void dns_transaction_gc(DnsTransaction *t) {
assert(t);
if (t->block_gc > 0)
return;
if (set_isempty(t->queries) && set_isempty(t->zone_items))
dns_transaction_free(t);
}
int scope_abandon(Scope *s) {
assert(s);
if (!IN_SET(s->state, SCOPE_RUNNING, SCOPE_ABANDONED))
return -ESTALE;
free(s->controller);
s->controller = NULL;
/* The client is no longer watching the remaining processes,
* so let's step in here, under the assumption that the
* remaining processes will be sooner or later reassigned to
* us as parent. */
unit_tidy_watch_pids(UNIT(s), 0, 0);
unit_watch_all_pids(UNIT(s));
/* If the PID set is empty now, then let's finish this off */
if (set_isempty(UNIT(s)->pids))
scope_notify_cgroup_empty_event(UNIT(s));
else
scope_set_state(s, SCOPE_ABANDONED);
return 0;
}
static int scope_verify(Scope *s) {
assert(s);
if (UNIT(s)->load_state != UNIT_LOADED)
return 0;
if (set_isempty(UNIT(s)->pids) && UNIT(s)->manager->n_reloading <= 0) {
log_unit_error(UNIT(s), "Scope has no PIDs. Refusing.");
return -EINVAL;
}
return 0;
}
/*
* db_set_isempty() - This function checks to see if a set is empty. The set
* (or sequence) must have no elements at all in order for this to be true.
* If this is a sequence and there are only NULL elements, this function
* will return false since NULL elements are still considered valid elements
* for sequences.
* return : non-zero if the set has no elements
* set(in): set descriptor
*/
int
db_set_isempty (DB_SET * set)
{
int retval;
CHECK_CONNECT_FALSE ();
CHECK_1ARG_FALSE (set);
/* allow all types */
retval = (set_isempty (set)) ? 1 : 0;
return (retval);
}
static int scope_verify(Scope *s) {
assert(s);
if (UNIT(s)->load_state != UNIT_LOADED)
return 0;
if (set_isempty(UNIT(s)->pids) &&
!MANAGER_IS_RELOADING(UNIT(s)->manager) &&
!unit_has_name(UNIT(s), SPECIAL_INIT_SCOPE)) {
log_unit_error(UNIT(s), "Scope has no PIDs. Refusing.");
return -ENOENT;
}
return 0;
}
static void scope_sigchld_event(Unit *u, pid_t pid, int code, int status) {
/* If we get a SIGCHLD event for one of the processes we were
interested in, then we look for others to watch, under the
assumption that we'll sooner or later get a SIGCHLD for
them, as the original process we watched was probably the
parent of them, and they are hence now our children. */
unit_tidy_watch_pids(u, 0, 0);
unit_watch_all_pids(u);
/* If the PID set is empty now, then let's finish this off */
if (set_isempty(u->pids))
scope_notify_cgroup_empty_event(u);
}
static int scope_verify(Scope *s) {
assert(s);
if (UNIT(s)->load_state != UNIT_LOADED)
return 0;
if (set_isempty(UNIT(s)->pids) &&
!manager_is_reloading_or_reexecuting(UNIT(s)->manager) &&
!unit_has_name(UNIT(s), SPECIAL_INIT_SCOPE)) {
log_unit_error(UNIT(s), "Scope has no PIDs. Refusing.");
return -EINVAL;
}
return 0;
}
int automount_send_ready(Automount *a, int status) {
int ioctl_fd, r;
unsigned token;
assert(a);
assert(status <= 0);
if (set_isempty(a->tokens))
return 0;
ioctl_fd = open_ioctl_fd(UNIT(a)->manager->dev_autofs_fd, a->where, a->dev_id);
if (ioctl_fd < 0) {
r = ioctl_fd;
goto fail;
}
if (status)
log_debug_unit(UNIT(a)->id, "Sending failure: %s", strerror(-status));
else
log_debug_unit(UNIT(a)->id, "Sending success.");
r = 0;
/* Autofs thankfully does not hand out 0 as a token */
while ((token = PTR_TO_UINT(set_steal_first(a->tokens)))) {
int k;
/* Autofs fun fact II:
*
* if you pass a positive status code here, the kernel will
* freeze! Yay! */
k = autofs_send_ready(UNIT(a)->manager->dev_autofs_fd,
ioctl_fd,
token,
status);
if (k < 0)
r = k;
}
fail:
if (ioctl_fd >= 0)
close_nointr_nofail(ioctl_fd);
return r;
}
static int automount_send_ready(Automount *a, Set *tokens, int status) {
_cleanup_close_ int ioctl_fd = -1;
unsigned token;
int r;
assert(a);
assert(status <= 0);
if (set_isempty(tokens))
return 0;
ioctl_fd = open_ioctl_fd(UNIT(a)->manager->dev_autofs_fd, a->where, a->dev_id);
if (ioctl_fd < 0)
return ioctl_fd;
if (status != 0)
log_unit_debug_errno(UNIT(a), status, "Sending failure: %m");
else
log_unit_debug(UNIT(a), "Sending success.");
r = 0;
/* Autofs thankfully does not hand out 0 as a token */
while ((token = PTR_TO_UINT(set_steal_first(tokens)))) {
int k;
/* Autofs fun fact II:
*
* if you pass a positive status code here, the kernel will
* freeze! Yay! */
k = autofs_send_ready(UNIT(a)->manager->dev_autofs_fd,
ioctl_fd,
token,
status);
if (k < 0)
r = k;
}
return r;
}
bool exit_status_set_is_empty(ExitStatusSet *x) {
if (!x)
return true;
return set_isempty(x->status) && set_isempty(x->signal);
}
/* Use this function only if you do not have direct access to /proc/self/mountinfo but the caller can open it
* for you. This is the case when /proc is masked or not mounted. Otherwise, use bind_remount_recursive. */
int bind_remount_recursive_with_mountinfo(
const char *prefix,
unsigned long new_flags,
unsigned long flags_mask,
char **blacklist,
FILE *proc_self_mountinfo) {
_cleanup_set_free_free_ Set *done = NULL;
_cleanup_free_ char *cleaned = NULL;
int r;
assert(proc_self_mountinfo);
/* Recursively remount a directory (and all its submounts) read-only or read-write. If the directory is already
* mounted, we reuse the mount and simply mark it MS_BIND|MS_RDONLY (or remove the MS_RDONLY for read-write
* operation). If it isn't we first make it one. Afterwards we apply MS_BIND|MS_RDONLY (or remove MS_RDONLY) to
* all submounts we can access, too. When mounts are stacked on the same mount point we only care for each
* individual "top-level" mount on each point, as we cannot influence/access the underlying mounts anyway. We
* do not have any effect on future submounts that might get propagated, they migt be writable. This includes
* future submounts that have been triggered via autofs.
*
* If the "blacklist" parameter is specified it may contain a list of subtrees to exclude from the
* remount operation. Note that we'll ignore the blacklist for the top-level path. */
cleaned = strdup(prefix);
if (!cleaned)
return -ENOMEM;
path_simplify(cleaned, false);
done = set_new(&path_hash_ops);
if (!done)
return -ENOMEM;
for (;;) {
_cleanup_set_free_free_ Set *todo = NULL;
bool top_autofs = false;
char *x;
unsigned long orig_flags;
todo = set_new(&path_hash_ops);
if (!todo)
return -ENOMEM;
rewind(proc_self_mountinfo);
for (;;) {
_cleanup_free_ char *path = NULL, *p = NULL, *type = NULL;
int k;
k = fscanf(proc_self_mountinfo,
"%*s " /* (1) mount id */
"%*s " /* (2) parent id */
"%*s " /* (3) major:minor */
"%*s " /* (4) root */
"%ms " /* (5) mount point */
"%*s" /* (6) mount options (superblock) */
"%*[^-]" /* (7) optional fields */
"- " /* (8) separator */
"%ms " /* (9) file system type */
"%*s" /* (10) mount source */
"%*s" /* (11) mount options (bind mount) */
"%*[^\n]", /* some rubbish at the end */
&path,
&type);
if (k != 2) {
if (k == EOF)
break;
continue;
}
r = cunescape(path, UNESCAPE_RELAX, &p);
if (r < 0)
return r;
if (!path_startswith(p, cleaned))
continue;
/* Ignore this mount if it is blacklisted, but only if it isn't the top-level mount we shall
* operate on. */
if (!path_equal(cleaned, p)) {
bool blacklisted = false;
char **i;
STRV_FOREACH(i, blacklist) {
if (path_equal(*i, cleaned))
continue;
if (!path_startswith(*i, cleaned))
continue;
if (path_startswith(p, *i)) {
blacklisted = true;
log_debug("Not remounting %s blacklisted by %s, called for %s", p, *i, cleaned);
break;
}
}
if (blacklisted)
continue;
}
/* Let's ignore autofs mounts. If they aren't
* triggered yet, we want to avoid triggering
* them, as we don't make any guarantees for
* future submounts anyway. If they are
* already triggered, then we will find
* another entry for this. */
if (streq(type, "autofs")) {
top_autofs = top_autofs || path_equal(cleaned, p);
continue;
}
if (!set_contains(done, p)) {
r = set_consume(todo, p);
p = NULL;
if (r == -EEXIST)
continue;
if (r < 0)
return r;
}
}
/* If we have no submounts to process anymore and if
* the root is either already done, or an autofs, we
* are done */
if (set_isempty(todo) &&
(top_autofs || set_contains(done, cleaned)))
return 0;
if (!set_contains(done, cleaned) &&
!set_contains(todo, cleaned)) {
/* The prefix directory itself is not yet a mount, make it one. */
if (mount(cleaned, cleaned, NULL, MS_BIND|MS_REC, NULL) < 0)
return -errno;
orig_flags = 0;
(void) get_mount_flags(cleaned, &orig_flags);
orig_flags &= ~MS_RDONLY;
if (mount(NULL, cleaned, NULL, (orig_flags & ~flags_mask)|MS_BIND|MS_REMOUNT|new_flags, NULL) < 0)
return -errno;
log_debug("Made top-level directory %s a mount point.", prefix);
r = set_put_strdup(done, cleaned);
if (r < 0)
return r;
}
while ((x = set_steal_first(todo))) {
r = set_consume(done, x);
if (IN_SET(r, 0, -EEXIST))
continue;
if (r < 0)
return r;
/* Deal with mount points that are obstructed by a later mount */
r = path_is_mount_point(x, NULL, 0);
if (IN_SET(r, 0, -ENOENT))
continue;
if (IN_SET(r, -EACCES, -EPERM)) {
/* Even if root user invoke this, submounts under private FUSE or NFS mount points
* may not be acceessed. E.g.,
*
* $ bindfs --no-allow-other ~/mnt/mnt ~/mnt/mnt
* $ bindfs --no-allow-other ~/mnt ~/mnt
*
* Then, root user cannot access the mount point ~/mnt/mnt.
* In such cases, the submounts are ignored, as we have no way to manage them. */
log_debug_errno(r, "Failed to determine '%s' is mount point or not, ignoring: %m", x);
continue;
}
if (r < 0)
return r;
/* Try to reuse the original flag set */
orig_flags = 0;
(void) get_mount_flags(x, &orig_flags);
orig_flags &= ~MS_RDONLY;
if (mount(NULL, x, NULL, (orig_flags & ~flags_mask)|MS_BIND|MS_REMOUNT|new_flags, NULL) < 0)
return -errno;
log_debug("Remounted %s read-only.", x);
}
}
static void wait_for_children(Set *pids, sigset_t *mask) {
usec_t until;
assert(mask);
if (set_isempty(pids))
return;
until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC;
for (;;) {
struct timespec ts;
int k;
usec_t n;
void *p;
Iterator i;
/* First, let the kernel inform us about killed
* children. Most processes will probably be our
* children, but some are not (might be our
* grandchildren instead...). */
for (;;) {
pid_t pid;
pid = waitpid(-1, NULL, WNOHANG);
if (pid == 0)
break;
if (pid < 0) {
if (errno == ECHILD)
break;
log_error("waitpid() failed: %m");
return;
}
set_remove(pids, ULONG_TO_PTR(pid));
}
/* Now explicitly check who might be remaining, who
* might not be our child. */
SET_FOREACH(p, pids, i) {
/* We misuse getpgid as a check whether a
* process still exists. */
if (getpgid((pid_t) PTR_TO_ULONG(p)) >= 0)
continue;
if (errno != ESRCH)
continue;
set_remove(pids, p);
}
if (set_isempty(pids))
return;
n = now(CLOCK_MONOTONIC);
if (n >= until)
return;
timespec_store(&ts, until - n);
k = sigtimedwait(mask, NULL, &ts);
if (k != SIGCHLD) {
if (k < 0 && errno != EAGAIN) {
log_error("sigtimedwait() failed: %m");
return;
}
if (k >= 0)
log_warning("sigtimedwait() returned unexpected signal.");
}
}
}
int dns_transaction_go(DnsTransaction *t) {
bool had_stream;
usec_t ts;
int r;
assert(t);
had_stream = !!t->stream;
dns_transaction_stop(t);
log_debug("Excercising transaction on scope %s on %s/%s",
dns_protocol_to_string(t->scope->protocol),
t->scope->link ? t->scope->link->name : "*",
t->scope->family == AF_UNSPEC ? "*" : af_to_name(t->scope->family));
if (t->n_attempts >= TRANSACTION_ATTEMPTS_MAX(t->scope->protocol)) {
dns_transaction_complete(t, DNS_TRANSACTION_ATTEMPTS_MAX_REACHED);
return 0;
}
if (t->scope->protocol == DNS_PROTOCOL_LLMNR && had_stream) {
/* If we already tried via a stream, then we don't
* retry on LLMNR. See RFC 4795, Section 2.7. */
dns_transaction_complete(t, DNS_TRANSACTION_ATTEMPTS_MAX_REACHED);
return 0;
}
assert_se(sd_event_now(t->scope->manager->event, clock_boottime_or_monotonic(), &ts) >= 0);
t->n_attempts++;
t->start_usec = ts;
t->received = dns_packet_unref(t->received);
t->cached = dns_answer_unref(t->cached);
t->cached_rcode = 0;
/* Check the cache, but only if this transaction is not used
* for probing or verifying a zone item. */
if (set_isempty(t->zone_items)) {
/* Before trying the cache, let's make sure we figured out a
* server to use. Should this cause a change of server this
* might flush the cache. */
dns_scope_get_dns_server(t->scope);
/* Let's then prune all outdated entries */
dns_cache_prune(&t->scope->cache);
r = dns_cache_lookup(&t->scope->cache, t->key, &t->cached_rcode, &t->cached);
if (r < 0)
return r;
if (r > 0) {
if (t->cached_rcode == DNS_RCODE_SUCCESS)
dns_transaction_complete(t, DNS_TRANSACTION_SUCCESS);
else
dns_transaction_complete(t, DNS_TRANSACTION_FAILURE);
return 0;
}
}
if (t->scope->protocol == DNS_PROTOCOL_LLMNR && !t->initial_jitter) {
usec_t jitter;
/* RFC 4795 Section 2.7 suggests all queries should be
* delayed by a random time from 0 to JITTER_INTERVAL. */
t->initial_jitter = true;
random_bytes(&jitter, sizeof(jitter));
jitter %= LLMNR_JITTER_INTERVAL_USEC;
r = sd_event_add_time(
t->scope->manager->event,
&t->timeout_event_source,
clock_boottime_or_monotonic(),
ts + jitter,
LLMNR_JITTER_INTERVAL_USEC,
on_transaction_timeout, t);
if (r < 0)
return r;
t->n_attempts = 0;
t->state = DNS_TRANSACTION_PENDING;
log_debug("Delaying LLMNR transaction for " USEC_FMT "us.", jitter);
return 0;
}
/* Otherwise, we need to ask the network */
r = dns_transaction_make_packet(t);
if (r == -EDOM) {
/* Not the right request to make on this network?
* (i.e. an A request made on IPv6 or an AAAA request
* made on IPv4, on LLMNR or mDNS.) */
dns_transaction_complete(t, DNS_TRANSACTION_NO_SERVERS);
return 0;
}
if (r < 0)
return r;
if (t->scope->protocol == DNS_PROTOCOL_LLMNR &&
(dns_name_endswith(DNS_RESOURCE_KEY_NAME(t->key), "in-addr.arpa") > 0 ||
dns_name_endswith(DNS_RESOURCE_KEY_NAME(t->key), "ip6.arpa") > 0)) {
/* RFC 4795, Section 2.4. says reverse lookups shall
* always be made via TCP on LLMNR */
r = dns_transaction_open_tcp(t);
} else {
/* Try via UDP, and if that fails due to large size try via TCP */
r = dns_transaction_emit(t);
if (r == -EMSGSIZE)
r = dns_transaction_open_tcp(t);
}
if (r == -ESRCH) {
/* No servers to send this to? */
dns_transaction_complete(t, DNS_TRANSACTION_NO_SERVERS);
return 0;
} else if (r < 0) {
if (t->scope->protocol != DNS_PROTOCOL_DNS) {
dns_transaction_complete(t, DNS_TRANSACTION_RESOURCES);
return 0;
}
/* Couldn't send? Try immediately again, with a new server */
dns_transaction_next_dns_server(t);
return dns_transaction_go(t);
}
r = sd_event_add_time(
t->scope->manager->event,
&t->timeout_event_source,
clock_boottime_or_monotonic(),
ts + transaction_get_resend_timeout(t), 0,
on_transaction_timeout, t);
if (r < 0)
return r;
t->state = DNS_TRANSACTION_PENDING;
return 1;
}
bool fdset_isempty(FDSet *fds) {
return set_isempty(MAKE_SET(fds));
}
static void test_struct(void) {
Prioq *q;
Set *s;
unsigned previous = 0, i;
int r;
srand(0);
q = prioq_new(test_compare);
assert_se(q);
s = set_new(test_hash, test_compare);
assert_se(s);
for (i = 0; i < SET_SIZE; i++) {
struct test *t;
t = new0(struct test, 1);
assert_se(t);
t->value = (unsigned) rand();
r = prioq_put(q, t, &t->idx);
assert_se(r >= 0);
if (i % 4 == 0) {
r = set_consume(s, t);
assert_se(r >= 0);
}
}
for (;;) {
struct test *t;
t = set_steal_first(s);
if (!t)
break;
r = prioq_remove(q, t, &t->idx);
assert_se(r > 0);
free(t);
}
for (i = 0; i < SET_SIZE * 3 / 4; i++) {
struct test *t;
assert_se(prioq_size(q) == (SET_SIZE * 3 / 4) - i);
t = prioq_pop(q);
assert_se(t);
assert_se(previous <= t->value);
previous = t->value;
free(t);
}
assert_se(prioq_isempty(q));
prioq_free(q);
assert_se(set_isempty(s));
set_free(s);
}
