X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fshared%2Futil.c;h=e1a11684565f4bc8780422af492238271a988fa1;hp=38134ae521b61963de65667dc7920dffc9d9bf32;hb=e0333c7314e89c0bc268bd20c5e247a7c907ab34;hpb=5b12334d35eadf1f45cc3d631fd1a2e72ffaea0a diff --git a/src/shared/util.c b/src/shared/util.c index 38134ae52..e1a116845 100644 --- a/src/shared/util.c +++ b/src/shared/util.c @@ -58,7 +58,13 @@ #include #include #include +#include #include +#undef basename + +#ifdef HAVE_SYS_AUXV_H +#include +#endif #include "macro.h" #include "util.h" @@ -77,6 +83,7 @@ #include "gunicode.h" #include "virt.h" #include "def.h" +#include "missing.h" int saved_argc = 0; char **saved_argv = NULL; @@ -85,7 +92,7 @@ static volatile unsigned cached_columns = 0; 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)) @@ -354,8 +361,23 @@ int safe_atod(const char *s, double *ret_d) { 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; @@ -364,70 +386,19 @@ char *split(const char *c, size_t *l, const char *separator, char **state) { 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; @@ -435,8 +406,7 @@ char *split_quoted(const char *c, size_t *l, char **state) { 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; @@ -449,14 +419,9 @@ int get_parent_of_pid(pid_t pid, pid_t *_ppid) { } 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 @@ -483,28 +448,17 @@ int get_parent_of_pid(pid_t pid, pid_t *_ppid) { } 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 @@ -561,6 +515,31 @@ char *truncate_nl(char *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; @@ -568,10 +547,7 @@ int get_process_comm(pid_t pid, char **name) { 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) @@ -589,10 +565,7 @@ int get_process_cmdline(pid_t pid, size_t max_length, bool comm_fallback, char * 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) @@ -711,10 +684,7 @@ int get_process_capeff(pid_t pid, char **capeff) { 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); } @@ -727,10 +697,7 @@ int get_process_exe(pid_t pid, char **name) { 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) @@ -821,28 +788,31 @@ char *strappend(const char *s, const char *suffix) { 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; } @@ -951,16 +921,9 @@ char *delete_chars(char *s, const char *bad) { } bool in_charset(const char *s, const char* charset) { - const char *i; - assert(s); assert(charset); - - for (i = s; *i; i++) - if (!strchr(charset, *i)) - return false; - - return true; + return s[strspn(s, charset)] == '\0'; } char *file_in_same_dir(const char *path, const char *filename) { @@ -2097,45 +2060,31 @@ int close_pipe(int p[]) { } 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; @@ -2146,46 +2095,31 @@ ssize_t loop_read(int fd, void *buf, size_t nbytes, bool do_poll) { } 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 (poll(&pollfd, 1, -1) < 0) { - if (errno == EINTR) - continue; + 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 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; @@ -2195,30 +2129,71 @@ ssize_t loop_write(int fd, const void *buf, size_t nbytes, bool do_poll) { return n; } -int parse_bytes(const char *t, off_t *bytes) { - static const struct { +int parse_size(const char *t, off_t base, off_t *size) { + + /* Soo, sometimes we want to parse IEC binary suffxies, and + * sometimes SI decimal suffixes. This function can parse + * both. Which one is the right way depends on the + * context. Wikipedia suggests that SI is customary for + * hardrware metrics and network speeds, while IEC is + * customary for most data sizes used by software and volatile + * (RAM) memory. Hence be careful which one you pick! + * + * In either case we use just K, M, G as suffix, and not Ki, + * Mi, Gi or so (as IEC would suggest). That's because that's + * frickin' ugly. But this means you really need to make sure + * to document which base you are parsing when you use this + * call. */ + + struct table { const char *suffix; unsigned long long factor; - } table[] = { - { "B", 1 }, - { "K", 1024ULL }, - { "M", 1024ULL*1024ULL }, - { "G", 1024ULL*1024ULL*1024ULL }, - { "T", 1024ULL*1024ULL*1024ULL*1024ULL }, - { "P", 1024ULL*1024ULL*1024ULL*1024ULL*1024ULL }, + }; + + static const struct table iec[] = { { "E", 1024ULL*1024ULL*1024ULL*1024ULL*1024ULL*1024ULL }, + { "P", 1024ULL*1024ULL*1024ULL*1024ULL*1024ULL }, + { "T", 1024ULL*1024ULL*1024ULL*1024ULL }, + { "G", 1024ULL*1024ULL*1024ULL }, + { "M", 1024ULL*1024ULL }, + { "K", 1024ULL }, + { "B", 1 }, + { "", 1 }, + }; + + static const struct table si[] = { + { "E", 1000ULL*1000ULL*1000ULL*1000ULL*1000ULL*1000ULL }, + { "P", 1000ULL*1000ULL*1000ULL*1000ULL*1000ULL }, + { "T", 1000ULL*1000ULL*1000ULL*1000ULL }, + { "G", 1000ULL*1000ULL*1000ULL }, + { "M", 1000ULL*1000ULL }, + { "K", 1000ULL }, + { "B", 1 }, { "", 1 }, }; + const struct table *table; const char *p; unsigned long long r = 0; + unsigned n_entries, start_pos = 0; assert(t); - assert(bytes); + assert(base == 1000 || base == 1024); + assert(size); + + if (base == 1000) { + table = si; + n_entries = ELEMENTSOF(si); + } else { + table = iec; + n_entries = ELEMENTSOF(iec); + } p = t; do { long long l; + unsigned long long l2; + double frac = 0; char *e; unsigned i; @@ -2234,14 +2209,32 @@ int parse_bytes(const char *t, off_t *bytes) { if (e == p) return -EINVAL; + if (*e == '.') { + e++; + if (*e >= '0' && *e <= '9') { + char *e2; + + /* strotoull itself would accept space/+/- */ + l2 = strtoull(e, &e2, 10); + + if (errno == ERANGE) + return -errno; + + /* Ignore failure. E.g. 10.M is valid */ + frac = l2; + for (; e < e2; e++) + frac /= 10; + } + } + e += strspn(e, WHITESPACE); - for (i = 0; i < ELEMENTSOF(table); i++) + for (i = start_pos; i < n_entries; i++) if (startswith(e, table[i].suffix)) { unsigned long long tmp; - if ((unsigned long long) l > ULLONG_MAX / table[i].factor) + if ((unsigned long long) l + (frac > 0) > ULLONG_MAX / table[i].factor) return -ERANGE; - tmp = l * table[i].factor; + tmp = l * table[i].factor + (unsigned long long) (frac * table[i].factor); if (tmp > ULLONG_MAX - r) return -ERANGE; @@ -2250,15 +2243,17 @@ int parse_bytes(const char *t, off_t *bytes) { return -ERANGE; p = e + strlen(table[i].suffix); + + start_pos = i + 1; break; } - if (i >= ELEMENTSOF(table)) + if (i >= n_entries) return -EINVAL; } while (*p); - *bytes = r; + *size = r; return 0; } @@ -2305,7 +2300,6 @@ bool is_device_path(const char *path) { int dir_is_empty(const char *path) { _cleanup_closedir_ DIR *d; - int r; d = opendir(path); if (!d) @@ -2313,11 +2307,11 @@ int dir_is_empty(const char *path) { 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; @@ -2345,42 +2339,60 @@ char* dirname_malloc(const char *path) { 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; + return errno == ENOENT ? -ENOSYS : -errno; - r = loop_read(fd, &ull, sizeof(ull), true); - if (r != sizeof(ull)) - goto fallback; - - 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]) { @@ -2409,7 +2421,7 @@ 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])); } } } @@ -2501,7 +2513,7 @@ int getttyname_malloc(int fd, char **r) { assert(r); k = ttyname_r(fd, path, sizeof(path)); - if (k != 0) + if (k > 0) return -k; char_array_0(path); @@ -2532,24 +2544,17 @@ int getttyname_harder(int fd, char **r) { } 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) @@ -2576,9 +2581,11 @@ int get_ctty_devnr(pid_t pid, dev_t *d) { } 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); @@ -2596,14 +2603,8 @@ int get_ctty(pid_t pid, dev_t *_devnr, char **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 @@ -2611,14 +2612,7 @@ int get_ctty(pid_t pid, dev_t *_devnr, char **r) { * vaguely useful. */ b = strdup(fn + 5); - if (!b) - return -ENOMEM; - - *r = b; - if (_devnr) - *_devnr = devnr; - - return 0; + goto finish; } if (startswith(s, "/dev/")) @@ -2629,8 +2623,8 @@ int get_ctty(pid_t pid, dev_t *_devnr, char **r) { p = s; b = strdup(p); - free(s); +finish: if (!b) return -ENOMEM; @@ -2659,14 +2653,15 @@ int rm_rf_children_dangerous(int fd, bool only_dirs, bool honour_sticky, struct 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; } @@ -2736,9 +2731,9 @@ int rm_rf_children_dangerous(int fd, bool only_dirs, bool honour_sticky, struct _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) { @@ -2978,24 +2973,6 @@ int status_printf(const char *status, bool ellipse, bool ephemeral, const char * 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, @@ -3214,19 +3191,27 @@ bool on_tty(void) { return cached_on_tty; } -int running_in_chroot(void) { - struct stat a = {}, b = {}; +int files_same(const char *filea, const char *fileb) { + struct stat a, b; - /* Only works as root */ - if (stat("/proc/1/root", &a) < 0) + if (stat(filea, &a) < 0) return -errno; - if (stat("/", &b) < 0) + if (stat(fileb, &b) < 0) return -errno; - return - a.st_dev != b.st_dev || - a.st_ino != b.st_ino; + return a.st_dev == b.st_dev && + a.st_ino == b.st_ino; +} + +int running_in_chroot(void) { + int ret; + + ret = files_same("/proc/1/root", "/"); + if (ret < 0) + return ret; + + return ret == 0; } static char *ascii_ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) { @@ -3468,7 +3453,7 @@ int wait_for_terminate_and_warn(const char *name, pid_t pid) { 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); @@ -3535,25 +3520,21 @@ int signal_from_string_try_harder(const char *s) { static char *tag_to_udev_node(const char *tagvalue, const char *by) { _cleanup_free_ char *t = NULL, *u = NULL; - char *dn; size_t enc_len; u = unquote(tagvalue, "\"\'"); - if (u == NULL) + if (!u) return NULL; enc_len = strlen(u) * 4 + 1; t = new(char, enc_len); - if (t == NULL) + if (!t) return NULL; if (encode_devnode_name(u, t, enc_len) < 0) return NULL; - if (asprintf(&dn, "/dev/disk/by-%s/%s", by, t) < 0) - return NULL; - - return dn; + return strjoin("/dev/disk/by-", by, "/", t, NULL); } char *fstab_node_to_udev_node(const char *p) { @@ -3700,111 +3681,123 @@ bool dirent_is_file_with_suffix(const struct dirent *de, const char *suffix) { return endswith(de->d_name, suffix); } -void execute_directory(const char *directory, DIR *d, char *argv[]) { - DIR *_d = NULL; - struct dirent *de; - Hashmap *pids = NULL; +void execute_directory(const char *directory, DIR *d, usec_t timeout, char *argv[]) { + pid_t executor_pid; + int r; assert(directory); - /* Executes all binaries in a directory in parallel and - * waits for them to finish. */ + /* Executes all binaries in a directory in parallel and waits + * for them to finish. Optionally a timeout is applied. */ - if (!d) { - if (!(_d = opendir(directory))) { + executor_pid = fork(); + if (executor_pid < 0) { + log_error("Failed to fork: %m"); + return; - if (errno == ENOENT) - return; + } else if (executor_pid == 0) { + _cleanup_hashmap_free_free_ Hashmap *pids = NULL; + _cleanup_closedir_ DIR *_d = NULL; + struct dirent *de; + sigset_t ss; - log_error("Failed to enumerate directory %s: %m", directory); - return; - } + /* We fork this all off from a child process so that + * we can somewhat cleanly make use of SIGALRM to set + * a time limit */ - d = _d; - } + reset_all_signal_handlers(); - if (!(pids = hashmap_new(trivial_hash_func, trivial_compare_func))) { - log_error("Failed to allocate set."); - goto finish; - } + assert_se(sigemptyset(&ss) == 0); + assert_se(sigprocmask(SIG_SETMASK, &ss, NULL) == 0); - while ((de = readdir(d))) { - char *path; - pid_t pid; - int k; + assert_se(prctl(PR_SET_PDEATHSIG, SIGTERM) == 0); - if (!dirent_is_file(de)) - continue; + if (!d) { + d = _d = opendir(directory); + if (!d) { + if (errno == ENOENT) + _exit(EXIT_SUCCESS); - if (asprintf(&path, "%s/%s", directory, de->d_name) < 0) { - log_oom(); - continue; + log_error("Failed to enumerate directory %s: %m", directory); + _exit(EXIT_FAILURE); + } } - if ((pid = fork()) < 0) { - log_error("Failed to fork: %m"); - free(path); - continue; + pids = hashmap_new(NULL, NULL); + if (!pids) { + log_oom(); + _exit(EXIT_FAILURE); } - if (pid == 0) { - char *_argv[2]; - /* Child */ + FOREACH_DIRENT(de, d, break) { + _cleanup_free_ char *path = NULL; + pid_t pid; - if (!argv) { - _argv[0] = path; - _argv[1] = NULL; - argv = _argv; - } else - argv[0] = path; + if (!dirent_is_file(de)) + continue; - execv(path, argv); + if (asprintf(&path, "%s/%s", directory, de->d_name) < 0) { + log_oom(); + _exit(EXIT_FAILURE); + } - log_error("Failed to execute %s: %m", path); - _exit(EXIT_FAILURE); - } + pid = fork(); + if (pid < 0) { + log_error("Failed to fork: %m"); + continue; + } else if (pid == 0) { + char *_argv[2]; - log_debug("Spawned %s as %lu", path, (unsigned long) pid); + assert_se(prctl(PR_SET_PDEATHSIG, SIGTERM) == 0); - if ((k = hashmap_put(pids, UINT_TO_PTR(pid), path)) < 0) { - log_error("Failed to add PID to set: %s", strerror(-k)); - free(path); - } - } + if (!argv) { + _argv[0] = path; + _argv[1] = NULL; + argv = _argv; + } else + argv[0] = path; - while (!hashmap_isempty(pids)) { - pid_t pid = PTR_TO_UINT(hashmap_first_key(pids)); - siginfo_t si = {}; - char *path; + execv(path, argv); + log_error("Failed to execute %s: %m", path); + _exit(EXIT_FAILURE); + } - if (waitid(P_PID, pid, &si, WEXITED) < 0) { - if (errno == EINTR) - continue; + log_debug("Spawned %s as " PID_FMT ".", path, pid); - log_error("waitid() failed: %m"); - goto finish; - } + r = hashmap_put(pids, UINT_TO_PTR(pid), path); + if (r < 0) { + log_oom(); + _exit(EXIT_FAILURE); + } - if ((path = hashmap_remove(pids, UINT_TO_PTR(si.si_pid)))) { - if (!is_clean_exit(si.si_code, si.si_status, NULL)) { - if (si.si_code == CLD_EXITED) - log_error("%s exited with exit status %i.", path, si.si_status); - else - log_error("%s terminated by signal %s.", path, signal_to_string(si.si_status)); - } else - log_debug("%s exited successfully.", path); + path = NULL; + } + + /* Abort execution of this process after the + * timout. We simply rely on SIGALRM as default action + * terminating the process, and turn on alarm(). */ + + if (timeout != (usec_t) -1) + alarm((timeout + USEC_PER_SEC - 1) / USEC_PER_SEC); - free(path); + while (!hashmap_isempty(pids)) { + _cleanup_free_ char *path = NULL; + pid_t pid; + + pid = PTR_TO_UINT(hashmap_first_key(pids)); + assert(pid > 0); + + path = hashmap_remove(pids, UINT_TO_PTR(pid)); + assert(path); + + wait_for_terminate_and_warn(path, pid); } - } -finish: - if (_d) - closedir(_d); + _exit(EXIT_SUCCESS); + } - if (pids) - hashmap_free_free(pids); + wait_for_terminate_and_warn(directory, executor_pid); } int kill_and_sigcont(pid_t pid, int sig) { @@ -3913,12 +3906,13 @@ char* hostname_cleanup(char *s, bool lowercase) { } 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; @@ -3930,13 +3924,16 @@ int pipe_eof(int fd) { } 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; @@ -3961,13 +3958,13 @@ int fopen_temporary(const char *path, FILE **_f, char **_temp_path) { if (!t) return -ENOMEM; - fn = path_get_file_name(path); - k = fn-path; + fn = basename(path); + k = fn - path; memcpy(t, path, k); 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; @@ -4136,7 +4133,7 @@ int symlink_atomic(const char *from, const char *to) { _cleanup_free_ char *t; const char *fn; size_t k; - unsigned long long ull; + uint64_t u; unsigned i; int r; @@ -4147,16 +4144,16 @@ int symlink_atomic(const char *from, const char *to) { if (!t) return -ENOMEM; - fn = path_get_file_name(to); + fn = basename(to); k = fn-to; memcpy(t, to, k); 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; @@ -4484,13 +4481,11 @@ int get_files_in_directory(const char *path, char ***list) { 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; @@ -4579,7 +4574,7 @@ char *strjoin(const char *x, ...) { } 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; @@ -4728,7 +4723,7 @@ static const char* const sched_policy_table[] = { DEFINE_STRING_TABLE_LOOKUP_WITH_FALLBACK(sched_policy, int, INT_MAX); -static const char* const rlimit_table[] = { +static const char* const rlimit_table[_RLIMIT_MAX] = { [RLIMIT_CPU] = "LimitCPU", [RLIMIT_FSIZE] = "LimitFSIZE", [RLIMIT_DATA] = "LimitDATA", @@ -4797,7 +4792,7 @@ static const char *const __signal_table[] = { 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); @@ -4935,15 +4930,15 @@ int fd_inc_sndbuf(int fd, size_t n) { 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; } @@ -4953,16 +4948,15 @@ int fd_inc_rcvbuf(int fd, size_t n) { 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; } @@ -5087,10 +5081,7 @@ int getenv_for_pid(pid_t pid, const char *field, char **_value) { 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) @@ -5154,7 +5145,7 @@ bool is_valid_documentation_url(const char *url) { } bool in_initrd(void) { - static __thread int saved = -1; + static int saved = -1; struct statfs s; if (saved >= 0) @@ -5219,10 +5210,10 @@ int make_console_stdio(void) { } int get_home_dir(char **_h) { - char *h; + struct passwd *p; const char *e; + char *h; uid_t u; - struct passwd *p; assert(_h); @@ -5265,6 +5256,53 @@ int get_home_dir(char **_h) { return 0; } +int get_shell(char **_s) { + struct passwd *p; + const char *e; + char *s; + uid_t u; + + assert(_s); + + /* Take the user specified one */ + e = getenv("SHELL"); + if (e) { + s = strdup(e); + if (!s) + return -ENOMEM; + + *_s = s; + return 0; + } + + /* Hardcode home directory for root to avoid NSS */ + u = getuid(); + if (u == 0) { + s = strdup("/bin/sh"); + if (!s) + return -ENOMEM; + + *_s = s; + return 0; + } + + /* Check the database... */ + errno = 0; + p = getpwuid(u); + if (!p) + return errno > 0 ? -errno : -ESRCH; + + if (!path_is_absolute(p->pw_shell)) + return -EINVAL; + + s = strdup(p->pw_shell); + if (!s) + return -ENOMEM; + + *_s = s; + return 0; +} + bool filename_is_safe(const char *p) { if (isempty(p)) @@ -5379,7 +5417,7 @@ bool is_locale_utf8(void) { goto out; } - if(streq(set, "UTF-8")) { + if (streq(set, "UTF-8")) { cached_answer = true; goto out; } @@ -5570,15 +5608,14 @@ int on_ac_power(void) { 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; @@ -5644,7 +5681,7 @@ static int search_and_fopen_internal(const char *path, const char *mode, char ** assert(mode); assert(_f); - if (!path_strv_canonicalize_uniq(search)) + if (!path_strv_canonicalize_absolute_uniq(search, NULL)) return -ENOMEM; STRV_FOREACH(i, search) { @@ -5791,10 +5828,18 @@ void* greedy_realloc(void **p, size_t *allocated, size_t need) { size_t a; void *q; + assert(p); + assert(allocated); + if (*allocated >= need) return *p; a = MAX(64u, need * 2); + + /* check for overflows */ + if (a < need) + return NULL; + q = realloc(*p, a); if (!q) return NULL; @@ -5804,6 +5849,25 @@ void* greedy_realloc(void **p, size_t *allocated, size_t need) { return q; } +void* greedy_realloc0(void **p, size_t *allocated, size_t need) { + size_t prev; + uint8_t *q; + + assert(p); + assert(allocated); + + prev = *allocated; + + q = greedy_realloc(p, allocated, need); + if (!q) + return NULL; + + if (*allocated > prev) + memzero(&q[prev], *allocated - prev); + + return q; +} + bool id128_is_valid(const char *s) { size_t i, l; @@ -5845,20 +5909,6 @@ bool id128_is_valid(const char *s) { 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; @@ -5891,7 +5941,7 @@ int split_pair(const char *s, const char *sep, char **l, char **r) { } int shall_restore_state(void) { - _cleanup_free_ char *line; + _cleanup_free_ char *line = NULL; char *w, *state; size_t l; int r; @@ -5902,19 +5952,46 @@ int shall_restore_state(void) { if (r == 0) /* Container ... */ return 1; - FOREACH_WORD_QUOTED(w, l, line, state) - if (l == 23 && memcmp(w, "systemd.restore_state=0", 23)) - return 0; + r = 1; - return 1; + FOREACH_WORD_QUOTED(w, l, line, state) { + const char *e; + char n[l+1]; + int k; + + memcpy(n, w, l); + n[l] = 0; + + e = startswith(n, "systemd.restore_state="); + if (!e) + continue; + + k = parse_boolean(e); + if (k >= 0) + r = k; + } + + return r; } int proc_cmdline(char **ret) { 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); @@ -5923,3 +6000,356 @@ int proc_cmdline(char **ret) { return 1; } + +int parse_proc_cmdline(int (*parse_item)(const char *key, const char *value)) { + _cleanup_free_ char *line = NULL; + char *w, *state; + size_t l; + int r; + + assert(parse_item); + + 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) { + char word[l+1], *value; + + memcpy(word, w, l); + word[l] = 0; + + /* Filter out arguments that are intended only for the + * initrd */ + if (!in_initrd() && startswith(word, "rd.")) + continue; + + value = strchr(word, '='); + if (value) + *(value++) = 0; + + r = parse_item(word, value); + if (r < 0) + return r; + } + + return 0; +} + +int container_get_leader(const char *machine, pid_t *pid) { + _cleanup_free_ char *s = NULL, *class = NULL; + const char *p; + pid_t leader; + int r; + + assert(machine); + assert(pid); + + p = strappenda("/run/systemd/machines/", machine); + r = parse_env_file(p, NEWLINE, "LEADER", &s, "CLASS", &class, NULL); + if (r == -ENOENT) + return -EHOSTDOWN; + if (r < 0) + return r; + if (!s) + return -EIO; + + if (!streq_ptr(class, "container")) + return -EIO; + + r = parse_pid(s, &leader); + if (r < 0) + return r; + if (leader <= 1) + return -EIO; + + *pid = leader; + return 0; +} + +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(pidns_fd); + assert(mntns_fd); + assert(root_fd); + + 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"); + rfd = open(root, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); + if (rfd < 0) + return -errno; + + *pidns_fd = pidnsfd; + *mntns_fd = mntnsfd; + *root_fd = rfd; + pidnsfd = -1; + mntnsfd = -1; + + return 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(pidns_fd, CLONE_NEWPID) < 0) + return -errno; + + if (setns(mntns_fd, CLONE_NEWNS) < 0) + return -errno; + + if (fchdir(root_fd) < 0) + return -errno; + + if (chroot(".") < 0) + return -errno; + + if (setresgid(0, 0, 0) < 0) + return -errno; + + if (setresuid(0, 0, 0) < 0) + return -errno; + + 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; +} + +uint64_t physical_memory(void) { + long mem; + + /* We return this as uint64_t in case we are running as 32bit + * process on a 64bit kernel with huge amounts of memory */ + + mem = sysconf(_SC_PHYS_PAGES); + assert(mem > 0); + + return (uint64_t) mem * (uint64_t) page_size(); +} + +char* mount_test_option(const char *haystack, const char *needle) { + + struct mntent me = { + .mnt_opts = (char*) haystack + }; + + assert(needle); + + /* Like glibc's hasmntopt(), but works on a string, not a + * struct mntent */ + + if (!haystack) + return NULL; + + return hasmntopt(&me, needle); +}