int
__pthread_setaffinity_new (pthread_t th, size_t cpusetsize,
const cpu_set_t *cpuset)
{
const struct pthread *pd = (const struct pthread *) th;
INTERNAL_SYSCALL_DECL (err);
int res;
if (__builtin_expect (__kernel_cpumask_size == 0, 0))
{
res = __determine_cpumask_size (pd->tid);
if (res != 0)
return res;
}
/* We now know the size of the kernel cpumask_t. Make sure the user
does not request to set a bit beyond that. */
for (size_t cnt = __kernel_cpumask_size; cnt < cpusetsize; ++cnt)
if (((char *) cpuset)[cnt] != '\0')
/* Found a nonzero byte. This means the user request cannot be
fulfilled. */
return EINVAL;
res = INTERNAL_SYSCALL (sched_setaffinity, err, 3, pd->tid, cpusetsize,
cpuset);
#ifdef RESET_VGETCPU_CACHE
if (!INTERNAL_SYSCALL_ERROR_P (res, err))
RESET_VGETCPU_CACHE ();
#endif
return (INTERNAL_SYSCALL_ERROR_P (res, err)
? INTERNAL_SYSCALL_ERRNO (res, err)
: 0);
}
/* Get information about the file NAME in BUF. */
int
__fxstatat (int vers, int fd, const char *file, struct stat *st, int flag)
{
int result;
INTERNAL_SYSCALL_DECL (err);
#ifdef STAT_IS_KERNEL_STAT
# define kst (*st)
#else
struct kernel_stat kst;
#endif
result = INTERNAL_SYSCALL (newfstatat, err, 4, fd, file, &kst, flag);
if (!__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 1))
{
#ifdef STAT_IS_KERNEL_STAT
return 0;
#else
return __xstat_conv (vers, &kst, st);
#endif
}
else
return INLINE_SYSCALL_ERROR_RETURN_VALUE (INTERNAL_SYSCALL_ERRNO (result,
err));
}
pthread_mutex_timedlock (
pthread_mutex_t *mutex,
const struct timespec *abstime)
{
int oldval;
pid_t id = THREAD_GETMEM (THREAD_SELF, tid);
int result = 0;
/* We must not check ABSTIME here. If the thread does not block
abstime must not be checked for a valid value. */
switch (__builtin_expect (PTHREAD_MUTEX_TYPE (mutex),
PTHREAD_MUTEX_TIMED_NP))
{
/* Recursive mutex. */
case PTHREAD_MUTEX_RECURSIVE_NP:
/* Check whether we already hold the mutex. */
if (mutex->__data.__owner == id)
{
/* Just bump the counter. */
if (__builtin_expect (mutex->__data.__count + 1 == 0, 0))
/* Overflow of the counter. */
return EAGAIN;
++mutex->__data.__count;
goto out;
}
/* We have to get the mutex. */
result = lll_timedlock (mutex->__data.__lock, abstime,
PTHREAD_MUTEX_PSHARED (mutex));
if (result != 0)
goto out;
/* Only locked once so far. */
mutex->__data.__count = 1;
break;
/* Error checking mutex. */
case PTHREAD_MUTEX_ERRORCHECK_NP:
/* Check whether we already hold the mutex. */
if (__builtin_expect (mutex->__data.__owner == id, 0))
return EDEADLK;
/* FALLTHROUGH */
case PTHREAD_MUTEX_TIMED_NP:
simple:
/* Normal mutex. */
result = lll_timedlock (mutex->__data.__lock, abstime,
PTHREAD_MUTEX_PSHARED (mutex));
break;
case PTHREAD_MUTEX_ADAPTIVE_NP:
if (! __is_smp)
goto simple;
if (lll_trylock (mutex->__data.__lock) != 0)
{
int cnt = 0;
int max_cnt = MIN (MAX_ADAPTIVE_COUNT,
mutex->__data.__spins * 2 + 10);
do
{
if (cnt++ >= max_cnt)
{
result = lll_timedlock (mutex->__data.__lock, abstime,
PTHREAD_MUTEX_PSHARED (mutex));
break;
}
#ifdef BUSY_WAIT_NOP
BUSY_WAIT_NOP;
#endif
}
while (lll_trylock (mutex->__data.__lock) != 0);
mutex->__data.__spins += (cnt - mutex->__data.__spins) / 8;
}
break;
case PTHREAD_MUTEX_ROBUST_RECURSIVE_NP:
case PTHREAD_MUTEX_ROBUST_ERRORCHECK_NP:
case PTHREAD_MUTEX_ROBUST_NORMAL_NP:
case PTHREAD_MUTEX_ROBUST_ADAPTIVE_NP:
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending,
&mutex->__data.__list.__next);
oldval = mutex->__data.__lock;
do
{
again:
if ((oldval & FUTEX_OWNER_DIED) != 0)
{
/* The previous owner died. Try locking the mutex. */
int newval = id | (oldval & FUTEX_WAITERS);
newval
= atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
newval, oldval);
if (newval != oldval)
{
oldval = newval;
goto again;
}
/* We got the mutex. */
mutex->__data.__count = 1;
/* But it is inconsistent unless marked otherwise. */
mutex->__data.__owner = PTHREAD_MUTEX_INCONSISTENT;
ENQUEUE_MUTEX (mutex);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
/* Note that we deliberately exit here. If we fall
through to the end of the function __nusers would be
incremented which is not correct because the old
owner has to be discounted. */
return EOWNERDEAD;
}
/* Check whether we already hold the mutex. */
if (__builtin_expect ((oldval & FUTEX_TID_MASK) == id, 0))
{
int kind = PTHREAD_MUTEX_TYPE (mutex);
if (kind == PTHREAD_MUTEX_ROBUST_ERRORCHECK_NP)
{
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending,
NULL);
return EDEADLK;
}
if (kind == PTHREAD_MUTEX_ROBUST_RECURSIVE_NP)
{
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending,
NULL);
/* Just bump the counter. */
if (__builtin_expect (mutex->__data.__count + 1 == 0, 0))
/* Overflow of the counter. */
return EAGAIN;
++mutex->__data.__count;
return 0;
}
}
result = lll_robust_timedlock (mutex->__data.__lock, abstime, id,
PTHREAD_ROBUST_MUTEX_PSHARED (mutex));
if (__builtin_expect (mutex->__data.__owner
== PTHREAD_MUTEX_NOTRECOVERABLE, 0))
{
/* This mutex is now not recoverable. */
mutex->__data.__count = 0;
lll_unlock (mutex->__data.__lock,
PTHREAD_ROBUST_MUTEX_PSHARED (mutex));
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
return ENOTRECOVERABLE;
}
if (result == ETIMEDOUT || result == EINVAL)
goto out;
oldval = result;
}
while ((oldval & FUTEX_OWNER_DIED) != 0);
mutex->__data.__count = 1;
ENQUEUE_MUTEX (mutex);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
break;
case PTHREAD_MUTEX_PI_RECURSIVE_NP:
case PTHREAD_MUTEX_PI_ERRORCHECK_NP:
case PTHREAD_MUTEX_PI_NORMAL_NP:
case PTHREAD_MUTEX_PI_ADAPTIVE_NP:
case PTHREAD_MUTEX_PI_ROBUST_RECURSIVE_NP:
case PTHREAD_MUTEX_PI_ROBUST_ERRORCHECK_NP:
case PTHREAD_MUTEX_PI_ROBUST_NORMAL_NP:
case PTHREAD_MUTEX_PI_ROBUST_ADAPTIVE_NP:
{
int kind = mutex->__data.__kind & PTHREAD_MUTEX_KIND_MASK_NP;
int robust = mutex->__data.__kind & PTHREAD_MUTEX_ROBUST_NORMAL_NP;
if (robust)
/* Note: robust PI futexes are signaled by setting bit 0. */
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending,
(void *) (((uintptr_t) &mutex->__data.__list.__next)
| 1));
oldval = mutex->__data.__lock;
/* Check whether we already hold the mutex. */
if (__builtin_expect ((oldval & FUTEX_TID_MASK) == id, 0))
{
if (kind == PTHREAD_MUTEX_ERRORCHECK_NP)
{
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
return EDEADLK;
}
if (kind == PTHREAD_MUTEX_RECURSIVE_NP)
{
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
/* Just bump the counter. */
if (__builtin_expect (mutex->__data.__count + 1 == 0, 0))
/* Overflow of the counter. */
return EAGAIN;
++mutex->__data.__count;
return 0;
}
}
oldval = atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
id, 0);
if (oldval != 0)
{
/* The mutex is locked. The kernel will now take care of
everything. The timeout value must be a relative value.
Convert it. */
int private = (robust
? PTHREAD_ROBUST_MUTEX_PSHARED (mutex)
: PTHREAD_MUTEX_PSHARED (mutex));
INTERNAL_SYSCALL_DECL (__err);
int e = INTERNAL_SYSCALL (futex, __err, 4, &mutex->__data.__lock,
__lll_private_flag (FUTEX_LOCK_PI,
private), 1,
abstime);
if (INTERNAL_SYSCALL_ERROR_P (e, __err))
{
if (INTERNAL_SYSCALL_ERRNO (e, __err) == ETIMEDOUT)
return ETIMEDOUT;
if (INTERNAL_SYSCALL_ERRNO (e, __err) == ESRCH
|| INTERNAL_SYSCALL_ERRNO (e, __err) == EDEADLK)
{
assert (INTERNAL_SYSCALL_ERRNO (e, __err) != EDEADLK
|| (kind != PTHREAD_MUTEX_ERRORCHECK_NP
&& kind != PTHREAD_MUTEX_RECURSIVE_NP));
/* ESRCH can happen only for non-robust PI mutexes where
the owner of the lock died. */
assert (INTERNAL_SYSCALL_ERRNO (e, __err) != ESRCH
|| !robust);
/* Delay the thread until the timeout is reached.
Then return ETIMEDOUT. */
struct timespec reltime;
struct timespec now;
INTERNAL_SYSCALL (clock_gettime, __err, 2, CLOCK_REALTIME,
&now);
reltime.tv_sec = abstime->tv_sec - now.tv_sec;
reltime.tv_nsec = abstime->tv_nsec - now.tv_nsec;
if (reltime.tv_nsec < 0)
{
reltime.tv_nsec += 1000000000;
--reltime.tv_sec;
}
if (reltime.tv_sec >= 0)
while (nanosleep_not_cancel (&reltime, &reltime) != 0)
continue;
return ETIMEDOUT;
}
return INTERNAL_SYSCALL_ERRNO (e, __err);
}
oldval = mutex->__data.__lock;
assert (robust || (oldval & FUTEX_OWNER_DIED) == 0);
}
if (__builtin_expect (oldval & FUTEX_OWNER_DIED, 0))
{
atomic_and (&mutex->__data.__lock, ~FUTEX_OWNER_DIED);
/* We got the mutex. */
mutex->__data.__count = 1;
/* But it is inconsistent unless marked otherwise. */
mutex->__data.__owner = PTHREAD_MUTEX_INCONSISTENT;
ENQUEUE_MUTEX_PI (mutex);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
/* Note that we deliberately exit here. If we fall
through to the end of the function __nusers would be
incremented which is not correct because the old owner
has to be discounted. */
return EOWNERDEAD;
}
if (robust
&& __builtin_expect (mutex->__data.__owner
== PTHREAD_MUTEX_NOTRECOVERABLE, 0))
{
/* This mutex is now not recoverable. */
mutex->__data.__count = 0;
INTERNAL_SYSCALL_DECL (__err);
INTERNAL_SYSCALL (futex, __err, 4, &mutex->__data.__lock,
__lll_private_flag (FUTEX_UNLOCK_PI,
PTHREAD_ROBUST_MUTEX_PSHARED (mutex)),
0, 0);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
return ENOTRECOVERABLE;
}
mutex->__data.__count = 1;
if (robust)
{
ENQUEUE_MUTEX_PI (mutex);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
}
}
break;
case PTHREAD_MUTEX_PP_RECURSIVE_NP:
case PTHREAD_MUTEX_PP_ERRORCHECK_NP:
case PTHREAD_MUTEX_PP_NORMAL_NP:
case PTHREAD_MUTEX_PP_ADAPTIVE_NP:
{
int kind = mutex->__data.__kind & PTHREAD_MUTEX_KIND_MASK_NP;
oldval = mutex->__data.__lock;
/* Check whether we already hold the mutex. */
if (mutex->__data.__owner == id)
{
if (kind == PTHREAD_MUTEX_ERRORCHECK_NP)
return EDEADLK;
if (kind == PTHREAD_MUTEX_RECURSIVE_NP)
{
/* Just bump the counter. */
if (__builtin_expect (mutex->__data.__count + 1 == 0, 0))
/* Overflow of the counter. */
return EAGAIN;
++mutex->__data.__count;
return 0;
}
}
int oldprio = -1, ceilval;
do
{
int ceiling = (oldval & PTHREAD_MUTEX_PRIO_CEILING_MASK)
>> PTHREAD_MUTEX_PRIO_CEILING_SHIFT;
if (__pthread_current_priority () > ceiling)
{
result = EINVAL;
failpp:
if (oldprio != -1)
__pthread_tpp_change_priority (oldprio, -1);
return result;
}
result = __pthread_tpp_change_priority (oldprio, ceiling);
if (result)
return result;
ceilval = ceiling << PTHREAD_MUTEX_PRIO_CEILING_SHIFT;
oldprio = ceiling;
oldval
= atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
ceilval | 1, ceilval);
if (oldval == ceilval)
break;
do
{
oldval
= atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
ceilval | 2,
ceilval | 1);
if ((oldval & PTHREAD_MUTEX_PRIO_CEILING_MASK) != ceilval)
break;
if (oldval != ceilval)
{
/* Reject invalid timeouts. */
if (abstime->tv_nsec < 0 || abstime->tv_nsec >= 1000000000)
{
result = EINVAL;
goto failpp;
}
struct timeval tv;
struct timespec rt;
/* Get the current time. */
(void) gettimeofday (&tv, NULL);
/* Compute relative timeout. */
rt.tv_sec = abstime->tv_sec - tv.tv_sec;
rt.tv_nsec = abstime->tv_nsec - tv.tv_usec * 1000;
if (rt.tv_nsec < 0)
{
rt.tv_nsec += 1000000000;
--rt.tv_sec;
}
/* Already timed out? */
if (rt.tv_sec < 0)
{
result = ETIMEDOUT;
goto failpp;
}
lll_futex_timed_wait (&mutex->__data.__lock,
ceilval | 2, &rt,
PTHREAD_MUTEX_PSHARED (mutex));
}
}
while (atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
ceilval | 2, ceilval)
!= ceilval);
}
while ((oldval & PTHREAD_MUTEX_PRIO_CEILING_MASK) != ceilval);
assert (mutex->__data.__owner == 0);
mutex->__data.__count = 1;
}
break;
default:
/* Correct code cannot set any other type. */
return EINVAL;
}
/* Get information about the file NAME relative to FD in ST. */
int
__fxstatat (int vers, int fd, const char *file, struct stat *st, int flag)
{
if (vers != _STAT_VER_KERNEL && vers != _STAT_VER_LINUX)
{
__set_errno (EINVAL);
return -1;
}
int res;
#ifdef __NR_newfstatat
# ifndef __ASSUME_ATFCTS
if (__have_atfcts >= 0)
# endif
{
res = INLINE_SYSCALL (newfstatat, 4, fd, file, st, flag);
# ifndef __ASSUME_ATFCTS
if (res == -1 && errno == ENOSYS)
__have_atfcts = -1;
else
# endif
return res;
}
#endif
#ifndef __ASSUME_ATFCTS
if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
{
__set_errno (EINVAL);
return -1;
}
char *buf = NULL;
if (fd != AT_FDCWD && file[0] != '/')
{
size_t filelen = strlen (file);
if (__builtin_expect (filelen == 0, 0))
{
__set_errno (ENOENT);
return -1;
}
static const char procfd[] = "/proc/self/fd/%d/%s";
/* Buffer for the path name we are going to use. It consists of
- the string /proc/self/fd/
- the file descriptor number
- the file name provided.
The final NUL is included in the sizeof. A bit of overhead
due to the format elements compensates for possible negative
numbers. */
size_t buflen = sizeof (procfd) + sizeof (int) * 3 + filelen;
buf = alloca (buflen);
__snprintf (buf, buflen, procfd, fd, file);
file = buf;
}
INTERNAL_SYSCALL_DECL (err);
if (flag & AT_SYMLINK_NOFOLLOW)
res = INTERNAL_SYSCALL (lstat, err, 2, file, CHECK_1 (st));
else
res = INTERNAL_SYSCALL (stat, err, 2, file, CHECK_1 (st));
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (res, err), 0))
{
__atfct_seterrno (INTERNAL_SYSCALL_ERRNO (res, err), fd, buf);
res = -1;
}
return res;
#endif
}
/* Get information about the file NAME relative to FD in ST. */
int
__fxstatat (int vers, int fd, const char *file, struct stat *st, int flag)
{
int result;
INTERNAL_SYSCALL_DECL (err);
struct stat64 st64;
#ifdef __NR_fstatat64
# ifndef __ASSUME_ATFCTS
if (__have_atfcts >= 0)
# endif
{
result = INTERNAL_SYSCALL (fstatat64, err, 4, fd, file, &st64, flag);
# ifndef __ASSUME_ATFCTS
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 1)
&& INTERNAL_SYSCALL_ERRNO (result, err) == ENOSYS)
__have_atfcts = -1;
else
# endif
if (!__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return __xstat32_conv (vers, &st64, st);
else
{
__set_errno (INTERNAL_SYSCALL_ERRNO (result, err));
return -1;
}
}
#endif
#ifndef __ASSUME_ATFCTS
if (__builtin_expect (flag & ~AT_SYMLINK_NOFOLLOW, 0))
{
__set_errno (EINVAL);
return -1;
}
char *buf = NULL;
if (fd != AT_FDCWD && file[0] != '/')
{
size_t filelen = strlen (file);
static const char procfd[] = "/proc/self/fd/%d/%s";
/* Buffer for the path name we are going to use. It consists of
- the string /proc/self/fd/
- the file descriptor number
- the file name provided.
The final NUL is included in the sizeof. A bit of overhead
due to the format elements compensates for possible negative
numbers. */
size_t buflen = sizeof (procfd) + sizeof (int) * 3 + filelen;
buf = alloca (buflen);
__snprintf (buf, buflen, procfd, fd, file);
file = buf;
}
# if __ASSUME_STAT64_SYSCALL == 0
struct kernel_stat kst;
# endif
if (vers == _STAT_VER_KERNEL)
{
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat, err, 2, CHECK_STRING (file),
CHECK_1 ((struct kernel_stat *) st));
else
result = INTERNAL_SYSCALL (stat, err, 2, CHECK_STRING (file),
CHECK_1 ((struct kernel_stat *) st));
goto out;
}
# if __ASSUME_STAT64_SYSCALL > 0
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat64, err, 2, CHECK_STRING (file),
__ptrvalue (&st64));
else
result = INTERNAL_SYSCALL (stat64, err, 2, CHECK_STRING (file),
__ptrvalue (&st64));
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return __xstat32_conv (vers, &st64, st);
# else
# if defined __NR_stat64
/* To support 32 bit UIDs, we have to use stat64. The normal stat
call only returns 16 bit UIDs. */
if (! __have_no_stat64)
{
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat64, err, 2, CHECK_STRING (file),
__ptrvalue (&st64));
else
result = INTERNAL_SYSCALL (stat64, err, 2, CHECK_STRING (file),
__ptrvalue (&st64));
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
result = __xstat32_conv (vers, &st64, st);
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1)
|| INTERNAL_SYSCALL_ERRNO (result, err) != ENOSYS)
goto out;
__have_no_stat64 = 1;
}
# endif
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat, err, 2, CHECK_STRING (file),
__ptrvalue (&kst));
else
result = INTERNAL_SYSCALL (stat, err, 2, CHECK_STRING (file),
__ptrvalue (&kst));
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return __xstat_conv (vers, &kst, st);
# endif /* __ASSUME_STAT64_SYSCALL */
out:
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 0))
{
__atfct_seterrno (INTERNAL_SYSCALL_ERRNO (result, err), fd, buf);
result = -1;
}
return result;
#endif
}
int
pthread_cancel (pthread_t th)
{
volatile struct pthread *pd = (volatile struct pthread *) th;
/* Make sure the descriptor is valid. */
if (INVALID_TD_P (pd))
/* Not a valid thread handle. */
return ESRCH;
#ifdef SHARED
pthread_cancel_init ();
#endif
int result = 0;
int oldval;
int newval;
do
{
again:
oldval = pd->cancelhandling;
newval = oldval | CANCELING_BITMASK | CANCELED_BITMASK;
/* Avoid doing unnecessary work. The atomic operation can
potentially be expensive if the bug has to be locked and
remote cache lines have to be invalidated. */
if (oldval == newval)
break;
/* If the cancellation is handled asynchronously just send a
signal. We avoid this if possible since it's more
expensive. */
if (CANCEL_ENABLED_AND_CANCELED_AND_ASYNCHRONOUS (newval))
{
/* Mark the cancellation as "in progress". */
if (atomic_compare_and_exchange_bool_acq (&pd->cancelhandling,
oldval | CANCELING_BITMASK,
oldval))
goto again;
#ifdef SIGCANCEL
/* The cancellation handler will take care of marking the
thread as canceled. */
pid_t pid = getpid ();
INTERNAL_SYSCALL_DECL (err);
int val = INTERNAL_SYSCALL_CALL (tgkill, err, pid, pd->tid,
SIGCANCEL);
if (INTERNAL_SYSCALL_ERROR_P (val, err))
result = INTERNAL_SYSCALL_ERRNO (val, err);
#else
/* It should be impossible to get here at all, since
pthread_setcanceltype should never have allowed
PTHREAD_CANCEL_ASYNCHRONOUS to be set. */
abort ();
#endif
break;
}
/* A single-threaded process should be able to kill itself, since
there is nothing in the POSIX specification that says that it
cannot. So we set multiple_threads to true so that cancellation
points get executed. */
THREAD_SETMEM (THREAD_SELF, header.multiple_threads, 1);
#ifndef TLS_MULTIPLE_THREADS_IN_TCB
__pthread_multiple_threads = *__libc_multiple_threads_ptr = 1;
#endif
}
/* Mark the thread as canceled. This has to be done
atomically since other bits could be modified as well. */
while (atomic_compare_and_exchange_bool_acq (&pd->cancelhandling, newval,
oldval));
return result;
}
int
fchownat (int fd, const char *file, uid_t owner, gid_t group, int flag)
{
if (flag & ~AT_SYMLINK_NOFOLLOW)
{
__set_errno (EINVAL);
return -1;
}
char *buf = NULL;
if (fd != AT_FDCWD && file[0] != '/')
{
size_t filelen = strlen (file);
static const char procfd[] = "/proc/self/fd/%d/%s";
/* Buffer for the path name we are going to use. It consists of
- the string /proc/self/fd/
- the file descriptor number
- the file name provided.
The final NUL is included in the sizeof. A bit of overhead
due to the format elements compensates for possible negative
numbers. */
size_t buflen = sizeof (procfd) + sizeof (int) * 3 + filelen;
buf = alloca (buflen);
__snprintf (buf, buflen, procfd, fd, file);
file = buf;
}
int result;
INTERNAL_SYSCALL_DECL (err);
#if __ASSUME_LCHOWN_SYSCALL
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lchown, err, 3, file, owner, group);
else
result = INTERNAL_SYSCALL (chown, err, 3, file, owner, group);
#else
char link[PATH_MAX + 2];
char path[2 * PATH_MAX + 4];
int loopct;
size_t filelen;
static int libc_old_chown = 0 /* -1=old linux, 1=new linux, 0=unknown */;
if (libc_old_chown == 1)
{
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lchown, err, 3, __ptrvalue (file), owner,
group);
else
result = INTERNAL_SYSCALL (chown, err, 3, __ptrvalue (file), owner,
group);
goto out;
}
# ifdef __NR_lchown
if (flag & AT_SYMLINK_NOFOLLOW)
{
result = INTERNAL_SYSCALL (lchown, err, 3, __ptrvalue (file), owner,
group);
goto out;
}
if (libc_old_chown == 0)
{
result = INTERNAL_SYSCALL (chown, err, 3, __ptrvalue (file), owner,
group);
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return result;
if (INTERNAL_SYSCALL_ERRNO (result, err) != ENOSYS)
{
libc_old_chown = 1;
goto fail;
}
libc_old_chown = -1;
}
# else
if (flag & AT_SYMLINK_NOFOLLOW)
{
result = INTERNAL_SYSCALL (chown, err, 3, __ptrvalue (file), owner,
group);
goto out;
}
# endif
result = __readlink (file, link, PATH_MAX + 1);
if (result == -1)
{
# ifdef __NR_lchown
result = INTERNAL_SYSCALL (lchown, err, 3, __ptrvalue (file), owner,
group);
# else
result = INTERNAL_SYSCALL (chown, err, 3, __ptrvalue (file), owner,
group);
# endif
goto out;
}
filelen = strlen (file) + 1;
if (filelen > sizeof (path))
{
errno = ENAMETOOLONG;
return -1;
}
memcpy (path, file, filelen);
/* 'The system has an arbitrary limit...' In practise, we'll hit
ENAMETOOLONG before this, usually. */
for (loopct = 0; loopct < 128; ++loopct)
{
size_t linklen;
if (result >= PATH_MAX + 1)
{
errno = ENAMETOOLONG;
return -1;
}
link[result] = 0; /* Null-terminate string, just-in-case. */
linklen = strlen (link) + 1;
if (link[0] == '/')
memcpy (path, link, linklen);
else
{
filelen = strlen (path);
while (filelen > 1 && path[filelen - 1] == '/')
--filelen;
while (filelen > 0 && path[filelen - 1] != '/')
--filelen;
if (filelen + linklen > sizeof (path))
{
errno = ENAMETOOLONG;
return -1;
}
memcpy (path + filelen, link, linklen);
}
result = __readlink (path, link, PATH_MAX + 1);
if (result == -1)
{
# ifdef __NR_lchown
result = INTERNAL_SYSCALL (lchown, err, 3, path, owner, group);
# else
result = INTERNAL_SYSCALL (chown, err, 3, path, owner, group);
# endif
goto out;
}
}
__set_errno (ELOOP);
return -1;
out:
#endif
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 0))
{
#if !__ASSUME_LCHOWN_SYSCALL
fail:
#endif
__atfct_seterrno (INTERNAL_SYSCALL_ERRNO (result, err), fd, buf);
result = -1;
}
return result;
}
/* Get information about the file NAME in BUF. */
int
__fxstatat (int vers, int fd, const char *file, struct stat *st, int flag)
{
int result;
INTERNAL_SYSCALL_DECL (err);
#ifdef STAT_IS_KERNEL_STAT
# define kst (*st)
#else
struct kernel_stat kst;
#endif
#ifdef __NR_newfstatat
# ifndef __ASSUME_ATFCTS
if (__have_atfcts >= 0)
# endif
{
result = INTERNAL_SYSCALL (newfstatat, err, 4, fd, file, &kst, flag);
# ifndef __ASSUME_ATFCTS
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 1)
&& INTERNAL_SYSCALL_ERRNO (result, err) == ENOSYS)
__have_atfcts = -1;
else
# endif
if (!__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 1))
{
#ifdef STAT_IS_KERNEL_STAT
return 0;
#else
return __xstat_conv (vers, &kst, st);
#endif
}
else
{
__set_errno (INTERNAL_SYSCALL_ERRNO (result, err));
return -1;
}
}
#endif
if (flag & ~AT_SYMLINK_NOFOLLOW)
{
__set_errno (EINVAL);
return -1;
}
char *buf = NULL;
if (fd != AT_FDCWD && file[0] != '/')
{
size_t filelen = strlen (file);
if (__builtin_expect (filelen == 0, 0))
{
__set_errno (ENOENT);
return -1;
}
static const char procfd[] = "/proc/self/fd/%d/%s";
/* Buffer for the path name we are going to use. It consists of
- the string /proc/self/fd/
- the file descriptor number
- the file name provided.
The final NUL is included in the sizeof. A bit of overhead
due to the format elements compensates for possible negative
numbers. */
size_t buflen = sizeof (procfd) + sizeof (int) * 3 + filelen;
buf = alloca (buflen);
__snprintf (buf, buflen, procfd, fd, file);
file = buf;
}
if (vers == _STAT_VER_KERNEL)
{
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat, err, 2, file,
(struct kernel_stat *) st);
else
result = INTERNAL_SYSCALL (stat, err, 2, file,
(struct kernel_stat *) st);
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return result;
}
#ifdef STAT_IS_KERNEL_STAT
else
{
__set_errno (EINVAL);
return -1;
}
#else
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat, err, 2, file, &kst);
else
result = INTERNAL_SYSCALL (stat, err, 2, file, &kst);
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return __xstat_conv (vers, &kst, st);
#endif
__atfct_seterrno (INTERNAL_SYSCALL_ERRNO (result, err), fd, buf);
return -1;
}
void
__pthread_initialize_minimal_internal (void)
{
#ifndef SHARED
/* Unlike in the dynamically linked case the dynamic linker has not
taken care of initializing the TLS data structures. */
__libc_setup_tls (TLS_TCB_SIZE, TLS_TCB_ALIGN);
/* We must prevent gcc from being clever and move any of the
following code ahead of the __libc_setup_tls call. This function
will initialize the thread register which is subsequently
used. */
__asm __volatile ("");
#endif
/* Minimal initialization of the thread descriptor. */
struct pthread *pd = THREAD_SELF;
__pthread_initialize_pids (pd);
THREAD_SETMEM (pd, specific[0], &pd->specific_1stblock[0]);
THREAD_SETMEM (pd, user_stack, true);
if (LLL_LOCK_INITIALIZER != 0)
THREAD_SETMEM (pd, lock, LLL_LOCK_INITIALIZER);
#if HP_TIMING_AVAIL
THREAD_SETMEM (pd, cpuclock_offset, GL(dl_cpuclock_offset));
#endif
/* Initialize the robust mutex data. */
{
#ifdef __PTHREAD_MUTEX_HAVE_PREV
pd->robust_prev = &pd->robust_head;
#endif
pd->robust_head.list = &pd->robust_head;
#ifdef __NR_set_robust_list
pd->robust_head.futex_offset = (offsetof (pthread_mutex_t, __data.__lock)
- offsetof (pthread_mutex_t,
__data.__list.__next));
INTERNAL_SYSCALL_DECL (err);
int res = INTERNAL_SYSCALL (set_robust_list, err, 2, &pd->robust_head,
sizeof (struct robust_list_head));
if (INTERNAL_SYSCALL_ERROR_P (res, err))
#endif
set_robust_list_not_avail ();
}
#ifdef __NR_futex
# ifndef __ASSUME_PRIVATE_FUTEX
/* Private futexes are always used (at least internally) so that
doing the test once this early is beneficial. */
{
int word = 0;
INTERNAL_SYSCALL_DECL (err);
word = INTERNAL_SYSCALL (futex, err, 3, &word,
FUTEX_WAKE | FUTEX_PRIVATE_FLAG, 1);
if (!INTERNAL_SYSCALL_ERROR_P (word, err))
THREAD_SETMEM (pd, header.private_futex, FUTEX_PRIVATE_FLAG);
}
/* Private futexes have been introduced earlier than the
FUTEX_CLOCK_REALTIME flag. We don't have to run the test if we
know the former are not supported. This also means we know the
kernel will return ENOSYS for unknown operations. */
if (THREAD_GETMEM (pd, header.private_futex) != 0)
# endif
# ifndef __ASSUME_FUTEX_CLOCK_REALTIME
{
int word = 0;
/* NB: the syscall actually takes six parameters. The last is the
bit mask. But since we will not actually wait at all the value
is irrelevant. Given that passing six parameters is difficult
on some architectures we just pass whatever random value the
calling convention calls for to the kernel. It causes no harm. */
INTERNAL_SYSCALL_DECL (err);
word = INTERNAL_SYSCALL (futex, err, 5, &word,
FUTEX_WAIT_BITSET | FUTEX_CLOCK_REALTIME
| FUTEX_PRIVATE_FLAG, 1, NULL, 0);
assert (INTERNAL_SYSCALL_ERROR_P (word, err));
if (INTERNAL_SYSCALL_ERRNO (word, err) != ENOSYS)
__set_futex_clock_realtime ();
}
# endif
#endif
/* Set initial thread's stack block from 0 up to __libc_stack_end.
It will be bigger than it actually is, but for unwind.c/pt-longjmp.c
purposes this is good enough. */
THREAD_SETMEM (pd, stackblock_size, (size_t) __libc_stack_end);
/* Initialize the list of all running threads with the main thread. */
INIT_LIST_HEAD (&__stack_user);
list_add (&pd->list, &__stack_user);
/* Before initializing __stack_user, the debugger could not find us and
had to set __nptl_initial_report_events. Propagate its setting. */
THREAD_SETMEM (pd, report_events, __nptl_initial_report_events);
#if defined SIGCANCEL || defined SIGSETXID
struct sigaction sa;
__sigemptyset (&sa.sa_mask);
# ifdef SIGCANCEL
/* Install the cancellation signal handler. If for some reason we
cannot install the handler we do not abort. Maybe we should, but
it is only asynchronous cancellation which is affected. */
sa.sa_sigaction = sigcancel_handler;
sa.sa_flags = SA_SIGINFO;
(void) __libc_sigaction (SIGCANCEL, &sa, NULL);
# endif
# ifdef SIGSETXID
/* Install the handle to change the threads' uid/gid. */
sa.sa_sigaction = sighandler_setxid;
sa.sa_flags = SA_SIGINFO | SA_RESTART;
(void) __libc_sigaction (SIGSETXID, &sa, NULL);
# endif
/* The parent process might have left the signals blocked. Just in
case, unblock it. We reuse the signal mask in the sigaction
structure. It is already cleared. */
# ifdef SIGCANCEL
__sigaddset (&sa.sa_mask, SIGCANCEL);
# endif
# ifdef SIGSETXID
__sigaddset (&sa.sa_mask, SIGSETXID);
# endif
{
INTERNAL_SYSCALL_DECL (err);
(void) INTERNAL_SYSCALL (rt_sigprocmask, err, 4, SIG_UNBLOCK, &sa.sa_mask,
NULL, _NSIG / 8);
}
#endif
/* Get the size of the static and alignment requirements for the TLS
block. */
size_t static_tls_align;
_dl_get_tls_static_info (&__static_tls_size, &static_tls_align);
/* Make sure the size takes all the alignments into account. */
if (STACK_ALIGN > static_tls_align)
static_tls_align = STACK_ALIGN;
__static_tls_align_m1 = static_tls_align - 1;
__static_tls_size = roundup (__static_tls_size, static_tls_align);
/* Determine the default allowed stack size. This is the size used
in case the user does not specify one. */
struct rlimit limit;
if (__getrlimit (RLIMIT_STACK, &limit) != 0
|| limit.rlim_cur == RLIM_INFINITY)
/* The system limit is not usable. Use an architecture-specific
default. */
limit.rlim_cur = ARCH_STACK_DEFAULT_SIZE;
else if (limit.rlim_cur < PTHREAD_STACK_MIN)
/* The system limit is unusably small.
Use the minimal size acceptable. */
limit.rlim_cur = PTHREAD_STACK_MIN;
/* Make sure it meets the minimum size that allocate_stack
(allocatestack.c) will demand, which depends on the page size. */
const uintptr_t pagesz = GLRO(dl_pagesize);
const size_t minstack = pagesz + __static_tls_size + MINIMAL_REST_STACK;
if (limit.rlim_cur < minstack)
limit.rlim_cur = minstack;
/* Round the resource limit up to page size. */
limit.rlim_cur = ALIGN_UP (limit.rlim_cur, pagesz);
lll_lock (__default_pthread_attr_lock, LLL_PRIVATE);
__default_pthread_attr.stacksize = limit.rlim_cur;
__default_pthread_attr.guardsize = GLRO (dl_pagesize);
lll_unlock (__default_pthread_attr_lock, LLL_PRIVATE);
#ifdef SHARED
/* Transfer the old value from the dynamic linker's internal location. */
*__libc_dl_error_tsd () = *(*GL(dl_error_catch_tsd)) ();
GL(dl_error_catch_tsd) = &__libc_dl_error_tsd;
/* Make __rtld_lock_{,un}lock_recursive use pthread_mutex_{,un}lock,
keep the lock count from the ld.so implementation. */
GL(dl_rtld_lock_recursive) = (void *) __pthread_mutex_lock;
GL(dl_rtld_unlock_recursive) = (void *) __pthread_mutex_unlock;
unsigned int rtld_lock_count = GL(dl_load_lock).mutex.__data.__count;
GL(dl_load_lock).mutex.__data.__count = 0;
while (rtld_lock_count-- > 0)
__pthread_mutex_lock (&GL(dl_load_lock).mutex);
GL(dl_make_stack_executable_hook) = &__make_stacks_executable;
#endif
GL(dl_init_static_tls) = &__pthread_init_static_tls;
GL(dl_wait_lookup_done) = &__wait_lookup_done;
/* Register the fork generation counter with the libc. */
#ifndef TLS_MULTIPLE_THREADS_IN_TCB
__libc_multiple_threads_ptr =
#endif
__libc_pthread_init (&__fork_generation, __reclaim_stacks,
ptr_pthread_functions);
/* Determine whether the machine is SMP or not. */
__is_smp = is_smp_system ();
}
int
timer_create (clockid_t clock_id, struct sigevent *evp, timer_t *timerid)
{
#undef timer_create
{
clockid_t syscall_clockid = (clock_id == CLOCK_PROCESS_CPUTIME_ID
? MAKE_PROCESS_CPUCLOCK (0, CPUCLOCK_SCHED)
: clock_id == CLOCK_THREAD_CPUTIME_ID
? MAKE_THREAD_CPUCLOCK (0, CPUCLOCK_SCHED)
: clock_id);
/* If the user wants notification via a thread we need to handle
this special. */
if (evp == NULL
|| __builtin_expect (evp->sigev_notify != SIGEV_THREAD, 1))
{
struct sigevent local_evp;
/* We avoid allocating too much memory by basically
using struct timer as a derived class with the
first two elements being in the superclass. We only
need these two elements here. */
struct timer *newp = (struct timer *) malloc (offsetof (struct timer,
thrfunc));
if (newp == NULL)
/* No more memory. */
return -1;
if (evp == NULL)
{
/* The kernel has to pass up the timer ID which is a
userlevel object. Therefore we cannot leave it up to
the kernel to determine it. */
local_evp.sigev_notify = SIGEV_SIGNAL;
local_evp.sigev_signo = SIGALRM;
local_evp.sigev_value.sival_ptr = newp;
evp = &local_evp;
}
kernel_timer_t ktimerid;
int retval = INLINE_SYSCALL (timer_create, 3, syscall_clockid, evp,
&ktimerid);
if (retval != -1)
{
newp->sigev_notify = (evp != NULL
? evp->sigev_notify : SIGEV_SIGNAL);
newp->ktimerid = ktimerid;
*timerid = (timer_t) newp;
}
else
{
/* Cannot allocate the timer, fail. */
free (newp);
retval = -1;
}
return retval;
}
else
{
/* Create the helper thread. */
pthread_once (&__helper_once, __start_helper_thread);
if (__helper_tid == 0)
{
/* No resources to start the helper thread. */
__set_errno (EAGAIN);
return -1;
}
struct timer *newp;
newp = (struct timer *) malloc (sizeof (struct timer));
if (newp == NULL)
return -1;
/* Copy the thread parameters the user provided. */
newp->sival = evp->sigev_value;
newp->thrfunc = evp->sigev_notify_function;
newp->sigev_notify = SIGEV_THREAD;
/* We cannot simply copy the thread attributes since the
implementation might keep internal information for
each instance. */
(void) pthread_attr_init (&newp->attr);
if (evp->sigev_notify_attributes != NULL)
{
struct pthread_attr *nattr;
struct pthread_attr *oattr;
nattr = (struct pthread_attr *) &newp->attr;
oattr = (struct pthread_attr *) evp->sigev_notify_attributes;
nattr->schedparam = oattr->schedparam;
nattr->schedpolicy = oattr->schedpolicy;
nattr->flags = oattr->flags;
nattr->guardsize = oattr->guardsize;
nattr->stackaddr = oattr->stackaddr;
nattr->stacksize = oattr->stacksize;
}
/* In any case set the detach flag. */
(void) pthread_attr_setdetachstate (&newp->attr,
PTHREAD_CREATE_DETACHED);
/* Create the event structure for the kernel timer. */
struct sigevent sev =
{ .sigev_value.sival_ptr = newp,
.sigev_signo = SIGTIMER,
.sigev_notify = SIGEV_SIGNAL | SIGEV_THREAD_ID,
._sigev_un = { ._pad = { [0] = __helper_tid } } };
/* Create the timer. */
INTERNAL_SYSCALL_DECL (err);
int res;
res = INTERNAL_SYSCALL (timer_create, err, 3,
syscall_clockid, &sev, &newp->ktimerid);
if (! INTERNAL_SYSCALL_ERROR_P (res, err))
{
/* Add to the queue of active timers with thread
delivery. */
pthread_mutex_lock (&__active_timer_sigev_thread_lock);
newp->next = __active_timer_sigev_thread;
__active_timer_sigev_thread = newp;
pthread_mutex_unlock (&__active_timer_sigev_thread_lock);
*timerid = (timer_t) newp;
return 0;
}
/* Free the resources. */
free (newp);
__set_errno (INTERNAL_SYSCALL_ERRNO (res, err));
return -1;
}
}
int
clock_getcpuclockid (pid_t pid, clockid_t *clock_id)
{
#ifdef __NR_clock_getres
/* The clockid_t value is a simple computation from the PID.
But we do a clock_getres call to validate it. */
const clockid_t pidclock = MAKE_PROCESS_CPUCLOCK (pid, CPUCLOCK_SCHED);
# if !(__ASSUME_POSIX_CPU_TIMERS > 0)
extern int __libc_missing_posix_cpu_timers attribute_hidden;
# if !(__ASSUME_POSIX_TIMERS > 0)
extern int __libc_missing_posix_timers attribute_hidden;
if (__libc_missing_posix_timers && !__libc_missing_posix_cpu_timers)
__libc_missing_posix_cpu_timers = 1;
# endif
if (!__libc_missing_posix_cpu_timers)
# endif
{
INTERNAL_SYSCALL_DECL (err);
int r = INTERNAL_SYSCALL (clock_getres, err, 2, pidclock, NULL);
if (!INTERNAL_SYSCALL_ERROR_P (r, err))
{
*clock_id = pidclock;
return 0;
}
# if !(__ASSUME_POSIX_TIMERS > 0)
if (INTERNAL_SYSCALL_ERRNO (r, err) == ENOSYS)
{
/* The kernel doesn't support these calls at all. */
__libc_missing_posix_timers = 1;
__libc_missing_posix_cpu_timers = 1;
}
else
# endif
if (INTERNAL_SYSCALL_ERRNO (r, err) == EINVAL)
{
# if !(__ASSUME_POSIX_CPU_TIMERS > 0)
if (pidclock == MAKE_PROCESS_CPUCLOCK (0, CPUCLOCK_SCHED)
|| INTERNAL_SYSCALL_ERROR_P (INTERNAL_SYSCALL
(clock_getres, err, 2,
MAKE_PROCESS_CPUCLOCK
(0, CPUCLOCK_SCHED), NULL),
err))
/* The kernel doesn't support these clocks at all. */
__libc_missing_posix_cpu_timers = 1;
else
# endif
/* The clock_getres system call checked the PID for us. */
return ESRCH;
}
else
return INTERNAL_SYSCALL_ERRNO (r, err);
}
#endif
/* We don't allow any process ID but our own. */
if (pid != 0 && pid != getpid ())
return EPERM;
#ifdef CLOCK_PROCESS_CPUTIME_ID
if (HAS_CPUCLOCK)
{
/* Store the number. */
*clock_id = CLOCK_PROCESS_CPUTIME_ID;
return 0;
}
#endif
/* We don't have a timer for that. */
return ENOENT;
}
static int
__pthread_mutex_lock_full (pthread_mutex_t *mutex)
{
int oldval;
pid_t id = THREAD_GETMEM (THREAD_SELF, tid);
switch (PTHREAD_MUTEX_TYPE (mutex))
{
case PTHREAD_MUTEX_ROBUST_RECURSIVE_NP:
case PTHREAD_MUTEX_ROBUST_ERRORCHECK_NP:
case PTHREAD_MUTEX_ROBUST_NORMAL_NP:
case PTHREAD_MUTEX_ROBUST_ADAPTIVE_NP:
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending,
&mutex->__data.__list.__next);
oldval = mutex->__data.__lock;
do
{
again:
if ((oldval & FUTEX_OWNER_DIED) != 0)
{
/* The previous owner died. Try locking the mutex. */
int newval = id;
#ifdef NO_INCR
newval |= FUTEX_WAITERS;
#else
newval |= (oldval & FUTEX_WAITERS);
#endif
newval
= atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
newval, oldval);
if (newval != oldval)
{
oldval = newval;
goto again;
}
/* We got the mutex. */
mutex->__data.__count = 1;
/* But it is inconsistent unless marked otherwise. */
mutex->__data.__owner = PTHREAD_MUTEX_INCONSISTENT;
ENQUEUE_MUTEX (mutex);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
/* Note that we deliberately exit here. If we fall
through to the end of the function __nusers would be
incremented which is not correct because the old
owner has to be discounted. If we are not supposed
to increment __nusers we actually have to decrement
it here. */
#ifdef NO_INCR
--mutex->__data.__nusers;
#endif
return EOWNERDEAD;
}
/* Check whether we already hold the mutex. */
if (__builtin_expect ((oldval & FUTEX_TID_MASK) == id, 0))
{
int kind = PTHREAD_MUTEX_TYPE (mutex);
if (kind == PTHREAD_MUTEX_ROBUST_ERRORCHECK_NP)
{
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending,
NULL);
return EDEADLK;
}
if (kind == PTHREAD_MUTEX_ROBUST_RECURSIVE_NP)
{
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending,
NULL);
/* Just bump the counter. */
if (__builtin_expect (mutex->__data.__count + 1 == 0, 0))
/* Overflow of the counter. */
return EAGAIN;
++mutex->__data.__count;
return 0;
}
}
oldval = LLL_ROBUST_MUTEX_LOCK (mutex, id);
if (__builtin_expect (mutex->__data.__owner
== PTHREAD_MUTEX_NOTRECOVERABLE, 0))
{
/* This mutex is now not recoverable. */
mutex->__data.__count = 0;
lll_unlock (mutex->__data.__lock,
PTHREAD_ROBUST_MUTEX_PSHARED (mutex));
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
return ENOTRECOVERABLE;
}
}
while ((oldval & FUTEX_OWNER_DIED) != 0);
mutex->__data.__count = 1;
ENQUEUE_MUTEX (mutex);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
break;
case PTHREAD_MUTEX_PI_RECURSIVE_NP:
case PTHREAD_MUTEX_PI_ERRORCHECK_NP:
case PTHREAD_MUTEX_PI_NORMAL_NP:
case PTHREAD_MUTEX_PI_ADAPTIVE_NP:
case PTHREAD_MUTEX_PI_ROBUST_RECURSIVE_NP:
case PTHREAD_MUTEX_PI_ROBUST_ERRORCHECK_NP:
case PTHREAD_MUTEX_PI_ROBUST_NORMAL_NP:
case PTHREAD_MUTEX_PI_ROBUST_ADAPTIVE_NP:
{
int kind = mutex->__data.__kind & PTHREAD_MUTEX_KIND_MASK_NP;
int robust = mutex->__data.__kind & PTHREAD_MUTEX_ROBUST_NORMAL_NP;
if (robust)
/* Note: robust PI futexes are signaled by setting bit 0. */
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending,
(void *) (((uintptr_t) &mutex->__data.__list.__next)
| 1));
oldval = mutex->__data.__lock;
/* Check whether we already hold the mutex. */
if (__builtin_expect ((oldval & FUTEX_TID_MASK) == id, 0))
{
if (kind == PTHREAD_MUTEX_ERRORCHECK_NP)
{
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
return EDEADLK;
}
if (kind == PTHREAD_MUTEX_RECURSIVE_NP)
{
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
/* Just bump the counter. */
if (__builtin_expect (mutex->__data.__count + 1 == 0, 0))
/* Overflow of the counter. */
return EAGAIN;
++mutex->__data.__count;
return 0;
}
}
int newval = id;
#ifdef NO_INCR
newval |= FUTEX_WAITERS;
#endif
oldval = atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
newval, 0);
if (oldval != 0)
{
/* The mutex is locked. The kernel will now take care of
everything. */
int private = (robust
? PTHREAD_ROBUST_MUTEX_PSHARED (mutex)
: PTHREAD_MUTEX_PSHARED (mutex));
INTERNAL_SYSCALL_DECL (__err);
int e = INTERNAL_SYSCALL (futex, __err, 4, &mutex->__data.__lock,
__lll_private_flag (FUTEX_LOCK_PI,
private), 1, 0);
if (INTERNAL_SYSCALL_ERROR_P (e, __err)
&& (INTERNAL_SYSCALL_ERRNO (e, __err) == ESRCH
|| INTERNAL_SYSCALL_ERRNO (e, __err) == EDEADLK))
{
assert (INTERNAL_SYSCALL_ERRNO (e, __err) != EDEADLK
|| (kind != PTHREAD_MUTEX_ERRORCHECK_NP
&& kind != PTHREAD_MUTEX_RECURSIVE_NP));
/* ESRCH can happen only for non-robust PI mutexes where
the owner of the lock died. */
assert (INTERNAL_SYSCALL_ERRNO (e, __err) != ESRCH || !robust);
/* Delay the thread indefinitely. */
while (1)
pause_not_cancel ();
}
oldval = mutex->__data.__lock;
assert (robust || (oldval & FUTEX_OWNER_DIED) == 0);
}
if (__builtin_expect (oldval & FUTEX_OWNER_DIED, 0))
{
atomic_and (&mutex->__data.__lock, ~FUTEX_OWNER_DIED);
/* We got the mutex. */
mutex->__data.__count = 1;
/* But it is inconsistent unless marked otherwise. */
mutex->__data.__owner = PTHREAD_MUTEX_INCONSISTENT;
ENQUEUE_MUTEX_PI (mutex);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
/* Note that we deliberately exit here. If we fall
through to the end of the function __nusers would be
incremented which is not correct because the old owner
has to be discounted. If we are not supposed to
increment __nusers we actually have to decrement it here. */
#ifdef NO_INCR
--mutex->__data.__nusers;
#endif
return EOWNERDEAD;
}
if (robust
&& __builtin_expect (mutex->__data.__owner
== PTHREAD_MUTEX_NOTRECOVERABLE, 0))
{
/* This mutex is now not recoverable. */
mutex->__data.__count = 0;
INTERNAL_SYSCALL_DECL (__err);
INTERNAL_SYSCALL (futex, __err, 4, &mutex->__data.__lock,
__lll_private_flag (FUTEX_UNLOCK_PI,
PTHREAD_ROBUST_MUTEX_PSHARED (mutex)),
0, 0);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
return ENOTRECOVERABLE;
}
mutex->__data.__count = 1;
if (robust)
{
ENQUEUE_MUTEX_PI (mutex);
THREAD_SETMEM (THREAD_SELF, robust_head.list_op_pending, NULL);
}
}
break;
case PTHREAD_MUTEX_PP_RECURSIVE_NP:
case PTHREAD_MUTEX_PP_ERRORCHECK_NP:
case PTHREAD_MUTEX_PP_NORMAL_NP:
case PTHREAD_MUTEX_PP_ADAPTIVE_NP:
{
int kind = mutex->__data.__kind & PTHREAD_MUTEX_KIND_MASK_NP;
oldval = mutex->__data.__lock;
/* Check whether we already hold the mutex. */
if (mutex->__data.__owner == id)
{
if (kind == PTHREAD_MUTEX_ERRORCHECK_NP)
return EDEADLK;
if (kind == PTHREAD_MUTEX_RECURSIVE_NP)
{
/* Just bump the counter. */
if (__builtin_expect (mutex->__data.__count + 1 == 0, 0))
/* Overflow of the counter. */
return EAGAIN;
++mutex->__data.__count;
return 0;
}
}
int oldprio = -1, ceilval;
do
{
int ceiling = (oldval & PTHREAD_MUTEX_PRIO_CEILING_MASK)
>> PTHREAD_MUTEX_PRIO_CEILING_SHIFT;
if (__pthread_current_priority () > ceiling)
{
if (oldprio != -1)
__pthread_tpp_change_priority (oldprio, -1);
return EINVAL;
}
int retval = __pthread_tpp_change_priority (oldprio, ceiling);
if (retval)
return retval;
ceilval = ceiling << PTHREAD_MUTEX_PRIO_CEILING_SHIFT;
oldprio = ceiling;
oldval
= atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
#ifdef NO_INCR
ceilval | 2,
#else
ceilval | 1,
#endif
ceilval);
if (oldval == ceilval)
break;
do
{
oldval
= atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
ceilval | 2,
ceilval | 1);
if ((oldval & PTHREAD_MUTEX_PRIO_CEILING_MASK) != ceilval)
break;
if (oldval != ceilval)
lll_futex_wait (&mutex->__data.__lock, ceilval | 2,
PTHREAD_MUTEX_PSHARED (mutex));
}
while (atomic_compare_and_exchange_val_acq (&mutex->__data.__lock,
ceilval | 2, ceilval)
!= ceilval);
}
while ((oldval & PTHREAD_MUTEX_PRIO_CEILING_MASK) != ceilval);
assert (mutex->__data.__owner == 0);
mutex->__data.__count = 1;
}
break;
default:
/* Correct code cannot set any other type. */
return EINVAL;
}
hp_timing_t
__get_clockfreq (void)
{
/* We read the information from the /proc filesystem. /proc/cpuinfo
contains at least one line like:
timebase : 33333333
We search for this line and convert the number into an integer. */
static hp_timing_t timebase_freq;
hp_timing_t result = 0L;
/* If this function was called before, we know the result. */
if (timebase_freq != 0)
return timebase_freq;
/* If we can use the vDSO to obtain the timebase even better. */
#ifdef SHARED
INTERNAL_SYSCALL_DECL (err);
timebase_freq =
INTERNAL_VSYSCALL_NO_SYSCALL_FALLBACK (get_tbfreq, err, hp_timing_t, 0);
if (INTERNAL_SYSCALL_ERROR_P (timebase_freq, err)
&& INTERNAL_SYSCALL_ERRNO (timebase_freq, err) == ENOSYS)
#endif
{
int fd = __open ("/proc/cpuinfo", O_RDONLY);
if (__builtin_expect (fd != -1, 1))
{
/* The timebase will be in the 1st 1024 bytes for systems with up
to 8 processors. If the first read returns less then 1024
bytes read, we have the whole cpuinfo and can start the scan.
Otherwise we will have to read more to insure we have the
timebase value in the scan. */
char buf[1024];
ssize_t n;
n = __read (fd, buf, sizeof (buf));
if (n == sizeof (buf))
{
/* We are here because the 1st read returned exactly sizeof
(buf) bytes. This implies that we are not at EOF and may
not have read the timebase value yet. So we need to read
more bytes until we know we have EOF. We copy the lower
half of buf to the upper half and read sizeof (buf)/2
bytes into the lower half of buf and repeat until we
reach EOF. We can assume that the timebase will be in
the last 512 bytes of cpuinfo, so two 512 byte half_bufs
will be sufficient to contain the timebase and will
handle the case where the timebase spans the half_buf
boundry. */
const ssize_t half_buf = sizeof (buf) / 2;
while (n >= half_buf)
{
memcpy (buf, buf + half_buf, half_buf);
n = __read (fd, buf + half_buf, half_buf);
}
if (n >= 0)
n += half_buf;
}
if (__builtin_expect (n, 1) > 0)
{
char *mhz = memmem (buf, n, "timebase", 7);
if (__builtin_expect (mhz != NULL, 1))
{
char *endp = buf + n;
/* Search for the beginning of the string. */
while (mhz < endp && (*mhz < '0' || *mhz > '9')
&& *mhz != '\n')
++mhz;
while (mhz < endp && *mhz != '\n')
{
if (*mhz >= '0' && *mhz <= '9')
{
result *= 10;
result += *mhz - '0';
}
++mhz;
}
}
timebase_freq = result;
}
__close (fd);
}
}
return timebase_freq;
}
static int
do_clone (struct pthread *pd, const struct pthread_attr *attr,
int clone_flags, int (*fct) (void *), STACK_VARIABLES_PARMS,
int stopped)
{
#ifdef PREPARE_CREATE
PREPARE_CREATE;
#endif
if (__builtin_expect (stopped != 0, 0))
/* We make sure the thread does not run far by forcing it to get a
lock. We lock it here too so that the new thread cannot continue
until we tell it to. */
lll_lock (pd->lock, LLL_PRIVATE);
/* One more thread. We cannot have the thread do this itself, since it
might exist but not have been scheduled yet by the time we've returned
and need to check the value to behave correctly. We must do it before
creating the thread, in case it does get scheduled first and then
might mistakenly think it was the only thread. In the failure case,
we momentarily store a false value; this doesn't matter because there
is no kosher thing a signal handler interrupting us right here can do
that cares whether the thread count is correct. */
atomic_increment (&__nptl_nthreads);
int rc = ARCH_CLONE (fct, STACK_VARIABLES_ARGS, clone_flags,
pd, &pd->tid, TLS_VALUE, &pd->tid);
if (__builtin_expect (rc == -1, 0))
{
atomic_decrement (&__nptl_nthreads); /* Oops, we lied for a second. */
/* Perhaps a thread wants to change the IDs and if waiting
for this stillborn thread. */
if (__builtin_expect (atomic_exchange_acq (&pd->setxid_futex, 0)
== -2, 0))
lll_futex_wake (&pd->setxid_futex, 1, LLL_PRIVATE);
/* Free the resources. */
__deallocate_stack (pd);
/* We have to translate error codes. */
return errno == ENOMEM ? EAGAIN : errno;
}
/* Now we have the possibility to set scheduling parameters etc. */
if (__builtin_expect (stopped != 0, 0))
{
INTERNAL_SYSCALL_DECL (err);
int res = 0;
/* Set the affinity mask if necessary. */
if (attr->cpuset != NULL)
{
res = INTERNAL_SYSCALL (sched_setaffinity, err, 3, pd->tid,
attr->cpusetsize, attr->cpuset);
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (res, err), 0))
{
/* The operation failed. We have to kill the thread. First
send it the cancellation signal. */
INTERNAL_SYSCALL_DECL (err2);
err_out:
(void) INTERNAL_SYSCALL (tgkill, err2, 3,
THREAD_GETMEM (THREAD_SELF, pid),
pd->tid, SIGCANCEL);
/* We do not free the stack here because the canceled thread
itself will do this. */
return (INTERNAL_SYSCALL_ERROR_P (res, err)
? INTERNAL_SYSCALL_ERRNO (res, err)
: 0);
}
}
/* Set the scheduling parameters. */
if ((attr->flags & ATTR_FLAG_NOTINHERITSCHED) != 0)
{
res = INTERNAL_SYSCALL (sched_setscheduler, err, 3, pd->tid,
pd->schedpolicy, &pd->schedparam);
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (res, err), 0))
goto err_out;
}
}
/* We now have for sure more than one thread. The main thread might
not yet have the flag set. No need to set the global variable
again if this is what we use. */
THREAD_SETMEM (THREAD_SELF, header.multiple_threads, 1);
return 0;
}
/* Get information about the file NAME in BUF. */
int
__fxstatat (int vers, int fd, const char *file, struct stat *st, int flag)
{
INTERNAL_SYSCALL_DECL (err);
int result, errno_out;
/* ??? The __fxstatat entry point is new enough that it must be using
vers == _STAT_VER_KERNEL64. For the benefit of dl-fxstatat64.c, we
cannot actually check this, lest the compiler not optimize the rest
of the function away. */
#ifdef __NR_fstatat64
if (__have_atfcts >= 0)
{
result = INTERNAL_SYSCALL (fstatat64, err, 4, fd, file, st, flag);
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return result;
errno_out = INTERNAL_SYSCALL_ERRNO (result, err);
#ifndef __ASSUME_ATFCTS
if (errno_out == ENOSYS)
__have_atfcts = -1;
else
#endif
{
__set_errno (errno_out);
return -1;
}
}
#endif /* __NR_fstatat64 */
if (flag & ~AT_SYMLINK_NOFOLLOW)
{
__set_errno (EINVAL);
return -1;
}
char *buf = NULL;
if (fd != AT_FDCWD && file[0] != '/')
{
size_t filelen = strlen (file);
if (__builtin_expect (filelen == 0, 0))
{
__set_errno (ENOENT);
return -1;
}
static const char procfd[] = "/proc/self/fd/%d/%s";
/* Buffer for the path name we are going to use. It consists of
- the string /proc/self/fd/
- the file descriptor number
- the file name provided.
The final NUL is included in the sizeof. A bit of overhead
due to the format elements compensates for possible negative
numbers. */
size_t buflen = sizeof (procfd) + sizeof (int) * 3 + filelen;
buf = alloca (buflen);
__snprintf (buf, buflen, procfd, fd, file);
file = buf;
}
#ifdef __NR_stat64
if (!__libc_missing_axp_stat64)
{
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat64, err, 2, file, st);
else
result = INTERNAL_SYSCALL (stat64, err, 2, file, st);
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return result;
errno_out = INTERNAL_SYSCALL_ERRNO (result, err);
# if __ASSUME_STAT64_SYSCALL == 0
if (errno_out == ENOSYS)
__libc_missing_axp_stat64 = 1;
else
# endif
goto fail;
}
#endif /* __NR_stat64 */
struct kernel_stat kst;
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat, err, 2, file, &kst);
else
result = INTERNAL_SYSCALL (stat, err, 2, file, &kst);
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return __xstat_conv (vers, &kst, st);
errno_out = INTERNAL_SYSCALL_ERRNO (result, err);
fail:
__atfct_seterrno (errno_out, fd, buf);
return -1;
}
int
fchownat (int fd, const char *file, uid_t owner, gid_t group, int flag)
{
int result;
#ifdef __NR_fchownat
# ifndef __ASSUME_ATFCTS
if (__have_atfcts >= 0)
# endif
{
result = INLINE_SYSCALL (fchownat, 5, fd, file, owner, group, flag);
# ifndef __ASSUME_ATFCTS
if (result == -1 && errno == ENOSYS)
__have_atfcts = -1;
else
# endif
return result;
}
#endif
#ifndef __ASSUME_ATFCTS
if (flag & ~AT_SYMLINK_NOFOLLOW)
{
__set_errno (EINVAL);
return -1;
}
char *buf = NULL;
if (fd != AT_FDCWD && file[0] != '/')
{
size_t filelen = strlen (file);
if (__builtin_expect (filelen == 0, 0))
{
__set_errno (ENOENT);
return -1;
}
static const char procfd[] = "/proc/self/fd/%d/%s";
/* Buffer for the path name we are going to use. It consists of
- the string /proc/self/fd/
- the file descriptor number
- the file name provided.
The final NUL is included in the sizeof. A bit of overhead
due to the format elements compensates for possible negative
numbers. */
size_t buflen = sizeof (procfd) + sizeof (int) * 3 + filelen;
buf = alloca (buflen);
__snprintf (buf, buflen, procfd, fd, file);
file = buf;
}
INTERNAL_SYSCALL_DECL (err);
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lchown32, err, 3, CHECK_STRING (file), owner,
group);
else
result = INTERNAL_SYSCALL (chown32, err, 3, CHECK_STRING (file), owner,
group);
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 0))
{
__atfct_seterrno (INTERNAL_SYSCALL_ERRNO (result, err), fd, buf);
return -1;
}
return result;
#endif
}
int
__fxstatat64 (int vers, int fd, const char *file, struct stat64 *st, int flag)
{
if (__builtin_expect (vers != _STAT_VER_LINUX, 0))
{
__set_errno (EINVAL);
return -1;
}
int result;
INTERNAL_SYSCALL_DECL (err);
#ifdef __NR_fstatat64
# ifndef __ASSUME_ATFCTS
if (__have_atfcts >= 0)
# endif
{
result = INTERNAL_SYSCALL (fstatat64, err, 4, fd, file, st, flag);
# ifndef __ASSUME_ATFCTS
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 1)
&& INTERNAL_SYSCALL_ERRNO (result, err) == ENOSYS)
__have_atfcts = -1;
else
# endif
if (!__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return 0;
else
{
__set_errno (INTERNAL_SYSCALL_ERRNO (result, err));
return -1;
}
}
#endif
#ifndef __ASSUME_ATFCTS
if (flag & ~AT_SYMLINK_NOFOLLOW)
{
__set_errno (EINVAL);
return -1;
}
char *buf = NULL;
if (fd != AT_FDCWD && file[0] != '/')
{
size_t filelen = strlen (file);
static const char procfd[] = "/proc/self/fd/%d/%s";
/* Buffer for the path name we are going to use. It consists of
- the string /proc/self/fd/
- the file descriptor number
- the file name provided.
The final NUL is included in the sizeof. A bit of overhead
due to the format elements compensates for possible negative
numbers. */
size_t buflen = sizeof (procfd) + sizeof (int) * 3 + filelen;
buf = alloca (buflen);
__snprintf (buf, buflen, procfd, fd, file);
file = buf;
}
# if __ASSUME_STAT64_SYSCALL > 0
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat64, err, 2, CHECK_STRING (file),
CHECK_1 (st));
else
result = INTERNAL_SYSCALL (stat64, err, 2, CHECK_STRING (file),
CHECK_1 (st));
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
{
# if defined _HAVE_STAT64___ST_INO && __ASSUME_ST_INO_64_BIT == 0
if (st->__st_ino != (__ino_t) st->st_ino)
st->st_ino = st->__st_ino;
# endif
return result;
}
# else
struct kernel_stat kst;
# ifdef __NR_stat64
if (! __have_no_stat64)
{
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat64, err, 2, CHECK_STRING (file),
CHECK_1 (st));
else
result = INTERNAL_SYSCALL (stat64, err, 2, CHECK_STRING (file),
CHECK_1 (st));
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
{
# if defined _HAVE_STAT64___ST_INO && __ASSUME_ST_INO_64_BIT == 0
if (st->__st_ino != (__ino_t) st->st_ino)
st->st_ino = st->__st_ino;
# endif
return result;
}
if (INTERNAL_SYSCALL_ERRNO (result, err) != ENOSYS)
goto fail;
__have_no_stat64 = 1;
}
# endif
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lstat, err, 2, CHECK_STRING (file),
__ptrvalue (&kst));
else
result = INTERNAL_SYSCALL (stat, err, 2, CHECK_STRING (file),
__ptrvalue (&kst));
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return __xstat64_conv (vers, &kst, st);
fail:
# endif
__atfct_seterrno (INTERNAL_SYSCALL_ERRNO (result, err), fd, buf);
return -1;
#endif
}
int
pthread_cancel (
pthread_t th)
{
volatile struct pthread *pd = (volatile struct pthread *) th;
/* Make sure the descriptor is valid. */
if (INVALID_TD_P (pd))
/* Not a valid thread handle. */
return ESRCH;
#ifdef SHARED
pthread_cancel_init ();
#endif
int result = 0;
int oldval;
int newval;
do
{
again:
oldval = pd->cancelhandling;
newval = oldval | CANCELING_BITMASK | CANCELED_BITMASK;
/* Avoid doing unnecessary work. The atomic operation can
potentially be expensive if the bug has to be locked and
remote cache lines have to be invalidated. */
if (oldval == newval)
break;
/* If the cancellation is handled asynchronously just send a
signal. We avoid this if possible since it's more
expensive. */
if (CANCEL_ENABLED_AND_CANCELED_AND_ASYNCHRONOUS (newval))
{
/* Mark the cancellation as "in progress". */
if (atomic_compare_and_exchange_bool_acq (&pd->cancelhandling,
oldval | CANCELING_BITMASK,
oldval))
goto again;
/* The cancellation handler will take care of marking the
thread as canceled. */
INTERNAL_SYSCALL_DECL (err);
/* One comment: The PID field in the TCB can temporarily be
changed (in fork). But this must not affect this code
here. Since this function would have to be called while
the thread is executing fork, it would have to happen in
a signal handler. But this is no allowed, pthread_cancel
is not guaranteed to be async-safe. */
int val;
#if defined(__ASSUME_TGKILL) && __ASSUME_TGKILL
val = INTERNAL_SYSCALL (tgkill, err, 3,
THREAD_GETMEM (THREAD_SELF, pid), pd->tid,
SIGCANCEL);
#else
# ifdef __NR_tgkill
val = INTERNAL_SYSCALL (tgkill, err, 3,
THREAD_GETMEM (THREAD_SELF, pid), pd->tid,
SIGCANCEL);
if (INTERNAL_SYSCALL_ERROR_P (val, err)
&& INTERNAL_SYSCALL_ERRNO (val, err) == ENOSYS)
# endif
val = INTERNAL_SYSCALL (tkill, err, 2, pd->tid, SIGCANCEL);
#endif
if (INTERNAL_SYSCALL_ERROR_P (val, err))
result = INTERNAL_SYSCALL_ERRNO (val, err);
break;
}
}
/* Mark the thread as canceled. This has to be done
atomically since other bits could be modified as well. */
while (atomic_compare_and_exchange_bool_acq (&pd->cancelhandling, newval,
oldval));
return result;
}
int
fchownat (int fd, const char *file, uid_t owner, gid_t group, int flag)
{
if (flag & ~AT_SYMLINK_NOFOLLOW)
{
__set_errno (EINVAL);
return -1;
}
char *buf = NULL;
if (fd != AT_FDCWD && file[0] != '/')
{
size_t filelen = strlen (file);
if (__builtin_expect (filelen == 0, 0))
{
__set_errno (ENOENT);
return -1;
}
static const char procfd[] = "/proc/self/fd/%d/%s";
/* Buffer for the path name we are going to use. It consists of
- the string /proc/self/fd/
- the file descriptor number
- the file name provided.
The final NUL is included in the sizeof. A bit of overhead
due to the format elements compensates for possible negative
numbers. */
size_t buflen = sizeof (procfd) + sizeof (int) * 3 + filelen;
buf = alloca (buflen);
__snprintf (buf, buflen, procfd, fd, file);
file = buf;
}
int result;
INTERNAL_SYSCALL_DECL (err);
#if __ASSUME_32BITUIDS > 0
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lchown32, err, 3, CHECK_STRING (file), owner,
group);
else
result = INTERNAL_SYSCALL (chown32, err, 3, CHECK_STRING (file), owner,
group);
#else
# ifdef __NR_chown32
if (__libc_missing_32bit_uids <= 0)
{
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lchown32, err, 3, CHECK_STRING (file),
owner, group);
else
result = INTERNAL_SYSCALL (chown32, err, 3, CHECK_STRING (file), owner,
group);
if (__builtin_expect (!INTERNAL_SYSCALL_ERROR_P (result, err), 1))
return result;
if (INTERNAL_SYSCALL_ERRNO (result, err) != ENOSYS)
goto fail;
__libc_missing_32bit_uids = 1;
}
# endif /* __NR_chown32 */
if (((owner + 1) > (gid_t) ((__kernel_uid_t) -1U))
|| ((group + 1) > (gid_t) ((__kernel_gid_t) -1U)))
{
__set_errno (EINVAL);
return -1;
}
if (flag & AT_SYMLINK_NOFOLLOW)
result = INTERNAL_SYSCALL (lchown, err, 3, CHECK_STRING (file), owner,
group);
else
result = INTERNAL_SYSCALL (chown, err, 3, CHECK_STRING (file), owner,
group);
#endif
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (result, err), 0))
{
fail:
__atfct_seterrno (INTERNAL_SYSCALL_ERRNO (result, err), fd, buf);
result = -1;
}
return result;
}
static int
create_thread (struct pthread *pd, const struct pthread_attr *attr,
bool stopped_start, STACK_VARIABLES_PARMS, bool *thread_ran)
{
/* Determine whether the newly created threads has to be started
stopped since we have to set the scheduling parameters or set the
affinity. */
if (attr != NULL
&& (__glibc_unlikely (attr->cpuset != NULL)
|| __glibc_unlikely ((attr->flags & ATTR_FLAG_NOTINHERITSCHED) != 0)))
stopped_start = true;
pd->stopped_start = stopped_start;
if (__glibc_unlikely (stopped_start))
/* We make sure the thread does not run far by forcing it to get a
lock. We lock it here too so that the new thread cannot continue
until we tell it to. */
lll_lock (pd->lock, LLL_PRIVATE);
/* We rely heavily on various flags the CLONE function understands:
CLONE_VM, CLONE_FS, CLONE_FILES
These flags select semantics with shared address space and
file descriptors according to what POSIX requires.
CLONE_SIGHAND, CLONE_THREAD
This flag selects the POSIX signal semantics and various
other kinds of sharing (itimers, POSIX timers, etc.).
CLONE_SETTLS
The sixth parameter to CLONE determines the TLS area for the
new thread.
CLONE_PARENT_SETTID
The kernels writes the thread ID of the newly created thread
into the location pointed to by the fifth parameters to CLONE.
Note that it would be semantically equivalent to use
CLONE_CHILD_SETTID but it is be more expensive in the kernel.
CLONE_CHILD_CLEARTID
The kernels clears the thread ID of a thread that has called
sys_exit() in the location pointed to by the seventh parameter
to CLONE.
The termination signal is chosen to be zero which means no signal
is sent. */
const int clone_flags = (CLONE_VM | CLONE_FS | CLONE_FILES | CLONE_SYSVSEM
| CLONE_SIGHAND | CLONE_THREAD
| CLONE_SETTLS | CLONE_PARENT_SETTID
| CLONE_CHILD_CLEARTID
| 0);
TLS_DEFINE_INIT_TP (tp, pd);
if (__glibc_unlikely (ARCH_CLONE (&start_thread, STACK_VARIABLES_ARGS,
clone_flags, pd, &pd->tid, tp, &pd->tid)
== -1))
return errno;
/* It's started now, so if we fail below, we'll have to cancel it
and let it clean itself up. */
*thread_ran = true;
/* Now we have the possibility to set scheduling parameters etc. */
if (attr != NULL)
{
INTERNAL_SYSCALL_DECL (err);
int res;
/* Set the affinity mask if necessary. */
if (attr->cpuset != NULL)
{
assert (stopped_start);
res = INTERNAL_SYSCALL (sched_setaffinity, err, 3, pd->tid,
attr->cpusetsize, attr->cpuset);
if (__glibc_unlikely (INTERNAL_SYSCALL_ERROR_P (res, err)))
err_out:
{
/* The operation failed. We have to kill the thread.
We let the normal cancellation mechanism do the work. */
INTERNAL_SYSCALL_DECL (err2);
(void) INTERNAL_SYSCALL (tgkill, err2, 3,
THREAD_GETMEM (THREAD_SELF, pid),
pd->tid, SIGCANCEL);
return INTERNAL_SYSCALL_ERRNO (res, err);
}
}
/* Set the scheduling parameters. */
if ((attr->flags & ATTR_FLAG_NOTINHERITSCHED) != 0)
{
assert (stopped_start);
res = INTERNAL_SYSCALL (sched_setscheduler, err, 3, pd->tid,
pd->schedpolicy, &pd->schedparam);
if (__glibc_unlikely (INTERNAL_SYSCALL_ERROR_P (res, err)))
goto err_out;
}
}
return 0;
}
static int
do_clone (struct pthread *pd, const struct pthread_attr *attr,
int clone_flags, int (*fct) (void *), STACK_VARIABLES_PARMS,
int stopped)
{
#ifdef PREPARE_CREATE
PREPARE_CREATE;
#endif
if (stopped)
/* We Make sure the thread does not run far by forcing it to get a
lock. We lock it here too so that the new thread cannot continue
until we tell it to. */
lll_lock (pd->lock);
/* One more thread. We cannot have the thread do this itself, since it
might exist but not have been scheduled yet by the time we've returned
and need to check the value to behave correctly. We must do it before
creating the thread, in case it does get scheduled first and then
might mistakenly think it was the only thread. In the failure case,
we momentarily store a false value; this doesn't matter because there
is no kosher thing a signal handler interrupting us right here can do
that cares whether the thread count is correct. */
atomic_increment (&__nptl_nthreads);
if (ARCH_CLONE (fct, STACK_VARIABLES_ARGS, clone_flags,
pd, &pd->tid, TLS_VALUE, &pd->tid) == -1)
{
atomic_decrement (&__nptl_nthreads); /* Oops, we lied for a second. */
/* Failed. If the thread is detached, remove the TCB here since
the caller cannot do this. The caller remembered the thread
as detached and cannot reverify that it is not since it must
not access the thread descriptor again. */
if (IS_DETACHED (pd))
__deallocate_stack (pd);
return errno;
}
/* Now we have the possibility to set scheduling parameters etc. */
if (__builtin_expect (stopped != 0, 0))
{
INTERNAL_SYSCALL_DECL (err);
int res = 0;
/* Set the affinity mask if necessary. */
if (attr->cpuset != NULL)
{
res = INTERNAL_SYSCALL (sched_setaffinity, err, 3, pd->tid,
sizeof (cpu_set_t), attr->cpuset);
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (res, err), 0))
{
/* The operation failed. We have to kill the thread. First
send it the cancellation signal. */
INTERNAL_SYSCALL_DECL (err2);
err_out:
#if __ASSUME_TGKILL
(void) INTERNAL_SYSCALL (tgkill, err2, 3,
THREAD_GETMEM (THREAD_SELF, pid),
pd->tid, SIGCANCEL);
#else
(void) INTERNAL_SYSCALL (tkill, err2, 2, pd->tid, SIGCANCEL);
#endif
return (INTERNAL_SYSCALL_ERROR_P (res, err)
? INTERNAL_SYSCALL_ERRNO (res, err)
: 0);
}
}
/* Set the scheduling parameters. */
if ((attr->flags & ATTR_FLAG_NOTINHERITSCHED) != 0)
{
res = INTERNAL_SYSCALL (sched_setscheduler, err, 3, pd->tid,
pd->schedpolicy, &pd->schedparam);
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (res, err), 0))
goto err_out;
}
}
/* We now have for sure more than one thread. The main thread might
not yet have the flag set. No need to set the global variable
again if this is what we use. */
THREAD_SETMEM (THREAD_SELF, header.multiple_threads, 1);
return 0;
}
static int
do_clone (struct pthread *pd, const struct pthread_attr *attr,
int clone_flags, int (*fct) (void *), STACK_VARIABLES_PARMS,
int stopped)
{
#if 0
PREPARE_CREATE;
#endif
if (__builtin_expect (stopped != 0, 0))
/* We make sure the thread does not run far by forcing it to get a
lock. We lock it here too so that the new thread cannot continue
until we tell it to. */
lll_lock (pd->lock, LLL_PRIVATE);
/* One more thread. We cannot have the thread do this itself, since it
might exist but not have been scheduled yet by the time we've returned
and need to check the value to behave correctly. We must do it before
creating the thread, in case it does get scheduled first and then
might mistakenly think it was the only thread. In the failure case,
we momentarily store a false value; this doesn't matter because there
is no kosher thing a signal handler interrupting us right here can do
that cares whether the thread count is correct. */
atomic_increment (&__nptl_nthreads);
#if !defined(__native_client__) && !defined(__ZRT_HOST)
#error "This code was changed to work only in Native Client"
#endif
/* Native Client does not have a notion of a thread ID, so we make
one up. This must be small enough to leave space for number identifying
the clock. Use CLOCK_IDFIELD_SIZE to guarantee that. */
pd->tid = ((unsigned int) pd) >> CLOCK_IDFIELD_SIZE;
/* Native Client syscall thread_create does not push return address onto stack
as opposed to the kernel. We emulate this behavior on x86-64 to meet the
ABI requirement ((%rsp + 8) mod 16 == 0). On x86-32 the attribute
'force_align_arg_pointer' does the same for start_thread (). */
#ifdef __x86_64__
STACK_VARIABLES_ARGS -= 8;
#endif
if (__nacl_irt_thread_create (fct, STACK_VARIABLES_ARGS, pd) != 0)
{
pd->tid = 0;
atomic_decrement (&__nptl_nthreads); /* Oops, we lied for a second. */
/* Failed. If the thread is detached, remove the TCB here since
the caller cannot do this. The caller remembered the thread
as detached and cannot reverify that it is not since it must
not access the thread descriptor again. */
if (IS_DETACHED (pd))
__deallocate_stack (pd);
/* We have to translate error codes. */
return errno == ENOMEM ? EAGAIN : errno;
}
/* Now we have the possibility to set scheduling parameters etc. */
if (__builtin_expect (stopped != 0, 0))
{
INTERNAL_SYSCALL_DECL (err);
int res = 0;
/* Set the affinity mask if necessary. */
if (attr->cpuset != NULL)
{
res = INTERNAL_SYSCALL (sched_setaffinity, err, 3, pd->tid,
attr->cpusetsize, attr->cpuset);
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (res, err), 0))
{
/* The operation failed. We have to kill the thread. First
send it the cancellation signal. */
INTERNAL_SYSCALL_DECL (err2);
err_out:
#if __ASSUME_TGKILL
(void) INTERNAL_SYSCALL (tgkill, err2, 3,
THREAD_GETMEM (THREAD_SELF, pid),
pd->tid, SIGCANCEL);
#else
(void) INTERNAL_SYSCALL (tkill, err2, 2, pd->tid, SIGCANCEL);
#endif
return (INTERNAL_SYSCALL_ERROR_P (res, err)
? INTERNAL_SYSCALL_ERRNO (res, err)
: 0);
}
}
/* Set the scheduling parameters. */
if ((attr->flags & ATTR_FLAG_NOTINHERITSCHED) != 0)
{
res = INTERNAL_SYSCALL (sched_setscheduler, err, 3, pd->tid,
pd->schedpolicy, &pd->schedparam);
if (__builtin_expect (INTERNAL_SYSCALL_ERROR_P (res, err), 0))
goto err_out;
}
}
/* We now have for sure more than one thread. The main thread might
not yet have the flag set. No need to set the global variable
again if this is what we use. */
THREAD_SETMEM (THREAD_SELF, header.multiple_threads, 1);
return 0;
}
