April 2026 - dev

[PATCH v8 0/3] vhost-user,udp: Handle multiple iovec entries per virtqueue element
by Laurent Vivier 13 May '26

13 May '26

Some virtio-net drivers (notably iPXE) provide descriptors where the vnet header and the frame payload are in separate buffers, resulting in two iovec entries per virtqueue element. Currently, the RX (host to guest) path assumes a single iovec per element, which triggers: ASSERTION FAILED in virtqueue_map_desc (virtio.c:403): num_sg < max_num_sg This series reworks the UDP vhost-user receive path to support multiple iovec entries per element, fixing the iPXE crash. This series only addresses the UDP path. TCP vhost-user will be updated to use multi-iov elements in a subsequent series. Based-on: 20260416155721.3807225-1-lvivier(a)redhat.com v8: - Add Reviewed-by from David for 1/3 and 2/3 - Remove VLA to use fixed size (VIRTQUEUE_MAX_SIZE) - Rename udp_frame to datagram - Use sizeof(uh) instead of sizeof(struct udphdr) - Push back uh in memory in udp_vu_csum() v7: - In udp_vu_sock_to_tap(), introduce iov_still_needed variable - Fix iov_tail @data doc comments in udp_vu_prepare()/udp_vu_csum() - Don't rewrap the virtio checksum comment in udp_update_hdr6() - Add NULL check for uh in udp_vu_csum() v6: - Rebased on top of [PATCH 00/10] vhost-user: Preparatory series for multiple iovec entries per virtqueue element v5: - This version doesn't change the padding system regarding v4, it's a complex task that will be addressed in another version - reorder patches and add new patches - remove IOV_PUT_HEADER()/with_header() and introduce IOV_PUSH_HEADER() - don't use the iov_tail to provide the headers to the functions - move vu_set_vnethdr() to vu_flush(), extract vu_queue_notify() - move vu_flush() inside loop in tcp_vu_data_from_sock() to flush data by frame and not by full data length v4: - rebase - replace ASSERT() by assert() v3: - include the series "Decouple iovec management from virtqueues elements" - because of this series, drop: "vu_common: Accept explicit iovec counts in vu_set_element()" "vu_common: Accept explicit iovec count per element in vu_init_elem()" "vu_common: Prepare to use multibuffer with guest RX" "vhost-user,udp: Use 2 iovec entries per element" - drop "vu_common: Pass iov_tail to vu_set_vnethdr()" as the specs insures a buffer is big enough to contain vnet header - introduce "with_header()" and merge "udp: Pass iov_tail to udp_update_hdr4()/udp_update_hdr6()" and "udp_vu: Use iov_tail in udp_vu_prepare()" to use it v2: - add iov_truncate(), iov_memset() - remove iov_tail_truncate() and iov_tail_zero_end() - manage 802.3 minimum frame size Laurent Vivier (3): udp_vu: Allow virtqueue elements with multiple iovec entries iov: Introduce IOV_PUSH_HEADER() macro udp: Pass iov_tail to udp_update_hdr4()/udp_update_hdr6() iov.c | 22 ++++++++++ iov.h | 11 +++++ udp.c | 70 ++++++++++++++++------------- udp_internal.h | 4 +- udp_vu.c | 116 ++++++++++++++++++++++++++----------------------- 5 files changed, 135 insertions(+), 88 deletions(-) -- 2.53.0

3 8

[PATCH v3 00/10] vhost-user: Preparatory series for multiple iovec entries per virtqueue element
by Laurent Vivier 13 May '26

13 May '26

Currently, the vhost-user path assumes each virtqueue element contains exactly one iovec entry covering the entire frame. This assumption breaks as some virtio-net drivers (notably iPXE) provide descriptors where the vnet header and the frame payload are in separate buffers, resulting in two iovec entries per virtqueue element. This series refactors the vhost-user data path so that frame lengths, header sizes, and padding are tracked and passed explicitly rather than being derived from iovec sizes. This decoupling is a prerequisite for correctly handling padding of multi-buffer frames. The changes in this series can be split in 3 groups: - New iov helpers (patches 1-2): iov_memset() and iov_memcpy() operate across iovec boundaries. These are needed by the final patch to pad and copy frame data when a frame spans multiple iovec entries. - Structural refactoring (patches 3-5): Move vnethdr setup into vu_flush(), separate virtqueue management from socket I/O in the UDP path, and pass iov arrays explicitly instead of using file-scoped state. These changes make it possible to pass explicit frame lengths through the stack, which is required to pad frames independently of iovec layout. - Explicit length passing throughout the stack (patches 6-10): Thread explicit L4, L2, frame, and data lengths through checksum, pcap, vu_flush(), and tcp_fill_headers(), replacing lengths that were previously derived from iovec sizes. With lengths tracked explicitly, the final patch can centralise Ethernet frame padding into vu_collect() and a new vu_pad() helper that correctly pads frames spanning multiple iovec entries. v3: - csum_udp4()/csum_udp6()/udp_vu_csum receive payload length (dlen) rather than l4len - Add a length parameter to write_remainder() and use it in pcap_frame() v2: - Rename iov_memcopy() to iov_memcpy() and use clearer parameter names - Use clearer code in pcap_frame() - Add braces around bodies in pcap.c and tcp_vu.c for style consistency - Extract l2len variable in tap_add_packet() and tcp_vu_send_flag() to avoid repeating the same expression - Fix indentation alignment of iov_skip_bytes() arguments in tcp_vu_c - Introduce fill_size variable in vu_flush() - Reposition comment for ETH_ZLEN in vu_collect() Laurent Vivier (10): iov: Introduce iov_memset() iov: Add iov_memcpy() to copy data between iovec arrays vu_common: Move vnethdr setup into vu_flush() udp_vu: Move virtqueue management from udp_vu_sock_recv() to its caller udp_vu: Pass iov explicitly to helpers instead of using file-scoped array checksum: Pass explicit L4 length to checksum functions pcap: Pass explicit L2 length to pcap_iov() vu_common: Pass explicit frame length to vu_flush() tcp: Pass explicit data length to tcp_fill_headers() vhost-user: Centralise Ethernet frame padding in vu_collect() and vu_pad() checksum.c | 43 +++++++----- checksum.h | 6 +- iov.c | 78 ++++++++++++++++++++++ iov.h | 5 ++ pcap.c | 28 +++++--- pcap.h | 2 +- tap.c | 10 +-- tcp.c | 14 ++-- tcp_buf.c | 3 +- tcp_internal.h | 2 +- tcp_vu.c | 66 ++++++++++--------- udp.c | 5 +- udp_vu.c | 173 +++++++++++++++++++++++++------------------------ util.c | 31 +++++++-- util.h | 3 +- vu_common.c | 58 ++++++++++------- vu_common.h | 5 +- 17 files changed, 338 insertions(+), 194 deletions(-) -- 2.53.0

3 20

[PATCH v6 0/4] vhost-user,tcp: Handle multiple iovec entries per virtqueue element
by Laurent Vivier 11 May '26

11 May '26

This is the TCP counterpart to the UDP multi-iov series. It converts the TCP vhost-user receive path from direct pointer arithmetic (via vu_eth(), vu_ip(), etc.) to the iov_tail abstraction, removing the assumption that all headers reside in a single contiguous buffer. With this series applied, the TCP path correctly handles virtio-net drivers that provide multiple buffers per virtqueue element (e.g. iPXE provides the vnet header in the first buffer and the frame payload in a second one), matching the support already present in the UDP path. Based-on: 20260416160926.3822963-1-lvivier(a)redhat.com v6: - Rebase on v8 of UDP series (tcp_update_csum() takes dlen rather than l4len) v5: - Use l2len variable for pcap_iov() length in tcp_vu_send_flag() - Add braces - Move pcap_iov() before vu_flush() - Remove vu_flush() from tcp_vu_send_dup(), let the caller handle it v4: - fix error during rebase, s/vu_pad_len/vu_pad/ v3: - Rebased on top of [PATCH 00/10] vhost-user: Preparatory series for multiple iovec entries per virtqueue element v2: - add "tcp: Encode checksum computation flags in a single parameter" - remove IOV_PUT_HEADER()/with_header() and use IOV_PUSH_HEADER() - don't use the iov_tail to provide the headers to the functions Laurent Vivier (4): tcp: Encode checksum computation flags in a single parameter tcp_vu: Build headers on the stack and write them into the iovec tcp_vu: Support multibuffer frames in tcp_vu_sock_recv() tcp_vu: Support multibuffer frames in tcp_vu_send_flag() iov.c | 1 - tcp.c | 25 +-- tcp_buf.c | 23 +-- tcp_internal.h | 7 +- tcp_vu.c | 403 +++++++++++++++++++++++++++++-------------------- vu_common.h | 20 --- 6 files changed, 270 insertions(+), 209 deletions(-) -- 2.53.0

3 13

[PATCH v3] tcp: Use SO_MEMINFO for accurate send buffer overhead accounting
by Jon Maloy 07 May '26

07 May '26

The TCP window advertised to the guest/container must balance two competing needs: large enough to trigger kernel socket buffer auto-tuning, but not so large that sendmsg() partially fails, causing retransmissions. The current approach uses the difference (SNDBUF_GET() - SIOCOUTQ), but SNDBUF_GET() returns a scaled value that only roughly accounts for per-skb overhead. The clamped_scale approximation doesn't accurately track the actual per-segment overhead, which can lead to both excessive retransmissions and reduced throughput. We now introduce the use of SO_MEMINFO to obtain SK_MEMINFO_SNDBUF and SK_MEMINFO_WMEM_QUEUED from the kernel. The latter is presented in the kernel's own accounting units, i.e. including the sk_buff overhead, and matches exactly what the kernel's own sk_stream_memory_free() function is using. When data is queued and the overhead ratio is observable, we calculate the per-segment overhead as (wmem_queued - sendq) / num_segments, then determine how many additional segments should fit in the remaining buffer space, considering the calculated per-mss overhead. This approach treats segments as discrete quantities, and produces a more accurate estimate of available buffer space than a linear scaling factor does. When the ratio cannot be observed, e.g. because the queue is empty or we are in a transient state, we fall back to the existing clamped_scale calculation (scaling between 100% and 75% of buffer capacity). When SO_MEMINFO succeeds, we also use SK_MEMINFO_SNDBUF directly to set SNDBUF, avoiding a separate SO_SNDBUF getsockopt() call. If SO_MEMINFO is unavailable, we fall back to the pre-existing SNDBUF_GET() - SIOCOUTQ calculation. Link: https://bugs.passt.top/show_bug.cgi?id=138 Link: https://github.com/containers/podman/issues/28219 Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> --- v2: Updated according to feedback from Stefano. Segment-based discrete overhead calculation instead of linear ratio. v3: Addressed Stefano's v2 feedback: - Extracted window calculation into tcp_wnd_from_sndbuf() - Use wmem_queued instead of SIOCOUTQ for fallback and SWS check --- tcp.c | 137 ++++++++++++++++++++++++++++++++++------------------- tcp_conn.h | 2 +- 2 files changed, 89 insertions(+), 50 deletions(-) diff --git a/tcp.c b/tcp.c index 43b8fdb..61160cf 100644 --- a/tcp.c +++ b/tcp.c @@ -295,6 +295,7 @@ #include <arpa/inet.h> #include <linux/sockios.h> +#include <linux/sock_diag.h> #include "checksum.h" #include "util.h" @@ -1017,6 +1018,90 @@ size_t tcp_fill_headers(const struct ctx *c, struct tcp_tap_conn *conn, return MAX(l3len + sizeof(struct ethhdr), ETH_ZLEN); } +/** + * tcp_wnd_from_sndbuf() - Calculate window from available send buffer space + * @s: Socket file descriptor + * @conn: Connection pointer + * @tinfo: tcp_info from kernel + * + * Return: window value to advertise, not scaled + */ +static uint32_t tcp_wnd_from_sndbuf(int s, struct tcp_tap_conn *conn, + const struct tcp_info_linux *tinfo) +{ + uint32_t rtt_ms_ceiling = DIV_ROUND_UP(tinfo->tcpi_rtt, 1000); + uint32_t mem[SK_MEMINFO_VARS]; + socklen_t mem_sl = sizeof(mem); + int mss = MSS_GET(conn); + uint32_t limit, sendq; + + if (ioctl(s, SIOCOUTQ, &sendq)) { + debug_perror("SIOCOUTQ on socket %i, assuming 0", s); + sendq = 0; + } + + if (getsockopt(s, SOL_SOCKET, SO_MEMINFO, &mem, &mem_sl)) { + tcp_get_sndbuf(conn); + + if (sendq > SNDBUF_GET(conn)) /* Due to memory pressure? */ + limit = 0; + else + limit = SNDBUF_GET(conn) - sendq; + } else { + uint32_t sndbuf = mem[SK_MEMINFO_SNDBUF]; + uint32_t wmemq = mem[SK_MEMINFO_WMEM_QUEUED]; + uint32_t scaled = clamped_scale(sndbuf, sndbuf, SNDBUF_SMALL, + SNDBUF_BIG, 75); + + SNDBUF_SET(conn, MIN(INT_MAX, scaled)); + + if (wmemq > sndbuf) { + limit = 0; + } else if (!sendq || !mss || wmemq <= sendq) { + limit = SNDBUF_GET(conn) - wmemq; + } else { + uint32_t used_segs = MAX(sendq / mss, 1); + uint32_t overhead = (wmemq - sendq) / used_segs; + uint32_t remaining = sndbuf - wmemq; + uint32_t avail_segs = remaining / (mss + overhead); + + limit = avail_segs * mss; + } + } + + /* If the sender uses mechanisms to prevent Silly Window + * Syndrome (SWS, described in RFC 813 Section 3) it's critical + * that, should the window ever become less than the MSS, we + * advertise a new value once it increases again to be above it. + * + * The mechanism to avoid SWS in the kernel is, implicitly, + * implemented by Nagle's algorithm (which was proposed after + * RFC 813). + * + * To this end, for simplicity, approximate a window value below + * the MSS to zero, as we already have mechanisms in place to + * force updates after the window becomes zero. This matches the + * suggestion from RFC 813, Section 4. + * + * But don't do this if, either: + * + * - there's nothing in the outbound queue: the size of the + * sending buffer is limiting us, and it won't increase if we + * don't send data, so there's no point in waiting, or + * + * - we haven't sent data in a while (somewhat arbitrarily, ten + * times the RTT), as that might indicate that the receiver + * will only process data in batches that are large enough, + * but we won't send enough to fill one because we're stuck + * with pending data in the outbound queue + */ + if (limit < (uint32_t)MSS_GET(conn) && sendq && + tinfo->tcpi_last_data_sent < rtt_ms_ceiling * 10) + limit = 0; + + return MIN(tinfo->tcpi_snd_wnd, limit); +} + /** * tcp_update_seqack_wnd() - Update ACK sequence and window to guest/tap * @c: Execution context @@ -1124,56 +1209,10 @@ int tcp_update_seqack_wnd(const struct ctx *c, struct tcp_tap_conn *conn, } } - if ((conn->flags & LOCAL) || tcp_rtt_dst_low(conn)) { + if ((conn->flags & LOCAL) || tcp_rtt_dst_low(conn)) new_wnd_to_tap = tinfo->tcpi_snd_wnd; - } else { - unsigned rtt_ms_ceiling = DIV_ROUND_UP(tinfo->tcpi_rtt, 1000); - uint32_t sendq; - int limit; - - if (ioctl(s, SIOCOUTQ, &sendq)) { - debug_perror("SIOCOUTQ on socket %i, assuming 0", s); - sendq = 0; - } - tcp_get_sndbuf(conn); - - if ((int)sendq > SNDBUF_GET(conn)) /* Due to memory pressure? */ - limit = 0; - else - limit = SNDBUF_GET(conn) - (int)sendq; - - /* If the sender uses mechanisms to prevent Silly Window - * Syndrome (SWS, described in RFC 813 Section 3) it's critical - * that, should the window ever become less than the MSS, we - * advertise a new value once it increases again to be above it. - * - * The mechanism to avoid SWS in the kernel is, implicitly, - * implemented by Nagle's algorithm (which was proposed after - * RFC 813). - * - * To this end, for simplicity, approximate a window value below - * the MSS to zero, as we already have mechanisms in place to - * force updates after the window becomes zero. This matches the - * suggestion from RFC 813, Section 4. - * - * But don't do this if, either: - * - * - there's nothing in the outbound queue: the size of the - * sending buffer is limiting us, and it won't increase if we - * don't send data, so there's no point in waiting, or - * - * - we haven't sent data in a while (somewhat arbitrarily, ten - * times the RTT), as that might indicate that the receiver - * will only process data in batches that are large enough, - * but we won't send enough to fill one because we're stuck - * with pending data in the outbound queue - */ - if (limit < MSS_GET(conn) && sendq && - tinfo->tcpi_last_data_sent < rtt_ms_ceiling * 10) - limit = 0; - - new_wnd_to_tap = MIN((int)tinfo->tcpi_snd_wnd, limit); - } + else + new_wnd_to_tap = tcp_wnd_from_sndbuf(s, conn, tinfo); new_wnd_to_tap = MIN(new_wnd_to_tap, MAX_WINDOW); if (!(conn->events & ESTABLISHED)) diff --git a/tcp_conn.h b/tcp_conn.h index 6985426..9f5bee0 100644 --- a/tcp_conn.h +++ b/tcp_conn.h @@ -98,7 +98,7 @@ struct tcp_tap_conn { #define SNDBUF_BITS 24 unsigned int sndbuf :SNDBUF_BITS; #define SNDBUF_SET(conn, bytes) (conn->sndbuf = ((bytes) >> (32 - SNDBUF_BITS))) -#define SNDBUF_GET(conn) (conn->sndbuf << (32 - SNDBUF_BITS)) +#define SNDBUF_GET(conn) ((uint32_t)(conn->sndbuf << (32 - SNDBUF_BITS))) uint8_t seq_dup_ack_approx; -- 2.52.0

2 1

[PATCH v2 00/13] Improvements to static checker invocation
by David Gibson 06 May '26

06 May '26

While working on pesto, I ran into a number of awkward errors with the static checkers. This series reworks the invocation of the checkers in a way that will let us deal with that. As a bonus, it also gives us static checking for passt-repair and qrap. It also makes a few other cleanups to the Makefile that seemed natural along the way. v2: - Fixed nasty test failure in test/build/build.py David Gibson (13): Makefile: Use make variables for static checker configuration cppcheck: Split out essential defines into a BASE_CPPFLAGS variable Makefile: Remove preprocessor flags from $(FLAGS) Makefile: Remove non-standard $(FLAGS) variable Makefile: Make conditional definition of $(BIN) clearer Makefile: Use common binary compilation rule Makefile: Remove unhelpful $(HEADERS) variable Makefile: Add header dependencies for secondary binaries Makefile: Split static checker targets passt-repair: Split out inotify handling to its own function passt-repair: Simplify construction of Unix path from inotify passt-repair: Run static checkers qrap: Run static checkers Makefile | 106 ++++++++++++++++----------- linux_dep.h | 2 +- passt-repair.c | 171 ++++++++++++++++++++++++-------------------- qrap.c | 42 ++++++----- test/build/build.py | 4 +- 5 files changed, 190 insertions(+), 135 deletions(-) -- 2.53.0

3 30

[PATCH v5 01/18] conf, fwd: Stricter rule checking in fwd_rule_add()
by David Gibson 04 May '26

04 May '26

Although fwd_rule_add() performs some sanity checks on the rule it is given, there are invalid rules we don't check for, assuming that its callers will do that. That won't be enough when we can get rules inserted by a dynamic update client without going through the existing parsing code. So, add stricter checks to fwd_rule_add(), which is now possible thanks to the capabilities bits in the struct fwd_table. Where those duplicate existing checks in the callers, remove the old copies. Signed-off-by: David Gibson <david(a)gibson.dropbear.id.au> --- conf.c | 21 --------------------- fwd.c | 52 +++++++++++++++++++++++++++++++++++++++++++++++----- 2 files changed, 47 insertions(+), 26 deletions(-) diff --git a/conf.c b/conf.c index 6e884e54..b470b0d8 100644 --- a/conf.c +++ b/conf.c @@ -176,8 +176,6 @@ static void conf_ports_range_except(struct fwd_table *fwd, uint8_t proto, die("Invalid interface name: %s", ifname); } - assert(first != 0); - for (base = first; base <= last; base++) { if (exclude && bitmap_isset(exclude, base)) continue; @@ -310,10 +308,6 @@ static void conf_ports_spec(struct fwd_table *fwd, uint8_t proto, if (p != ep) /* Garbage after the ranges */ goto bad; - if (orig_range.first == 0) { - die("Can't forward port 0 included in '%s'", spec); - } - conf_ports_range_except(fwd, proto, addr, ifname, orig_range.first, orig_range.last, exclude, @@ -356,11 +350,6 @@ static void conf_ports(char optname, const char *optarg, struct fwd_table *fwd) return; } - if (proto == IPPROTO_TCP && !(fwd->caps & FWD_CAP_TCP)) - die("TCP port forwarding requested but TCP is disabled"); - if (proto == IPPROTO_UDP && !(fwd->caps & FWD_CAP_UDP)) - die("UDP port forwarding requested but UDP is disabled"); - strncpy(buf, optarg, sizeof(buf) - 1); if ((spec = strchr(buf, '/'))) { @@ -405,16 +394,6 @@ static void conf_ports(char optname, const char *optarg, struct fwd_table *fwd) addr = NULL; } - if (addr) { - if (!(fwd->caps & FWD_CAP_IPV4) && inany_v4(addr)) { - die("IPv4 is disabled, can't use -%c %s", - optname, optarg); - } else if (!(fwd->caps & FWD_CAP_IPV6) && !inany_v4(addr)) { - die("IPv6 is disabled, can't use -%c %s", - optname, optarg); - } - } - if (optname == 'T' || optname == 'U') { assert(!addr && !ifname); diff --git a/fwd.c b/fwd.c index c7fd1a9d..979c1494 100644 --- a/fwd.c +++ b/fwd.c @@ -367,17 +367,59 @@ int fwd_rule_add(struct fwd_table *fwd, const struct fwd_rule *new) new->first, new->last); return -EINVAL; } + if (!new->first) { + warn("Forwarding rule attempts to map from port 0"); + return -EINVAL; + } + if (!new->to || + (in_port_t)(new->to + new->last - new->first) < new->to) { + warn("Forwarding rule attempts to map to port 0"); + return -EINVAL; + } if (new->flags & ~allowed_flags) { warn("Rule has invalid flags 0x%hhx", new->flags & ~allowed_flags); return -EINVAL; } - if (new->flags & FWD_DUAL_STACK_ANY && - !inany_equals(&new->addr, &inany_any6)) { - char astr[INANY_ADDRSTRLEN]; + if (new->flags & FWD_DUAL_STACK_ANY) { + if (!inany_equals(&new->addr, &inany_any6)) { + char astr[INANY_ADDRSTRLEN]; - warn("Dual stack rule has non-wildcard address %s", - inany_ntop(&new->addr, astr, sizeof(astr))); + warn("Dual stack rule has non-wildcard address %s", + inany_ntop(&new->addr, astr, sizeof(astr))); + return -EINVAL; + } + if (!(fwd->caps & FWD_CAP_IPV4)) { + warn("Dual stack forward, but IPv4 not enabled"); + return -EINVAL; + } + if (!(fwd->caps & FWD_CAP_IPV6)) { + warn("Dual stack forward, but IPv6 not enabled"); + return -EINVAL; + } + } else { + if (inany_v4(&new->addr) && !(fwd->caps & FWD_CAP_IPV4)) { + warn("IPv4 forward, but IPv4 not enabled"); + return -EINVAL; + } + if (!inany_v4(&new->addr) && !(fwd->caps & FWD_CAP_IPV6)) { + warn("IPv6 forward, but IPv6 not enabled"); + return -EINVAL; + } + } + if (new->proto == IPPROTO_TCP) { + if (!(fwd->caps & FWD_CAP_TCP)) { + warn("Can't add TCP forwarding rule, TCP not enabled"); + return -EINVAL; + } + } else if (new->proto == IPPROTO_UDP) { + if (!(fwd->caps & FWD_CAP_UDP)) { + warn("Can't add UDP forwarding rule, UDP not enabled"); + return -EINVAL; + } + } else { + warn("Unsupported protocol 0x%hhx (%s) for forwarding rule", + new->proto, ipproto_name(new->proto)); return -EINVAL; } -- 2.53.0

3 4

[PATCH v5 00/18] RFC: Dynamic configuration update implementation
by David Gibson 03 May '26

03 May '26

Here's the next draft of dynamic configuration updates. This now can successfully update rules, though I've not tested it very extensively. Patches 1..8/18 are preliminary reworks that make sense even without pesto - feel free to apply if you're happy with them. I don't think the rest should be applied yet; we need to at least harden it so passt can't be blocked indefinitely by a client which sends a partial update then waits. Based on my earlier series reworking static checking invocation. TODO: - Don't allow a client which sends a partial configuration then blocks also block passt - Allow pesto to clear existing configuration, not just add - Allow pesto selectively delete existing rules, not just add Changes in v5: * If multiple clients connect at once, they're now blocked until the first one finishes, instead of later ones being discarded Changes in v4: * Merged with remainder of forward rule parsing rework series * Fix some bugs in rule checking pointed out by Laurent * Significantly cleaned up option parsing code * Changed from replacing all existing rules to adding new rules (clear and remove still TBD) * Somewhat simplified protocol (pif names and rules sent in a single pass) * pesto is now allocation free * Fixed commit message and style nits pointed out by Stefano Changes in v3: * Removed already applied ASSERT() rename * Renamed serialisation functions * Incorporated Stefano's extensions, reworked and fixed * Several additional cleanups / preliminary reworks Changes in v2: * Removed already applied cleanups * Reworked assert() patch to handle -DNDEBUG properly * Numerous extra patches: * Factored out serialisation helpers and use them for migration as well * Reworked to allow ip.[ch] and inany.[ch] to be shared with pesto * Reworks to share some forwarding rule datatypes with pesto * Implemented sending pif names and current ruleset to pesto David Gibson (18): conf, fwd: Stricter rule checking in fwd_rule_add() fwd_rule: Move ephemeral port probing to fwd_rule.c fwd, conf: Move rule parsing code to fwd_rule.[ch] fwd_rule: Move conflict checking back within fwd_rule_add() fwd: Generalise fwd_rules_info() pif: Limit pif names to 128 bytes fwd_rule: Fix some format specifiers tap, repair: Use SOCK_NONBLOCK and SOCK_CLOEXEC on Unix sockets pesto: Introduce stub configuration tool pesto, log: Share log.h (but not log.c) with pesto tool pesto, conf: Have pesto connect to passt and check versions pesto: Expose list of pifs to pesto and optionally display ip: Prepare ip.[ch] for sharing with pesto tool inany: Prepare inany.[ch] for sharing with pesto tool pesto: Read current ruleset from passt/pasta and optionally display it pesto: Parse and add new rules from command line pesto, conf: Send updated rules from pesto back to passt/pasta conf, fwd: Allow switching to new rules received from pesto .gitignore | 2 + Makefile | 54 ++-- common.h | 122 +++++++++ conf.c | 686 ++++++++++++++++++++++----------------------------- conf.h | 2 + epoll_type.h | 4 + flow.c | 4 +- fwd.c | 169 ++++--------- fwd.h | 41 +-- fwd_rule.c | 603 ++++++++++++++++++++++++++++++++++++++++++-- fwd_rule.h | 66 ++++- inany.c | 19 +- inany.h | 17 +- ip.c | 56 +---- ip.h | 4 +- lineread.c | 2 +- log.h | 59 ++++- passt.1 | 5 + passt.c | 8 + passt.h | 8 + pesto.1 | 46 ++++ pesto.c | 470 +++++++++++++++++++++++++++++++++++ pesto.h | 55 +++++ pif.c | 2 +- pif.h | 8 +- repair.c | 9 +- serialise.c | 7 + serialise.h | 1 + siphash.h | 13 + tap.c | 64 ++++- util.c | 2 +- util.h | 110 +-------- 32 files changed, 1921 insertions(+), 797 deletions(-) create mode 100644 common.h create mode 100644 pesto.1 create mode 100644 pesto.c create mode 100644 pesto.h -- 2.53.0

3 28

[PATCH v5 03/18] fwd, conf: Move rule parsing code to fwd_rule.[ch]
by David Gibson 03 May '26

03 May '26

The code parsing command line options into forwarding rules has now been decoupled from most of passt/pasta's internals. This is good, because we'll soon want to share it with a configuration update client. Make the next step by moving this code into fwd_rule.[ch]. Signed-off-by: David Gibson <david(a)gibson.dropbear.id.au> Reviewd-by: Laurent Vivier <lvivier(a)redhat.com> --- conf.c | 376 +------------------------------------------ fwd.c | 94 ----------- fwd.h | 33 ---- fwd_rule.c | 464 ++++++++++++++++++++++++++++++++++++++++++++++++++++- fwd_rule.h | 36 ++++- 5 files changed, 502 insertions(+), 501 deletions(-) diff --git a/conf.c b/conf.c index b470b0d8..5aacfe0f 100644 --- a/conf.c +++ b/conf.c @@ -13,7 +13,6 @@ */ #include <arpa/inet.h> -#include <ctype.h> #include <errno.h> #include <fcntl.h> #include <getopt.h> @@ -66,365 +65,6 @@ const char *pasta_default_ifn = "tap0"; -/** - * port_range() - Represents a non-empty range of ports - * @first: First port number in the range - * @last: Last port number in the range (inclusive) - * - * Invariant: @last >= @first - */ -struct port_range { - in_port_t first, last; -}; - -/** - * parse_port_range() - Parse a range of port numbers '<first>[-<last>]' - * @s: String to parse - * @endptr: Update to the character after the parsed range (similar to - * strtol() etc.) - * @range: Update with the parsed values on success - * - * Return: -EINVAL on parsing error, -ERANGE on out of range port - * numbers, 0 on success - */ -static int parse_port_range(const char *s, const char **endptr, - struct port_range *range) -{ - unsigned long first, last; - char *ep; - - last = first = strtoul(s, &ep, 10); - if (ep == s) /* Parsed nothing */ - return -EINVAL; - if (*ep == '-') { /* we have a last value too */ - const char *lasts = ep + 1; - last = strtoul(lasts, &ep, 10); - if (ep == lasts) /* Parsed nothing */ - return -EINVAL; - } - - if ((last < first) || (last >= NUM_PORTS)) - return -ERANGE; - - range->first = first; - range->last = last; - *endptr = ep; - - return 0; -} - -/** - * parse_keyword() - Parse a literal keyword - * @s: String to parse - * @endptr: Update to the character after the keyword - * @kw: Keyword to accept - * - * Return: 0, if @s starts with @kw, -EINVAL if it does not - */ -static int parse_keyword(const char *s, const char **endptr, const char *kw) -{ - size_t len = strlen(kw); - - if (strlen(s) < len) - return -EINVAL; - - if (memcmp(s, kw, len)) - return -EINVAL; - - *endptr = s + len; - return 0; -} - -/** - * conf_ports_range_except() - Set up forwarding for a range of ports minus a - * bitmap of exclusions - * @fwd: Forwarding table to be updated - * @proto: Protocol to forward - * @addr: Listening address - * @ifname: Listening interface - * @first: First port to forward - * @last: Last port to forward - * @exclude: Bitmap of ports to exclude (may be NULL) - * @to: Port to translate @first to when forwarding - * @flags: Flags for forwarding entries - */ -static void conf_ports_range_except(struct fwd_table *fwd, uint8_t proto, - const union inany_addr *addr, - const char *ifname, - uint16_t first, uint16_t last, - const uint8_t *exclude, uint16_t to, - uint8_t flags) -{ - struct fwd_rule rule = { - .addr = addr ? *addr : inany_any6, - .ifname = { 0 }, - .proto = proto, - .flags = flags, - }; - char rulestr[FWD_RULE_STRLEN]; - unsigned delta = to - first; - unsigned base, i; - - if (!addr) - rule.flags |= FWD_DUAL_STACK_ANY; - if (ifname) { - int ret; - - ret = snprintf(rule.ifname, sizeof(rule.ifname), - "%s", ifname); - if (ret <= 0 || (size_t)ret >= sizeof(rule.ifname)) - die("Invalid interface name: %s", ifname); - } - - for (base = first; base <= last; base++) { - if (exclude && bitmap_isset(exclude, base)) - continue; - - for (i = base; i <= last; i++) { - if (exclude && bitmap_isset(exclude, i)) - break; - } - - rule.first = base; - rule.last = i - 1; - rule.to = base + delta; - - fwd_rule_conflict_check(&rule, fwd->rules, fwd->count); - if (fwd_rule_add(fwd, &rule) < 0) - goto fail; - - base = i - 1; - } - return; - -fail: - die("Unable to add rule %s", - fwd_rule_fmt(&rule, rulestr, sizeof(rulestr))); -} - -/* - * for_each_chunk - Step through delimited chunks of a string - * @p_: Pointer to start of each chunk (updated) - * @ep_: Pointer to end of each chunk (updated) - * @s_: String to step through - * @sep_: String of all allowed delimiters - */ -#define for_each_chunk(p_, ep_, s_, sep_) \ - for ((p_) = (s_); \ - (ep_) = (p_) + strcspn((p_), (sep_)), *(p_); \ - (p_) = *(ep_) ? (ep_) + 1 : (ep_)) - -/** - * conf_ports_spec() - Parse port range(s) specifier - * @fwd: Forwarding table to be updated - * @proto: Protocol to forward - * @addr: Listening address for forwarding - * @ifname: Interface name for listening - * @spec: Port range(s) specifier - */ -static void conf_ports_spec(struct fwd_table *fwd, uint8_t proto, - const union inany_addr *addr, const char *ifname, - const char *spec) -{ - uint8_t exclude[PORT_BITMAP_SIZE] = { 0 }; - bool exclude_only = true; - const char *p, *ep; - uint8_t flags = 0; - unsigned i; - - if (!strcmp(spec, "all")) { - /* Treat "all" as equivalent to "": all non-ephemeral ports */ - spec = ""; - } - - /* Parse excluded ranges and "auto" in the first pass */ - for_each_chunk(p, ep, spec, ",") { - struct port_range xrange; - - if (isdigit(*p)) { - /* Include range, parse later */ - exclude_only = false; - continue; - } - - if (parse_keyword(p, &p, "auto") == 0) { - if (p != ep) /* Garbage after the keyword */ - goto bad; - - if (!(fwd->caps & FWD_CAP_SCAN)) { - die( -"'auto' port forwarding is only allowed for pasta"); - } - - flags |= FWD_SCAN; - continue; - } - - /* Should be an exclude range */ - if (*p != '~') - goto bad; - p++; - - if (parse_port_range(p, &p, &xrange)) - goto bad; - if (p != ep) /* Garbage after the range */ - goto bad; - - for (i = xrange.first; i <= xrange.last; i++) - bitmap_set(exclude, i); - } - - if (exclude_only) { - /* Exclude ephemeral ports */ - fwd_port_map_ephemeral(exclude); - - conf_ports_range_except(fwd, proto, addr, ifname, - 1, NUM_PORTS - 1, exclude, - 1, flags | FWD_WEAK); - return; - } - - /* Now process base ranges, skipping exclusions */ - for_each_chunk(p, ep, spec, ",") { - struct port_range orig_range, mapped_range; - - if (!isdigit(*p)) - /* Already parsed */ - continue; - - if (parse_port_range(p, &p, &orig_range)) - goto bad; - - if (*p == ':') { /* There's a range to map to as well */ - if (parse_port_range(p + 1, &p, &mapped_range)) - goto bad; - if ((mapped_range.last - mapped_range.first) != - (orig_range.last - orig_range.first)) - goto bad; - } else { - mapped_range = orig_range; - } - - if (p != ep) /* Garbage after the ranges */ - goto bad; - - conf_ports_range_except(fwd, proto, addr, ifname, - orig_range.first, orig_range.last, - exclude, - mapped_range.first, flags); - } - - return; -bad: - die("Invalid port specifier '%s'", spec); -} - -/** - * conf_ports() - Parse port configuration options, initialise UDP/TCP sockets - * @optname: Short option name, t, T, u, or U - * @optarg: Option argument (port specification) - * @fwd: Forwarding table to be updated - */ -static void conf_ports(char optname, const char *optarg, struct fwd_table *fwd) -{ - union inany_addr addr_buf = inany_any6, *addr = &addr_buf; - char buf[BUFSIZ], *spec, *ifname = NULL; - uint8_t proto; - - if (optname == 't' || optname == 'T') - proto = IPPROTO_TCP; - else if (optname == 'u' || optname == 'U') - proto = IPPROTO_UDP; - else - assert(0); - - if (!strcmp(optarg, "none")) { - unsigned i; - - for (i = 0; i < fwd->count; i++) { - if (fwd->rules[i].proto == proto) { - die("-%c none conflicts with previous options", - optname); - } - } - return; - } - - strncpy(buf, optarg, sizeof(buf) - 1); - - if ((spec = strchr(buf, '/'))) { - *spec = 0; - spec++; - - if (optname != 't' && optname != 'u') - die("Listening address not allowed for -%c %s", - optname, optarg); - - if ((ifname = strchr(buf, '%'))) { - *ifname = 0; - ifname++; - - /* spec is already advanced one past the '/', - * so the length of the given ifname is: - * (spec - ifname - 1) - */ - if (spec - ifname - 1 >= IFNAMSIZ) { - die("Interface name '%s' is too long (max %u)", - ifname, IFNAMSIZ - 1); - } - } - - if (ifname == buf + 1) { /* Interface without address */ - addr = NULL; - } else { - char *p = buf; - - /* Allow square brackets for IPv4 too for convenience */ - if (*p == '[' && p[strlen(p) - 1] == ']') { - p[strlen(p) - 1] = '\0'; - p++; - } - - if (!inany_pton(p, addr)) - die("Bad forwarding address '%s'", p); - } - } else { - spec = buf; - - addr = NULL; - } - - if (optname == 'T' || optname == 'U') { - assert(!addr && !ifname); - - if (!(fwd->caps & FWD_CAP_IFNAME)) { - warn( -"SO_BINDTODEVICE unavailable, forwarding only 127.0.0.1 and ::1 for '-%c %s'", - optname, optarg); - - if (fwd->caps & FWD_CAP_IPV4) { - conf_ports_spec(fwd, proto, - &inany_loopback4, NULL, spec); - } - if (fwd->caps & FWD_CAP_IPV6) { - conf_ports_spec(fwd, proto, - &inany_loopback6, NULL, spec); - } - return; - } - - ifname = "lo"; - } - - if (ifname && !(fwd->caps & FWD_CAP_IFNAME)) { - die( -"Device binding for '-%c %s' unsupported (requires kernel 5.7+)", - optname, optarg); - } - - conf_ports_spec(fwd, proto, addr, ifname, spec); -} - /** * add_dns4() - Possibly add the IPv4 address of a DNS resolver to configuration * @c: Execution context @@ -2160,16 +1800,16 @@ void conf(struct ctx *c, int argc, char **argv) if (name == 't') { opt_t = true; - conf_ports(name, optarg, c->fwd[PIF_HOST]); + fwd_rule_parse(name, optarg, c->fwd[PIF_HOST]); } else if (name == 'u') { opt_u = true; - conf_ports(name, optarg, c->fwd[PIF_HOST]); + fwd_rule_parse(name, optarg, c->fwd[PIF_HOST]); } else if (name == 'T') { opt_T = true; - conf_ports(name, optarg, c->fwd[PIF_SPLICE]); + fwd_rule_parse(name, optarg, c->fwd[PIF_SPLICE]); } else if (name == 'U') { opt_U = true; - conf_ports(name, optarg, c->fwd[PIF_SPLICE]); + fwd_rule_parse(name, optarg, c->fwd[PIF_SPLICE]); } } while (name != -1); @@ -2221,13 +1861,13 @@ void conf(struct ctx *c, int argc, char **argv) if (c->mode == MODE_PASTA) { if (!opt_t) - conf_ports('t', "auto", c->fwd[PIF_HOST]); + fwd_rule_parse('t', "auto", c->fwd[PIF_HOST]); if (!opt_T) - conf_ports('T', "auto", c->fwd[PIF_SPLICE]); + fwd_rule_parse('T', "auto", c->fwd[PIF_SPLICE]); if (!opt_u) - conf_ports('u', "auto", c->fwd[PIF_HOST]); + fwd_rule_parse('u', "auto", c->fwd[PIF_HOST]); if (!opt_U) - conf_ports('U', "auto", c->fwd[PIF_SPLICE]); + fwd_rule_parse('U', "auto", c->fwd[PIF_SPLICE]); } if (!c->quiet) diff --git a/fwd.c b/fwd.c index a6d75b74..728a783c 100644 --- a/fwd.c +++ b/fwd.c @@ -275,100 +275,6 @@ void fwd_rule_init(struct ctx *c) c->fwd[PIF_SPLICE] = &fwd_out; } -/** - * fwd_rule_add() - Validate and add a rule to a forwarding table - * @fwd: Table to add to - * @new: Rule to add - * - * Return: 0 on success, negative error code on failure - */ -int fwd_rule_add(struct fwd_table *fwd, const struct fwd_rule *new) -{ - /* Flags which can be set from the caller */ - const uint8_t allowed_flags = FWD_WEAK | FWD_SCAN | FWD_DUAL_STACK_ANY; - unsigned num = (unsigned)new->last - new->first + 1; - unsigned port; - - if (new->first > new->last) { - warn("Rule has invalid port range %u-%u", - new->first, new->last); - return -EINVAL; - } - if (!new->first) { - warn("Forwarding rule attempts to map from port 0"); - return -EINVAL; - } - if (!new->to || - (in_port_t)(new->to + new->last - new->first) < new->to) { - warn("Forwarding rule attempts to map to port 0"); - return -EINVAL; - } - if (new->flags & ~allowed_flags) { - warn("Rule has invalid flags 0x%hhx", - new->flags & ~allowed_flags); - return -EINVAL; - } - if (new->flags & FWD_DUAL_STACK_ANY) { - if (!inany_equals(&new->addr, &inany_any6)) { - char astr[INANY_ADDRSTRLEN]; - - warn("Dual stack rule has non-wildcard address %s", - inany_ntop(&new->addr, astr, sizeof(astr))); - return -EINVAL; - } - if (!(fwd->caps & FWD_CAP_IPV4)) { - warn("Dual stack forward, but IPv4 not enabled"); - return -EINVAL; - } - if (!(fwd->caps & FWD_CAP_IPV6)) { - warn("Dual stack forward, but IPv6 not enabled"); - return -EINVAL; - } - } else { - if (inany_v4(&new->addr) && !(fwd->caps & FWD_CAP_IPV4)) { - warn("IPv4 forward, but IPv4 not enabled"); - return -EINVAL; - } - if (!inany_v4(&new->addr) && !(fwd->caps & FWD_CAP_IPV6)) { - warn("IPv6 forward, but IPv6 not enabled"); - return -EINVAL; - } - } - if (new->proto == IPPROTO_TCP) { - if (!(fwd->caps & FWD_CAP_TCP)) { - warn("Can't add TCP forwarding rule, TCP not enabled"); - return -EINVAL; - } - } else if (new->proto == IPPROTO_UDP) { - if (!(fwd->caps & FWD_CAP_UDP)) { - warn("Can't add UDP forwarding rule, UDP not enabled"); - return -EINVAL; - } - } else { - warn("Unsupported protocol 0x%hhx (%s) for forwarding rule", - new->proto, ipproto_name(new->proto)); - return -EINVAL; - } - - if (fwd->count >= ARRAY_SIZE(fwd->rules)) { - warn("Too many rules (maximum %u)", ARRAY_SIZE(fwd->rules)); - return -ENOSPC; - } - if ((fwd->sock_count + num) > ARRAY_SIZE(fwd->socks)) { - warn("Rules require too many listening sockets (maximum %u)", - ARRAY_SIZE(fwd->socks)); - return -ENOSPC; - } - - fwd->rulesocks[fwd->count] = &fwd->socks[fwd->sock_count]; - for (port = new->first; port <= new->last; port++) - fwd->rulesocks[fwd->count][port - new->first] = -1; - - fwd->rules[fwd->count++] = *new; - fwd->sock_count += num; - return 0; -} - /** * fwd_rule_match() - Does a prospective flow match a given forwarding rule? * @rule: Forwarding rule diff --git a/fwd.h b/fwd.h index e664d1d0..8f845d09 100644 --- a/fwd.h +++ b/fwd.h @@ -20,8 +20,6 @@ struct flowside; -#define FWD_RULE_BITS 8 -#define MAX_FWD_RULES MAX_FROM_BITS(FWD_RULE_BITS) #define FWD_NO_HINT (-1) /** @@ -36,36 +34,6 @@ struct fwd_listen_ref { unsigned rule :FWD_RULE_BITS; }; -/* Maximum number of listening sockets (per pif) - * - * Rationale: This lets us listen on every port for two addresses and two - * protocols (which we need for -T auto -U auto without SO_BINDTODEVICE), plus a - * comfortable number of extras. - */ -#define MAX_LISTEN_SOCKS (NUM_PORTS * 5) - -/** - * struct fwd_table - Forwarding state (per initiating pif) - * @caps: Forwarding capabilities for this initiating pif - * @count: Number of forwarding rules - * @rules: Array of forwarding rules - * @rulesocks: Parallel array of @rules (@count valid entries) of pointers to - * @socks entries giving the start of the corresponding rule's - * sockets within the larger array - * @sock_count: Number of entries used in @socks (for all rules combined) - * @socks: Listening sockets for forwarding - */ -struct fwd_table { - uint32_t caps; - unsigned count; - struct fwd_rule rules[MAX_FWD_RULES]; - int *rulesocks[MAX_FWD_RULES]; - unsigned sock_count; - int socks[MAX_LISTEN_SOCKS]; -}; - -#define PORT_BITMAP_SIZE DIV_ROUND_UP(NUM_PORTS, 8) - /** * struct fwd_scan - Port scanning state for a protocol+direction * @scan4: /proc/net fd to scan for IPv4 ports when in AUTO mode @@ -81,7 +49,6 @@ struct fwd_scan { #define FWD_PORT_SCAN_INTERVAL 1000 /* ms */ void fwd_rule_init(struct ctx *c); -int fwd_rule_add(struct fwd_table *fwd, const struct fwd_rule *new); const struct fwd_rule *fwd_rule_search(const struct fwd_table *fwd, const struct flowside *ini, uint8_t proto, int hint); diff --git a/fwd_rule.c b/fwd_rule.c index 9d489827..cd3dec04 100644 --- a/fwd_rule.c +++ b/fwd_rule.c @@ -15,6 +15,7 @@ * Author: David Gibson <david(a)gibson.dropbear.id.au> */ +#include <ctype.h> #include <errno.h> #include <fcntl.h> #include <stdio.h> @@ -89,7 +90,7 @@ parse_err: * fwd_port_map_ephemeral() - Mark ephemeral ports in a bitmap * @map: Bitmap to update */ -void fwd_port_map_ephemeral(uint8_t *map) +static void fwd_port_map_ephemeral(uint8_t *map) { unsigned port; @@ -123,6 +124,7 @@ const union inany_addr *fwd_rule_addr(const struct fwd_rule *rule) */ __attribute__((noinline)) #endif +/* cppcheck-suppress staticFunction */ const char *fwd_rule_fmt(const struct fwd_rule *rule, char *dst, size_t size) { const char *percent = *rule->ifname ? "%" : ""; @@ -199,8 +201,8 @@ static bool fwd_rule_conflicts(const struct fwd_rule *a, const struct fwd_rule * * @rules: Existing rules against which to test * @count: Number of rules in @rules */ -void fwd_rule_conflict_check(const struct fwd_rule *new, - const struct fwd_rule *rules, size_t count) +static void fwd_rule_conflict_check(const struct fwd_rule *new, + const struct fwd_rule *rules, size_t count) { unsigned i; @@ -215,3 +217,459 @@ void fwd_rule_conflict_check(const struct fwd_rule *new, fwd_rule_fmt(&rules[i], rulestr, sizeof(rulestr))); } } + +/** + * fwd_rule_add() - Validate and add a rule to a forwarding table + * @fwd: Table to add to + * @new: Rule to add + * + * Return: 0 on success, negative error code on failure + */ +static int fwd_rule_add(struct fwd_table *fwd, const struct fwd_rule *new) +{ + /* Flags which can be set from the caller */ + const uint8_t allowed_flags = FWD_WEAK | FWD_SCAN | FWD_DUAL_STACK_ANY; + unsigned num = (unsigned)new->last - new->first + 1; + unsigned port; + + if (new->first > new->last) { + warn("Rule has invalid port range %u-%u", + new->first, new->last); + return -EINVAL; + } + if (!new->first) { + warn("Forwarding rule attempts to map from port 0"); + return -EINVAL; + } + if (!new->to || + (in_port_t)(new->to + new->last - new->first) < new->to) { + warn("Forwarding rule attempts to map to port 0"); + return -EINVAL; + } + if (new->flags & ~allowed_flags) { + warn("Rule has invalid flags 0x%hhx", + new->flags & ~allowed_flags); + return -EINVAL; + } + if (new->flags & FWD_DUAL_STACK_ANY) { + if (!inany_equals(&new->addr, &inany_any6)) { + char astr[INANY_ADDRSTRLEN]; + + warn("Dual stack rule has non-wildcard address %s", + inany_ntop(&new->addr, astr, sizeof(astr))); + return -EINVAL; + } + if (!(fwd->caps & FWD_CAP_IPV4)) { + warn("Dual stack forward, but IPv4 not enabled"); + return -EINVAL; + } + if (!(fwd->caps & FWD_CAP_IPV6)) { + warn("Dual stack forward, but IPv6 not enabled"); + return -EINVAL; + } + } else { + if (inany_v4(&new->addr) && !(fwd->caps & FWD_CAP_IPV4)) { + warn("IPv4 forward, but IPv4 not enabled"); + return -EINVAL; + } + if (!inany_v4(&new->addr) && !(fwd->caps & FWD_CAP_IPV6)) { + warn("IPv6 forward, but IPv6 not enabled"); + return -EINVAL; + } + } + if (new->proto == IPPROTO_TCP) { + if (!(fwd->caps & FWD_CAP_TCP)) { + warn("Can't add TCP forwarding rule, TCP not enabled"); + return -EINVAL; + } + } else if (new->proto == IPPROTO_UDP) { + if (!(fwd->caps & FWD_CAP_UDP)) { + warn("Can't add UDP forwarding rule, UDP not enabled"); + return -EINVAL; + } + } else { + warn("Unsupported protocol 0x%hhx (%s) for forwarding rule", + new->proto, ipproto_name(new->proto)); + return -EINVAL; + } + + if (fwd->count >= ARRAY_SIZE(fwd->rules)) { + warn("Too many rules (maximum %u)", ARRAY_SIZE(fwd->rules)); + return -ENOSPC; + } + if ((fwd->sock_count + num) > ARRAY_SIZE(fwd->socks)) { + warn("Rules require too many listening sockets (maximum %u)", + ARRAY_SIZE(fwd->socks)); + return -ENOSPC; + } + + fwd->rulesocks[fwd->count] = &fwd->socks[fwd->sock_count]; + for (port = new->first; port <= new->last; port++) + fwd->rulesocks[fwd->count][port - new->first] = -1; + + fwd->rules[fwd->count++] = *new; + fwd->sock_count += num; + return 0; +} + +/** + * port_range() - Represents a non-empty range of ports + * @first: First port number in the range + * @last: Last port number in the range (inclusive) + * + * Invariant: @last >= @first + */ +struct port_range { + in_port_t first, last; +}; + +/** + * parse_port_range() - Parse a range of port numbers '<first>[-<last>]' + * @s: String to parse + * @endptr: Update to the character after the parsed range (similar to + * strtol() etc.) + * @range: Update with the parsed values on success + * + * Return: -EINVAL on parsing error, -ERANGE on out of range port + * numbers, 0 on success + */ +static int parse_port_range(const char *s, const char **endptr, + struct port_range *range) +{ + unsigned long first, last; + char *ep; + + last = first = strtoul(s, &ep, 10); + if (ep == s) /* Parsed nothing */ + return -EINVAL; + if (*ep == '-') { /* we have a last value too */ + const char *lasts = ep + 1; + last = strtoul(lasts, &ep, 10); + if (ep == lasts) /* Parsed nothing */ + return -EINVAL; + } + + if ((last < first) || (last >= NUM_PORTS)) + return -ERANGE; + + range->first = first; + range->last = last; + *endptr = ep; + + return 0; +} + +/** + * parse_keyword() - Parse a literal keyword + * @s: String to parse + * @endptr: Update to the character after the keyword + * @kw: Keyword to accept + * + * Return: 0, if @s starts with @kw, -EINVAL if it does not + */ +static int parse_keyword(const char *s, const char **endptr, const char *kw) +{ + size_t len = strlen(kw); + + if (strlen(s) < len) + return -EINVAL; + + if (memcmp(s, kw, len)) + return -EINVAL; + + *endptr = s + len; + return 0; +} + +/** + * fwd_rule_range_except() - Set up forwarding for a range of ports minus a + * bitmap of exclusions + * @fwd: Forwarding table to be updated + * @proto: Protocol to forward + * @addr: Listening address + * @ifname: Listening interface + * @first: First port to forward + * @last: Last port to forward + * @exclude: Bitmap of ports to exclude (may be NULL) + * @to: Port to translate @first to when forwarding + * @flags: Flags for forwarding entries + */ +static void fwd_rule_range_except(struct fwd_table *fwd, uint8_t proto, + const union inany_addr *addr, + const char *ifname, + uint16_t first, uint16_t last, + const uint8_t *exclude, uint16_t to, + uint8_t flags) +{ + struct fwd_rule rule = { + .addr = addr ? *addr : inany_any6, + .ifname = { 0 }, + .proto = proto, + .flags = flags, + }; + char rulestr[FWD_RULE_STRLEN]; + unsigned delta = to - first; + unsigned base, i; + + if (!addr) + rule.flags |= FWD_DUAL_STACK_ANY; + if (ifname) { + int ret; + + ret = snprintf(rule.ifname, sizeof(rule.ifname), + "%s", ifname); + if (ret <= 0 || (size_t)ret >= sizeof(rule.ifname)) + die("Invalid interface name: %s", ifname); + } + + for (base = first; base <= last; base++) { + if (exclude && bitmap_isset(exclude, base)) + continue; + + for (i = base; i <= last; i++) { + if (exclude && bitmap_isset(exclude, i)) + break; + } + + rule.first = base; + rule.last = i - 1; + rule.to = base + delta; + + fwd_rule_conflict_check(&rule, fwd->rules, fwd->count); + if (fwd_rule_add(fwd, &rule) < 0) + goto fail; + + base = i - 1; + } + return; + +fail: + die("Unable to add rule %s", + fwd_rule_fmt(&rule, rulestr, sizeof(rulestr))); +} + +/* + * for_each_chunk - Step through delimited chunks of a string + * @p_: Pointer to start of each chunk (updated) + * @ep_: Pointer to end of each chunk (updated) + * @s_: String to step through + * @sep_: String of all allowed delimiters + */ +#define for_each_chunk(p_, ep_, s_, sep_) \ + for ((p_) = (s_); \ + (ep_) = (p_) + strcspn((p_), (sep_)), *(p_); \ + (p_) = *(ep_) ? (ep_) + 1 : (ep_)) + +/** + * fwd_rule_parse_ports() - Parse port range(s) specifier + * @fwd: Forwarding table to be updated + * @proto: Protocol to forward + * @addr: Listening address for forwarding + * @ifname: Interface name for listening + * @spec: Port range(s) specifier + */ +static void fwd_rule_parse_ports(struct fwd_table *fwd, uint8_t proto, + const union inany_addr *addr, + const char *ifname, + const char *spec) +{ + uint8_t exclude[PORT_BITMAP_SIZE] = { 0 }; + bool exclude_only = true; + const char *p, *ep; + uint8_t flags = 0; + unsigned i; + + if (!strcmp(spec, "all")) { + /* Treat "all" as equivalent to "": all non-ephemeral ports */ + spec = ""; + } + + /* Parse excluded ranges and "auto" in the first pass */ + for_each_chunk(p, ep, spec, ",") { + struct port_range xrange; + + if (isdigit(*p)) { + /* Include range, parse later */ + exclude_only = false; + continue; + } + + if (parse_keyword(p, &p, "auto") == 0) { + if (p != ep) /* Garbage after the keyword */ + goto bad; + + if (!(fwd->caps & FWD_CAP_SCAN)) { + die( +"'auto' port forwarding is only allowed for pasta"); + } + + flags |= FWD_SCAN; + continue; + } + + /* Should be an exclude range */ + if (*p != '~') + goto bad; + p++; + + if (parse_port_range(p, &p, &xrange)) + goto bad; + if (p != ep) /* Garbage after the range */ + goto bad; + + for (i = xrange.first; i <= xrange.last; i++) + bitmap_set(exclude, i); + } + + if (exclude_only) { + /* Exclude ephemeral ports */ + fwd_port_map_ephemeral(exclude); + + fwd_rule_range_except(fwd, proto, addr, ifname, + 1, NUM_PORTS - 1, exclude, + 1, flags | FWD_WEAK); + return; + } + + /* Now process base ranges, skipping exclusions */ + for_each_chunk(p, ep, spec, ",") { + struct port_range orig_range, mapped_range; + + if (!isdigit(*p)) + /* Already parsed */ + continue; + + if (parse_port_range(p, &p, &orig_range)) + goto bad; + + if (*p == ':') { /* There's a range to map to as well */ + if (parse_port_range(p + 1, &p, &mapped_range)) + goto bad; + if ((mapped_range.last - mapped_range.first) != + (orig_range.last - orig_range.first)) + goto bad; + } else { + mapped_range = orig_range; + } + + if (p != ep) /* Garbage after the ranges */ + goto bad; + + fwd_rule_range_except(fwd, proto, addr, ifname, + orig_range.first, orig_range.last, + exclude, + mapped_range.first, flags); + } + + return; +bad: + die("Invalid port specifier '%s'", spec); +} + +/** + * fwd_rule_parse() - Parse port configuration option + * @optname: Short option name, t, T, u, or U + * @optarg: Option argument (port specification) + * @fwd: Forwarding table to be updated + */ +void fwd_rule_parse(char optname, const char *optarg, struct fwd_table *fwd) +{ + union inany_addr addr_buf = inany_any6, *addr = &addr_buf; + char buf[BUFSIZ], *spec, *ifname = NULL; + uint8_t proto; + + if (optname == 't' || optname == 'T') + proto = IPPROTO_TCP; + else if (optname == 'u' || optname == 'U') + proto = IPPROTO_UDP; + else + assert(0); + + if (!strcmp(optarg, "none")) { + unsigned i; + + for (i = 0; i < fwd->count; i++) { + if (fwd->rules[i].proto == proto) { + die("-%c none conflicts with previous options", + optname); + } + } + return; + } + + strncpy(buf, optarg, sizeof(buf) - 1); + + if ((spec = strchr(buf, '/'))) { + *spec = 0; + spec++; + + if (optname != 't' && optname != 'u') + die("Listening address not allowed for -%c %s", + optname, optarg); + + if ((ifname = strchr(buf, '%'))) { + *ifname = 0; + ifname++; + + /* spec is already advanced one past the '/', + * so the length of the given ifname is: + * (spec - ifname - 1) + */ + if (spec - ifname - 1 >= IFNAMSIZ) { + die("Interface name '%s' is too long (max %u)", + ifname, IFNAMSIZ - 1); + } + } + + if (ifname == buf + 1) { /* Interface without address */ + addr = NULL; + } else { + char *p = buf; + + /* Allow square brackets for IPv4 too for convenience */ + if (*p == '[' && p[strlen(p) - 1] == ']') { + p[strlen(p) - 1] = '\0'; + p++; + } + + if (!inany_pton(p, addr)) + die("Bad forwarding address '%s'", p); + } + } else { + spec = buf; + + addr = NULL; + } + + if (optname == 'T' || optname == 'U') { + assert(!addr && !ifname); + + if (!(fwd->caps & FWD_CAP_IFNAME)) { + warn( +"SO_BINDTODEVICE unavailable, forwarding only 127.0.0.1 and ::1 for '-%c %s'", + optname, optarg); + + if (fwd->caps & FWD_CAP_IPV4) { + fwd_rule_parse_ports(fwd, proto, + &inany_loopback4, NULL, + spec); + } + if (fwd->caps & FWD_CAP_IPV6) { + fwd_rule_parse_ports(fwd, proto, + &inany_loopback6, NULL, + spec); + } + return; + } + + ifname = "lo"; + } + + if (ifname && !(fwd->caps & FWD_CAP_IFNAME)) { + die( +"Device binding for '-%c %s' unsupported (requires kernel 5.7+)", + optname, optarg); + } + + fwd_rule_parse_ports(fwd, proto, addr, ifname, spec); +} diff --git a/fwd_rule.h b/fwd_rule.h index 5c7b67aa..f0f4efda 100644 --- a/fwd_rule.h +++ b/fwd_rule.h @@ -19,6 +19,7 @@ /* Number of ports for both TCP and UDP */ #define NUM_PORTS (1U << 16) +#define PORT_BITMAP_SIZE DIV_ROUND_UP(NUM_PORTS, 8) /* Forwarding capability bits */ #define FWD_CAP_IPV4 BIT(0) @@ -54,8 +55,38 @@ struct fwd_rule { uint8_t flags; }; +#define FWD_RULE_BITS 8 +#define MAX_FWD_RULES MAX_FROM_BITS(FWD_RULE_BITS) + +/* Maximum number of listening sockets (per pif) + * + * Rationale: This lets us listen on every port for two addresses and two + * protocols (which we need for -T auto -U auto without SO_BINDTODEVICE), plus a + * comfortable number of extras. + */ +#define MAX_LISTEN_SOCKS (NUM_PORTS * 5) + +/** + * struct fwd_table - Forwarding state (per initiating pif) + * @caps: Forwarding capabilities for this initiating pif + * @count: Number of forwarding rules + * @rules: Array of forwarding rules + * @rulesocks: Parallel array of @rules (@count valid entries) of pointers to + * @socks entries giving the start of the corresponding rule's + * sockets within the larger array + * @sock_count: Number of entries used in @socks (for all rules combined) + * @socks: Listening sockets for forwarding + */ +struct fwd_table { + uint32_t caps; + unsigned count; + struct fwd_rule rules[MAX_FWD_RULES]; + int *rulesocks[MAX_FWD_RULES]; + unsigned sock_count; + int socks[MAX_LISTEN_SOCKS]; +}; + void fwd_probe_ephemeral(void); -void fwd_port_map_ephemeral(uint8_t *map); #define FWD_RULE_STRLEN \ (IPPROTO_STRLEN - 1 \ @@ -67,7 +98,6 @@ void fwd_port_map_ephemeral(uint8_t *map); const union inany_addr *fwd_rule_addr(const struct fwd_rule *rule); const char *fwd_rule_fmt(const struct fwd_rule *rule, char *dst, size_t size); void fwd_rules_info(const struct fwd_rule *rules, size_t count); -void fwd_rule_conflict_check(const struct fwd_rule *new, - const struct fwd_rule *rules, size_t count); +void fwd_rule_parse(char optname, const char *optarg, struct fwd_table *fwd); #endif /* FWD_RULE_H */ -- 2.53.0

3 3

[PATCH v4 00/17] RFC: Dynamic configuration update implementation
by David Gibson 25 Apr '26

25 Apr '26

Here's the next draft of dynamic configuration updates. This now can successfully update rules, though I've not tested it very extensively. Patches 1..7/17 are preliminary reworks that make sense even without pesto - feel free to apply if you're happy with them. I don't think the rest should be applied yet; we need to at least harden it so passt can't be blocked indefinitely by a client which sends a partial update then waits. Based on my earlier series reworking static checking invocation. TODO: - Have pestos connecting to passt which already have a client attached block, instead of being rejected - Don't allow a client which sends a partial configuration then blocks also block passt - Allow pesto to clear existing configuration, not just add - Allow pesto selectively delete existing rules, not just add Changes in v4: * Merged with remainder of forward rule parsing rework series * Fix some bugs in rule checking pointed out by Laurent * Significantly cleaned up option parsing code * Changed from replacing all existing rules to adding new rules (clear and remove still TBD) * Somewhat simplified protocol (pif names and rules sent in a single pass) * pesto is now allocation free * Fixed commit message and style nits pointed out by Stefano Changes in v3: * Removed already applied ASSERT() rename * Renamed serialisation functions * Incorporated Stefano's extensions, reworked and fixed * Several additional cleanups / preliminary reworks Changes in v2: * Removed already applied cleanups * Reworked assert() patch to handle -DNDEBUG properly * Numerous extra patches: * Factored out serialisation helpers and use them for migration as well * Reworked to allow ip.[ch] and inany.[ch] to be shared with pesto * Reworks to share some forwarding rule datatypes with pesto * Implemented sending pif names and current ruleset to pesto David Gibson (17): conf, fwd: Stricter rule checking in fwd_rule_add() fwd_rule: Move ephemeral port probing to fwd_rule.c fwd, conf: Move rule parsing code to fwd_rule.[ch] fwd_rule: Move conflict checking back within fwd_rule_add() fwd: Generalise fwd_rules_info() pif: Limit pif names to 128 bytes fwd_rule: Fix some format specifiers pesto: Introduce stub configuration tool pesto, log: Share log.h (but not log.c) with pesto tool pesto, conf: Have pesto connect to passt and check versions pesto: Expose list of pifs to pesto and optionally display ip: Prepare ip.[ch] for sharing with pesto tool inany: Prepare inany.[ch] for sharing with pesto tool pesto: Read current ruleset from passt/pasta and optionally display it pesto: Parse and add new rules from command line pesto, conf: Send updated rules from pesto back to passt/pasta conf, fwd: Allow switching to new rules received from pesto .gitignore | 2 + Makefile | 54 +++-- common.h | 122 ++++++++++ conf.c | 672 +++++++++++++++++++++------------------------------ conf.h | 2 + epoll_type.h | 4 + flow.c | 4 +- fwd.c | 169 ++++--------- fwd.h | 41 +--- fwd_rule.c | 603 ++++++++++++++++++++++++++++++++++++++++++--- fwd_rule.h | 66 ++++- inany.c | 19 +- inany.h | 17 +- ip.c | 56 +---- ip.h | 4 +- lineread.c | 2 +- log.h | 59 ++++- passt.1 | 5 + passt.c | 8 + passt.h | 8 + pesto.1 | 46 ++++ pesto.c | 470 +++++++++++++++++++++++++++++++++++ pesto.h | 55 +++++ pif.c | 2 +- pif.h | 8 +- serialise.c | 7 + serialise.h | 1 + siphash.h | 13 + tap.c | 52 ++++ util.h | 110 +-------- 30 files changed, 1889 insertions(+), 792 deletions(-) create mode 100644 common.h create mode 100644 pesto.1 create mode 100644 pesto.c create mode 100644 pesto.h -- 2.53.0

2 18

[PATCH v2] tcp: Use SO_MEMINFO for accurate send buffer overhead accounting
by Jon Maloy 25 Apr '26

25 Apr '26

The TCP window advertised to the guest/container must balance two competing needs: large enough to trigger kernel socket buffer auto-tuning, but not so large that sendmsg() partially fails, causing retransmissions. The current approach uses the difference (SNDBUF_GET() - SIOCOUTQ), but SNDBUF_GET() returns a scaled value that only roughly accounts for per-skb overhead. The clamped_scale approximation doesn't accurately track the actual per-segment overhead, which can lead to both excessive retransmissions and reduced throughput. We now introduce the use of SO_MEMINFO to obtain SK_MEMINFO_SNDBUF and SK_MEMINFO_WMEM_QUEUED from the kernel. The latter is presented in the kernel's own accounting units, i.e. including the sk_buff overhead, and matches exactly what the kernel's own sk_stream_memory_free() function is using. When data is queued and the overhead ratio is observable, we calculate the per-segment overhead as (wmem_queued - sendq) / num_segments, then determine how many additional segments should fit in the remaining buffer space, considering the calculated per-mss overhead. This approach treats segments as discrete quantities, and produces a more accurate estimate of available buffer space than a linear scaling factor does. When the ratio cannot be observed, e.g. because the queue is empty or we are in a transient state, we fall back to the existing clamped_scale calculation (scaling between 100% and 75% of buffer capacity). When SO_MEMINFO succeeds, we also use SK_MEMINFO_SNDBUF directly to set SNDBUF, avoiding a separate SO_SNDBUF getsockopt() call. If SO_MEMINFO is unavailable, we fall back to the pre-existing SNDBUF_GET() - SIOCOUTQ calculation. Link: https://bugs.passt.top/show_bug.cgi?id=138 Signed-off-by: Jon Maloy <jmaloy(a)redhat.com> --- v2: Updated according to feedback from Stefano. My own measurements indicate that this approach largely solves both the retransmission and throughput issues observed with the previous version. --- tcp.c | 42 ++++++++++++++++++++++++++++++++++-------- tcp_conn.h | 2 +- 2 files changed, 35 insertions(+), 9 deletions(-) diff --git a/tcp.c b/tcp.c index 43b8fdb..2ba08fd 100644 --- a/tcp.c +++ b/tcp.c @@ -295,6 +295,7 @@ #include <arpa/inet.h> #include <linux/sockios.h> +#include <linux/sock_diag.h> #include "checksum.h" #include "util.h" @@ -1128,19 +1129,44 @@ int tcp_update_seqack_wnd(const struct ctx *c, struct tcp_tap_conn *conn, new_wnd_to_tap = tinfo->tcpi_snd_wnd; } else { unsigned rtt_ms_ceiling = DIV_ROUND_UP(tinfo->tcpi_rtt, 1000); + uint32_t mem[SK_MEMINFO_VARS]; + socklen_t mem_sl = sizeof(mem); + int mss = MSS_GET(conn); uint32_t sendq; - int limit; + uint32_t limit; if (ioctl(s, SIOCOUTQ, &sendq)) { debug_perror("SIOCOUTQ on socket %i, assuming 0", s); sendq = 0; } - tcp_get_sndbuf(conn); - if ((int)sendq > SNDBUF_GET(conn)) /* Due to memory pressure? */ - limit = 0; - else - limit = SNDBUF_GET(conn) - (int)sendq; + if (getsockopt(s, SOL_SOCKET, SO_MEMINFO, &mem, &mem_sl)) { + tcp_get_sndbuf(conn); + if (sendq > SNDBUF_GET(conn)) + limit = 0; + else + limit = SNDBUF_GET(conn) - sendq; + } else { + uint32_t sb = mem[SK_MEMINFO_SNDBUF]; + uint32_t wq = mem[SK_MEMINFO_WMEM_QUEUED]; + uint32_t cs = clamped_scale(sb, sb, SNDBUF_SMALL, + SNDBUF_BIG, 75); + + SNDBUF_SET(conn, MIN(INT_MAX, cs)); + + if (wq > sb) { + limit = 0; + } else if (!sendq || wq <= sendq || !mss) { + limit = SNDBUF_GET(conn) - sendq; + } else { + uint32_t nsegs = MAX(sendq / mss, 1); + uint32_t overhead = (wq - sendq) / nsegs; + uint32_t remaining = sb - wq; + + nsegs = remaining / (mss + overhead); + limit = nsegs * mss; + } + } /* If the sender uses mechanisms to prevent Silly Window * Syndrome (SWS, described in RFC 813 Section 3) it's critical @@ -1168,11 +1194,11 @@ int tcp_update_seqack_wnd(const struct ctx *c, struct tcp_tap_conn *conn, * but we won't send enough to fill one because we're stuck * with pending data in the outbound queue */ - if (limit < MSS_GET(conn) && sendq && + if (limit < (uint32_t)MSS_GET(conn) && sendq && tinfo->tcpi_last_data_sent < rtt_ms_ceiling * 10) limit = 0; - new_wnd_to_tap = MIN((int)tinfo->tcpi_snd_wnd, limit); + new_wnd_to_tap = MIN(tinfo->tcpi_snd_wnd, limit); } new_wnd_to_tap = MIN(new_wnd_to_tap, MAX_WINDOW); diff --git a/tcp_conn.h b/tcp_conn.h index 6985426..9f5bee0 100644 --- a/tcp_conn.h +++ b/tcp_conn.h @@ -98,7 +98,7 @@ struct tcp_tap_conn { #define SNDBUF_BITS 24 unsigned int sndbuf :SNDBUF_BITS; #define SNDBUF_SET(conn, bytes) (conn->sndbuf = ((bytes) >> (32 - SNDBUF_BITS))) -#define SNDBUF_GET(conn) (conn->sndbuf << (32 - SNDBUF_BITS)) +#define SNDBUF_GET(conn) ((uint32_t)(conn->sndbuf << (32 - SNDBUF_BITS))) uint8_t seq_dup_ack_approx; -- 2.52.0

2 1