Merge c4d274dde4 into f6793bbcf0

"nsenter -a" doesn't migrate the specified process into the target
cgroup (it really should). Thus the cgroup will remain in a cgroup
that is (due to cgroup ns) outside our visibility. The kernel will
report the cgroup path of such cgroups as starting with "/../". Detect
that and print a reasonably error message instead of trying to resolve
that.

src/basic/cgroup-util.c

@@ -803,6 +803,10 @@ int cg_pid_get_path(const char *controller, pid_t pid, char **ret_path) {
                 if (!path)
                         return -ENOMEM;
 
+                /* Refuse cgroup paths from outside our cgroup namespace */
+                if (startswith(path, "/../"))
+                        return -EUNATCH;
+
                 /* Truncate suffix indicating the process is a zombie */
                 e = endswith(path, " (deleted)");
                 if (e)

src/basic/process-util.c

@@ -102,8 +102,8 @@ int pid_get_comm(pid_t pid, char **ret) {
         _cleanup_free_ char *escaped = NULL, *comm = NULL;
         int r;
 
-        assert(ret);
         assert(pid >= 0);
+        assert(ret);
 
         if (pid == 0 || pid == getpid_cached()) {
                 comm = new0(char, TASK_COMM_LEN + 1); /* Must fit in 16 byte according to prctl(2) */
@@ -143,6 +143,9 @@ int pidref_get_comm(const PidRef *pid, char **ret) {
         if (!pidref_is_set(pid))
                 return -ESRCH;
 
+        if (pidref_is_remote(pid))
+                return -EREMOTE;
+
         r = pid_get_comm(pid->pid, &comm);
         if (r < 0)
                 return r;
@@ -289,6 +292,9 @@ int pidref_get_cmdline(const PidRef *pid, size_t max_columns, ProcessCmdlineFlag
         if (!pidref_is_set(pid))
                 return -ESRCH;
 
+        if (pidref_is_remote(pid))
+                return -EREMOTE;
+
         r = pid_get_cmdline(pid->pid, max_columns, flags, &s);
         if (r < 0)
                 return r;
@@ -331,6 +337,9 @@ int pidref_get_cmdline_strv(const PidRef *pid, ProcessCmdlineFlags flags, char *
         if (!pidref_is_set(pid))
                 return -ESRCH;
 
+        if (pidref_is_remote(pid))
+                return -EREMOTE;
+
         r = pid_get_cmdline_strv(pid->pid, flags, &args);
         if (r < 0)
                 return r;
@@ -477,6 +486,9 @@ int pidref_is_kernel_thread(const PidRef *pid) {
         if (!pidref_is_set(pid))
                 return -ESRCH;
 
+        if (pidref_is_remote(pid))
+                return -EREMOTE;
+
         result = pid_is_kernel_thread(pid->pid);
         if (result < 0)
                 return result;
@@ -594,6 +606,9 @@ int pidref_get_uid(const PidRef *pid, uid_t *ret) {
         if (!pidref_is_set(pid))
                 return -ESRCH;
 
+        if (pidref_is_remote(pid))
+                return -EREMOTE;
+
         r = pid_get_uid(pid->pid, &uid);
         if (r < 0)
                 return r;
@@ -794,6 +809,9 @@ int pidref_get_start_time(const PidRef *pid, usec_t *ret) {
         if (!pidref_is_set(pid))
                 return -ESRCH;
 
+        if (pidref_is_remote(pid))
+                return -EREMOTE;
+
         r = pid_get_start_time(pid->pid, ret ? &t : NULL);
         if (r < 0)
                 return r;
@@ -1093,6 +1111,9 @@ int pidref_is_my_child(const PidRef *pid) {
         if (!pidref_is_set(pid))
                 return -ESRCH;
 
+        if (pidref_is_remote(pid))
+                return -EREMOTE;
+
         result = pid_is_my_child(pid->pid);
         if (result < 0)
                 return result;
@@ -1128,6 +1149,9 @@ int pidref_is_unwaited(const PidRef *pid) {
         if (!pidref_is_set(pid))
                 return -ESRCH;
 
+        if (pidref_is_remote(pid))
+                return -EREMOTE;
+
         if (pid->pid == 1 || pidref_is_self(pid))
                 return true;
 
@@ -1169,6 +1193,9 @@ int pidref_is_alive(const PidRef *pidref) {
         if (!pidref_is_set(pidref))
                 return -ESRCH;
 
+        if (pidref_is_remote(pidref))
+                return -EREMOTE;
+
         result = pid_is_alive(pidref->pid);
         if (result < 0) {
                 assert(result != -ESRCH);

src/cryptenroll/cryptenroll-fido2.c

@@ -193,7 +193,7 @@ int enroll_fido2(
                 fflush(stdout);
 
                 fprintf(stderr,
-                        "\nPlease save this FIDO2 credential ID. It is required when unloocking the volume\n"
+                        "\nPlease save this FIDO2 credential ID. It is required when unlocking the volume\n"
                         "using the associated FIDO2 keyslot which we just created. To configure automatic\n"
                         "unlocking using this FIDO2 token, add an appropriate entry to your /etc/crypttab\n"
                         "file, see %s for details.\n", link);

src/machine/machine-varlink.c

@@ -12,6 +12,7 @@
 #include "json-util.h"
 #include "machine-varlink.h"
 #include "machine.h"
+#include "mount-util.h"
 #include "path-util.h"
 #include "pidref.h"
 #include "process-util.h"
@@ -570,3 +571,100 @@ int vl_method_open(sd_varlink *link, sd_json_variant *parameters, sd_varlink_met
 
         return sd_varlink_reply(link, v);
 }
+
+typedef struct MachineMountParameters {
+        const char *name;
+        PidRef pidref;
+        char *src, *dest;
+        bool read_only, mkdir;
+} MachineMountParameters;
+
+static void machine_mount_paramaters_done(MachineMountParameters *p) {
+        assert(p);
+        pidref_done(&p->pidref);
+        free(p->src);
+        free(p->dest);
+}
+
+int vl_method_mount(sd_varlink *link, sd_json_variant *parameters, sd_varlink_method_flags_t flags, void *userdata) {
+        static const sd_json_dispatch_field dispatch_table[] = {
+                VARLINK_DISPATCH_MACHINE_LOOKUP_FIELDS(MachineOpenParameters),
+                { "source",              SD_JSON_VARIANT_STRING,  json_dispatch_path,       offsetof(MachineMountParameters, src),       SD_JSON_MANDATORY },
+                { "destination",         SD_JSON_VARIANT_STRING,  json_dispatch_path,       offsetof(MachineMountParameters, dest),      0                 },
+                { "readOnly",            SD_JSON_VARIANT_BOOLEAN, sd_json_dispatch_stdbool, offsetof(MachineMountParameters, read_only), 0                 },
+                { "makeFileOrDirectory", SD_JSON_VARIANT_BOOLEAN, sd_json_dispatch_stdbool, offsetof(MachineMountParameters, mkdir),     0                 },
+                VARLINK_DISPATCH_POLKIT_FIELD,
+                {}
+        };
+
+        Manager *manager = ASSERT_PTR(userdata);
+        _cleanup_(machine_mount_paramaters_done) MachineMountParameters p = { .pidref = PIDREF_NULL };
+        MountInNamespaceFlags mount_flags = 0;
+        uid_t uid_shift;
+        int r;
+
+        assert(link);
+        assert(parameters);
+
+        r = sd_varlink_dispatch(link, parameters, dispatch_table, &p);
+        if (r != 0)
+                return r;
+
+        /* There is no need for extra validation since path_is_absolute() does path_is_valid() and path_is_absolute().*/
+        const char *dest = p.dest ?: p.src;
+
+        Machine *machine;
+        r = lookup_machine_by_name_or_pidref(link, manager, p.name, &p.pidref, &machine);
+        if (r == -ESRCH)
+                return sd_varlink_error(link, "io.systemd.Machine.NoSuchMachine", NULL);
+        if (r != 0)
+                return r;
+
+        if (machine->class != MACHINE_CONTAINER)
+                return sd_varlink_error(link, "io.systemd.Machine.NotSupported", NULL);
+
+        r = varlink_verify_polkit_async(
+                        link,
+                        manager->bus,
+                        "org.freedesktop.machine1.manage-machines",
+                        (const char**) STRV_MAKE("name", machine->name,
+                                                 "verb", "bind", // TODO(ikruglov): mount seems a better verb here, but it's bind in DBus
+                                                 "src", p.src,
+                                                 "dest", dest),
+                        &manager->polkit_registry);
+        if (r <= 0)
+                return r;
+
+        r = machine_get_uid_shift(machine, &uid_shift);
+        if (r < 0)
+                return log_debug_errno(r, "Failed to get machine UID shift: %m");
+        if (uid_shift != 0) {
+                log_debug_errno(EOPNOTSUPP, "Can't bind mount on container with user namespacing applied.");
+                return sd_varlink_error(link, "io.systemd.Machine.NotSupported", NULL);
+        }
+
+        if (p.read_only)
+                mount_flags |= MOUNT_IN_NAMESPACE_READ_ONLY;
+        if (p.mkdir)
+                mount_flags |= MOUNT_IN_NAMESPACE_MAKE_FILE_OR_DIRECTORY;
+
+        const char *propagate_directory = strjoina("/run/systemd/nspawn/propagate/", machine->name);
+        if (!propagate_directory)
+                return -ENOMEM;
+
+        /* TODO(ikruglov): base on implementation of mount_in_namespace() it
+         * feels better to convert this in an Operation() so we can apply
+         * `manager->n_operations >= OPERATIONS_MAX)`.  But it will require
+         * some surgery. */
+        r = bind_mount_in_namespace(
+                        &machine->leader,
+                        propagate_directory,
+                        "/run/host/incoming/",
+                        p.src,
+                        dest,
+                        mount_flags);
+        if (r < 0)
+                return log_debug_errno(r, "Failed to mount %s on %s in machine's namespace: %m", p.src, dest);
+
+        return sd_varlink_reply(link, NULL);
+}

src/machine/machine-varlink.h

@@ -25,3 +25,4 @@ int vl_method_unregister_internal(sd_varlink *link, sd_json_variant *parameters,
 int vl_method_terminate_internal(sd_varlink *link, sd_json_variant *parameters, sd_varlink_method_flags_t flags, void *userdata);
 int vl_method_kill(sd_varlink *link, sd_json_variant *parameters, sd_varlink_method_flags_t flags, void *userdata);
 int vl_method_open(sd_varlink *link, sd_json_variant *parameters, sd_varlink_method_flags_t flags, void *userdata);
+int vl_method_mount(sd_varlink *link, sd_json_variant *parameters, sd_varlink_method_flags_t flags, void *userdata);

src/machine/machined-varlink.c

@@ -774,6 +774,7 @@ static int manager_varlink_init_machine(Manager *m) {
                         "io.systemd.Machine.Terminate",   vl_method_terminate,
                         "io.systemd.Machine.Kill",        vl_method_kill,
                         "io.systemd.Machine.Open",        vl_method_open,
+                        "io.systemd.Machine.Mount",       vl_method_mount,
                         "io.systemd.MachineImage.List",   vl_method_list_images,
                         "io.systemd.MachineImage.Update", vl_method_update_image,
                         "io.systemd.MachineImage.Clone",  vl_method_clone_image,

src/shared/killall.c

@@ -46,13 +46,17 @@ static bool argv_has_at(pid_t pid) {
         return c == '@';
 }
 
-static bool is_survivor_cgroup(const PidRef *pid) {
+static bool is_in_survivor_cgroup(const PidRef *pid) {
         _cleanup_free_ char *cgroup_path = NULL;
         int r;
 
         assert(pidref_is_set(pid));
 
         r = cg_pidref_get_path(/* root= */ NULL, pid, &cgroup_path);
+        if (r == -EUNATCH) {
+                log_warning_errno(r, "Process " PID_FMT " appears to originate in foreign namespace, ignoring.", pid->pid);
+                return true;
+        }
         if (r < 0) {
                 log_warning_errno(r, "Failed to get cgroup path of process " PID_FMT ", ignoring: %m", pid->pid);
                 return false;
@@ -86,7 +90,7 @@ static bool ignore_proc(const PidRef *pid, bool warn_rootfs) {
                 return true; /* also ignore processes where we can't determine this */
 
         /* Ignore processes that are part of a cgroup marked with the user.survive_final_kill_signal xattr */
-        if (is_survivor_cgroup(pid))
+        if (is_in_survivor_cgroup(pid))
                 return true;
 
         r = pidref_get_uid(pid, &uid);

src/test/test-audit-util.c

@@ -7,24 +7,26 @@ TEST(audit_loginuid_from_pid) {
         _cleanup_(pidref_done) PidRef self = PIDREF_NULL, pid1 = PIDREF_NULL;
         int r;
 
-        assert_se(pidref_set_self(&self) >= 0);
-        assert_se(pidref_set_pid(&pid1, 1) >= 0);
+        ASSERT_OK(pidref_set_self(&self));
+        ASSERT_OK(pidref_set_pid(&pid1, 1));
 
         uid_t uid;
         r = audit_loginuid_from_pid(&self, &uid);
-        assert_se(r >= 0 || r == -ENODATA);
+        if (r != -ENODATA)
+                ASSERT_OK(r);
         if (r >= 0)
                 log_info("self audit login uid: " UID_FMT, uid);
 
-        assert_se(audit_loginuid_from_pid(&pid1, &uid) == -ENODATA);
+        ASSERT_ERROR(audit_loginuid_from_pid(&pid1, &uid), ENODATA);
 
         uint32_t sessionid;
         r = audit_session_from_pid(&self, &sessionid);
-        assert_se(r >= 0 || r == -ENODATA);
+        if (r != -ENODATA)
+                ASSERT_OK(r);
         if (r >= 0)
                 log_info("self audit session id: %" PRIu32, sessionid);
 
-        assert_se(audit_session_from_pid(&pid1, &sessionid) == -ENODATA);
+        ASSERT_ERROR(audit_session_from_pid(&pid1, &sessionid), ENODATA);
 }
 
 static int intro(void) {