- dev - passt

[PATCH] pasta: Add --no-pidns to keep spawned command in caller's PID namespace
by Christian Korneck 17 Sep '26

17 Sep '26

This is to allow running pasta inside a container without unmasking /proc for the whole container (Docker's --security-opt systempaths=unconfined, Podman's --security-opt unmask=ALL), which is undesirable as it exposes /proc/sysrq-trigger and other masked paths. In spawn mode, pasta clones the command with CLONE_NEWPID and mounts a new procfs instance on /proc, so that it matches the new PID namespace. Mounting procfs in a new user namespace requires a fully visible, unobstructed procfs. Container runtimes deliberately obstruct /proc (Docker, for example, masks /proc/kcore and friends and mounts /proc/sys read-only), so the mount is refused: Couldn't mount /proc: Operation not permitted We only warn and continue, leaving the command in a new PID namespace while the visible /proc still numbers processes in the outer one. Anything resolving its own PID through /proc then fails, for example bubblewrap: bwrap: open /proc/22/ns/ns failed: No such file or directory Add a --no-pidns option: skip CLONE_NEWPID for the spawned command and don't mount /proc, which is then not needed. User, network, mount, UTS and IPC namespaces, --config-net and port forwarding are unaffected. The option is rejected together with PID or --netns, as it only makes sense when we spawn the command ourselves. Add a test checking that, by default, the command runs in a new PID namespace, and that --no-pidns keeps it in the caller's one. Signed-off-by: Christian Korneck <christian(a)korneck.de> --- conf.c | 11 +++++++++++ passt.1 | 12 ++++++++++++ passt.h | 2 ++ pasta.c | 18 +++++++++++------- test/pasta_options/no_pidns | 25 +++++++++++++++++++++++++ test/run | 1 + 6 files changed, 62 insertions(+), 7 deletions(-) create mode 100644 test/pasta_options/no_pidns diff --git a/conf.c b/conf.c index faf2681..16b2f40 100644 --- a/conf.c +++ b/conf.c @@ -742,6 +742,7 @@ pasta_opts: " implied if PATH or NAME are given without --userns\n" " --no-netns-quit Don't quit if filesystem-bound target\n" " network namespace is deleted\n" + " --no-pidns Don't spawn command in a new PID namespace\n" " --config-net Configure tap interface in namespace\n" " --no-copy-routes DEPRECATED:\n" " Don't copy all routes to namespace\n" @@ -1348,6 +1349,7 @@ void conf(struct ctx *c, int argc, char **argv) {"stats", required_argument, NULL, 31 }, {"conf-path", required_argument, NULL, 'c' }, {"chroot-fallback", no_argument, NULL, 32 }, + {"no-pidns", no_argument, NULL, 33 }, { 0 }, }; const char *optstring = "+dqfel:hs:c:F:I:p:P:m:a:n:M:g:i:o:D:S:H:461t:u:T:U:"; @@ -1589,6 +1591,12 @@ void conf(struct ctx *c, int argc, char **argv) case 32: c->chroot_fallback = true; break; + case 33: + if (c->mode != MODE_PASTA) + die("--no-pidns is for pasta mode only"); + + c->no_pidns = true; + break; case 'd': c->debug = 1; c->quiet = 0; @@ -1948,6 +1956,9 @@ void conf(struct ctx *c, int argc, char **argv) else if (optind != argc) die("Extra non-option argument: %s", argv[optind]); + if (c->no_pidns && *netns) + die("--no-pidns is incompatible with PID or --netns"); + conf_open_files(c); /* Before any possible setuid() / setgid() */ isolate_user(c, uid, gid, !netns_only, userns); diff --git a/passt.1 b/passt.1 index 53e072a..0b780fe 100644 --- a/passt.1 +++ b/passt.1 @@ -704,6 +704,18 @@ is bound to the filesystem, and the given path is deleted, or if the target network namespace is represented by a procfs entry, and that entry is deleted, representing the fact that a process with the given PID terminated. +.TP +.BR \-\-no-pidns +Don't create a new PID namespace for the spawned command or shell: keep it in +the PID namespace \fBpasta\fR itself runs in, and don't mount a new +\fIprocfs\fR instance on \fI/proc\fR for it. This is useful in environments +where mounting \fIprocfs\fR is not permitted, such as containers, where the +command would otherwise get a \fI/proc\fR view that doesn't match its own PID +namespace. Note that, without a PID namespace, processes started by the +command are not terminated once the command exits. + +This option can't be specified with a PID or with \-\-netns. + .TP .BR \-\-config-net Configure networking in the namespace: set up addresses and routes as configured diff --git a/passt.h b/passt.h index 51ccd4f..afd8e9f 100644 --- a/passt.h +++ b/passt.h @@ -195,6 +195,7 @@ struct ip6_ctx { * @pasta_ifn: Name of namespace interface for pasta * @pasta_ifi: Index of namespace interface for pasta * @pasta_conf_ns: Configure namespace after creating it + * @no_pidns: Don't create a new PID namespace for spawned command * @fwd: Forwarding tables * @fwd_pending: Pending forward tables * @no_tcp: Disable TCP operation @@ -278,6 +279,7 @@ struct ctx { char pasta_ifn[IF_NAMESIZE]; unsigned int pasta_ifi; int pasta_conf_ns; + bool no_pidns; struct fwd_table *fwd[PIF_NUM_TYPES]; struct fwd_table *fwd_pending[PIF_NUM_TYPES]; diff --git a/pasta.c b/pasta.c index 5aa56b7..2d916e1 100644 --- a/pasta.c +++ b/pasta.c @@ -194,15 +194,17 @@ static int pasta_spawn_cmd(void *arg) if (prctl(PR_SET_PDEATHSIG, SIGKILL)) die_perror("Couldn't set PR_SET_PDEATHSIG"); - /* We run in a detached PID and mount namespace: mount /proc over */ - if (mount("", "/proc", "proc", 0, NULL)) + a = (const struct pasta_spawn_cmd_arg *)arg; + + /* We run in a detached mount namespace, and, unless --no-pidns was + * given, in a detached PID namespace: mount /proc over + */ + if (!a->c->no_pidns && mount("", "/proc", "proc", 0, NULL)) warn_perror("Couldn't mount /proc"); if (write_file("/proc/sys/net/ipv4/ping_group_range", "0 0")) warn("Cannot set ping_group_range, ICMP requests might fail"); - a = (const struct pasta_spawn_cmd_arg *)arg; - conf_hostname_len = strlen(a->c->hostname); if (conf_hostname_len > 0) { if (sethostname(a->c->hostname, conf_hostname_len)) @@ -243,6 +245,7 @@ static int pasta_spawn_cmd(void *arg) void pasta_start_ns(struct ctx *c, uid_t uid, gid_t gid, bool config_idmaps, int argc, char *argv[]) { + int flags = CLONE_NEWIPC | CLONE_NEWNET | CLONE_NEWUTS | CLONE_NEWNS; char ns_fn_stack[NS_FN_STACK_SIZE] __attribute__ ((aligned(__alignof__(max_align_t)))); struct pasta_spawn_cmd_arg arg = { @@ -293,10 +296,11 @@ void pasta_start_ns(struct ctx *c, uid_t uid, gid_t gid, bool config_idmaps, sigaddset(&set, SIGUSR1); sigprocmask(SIG_BLOCK, &set, NULL); + if (!c->no_pidns) + flags |= CLONE_NEWPID; + pasta_child_pid = do_clone(pasta_spawn_cmd, ns_fn_stack, - sizeof(ns_fn_stack), - CLONE_NEWIPC | CLONE_NEWPID | CLONE_NEWNET | - CLONE_NEWUTS | CLONE_NEWNS | SIGCHLD, + sizeof(ns_fn_stack), flags | SIGCHLD, (void *)&arg); if (pasta_child_pid == -1) diff --git a/test/pasta_options/no_pidns b/test/pasta_options/no_pidns new file mode 100644 index 0000000..f565b2f --- /dev/null +++ b/test/pasta_options/no_pidns @@ -0,0 +1,25 @@ +# SPDX-License-Identifier: GPL-2.0-or-later +# +# PASST - Plug A Simple Socket Transport +# for qemu/UNIX domain socket mode +# +# PASTA - Pack A Subtle Tap Abstraction +# for network namespace/tap device mode +# +# test/pasta_options/no_pidns - Check --no-pidns handling + +htools readlink + +test Command is spawned in a new PID namespace by default +set PIDNS __STATEDIR__/pidns +set OUT __STATEDIR__/no-pidns.out +set ERR __STATEDIR__/no-pidns.err + +passt readlink /proc/self/ns/pid > __PIDNS__ +passt ./pasta -q -- readlink /proc/self/ns/pid > __OUT__ 2> __ERR__ +check [ "$(cat __PIDNS__)" != "$(cat __OUT__)" ] + +test --no-pidns keeps command in the caller's PID namespace +passt ./pasta -q --no-pidns -- readlink /proc/self/ns/pid > __OUT__ 2> __ERR__ +check [ "$(cat __PIDNS__)" = "$(cat __OUT__)" ] +check [ ! -s __ERR__ ] diff --git a/test/run b/test/run index c4073cc..14ddbbf 100755 --- a/test/run +++ b/test/run @@ -86,6 +86,7 @@ run() { setup pasta_options test pasta_options/log_to_file test pasta_options/netns_only + test pasta_options/no_pidns teardown pasta_options setup build -- 2.55.0

3 9

[PATCH v3 0/2] [PATCH 0/2] dhcpv6: Add --dhcpv6-opt for custom DHCPv6 options
by Anshu Kumari 10 Sep '26

10 Sep '26

This series adds a --dhcpv6-opt CODE,VALUE command-line option to inject custom options into DHCPv6 replies, complementing the existing --dhcp-opt support for DHCPv4. The primary use case is UEFI HTTP Boot, which requires Vendor Class (option 16) with the correct enterprise-number + length-prefixed wire encoding per RFC 8415 Section 21.16, and Boot File URL (option 59). Value formats are determined automatically from the option code via a type table. Supported types include plain strings, IPv6 addresses (single and list), 8/16/32-bit integers, vendor class (ENTERPRISE:DATA), and length-prefixed string lists. Anshu Kumari (2): dhcpv6: Add --dhcpv6-opt with option type table and value parser dhcpv6: Inject custom options into DHCPv6 replies conf.c | 25 +++- dhcpv6.c | 363 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ dhcpv6.h | 2 + passt.1 | 31 +++++ 4 files changed, 420 insertions(+), 1 deletion(-) -- 2.55.0

3 6

[PATCH v6 0/6] Add --dhcp-boot and --dhcp-opt options
by Anshu Kumari 10 Sep '26

10 Sep '26

This series adds support for custom DHCP options in passt, enabling network boot (PXE/UEFI HTTP Boot) and arbitrary DHCP option injection. Two new command-line flags are introduced: --dhcp-boot URL Sets the boot file URL (DHCP option 67 and the legacy boot file field) --dhcp-opt CODE,VALUE Sets any DHCP option by numeric code, with type-aware parsing per RFC 2132 The DHCP reply path is extended with option overload support (RFC 2132 option 52), allowing options to overflow into the file and sname fields when the standard options area is full. RFC 3396 option splitting for concatenation-requiring options has also been introduce as a separate patch. Anshu Kumari (6): dhcp: Refactor fill_one() to operate on a generic buffer dhcp: Add option state management with enum opt_state dhcp: Add option overload dhcp: Add --dhcp-opt with option table and value parser dhcp: Add --dhcp-boot command-line option dhcp: Add RFC 3396 option splitting for concatenation-requiring options conf.c | 31 +++- dhcp.c | 561 ++++++++++++++++++++++++++++++++++++++++++++++++++------ dhcp.h | 2 + passt.1 | 49 +++++ 4 files changed, 588 insertions(+), 55 deletions(-) -- 2.55.0

3 16

[PATCH] RFC: Don't override system's default ping_group_range for pasta
by David Gibson 09 Sep '26

09 Sep '26

When spawning a command, pasta sets the net.ipv4.ping_group_range sysctl to 0 0, meaning only group 0 can use ping sockets within the namespace. Since group 0 is the only one we map in the userns, that's equivalent to anyone being able to use ping sockets. Although the kernel default for this is 1 0 (nobody can use ping sockets), common distros - at least ones using systemd or even just systemd-udevd - appear to set it to "0 2147483647" meaning effectively anyone can use ping sockets. There's no obvious reason that we need to override the system's default behaviour here. The override was introduced in 32d07f5e5 ("passt, pasta: Completely avoid dynamic memory allocation") as part of a large chunk which kind of looks like it was meant to be in another patch, so the git history isn't particularly informative. Signed-off-by: David Gibson <david(a)gibson.dropbear.id.au> --- pasta.c | 3 --- 1 file changed, 3 deletions(-) diff --git a/pasta.c b/pasta.c index 5aa56b78..4248b508 100644 --- a/pasta.c +++ b/pasta.c @@ -198,9 +198,6 @@ static int pasta_spawn_cmd(void *arg) if (mount("", "/proc", "proc", 0, NULL)) warn_perror("Couldn't mount /proc"); - if (write_file("/proc/sys/net/ipv4/ping_group_range", "0 0")) - warn("Cannot set ping_group_range, ICMP requests might fail"); - a = (const struct pasta_spawn_cmd_arg *)arg; conf_hostname_len = strlen(a->c->hostname); -- 2.55.0

2 4

[PATCH 3/3] Optimize route dependency solver.
by Martin Schitter 08 Sep '26

08 Sep '26

Avoid retransmitting routes that explicitly got reported as already existing. Never make more resolution attempts than there are still unresolved dependency errors. In the context of my current mesh network with ~700 route entries these changes reduce the required processing time from 5.5s to 15ms. --- netlink.c | 25 ++++++++++++++++++------- 1 file changed, 18 insertions(+), 7 deletions(-) diff --git a/netlink.c b/netlink.c index 2b49fe3..1cbe5e5 100644 --- a/netlink.c +++ b/netlink.c @@ -703,13 +703,12 @@ int nl_route_dup(int s_src, unsigned int ifi_src, /* Routes might have dependencies between each other, and the kernel * processes RTM_NEWROUTE messages sequentially. For n routes, we might - * need to send the requests up to n times to get all of them inserted. - * Routes that have been already inserted will return -EEXIST, but we - * can safely ignore that and repeat the requests. This avoids the need - * to calculate dependencies: let the kernel do that. + * need to send the requests up to n times in the worst case to get all + * of them inserted. */ clock_gettime(CLOCK_MONOTONIC, &start); - for (i = 0; i < dup_routes; i++) { + for (i = dup_routes; i > 0; i--) { + int dep_errors = 0; for (nh = (struct nlmsghdr *)buf, left = nlmsgs_size; NLMSG_OK(nh, left); nh = NLMSG_NEXT(nh, left)) { @@ -722,10 +721,22 @@ int nl_route_dup(int s_src, unsigned int ifi_src, rc = nl_do(s_dst, nh, RTM_NEWROUTE, (flags & ~NLM_F_DUMP_FILTERED) | NLM_F_CREATE, nh->nlmsg_len); - if (rc < 0 && rc != -EEXIST && - rc != -ENETUNREACH && rc != -EHOSTUNREACH) + + if ( rc == -EEXIST) { + /* Exclude existing routes from further retry attempts */ + nh->nlmsg_type = NLMSG_NOOP; + } + if ( rc == -ENETUNREACH || rc == -EHOSTUNREACH){ + dep_errors++; + continue; + } + if (rc < 0) return rc; } + debug("route dependency errors: %d", dep_errors); + /* Avoid having much more resolution attempts than + * there are still unresolved dependency errors */ + i = MIN(i, dep_errors++); } clock_gettime(CLOCK_MONOTONIC, &now); debug("route dependency handling time: %f s", -- 2.53.0

1 0

Fix for Bug #221 -- Issue concerning bigger routing tables
by Martin Schitter 08 Sep '26

08 Sep '26

This is a rebased version of the three patches to fix #221. Only patch 3/3 got slightly modified becaus of merge conflicts. [PATCH 1/3] Fix handling of netlink multipart route dumps. [PATCH 2/3] Timekeeping: Resolving Route Dependencies [PATCH 3/3] Optimize route dependency solver.

1 2

[PATCH] contrib/apparmor: add missing setfcap capability
by Sevinj Aghayeva 08 Sep '26

08 Sep '26

Since Linux 5.12, writing a mapping from UID 0 to /proc/self/uid_map requires CAP_SETFCAP. isolation.c already retains this capability for the case where pasta spawns a child from a non-init user namespace, but the AppArmor profile doesn't grant it, so the write is denied whenever the profile is enforced. Add setfcap to the AppArmor abstraction to match what isolation.c expects. Link: https://bugs.passt.top/show_bug.cgi?id=172 Signed-off-by: Sevinj Aghayeva <sevinj.aghayeva(a)gmail.com> --- contrib/apparmor/abstractions/passt | 1 + 1 file changed, 1 insertion(+) diff --git a/contrib/apparmor/abstractions/passt b/contrib/apparmor/abstractions/passt index 85bd1ee..f4570c1 100644 --- a/contrib/apparmor/abstractions/passt +++ b/contrib/apparmor/abstractions/passt @@ -24,6 +24,7 @@ capability setpcap, capability net_admin, capability sys_ptrace, + capability setfcap, userns, / r, # isolate_prefork(), isolation.c -- 2.43.0

2 1

[PATCH] vhost_user: Reset vq enable flag in vu_cleanup()
by Laurent Vivier 08 Sep '26

08 Sep '26

vu_cleanup() resets most virtqueue state (started, notification, file descriptors) but does not reset the enable flag. After a QEMU disconnect and reconnect, the stale enable flag causes vu_set_vring_enable_exec() to hit its early return check (vq->enable == enable). Reset vq->enable to false in vu_cleanup() alongside the other virtqueue state. Signed-off-by: Laurent Vivier <lvivier(a)redhat.com> Reviewed-by: David Gibson <david(a)gibson.dropbear.id.au> --- vhost_user.c | 1 + 1 file changed, 1 insertion(+) diff --git a/vhost_user.c b/vhost_user.c index a1259c2624c0..272295f29db4 100644 --- a/vhost_user.c +++ b/vhost_user.c @@ -1081,6 +1081,7 @@ void vu_cleanup(struct vu_dev *vdev) struct vu_virtq *vq = &vdev->vq[i]; vq->started = false; + vq->enable = false; vq->notification = true; if (vq->call_fd != -1) { -- 2.55.0

2 1

[PATCH 0/5] Settings for Zed editor and minor fixups
by David Gibson 08 Sep '26

08 Sep '26

I've been playing with the Zed editor again. It's default settings don't quite match passt's style, so add a config file to make it do the right thing. Also update some of the config for clangd to suppress bogus warnings the way Zed invokes it. Between them these also found a handful of minor but real code warts, so fix those, too. David Gibson (5): clangd: Add _GNU_SOURCE to default clangd options util: Eliminate a stray trailing whitespace util: Add missing O_CLOEXEC for !HAS_GETRANDOM path util, pasta: Remove some unneeded #includes Add Zed editor settings .clangd | 2 +- .zed/settings.json | 4 ++++ pasta.c | 1 - util.c | 7 ++----- 4 files changed, 7 insertions(+), 7 deletions(-) create mode 100644 .zed/settings.json -- 2.53.0

3 6

[PATCH] udp_flow: Remove obsolete comment
by David Gibson 08 Sep '26

08 Sep '26

This comment was true - although the upshot was not particularly clear - when it was introduced. However, for other reaosns we now *do* always populate ini->oaddr from flow_initiate_sa(). So rather than explaning why we can't, we can actually validate ini->oaddr. Signed-off-by: David Gibson <david(a)gibson.dropbear.id.au> --- udp_flow.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/udp_flow.c b/udp_flow.c index f59649f6..a29f3562 100644 --- a/udp_flow.c +++ b/udp_flow.c @@ -247,11 +247,8 @@ flow_sidx_t udp_flow_from_sock(const struct ctx *c, uint8_t pif, ini = flow_initiate_sa(flow, pif, s_in, dst, port); if (!inany_is_unicast(&ini->eaddr) || + inany_is_unspecified(&ini->oaddr) || ini->eport == 0 || ini->oport == 0) { - /* In principle ini->oddr also must be specified, but when we've - * been initiated from a socket bound to 0.0.0.0 or ::, we don't - * know our address, so we have to leave it unpopulated. - */ flow_err_ratelimit(flow, now, "Invalid endpoint on UDP recvfrom()"); flow_alloc_cancel(flow); -- 2.55.0

2 1