- dev - passt

[PATCH v5 1/5] Show debug message whenever we learn a new guest MAC address
by Volker Diels-Grabsch 13 Sep '25

13 Sep '25

Signed-off-by: Volker Diels-Grabsch <v(a)njh.eu> --- tap.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/tap.c b/tap.c index 7ba6399..3c08126 100644 --- a/tap.c +++ b/tap.c @@ -1096,7 +1096,11 @@ void tap_add_packet(struct ctx *c, struct iov_tail *data, return; if (memcmp(c->guest_mac, eh->h_source, ETH_ALEN)) { + char bufmac[ETH_ADDRSTRLEN]; + memcpy(c->guest_mac, eh->h_source, ETH_ALEN); + debug("Guest MAC address: %s", + eth_ntop(c->guest_mac, bufmac, sizeof(bufmac))); proto_update_l2_buf(c->guest_mac, NULL); } -- 2.47.3

3 11

[PATCH v2] tap: Drop frames if no client connected
by Yumei Huang 12 Sep '25

12 Sep '25

If no client is attached, discard outgoing frames and report them as sent. Without this we will get an EBADF in either writev() (pasta) or sendmsg() (passt). That's basically harmless, but a bit ugly. Explicitly catching this case results in behaviour that's probably a bit clearer to debug if we hit it. There are several different approaches we can take here. Here's some reasoning as David explained: * Don't listen() until the tap connection is ready - It's not clear that the host rejecting the connection is better than the host accepting, then the connection stalling until the guest is ready. - Would require substantial rework because we currently listen() as we parse the command line and don't store the information we'd need to do it later. * Don't accept() until the tap connection is ready - To the peer, will behave basically the same as this patch - the host will complete the TCP handshake, then the connection will stall until the guest is ready. - More work to implement, because essentially every sock-side handler has to check fd_tap and abort early * Drop packets in tap_send_frames(), but return 0 - To the peer, would behave basically the same - Would make the TCP code do a bunch of busy work attempting to resend, probably to no avail - Handling of errors returned by tap_send_frames() is on the basis that it's probably a transient fault (buffer full) and we want to resend very soon. That approach doesn't make sense for a missing guest. Suggested-by: David Gibson <david(a)gibson.dropbear.id.au> Signed-off-by: Yumei Huang <yuhuang(a)redhat.com> --- tap.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/tap.c b/tap.c index 7ba6399..2e371b3 100644 --- a/tap.c +++ b/tap.c @@ -507,13 +507,16 @@ static size_t tap_send_frames_passt(const struct ctx *c, * @iov must have total length @bufs_per_frame * @nframes, with each set of * @bufs_per_frame contiguous buffers representing a single frame. * - * Return: number of frames actually sent + * Return: number of frames actually sent, or accounted as sent */ size_t tap_send_frames(const struct ctx *c, const struct iovec *iov, size_t bufs_per_frame, size_t nframes) { size_t m; + if (c->fd_tap == -1) + return nframes; + if (!nframes) return 0; -- 2.47.0

1 0

[PATCH] Add CONTRIBUTING.md
by Yumei Huang 12 Sep '25

12 Sep '25

Signed-off-by: Yumei Huang <yuhuang(a)redhat.com> --- CONTRIBUTING.md | 86 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 86 insertions(+) create mode 100644 CONTRIBUTING.md diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md new file mode 100644 index 0000000..70e35b1 --- /dev/null +++ b/CONTRIBUTING.md @@ -0,0 +1,86 @@ +# Contributing to passt + +Thank you for your interest in contributing! This document explains how to prepare patches and participate in the email-based review process. + +## Workflow + +### 1. Clone the project + + git clone git://passt.top/passt + +### 2. Create a Branch + + cd passt + git checkout -b my-feature-branch + +Work on a separate branch instead of committing directly to `master`. + +### 3. Make Changes and Commit + +* Edit the source code or documentation. + +* Stage your changes: + + git add <file1> <file2> ... + +* Commit with a clear message containing `Signed-off-by:` tag: + + git commit -s + + Commit message format: + + Subsystem: Brief summary (max ~50 chars) + + More detailed explanation if needed, wrapped at 72 chars. + + Signed-off-by: Your Name <author@email> + + If there are some references, use "Links: " tag for simplicity. + +### 4. Generate Patches + +Use `git format-path` to generate patch(es): + + git format-patch origin/master + +This will generate numbered patch files such as 0001-...patch, 0002-...patch, etc. + +If you send a series of patches, use the `--cover-letter` option with `git format-patch`: + + git format-patch origin/main --cover-letter + +This will generate a cover letter besides your patches. You can edit the cover letter before sending. + +### 5. Send Patches + +Use `git send-email` to send patches directly to the mailing list: + + git send-email --to=passt-dev(a)passt.top 000*.patch + +If there are CCs (e.g. maintainers, reviewers), you can add them with `--cc`: + + git send-email --to=passt-dev(a)passt.top --cc=maintainer(a)example.com 000*.patch + +### 6. Responding to Review Feedback + +* Be open to feedback on both code and documentation. + +* Update your patch as needed, and regenerate patches: + + git add <file1> <file2> ... + git commit --amend + git format-patch -v2 origin/master + +* Send the revised patches + + git send-email --to=passt-dev(a)passt.top v2-000*.patch + +### 7. Tips and Best Practices + +* Keep changes focused and easy to review. + +* Test your changes thoroughly. + +* Include documentation updates if your change affects usage or APIs. + +Thank you for helping improve passt! Your contributions make a big difference. -- 2.47.0

1 0

passt: new version 2025_09_11.6cbcccc available
by Stefano Brivio 12 Sep '25

12 Sep '25

The new version with tag 2025_09_11.6cbcccc includes the following changes: 6cbcccc tcp: Store the owner connections for flags frames cd2e886 Reduce tcp_buf_discard size 8d2f8c4 tcp: Don't send FIN segment to guest yet if we have pending unacknowledged data bde1847 tcp: Fast re-transmit if half-closed, make TAP_FIN_RCVD path consistent 660cd69 tcp: Cast operands of sequence comparison macros to uint32_t before using them 25f9354 tcp: Don't try to transmit right after the peer shrank the window to zero c62fb08 tcp: Fix closing logic for half-closed connections e86d480 tcp: Rewind sequence when guest shrinks window to zero 1d502be tcp: Factor sequence rewind for retransmissions into a new function 2e3d93b tcp: FIN flags have to be retransmitted as well d363fb7 test: Fix the download link for debian-11-generic-ppc64el image c10d8c9 tcp_vu: Pass virtqueue pointer to tcp_vu_sock_recv() ed18d4c udp_vu: Pass virtqueue pointer to udp_vu_sock_recv() 6239915 vhost-user: Fix VHOST_USER_GET_QUEUE_NUM to return number of queues f9ee749 Add missing explicit PSH assignment 83afb88 Fix typo in doc comment 00e3580 test: Explicit specify forwarding ports for pasta in log rotation tests 3c44ef8 test: Allow exeter & podman tests to be parallel executed with BATS ffe34d2 test: Convert build tests to exeter a58e60e test: Run static checkers as exeter tests a283ef4 test: Extend test scripts to allow running exeter tests. ca38be0 packet: Add support for multi-vector packets 3e43e1a packet: Refactor vhost-user memory region handling 1602aa2 packet: remove unused parameter from PACKET_POOL_DECL() 7ae35a9 packet: remove PACKET_POOL() and PACKET_POOL_P() 2eb845a ndp: use iov_tail rather than pool 76de6f5 icmp: use iov_tail rather than pool 42a108b dhcpv6: use iov_tail rather than pool 3a261fd dhcp: use iov_tail rather than pool c977d1f arp: use iov_tail rather than pool 7e25351 packet: rename packet_data() to packet_get() 9505908 tap: Convert tap6_handler() to iov_tail a26c608 tap: Convert tap4_handler() to iov_tail 20cd6d0 ip: Use iov_tail in ipv6_l4hdr() 84a4d3e dhcp: Convert to iov_tail feb3330 dhcpv6: Use iov_tail in dhcpv6_opt() c4cad31 dhcpv6: Convert to iov_tail 54f15c6 dhcpv6: Extract sending of NotOnLink status 1932832 dhcpv6: move offset initialization out of dhcpv6_opt() d2c33f4 tcp: Convert tcp_data_from_tap() to use iov_tail 87cc7ab tcp: Convert tcp_tap_handler() to use iov_tail d9604f0 udp: Convert to iov_tail e45bf13 icmp: Convert to iov_tail f8860bb ndp: Convert to iov_tail 1fc944c arp: Convert to iov_tail 6bada9a packet: Add packet_data() de469a3 packet: Use iov_tail with packet_add() 720d8fc tap: Use iov_tail with tap_add_packet() 70b9c0c iov: Update IOV_REMOVE_HEADER() and IOV_PEEK_HEADER() 066e6b1 iov: Introduce iov_tail_clone() and iov_drop_header(). ea3dd28 arp: Don't mix incoming and outgoing buffers eef5bb8 build: Fix errors of TCP_REPAIR_* undeclared b4fc6cd treewide: Flush pcap and log files, if used, before exiting https://passt.top/passt/log/?qt=range&q=2025_08_05.309eefd..2025_09_11.6cbc… Packages: - Alpine Linux: https://pkgs.alpinelinux.org/packages?name=passt - Arch Linux: https://www.archlinux.org/packages/extra/x86_64/passt/ https://archlinuxarm.org/packages/aarch64/passt https://archlinuxarm.org/packages/armv7h/passt - Chimera: https://pkgs.chimera-linux.org/packages?name=passt - Clear Linux: https://github.com/clearlinux-pkgs/passt/ - Copr (CentOS Stream, EPEL, Fedora, Mageia): https://copr.fedorainfracloud.org/coprs/sbrivio/passt/build/9548830/ permanent mirror: https://passt.top/builds/copr/0^20250911.g6cbcccc/ - Debian tracker: https://tracker.debian.org/pkg/passt - Fedora updates: https://bodhi.fedoraproject.org/updates/?packages=passt - Gentoo versions: https://packages.gentoo.org/packages/net-misc/passt - GNU Guix: https://packages.guix.gnu.org/packages/passt/ - Homebrew: https://formulae.brew.sh/formula/passt - NixOS: https://github.com/NixOS/nixpkgs/tree/nixos-unstable/pkgs/by-name/pa/passt - openSUSE: https://software.opensuse.org/package/passt - OpenMandriva: https://github.com/OpenMandrivaAssociation/passt/tree/master - PLD Linux: https://git.pld-linux.org/cgi-bin/gitweb.cgi?p=packages/passt.git - Solus: https://github.com/getsolus/packages/tree/main/packages/p/passt - Ubuntu tracker: https://launchpad.net/ubuntu/+source/passt - Void Linux: https://voidlinux.org/packages/?q=passt - Static builds: - Package for other RPM-based distributions, x86_64 only: https://passt.top/builds/latest/x86_64/passt-g6cbcccc-1.x86_64.rpm - x86_64 static binaries: https://passt.top/builds/latest/x86_64/ - Debian package, from x86_64 static build: https://passt.top/builds/latest/x86_64/passt_6cbcccc-1_all.deb -- Stefano

1 0

[PATCH v4] Send an initial ARP and NDP request to resolve the guest IP
by Volker Diels-Grabsch 12 Sep '25

12 Sep '25

Changes from PATCH v3: (... which wasn't marked as "v3", sorry for that) 1. Simply remove the "e.g." part from the comment, as it's misleading anyways 2. Make "bufmac[]" as local as possible 3. Demote the "Sending initial ..." log message from info() to debug() 4. Log "Sending initial ..." separately for ARP and NDP (because of 5. and 6.) 5. Send initial ARP request only if IPv4 is enabled 6. Send initial NDP request only if IPv6, NDP and ICMP are enabled 7. Update description of "--no-ndp" in man page 8. Fix and update description of "--no-icmp" in man page 9. Break long line for MAC_BROADCAST One final thought regarding 8.: Would it make sense to state in the man page that "--no-icmp" implies "--no-ndp", and to implement it rigorously that way? Or, would that be going to far?

2 5

[PATCH v7 0/8] Use true MAC address of LAN local remote hosts
by Jon Maloy 12 Sep '25

12 Sep '25

Bug #120 asks us to use the true MAC addresses of LAN local remote hosts, since some programs need this information. These commits introduces this for ARP, NDP, UDP, TCP and ICMP. --- v3: Updated according to feedback from Stefano and David: - Made the ARP/NDP lookup call filter out the requested address by itself, qualified by the index if the template interface - Moved the flow specific MAC address from struct flowside to struct flow_common. v4: - Updated according to feedback from David and Stefan - Added a cache table for ARP/NDP table contents v5: - Updated according to feedback from David and Stefan - Added cache table entries to FIFO/LRU queue - New criteria for when to consult ARP/NDP v6: - Simplified and merged mac cache table commits - Other changes after feedback from David. v7: - Fixes in patch #2 based on feedback from David and Stefano. Jon Maloy (8): netlink: add function to extract MAC addresses from NDP/ARP table fwd: Added cache table for ARP/NDP contents arp/ndp: respond with true MAC address of LAN local remote hosts flow: add MAC address of LAN local remote hosts to flow udp: forward external source MAC address through tap interface tcp: forward external source MAC address through tap interface tap: change signature of function tap_push_l2h() icmp: let icmp use mac address from flowside structure arp.c | 9 ++- conf.c | 1 + flow.c | 2 + flow.h | 2 + fwd.c | 184 +++++++++++++++++++++++++++++++++++++++++++++++-- fwd.h | 6 ++ icmp.c | 8 ++- inany.c | 1 + ndp.c | 10 ++- netlink.c | 82 ++++++++++++++++++++++ netlink.h | 2 + passt.c | 9 ++- passt.h | 3 +- pasta.c | 2 +- tap.c | 24 ++++--- tap.h | 7 +- tcp.c | 18 +++-- tcp.h | 2 +- tcp_buf.c | 37 +++++----- tcp_internal.h | 4 +- tcp_vu.c | 5 +- udp.c | 57 +++++++++------ udp.h | 2 +- util.c | 12 ++++ util.h | 1 + 25 files changed, 407 insertions(+), 83 deletions(-) -- 2.50.1

2 9

[PATCH] tap: Drop frames if no client connected
by Yumei Huang 12 Sep '25

12 Sep '25

If no client is attached, discard outgoing frames and report them as sent. This mimics the behavior of a physical host with its network cable unplugged. Suggested-by: David Gibson <david(a)gibson.dropbear.id.au> Signed-off-by: Yumei Huang <yuhuang(a)redhat.com> --- tap.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/tap.c b/tap.c index 7ba6399..e01219d 100644 --- a/tap.c +++ b/tap.c @@ -507,13 +507,17 @@ static size_t tap_send_frames_passt(const struct ctx *c, * @iov must have total length @bufs_per_frame * @nframes, with each set of * @bufs_per_frame contiguous buffers representing a single frame. * - * Return: number of frames actually sent + * Return: number of frames actually sent, or accounted as sent */ size_t tap_send_frames(const struct ctx *c, const struct iovec *iov, size_t bufs_per_frame, size_t nframes) { size_t m; + if (c->fd_tap == -1) + /* If no client connected, account the frames have been sent */ + return nframes; + if (!nframes) return 0; -- 2.47.0

3 3

[PATCH v3] tcp: Store the owner connections for flags frames
by Yumei Huang 11 Sep '25

11 Sep '25

There is an issue reported by Volker Diels-Grabsch and Boleyn Su. A segmentation fault occurs when executing the following command: (sleep 0.1; ssh -p 22000 127.0.0.1) & passt -f -t 22000:22 It's caused by commit 78da088f7bab ("tcp: unify payload and flags l2 frames array"). Fix it by storing the owner connections of flags frames into tcp_frame_conns[] array. Reported-by: Volker Diels-Grabsch <v(a)njh.eu> Reported-by: Boleyn Su <boleyn.su(a)gmail.com> Suggested-by: David Gibson <david(a)gibson.dropbear.id.au> Fixes: 78da088f7bab ("tcp: unify payload and flags l2 frames array") Signed-off-by: Yumei Huang <yuhuang(a)redhat.com> --- tcp_buf.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/tcp_buf.c b/tcp_buf.c index bc898de..d63c18d 100644 --- a/tcp_buf.c +++ b/tcp_buf.c @@ -209,13 +209,14 @@ int tcp_buf_send_flag(const struct ctx *c, struct tcp_tap_conn *conn, int flags) if (ret <= 0) return ret; - tcp_payload_used++; + tcp_frame_conns[tcp_payload_used++] = conn; l4len = optlen + sizeof(struct tcphdr); iov[TCP_IOV_PAYLOAD].iov_len = l4len; tcp_l2_buf_fill_headers(conn, iov, NULL, seq, false); if (flags & DUP_ACK) { - struct iovec *dup_iov = tcp_l2_iov[tcp_payload_used++]; + struct iovec *dup_iov = tcp_l2_iov[tcp_payload_used]; + tcp_frame_conns[tcp_payload_used++] = conn; memcpy(dup_iov[TCP_IOV_TAP].iov_base, iov[TCP_IOV_TAP].iov_base, iov[TCP_IOV_TAP].iov_len); -- 2.47.0

3 2

[PATCH] Reduce tcp_buf_discard size
by xugu＠redhat.com 11 Sep '25

11 Sep '25

From: Xun Gu <xugu(a)redhat.com> On kernels without SO_PEEK_OFF, a 16MB static buffer is used to discard sent data. This patch reduces the buffer to 1MB. Larger discards are now handled by using multiple iovec entries pointing to the same 1MB buffer. Signed-off-by: Xun Gu <xugu(a)redhat.com> --- tcp.c | 66 +++++++++++++++++++++++++++++++++++++++++++++++++- tcp_buf.c | 18 +++++--------- tcp_internal.h | 7 +++++- tcp_vu.c | 17 ++++--------- 4 files changed, 82 insertions(+), 26 deletions(-) diff --git a/tcp.c b/tcp.c index a27b069..253cdb3 100644 --- a/tcp.c +++ b/tcp.c @@ -399,7 +399,7 @@ static int tcp_sock_ns [NUM_PORTS][IP_VERSIONS]; */ static union inany_addr low_rtt_dst[LOW_RTT_TABLE_SIZE]; -char tcp_buf_discard [MAX_WINDOW]; +char tcp_buf_discard [BUF_DISCARD_SIZE]; /* Does the kernel support TCP_PEEK_OFF? */ bool peek_offset_cap; @@ -3766,3 +3766,67 @@ fail: return 0; } + +/** + * tcp_prepare_iov() - Prepare iov according to kernel capability + * @msg: Message header to update + * @iov: iovec to receive TCP payload and data to discard + * @already_sent: Bytes sent after the last acknowledged one + * @payload_iov_cnt: Number of TCP payload iovec entries + * + * Return: 0 on success, -1 if already_sent cannot be discarded fully + */ +int tcp_prepare_iov(struct msghdr *msg, struct iovec *iov, + uint32_t already_sent, int payload_iov_cnt) +{ + /* + * IOV layout + * |- tcp_buf_discard -|---------- TCP data slots ------------| + * + * with discarded data: + * |------ddddddddddddd|ttttttttttttt-------------------------| + * ^ + * | + * msg_iov + * + * without discarded data: + * |-------------------|ttttttttttttt-------------------------| + * ^ + * | + * msg_iov + * d: discard data + * t: TCP data + */ + if (peek_offset_cap) { + msg->msg_iov = iov + DISCARD_IOV_NUM; + msg->msg_iovlen = payload_iov_cnt; + } else { + int discard_cnt, discard_iov_rem; + struct iovec *iov_start; + int i; + + discard_cnt = DIV_ROUND_UP(already_sent, BUF_DISCARD_SIZE); + if (discard_cnt > DISCARD_IOV_NUM) { + debug("Failed to discard %u already sent bytes", + already_sent); + return -1; + } + + discard_iov_rem = already_sent % BUF_DISCARD_SIZE; + + iov_start = iov + (DISCARD_IOV_NUM - discard_cnt); + + /* Multiple iov entries pointing to the same buffer */ + for (i = 0; i < discard_cnt; i++) { + iov_start[i].iov_base = tcp_buf_discard; + iov_start[i].iov_len = BUF_DISCARD_SIZE; + } + if (discard_iov_rem) + iov[DISCARD_IOV_NUM - 1].iov_len = discard_iov_rem; + + msg->msg_iov = iov_start; + msg->msg_iovlen = discard_cnt + payload_iov_cnt; + } + + return 0; +} diff --git a/tcp_buf.c b/tcp_buf.c index bc898de..4ebb013 100644 --- a/tcp_buf.c +++ b/tcp_buf.c @@ -60,7 +60,7 @@ static struct tcp_tap_conn *tcp_frame_conns[TCP_FRAMES_MEM]; static unsigned int tcp_payload_used; /* recvmsg()/sendmsg() data for tap */ -static struct iovec iov_sock [TCP_FRAMES_MEM + 1]; +static struct iovec iov_sock [TCP_FRAMES_MEM + DISCARD_IOV_NUM]; static struct iovec tcp_l2_iov[TCP_FRAMES_MEM][TCP_NUM_IOVS]; @@ -326,15 +326,9 @@ int tcp_buf_data_from_sock(const struct ctx *c, struct tcp_tap_conn *conn) iov_rem = (wnd_scaled - already_sent) % mss; } - /* Prepare iov according to kernel capability */ - if (!peek_offset_cap) { - mh_sock.msg_iov = iov_sock; - iov_sock[0].iov_base = tcp_buf_discard; - iov_sock[0].iov_len = already_sent; - mh_sock.msg_iovlen = fill_bufs + 1; - } else { - mh_sock.msg_iov = &iov_sock[1]; - mh_sock.msg_iovlen = fill_bufs; + if (tcp_prepare_iov(&mh_sock, iov_sock, already_sent, fill_bufs)) { + tcp_rst(c, conn); + return -1; } if (tcp_payload_used + fill_bufs > TCP_FRAMES_MEM) { @@ -344,12 +338,12 @@ int tcp_buf_data_from_sock(const struct ctx *c, struct tcp_tap_conn *conn) tcp_payload_used = 0; } - for (i = 0, iov = iov_sock + 1; i < fill_bufs; i++, iov++) { + for (i = 0, iov = iov_sock + DISCARD_IOV_NUM; i < fill_bufs; i++, iov++) { iov->iov_base = &tcp_payload[tcp_payload_used + i].data; iov->iov_len = mss; } if (iov_rem) - iov_sock[fill_bufs].iov_len = iov_rem; + iov_sock[fill_bufs + DISCARD_IOV_NUM - 1].iov_len = iov_rem; /* Receive into buffers, don't dequeue until acknowledged by guest. */ do diff --git a/tcp_internal.h b/tcp_internal.h index 9dae688..d0009f8 100644 --- a/tcp_internal.h +++ b/tcp_internal.h @@ -9,6 +9,9 @@ #define MAX_WS 8 #define MAX_WINDOW (1 << (16 + (MAX_WS))) +#define BUF_DISCARD_SIZE (1 << 20) +#define DISCARD_IOV_NUM DIV_ROUND_UP(MAX_WINDOW, BUF_DISCARD_SIZE) + #define MSS4 ROUND_DOWN(IP_MAX_MTU - \ sizeof(struct tcphdr) - \ sizeof(struct iphdr), \ @@ -139,7 +142,7 @@ struct tcp_syn_opts { .ws = TCP_OPT_WS(ws_), \ }) -extern char tcp_buf_discard [MAX_WINDOW]; +extern char tcp_buf_discard [BUF_DISCARD_SIZE]; void conn_flag_do(const struct ctx *c, struct tcp_tap_conn *conn, unsigned long flag); @@ -180,4 +183,6 @@ int tcp_prepare_flags(const struct ctx *c, struct tcp_tap_conn *conn, size_t *optlen); int tcp_set_peek_offset(const struct tcp_tap_conn *conn, int offset); +int tcp_prepare_iov(struct msghdr *msg, struct iovec *iov, + uint32_t already_sent, int payload_iov_cnt); #endif /* TCP_INTERNAL_H */ diff --git a/tcp_vu.c b/tcp_vu.c index cb39bc2..097ca13 100644 --- a/tcp_vu.c +++ b/tcp_vu.c @@ -35,7 +35,7 @@ #include "vu_common.h" #include <time.h> -static struct iovec iov_vu[VIRTQUEUE_MAX_SIZE + 1]; +static struct iovec iov_vu[VIRTQUEUE_MAX_SIZE + DISCARD_IOV_NUM]; static struct vu_virtq_element elem[VIRTQUEUE_MAX_SIZE]; static int head[VIRTQUEUE_MAX_SIZE + 1]; @@ -200,7 +200,7 @@ static ssize_t tcp_vu_sock_recv(const struct ctx *c, hdrlen = tcp_vu_hdrlen(v6); - vu_init_elem(elem, &iov_vu[1], VIRTQUEUE_MAX_SIZE); + vu_init_elem(elem, &iov_vu[DISCARD_IOV_NUM], VIRTQUEUE_MAX_SIZE); elem_cnt = 0; *head_cnt = 0; @@ -228,16 +228,9 @@ static ssize_t tcp_vu_sock_recv(const struct ctx *c, elem_cnt += cnt; } - if (peek_offset_cap) { - mh_sock.msg_iov = iov_vu + 1; - mh_sock.msg_iovlen = elem_cnt; - } else { - iov_vu[0].iov_base = tcp_buf_discard; - iov_vu[0].iov_len = already_sent; - - mh_sock.msg_iov = iov_vu; - mh_sock.msg_iovlen = elem_cnt + 1; - } + if (tcp_prepare_iov(&mh_sock, iov_vu, already_sent, elem_cnt)) + /* Expect caller to do a TCP reset */ + return -1; do ret = recvmsg(s, &mh_sock, MSG_PEEK); -- 2.51.0

2 3

[PATCH v4 0/8] tcp: Fixes for issues uncovered by tests with 6.17-rc1 kernels
by Stefano Brivio 11 Sep '25

11 Sep '25

Starting from Linux kernel commit 1d2fbaad7cd8 ("tcp: stronger sk_rcvbuf checks"), window limits are enforced more aggressively with a bigger amount of zero-window updates compared to what happened with e2142825c120 ("net: tcp: send zero-window ACK when no memory") alone, and occasional duplicate ACKs can now be seen also for local transfers with default (208 KiB) socket buffer sizes. Paul reports that, with 6.17-rc1-ish kernels, Podman tests for the pasta integration occasionally fail on the "TCP/IPv4 large transfer, tap" case. While playing with a reproducer that seems to be matching those failures: while true; do ./pasta --trace -l /tmp/pasta.log -p /tmp/pasta.pcap --config-net -t 5555 -- socat TCP-LISTEN:5555 OPEN:/tmp/large.rcv,trunc & (sleep 0.3; socat -T2 OPEN:large.bin TCP:88.198.0.164:5555; ); wait; diff large.bin /tmp/large.rcv || break; done and a kernel including that commit, I hit a few different failures, that should be fixed by this series. Paul tested v1 of this series and found an additional failure (transfer timeout), which I could reproduce with a slightly different command: while true; do ./pasta --trace -l /tmp/pasta.log -p /tmp/pasta.pcap --config-net -t 5555 -- socat TCP-LISTEN:5555 EXEC:./write.sh & (sleep 0.3; socat -T2 OPEN:large.bin TCP:88.198.0.164:5555; ); wait; diff large.bin /tmp/large.rcv || break; done where write.sh is simply: #!/bin/sh cat > /tmp/large.rcv so that the connection is not half-closed starting from the beginning, because socat can't make assumptions about the unidirectional nature of the traffic. This should now be fixed as well by the new version of patch 3/8. v4: - add patch 8/8 v3: - add patch 6/7 - in 7/7, check dlen <= 1 for keep-alive segments, instead of len <= 1 v2: in 3/6, rewind sequence also if the zero-window update comes in the middle of a batch with non-zero window updates Stefano Brivio (8): tcp: FIN flags have to be retransmitted as well tcp: Factor sequence rewind for retransmissions into a new function tcp: Rewind sequence when guest shrinks window to zero tcp: Fix closing logic for half-closed connections tcp: Don't try to transmit right after the peer shrank the window to zero tcp: Cast operands of sequence comparison macros to uint32_t before using them tcp: Fast re-transmit if half-closed, make TAP_FIN_RCVD path consistent tcp: Don't send FIN segment to guest yet if we have pending unacknowledged data tcp.c | 142 ++++++++++++++++++++++++++++++++++++++----------- tcp_buf.c | 5 +- tcp_internal.h | 12 +++-- 3 files changed, 122 insertions(+), 37 deletions(-) -- 2.43.0

4 26