#include <limits.h>
#include <langinfo.h>
#include <locale.h>
+#include <sys/personality.h>
#include <libgen.h>
#undef basename
+#ifdef HAVE_SYS_AUXV_H
+#include <sys/auxv.h>
+#endif
+
#include "macro.h"
#include "util.h"
#include "ioprio.h"
static volatile unsigned cached_lines = 0;
size_t page_size(void) {
- static __thread size_t pgsz = 0;
+ static thread_local size_t pgsz = 0;
long r;
if (_likely_(pgsz > 0))
return 0;
}
+static size_t strcspn_escaped(const char *s, const char *reject) {
+ bool escaped = false;
+ size_t n;
+
+ for (n=0; s[n]; n++) {
+ if (escaped)
+ escaped = false;
+ else if (s[n] == '\\')
+ escaped = true;
+ else if (strchr(reject, s[n]))
+ return n;
+ }
+ return n;
+}
+
/* Split a string into words. */
-char *split(const char *c, size_t *l, const char *separator, char **state) {
+char *split(const char *c, size_t *l, const char *separator, bool quoted, char **state) {
char *current;
current = *state ? *state : (char*) c;
return NULL;
current += strspn(current, separator);
- *l = strcspn(current, separator);
- *state = current+*l;
-
- return (char*) current;
-}
-
-/* Split a string into words, but consider strings enclosed in '' and
- * "" as words even if they include spaces. */
-char *split_quoted(const char *c, size_t *l, char **state) {
- const char *current, *e;
- bool escaped = false;
-
- assert(c);
- assert(l);
- assert(state);
-
- current = *state ? *state : c;
-
- current += strspn(current, WHITESPACE);
-
- if (*current == 0)
+ if (!*current)
return NULL;
- else if (*current == '\'') {
- current ++;
-
- for (e = current; *e; e++) {
- if (escaped)
- escaped = false;
- else if (*e == '\\')
- escaped = true;
- else if (*e == '\'')
- break;
- }
-
- *l = e-current;
- *state = (char*) (*e == 0 ? e : e+1);
-
- } else if (*current == '\"') {
- current ++;
-
- for (e = current; *e; e++) {
- if (escaped)
- escaped = false;
- else if (*e == '\\')
- escaped = true;
- else if (*e == '\"')
- break;
- }
-
- *l = e-current;
- *state = (char*) (*e == 0 ? e : e+1);
-
+ if (quoted && strchr("\'\"", *current)) {
+ char quotechar = *(current++);
+ *l = strcspn_escaped(current, (char[]){quotechar, '\0'});
+ *state = current+*l+1;
+ } else if (quoted) {
+ *l = strcspn_escaped(current, separator);
+ *state = current+*l;
} else {
- for (e = current; *e; e++) {
- if (escaped)
- escaped = false;
- else if (*e == '\\')
- escaped = true;
- else if (strchr(WHITESPACE, *e))
- break;
- }
- *l = e-current;
- *state = (char*) e;
+ *l = strcspn(current, separator);
+ *state = current+*l;
}
return (char*) current;
int get_parent_of_pid(pid_t pid, pid_t *_ppid) {
int r;
- _cleanup_fclose_ FILE *f = NULL;
- char line[LINE_MAX];
+ _cleanup_free_ char *line = NULL;
long unsigned ppid;
const char *p;
}
p = procfs_file_alloca(pid, "stat");
- f = fopen(p, "re");
- if (!f)
- return -errno;
-
- if (!fgets(line, sizeof(line), f)) {
- r = feof(f) ? -EIO : -errno;
+ r = read_one_line_file(p, &line);
+ if (r < 0)
return r;
- }
/* Let's skip the pid and comm fields. The latter is enclosed
* in () but does not escape any () in its value, so let's
}
int get_starttime_of_pid(pid_t pid, unsigned long long *st) {
- _cleanup_fclose_ FILE *f = NULL;
- char line[LINE_MAX];
+ int r;
+ _cleanup_free_ char *line = NULL;
const char *p;
assert(pid >= 0);
assert(st);
- if (pid == 0)
- p = "/proc/self/stat";
- else
- p = procfs_file_alloca(pid, "stat");
-
- f = fopen(p, "re");
- if (!f)
- return errno == ENOENT ? -ESRCH : -errno;
-
- if (!fgets(line, sizeof(line), f)) {
- if (ferror(f))
- return -errno;
-
- return -EIO;
- }
+ p = procfs_file_alloca(pid, "stat");
+ r = read_one_line_file(p, &line);
+ if (r < 0)
+ return r;
/* Let's skip the pid and comm fields. The latter is enclosed
* in () but does not escape any () in its value, so let's
return s;
}
+int get_process_state(pid_t pid) {
+ const char *p;
+ char state;
+ int r;
+ _cleanup_free_ char *line = NULL;
+
+ assert(pid >= 0);
+
+ p = procfs_file_alloca(pid, "stat");
+ r = read_one_line_file(p, &line);
+ if (r < 0)
+ return r;
+
+ p = strrchr(line, ')');
+ if (!p)
+ return -EIO;
+
+ p++;
+
+ if (sscanf(p, " %c", &state) != 1)
+ return -EIO;
+
+ return (unsigned char) state;
+}
+
int get_process_comm(pid_t pid, char **name) {
const char *p;
int r;
assert(name);
assert(pid >= 0);
- if (pid == 0)
- p = "/proc/self/comm";
- else
- p = procfs_file_alloca(pid, "comm");
+ p = procfs_file_alloca(pid, "comm");
r = read_one_line_file(p, name);
if (r == -ENOENT)
assert(line);
assert(pid >= 0);
- if (pid == 0)
- p = "/proc/self/cmdline";
- else
- p = procfs_file_alloca(pid, "cmdline");
+ p = procfs_file_alloca(pid, "cmdline");
f = fopen(p, "re");
if (!f)
assert(capeff);
assert(pid >= 0);
- if (pid == 0)
- p = "/proc/self/status";
- else
- p = procfs_file_alloca(pid, "status");
+ p = procfs_file_alloca(pid, "status");
return get_status_field(p, "\nCapEff:", capeff);
}
assert(pid >= 0);
assert(name);
- if (pid == 0)
- p = "/proc/self/exe";
- else
- p = procfs_file_alloca(pid, "exe");
+ p = procfs_file_alloca(pid, "exe");
r = readlink_malloc(p, name);
if (r < 0)
return strnappend(s, suffix, suffix ? strlen(suffix) : 0);
}
-int readlink_malloc(const char *p, char **r) {
+int readlink_malloc(const char *p, char **ret) {
size_t l = 100;
+ int r;
assert(p);
- assert(r);
+ assert(ret);
for (;;) {
char *c;
ssize_t n;
- if (!(c = new(char, l)))
+ c = new(char, l);
+ if (!c)
return -ENOMEM;
- if ((n = readlink(p, c, l-1)) < 0) {
- int ret = -errno;
+ n = readlink(p, c, l-1);
+ if (n < 0) {
+ r = -errno;
free(c);
- return ret;
+ return r;
}
if ((size_t) n < l-1) {
c[n] = 0;
- *r = c;
+ *ret = c;
return 0;
}
}
ssize_t loop_read(int fd, void *buf, size_t nbytes, bool do_poll) {
- uint8_t *p;
+ uint8_t *p = buf;
ssize_t n = 0;
assert(fd >= 0);
assert(buf);
- p = buf;
-
while (nbytes > 0) {
ssize_t k;
- if ((k = read(fd, p, nbytes)) <= 0) {
-
- if (k < 0 && errno == EINTR)
- continue;
-
- if (k < 0 && errno == EAGAIN && do_poll) {
- struct pollfd pollfd = {
- .fd = fd,
- .events = POLLIN,
- };
-
- if (poll(&pollfd, 1, -1) < 0) {
- if (errno == EINTR)
- continue;
+ k = read(fd, p, nbytes);
+ if (k < 0 && errno == EINTR)
+ continue;
- return n > 0 ? n : -errno;
- }
+ if (k < 0 && errno == EAGAIN && do_poll) {
- /* We knowingly ignore the revents value here,
- * and expect that any error/EOF is reported
- * via read()/write()
- */
+ /* We knowingly ignore any return value here,
+ * and expect that any error/EOF is reported
+ * via read() */
- continue;
- }
+ fd_wait_for_event(fd, POLLIN, (usec_t) -1);
+ continue;
+ }
+ if (k <= 0)
return n > 0 ? n : (k < 0 ? -errno : 0);
- }
p += k;
nbytes -= k;
}
ssize_t loop_write(int fd, const void *buf, size_t nbytes, bool do_poll) {
- const uint8_t *p;
+ const uint8_t *p = buf;
ssize_t n = 0;
assert(fd >= 0);
assert(buf);
- p = buf;
-
while (nbytes > 0) {
ssize_t k;
k = write(fd, p, nbytes);
- if (k <= 0) {
-
- if (k < 0 && errno == EINTR)
- continue;
-
- if (k < 0 && errno == EAGAIN && do_poll) {
- struct pollfd pollfd = {
- .fd = fd,
- .events = POLLOUT,
- };
+ if (k < 0 && errno == EINTR)
+ continue;
- if (poll(&pollfd, 1, -1) < 0) {
- if (errno == EINTR)
- continue;
+ if (k < 0 && errno == EAGAIN && do_poll) {
- return n > 0 ? n : -errno;
- }
+ /* We knowingly ignore any return value here,
+ * and expect that any error/EOF is reported
+ * via write() */
- /* We knowingly ignore the revents value here,
- * and expect that any error/EOF is reported
- * via read()/write()
- */
-
- continue;
- }
+ fd_wait_for_event(fd, POLLOUT, (usec_t) -1);
+ continue;
+ }
+ if (k <= 0)
return n > 0 ? n : (k < 0 ? -errno : 0);
- }
p += k;
nbytes -= k;
int dir_is_empty(const char *path) {
_cleanup_closedir_ DIR *d;
- int r;
d = opendir(path);
if (!d)
for (;;) {
struct dirent *de;
- union dirent_storage buf;
- r = readdir_r(d, &buf.de, &de);
- if (r > 0)
- return -r;
+ errno = 0;
+ de = readdir(d);
+ if (!de && errno != 0)
+ return -errno;
if (!de)
return 1;
return dir;
}
-unsigned long long random_ull(void) {
+int dev_urandom(void *p, size_t n) {
_cleanup_close_ int fd;
- uint64_t ull;
- ssize_t r;
+ ssize_t k;
fd = open("/dev/urandom", O_RDONLY|O_CLOEXEC|O_NOCTTY);
if (fd < 0)
- goto fallback;
-
- r = loop_read(fd, &ull, sizeof(ull), true);
- if (r != sizeof(ull))
- goto fallback;
+ return errno == ENOENT ? -ENOSYS : -errno;
- return ull;
+ k = loop_read(fd, p, n, true);
+ if (k < 0)
+ return (int) k;
+ if ((size_t) k != n)
+ return -EIO;
-fallback:
- return random() * RAND_MAX + random();
+ return 0;
}
-unsigned random_u(void) {
- _cleanup_close_ int fd;
- unsigned u;
- ssize_t r;
+void random_bytes(void *p, size_t n) {
+ static bool srand_called = false;
+ uint8_t *q;
+ int r;
- fd = open("/dev/urandom", O_RDONLY|O_CLOEXEC|O_NOCTTY);
- if (fd < 0)
- goto fallback;
+ r = dev_urandom(p, n);
+ if (r >= 0)
+ return;
+
+ /* If some idiot made /dev/urandom unavailable to us, he'll
+ * get a PRNG instead. */
- r = loop_read(fd, &u, sizeof(u), true);
- if (r != sizeof(u))
- goto fallback;
+ if (!srand_called) {
+ unsigned x = 0;
- return u;
+#ifdef HAVE_SYS_AUXV_H
+ /* The kernel provides us with a bit of entropy in
+ * auxv, so let's try to make use of that to seed the
+ * pseudo-random generator. It's better than
+ * nothing... */
-fallback:
- return random() * RAND_MAX + random();
+ void *auxv;
+
+ auxv = (void*) getauxval(AT_RANDOM);
+ if (auxv)
+ x ^= *(unsigned*) auxv;
+#endif
+
+ x ^= (unsigned) now(CLOCK_REALTIME);
+ x ^= (unsigned) gettid();
+
+ srand(x);
+ srand_called = true;
+ }
+
+ for (q = p; q < (uint8_t*) p + n; q ++)
+ *q = rand();
}
void rename_process(const char name[8]) {
if (!saved_argv[i])
break;
- memset(saved_argv[i], 0, strlen(saved_argv[i]));
+ memzero(saved_argv[i], strlen(saved_argv[i]));
}
}
}
}
int get_ctty_devnr(pid_t pid, dev_t *d) {
- _cleanup_fclose_ FILE *f = NULL;
- char line[LINE_MAX], *p;
+ int r;
+ _cleanup_free_ char *line = NULL;
+ const char *p;
unsigned long ttynr;
- const char *fn;
assert(pid >= 0);
- if (pid == 0)
- fn = "/proc/self/stat";
- else
- fn = procfs_file_alloca(pid, "stat");
-
- f = fopen(fn, "re");
- if (!f)
- return -errno;
-
- if (!fgets(line, sizeof(line), f))
- return feof(f) ? -EIO : -errno;
+ p = procfs_file_alloca(pid, "stat");
+ r = read_one_line_file(p, &line);
+ if (r < 0)
+ return r;
p = strrchr(line, ')');
if (!p)
}
int get_ctty(pid_t pid, dev_t *_devnr, char **r) {
- int k;
- char fn[sizeof("/dev/char/")-1 + 2*DECIMAL_STR_MAX(unsigned) + 1 + 1], *s, *b, *p;
+ char fn[sizeof("/dev/char/")-1 + 2*DECIMAL_STR_MAX(unsigned) + 1 + 1], *b = NULL;
+ _cleanup_free_ char *s = NULL;
+ const char *p;
dev_t devnr;
+ int k;
assert(r);
/* This is an ugly hack */
if (major(devnr) == 136) {
- if (asprintf(&b, "pts/%lu", (unsigned long) minor(devnr)) < 0)
- return -ENOMEM;
-
- *r = b;
- if (_devnr)
- *_devnr = devnr;
-
- return 0;
+ asprintf(&b, "pts/%lu", (unsigned long) minor(devnr));
+ goto finish;
}
/* Probably something like the ptys which have no
* vaguely useful. */
b = strdup(fn + 5);
- if (!b)
- return -ENOMEM;
-
- *r = b;
- if (_devnr)
- *_devnr = devnr;
-
- return 0;
+ goto finish;
}
if (startswith(s, "/dev/"))
p = s;
b = strdup(p);
- free(s);
+finish:
if (!b)
return -ENOMEM;
for (;;) {
struct dirent *de;
- union dirent_storage buf;
bool is_dir, keep_around;
struct stat st;
int r;
- r = readdir_r(d, &buf.de, &de);
- if (r != 0 && ret == 0) {
- ret = -r;
+ errno = 0;
+ de = readdir(d);
+ if (!de && errno != 0) {
+ if (ret == 0)
+ ret = -errno;
break;
}
_pure_ static int is_temporary_fs(struct statfs *s) {
assert(s);
- return
- F_TYPE_EQUAL(s->f_type, TMPFS_MAGIC) ||
- F_TYPE_EQUAL(s->f_type, RAMFS_MAGIC);
+
+ return F_TYPE_EQUAL(s->f_type, TMPFS_MAGIC) ||
+ F_TYPE_EQUAL(s->f_type, RAMFS_MAGIC);
}
int rm_rf_children(int fd, bool only_dirs, bool honour_sticky, struct stat *root_dev) {
return r;
}
-int status_welcome(void) {
- _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
- int r;
-
- r = parse_env_file("/etc/os-release", NEWLINE,
- "PRETTY_NAME", &pretty_name,
- "ANSI_COLOR", &ansi_color,
- NULL);
-
- if (r < 0 && r != -ENOENT)
- log_warning("Failed to read /etc/os-release: %s", strerror(-r));
-
- return status_printf(NULL, false, false,
- "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
- isempty(ansi_color) ? "1" : ansi_color,
- isempty(pretty_name) ? "Linux" : pretty_name);
-}
-
char *replace_env(const char *format, char **env) {
enum {
WORD,
return -EPROTO;
}
-_noreturn_ void freeze(void) {
+noreturn void freeze(void) {
/* Make sure nobody waits for us on a socket anymore */
close_all_fds(NULL, 0);
}
int pipe_eof(int fd) {
- int r;
struct pollfd pollfd = {
.fd = fd,
.events = POLLIN|POLLHUP,
};
+ int r;
+
r = poll(&pollfd, 1, 0);
if (r < 0)
return -errno;
}
int fd_wait_for_event(int fd, int event, usec_t t) {
- int r;
+
struct pollfd pollfd = {
.fd = fd,
.events = event,
};
- r = poll(&pollfd, 1, t == (usec_t) -1 ? -1 : (int) (t / USEC_PER_MSEC));
+ struct timespec ts;
+ int r;
+
+ r = ppoll(&pollfd, 1, t == (usec_t) -1 ? NULL : timespec_store(&ts, t), NULL);
if (r < 0)
return -errno;
t[k] = '.';
stpcpy(stpcpy(t+k+1, fn), "XXXXXX");
- fd = mkostemp(t, O_WRONLY|O_CLOEXEC);
+ fd = mkostemp_safe(t, O_WRONLY|O_CLOEXEC);
if (fd < 0) {
free(t);
return -errno;
_cleanup_free_ char *t;
const char *fn;
size_t k;
- unsigned long long ull;
+ uint64_t u;
unsigned i;
int r;
t[k] = '.';
x = stpcpy(t+k+1, fn);
- ull = random_ull();
+ u = random_u64();
for (i = 0; i < 16; i++) {
- *(x++) = hexchar(ull & 0xF);
- ull >>= 4;
+ *(x++) = hexchar(u & 0xF);
+ u >>= 4;
}
*x = 0;
for (;;) {
struct dirent *de;
- union dirent_storage buf;
- int k;
- k = readdir_r(d, &buf.de, &de);
- assert(k >= 0);
- if (k > 0)
- return -k;
+ errno = 0;
+ de = readdir(d);
+ if (!de && errno != 0)
+ return -errno;
if (!de)
break;
}
bool is_main_thread(void) {
- static __thread int cached = 0;
+ static thread_local int cached = 0;
if (_unlikely_(cached == 0))
cached = getpid() == gettid() ? 1 : -1;
DEFINE_PRIVATE_STRING_TABLE_LOOKUP(__signal, int);
const char *signal_to_string(int signo) {
- static __thread char buf[sizeof("RTMIN+")-1 + DECIMAL_STR_MAX(int) + 1];
+ static thread_local char buf[sizeof("RTMIN+")-1 + DECIMAL_STR_MAX(int) + 1];
const char *name;
name = __signal_to_string(signo);
socklen_t l = sizeof(value);
r = getsockopt(fd, SOL_SOCKET, SO_SNDBUF, &value, &l);
- if (r >= 0 &&
- l == sizeof(value) &&
- (size_t) value >= n*2)
+ if (r >= 0 && l == sizeof(value) && (size_t) value >= n*2)
return 0;
+ /* If we have the privileges we will ignore the kernel limit. */
+
value = (int) n;
- r = setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &value, sizeof(value));
- if (r < 0)
- return -errno;
+ if (setsockopt(fd, SOL_SOCKET, SO_SNDBUFFORCE, &value, sizeof(value)) < 0)
+ if (setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &value, sizeof(value)) < 0)
+ return -errno;
return 1;
}
socklen_t l = sizeof(value);
r = getsockopt(fd, SOL_SOCKET, SO_RCVBUF, &value, &l);
- if (r >= 0 &&
- l == sizeof(value) &&
- (size_t) value >= n*2)
+ if (r >= 0 && l == sizeof(value) && (size_t) value >= n*2)
return 0;
- value = (int) n;
- r = setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &value, sizeof(value));
- if (r < 0)
- return -errno;
+ /* If we have the privileges we will ignore the kernel limit. */
+ value = (int) n;
+ if (setsockopt(fd, SOL_SOCKET, SO_RCVBUFFORCE, &value, sizeof(value)) < 0)
+ if (setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &value, sizeof(value)) < 0)
+ return -errno;
return 1;
}
assert(field);
assert(_value);
- if (pid == 0)
- path = "/proc/self/environ";
- else
- path = procfs_file_alloca(pid, "environ");
+ path = procfs_file_alloca(pid, "environ");
f = fopen(path, "re");
if (!f)
}
bool in_initrd(void) {
- static __thread int saved = -1;
+ static int saved = -1;
struct statfs s;
if (saved >= 0)
for (;;) {
struct dirent *de;
- union dirent_storage buf;
_cleanup_close_ int fd = -1, device = -1;
char contents[6];
ssize_t n;
- int k;
- k = readdir_r(d, &buf.de, &de);
- if (k != 0)
- return -k;
+ errno = 0;
+ de = readdir(d);
+ if (!de && errno != 0)
+ return -errno;
if (!de)
break;
assert(mode);
assert(_f);
- if (!path_strv_canonicalize_uniq(search))
+ if (!path_strv_canonicalize_absolute_uniq(search, NULL))
return -ENOMEM;
STRV_FOREACH(i, search) {
return NULL;
if (*allocated > prev)
- memset(&q[prev], 0, *allocated - prev);
+ memzero(&q[prev], *allocated - prev);
return q;
}
return true;
}
-void parse_user_at_host(char *arg, char **user, char **host) {
- assert(arg);
- assert(user);
- assert(host);
-
- *host = strchr(arg, '@');
- if (*host == NULL)
- *host = arg;
- else {
- *host[0]++ = '\0';
- *user = arg;
- }
-}
-
int split_pair(const char *s, const char *sep, char **l, char **r) {
char *x, *a, *b;
int r;
if (detect_container(NULL) > 0) {
- *ret = NULL;
- return 0;
+ char *buf = NULL, *p;
+ size_t sz = 0;
+
+ r = read_full_file("/proc/1/cmdline", &buf, &sz);
+ if (r < 0)
+ return r;
+
+ for (p = buf; p + 1 < buf + sz; p++)
+ if (*p == 0)
+ *p = ' ';
+
+ *p = 0;
+ *ret = buf;
+ return 1;
}
r = read_one_line_file("/proc/cmdline", ret);
return 1;
}
+int parse_proc_cmdline(int (*parse_word)(const char *word)) {
+ _cleanup_free_ char *line = NULL;
+ char *w, *state;
+ size_t l;
+ int r;
+
+ r = proc_cmdline(&line);
+ if (r < 0)
+ log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r));
+ if (r <= 0)
+ return 0;
+
+ FOREACH_WORD_QUOTED(w, l, line, state) {
+ _cleanup_free_ char *word;
+
+ word = strndup(w, l);
+ if (!word)
+ return log_oom();
+
+ r = parse_word(word);
+ if (r < 0) {
+ log_error("Failed on cmdline argument %s: %s", word, strerror(-r));
+ return r;
+ }
+ }
+
+ return 0;
+}
+
int container_get_leader(const char *machine, pid_t *pid) {
_cleanup_free_ char *s = NULL, *class = NULL;
const char *p;
return 0;
}
-int namespace_open(pid_t pid, int *namespace_fd, int *root_fd) {
- _cleanup_close_ int nsfd = -1;
- const char *ns, *root;
+int namespace_open(pid_t pid, int *pidns_fd, int *mntns_fd, int *root_fd) {
+ _cleanup_close_ int pidnsfd = -1, mntnsfd = -1;
+ const char *pidns, *mntns, *root;
int rfd;
assert(pid >= 0);
- assert(namespace_fd);
+ assert(pidns_fd);
+ assert(mntns_fd);
assert(root_fd);
- ns = procfs_file_alloca(pid, "ns/mnt");
- nsfd = open(ns, O_RDONLY|O_NOCTTY|O_CLOEXEC);
- if (nsfd < 0)
+ mntns = procfs_file_alloca(pid, "ns/mnt");
+ mntnsfd = open(mntns, O_RDONLY|O_NOCTTY|O_CLOEXEC);
+ if (mntnsfd < 0)
+ return -errno;
+
+ pidns = procfs_file_alloca(pid, "ns/pid");
+ pidnsfd = open(pidns, O_RDONLY|O_NOCTTY|O_CLOEXEC);
+ if (pidnsfd < 0)
return -errno;
root = procfs_file_alloca(pid, "root");
if (rfd < 0)
return -errno;
- *namespace_fd = nsfd;
+ *pidns_fd = pidnsfd;
+ *mntns_fd = mntnsfd;
*root_fd = rfd;
- nsfd = -1;
+ pidnsfd = -1;
+ mntnsfd = -1;
return 0;
}
-int namespace_enter(int namespace_fd, int root_fd) {
- assert(namespace_fd >= 0);
+int namespace_enter(int pidns_fd, int mntns_fd, int root_fd) {
+ assert(pidns_fd >= 0);
+ assert(mntns_fd >= 0);
assert(root_fd >= 0);
- if (setns(namespace_fd, CLONE_NEWNS) < 0)
+ if (setns(pidns_fd, CLONE_NEWPID) < 0)
+ return -errno;
+
+ if (setns(mntns_fd, CLONE_NEWNS) < 0)
return -errno;
if (fchdir(root_fd) < 0)
return 0;
}
+
+bool pid_is_unwaited(pid_t pid) {
+ /* Checks whether a PID is still valid at all, including a zombie */
+
+ if (pid <= 0)
+ return false;
+
+ if (kill(pid, 0) >= 0)
+ return true;
+
+ return errno != ESRCH;
+}
+
+bool pid_is_alive(pid_t pid) {
+ int r;
+
+ /* Checks whether a PID is still valid and not a zombie */
+
+ if (pid <= 0)
+ return false;
+
+ r = get_process_state(pid);
+ if (r == -ENOENT || r == 'Z')
+ return false;
+
+ return true;
+}
+
+int getpeercred(int fd, struct ucred *ucred) {
+ socklen_t n = sizeof(struct ucred);
+ struct ucred u;
+ int r;
+
+ assert(fd >= 0);
+ assert(ucred);
+
+ r = getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &u, &n);
+ if (r < 0)
+ return -errno;
+
+ if (n != sizeof(struct ucred))
+ return -EIO;
+
+ /* Check if the data is actually useful and not suppressed due
+ * to namespacing issues */
+ if (u.pid <= 0)
+ return -ENODATA;
+
+ *ucred = u;
+ return 0;
+}
+
+int getpeersec(int fd, char **ret) {
+ socklen_t n = 64;
+ char *s;
+ int r;
+
+ assert(fd >= 0);
+ assert(ret);
+
+ s = new0(char, n);
+ if (!s)
+ return -ENOMEM;
+
+ r = getsockopt(fd, SOL_SOCKET, SO_PEERSEC, s, &n);
+ if (r < 0) {
+ free(s);
+
+ if (errno != ERANGE)
+ return -errno;
+
+ s = new0(char, n);
+ if (!s)
+ return -ENOMEM;
+
+ r = getsockopt(fd, SOL_SOCKET, SO_PEERSEC, s, &n);
+ if (r < 0) {
+ free(s);
+ return -errno;
+ }
+ }
+
+ if (isempty(s)) {
+ free(s);
+ return -ENOTSUP;
+ }
+
+ *ret = s;
+ return 0;
+}
+
+/* This is much like like mkostemp() but is subject to umask(). */
+int mkostemp_safe(char *pattern, int flags) {
+ _cleanup_umask_ mode_t u;
+ int fd;
+
+ assert(pattern);
+
+ u = umask(077);
+
+ fd = mkostemp(pattern, flags);
+ if (fd < 0)
+ return -errno;
+
+ return fd;
+}
+
+int open_tmpfile(const char *path, int flags) {
+ char *p;
+ int fd;
+
+ assert(path);
+
+#ifdef O_TMPFILE
+ /* Try O_TMPFILE first, if it is supported */
+ fd = open(path, flags|O_TMPFILE, S_IRUSR|S_IWUSR);
+ if (fd >= 0)
+ return fd;
+#endif
+
+ /* Fall back to unguessable name + unlinking */
+ p = strappenda(path, "/systemd-tmp-XXXXXX");
+
+ fd = mkostemp_safe(p, flags);
+ if (fd < 0)
+ return fd;
+
+ unlink(p);
+ return fd;
+}
+
+int fd_warn_permissions(const char *path, int fd) {
+ struct stat st;
+
+ if (fstat(fd, &st) < 0)
+ return -errno;
+
+ if (st.st_mode & 0111)
+ log_warning("Configuration file %s is marked executable. Please remove executable permission bits. Proceeding anyway.", path);
+
+ if (st.st_mode & 0002)
+ log_warning("Configuration file %s is marked world-writable. Please remove world writability permission bits. Proceeding anyway.", path);
+
+ if (getpid() == 1 && (st.st_mode & 0044) != 0044)
+ log_warning("Configuration file %s is marked world-inaccessible. This has no effect as configuration data is accessible via APIs without restrictions. Proceeding anyway.", path);
+
+ return 0;
+}
+
+unsigned long personality_from_string(const char *p) {
+
+ /* Parse a personality specifier. We introduce our own
+ * identifiers that indicate specific ABIs, rather than just
+ * hints regarding the register size, since we want to keep
+ * things open for multiple locally supported ABIs for the
+ * same register size. We try to reuse the ABI identifiers
+ * used by libseccomp. */
+
+#if defined(__x86_64__)
+
+ if (streq(p, "x86"))
+ return PER_LINUX32;
+
+ if (streq(p, "x86-64"))
+ return PER_LINUX;
+
+#elif defined(__i386__)
+
+ if (streq(p, "x86"))
+ return PER_LINUX;
+#endif
+
+ /* personality(7) documents that 0xffffffffUL is used for
+ * querying the current personality, hence let's use that here
+ * as error indicator. */
+ return 0xffffffffUL;
+}
+
+const char* personality_to_string(unsigned long p) {
+
+#if defined(__x86_64__)
+
+ if (p == PER_LINUX32)
+ return "x86";
+
+ if (p == PER_LINUX)
+ return "x86-64";
+
+#elif defined(__i386__)
+
+ if (p == PER_LINUX)
+ return "x86";
+#endif
+
+ return NULL;
+}
~ianmdlvl / elogind.git / blobdiff