3_bidir6 | 58 +++ test/migrate/iperf3_in4 | 50 ++ test/migrate/iperf3_in6 | 58 +++ test/migrate/iperf3_out4 | 50 ++ test/migrate/iperf3_out6 | 58 +++ test/migrate/rampstream_in | 10 +- test/migrate/rampstream_out | 6 +- test/run | 34 +- vhost_user.c | 1 + 24 files changed, 2176 insertions(+), 38 deletions(-) create mode 100644 test/migrate/basic create mode 100644 test/migrate/bidirectional create mode 100644 test/migrate/iperf3_bidir6 create mode 100644 test/migrate/iperf3_in4 create mode 100644 test/migrate/iperf3_in6 create mode 100644 test/migrate/iperf3_out4 create mode 100644 test/migrate/iperf3_out6 -- 2.48.1

2 6