Signed-off-by: Stefano Brivio --- slirp4netns.sh | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/slirp4netns.sh b/slirp4netns.sh index de74281..e6a6049 100755 --- a/slirp4netns.sh +++ b/slirp4netns.sh @@ -17,7 +17,10 @@ PASTA_PID="$(mktemp)" PASTA_OPTS="-q --ipv4-only -a 10.0.2.0 -n 24 -g 10.0.2.2 -m 1500 --no-ndp --no-dhcpv6 --no-dhcp -P ${PASTA_PID}" +PASTA="$(command -v ./pasta || command -v pasta || :)" + USAGE_RET=1 +NOTFOUND_RET=127 # add() - Add single option to $PASTA_OPTS # $1: Option name, with or without argument @@ -161,6 +164,8 @@ no_map_gw=0 EFD=0 RFD=0 +[ -z "${PASTA}" ] && echo "pasta command not found" && exit ${NOTFOUND_RET} + while getopts ce:r:m:6a:hv-: OPT 2>/dev/null; do if [ "${OPT}" = "-" ]; then OPT="${OPTARG%%[= ]*}" @@ -198,7 +203,7 @@ if [ ${v6} -eq 1 ]; then add "-a $(gen_addr6) -g fd00::2 -D fd00::3" fi -./pasta ${PASTA_OPTS} ${ns_spec} 2>/dev/null && \ +${PASTA} ${PASTA_OPTS} ${ns_spec} && \ [ ${RFD} -ne 0 ] && echo "1" >&${RFD} trap "kill $(cat ${PASTA_PID}); rm ${PASTA_PID}" INT TERM -- 2.34.1

To reach (at least) a conceptually equivalent security level as implemented by --enable-sandbox in slirp4netns, we need to create a new mount namespace and pivot_root() into a new (empty) mountpoint, so that passt and pasta can't access any filesystem resource after initialisation. While at it, also detach IPC, PID (only for passt, to prevent vulnerabilities based on the knowledge of a target PID), and UTS namespaces. With this approach, if we apply the seccomp filters right after the configuration step, the number of allowed syscalls grows further. To prevent this, defer the application of seccomp policies after the initialisation phase, before the main loop, that's where we expect bad things to happen, potentially. This way, we get back to 22 allowed syscalls for passt and 34 for pasta, on x86_64. While at it, move #syscalls notes to specific code paths wherever it conceptually makes sense. We have to open all the file handles we'll ever need before sandboxing: - the packet capture file can only be opened once, drop instance numbers from the default path and use the (pre-sandbox) PID instead - /proc/net/tcp{,v6} and /proc/net/udp{,v6}, for automatic detection of bound ports in pasta mode, are now opened only once, before sandboxing, and their handles are stored in the execution context - the UNIX domain socket for passt is also bound only once, before sandboxing: to reject clients after the first one, instead of closing the listening socket, keep it open, accept and immediately discard new connection if we already have a valid one Clarify the (unchanged) behaviour for --netns-only in the man page. To actually make passt and pasta processes run in a separate PID namespace, we need to unshare(CLONE_NEWPID) before forking to background (if configured to do so). Introduce a small daemon() implementation, __daemon(), that additionally saves the PID file before forking. While running in foreground, the process itself can't move to a new PID namespace (a process can't change the notion of its own PID): mention that in the man page. For some reason, fork() in a detached PID namespace causes SIGTERM and SIGQUIT to be ignored, even if the handler is still reported as SIG_DFL: add a signal handler that just exits. We can now drop most of the pasta_child_handler() implementation, that took care of terminating all processes running in the same namespace, if pasta started a shell: the shell itself is now the init process in that namespace, and all children will terminate once the init process exits. Issuing 'echo $$' in a detached PID namespace won't return the actual namespace PID as seen from the init namespace: adapt demo and test setup scripts to reflect that. Signed-off-by: Stefano Brivio --- README.md | 5 +- conf.c | 45 +++++++------ passt.1 | 15 +++-- passt.c | 126 ++++++++++++++++++++++-------------- passt.h | 7 +- pasta.c | 165 +++++++++++++++++------------------------------- pcap.c | 5 +- pcap.h | 2 +- slirp4netns.sh | 2 +- tap.c | 58 ++++++++--------- tcp.c | 13 ++-- test/demo/passt | 3 +- test/demo/pasta | 5 +- test/lib/setup | 28 ++++---- udp.c | 7 +- util.c | 129 ++++++++++++++++++++++++++++++++----- util.h | 12 +++- 17 files changed, 365 insertions(+), 262 deletions(-) diff --git a/README.md b/README.md index d16b705..1c8baf3 100644 --- a/README.md +++ b/README.md @@ -232,9 +232,10 @@ speeding up local connections, and usually requiring NAT. _pasta_: `seccomp`](/passt/tree/seccomp.sh)) * ✅ root operation not allowed outside user namespaces * ✅ all capabilities dropped, other than `CAP_NET_BIND_SERVICE` (if granted) +* ✅ with default options, user, mount, IPC, UTS, PID namespaces are detached * ✅ no external dependencies (other than a standard C library) -* ✅ restrictive seccomp profiles (50 syscalls allowed for _passt_, 62 for - _pasta_) +* ✅ restrictive seccomp profiles (22 syscalls allowed for _passt_, 34 for + _pasta_ on x86_64) * ✅ static checkers in continuous integration (clang-tidy, cppcheck) * 🛠️ rework of TCP state machine (flags instead of states), TCP timers, and code de-duplication diff --git a/conf.c b/conf.c index abe63a1..732d918 100644 --- a/conf.c +++ b/conf.c @@ -10,8 +10,6 @@ * * Copyright (c) 2020-2021 Red Hat GmbH * Author: Stefano Brivio - * - * #syscalls stat|statx */ #include @@ -46,31 +44,31 @@ */ void get_bound_ports(struct ctx *c, int ns, uint8_t proto) { - uint8_t *udp_map, *udp_exclude, *tcp_map, *tcp_exclude; + uint8_t *udp_map, *udp_excl, *tcp_map, *tcp_excl; if (ns) { udp_map = c->udp.port_to_tap; - udp_exclude = c->udp.port_to_init; + udp_excl = c->udp.port_to_init; tcp_map = c->tcp.port_to_tap; - tcp_exclude = c->tcp.port_to_init; + tcp_excl = c->tcp.port_to_init; } else { udp_map = c->udp.port_to_init; - udp_exclude = c->udp.port_to_tap; + udp_excl = c->udp.port_to_tap; tcp_map = c->tcp.port_to_init; - tcp_exclude = c->tcp.port_to_tap; + tcp_excl = c->tcp.port_to_tap; } if (proto == IPPROTO_UDP) { memset(udp_map, 0, USHRT_MAX / 8); - procfs_scan_listen("udp", udp_map, udp_exclude); - procfs_scan_listen("udp6", udp_map, udp_exclude); + procfs_scan_listen(c, IPPROTO_UDP, V4, ns, udp_map, udp_excl); + procfs_scan_listen(c, IPPROTO_UDP, V6, ns, udp_map, udp_excl); - procfs_scan_listen("tcp", udp_map, udp_exclude); - procfs_scan_listen("tcp6", udp_map, udp_exclude); + procfs_scan_listen(c, IPPROTO_TCP, V4, ns, udp_map, udp_excl); + procfs_scan_listen(c, IPPROTO_TCP, V6, ns, udp_map, udp_excl); } else if (proto == IPPROTO_TCP) { memset(tcp_map, 0, USHRT_MAX / 8); - procfs_scan_listen("tcp", tcp_map, tcp_exclude); - procfs_scan_listen("tcp6", tcp_map, tcp_exclude); + procfs_scan_listen(c, IPPROTO_TCP, V4, ns, tcp_map, tcp_excl); + procfs_scan_listen(c, IPPROTO_TCP, V6, ns, tcp_map, tcp_excl); } } @@ -367,7 +365,7 @@ static int conf_ns_check(void *arg) static int conf_ns_opt(struct ctx *c, char *nsdir, char *conf_userns, const char *optarg) { - int ufd = 0, nfd = 0, try, ret, netns_only_reset = c->netns_only; + int ufd = -1, nfd = -1, try, ret, netns_only_reset = c->netns_only; char userns[PATH_MAX] = { 0 }, netns[PATH_MAX]; char *endptr; pid_t pid; @@ -416,7 +414,7 @@ static int conf_ns_opt(struct ctx *c, nfd = open(netns, O_RDONLY); - if (nfd >= 0 && ufd >= 0) { + if (nfd >= 0 && (ufd >= 0 || c->netns_only)) { c->pasta_netns_fd = nfd; c->pasta_userns_fd = ufd; @@ -425,10 +423,10 @@ static int conf_ns_opt(struct ctx *c, return 0; } - if (nfd > 0) + if (nfd >= 0) close(nfd); - if (ufd > 0) + if (ufd >= 0) close(ufd); } @@ -565,9 +563,9 @@ static void usage(const char *name) info( " if FILE is not given, log to:"); if (strstr(name, "pasta") || strstr(name, "passt4netns")) - info(" /tmp/pasta_ISO8601-TIMESTAMP_INSTANCE-NUMBER.pcap"); + info(" /tmp/pasta_ISO8601-TIMESTAMP_PID.pcap"); else - info(" /tmp/passt_ISO8601-TIMESTAMP_INSTANCE-NUMBER.pcap"); + info(" /tmp/passt_ISO8601-TIMESTAMP_PID.pcap"); info( " -P, --pid FILE Write own PID to the given file"); info( " -m, --mtu MTU Assign MTU via DHCP/NDP"); @@ -664,7 +662,7 @@ pasta_opts: info( " SPEC is as described above"); info( " default: auto"); info( " --userns NSPATH Target user namespace to join"); - info( " --netns-only Don't join or create user namespace"); + info( " --netns-only Don't join existing user namespace"); info( " implied if PATH or NAME are given without --userns"); info( " --nsrun-dir Directory for nsfs mountpoints"); info( " default: " NETNS_RUN_DIR); @@ -1170,7 +1168,7 @@ void conf(struct ctx *c, int argc, char **argv) usage(argv[0]); } - if (c->mode == MODE_PASTA && c->pasta_netns_fd <= 0) + if (c->mode == MODE_PASTA && c->pasta_netns_fd == -1) pasta_start_ns(c); if (nl_sock_init(c)) { @@ -1216,6 +1214,11 @@ void conf(struct ctx *c, int argc, char **argv) c->tcp.init_detect_ports = c->udp.init_detect_ports = 0; if (c->mode == MODE_PASTA) { + c->proc_net_tcp[V4][0] = c->proc_net_tcp[V4][1] = -1; + c->proc_net_tcp[V6][0] = c->proc_net_tcp[V6][1] = -1; + c->proc_net_udp[V4][0] = c->proc_net_udp[V4][1] = -1; + c->proc_net_udp[V6][0] = c->proc_net_udp[V6][1] = -1; + if (!tcp_tap || tcp_tap == PORT_AUTO) { c->tcp.ns_detect_ports = 1; ns_ports_arg.proto = IPPROTO_TCP; diff --git a/passt.1 b/passt.1 index b0d7d87..92681f6 100644 --- a/passt.1 +++ b/passt.1 @@ -80,7 +80,8 @@ Don't print informational messages. .TP .BR \-f ", " \-\-foreground -Don't run in background. +Don't run in background. This implies that the process is not moved to a +detached PID namespace after starting, because the PID itself cannot change. Default is to fork into background. .TP @@ -100,14 +101,13 @@ Capture tap-facing (that is, guest-side or namespace-side) network packets to If \fIfile\fR is not given, capture packets to - \fB/tmp/passt_\fIISO8601-timestamp\fR_\fIinstance-number\fB.pcap\fR + \fB/tmp/passt_\fIISO8601-timestamp\fR_\fIPID\fB.pcap\fR in \fBpasst\fR mode and to - \fB/tmp/pasta_\fIISO8601-timestamp\fR_\fIinstance-number\fB.pcap\fR + \fB/tmp/pasta_\fIISO8601-timestamp\fR_\fIPID\fB.pcap\fR -in \fBpasta\fR mode, where \fIinstance-number\fR is a progressive count of -other detected instances running on the same host. +in \fBpasta\fR mode, where \fIPID\fR is the ID of the running process. .TP .BR \-P ", " \-\-pid " " \fIfile @@ -379,8 +379,9 @@ This option requires PID, PATH or NAME to be specified. .TP .BR \-\-netns-only -Join or create only the network namespace, not a user namespace. This is implied -if PATH or NAME are given without \-\-userns. +Join only a target network namespace, not a user namespace, and don't create one +for sandboxing purposes either. This is implied if PATH or NAME are given +without \-\-userns. .TP .BR \-\-nsrun-dir " " \fIpath diff --git a/passt.c b/passt.c index a8bb88e..508d525 100644 --- a/passt.c +++ b/passt.c @@ -30,7 +30,9 @@ #include #include #include +#include #include +#include #include #include #include @@ -53,7 +55,6 @@ #include #include #include -#include #include #include "util.h" @@ -228,42 +229,61 @@ static void check_root(void) } /** - * drop_caps() - Drop capabilities we might have except for CAP_NET_BIND_SERVICE + * sandbox() - Unshare IPC, mount, PID, UTS, and user namespaces, "unmount" root + * + * Return: negative error code on failure, zero on success */ -static void drop_caps(void) +static int sandbox(struct ctx *c) { - int i; + int flags = CLONE_NEWIPC | CLONE_NEWNS | CLONE_NEWUTS; - for (i = 0; i < 64; i++) { - if (i == CAP_NET_BIND_SERVICE) - continue; + errno = 0; - prctl(PR_CAPBSET_DROP, i, 0, 0, 0); + if (!c->netns_only) { + if (c->pasta_userns_fd == -1) + flags |= CLONE_NEWUSER; + else + setns(c->pasta_userns_fd, CLONE_NEWUSER); } -} -/** - * pid_file() - Write own PID to file, if configured - * @c: Execution context - */ -static void pid_file(struct ctx *c) { - char pid_buf[12]; - int pid_fd, n; + c->pasta_userns_fd = -1; - if (!*c->pid_file) - return; + /* If we run in foreground, we have no chance to actually move to a new + * PID namespace. For passt, use CLONE_NEWPID anyway, in case somebody + * ever gets around seccomp profiles -- there's no harm in passing it. + */ + if (!c->foreground || c->mode == MODE_PASST) + flags |= CLONE_NEWPID; - pid_fd = open(c->pid_file, O_CREAT | O_WRONLY, S_IRUSR | S_IWUSR); - if (pid_fd < 0) - return; + unshare(flags); - n = snprintf(pid_buf, sizeof(pid_buf), "%i\n", getpid()); + mount("", "/", "", MS_UNBINDABLE | MS_REC, NULL); + mount("", TMPDIR, "tmpfs", MS_NODEV | MS_NOEXEC | MS_NOSUID | MS_RDONLY, + "nr_inodes=2,nr_blocks=0"); + chdir(TMPDIR); + syscall(SYS_pivot_root, ".", "."); + umount2(".", MNT_DETACH | UMOUNT_NOFOLLOW); - if (write(pid_fd, pid_buf, n) < 0) { - perror("PID file write"); - exit(EXIT_FAILURE); - } - close(pid_fd); + if (errno) + return -errno; + + drop_caps(); /* Relative to the new user namespace this time. */ + + return 0; +} + +/** + * exit_handler() - Signal handler for SIGQUIT and SIGTERM + * @unused: Unused, handler deals with SIGQUIT and SIGTERM only + * + * TODO: After unsharing the PID namespace and forking, SIG_DFL for SIGTERM and + * SIGQUIT unexpectedly doesn't cause the process to terminate, figure out why. + */ +void exit_handler(int signal) +{ + (void)signal; + + exit(EXIT_SUCCESS); } /** @@ -273,36 +293,36 @@ static void pid_file(struct ctx *c) { * * Return: non-zero on failure * - * #syscalls read write open|openat close fork|clone dup2|dup3 ioctl writev - * #syscalls socket bind connect getsockopt setsockopt recvfrom sendto shutdown - * #syscalls accept4 accept listen set_robust_list getrlimit setrlimit - * #syscalls openat fcntl lseek clone setsid exit exit_group getpid chdir - * #syscalls epoll_ctl epoll_create1 epoll_wait|epoll_pwait epoll_pwait - * #syscalls prlimit64 clock_gettime fstat|newfstat newfstatat syslog - * #syscalls ppc64le:_llseek ppc64le:recv ppc64le:send ppc64le:getuid - * #syscalls ppc64:_llseek ppc64:recv ppc64:send ppc64:getuid ppc64:ugetrlimit - * #syscalls s390x:socketcall s390x:sigreturn - * #syscalls:pasta rt_sigreturn|sigreturn ppc64:sigreturn ppc64:fcntl64 + * #syscalls read write writev + * #syscalls socket bind connect getsockopt setsockopt s390x:socketcall close + * #syscalls recvfrom sendto shutdown ppc64le:recv ppc64le:send + * #syscalls accept4|accept listen + * #syscalls epoll_ctl epoll_wait|epoll_pwait epoll_pwait clock_gettime */ int main(int argc, char **argv) { + int nfds, i, devnull_fd = -1, pidfile_fd = -1; struct epoll_event events[EPOLL_EVENTS]; struct ctx c = { 0 }; struct rlimit limit; struct timespec now; + struct sigaction sa; char *log_name; - int nfds, i; #ifndef PASST_LEGACY_NO_OPTIONS check_root(); #endif drop_caps(); - if (strstr(argv[0], "pasta") || strstr(argv[0], "passt4netns")) { - struct sigaction sa; + c.pasta_userns_fd = c.pasta_netns_fd = c.fd_tap = c.fd_tap_listen = -1; + + sigemptyset(&sa.sa_mask); + sa.sa_flags = 0; + sa.sa_handler = exit_handler; + sigaction(SIGTERM, &sa, NULL); + sigaction(SIGQUIT, &sa, NULL); - sigemptyset(&sa.sa_mask); - sa.sa_flags = 0; + if (strstr(argv[0], "pasta") || strstr(argv[0], "passt4netns")) { sa.sa_handler = pasta_child_handler; sigaction(SIGCHLD, &sa, NULL); signal(SIGPIPE, SIG_IGN); @@ -323,8 +343,6 @@ int main(int argc, char **argv) conf(&c, argc, argv); - seccomp(&c); - if (!c.debug && (c.stderr || isatty(fileno(stdout)))) __openlog(log_name, LOG_PERROR, LOG_DAEMON); @@ -369,12 +387,26 @@ int main(int argc, char **argv) else __setlogmask(LOG_UPTO(LOG_INFO)); - if (!c.foreground && daemon(0, 0)) { - perror("daemon"); + pcap_init(&c); + + if (!c.foreground) + devnull_fd = open("/dev/null", O_RDWR); + + if (*c.pid_file) + pidfile_fd = open(c.pid_file, + O_CREAT | O_WRONLY, S_IRUSR | S_IWUSR); + + if (sandbox(&c)) { + err("Failed to sandbox process, exiting\n"); exit(EXIT_FAILURE); } - pid_file(&c); + if (!c.foreground) + __daemon(pidfile_fd, devnull_fd); + else + write_pidfile(pidfile_fd, getpid()); + + seccomp(&c); timer_init(&c, &now); loop: diff --git a/passt.h b/passt.h index 0ef1897..d7011da 100644 --- a/passt.h +++ b/passt.h @@ -99,8 +99,10 @@ enum passt_modes { * @pcap: Path for packet capture file * @pid_file: Path to PID file, empty string if not configured * @pasta_netns_fd: File descriptor for network namespace in pasta mode - * @pasta_userns_fd: File descriptor for user namespace in pasta mode + * @pasta_userns_fd: Descriptor for user namespace to join, -1 once joined * @netns_only: In pasta mode, don't join or create a user namespace + * @proc_net_tcp: Stored handles for /proc/net/tcp{,6} in init and ns + * @proc_net_udp: Stored handles for /proc/net/udp{,6} in init and ns * @epollfd: File descriptor for epoll instance * @fd_tap_listen: File descriptor for listening AF_UNIX socket, if any * @fd_tap: File descriptor for AF_UNIX socket or tuntap device @@ -155,6 +157,9 @@ struct ctx { int pasta_userns_fd; int netns_only; + int proc_net_tcp[IP_VERSIONS][2]; + int proc_net_udp[IP_VERSIONS][2]; + int epollfd; int fd_tap_listen; int fd_tap; diff --git a/pasta.c b/pasta.c index bce30d4..972cbcf 100644 --- a/pasta.c +++ b/pasta.c @@ -11,9 +11,8 @@ * Copyright (c) 2020-2021 Red Hat GmbH * Author: Stefano Brivio * - * #syscalls:pasta clone unshare waitid kill execve exit_group rt_sigprocmask - * #syscalls:pasta geteuid getdents64|getdents readlink|readlinkat setsid - * #syscalls:pasta nanosleep clock_nanosleep + * #syscalls:pasta clone waitid exit exit_group rt_sigprocmask + * #syscalls:pasta rt_sigreturn|sigreturn ppc64:sigreturn s390x:sigreturn */ #include @@ -40,75 +39,8 @@ #include "passt.h" #include "netlink.h" -/* PID of child, in case we created a namespace, and its procfs link */ +/* PID of child, in case we created a namespace */ static int pasta_child_pid; -static char pasta_child_ns[PATH_MAX]; - -/** - * pasta_ns_cleanup() - Look for processes in namespace, terminate them - */ -static void pasta_ns_cleanup(void) -{ - char proc_path[PATH_MAX], ns_link[PATH_MAX], buf[BUFSIZ]; - int recheck = 0, found = 0, waited = 0; - int dir_fd, n; - - if (!*pasta_child_ns) - return; - -loop: - if ((dir_fd = open("/proc", O_RDONLY | O_DIRECTORY)) < 0) - return; - - while ((n = syscall(SYS_getdents64, dir_fd, buf, BUFSIZ)) > 0) { - struct dirent *dp = (struct dirent *)buf; - int pos = 0; - - while (dp->d_reclen && pos < n) { - pid_t pid; - - errno = 0; - pid = strtol(dp->d_name, NULL, 0); - if (!pid || errno) - goto next; - - snprintf(proc_path, PATH_MAX, "/proc/%i/ns/net", pid); - if (readlink(proc_path, ns_link, PATH_MAX) < 0) - goto next; - - if (!strncmp(ns_link, pasta_child_ns, PATH_MAX)) { - found = 1; - if (waited) - kill(pid, SIGKILL); - else - kill(pid, SIGQUIT); - } -next: - dp = (struct dirent *)(buf + (pos += dp->d_reclen)); - } - } - - close(dir_fd); - - if (!found) - return; - - if (waited) { - if (recheck) { - info("Some processes in namespace didn't quit"); - } else { - found = 0; - recheck = 1; - goto loop; - } - return; - } - - info("Waiting for all processes in namespace to terminate"); - sleep(1); - waited = 1; - goto loop; -} /** * pasta_child_handler() - Exit once shell exits (if we started it), reap clones @@ -120,12 +52,14 @@ void pasta_child_handler(int signal) (void)signal; + if (signal != SIGCHLD) + return; + if (pasta_child_pid && !waitid(P_PID, pasta_child_pid, &infop, WEXITED | WNOHANG)) { - if (infop.si_pid == pasta_child_pid) { - pasta_ns_cleanup(); + if (infop.si_pid == pasta_child_pid) exit(EXIT_SUCCESS); - } + /* Nothing to do, detached PID namespace going away */ } waitid(P_ALL, 0, NULL, WEXITED | WNOHANG); @@ -163,45 +97,31 @@ netns: } /** - * pasta_start_ns() - Fork shell in new namespace if target ns is not given + * struct pasta_setup_ns_arg - Argument for pasta_setup_ns() * @c: Execution context + * @euid: Effective UID of caller */ -void pasta_start_ns(struct ctx *c) +struct pasta_setup_ns_arg { + struct ctx *c; + int euid; +}; + +/** + * pasta_setup_ns() - Map credentials, enable access to ping sockets, run shell + * @arg: See @pasta_setup_ns_arg + * + * Return: this function never returns + */ +static int pasta_setup_ns(void *arg) { - int euid = geteuid(), fd; + struct pasta_setup_ns_arg *a = (struct pasta_setup_ns_arg *)arg; char *shell; + int fd; - c->foreground = 1; - if (!c->debug) - c->quiet = 1; - - if ((pasta_child_pid = fork()) == -1) { - perror("fork"); - exit(EXIT_FAILURE); - } - - if (pasta_child_pid) { - char proc_path[PATH_MAX]; - - NS_CALL(pasta_wait_for_ns, c); - - snprintf(proc_path, PATH_MAX, "/proc/%i/ns/net", - pasta_child_pid); - if (readlink(proc_path, pasta_child_ns, PATH_MAX) < 0) - warn("Cannot read link to ns, won't clean up on exit"); - - return; - } - - if (unshare(CLONE_NEWNET | (c->netns_only ? 0 : CLONE_NEWUSER))) { - perror("unshare"); - exit(EXIT_FAILURE); - } - - if (!c->netns_only) { + if (!a->c->netns_only) { char buf[BUFSIZ]; - snprintf(buf, BUFSIZ, "%i %i %i", 0, euid, 1); + snprintf(buf, BUFSIZ, "%i %i %i", 0, a->euid, 1); fd = open("/proc/self/uid_map", O_WRONLY); if (write(fd, buf, strlen(buf)) < 0) @@ -234,6 +154,39 @@ void pasta_start_ns(struct ctx *c) exit(EXIT_FAILURE); } +/** + * pasta_start_ns() - Fork shell in new namespace if target ns is not given + * @c: Execution context + */ +void pasta_start_ns(struct ctx *c) +{ + struct pasta_setup_ns_arg arg = { .c = c, .euid = geteuid() }; + char ns_fn_stack[NS_FN_STACK_SIZE]; + + c->foreground = 1; + if (!c->debug) + c->quiet = 1; + + pasta_child_pid = clone(pasta_setup_ns, + ns_fn_stack + sizeof(ns_fn_stack) / 2, + (c->netns_only ? 0 : CLONE_NEWNET) | + CLONE_NEWIPC | CLONE_NEWPID | CLONE_NEWUSER | + CLONE_NEWUTS, + (void *)&arg); + + if (pasta_child_pid == -1) { + perror("clone"); + exit(EXIT_FAILURE); + } + + drop_caps(); + + if (pasta_child_pid) { + NS_CALL(pasta_wait_for_ns, c); + return; + } +} + /** * pasta_ns_conf() - Set up loopback and tap interfaces in namespace as needed * @c: Execution context diff --git a/pcap.c b/pcap.c index e00fc45..9c617ce 100644 --- a/pcap.c +++ b/pcap.c @@ -167,9 +167,8 @@ fail: /** * pcap_init() - Initialise pcap file * @c: Execution context - * @index: pcap name index: passt instance number or pasta netns socket */ -void pcap_init(struct ctx *c, int index) +void pcap_init(struct ctx *c) { struct timeval tv; @@ -196,7 +195,7 @@ void pcap_init(struct ctx *c, int index) snprintf(name + strlen(PCAP_PREFIX) + strlen(PCAP_ISO8601_STR), sizeof(name) - strlen(PCAP_PREFIX) - strlen(PCAP_ISO8601_STR), - "_%i.pcap", index); + "_%i.pcap", getpid()); strncpy(c->pcap, name, PATH_MAX); } diff --git a/pcap.h b/pcap.h index 26f4f35..73b5ed8 100644 --- a/pcap.h +++ b/pcap.h @@ -6,4 +6,4 @@ void pcap(char *pkt, size_t len); void pcapm(struct msghdr *mh); void pcapmm(struct mmsghdr *mmh, unsigned int vlen); -void pcap_init(struct ctx *c, int sock_index); +void pcap_init(struct ctx *c); diff --git a/slirp4netns.sh b/slirp4netns.sh index 518f581..7c2188d 100755 --- a/slirp4netns.sh +++ b/slirp4netns.sh @@ -10,7 +10,7 @@ # # slirp4netns.sh - Compatibility wrapper for pasta, behaving like slirp4netns(1) # -# WARNING: Draft quality, not really tested, --enable-sandbox not supported yet +# WARNING: Draft quality, not really tested # # Copyright (c) 2021 Red Hat GmbH # Author: Stefano Brivio diff --git a/tap.c b/tap.c index 22db9c5..38004a5 100644 --- a/tap.c +++ b/tap.c @@ -11,7 +11,6 @@ * Copyright (c) 2020-2021 Red Hat GmbH * Author: Stefano Brivio * - * #syscalls recvfrom sendto */ #include @@ -769,12 +768,10 @@ restart: } /** - * tap_sock_init_unix() - Create and bind AF_UNIX socket, listen for connection + * tap_sock_unix_init() - Create and bind AF_UNIX socket, listen for connection * @c: Execution context - * - * #syscalls:passt unlink|unlinkat */ -static void tap_sock_init_unix(struct ctx *c) +static void tap_sock_unix_init(struct ctx *c) { int fd = socket(AF_UNIX, SOCK_STREAM, 0), ex; struct epoll_event ev = { 0 }; @@ -783,11 +780,6 @@ static void tap_sock_init_unix(struct ctx *c) }; int i, ret; - if (c->fd_tap_listen != -1) { - epoll_ctl(c->epollfd, EPOLL_CTL_DEL, c->fd_tap_listen, &ev); - close(c->fd_tap_listen); - } - if (fd < 0) { perror("UNIX socket"); exit(EXIT_FAILURE); @@ -834,8 +826,6 @@ static void tap_sock_init_unix(struct ctx *c) S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH); #endif - pcap_init(c, i); - listen(fd, 0); ev.data.fd = c->fd_tap_listen = fd; @@ -852,19 +842,26 @@ static void tap_sock_init_unix(struct ctx *c) } /** - * tap_sock_accept_unix() - Accept connection on listening socket + * tap_sock_unix_new() - Handle new connection on listening socket * @c: Execution context */ -static void tap_sock_accept_unix(struct ctx *c) +static void tap_sock_unix_new(struct ctx *c) { struct epoll_event ev = { 0 }; int v = INT_MAX / 2; - c->fd_tap = accept(c->fd_tap_listen, NULL, NULL); + /* Another client is already connected: accept and close right away. */ + if (c->fd_tap != -1) { + int discard = accept4(c->fd_tap_listen, NULL, NULL, + SOCK_NONBLOCK); + + if (discard != -1) + close(discard); - epoll_ctl(c->epollfd, EPOLL_CTL_DEL, c->fd_tap_listen, &ev); - close(c->fd_tap_listen); - c->fd_tap_listen = -1; + return; + } + + c->fd_tap = accept4(c->fd_tap_listen, NULL, NULL, 0); if (!c->low_rmem) setsockopt(c->fd_tap, SOL_SOCKET, SO_RCVBUF, &v, sizeof(v)); @@ -884,8 +881,6 @@ static int tun_ns_fd = -1; * @c: Execution context * * Return: 0 - * - * #syscalls:pasta ioctl */ static int tap_ns_tun(void *arg) { @@ -907,7 +902,7 @@ static int tap_ns_tun(void *arg) * tap_sock_init_tun() - Set up tuntap file descriptor * @c: Execution context */ -static void tap_sock_init_tun(struct ctx *c) +static void tap_sock_tun_init(struct ctx *c) { struct epoll_event ev = { 0 }; @@ -919,8 +914,6 @@ static void tap_sock_init_tun(struct ctx *c) pasta_ns_conf(c); - pcap_init(c, c->pasta_netns_fd); - c->fd_tap = tun_ns_fd; ev.data.fd = c->fd_tap; @@ -937,12 +930,15 @@ void tap_sock_init(struct ctx *c) if (c->fd_tap != -1) { epoll_ctl(c->epollfd, EPOLL_CTL_DEL, c->fd_tap, NULL); close(c->fd_tap); + c->fd_tap = -1; } - if (c->mode == MODE_PASST) - tap_sock_init_unix(c); - else - tap_sock_init_tun(c); + if (c->mode == MODE_PASST) { + if (c->fd_tap_listen == -1) + tap_sock_unix_init(c); + } else { + tap_sock_tun_init(c); + } } /** @@ -955,18 +951,18 @@ void tap_sock_init(struct ctx *c) void tap_handler(struct ctx *c, int fd, uint32_t events, struct timespec *now) { if (fd == c->fd_tap_listen && events == EPOLLIN) { - tap_sock_accept_unix(c); + tap_sock_unix_new(c); return; } if (events & (EPOLLRDHUP | EPOLLHUP | EPOLLERR)) - goto fail; + goto reinit; if ((c->mode == MODE_PASST && tap_handler_passt(c, now)) || (c->mode == MODE_PASTA && tap_handler_pasta(c, now))) - goto fail; + goto reinit; return; -fail: +reinit: tap_sock_init(c); } diff --git a/tcp.c b/tcp.c index 723b18e..e4fac22 100644 --- a/tcp.c +++ b/tcp.c @@ -304,7 +304,7 @@ * - SPLICE_FIN_TO: FIN (EPOLLRDHUP) seen from connected socket * - SPLICE_FIN_BOTH: FIN (EPOLLRDHUP) seen from both sides * - * #syscalls pipe|pipe2 pipe2 + * #syscalls:pasta pipe2|pipe fcntl ppc64:fcntl64 */ #include @@ -3028,7 +3028,7 @@ static void tcp_conn_from_sock(struct ctx *c, union epoll_ref ref, * @ref: epoll reference * @events: epoll events bitmap * - * #syscalls splice + * #syscalls:pasta splice */ void tcp_sock_handler_splice(struct ctx *c, union epoll_ref ref, uint32_t events) @@ -3374,7 +3374,7 @@ static void tcp_set_pipe_size(struct ctx *c) smaller: for (i = 0; i < TCP_SPLICE_PIPE_POOL_SIZE * 2; i++) { - if (pipe(probe_pipe[i])) { + if (pipe2(probe_pipe[i], 0)) { i++; break; } @@ -3493,7 +3493,7 @@ static void tcp_sock_init_one(struct ctx *c, int ns, in_port_t port) * tcp_sock_init_ns() - Bind sockets in namespace for inbound connections * @arg: Execution context * - * Return: 0 on success, -1 on failure + * Return: 0 */ static int tcp_sock_init_ns(void *arg) { @@ -3560,8 +3560,7 @@ static int tcp_sock_refill(void *arg) int i, *p4, *p6; if (a->ns) { - if (ns_enter(a->c)) - return 0; + ns_enter(a->c); p4 = ns_sock_pool4; p6 = ns_sock_pool6; } else { @@ -3594,8 +3593,6 @@ static int tcp_sock_refill(void *arg) * @c: Execution context * * Return: 0 on success, -1 on failure - * - * #syscalls getrandom */ int tcp_sock_init(struct ctx *c, struct timespec *now) { diff --git a/test/demo/passt b/test/demo/passt index b5762aa..76aac86 100644 --- a/test/demo/passt +++ b/test/demo/passt @@ -84,7 +84,8 @@ say Now let's run 'passt' in the new namespace, and nl say enter this namespace from the guest terminal too. sleep 3 -pout TARGET_PID echo $$ +guest pstree -p | grep pasta +gout TARGET_PID pstree -p | grep pasta | sed -n 's/.*(\([0-9].*\))$/\1/p' sleep 1 passtb ./passt -f -t 5201,5203 diff --git a/test/demo/pasta b/test/demo/pasta index f8f0cd0..c136965 100644 --- a/test/demo/pasta +++ b/test/demo/pasta @@ -58,7 +58,8 @@ say For convenience, let's enter this namespace nl say from another terminal. sleep 3 -pout TARGET_PID echo $$ +ns pstree -p | grep pasta +nsout TARGET_PID pstree -p | grep pasta | sed -n 's/.*(\([0-9].*\))$/\1/p' sleep 1 ns nsenter -t __TARGET_PID__ -U -n --preserve-credentials @@ -172,7 +173,7 @@ sleep 2 passtb perf record -g ./pasta sleep 2 -pout TARGET_PID echo $$ +nsout TARGET_PID pstree -p | grep pasta | sed -n 's/.*(\([0-9].*\))$/\1/p' sleep 1 ns nsenter -t __TARGET_PID__ -U -n --preserve-credentials sleep 5 diff --git a/test/lib/setup b/test/lib/setup index ab51787..df21655 100755 --- a/test/lib/setup +++ b/test/lib/setup @@ -115,13 +115,14 @@ setup_passt_in_ns() { [ ${PCAP} -eq 1 ] && __opts="${__opts} -p /tmp/pasta_with_passt.pcap" [ ${DEBUG} -eq 1 ] && __opts="${__opts} -d" - pane_run PASST "./pasta ${__opts} -t 10001,10002,10011,10012 -T 10003,10013 -u 10001,10002,10011,10012 -U 10003,10013" + __pid_file="$(mktemp)" + pane_run PASST "./pasta ${__opts} -t 10001,10002,10011,10012 -T 10003,10013 -u 10001,10002,10011,10012 -U 10003,10013 -P ${__pid_file}" sleep 1 pane_run PASST '' pane_wait PASST - pane_run PASST 'echo $$' - pane_wait PASST - __ns_pid="$(pane_parse PASST)" + __pasta_pid="$(cat "${__pid_file}")" + __ns_pid="$(cat /proc/${__pasta_pid}/task/${__pasta_pid}/children | cut -f1 -d' ')" + rm "${__pid_file}" pane_run GUEST "nsenter -t ${__ns_pid} -U -n --preserve-credentials" pane_run NS "nsenter -t ${__ns_pid} -U -n --preserve-credentials" @@ -172,15 +173,18 @@ setup_two_guests() { # 10004 | as server | to init | to guest | to ns #2 # 10005 | | | as server | to ns #2 + __pid1_file="$(mktemp)" + __pid2_file="$(mktemp)" + __opts= [ ${PCAP} -eq 1 ] && __opts="${__opts} -p /tmp/pasta_1.pcap" [ ${DEBUG} -eq 1 ] && __opts="${__opts} -d" - pane_run PASST_1 "./pasta ${__opts} -t 10001,10002 -T 10003,10004 -u 10001,10002 -U 10003,10004" + pane_run PASST_1 "./pasta ${__opts} -P ${__pid1_file} -t 10001,10002 -T 10003,10004 -u 10001,10002 -U 10003,10004" __opts= [ ${PCAP} -eq 1 ] && __opts="${__opts} -p /tmp/pasta_2.pcap" [ ${DEBUG} -eq 1 ] && __opts="${__opts} -d" - pane_run PASST_2 "./pasta ${__opts} -t 10004,10005 -T 10003,10001 -u 10004,10005 -U 10003,10001" + pane_run PASST_2 "./pasta ${__opts} -P ${__pid2_file} -t 10004,10005 -T 10003,10001 -u 10004,10005 -U 10003,10001" sleep 1 pane_run PASST_1 '' @@ -188,12 +192,12 @@ setup_two_guests() { pane_wait PASST_1 pane_wait PASST_2 - pane_run PASST_1 'echo $$' - pane_run PASST_2 'echo $$' - pane_wait PASST_1 - pane_wait PASST_2 - __ns1_pid="$(pane_parse PASST_1)" - __ns2_pid="$(pane_parse PASST_2)" + __pasta1_pid="$(cat "${__pid1_file}")" + __ns1_pid="$(cat /proc/${__pasta1_pid}/task/${__pasta1_pid}/children | cut -f1 -d' ')" + rm "${__pid1_file}" + __pasta2_pid="$(cat "${__pid2_file}")" + __ns2_pid="$(cat /proc/${__pasta2_pid}/task/${__pasta2_pid}/children | cut -f1 -d' ')" + rm "${__pid2_file}" pane_run GUEST_1 "nsenter -t ${__ns1_pid} -U -n --preserve-credentials" pane_run GUEST_2 "nsenter -t ${__ns2_pid} -U -n --preserve-credentials" diff --git a/udp.c b/udp.c index e1a9ecb..348f695 100644 --- a/udp.c +++ b/udp.c @@ -529,7 +529,9 @@ static int udp_splice_connect_ns(void *arg) a = (struct udp_splice_connect_ns_arg *)arg; - ns_enter(a->c); + if (ns_enter(a->c)) + return 0; + a->s = udp_splice_connect(a->c, a->v6, a->bound_sock, a->src, a->dst, UDP_BACK_TO_INIT); @@ -1029,7 +1031,8 @@ int udp_sock_init_ns(void *arg) struct ctx *c = (struct ctx *)arg; int dst; - ns_enter(c); + if (ns_enter(c)) + return 0; for (dst = 0; dst < USHRT_MAX; dst++) { if (!bitmap_isset(c->udp.port_to_init, dst)) diff --git a/util.c b/util.c index 94d49a6..e9fca3b 100644 --- a/util.c +++ b/util.c @@ -16,6 +16,7 @@ #include #include #include +#include #include #include #include @@ -23,6 +24,7 @@ #include #include #include +#include #include #include #include @@ -32,6 +34,8 @@ #include #include +#include + #include "util.h" #include "passt.h" @@ -431,31 +435,51 @@ char *line_read(char *buf, size_t len, int fd) /** * procfs_scan_listen() - Set bits for listening TCP or UDP sockets from procfs - * @name: Corresponding name of file under /proc/net/ + * @proto: IPPROTO_TCP or IPPROTO_UDP + * @ip_version: IP version, V4 or V6 + * @ns: Use saved file descriptors for namespace if set * @map: Bitmap where numbers of ports in listening state will be set * @exclude: Bitmap of ports to exclude from setting (and clear) + * + * #syscalls:pasta lseek ppc64le:_llseek ppc64:_llseek */ -void procfs_scan_listen(char *name, uint8_t *map, uint8_t *exclude) +void procfs_scan_listen(struct ctx *c, uint8_t proto, int ip_version, int ns, + uint8_t *map, uint8_t *exclude) { - char line[BUFSIZ], path[PATH_MAX]; + char line[BUFSIZ], *path; unsigned long port; unsigned int state; - int fd; + int *fd; - snprintf(path, PATH_MAX, "/proc/net/%s", name); - if ((fd = open(path, O_RDONLY)) < 0) + if (proto == IPPROTO_TCP) { + fd = &c->proc_net_tcp[ip_version][ns]; + if (ip_version == V4) + path = "/proc/net/tcp"; + else + path = "/proc/net/tcp6"; + } else { + fd = &c->proc_net_udp[ip_version][ns]; + if (ip_version == V4) + path = "/proc/net/udp"; + else + path = "/proc/net/udp6"; + } + + if (*fd != -1) + lseek(*fd, 0, SEEK_SET); + else if ((*fd = open(path, O_RDONLY)) < 0) return; *line = 0; - line_read(line, sizeof(line), fd); - while (line_read(line, sizeof(line), fd)) { + line_read(line, sizeof(line), *fd); + while (line_read(line, sizeof(line), *fd)) { /* NOLINTNEXTLINE(cert-err34-c): != 2 if conversion fails */ if (sscanf(line, "%*u: %*x:%lx %*x:%*x %x", &port, &state) != 2) continue; /* See enum in kernel's include/net/tcp_states.h */ - if ((strstr(name, "tcp") && state != 0x0a) || - (strstr(name, "udp") && state != 0x07)) + if ((proto == IPPROTO_TCP && state != 0x0a) || + (proto == IPPROTO_UDP && state != 0x07)) continue; if (bitmap_isset(exclude, port)) @@ -463,25 +487,98 @@ void procfs_scan_listen(char *name, uint8_t *map, uint8_t *exclude) else bitmap_set(map, port); } +} - close(fd); +/** + * drop_caps() - Drop capabilities we might have except for CAP_NET_BIND_SERVICE + */ +void drop_caps(void) +{ + int i; + + for (i = 0; i < 64; i++) { + if (i == CAP_NET_BIND_SERVICE) + continue; + + prctl(PR_CAPBSET_DROP, i, 0, 0, 0); + } } /** - * ns_enter() - Enter configured network and user namespaces + * ns_enter() - Enter configured user (unless already joined) and network ns * @c: Execution context * - * Return: 0 on success, -1 on failure + * Return: 0, won't return on failure * * #syscalls:pasta setns */ int ns_enter(struct ctx *c) { - if (!c->netns_only && setns(c->pasta_userns_fd, CLONE_NEWUSER)) - return -errno; + if (!c->netns_only && + c->pasta_userns_fd != -1 && + setns(c->pasta_userns_fd, CLONE_NEWUSER)) + exit(EXIT_FAILURE); if (setns(c->pasta_netns_fd, CLONE_NEWNET)) - return -errno; + exit(EXIT_FAILURE); + + return 0; +} + +/** + * pid_file() - Write PID to file, if requested to do so, and close it + * @fd: Open PID file descriptor, closed on exit, -1 to skip writing it + * @pid: PID value to write + */ +void write_pidfile(int fd, pid_t pid) { + char pid_buf[12]; + int n; + + if (fd == -1) + return; + + n = snprintf(pid_buf, sizeof(pid_buf), "%i\n", pid); + + if (write(fd, pid_buf, n) < 0) { + perror("PID file write"); + exit(EXIT_FAILURE); + } + + close(fd); +} + +/** + * __daemon() - daemon()-like function writing PID file before parent exits + * @pidfile_fd: Open PID file descriptor + * @devnull_fd: Open file descriptor for /dev/null + * + * Return: child PID on success, won't return on failure + */ +int __daemon(int pidfile_fd, int devnull_fd) +{ + pid_t pid = fork(); + + if (pid == -1) { + perror("fork"); + exit(EXIT_FAILURE); + } + + if (pid) { + write_pidfile(pidfile_fd, pid); + exit(EXIT_SUCCESS); + } + + errno = 0; + + setsid(); + + dup2(devnull_fd, STDIN_FILENO); + dup2(devnull_fd, STDOUT_FILENO); + dup2(devnull_fd, STDERR_FILENO); + close(devnull_fd); + + if (errno) + exit(EXIT_FAILURE); return 0; } diff --git a/util.h b/util.h index add4c1e..b7852e9 100644 --- a/util.h +++ b/util.h @@ -54,6 +54,12 @@ void debug(const char *format, ...); #define STRINGIFY(x) #x #define STR(x) STRINGIFY(x) +#ifdef P_tmpdir +#define TMPDIR P_tmpdir +#else +#define TMPDIR "/tmp" +#endif + #define V4 0 #define V6 1 #define IP_VERSIONS 2 @@ -202,5 +208,9 @@ void bitmap_set(uint8_t *map, int bit); void bitmap_clear(uint8_t *map, int bit); int bitmap_isset(const uint8_t *map, int bit); char *line_read(char *buf, size_t len, int fd); -void procfs_scan_listen(char *name, uint8_t *map, uint8_t *exclude); +void procfs_scan_listen(struct ctx *c, uint8_t proto, int ip_version, int ns, + uint8_t *map, uint8_t *exclude); +void drop_caps(void); int ns_enter(struct ctx *c); +void write_pidfile(int fd, pid_t pid); +int __daemon(int pidfile_fd, int devnull_fd); -- 2.34.1

Nobody currently calls this as passt4netns, that was the name I used before 'pasta', drop any reference before it's too late. While at it, explicitly check that argc is bigger than or equal to one, just as a defensive measure: argv[0] being NULL is not an issue anyway. Signed-off-by: Stefano Brivio --- Makefile | 7 ++----- conf.c | 12 ++++++------ passt.c | 9 +++++++-- 3 files changed, 15 insertions(+), 13 deletions(-) diff --git a/Makefile b/Makefile index 5085578..8477cf0 100644 --- a/Makefile +++ b/Makefile @@ -62,7 +62,7 @@ endif prefix ?= /usr/local -all: passt pasta passt4netns qrap +all: passt pasta qrap avx2: CFLAGS += -Ofast -mavx2 -ftree-vectorize -funroll-loops avx2: clean all @@ -81,16 +81,13 @@ pasta: passt ln -s passt pasta ln -s passt.1 pasta.1 -passt4netns: passt - ln -s passt passt4netns - qrap: qrap.c passt.h $(CC) $(CFLAGS) \ qrap.c -o qrap .PHONY: clean clean: - -${RM} passt *.o seccomp.h qrap pasta pasta.1 passt4netns \ + -${RM} passt *.o seccomp.h qrap pasta pasta.1 \ passt.tar passt.tar.gz *.deb *.rpm install: passt pasta qrap diff --git a/conf.c b/conf.c index 732d918..2984ac2 100644 --- a/conf.c +++ b/conf.c @@ -532,7 +532,7 @@ static void conf_ip(struct ctx *c) */ static void usage(const char *name) { - if (strstr(name, "pasta") || strstr(name, "passt4netns")) { + if (strstr(name, "pasta")) { info("Usage: %s [OPTION]... [PID|PATH|NAME]", name); info(""); info("Without PID|PATH|NAME, run the default shell in a new"); @@ -550,7 +550,7 @@ static void usage(const char *name) info( " default: log to system logger only if started from a TTY"); info( " -h, --help Display this help message and exit"); - if (strstr(name, "pasta") || strstr(name, "passt4netns")) { + if (strstr(name, "pasta")) { info( " -I, --ns-ifname NAME namespace interface name"); info( " default: same interface name as external one"); } else { @@ -562,7 +562,7 @@ static void usage(const char *name) info( " -p, --pcap [FILE] Log tap-facing traffic to pcap file"); info( " if FILE is not given, log to:"); - if (strstr(name, "pasta") || strstr(name, "passt4netns")) + if (strstr(name, "pasta")) info(" /tmp/pasta_ISO8601-TIMESTAMP_PID.pcap"); else info(" /tmp/passt_ISO8601-TIMESTAMP_PID.pcap"); @@ -586,14 +586,14 @@ static void usage(const char *name) info( " -D, --dns ADDR Pass IPv4 or IPv6 address as DNS"); info( " can be specified multiple times"); info( " a single, empty option disables DNS information"); - if (strstr(name, "pasta") || strstr(name, "passt4netns")) + if (strstr(name, "pasta")) info( " default: don't send any addresses"); else info( " default: use addresses from /etc/resolv.conf"); info( " -S, --search LIST Space-separated list, search domains"); info( " a single, empty option disables the DNS search list"); - if (strstr(name, "pasta") || strstr(name, "passt4netns")) + if (strstr(name, "pasta")) info( " default: don't send any search list"); else info( " default: use search list from /etc/resolv.conf"); @@ -609,7 +609,7 @@ static void usage(const char *name) info( " -4, --ipv4-only Enable IPv4 operation only"); info( " -6, --ipv6-only Enable IPv6 operation only"); - if (strstr(name, "pasta") || strstr(name, "passt4netns")) + if (strstr(name, "pasta")) goto pasta_opts; info( " -t, --tcp-ports SPEC TCP port forwarding to guest"); diff --git a/passt.c b/passt.c index b5086d8..67ad1c7 100644 --- a/passt.c +++ b/passt.c @@ -322,16 +322,21 @@ int main(int argc, char **argv) sigaction(SIGTERM, &sa, NULL); sigaction(SIGQUIT, &sa, NULL); - if (strstr(argv[0], "pasta") || strstr(argv[0], "passt4netns")) { + if (argc < 1) + exit(EXIT_FAILURE); + + if (strstr(argv[0], "pasta")) { sa.sa_handler = pasta_child_handler; sigaction(SIGCHLD, &sa, NULL); signal(SIGPIPE, SIG_IGN); c.mode = MODE_PASTA; log_name = "pasta"; - } else { + } else if (strstr(argv[0], "passt")) { c.mode = MODE_PASST; log_name = "passt"; + } else { + exit(EXIT_FAILURE); } if (madvise(pkt_buf, TAP_BUF_BYTES, MADV_HUGEPAGE)) -- 2.34.1

Signed-off-by: Stefano Brivio --- conf.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/conf.c b/conf.c index 2984ac2..41895de 100644 --- a/conf.c +++ b/conf.c @@ -1239,5 +1239,6 @@ void conf(struct ctx *c, int argc, char **argv) } } - conf_print(c); + if (!c->quiet) + conf_print(c); } -- 2.34.1

For compatibility with libslirp/slirp4netns users: introduce a mechanism to map, in the UDP routines, an address facing guest or namespace to the first IPv4 or IPv6 address resulting from configuration as resolver. This can be enabled with the new --dns-forward option. This implies that sourcing and using DNS addresses and search lists, passed via command line or read from /etc/resolv.conf, is not bound anymore to DHCP/DHCPv6/NDP usage: for example, pasta users might just want to use addresses from /etc/resolv.conf as mapping target, while not passing DNS options via DHCP. Reflect this in all the involved code paths by differentiating DHCP/DHCPv6/NDP usage from DNS configuration per se, and in the new options --dhcp-dns, --dhcp-search for pasta, and --no-dhcp-dns, --no-dhcp-search for passt. This should be the last bit to enable substantial compatibility between slirp4netns.sh and slirp4netns(1): pass the --dns-forward option from the script too. Signed-off-by: Stefano Brivio --- conf.c | 102 +++++++++++++++++++++++++++++++++++++++---------- dhcp.c | 5 ++- dhcpv6.c | 7 ++++ ndp.c | 6 ++- passt.1 | 63 +++++++++++++++++++++++++----- passt.h | 14 +++++-- slirp4netns.sh | 2 +- udp.c | 16 ++++++++ 8 files changed, 177 insertions(+), 38 deletions(-) diff --git a/conf.c b/conf.c index 279fdfe..21e9bc0 100644 --- a/conf.c +++ b/conf.c @@ -279,7 +279,7 @@ static void get_dns(struct ctx *c) dns4_set = !c->v4 || !!*dns4; dns6_set = !c->v6 || !IN6_IS_ADDR_UNSPECIFIED(dns6); dnss_set = !!*s->n || c->no_dns_search; - dns_set = dns4_set || dns6_set || c->no_dns; + dns_set = (dns4_set && dns6_set) || c->no_dns; if (dns_set && dnss_set) return; @@ -583,21 +583,35 @@ static void usage(const char *name) info( " default: gateway from interface with default route"); info( " -i, --interface NAME Interface for addresses and routes"); info( " default: interface with first default route"); - info( " -D, --dns ADDR Pass IPv4 or IPv6 address as DNS"); + info( " -D, --dns ADDR Use IPv4 or IPv6 address as DNS"); info( " can be specified multiple times"); info( " a single, empty option disables DNS information"); if (strstr(name, "pasta")) - info( " default: don't send any addresses"); + info( " default: don't use any addresses"); else info( " default: use addresses from /etc/resolv.conf"); info( " -S, --search LIST Space-separated list, search domains"); info( " a single, empty option disables the DNS search list"); if (strstr(name, "pasta")) - info( " default: don't send any search list"); + info( " default: don't use any search list"); else info( " default: use search list from /etc/resolv.conf"); + if (strstr(name, "pasta")) + info(" --dhcp-dns: \tPass DNS list via DHCP/DHCPv6/NDP"); + else + info(" --no-dhcp-dns: No DNS list in DHCP/DHCPv6/NDP"); + + if (strstr(name, "pasta")) + info(" --dhcp-search: Pass list via DHCP/DHCPv6/NDP"); + else + info(" --no-dhcp-search: No list in DHCP/DHCPv6/NDP"); + + info( " --dns-forward ADDR Forward DNS queries sent to ADDR"); + info( " can be specified zero to two times (for IPv4 and IPv6)"); + info( " default: don't forward DNS queries"); + info( " --no-tcp Disable TCP protocol handler"); info( " --no-udp Disable UDP protocol handler"); info( " --no-icmp Disable ICMP/ICMPv6 protocol handler"); @@ -699,22 +713,18 @@ void conf_print(struct ctx *c) info(" router: %s", inet_ntop(AF_INET, &c->gw4, buf4, sizeof(buf4))); } - } - if (!c->no_dns && !(c->no_dhcp && c->no_ndp && c->no_dhcpv6)) { for (i = 0; c->dns4[i]; i++) { if (!i) - info(" DNS:"); + info("DNS:"); inet_ntop(AF_INET, &c->dns4[i], buf4, sizeof(buf4)); - info(" %s", buf4); + info(" %s", buf4); } - } - if (!c->no_dns_search && !(c->no_dhcp && c->no_ndp && c->no_dhcpv6)) { for (i = 0; *c->dns_search[i].n; i++) { if (!i) - info(" search:"); - info(" %s", c->dns_search[i].n); + info("DNS search list:"); + info(" %s", c->dns_search[i].n); } } @@ -728,7 +738,7 @@ void conf_print(struct ctx *c) else if (!c->no_dhcpv6) info("NDP:"); else - return; + goto dns6; info(" assign: %s", inet_ntop(AF_INET6, &c->addr6, buf6, sizeof(buf6))); @@ -737,17 +747,18 @@ void conf_print(struct ctx *c) info(" our link-local: %s", inet_ntop(AF_INET6, &c->addr6_ll, buf6, sizeof(buf6))); +dns6: for (i = 0; !IN6_IS_ADDR_UNSPECIFIED(&c->dns6[i]); i++) { if (!i) - info(" DNS:"); + info("DNS:"); inet_ntop(AF_INET6, &c->dns6[i], buf6, sizeof(buf6)); - info(" %s", buf6); + info(" %s", buf6); } for (i = 0; *c->dns_search[i].n; i++) { if (!i) - info(" search:"); - info(" %s", c->dns_search[i].n); + info("DNS search list:"); + info(" %s", c->dns_search[i].n); } } } @@ -797,6 +808,11 @@ void conf(struct ctx *c, int argc, char **argv) {"nsrun-dir", required_argument, NULL, 3 }, {"config-net", no_argument, &c->pasta_conf_ns, 1 }, {"ns-mac-addr", required_argument, NULL, 4 }, + {"dhcp-dns", no_argument, NULL, 5 }, + {"no-dhcp-dns", no_argument, NULL, 6 }, + {"dhcp-search", no_argument, NULL, 7 }, + {"no-dhcp-search", no_argument, NULL, 8 }, + {"dns-forward", required_argument, NULL, 9 }, { 0 }, }; struct get_bound_ports_ns_arg ns_ports_arg = { .c = c }; @@ -808,6 +824,9 @@ void conf(struct ctx *c, int argc, char **argv) int name, ret, mask, b, i; uint32_t *dns4 = c->dns4; + if (c->mode == MODE_PASTA) + c->no_dhcp_dns = c->no_dhcp_dns_search = 1; + do { enum conf_port_type *set = NULL; const char *optstring; @@ -873,6 +892,51 @@ void conf(struct ctx *c, int argc, char **argv) c->mac_guest[i] = b; } break; + case 5: + if (c->mode != MODE_PASTA) { + err("--dhcp-dns is for pasta mode only"); + usage(argv[0]); + } + c->no_dhcp_dns = 0; + break; + case 6: + if (c->mode != MODE_PASST) { + err("--no-dhcp-dns is for passt mode only"); + usage(argv[0]); + } + c->no_dhcp_dns = 1; + break; + case 7: + if (c->mode != MODE_PASTA) { + err("--dhcp-search is for pasta mode only"); + usage(argv[0]); + } + c->no_dhcp_dns_search = 0; + break; + case 8: + if (c->mode != MODE_PASST) { + err("--no-dhcp-search is for passt mode only"); + usage(argv[0]); + } + c->no_dhcp_dns_search = 1; + break; + case 9: + if (IN6_IS_ADDR_UNSPECIFIED(&c->dns6_fwd) && + inet_pton(AF_INET6, optarg, &c->dns6_fwd) && + !IN6_IS_ADDR_UNSPECIFIED(&c->dns6_fwd) && + !IN6_IS_ADDR_LOOPBACK(&c->dns6_fwd)) + break; + + if (c->dns4_fwd == INADDR_ANY && + inet_pton(AF_INET, optarg, &c->dns4_fwd) && + c->dns4_fwd != INADDR_ANY && + c->dns4_fwd != INADDR_BROADCAST && + c->dns4_fwd != INADDR_LOOPBACK) + break; + + err("Invalid DNS forwarding address: %s", optarg); + usage(argv[0]); + break; case 'd': if (c->debug) { err("Multiple --debug options given"); @@ -1189,10 +1253,6 @@ void conf(struct ctx *c, int argc, char **argv) if (!c->mtu) c->mtu = ROUND_DOWN(ETH_MAX_MTU - ETH_HLEN, sizeof(uint32_t)); - if (c->mode == MODE_PASTA && dns4 == c->dns4 && dns6 == c->dns6) - c->no_dns = 1; - if (c->mode == MODE_PASTA && dnss == c->dns_search) - c->no_dns_search = 1; get_dns(c); if (!*c->pasta_ifn) diff --git a/dhcp.c b/dhcp.c index a052397..ab1249c 100644 --- a/dhcp.c +++ b/dhcp.c @@ -333,12 +333,13 @@ int dhcp(struct ctx *c, struct ethhdr *eh, size_t len) opts[26].s[1] = c->mtu % 256; } - for (i = 0, opts[6].slen = 0; c->dns4[i]; i++) { + for (i = 0, opts[6].slen = 0; !c->no_dhcp_dns && c->dns4[i]; i++) { ((uint32_t *)opts[6].s)[i] = c->dns4[i]; opts[6].slen += sizeof(uint32_t); } - opt_set_dns_search(c, sizeof(m->o)); + if (!c->no_dhcp_dns_search) + opt_set_dns_search(c, sizeof(m->o)); uh->len = htons(len = offsetof(struct msg, o) + fill(m) + sizeof(*uh)); uh->check = 0; diff --git a/dhcpv6.c b/dhcpv6.c index e4113bc..b79a8e9 100644 --- a/dhcpv6.c +++ b/dhcpv6.c @@ -394,6 +394,9 @@ static size_t dhcpv6_dns_fill(struct ctx *c, char *buf, int offset) char *p = NULL; int i; + if (c->no_dhcp_dns) + goto search; + for (i = 0; !IN6_IS_ADDR_UNSPECIFIED(&c->dns6[i]); i++) { if (!i) { srv = (struct opt_dns_servers *)(buf + offset); @@ -410,6 +413,10 @@ static size_t dhcpv6_dns_fill(struct ctx *c, char *buf, int offset) if (srv) srv->hdr.l = htons(srv->hdr.l); +search: + if (c->no_dhcp_dns_search) + return offset; + for (i = 0; *c->dns_search[i].n; i++) { if (!i) { srch = (struct opt_dns_search *)(buf + offset); diff --git a/ndp.c b/ndp.c index 386098c..6b1c1a8 100644 --- a/ndp.c +++ b/ndp.c @@ -127,6 +127,9 @@ int ndp(struct ctx *c, struct ethhdr *eh, size_t len) p += 4; } + if (c->no_dhcp_dns) + goto dns_done; + for (n = 0; !IN6_IS_ADDR_UNSPECIFIED(&c->dns6[n]); n++); if (n) { *p++ = 25; /* RDNSS */ @@ -144,7 +147,7 @@ int ndp(struct ctx *c, struct ethhdr *eh, size_t len) dns_s_len += strlen(c->dns_search[n].n) + 2; } - if (dns_s_len) { + if (!c->no_dhcp_dns_search && dns_s_len) { *p++ = 31; /* DNSSL */ *p++ = (len + 8 - 1) / 8 + 1; /* length */ p += 2; /* reserved */ @@ -171,6 +174,7 @@ int ndp(struct ctx *c, struct ethhdr *eh, size_t len) p += 8 - dns_s_len % 8; } +dns_done: *p++ = 1; /* source ll */ *p++ = 1; /* length */ memcpy(p, c->mac, ETH_ALEN); diff --git a/passt.1 b/passt.1 index 92681f6..7070a31 100644 --- a/passt.1 +++ b/passt.1 @@ -165,19 +165,62 @@ Default is to use the interface with the first default route. .TP .BR \-D ", " \-\-dns " " \fIaddr -Assign IPv4 \fIaddr\fR via DHCP (option 23) or IPv6 \fIaddr\fR via NDP Router -Advertisement (option type 25) and DHCPv6 (option 23) as DNS resolver. +Use \fIaddr\fR (IPv4 or IPv6) for DHCP, DHCPv6, NDP or DNS forwarding, as +configured (see options \fB--no-dhcp-dns\fR, \fB--dhcp-dns\fR, +\fB--dns-forward\fR) instead of reading addresses from \fI/etc/resolv.conf\fR. This option can be specified multiple times, and a single, empty option disables -DNS options altogether. -In \fBpasst\fR mode, default is to use addresses from \fI/etc/resolv.conf\fR, -and, in \fBpasta\fR mode, no addresses are sent by default. +usage of DNS addresses altogether. + +.TP +.BR \-D ", " \-\-dns " " \fIaddr +Use \fIaddr\fR (IPv4 or IPv6) for DHCP, DHCPv6, NDP or DNS forwarding, as +configured (see options \fB--no-dhcp-dns\fR, \fB--dhcp-dns\fR, +\fB--dns-forward\fR) instead of reading addresses from \fI/etc/resolv.conf\fR. +This option can be specified multiple times, and a single, empty option disables +usage of DNS addresses altogether. + +.TP +.BR \-\-dns-forward " " \fIaddr +Map \fIaddr\fR (IPv4 or IPv6) as seen from guest or namespace to the first +configured DNS resolver (with corresponding IP version). Mapping is limited to +UDP traffic directed to port 53, and DNS answers are translated back with a +reverse mapping. +This option can be specified zero to two times (once for IPv4, once for IPv6). + .TP .BR \-S ", " \-\-search " " \fIlist -Assign space-separated \fIlist\fR via DHCP (option 119), via NDP Router -Advertisement (option type 31) and DHCPv6 (option 24) as DNS domain search list. -A single, empty option disables sending the DNS domain search list. -In \fBpasst\fR mode, default is to use the search list from -\fI/etc/resolv.conf\fR, and, in \fBpasta\fR mode, no list is sent by default. +Use space-separated \fIlist\fR for DHCP, DHCPv6, and NDP purposes, instead of +reading entries from \fI/etc/resolv.conf\fR. See options \fB--no-dhcp-search\fR +and \fB--dhcp-search\fR. A single, empty option disables the DNS domain search +list altogether. + +.TP +.BR \-\-no-dhcp-dns " " \fIaddr +In \fIpasst\fR mode, do not assign IPv4 addresses via DHCP (option 23) or IPv6 +addresses via NDP Router Advertisement (option type 25) and DHCPv6 (option 23) +as DNS resolvers. +By default, all the configured addresses are passed. + +.TP +.BR \-\-dhcp-dns " " \fIaddr +In \fIpasta\fR mode, assign IPv4 addresses via DHCP (option 23) or IPv6 +addresses via NDP Router Advertisement (option type 25) and DHCPv6 (option 23) +as DNS resolvers. +By default, configured addresses, if any, are not passed. + +.TP +.BR \-\-no-dhcp-search " " \fIaddr +In \fIpasst\fR mode, do not send the DNS domain search list addresses via DHCP +(option 119), via NDP Router Advertisement (option type 31) and DHCPv6 (option +24). +By default, the DNS domain search list resulting from configuration is passed. + +.TP +.BR \-\-dhcp-search " " \fIaddr +In \fIpasta\fR mode, send the DNS domain search list addresses via DHCP (option +119), via NDP Router Advertisement (option type 31) and DHCPv6 (option 24). +By default, the DNS domain search list resulting from configuration is not +passed. .TP .BR \-\-no-tcp diff --git a/passt.h b/passt.h index d7011da..2589ee7 100644 --- a/passt.h +++ b/passt.h @@ -114,6 +114,7 @@ enum passt_modes { * @mask4: IPv4 netmask, network order * @gw4: Default IPv4 gateway, network order * @dns4: IPv4 DNS addresses, zero-terminated, network order + * @dns4_fwd: Address forwarded (UDP) to first IPv4 DNS, network order * @dns_search: DNS search list * @v6: Enable IPv6 transport * @addr6: IPv6 address for external, routable interface @@ -121,7 +122,8 @@ enum passt_modes { * @addr6_seen: Latest IPv6 global/site address seen as source from tap * @addr6_ll_seen: Latest IPv6 link-local address seen as source from tap * @gw6: Default IPv6 gateway - * @dns4: IPv4 DNS addresses, zero-terminated + * @dns6: IPv6 DNS addresses, zero-terminated + * @dns6_fwd: Address forwarded (UDP) to first IPv6 DNS, network order * @ifi: Index of routable interface * @pasta_ifn: Name of namespace interface for pasta * @pasta_ifn: Index of namespace interface for pasta @@ -133,8 +135,10 @@ enum passt_modes { * @no_icmp: Disable ICMP operation * @icmp: Context for ICMP protocol handler * @mtu: MTU passed via DHCP/NDP - * @no_dns: Do not assign any DNS server via DHCP/DHCPv6/NDP - * @no_dns_search: Do not assign any DNS domain search via DHCP/DHCPv6/NDP + * @no_dns: Do not source/use DNS servers for any purpose + * @no_dns_search: Do not source/use domain search lists for any purpose + * @no_dhcp_dns: Do not assign any DNS server via DHCP/DHCPv6/NDP + * @no_dhcp_dns_search: Do not assign any DNS domain search via DHCP/DHCPv6/NDP * @no_dhcp: Disable DHCP server * @no_dhcpv6: Disable DHCPv6 server * @no_ndp: Disable NDP handler altogether @@ -172,6 +176,7 @@ struct ctx { uint32_t mask4; uint32_t gw4; uint32_t dns4[MAXNS + 1]; + uint32_t dns4_fwd; struct fqdn dns_search[MAXDNSRCH]; @@ -182,6 +187,7 @@ struct ctx { struct in6_addr addr6_ll_seen; struct in6_addr gw6; struct in6_addr dns6[MAXNS + 1]; + struct in6_addr dns6_fwd; unsigned int ifi; char pasta_ifn[IF_NAMESIZE]; @@ -198,6 +204,8 @@ struct ctx { int mtu; int no_dns; int no_dns_search; + int no_dhcp_dns; + int no_dhcp_dns_search; int no_dhcp; int no_dhcpv6; int no_ndp; diff --git a/slirp4netns.sh b/slirp4netns.sh index 1784926..ff12a52 100755 --- a/slirp4netns.sh +++ b/slirp4netns.sh @@ -16,7 +16,7 @@ # Author: Stefano Brivio PASTA_PID="$(mktemp)" -PASTA_OPTS="-q --ipv4-only -a 10.0.2.0 -n 24 -g 10.0.2.2 -m 1500 --no-ndp --no-dhcpv6 --no-dhcp -P ${PASTA_PID}" +PASTA_OPTS="-q --ipv4-only -a 10.0.2.0 -n 24 -g 10.0.2.2 --dns-forward 10.0.2.3 -m 1500 --no-ndp --no-dhcpv6 --no-dhcp -P ${PASTA_PID}" PASTA="$(command -v ./pasta || command -v pasta || :)" API_SOCKET= diff --git a/udp.c b/udp.c index 348f695..2fc52d3 100644 --- a/udp.c +++ b/udp.c @@ -718,6 +718,12 @@ void udp_sock_handler(struct ctx *c, union epoll_ref ref, uint32_t events, udp_tap_map[V6][src].loopback = 0; bitmap_set(udp_act[V6][UDP_ACT_TAP], src); + } else if (!IN6_IS_ADDR_UNSPECIFIED(&c->dns6_fwd) && + !memcmp(&b->s_in6.sin6_addr, &c->dns6_fwd, + sizeof(c->dns6_fwd)) && + ntohs(b->s_in6.sin6_port) == 53) { + b->ip6h.daddr = c->addr6_seen; + b->ip6h.saddr = c->dns6_fwd; } else { b->ip6h.daddr = c->addr6_seen; b->ip6h.saddr = b->s_in6.sin6_addr; @@ -797,6 +803,10 @@ void udp_sock_handler(struct ctx *c, union epoll_ref ref, uint32_t events, udp_tap_map[V4][src].loopback = 1; bitmap_set(udp_act[V4][UDP_ACT_TAP], src); + } else if (c->dns4_fwd && + s_addr == ntohl(c->dns4[0]) && + ntohs(b->s_in.sin_port) == 53) { + b->iph.saddr = c->dns4_fwd; } else { b->iph.saddr = b->s_in.sin_addr.s_addr; } @@ -958,6 +968,9 @@ int udp_tap_handler(struct ctx *c, int af, void *addr, s_in.sin_addr.s_addr = htonl(INADDR_LOOPBACK); else s_in.sin_addr.s_addr = c->addr4_seen; + } else if (s_in.sin_addr.s_addr == c->dns4_fwd && + ntohs(s_in.sin_port) == 53) { + s_in.sin_addr.s_addr = c->dns4[0]; } } else { s_in6 = (struct sockaddr_in6) { @@ -976,6 +989,9 @@ int udp_tap_handler(struct ctx *c, int af, void *addr, s_in6.sin6_addr = in6addr_loopback; else s_in6.sin6_addr = c->addr6_seen; + } else if (!memcmp(addr, &c->dns6_fwd, sizeof(c->dns6_fwd)) && + ntohs(s_in6.sin6_port) == 53) { + s_in6.sin6_addr = c->dns6[0]; } else if (IN6_IS_ADDR_LINKLOCAL(&s_in6.sin6_addr)) { bind_to = BIND_LL; } -- 2.34.1

In pasta mode, when we get data from sockets and write it as single frames to the tap device, we batch receive operations considerably, and then (conceptually) split the data in many smaller writes. It looked like an obvious choice, but performance is actually better if we receive data in many small frame-sized recvmsg()/recvmmsg(). The syscall overhead with the previous behaviour, observed by perf, comes predominantly from write operations, but receiving data in shorter chunks probably improves cache locality by a considerable amount. Signed-off-by: Stefano Brivio --- tcp.c | 36 ++++++++++++++++++++---------------- udp.c | 33 +++++++++++++++++---------------- 2 files changed, 37 insertions(+), 32 deletions(-) diff --git a/tcp.c b/tcp.c index e4fac22..a3a9dfd 100644 --- a/tcp.c +++ b/tcp.c @@ -343,7 +343,9 @@ #define MAX_TAP_CONNS (128 * 1024) #define MAX_SPLICE_CONNS (128 * 1024) -#define TCP_TAP_FRAMES 256 +#define TCP_TAP_FRAMES_MEM 256 +#define TCP_TAP_FRAMES \ + (c->mode == MODE_PASST ? TCP_TAP_FRAMES_MEM : 1) #define MAX_PIPE_SIZE (2UL * 1024 * 1024) @@ -609,7 +611,7 @@ static struct tcp4_l2_buf_t { #else } __attribute__ ((packed, aligned(__alignof__(unsigned int)))) #endif -tcp4_l2_buf[TCP_TAP_FRAMES]; +tcp4_l2_buf[TCP_TAP_FRAMES_MEM]; static unsigned int tcp4_l2_buf_used; static size_t tcp4_l2_buf_bytes; @@ -640,21 +642,21 @@ struct tcp6_l2_buf_t { #else } __attribute__ ((packed, aligned(__alignof__(unsigned int)))) #endif -tcp6_l2_buf[TCP_TAP_FRAMES]; +tcp6_l2_buf[TCP_TAP_FRAMES_MEM]; static unsigned int tcp6_l2_buf_used; static size_t tcp6_l2_buf_bytes; /* recvmsg()/sendmsg() data for tap */ static char tcp_buf_discard [MAX_WINDOW]; -static struct iovec iov_sock [TCP_TAP_FRAMES + 1]; +static struct iovec iov_sock [TCP_TAP_FRAMES_MEM + 1]; -static struct iovec tcp4_l2_iov_tap [TCP_TAP_FRAMES]; -static struct iovec tcp6_l2_iov_tap [TCP_TAP_FRAMES]; -static struct iovec tcp4_l2_flags_iov_tap [TCP_TAP_FRAMES]; -static struct iovec tcp6_l2_flags_iov_tap [TCP_TAP_FRAMES]; +static struct iovec tcp4_l2_iov_tap [TCP_TAP_FRAMES_MEM]; +static struct iovec tcp6_l2_iov_tap [TCP_TAP_FRAMES_MEM]; +static struct iovec tcp4_l2_flags_iov_tap [TCP_TAP_FRAMES_MEM]; +static struct iovec tcp6_l2_flags_iov_tap [TCP_TAP_FRAMES_MEM]; -static struct mmsghdr tcp_l2_mh_tap [TCP_TAP_FRAMES]; +static struct mmsghdr tcp_l2_mh_tap [TCP_TAP_FRAMES_MEM]; /* sendmsg() to socket */ static struct iovec tcp_tap_iov [UIO_MAXIOV]; @@ -688,7 +690,7 @@ static struct tcp4_l2_flags_buf_t { #else } __attribute__ ((packed, aligned(__alignof__(unsigned int)))) #endif -tcp4_l2_flags_buf[TCP_TAP_FRAMES]; +tcp4_l2_flags_buf[TCP_TAP_FRAMES_MEM]; static int tcp4_l2_flags_buf_used; @@ -717,7 +719,7 @@ static struct tcp6_l2_flags_buf_t { #else } __attribute__ ((packed, aligned(__alignof__(unsigned int)))) #endif -tcp6_l2_flags_buf[TCP_TAP_FRAMES]; +tcp6_l2_flags_buf[TCP_TAP_FRAMES_MEM]; static int tcp6_l2_flags_buf_used; @@ -916,7 +918,7 @@ void tcp_update_l2_buf(unsigned char *eth_d, unsigned char *eth_s, { int i; - for (i = 0; i < TCP_TAP_FRAMES; i++) { + for (i = 0; i < TCP_TAP_FRAMES_MEM; i++) { struct tcp4_l2_flags_buf_t *b4f = &tcp4_l2_flags_buf[i]; struct tcp6_l2_flags_buf_t *b6f = &tcp6_l2_flags_buf[i]; struct tcp4_l2_buf_t *b4 = &tcp4_l2_buf[i]; @@ -982,12 +984,13 @@ static void tcp_sock4_iov_init(void) }; } - for (i = 0, iov = tcp4_l2_iov_tap; i < TCP_TAP_FRAMES; i++, iov++) { + for (i = 0, iov = tcp4_l2_iov_tap; i < TCP_TAP_FRAMES_MEM; i++, iov++) { iov->iov_base = &tcp4_l2_buf[i].vnet_len; iov->iov_len = MSS_DEFAULT; } - for (i = 0, iov = tcp4_l2_flags_iov_tap; i < TCP_TAP_FRAMES; i++, iov++) + for (i = 0, iov = tcp4_l2_flags_iov_tap; i < TCP_TAP_FRAMES_MEM; + i++, iov++) iov->iov_base = &tcp4_l2_flags_buf[i].vnet_len; } @@ -1015,12 +1018,13 @@ static void tcp_sock6_iov_init(void) }; } - for (i = 0, iov = tcp6_l2_iov_tap; i < TCP_TAP_FRAMES; i++, iov++) { + for (i = 0, iov = tcp6_l2_iov_tap; i < TCP_TAP_FRAMES_MEM; i++, iov++) { iov->iov_base = &tcp6_l2_buf[i].vnet_len; iov->iov_len = MSS_DEFAULT; } - for (i = 0, iov = tcp6_l2_flags_iov_tap; i < TCP_TAP_FRAMES; i++, iov++) + for (i = 0, iov = tcp6_l2_flags_iov_tap; i < TCP_TAP_FRAMES_MEM; + i++, iov++) iov->iov_base = &tcp6_l2_flags_buf[i].vnet_len; } diff --git a/udp.c b/udp.c index 8129a89..d4f3714 100644 --- a/udp.c +++ b/udp.c @@ -118,7 +118,8 @@ #define UDP_CONN_TIMEOUT 180 /* s, timeout for ephemeral or local bind */ #define UDP_SPLICE_FRAMES 128 -#define UDP_TAP_FRAMES 128 +#define UDP_TAP_FRAMES_MEM 128 +#define UDP_TAP_FRAMES (c->mode == MODE_PASST ? UDP_TAP_FRAMES_MEM : 1) /** * struct udp_tap_port - Port tracking based on tap-facing source port @@ -204,7 +205,7 @@ static struct udp4_l2_buf_t { uint8_t data[USHRT_MAX - (sizeof(struct iphdr) + sizeof(struct udphdr))]; } __attribute__ ((packed, aligned(__alignof__(unsigned int)))) -udp4_l2_buf[UDP_TAP_FRAMES]; +udp4_l2_buf[UDP_TAP_FRAMES_MEM]; /** * udp6_l2_buf_t - Pre-cooked IPv6 packet buffers for tap connections @@ -234,23 +235,23 @@ struct udp6_l2_buf_t { #else } __attribute__ ((packed, aligned(__alignof__(unsigned int)))) #endif -udp6_l2_buf[UDP_TAP_FRAMES]; +udp6_l2_buf[UDP_TAP_FRAMES_MEM]; static struct sockaddr_storage udp_splice_namebuf; static uint8_t udp_splice_buf[UDP_SPLICE_FRAMES][USHRT_MAX]; /* recvmmsg()/sendmmsg() data for tap */ -static struct iovec udp4_l2_iov_sock [UDP_TAP_FRAMES]; -static struct iovec udp6_l2_iov_sock [UDP_TAP_FRAMES]; +static struct iovec udp4_l2_iov_sock [UDP_TAP_FRAMES_MEM]; +static struct iovec udp6_l2_iov_sock [UDP_TAP_FRAMES_MEM]; -static struct iovec udp4_l2_iov_tap [UDP_TAP_FRAMES]; -static struct iovec udp6_l2_iov_tap [UDP_TAP_FRAMES]; +static struct iovec udp4_l2_iov_tap [UDP_TAP_FRAMES_MEM]; +static struct iovec udp6_l2_iov_tap [UDP_TAP_FRAMES_MEM]; -static struct mmsghdr udp4_l2_mh_sock [UDP_TAP_FRAMES]; -static struct mmsghdr udp6_l2_mh_sock [UDP_TAP_FRAMES]; +static struct mmsghdr udp4_l2_mh_sock [UDP_TAP_FRAMES_MEM]; +static struct mmsghdr udp6_l2_mh_sock [UDP_TAP_FRAMES_MEM]; -static struct mmsghdr udp4_l2_mh_tap [UDP_TAP_FRAMES]; -static struct mmsghdr udp6_l2_mh_tap [UDP_TAP_FRAMES]; +static struct mmsghdr udp4_l2_mh_tap [UDP_TAP_FRAMES_MEM]; +static struct mmsghdr udp6_l2_mh_tap [UDP_TAP_FRAMES_MEM]; /* recvmmsg()/sendmmsg() data for "spliced" connections */ static struct iovec udp_splice_iov_recv [UDP_SPLICE_FRAMES]; @@ -310,7 +311,7 @@ void udp_update_l2_buf(unsigned char *eth_d, unsigned char *eth_s, { int i; - for (i = 0; i < UDP_TAP_FRAMES; i++) { + for (i = 0; i < UDP_TAP_FRAMES_MEM; i++) { struct udp4_l2_buf_t *b4 = &udp4_l2_buf[i]; struct udp6_l2_buf_t *b6 = &udp6_l2_buf[i]; @@ -354,7 +355,7 @@ static void udp_sock4_iov_init(void) }; } - for (i = 0, h = udp4_l2_mh_sock; i < UDP_TAP_FRAMES; i++, h++) { + for (i = 0, h = udp4_l2_mh_sock; i < UDP_TAP_FRAMES_MEM; i++, h++) { struct msghdr *mh = &h->msg_hdr; mh->msg_name = &udp4_l2_buf[i].s_in; @@ -366,7 +367,7 @@ static void udp_sock4_iov_init(void) mh->msg_iovlen = 1; } - for (i = 0, h = udp4_l2_mh_tap; i < UDP_TAP_FRAMES; i++, h++) { + for (i = 0, h = udp4_l2_mh_tap; i < UDP_TAP_FRAMES_MEM; i++, h++) { struct msghdr *mh = &h->msg_hdr; udp4_l2_iov_tap[i].iov_base = &udp4_l2_buf[i].vnet_len; @@ -394,7 +395,7 @@ static void udp_sock6_iov_init(void) }; } - for (i = 0, h = udp6_l2_mh_sock; i < UDP_TAP_FRAMES; i++, h++) { + for (i = 0, h = udp6_l2_mh_sock; i < UDP_TAP_FRAMES_MEM; i++, h++) { struct msghdr *mh = &h->msg_hdr; mh->msg_name = &udp6_l2_buf[i].s_in6; @@ -406,7 +407,7 @@ static void udp_sock6_iov_init(void) mh->msg_iovlen = 1; } - for (i = 0, h = udp6_l2_mh_tap; i < UDP_TAP_FRAMES; i++, h++) { + for (i = 0, h = udp6_l2_mh_tap; i < UDP_TAP_FRAMES_MEM; i++, h++) { struct msghdr *mh = &h->msg_hdr; udp6_l2_iov_tap[i].iov_base = &udp6_l2_buf[i].vnet_len; -- 2.34.1

Removing the needrestart package doesn't seem to work anymore, and I'm getting again prompts to restart services after installing gcc and make: export DEBIAN_FRONTEND=noninteractive before installing packages to avoid that. Signed-off-by: Stefano Brivio --- test/distro/ubuntu | 1 + 1 file changed, 1 insertion(+) diff --git a/test/distro/ubuntu b/test/distro/ubuntu index b67c1f3..781daab 100644 --- a/test/distro/ubuntu +++ b/test/distro/ubuntu @@ -187,6 +187,7 @@ host guestfish --rw -a __IMG__ -i copy-in __GUEST_FILES__ /root/ host ./qrap 5 qemu-system-s390x -m 2048 -smp 2 -serial stdio -nodefaults -nographic __IMG__ -net socket,fd=5 -net nic,model=virtio -device virtio-rng-ccw host service systemd-resolved stop +host export DEBIAN_FRONTEND=noninteractive host apt-get -y remove needrestart snapd host dhclient sleep 2 -- 2.34.1

Signed-off-by: Stefano Brivio --- passt.1 | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/passt.1 b/passt.1 index 485e1db..65b473b 100644 --- a/passt.1 +++ b/passt.1 @@ -1,5 +1,5 @@ .\" SPDX-License-Identifier: AGPL-3.0-or-later -.\" Copyright (c) 2020-2021 Red Hat GmbH +.\" Copyright (c) 2020-2022 Red Hat GmbH .\" Author: Stefano Brivio .TH passt 1 @@ -781,12 +781,13 @@ Stefano Brivio .SH REPORTING BUGS -No public bug tracker is available at this time. For the moment being, report -issues to Stefano Brivio . +Please report issues on the bug tracker at https://passt.top/passt/bugs, or +send a message to the passt-user@passt.top mailing list, see +https://passt.top/passt/lists. .SH COPYRIGHT -Copyright (c) 2020-2021 Red Hat GmbH. +Copyright (c) 2020-2022 Red Hat GmbH. \fBpasst\fR and \fBpasta\fR are free software: you can redistribute them and/or modify them under the terms of the GNU Affero General Public License as -- 2.34.1

The patch introduces a "pasta" networking mode for rootless container, similar to the existing slirp4netns mode. Notable differences are described in the commit message. Signed-off-by: Stefano Brivio --- ...001-libpod-Add-pasta-networking-mode.patch | 542 ++++++++++++++++++ 1 file changed, 542 insertions(+) create mode 100644 contrib/podman/0001-libpod-Add-pasta-networking-mode.patch diff --git a/contrib/podman/0001-libpod-Add-pasta-networking-mode.patch b/contrib/podman/0001-libpod-Add-pasta-networking-mode.patch new file mode 100644 index 0000000..98cb48b --- /dev/null +++ b/contrib/podman/0001-libpod-Add-pasta-networking-mode.patch @@ -0,0 +1,542 @@ +From bcfd618a316097e5d2e1a20703b11beeb21b6899 Mon Sep 17 00:00:00 2001 +From: Stefano Brivio +Date: Sat, 19 Feb 2022 04:54:09 +0100 +Subject: [PATCH] libpod: Add pasta networking mode + +Conceptually equivalent to networking by means of slirp4netns(1), +with a few practical differences: + +- pasta(1) forks to background once networking is configured in the + namespace and quits on its own once the namespace is deleted: + file descriptor synchronisation and PID tracking are not needed + +- port forwarding is configured via command line options at start-up, + instead of an API socket: this is taken care of right away as we're + about to start pasta + +- there's no need for further selection of port forwarding modes: + pasta behaves similarly to containers-rootlessport for local binds + (splice() instead of read()/write() pairs, without L2-L4 + translation), and keeps the original source address for non-local + connections like slirp4netns does + +- IPv6 is enabled by default, it's not an experimental feature. It + can be disabled using additional options as documented + +- by default, addresses and routes are copied from the host, that is, + container users will see the same IP address and routes as if they + were in the init namespace context. The interface name is also + sourced from the host upstream interface with the first default + route in the routing table. This is also configurable as documented + +- by default, the host is reachable using the gateway address from + the container, unless the --no-map-gw option is passed + +- sandboxing and seccomp(2) policies cannot be disabled + +See https://passt.top for more details about pasta. + +Signed-off-by: Stefano Brivio +--- +SPDX-FileCopyrightText: 2021-2022 Red Hat GmbH +SPDX-License-Identifier: Apache-2.0 + + docs/source/markdown/podman-create.1.md | 40 ++++++++++++- + docs/source/markdown/podman-pod-create.1.md | 33 +++++++++++ + docs/source/markdown/podman-run.1.md | 38 +++++++++++- + docs/source/markdown/podman.1.md | 6 +- + libpod/networking_linux.go | 6 +- + libpod/networking_pasta.go | 64 +++++++++++++++++++++ + pkg/namespaces/namespaces.go | 6 ++ + pkg/specgen/generate/namespaces.go | 10 ++++ + pkg/specgen/generate/pod_create.go | 6 ++ + pkg/specgen/namespaces.go | 18 +++++- + pkg/specgen/podspecgen.go | 2 +- + 11 files changed, 215 insertions(+), 14 deletions(-) + create mode 100644 libpod/networking_pasta.go + +diff --git a/docs/source/markdown/podman-create.1.md b/docs/source/markdown/podman-create.1.md +index 2a0f3b738..5cc03bff3 100644 +--- a/docs/source/markdown/podman-create.1.md ++++ b/docs/source/markdown/podman-create.1.md +@@ -699,12 +699,19 @@ Valid _mode_ values are: + - **interface_name**: Specify a name for the created network interface inside the container. + + For example to set a static ipv4 address and a static mac address, use `--network bridge:ip=10.88.0.10,mac=44:33:22:11:00:99`. ++ + - \[:OPTIONS,...]: Connect to a user-defined network; this is the network name or ID from a network created by **[podman network create](podman-network-create.1.md)**. Using the network name implies the bridge network mode. It is possible to specify the same options described under the bridge mode above. You can use the **--network** option multiple times to specify additional networks. ++ + - **none**: Create a network namespace for the container but do not configure network interfaces for it, thus the container has no network connectivity. ++ + - **container:**_id_: Reuse another container's network stack. ++ + - **host**: Do not create a network namespace, the container will use the host's network. Note: The host mode gives the container full access to local system services such as D-bus and is therefore considered insecure. ++ + - **ns:**_path_: Path to a network namespace to join. ++ + - **private**: Create a new namespace for the container. This will use the **bridge** mode for rootfull containers and **slirp4netns** for rootless ones. ++ + - **slirp4netns[:OPTIONS,...]**: use **slirp4netns**(1) to create a user network stack. This is the default for rootless containers. It is possible to specify these additional options: + - **allow_host_loopback=true|false**: Allow the slirp4netns to reach the host loopback IP (`10.0.2.2`, which is added to `/etc/hosts` as `host.containers.internal` for your convenience). Default is false. + - **mtu=MTU**: Specify the MTU to use for this network. (Default is `65520`). +@@ -718,6 +725,30 @@ Valid _mode_ values are: + Note: Rootlesskit changes the source IP address of incoming packets to an IP address in the container network namespace, usually `10.0.2.100`. If your application requires the real source IP address, e.g. web server logs, use the slirp4netns port handler. The rootlesskit port handler is also used for rootless containers when connected to user-defined networks. + - **port_handler=slirp4netns**: Use the slirp4netns port forwarding, it is slower than rootlesskit but preserves the correct source IP address. This port handler cannot be used for user-defined networks. + ++- **pasta[:OPTIONS,...]**: use **pasta**(1) to create a user-mode networking ++stack. By default, IPv4 and IPv6 addresses and routes, as well as the pod ++interface name, are copied from the host. If port forwarding isn't configured, ++ports will be forwarded dynamically as services are bound on either side (init ++namespace or container namespace). Port forwarding preserves the original source ++IP address. Options described in pasta(1) can be specified as comma-separated ++arguments. In terms of pasta(1) options, only **--config-net** is given by ++default, in order to configure networking when the container is started. Some ++examples: ++ - **pasta:--no-map-gw**: Don't allow the container to directly reach the host ++ using the gateway address, which would normally be mapped to a loopback or ++ link-local address. ++ - **pasta:--mtu,1500**: Specify a 1500 bytes MTU for the _tap_ interface in ++ the container. ++ - **pasta:--ipv4-only,-a,10.0.2.0,-n,24,-g,10.0.2.2,--dns-forward,10.0.2.3,-m,1500,--no-ndp,--no-dhcpv6,--no-dhcp**, ++ equivalent to default slirp4netns(1) options: disable IPv6, assign ++ `10.0.2.0/24` to the `tap0` interface in the container, with gateway ++ `10.0.2.3`, enable DNS forwarder reachable at `10.0.2.3`, set MTU to 1500 ++ bytes, disable NDP, DHCPv6 and DHCP support. ++ - **pasta:--no-map-gw,-I,tap0,--ipv4-only,-a,10.0.2.0,-n,24,-g,10.0.2.2,--dns-forward,10.0.2.3,--no-ndp,--no-dhcpv6,--no-dhcp**, ++ equivalent to default slirp4netns(1) options with Podman overrides: same as ++ above, but leave the MTU to 65520 bytes, and don't map the gateway address ++ from the container to a local address. ++ + #### **--network-alias**=*alias* + + Add a network-scoped alias for the container, setting the alias for all networks that the container joins. To set a name only for a specific network, use the alias option as described under the **--network** option. +@@ -1551,8 +1582,9 @@ In order for users to run rootless, there must be an entry for their username in + + Rootless Podman works better if the fuse-overlayfs and slirp4netns packages are installed. + The fuse-overlayfs package provides a userspace overlay storage driver, otherwise users need to use +-the vfs storage driver, which is diskspace expensive and does not perform well. slirp4netns is +-required for VPN, without it containers need to be run with the --network=host flag. ++the vfs storage driver, which is diskspace expensive and does not perform well. ++slirp4netns or pasta are required for VPN, without it containers need to be run ++with the --network=host flag. + + ## ENVIRONMENT + +@@ -1601,7 +1633,9 @@ page. + NOTE: Use the environment variable `TMPDIR` to change the temporary storage location of downloaded container images. Podman defaults to use `/var/tmp`. + + ## SEE ALSO +-**[podman(1)](podman.1.md)**, **[podman-save(1)](podman-save.1.md)**, **[podman-ps(1)](podman-ps.1.md)**, **[podman-attach(1)](podman-attach.1.md)**, **[podman-pod-create(1)](podman-pod-create.1.md)**, **[podman-port(1)](podman-port.1.md)**, **[podman-start(1)](podman-start.1.md)**, **[podman-kill(1)](podman-kill.1.md)**, **[podman-stop(1)](podman-stop.1.md)**, **[podman-generate-systemd(1)](podman-generate-systemd.1.md)**, **[podman-rm(1)](podman-rm.1.md)**, **[subgid(5)](https://www.unix.com/man-page/linux/5/subgid)**, **[subuid(5)](https://www.unix.com/man-page/linux/5/subuid)**, **[containers.conf(5)](https://github.com/containers/common/blob/main/docs/containers.conf.5.md)**, **[systemd.unit(5)](https://www.freedesktop.org/software/systemd/man/systemd.unit.html)**, **[setsebool(8)](https://man7.org/linux/man-pages/man8/setsebool.8.html)**, **[slirp4netns(1)](https://github.com/rootless-containers/slirp4netns/blob/master/slirp4netns.1..., **[fuse-overlayfs(1)](https://github.com/containers/fuse-overlayfs/blob/main/fuse-overlayfs.1.md)*..., **proc(5)**, **[conmon(8)](https://github.com/containers/conmon/blob/main/docs/conmon.8.md)**, **personality(2)** ++**[podman(1)](podman.1.md)**, **[podman-save(1)](podman-save.1.md)**, **[podman-ps(1)](podman-ps.1.md)**, **[podman-attach(1)](podman-attach.1.md)**, **[podman-pod-create(1)](podman-pod-create.1.md)**, **[podman-port(1)](podman-port.1.md)**, **[podman-start(1)](podman-start.1.md)**, **[podman-kill(1)](podman-kill.1.md)**, **[podman-stop(1)](podman-stop.1.md)**, **[podman-generate-systemd(1)](podman-generate-systemd.1.md)**, **[podman-rm(1)](podman-rm.1.md)**, **[subgid(5)](https://www.unix.com/man-page/linux/5/subgid)**, **[subuid(5)](https://www.unix.com/man-page/linux/5/subuid)**, **[containers.conf(5)](https://github.com/containers/common/blob/main/docs/containers.conf.5.md)**, **[systemd.unit(5)](https://www.freedesktop.org/software/systemd/man/systemd.unit.html)**, **[setsebool(8)](https://man7.org/linux/man-pages/man8/setsebool.8.html)**, **[slirp4netns(1)](https://github.com/rootless-containers/slirp4netns/blob/master/slirp4netns.1..., ++**[pasta(1)](https://passt.top/builds/latest/web/passt.1.html)**, ++**[fuse-overlayfs(1)](https://github.com/containers/fuse-overlayfs/blob/main/fuse-overlayfs.1.md)*..., **proc(5)**, **[conmon(8)](https://github.com/containers/conmon/blob/main/docs/conmon.8.md)**, **personality(2)** + + ## HISTORY + October 2017, converted from Docker documentation to Podman by Dan Walsh for Podman `` +diff --git a/docs/source/markdown/podman-pod-create.1.md b/docs/source/markdown/podman-pod-create.1.md +index 8088e1d62..c94ac6061 100644 +--- a/docs/source/markdown/podman-pod-create.1.md ++++ b/docs/source/markdown/podman-pod-create.1.md +@@ -175,12 +175,19 @@ Valid _mode_ values are: + - **interface_name**: Specify a name for the created network interface inside the container. + + For example to set a static ipv4 address and a static mac address, use `--network bridge:ip=10.88.0.10,mac=44:33:22:11:00:99`. ++ + - \[:OPTIONS,...]: Connect to a user-defined network; this is the network name or ID from a network created by **[podman network create](podman-network-create.1.md)**. Using the network name implies the bridge network mode. It is possible to specify the same options described under the bridge mode above. You can use the **--network** option multiple times to specify additional networks. ++ + - **none**: Create a network namespace for the container but do not configure network interfaces for it, thus the container has no network connectivity. ++ + - **container:**_id_: Reuse another container's network stack. ++ + - **host**: Do not create a network namespace, the container will use the host's network. Note: The host mode gives the container full access to local system services such as D-bus and is therefore considered insecure. ++ + - **ns:**_path_: Path to a network namespace to join. ++ + - **private**: Create a new namespace for the container. This will use the **bridge** mode for rootfull containers and **slirp4netns** for rootless ones. ++ + - **slirp4netns[:OPTIONS,...]**: use **slirp4netns**(1) to create a user network stack. This is the default for rootless containers. It is possible to specify these additional options: + - **allow_host_loopback=true|false**: Allow the slirp4netns to reach the host loopback IP (`10.0.2.2`, which is added to `/etc/hosts` as `host.containers.internal` for your convenience). Default is false. + - **mtu=MTU**: Specify the MTU to use for this network. (Default is `65520`). +@@ -194,6 +201,30 @@ Valid _mode_ values are: + Note: Rootlesskit changes the source IP address of incoming packets to an IP address in the container network namespace, usually `10.0.2.100`. If your application requires the real source IP address, e.g. web server logs, use the slirp4netns port handler. The rootlesskit port handler is also used for rootless containers when connected to user-defined networks. + - **port_handler=slirp4netns**: Use the slirp4netns port forwarding, it is slower than rootlesskit but preserves the correct source IP address. This port handler cannot be used for user-defined networks. + ++- **pasta[:OPTIONS,...]**: use **pasta**(1) to create a user-mode networking ++stack. By default, IPv4 and IPv6 addresses and routes, as well as the pod ++interface name, are copied from the host. If port forwarding isn't configured, ++ports will be forwarded dynamically as services are bound on either side (init ++namespace or container namespace). Port forwarding preserves the original source ++IP address. Options described in pasta(1) can be specified as comma-separated ++arguments. In terms of pasta(1) options, only **--config-net** is given by ++default, in order to configure networking when the container is started. Some ++examples: ++ - **pasta:--no-map-gw**: Don't allow the container to directly reach the host ++ using the gateway address, which would normally be mapped to a loopback or ++ link-local address. ++ - **pasta:--mtu,1500**: Specify a 1500 bytes MTU for the _tap_ interface in ++ the container. ++ - **pasta:--ipv4-only,-a,10.0.2.0,-n,24,-g,10.0.2.2,--dns-forward,10.0.2.3,-m,1500,--no-ndp,--no-dhcpv6,--no-dhcp**, ++ equivalent to default slirp4netns(1) options: disable IPv6, assign ++ `10.0.2.0/24` to the `tap0` interface in the container, with gateway ++ `10.0.2.3`, enable DNS forwarder reachable at `10.0.2.3`, set MTU to 1500 ++ bytes, disable NDP, DHCPv6 and DHCP support. ++ - **pasta:--no-map-gw,-I,tap0,--ipv4-only,-a,10.0.2.0,-n,24,-g,10.0.2.2,--dns-forward,10.0.2.3,--no-ndp,--no-dhcpv6,--no-dhcp**, ++ equivalent to default slirp4netns(1) options with Podman overrides: same as ++ above, but leave the MTU to 65520 bytes, and don't map the gateway address ++ from the container to a local address. ++ + #### **--network-alias**=*alias* + + Add a network-scoped alias for the pod, setting the alias for all networks that the pod joins. To set a name only for a specific network, use the alias option as described under the **--network** option. +@@ -527,6 +558,8 @@ $ podman pod create --network slirp4netns:outbound_addr=127.0.0.1,allow_host_loo + + $ podman pod create --network slirp4netns:cidr=192.168.0.0/24 + ++$ podman pod create --network pasta ++ + $ podman pod create --network net1:ip=10.89.1.5 --network net2:ip=10.89.10.10 + ``` + +diff --git a/docs/source/markdown/podman-run.1.md b/docs/source/markdown/podman-run.1.md +index 239cf3b83..7c12f5e88 100644 +--- a/docs/source/markdown/podman-run.1.md ++++ b/docs/source/markdown/podman-run.1.md +@@ -726,12 +726,19 @@ Valid _mode_ values are: + - **interface_name**: Specify a name for the created network interface inside the container. + + For example to set a static ipv4 address and a static mac address, use `--network bridge:ip=10.88.0.10,mac=44:33:22:11:00:99`. ++ + - \[:OPTIONS,...]: Connect to a user-defined network; this is the network name or ID from a network created by **[podman network create](podman-network-create.1.md)**. Using the network name implies the bridge network mode. It is possible to specify the same options described under the bridge mode above. You can use the **--network** option multiple times to specify additional networks. ++ + - **none**: Create a network namespace for the container but do not configure network interfaces for it, thus the container has no network connectivity. ++ + - **container:**_id_: Reuse another container's network stack. ++ + - **host**: Do not create a network namespace, the container will use the host's network. Note: The host mode gives the container full access to local system services such as D-bus and is therefore considered insecure. ++ + - **ns:**_path_: Path to a network namespace to join. ++ + - **private**: Create a new namespace for the container. This will use the **bridge** mode for rootfull containers and **slirp4netns** for rootless ones. ++ + - **slirp4netns[:OPTIONS,...]**: use **slirp4netns**(1) to create a user network stack. This is the default for rootless containers. It is possible to specify these additional options: + - **allow_host_loopback=true|false**: Allow the slirp4netns to reach the host loopback IP (`10.0.2.2`, which is added to `/etc/hosts` as `host.containers.internal` for your convenience). Default is false. + - **mtu=MTU**: Specify the MTU to use for this network. (Default is `65520`). +@@ -745,6 +752,30 @@ Valid _mode_ values are: + Note: Rootlesskit changes the source IP address of incoming packets to an IP address in the container network namespace, usually `10.0.2.100`. If your application requires the real source IP address, e.g. web server logs, use the slirp4netns port handler. The rootlesskit port handler is also used for rootless containers when connected to user-defined networks. + - **port_handler=slirp4netns**: Use the slirp4netns port forwarding, it is slower than rootlesskit but preserves the correct source IP address. This port handler cannot be used for user-defined networks. + ++- **pasta[:OPTIONS,...]**: use **pasta**(1) to create a user-mode networking ++stack. By default, IPv4 and IPv6 addresses and routes, as well as the pod ++interface name, are copied from the host. If port forwarding isn't configured, ++ports will be forwarded dynamically as services are bound on either side (init ++namespace or container namespace). Port forwarding preserves the original source ++IP address. Options described in pasta(1) can be specified as comma-separated ++arguments. In terms of pasta(1) options, only **--config-net** is given by ++default, in order to configure networking when the container is started. Some ++examples: ++ - **pasta:--no-map-gw**: Don't allow the container to directly reach the host ++ using the gateway address, which would normally be mapped to a loopback or ++ link-local address. ++ - **pasta:--mtu,1500**: Specify a 1500 bytes MTU for the _tap_ interface in ++ the container. ++ - **pasta:--ipv4-only,-a,10.0.2.0,-n,24,-g,10.0.2.2,--dns-forward,10.0.2.3,-m,1500,--no-ndp,--no-dhcpv6,--no-dhcp**, ++ equivalent to default slirp4netns(1) options: disable IPv6, assign ++ `10.0.2.0/24` to the `tap0` interface in the container, with gateway ++ `10.0.2.3`, enable DNS forwarder reachable at `10.0.2.3`, set MTU to 1500 ++ bytes, disable NDP, DHCPv6 and DHCP support. ++ - **pasta:--no-map-gw,-I,tap0,--ipv4-only,-a,10.0.2.0,-n,24,-g,10.0.2.2,--dns-forward,10.0.2.3,--no-ndp,--no-dhcpv6,--no-dhcp**, ++ equivalent to default slirp4netns(1) options with Podman overrides: same as ++ above, but leave the MTU to 65520 bytes, and don't map the gateway address ++ from the container to a local address. ++ + #### **--network-alias**=*alias* + + Add a network-scoped alias for the container, setting the alias for all networks that the container joins. To set a name only for a specific network, use the alias option as described under the **--network** option. +@@ -1935,8 +1966,9 @@ In order for users to run rootless, there must be an entry for their username in + + Rootless Podman works better if the fuse-overlayfs and slirp4netns packages are installed. + The **fuse-overlayfs** package provides a userspace overlay storage driver, otherwise users need to use +-the **vfs** storage driver, which is diskspace expensive and does not perform well. slirp4netns is +-required for VPN, without it containers need to be run with the **--network=host** flag. ++the **vfs** storage driver, which is diskspace expensive and does not perform ++well. slirp4netns or pasta are required for VPN, without it containers need to ++be run with the **--network=host** flag. + + ## ENVIRONMENT + +@@ -1983,7 +2015,7 @@ page. + NOTE: Use the environment variable `TMPDIR` to change the temporary storage location of downloaded container images. Podman defaults to use `/var/tmp`. + + ## SEE ALSO +-**[podman(1)](podman.1.md)**, **[podman-save(1)](podman-save.1.md)**, **[podman-ps(1)](podman-ps.1.md)**, **[podman-attach(1)](podman-attach.1.md)**, **[podman-pod-create(1)](podman-pod-create.1.md)**, **[podman-port(1)](podman-port.1.md)**, **[podman-start(1)](podman-start.1.md)**, **[podman-kill(1)](podman-kill.1.md)**, **[podman-stop(1)](podman-stop.1.md)**, **[podman-generate-systemd(1)](podman-generate-systemd.1.md)**, **[podman-rm(1)](podman-rm.1.md)**, **[subgid(5)](https://www.unix.com/man-page/linux/5/subgid)**, **[subuid(5)](https://www.unix.com/man-page/linux/5/subuid)**, **[containers.conf(5)](https://github.com/containers/common/blob/main/docs/containers.conf.5.md)**, **[systemd.unit(5)](https://www.freedesktop.org/software/systemd/man/systemd.unit.html)**, **[setsebool(8)](https://man7.org/linux/man-pages/man8/setsebool.8.html)**, **[slirp4netns(1)](https://github.com/rootless-containers/slirp4netns/blob/master/slirp4netns.1..., **[fuse-overlayfs(1)](https://github.com/containers/fuse-overlayfs/blob/main/fuse-overlayfs.1.md)*..., **proc(5)**, **[conmon(8)](https://github.com/containers/conmon/blob/main/docs/conmon.8.md)**, **personality(2)** ++**[podman(1)](podman.1.md)**, **[podman-save(1)](podman-save.1.md)**, **[podman-ps(1)](podman-ps.1.md)**, **[podman-attach(1)](podman-attach.1.md)**, **[podman-pod-create(1)](podman-pod-create.1.md)**, **[podman-port(1)](podman-port.1.md)**, **[podman-start(1)](podman-start.1.md)**, **[podman-kill(1)](podman-kill.1.md)**, **[podman-stop(1)](podman-stop.1.md)**, **[podman-generate-systemd(1)](podman-generate-systemd.1.md)**, **[podman-rm(1)](podman-rm.1.md)**, **[subgid(5)](https://www.unix.com/man-page/linux/5/subgid)**, **[subuid(5)](https://www.unix.com/man-page/linux/5/subuid)**, **[containers.conf(5)](https://github.com/containers/common/blob/main/docs/containers.conf.5.md)**, **[systemd.unit(5)](https://www.freedesktop.org/software/systemd/man/systemd.unit.html)**, **[setsebool(8)](https://man7.org/linux/man-pages/man8/setsebool.8.html)**, **[slirp4netns(1)](https://github.com/rootless-containers/slirp4netns/blob/master/slirp4netns.1..., **[pasta(1)](https://passt.top/builds/latest/web/passt.1.html)**, **[fuse-overlayfs(1)](https://github.com/containers/fuse-overlayfs/blob/main/fuse-overlayfs.1.md)*..., **proc(5)**, **[conmon(8)](https://github.com/containers/conmon/blob/main/docs/conmon.8.md)**, **personality(2)** + + ## HISTORY + September 2018, updated by Kunal Kushwaha `` +diff --git a/docs/source/markdown/podman.1.md b/docs/source/markdown/podman.1.md +index b318001e4..1ad808cba 100644 +--- a/docs/source/markdown/podman.1.md ++++ b/docs/source/markdown/podman.1.md +@@ -95,7 +95,7 @@ Set libpod namespace. Namespaces are used to separate groups of containers and p + When namespace is set, created containers and pods will join the given namespace, and only containers and pods in the given namespace will be visible to Podman. + + #### **--network-cmd-path**=*path* +-Path to the command binary to use for setting up a network. It is currently only used for setting up a slirp4netns network. If "" is used then the binary is looked up using the $PATH environment variable. ++Path to the command binary to use for setting up a network. It is currently only used for setting up a slirp4netns(1) or pasta(1) network. If "" is used then the binary is looked up using the $PATH environment variable. + + #### **--noout** + +@@ -409,7 +409,7 @@ See the `subuid(5)` and `subgid(5)` man pages for more information. + + Images are pulled under `XDG_DATA_HOME` when specified, otherwise in the home directory of the user under `.local/share/containers/storage`. + +-Currently the slirp4netns package is required to be installed to create a network device, otherwise rootless containers need to run in the network namespace of the host. ++Currently either slirp4netns or pasta are required to be installed to create a network device, otherwise rootless containers need to run in the network namespace of the host. + + In certain environments like HPC (High Performance Computing), users cannot take advantage of the additional UIDs and GIDs from the /etc/subuid and /etc/subgid systems. However, in this environment, rootless Podman can operate with a single UID. To make this work, set the `ignore_chown_errors` option in the /etc/containers/storage.conf or in ~/.config/containers/storage.conf files. This option tells Podman when pulling an image to ignore chown errors when attempting to change a file in a container image to match the non-root UID in the image. This means all files get saved as the user's UID. Note this could cause issues when running the container. + +@@ -422,7 +422,7 @@ The Network File System (NFS) and other distributed file systems (for example: L + For more information, please refer to the [Podman Troubleshooting Page](https://github.com/containers/podman/blob/main/troubleshooting.md). + + ## SEE ALSO +-**[containers-mounts.conf(5)](https://github.com/containers/common/blob/main/docs/containers-mounts.conf.5..., **[containers.conf(5)](https://github.com/containers/common/blob/main/docs/containers.conf.5.md)**, **[containers-registries.conf(5)](https://github.com/containers/image/blob/main/docs/containers-registries.con..., **[containers-storage.conf(5)](https://github.com/containers/storage/blob/main/docs/containers-storage.conf..., **[buildah(1)](https://github.com/containers/buildah/blob/main/docs/buildah.1.md)**, **oci-hooks(5)**, **[containers-policy.json(5)](https://github.com/containers/image/blob/main/docs/containers-policy.json.5...., **[crun(1)](https://github.com/containers/crun/blob/main/crun.1.md)**, **[runc(8)](https://github.com/opencontainers/runc/blob/master/man/runc.8.md)**, **[subuid(5)](https://www.unix.com/man-page/linux/5/subuid)**, **[subgid(5)](https://www.unix.com/man-page/linux/5/subgid)**, **[slirp4netns(1)](https://github.com/rootless-containers/slirp4netns/blob/master/slirp4netns.1..., **[conmon(8)](https://github.com/containers/conmon/blob/main/docs/conmon.8.md)** ++**[containers-mounts.conf(5)](https://github.com/containers/common/blob/main/docs/containers-mounts.conf.5..., **[containers.conf(5)](https://github.com/containers/common/blob/main/docs/containers.conf.5.md)**, **[containers-registries.conf(5)](https://github.com/containers/image/blob/main/docs/containers-registries.con..., **[containers-storage.conf(5)](https://github.com/containers/storage/blob/main/docs/containers-storage.conf..., **[buildah(1)](https://github.com/containers/buildah/blob/main/docs/buildah.1.md)**, **oci-hooks(5)**, **[containers-policy.json(5)](https://github.com/containers/image/blob/main/docs/containers-policy.json.5...., **[crun(1)](https://github.com/containers/crun/blob/main/crun.1.md)**, **[runc(8)](https://github.com/opencontainers/runc/blob/master/man/runc.8.md)**, **[subuid(5)](https://www.unix.com/man-page/linux/5/subuid)**, **[subgid(5)](https://www.unix.com/man-page/linux/5/subgid)**, **[slirp4netns(1)](https://github.com/rootless-containers/slirp4netns/blob/master/slirp4netns.1..., **[pasta(1)](https://passt.top/builds/latest/web/passt.1.html)**, **[conmon(8)](https://github.com/containers/conmon/blob/main/docs/conmon.8.md)** + + ## HISTORY + Dec 2016, Originally compiled by Dan Walsh +diff --git a/libpod/networking_linux.go b/libpod/networking_linux.go +index 19d5c7f76..183f815ba 100644 +--- a/libpod/networking_linux.go ++++ b/libpod/networking_linux.go +@@ -636,6 +636,9 @@ func (r *Runtime) configureNetNS(ctr *Container, ctrNS ns.NetNS) (status map[str + if ctr.config.NetMode.IsSlirp4netns() { + return nil, r.setupSlirp4netns(ctr, ctrNS) + } ++ if ctr.config.NetMode.IsPasta() { ++ return nil, r.setupPasta(ctr, ctrNS) ++ } + networks, err := ctr.networks() + if err != nil { + return nil, err +@@ -806,7 +809,8 @@ func (r *Runtime) teardownCNI(ctr *Container) error { + return err + } + +- if !ctr.config.NetMode.IsSlirp4netns() && len(networks) > 0 { ++ if !ctr.config.NetMode.IsSlirp4netns() && ++ !ctr.config.NetMode.IsPasta() && len(networks) > 0 { + netOpts, err := ctr.getNetworkOptions(networks) + if err != nil { + return err +diff --git a/libpod/networking_pasta.go b/libpod/networking_pasta.go +new file mode 100644 +index 000000000..71595c87c +--- /dev/null ++++ b/libpod/networking_pasta.go +@@ -0,0 +1,64 @@ ++// SPDX-License-Identifier: Apache-2.0 ++// ++// networking_pasta.go - Start pasta(1) to provide connectivity to the container ++// ++// Copyright (c) 2022 Red Hat GmbH ++// Author: Stefano Brivio ++// ++// +build linux ++ ++package libpod ++ ++import ( ++ "os/exec" ++ "fmt" ++ "strings" ++ ++ "github.com/containernetworking/plugins/pkg/ns" ++ "github.com/pkg/errors" ++ "github.com/sirupsen/logrus" ++) ++ ++func (r *Runtime) setupPasta(ctr *Container, netns ns.NetNS) error { ++ path := r.config.Engine.NetworkCmdPath ++ if path == "" { ++ var err error ++ path, err = exec.LookPath("pasta") ++ if err != nil { ++ logrus.Errorf("Could not find pasta, the network namespace won't be configured: %v", err) ++ return nil ++ } ++ } ++ ++ cmdArgs := []string{} ++ cmdArgs = append(cmdArgs, "--config-net") ++ ++ for _, i := range ctr.convertPortMappings() { ++ if i.Protocol == "tcp" { ++ cmdArgs = append(cmdArgs, "-t") ++ } else if i.Protocol == "udp" { ++ cmdArgs = append(cmdArgs, "-u") ++ } else { ++ logrus.Errorf("can't forward protocol: %s", i.Protocol) ++ return nil ++ } ++ ++ arg := fmt.Sprintf("%d:%d", i.HostPort, i.ContainerPort) ++ cmdArgs = append(cmdArgs, arg) ++ } ++ ++ cmdArgs = append(cmdArgs, ctr.config.NetworkOptions["pasta"]...) ++ ++ cmdArgs = append(cmdArgs, netns.Path()) ++ ++ logrus.Debugf("pasta arguments: %s", strings.Join(cmdArgs, " ")) ++ ++ // pasta forks once ready, and quits once we delete the target namespace ++ _, err := exec.Command(path, cmdArgs...).Output() ++ if err != nil { ++ return errors.Wrapf(err, "failed to start pasta: %s", ++ err.(*exec.ExitError).Stderr) ++ } ++ ++ return nil ++} +diff --git a/pkg/namespaces/namespaces.go b/pkg/namespaces/namespaces.go +index a7736aee0..0b2cb2b0b 100644 +--- a/pkg/namespaces/namespaces.go ++++ b/pkg/namespaces/namespaces.go +@@ -19,6 +19,7 @@ const ( + privateType = "private" + shareableType = "shareable" + slirpType = "slirp4netns" ++ pastaType = "pasta" + ) + + // CgroupMode represents cgroup mode in the container. +@@ -388,6 +389,11 @@ func (n NetworkMode) IsSlirp4netns() bool { + return n == slirpType || strings.HasPrefix(string(n), slirpType+":") + } + ++// IsPasta indicates if we are running a rootless network stack using pasta ++func (n NetworkMode) IsPasta() bool { ++ return n == pastaType || strings.HasPrefix(string(n), pastaType + ":") ++} ++ + // IsNS indicates a network namespace passed in by path (ns:<path>) + func (n NetworkMode) IsNS() bool { + return strings.HasPrefix(string(n), nsType) +diff --git a/pkg/specgen/generate/namespaces.go b/pkg/specgen/generate/namespaces.go +index 3f77cbe76..a72be1731 100644 +--- a/pkg/specgen/generate/namespaces.go ++++ b/pkg/specgen/generate/namespaces.go +@@ -258,6 +258,16 @@ func namespaceOptions(ctx context.Context, s *specgen.SpecGenerator, rt *libpod. + val = fmt.Sprintf("slirp4netns:%s", s.NetNS.Value) + } + toReturn = append(toReturn, libpod.WithNetNS(portMappings, expose, postConfigureNetNS, val, nil)) ++ case specgen.Pasta: ++ portMappings, expose, err := createPortMappings(ctx, s, imageData) ++ if err != nil { ++ return nil, err ++ } ++ val := "pasta" ++ if s.NetNS.Value != "" { ++ val = fmt.Sprintf("pasta:%s", s.NetNS.Value) ++ } ++ toReturn = append(toReturn, libpod.WithNetNS(portMappings, expose, postConfigureNetNS, val, nil)) + case specgen.Private: + fallthrough + case specgen.Bridge: +diff --git a/pkg/specgen/generate/pod_create.go b/pkg/specgen/generate/pod_create.go +index 68fda3ad7..0d64027a3 100644 +--- a/pkg/specgen/generate/pod_create.go ++++ b/pkg/specgen/generate/pod_create.go +@@ -232,6 +232,12 @@ func MapSpec(p *specgen.PodSpecGenerator) (*specgen.SpecGenerator, error) { + p.InfraContainerSpec.NetworkOptions = p.NetworkOptions + p.InfraContainerSpec.NetNS.NSMode = specgen.NamespaceMode("slirp4netns") + } ++ case specgen.Pasta: ++ logrus.Debugf("Pod will use pasta") ++ if p.InfraContainerSpec.NetNS.NSMode != "host" { ++ p.InfraContainerSpec.NetworkOptions = p.NetworkOptions ++ p.InfraContainerSpec.NetNS.NSMode = specgen.NamespaceMode("pasta") ++ } + case specgen.NoNetwork: + logrus.Debugf("Pod will not use networking") + if len(p.InfraContainerSpec.PortMappings) > 0 || +diff --git a/pkg/specgen/namespaces.go b/pkg/specgen/namespaces.go +index e672bc65f..c7d443661 100644 +--- a/pkg/specgen/namespaces.go ++++ b/pkg/specgen/namespaces.go +@@ -47,6 +47,9 @@ const ( + // be used. + // Only used with the network namespace, invalid otherwise. + Slirp NamespaceMode = "slirp4netns" ++ // Pasta indicates that a pasta network stack should be used. ++ // Only used with the network namespace, invalid otherwise. ++ Pasta NamespaceMode = "pasta" + // KeepId indicates a user namespace to keep the owner uid inside + // of the namespace itself. + // Only used with the user namespace, invalid otherwise. +@@ -135,7 +138,7 @@ func validateNetNS(n *Namespace) error { + return nil + } + switch n.NSMode { +- case Slirp: ++ case Slirp, Pasta: + break + case "", Default, Host, Path, FromContainer, FromPod, Private, NoNetwork, Bridge: + break +@@ -167,7 +170,7 @@ func (n *Namespace) validate() error { + switch n.NSMode { + case "", Default, Host, Path, FromContainer, FromPod, Private: + // Valid, do nothing +- case NoNetwork, Bridge, Slirp: ++ case NoNetwork, Bridge, Slirp, Pasta: + return errors.Errorf("cannot use network modes with non-network namespace") + default: + return errors.Errorf("invalid namespace type %s specified", n.NSMode) +@@ -281,6 +284,8 @@ func ParseNetworkNamespace(ns string, rootlessDefaultCNI bool) (Namespace, map[s + switch { + case ns == string(Slirp), strings.HasPrefix(ns, string(Slirp)+":"): + toReturn.NSMode = Slirp ++ case ns == string(Pasta), strings.HasPrefix(ns, string(Pasta) + ":"): ++ toReturn.NSMode = Pasta + case ns == string(FromPod): + toReturn.NSMode = FromPod + case ns == "" || ns == string(Default) || ns == string(Private): +@@ -349,6 +354,13 @@ func ParseNetworkFlag(networks []string) (Namespace, map[string]types.PerNetwork + networkOptions[parts[0]] = strings.Split(parts[1], ",") + } + toReturn.NSMode = Slirp ++ case ns == string(Pasta), strings.HasPrefix(ns, string(Pasta) + ":"): ++ parts := strings.SplitN(ns, ":", 2) ++ if len(parts) > 1 { ++ networkOptions = make(map[string][]string) ++ networkOptions[parts[0]] = strings.Split(parts[1], ",") ++ } ++ toReturn.NSMode = Pasta + case ns == string(FromPod): + toReturn.NSMode = FromPod + case ns == "" || ns == string(Default) || ns == string(Private): +@@ -425,7 +437,7 @@ func ParseNetworkFlag(networks []string) (Namespace, map[string]types.PerNetwork + if parts[0] == "" { + return toReturn, nil, nil, errors.Wrapf(define.ErrInvalidArg, "network name cannot be empty") + } +- if util.StringInSlice(parts[0], []string{string(Bridge), string(Slirp), string(FromPod), string(NoNetwork), ++ if util.StringInSlice(parts[0], []string{string(Bridge), string(Slirp), string(Pasta), string(FromPod), string(NoNetwork), + string(Default), string(Private), string(Path), string(FromContainer), string(Host)}) { + return toReturn, nil, nil, errors.Wrapf(define.ErrInvalidArg, "can only set extra network names, selected mode %s conflicts with bridge", parts[0]) + } +diff --git a/pkg/specgen/podspecgen.go b/pkg/specgen/podspecgen.go +index 759caa0c0..f95bbffc7 100644 +--- a/pkg/specgen/podspecgen.go ++++ b/pkg/specgen/podspecgen.go +@@ -93,7 +93,7 @@ type PodNetworkConfig struct { + // PortMappings is a set of ports to map into the infra container. + // As, by default, containers share their network with the infra + // container, this will forward the ports to the entire pod. +- // Only available if NetNS is set to Bridge or Slirp. ++ // Only available if NetNS is set to Bridge, Slirp, or Pasta. + // Optional. + PortMappings []types.PortMapping `json:"portmappings,omitempty"` + // Map of networks names to ids the container should join to. +-- +2.28.0 + -- 2.34.1