-/* Copyright (C) 2002, 2003 Free Software Foundation, Inc.
+/* Copyright (C) 2002-2014 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Ulrich Drepper <drepper@redhat.com>, 2002.
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
- License along with the GNU C Library; if not, write to the Free
- Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
- 02111-1307 USA. */
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
#include <assert.h>
#include <errno.h>
+#include <signal.h>
#include <stdint.h>
#include <string.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/param.h>
#include <dl-sysdep.h>
+#include <dl-tls.h>
#include <tls.h>
+#include <list.h>
+#include <lowlevellock.h>
+#include <kernel-features.h>
-
+#ifndef NEED_SEPARATE_REGISTER_STACK
/* Most architectures have exactly one stack pointer. Some have more. */
-#define STACK_VARIABLES void *stackaddr
+# define STACK_VARIABLES void *stackaddr = NULL
/* How to pass the values to the 'create_thread' function. */
-#define STACK_VARIABLES_ARGS stackaddr
+# define STACK_VARIABLES_ARGS stackaddr
/* How to declare function which gets there parameters. */
-#define STACK_VARIABLES_PARMS void *stackaddr
+# define STACK_VARIABLES_PARMS void *stackaddr
+
+/* How to declare allocate_stack. */
+# define ALLOCATE_STACK_PARMS void **stack
+
+/* This is how the function is called. We do it this way to allow
+ other variants of the function to have more parameters. */
+# define ALLOCATE_STACK(attr, pd) allocate_stack (attr, pd, &stackaddr)
+
+#else
+
+/* We need two stacks. The kernel will place them but we have to tell
+ the kernel about the size of the reserved address space. */
+# define STACK_VARIABLES void *stackaddr = NULL; size_t stacksize = 0
+
+/* How to pass the values to the 'create_thread' function. */
+# define STACK_VARIABLES_ARGS stackaddr, stacksize
+
+/* How to declare function which gets there parameters. */
+# define STACK_VARIABLES_PARMS void *stackaddr, size_t stacksize
+
+/* How to declare allocate_stack. */
+# define ALLOCATE_STACK_PARMS void **stack, size_t *stacksize
+
+/* This is how the function is called. We do it this way to allow
+ other variants of the function to have more parameters. */
+# define ALLOCATE_STACK(attr, pd) \
+ allocate_stack (attr, pd, &stackaddr, &stacksize)
+
+#endif
/* Default alignment of stack. */
#endif
+/* Newer kernels have the MAP_STACK flag to indicate a mapping is used for
+ a stack. Use it when possible. */
+#ifndef MAP_STACK
+# define MAP_STACK 0
+#endif
+/* This yields the pointer that TLS support code calls the thread pointer. */
+#if TLS_TCB_AT_TP
+# define TLS_TPADJ(pd) (pd)
+#elif TLS_DTV_AT_TP
+# define TLS_TPADJ(pd) ((struct pthread *)((char *) (pd) + TLS_PRE_TCB_SIZE))
+#endif
/* Cache handling for not-yet free stacks. */
static size_t stack_cache_actsize;
/* Mutex protecting this variable. */
-static lll_lock_t stack_cache_lock = LLL_LOCK_INITIALIZER;
+static int stack_cache_lock = LLL_LOCK_INITIALIZER;
/* List of queued stack frames. */
static LIST_HEAD (stack_cache);
/* List of the stacks in use. */
static LIST_HEAD (stack_used);
+/* We need to record what list operations we are going to do so that,
+ in case of an asynchronous interruption due to a fork() call, we
+ can correct for the work. */
+static uintptr_t in_flight_stack;
+
/* List of the threads with user provided stacks in use. No need to
initialize this, since it's done in __pthread_initialize_minimal. */
list_t __stack_user __attribute__ ((nocommon));
-hidden_def (__stack_user)
+hidden_data_def (__stack_user)
+#if COLORING_INCREMENT != 0
/* Number of threads created. */
static unsigned int nptl_ncreated;
+#endif
/* Check whether the stack is still used or not. */
-#define FREE_P(descr) ((descr)->tid == 0)
+#define FREE_P(descr) ((descr)->tid <= 0)
+
+
+static void
+stack_list_del (list_t *elem)
+{
+ in_flight_stack = (uintptr_t) elem;
+
+ atomic_write_barrier ();
+
+ list_del (elem);
+
+ atomic_write_barrier ();
+
+ in_flight_stack = 0;
+}
+
+
+static void
+stack_list_add (list_t *elem, list_t *list)
+{
+ in_flight_stack = (uintptr_t) elem | 1;
+
+ atomic_write_barrier ();
+
+ list_add (elem, list);
+
+ atomic_write_barrier ();
+
+ in_flight_stack = 0;
+}
/* We create a double linked list of all cache entries. Double linked
struct pthread *result = NULL;
list_t *entry;
- lll_lock (stack_cache_lock);
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
/* Search the cache for a matching entry. We search for the
smallest stack which has at least the required size. Note that
{
struct pthread *curr;
- curr = list_entry (entry, struct pthread, header.data.list);
+ curr = list_entry (entry, struct pthread, list);
if (FREE_P (curr) && curr->stackblock_size >= size)
{
if (curr->stackblock_size == size)
|| __builtin_expect (result->stackblock_size > 4 * size, 0))
{
/* Release the lock. */
- lll_unlock (stack_cache_lock);
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
return NULL;
}
+ /* Don't allow setxid until cloned. */
+ result->setxid_futex = -1;
+
/* Dequeue the entry. */
- list_del (&result->header.data.list);
+ stack_list_del (&result->list);
/* And add to the list of stacks in use. */
- list_add (&result->header.data.list, &stack_used);
+ stack_list_add (&result->list, &stack_used);
/* And decrease the cache size. */
stack_cache_actsize -= result->stackblock_size;
/* Release the lock early. */
- lll_unlock (stack_cache_lock);
-
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
+ /* Report size and location of the stack to the caller. */
*sizep = result->stackblock_size;
*memp = result->stackblock;
result->nextevent = NULL;
/* Clear the DTV. */
- dtv_t *dtv = GET_DTV (result);
- memset (dtv, '\0', (dtv[-1].counter + 1) * sizeof (dtv_t));
+ dtv_t *dtv = GET_DTV (TLS_TPADJ (result));
+ _dl_clear_dtv (dtv);
/* Re-initialize the TLS. */
- return _dl_allocate_tls_init (result);
+ _dl_allocate_tls_init (TLS_TPADJ (result));
+
+ return result;
+}
+
+
+/* Free stacks until cache size is lower than LIMIT. */
+void
+__free_stacks (size_t limit)
+{
+ /* We reduce the size of the cache. Remove the last entries until
+ the size is below the limit. */
+ list_t *entry;
+ list_t *prev;
+
+ /* Search from the end of the list. */
+ list_for_each_prev_safe (entry, prev, &stack_cache)
+ {
+ struct pthread *curr;
+
+ curr = list_entry (entry, struct pthread, list);
+ if (FREE_P (curr))
+ {
+ /* Unlink the block. */
+ stack_list_del (entry);
+
+ /* Account for the freed memory. */
+ stack_cache_actsize -= curr->stackblock_size;
+
+ /* Free the memory associated with the ELF TLS. */
+ _dl_deallocate_tls (TLS_TPADJ (curr), false);
+
+ /* Remove this block. This should never fail. If it does
+ something is really wrong. */
+ if (munmap (curr->stackblock, curr->stackblock_size) != 0)
+ abort ();
+
+ /* Maybe we have freed enough. */
+ if (stack_cache_actsize <= limit)
+ break;
+ }
+ }
}
/* Add a stack frame which is not used anymore to the stack. Must be
called with the cache lock held. */
static inline void
+__attribute ((always_inline))
queue_stack (struct pthread *stack)
{
/* We unconditionally add the stack to the list. The memory may
still be in use but it will not be reused until the kernel marks
the stack as not used anymore. */
- list_add (&stack->header.data.list, &stack_cache);
+ stack_list_add (&stack->list, &stack_cache);
stack_cache_actsize += stack->stackblock_size;
if (__builtin_expect (stack_cache_actsize > stack_cache_maxsize, 0))
- {
- /* We reduce the size of the cache. Remove the last entries
- until the size is below the limit. */
- list_t *entry;
- list_t *prev;
-
- /* Search from the end of the list. */
- list_for_each_prev_safe (entry, prev, &stack_cache)
- {
- struct pthread *curr;
-
- curr = list_entry (entry, struct pthread, header.data.list);
- if (FREE_P (curr))
- {
- /* Unlink the block. */
- list_del (entry);
-
- /* Account for the freed memory. */
- stack_cache_actsize -= curr->stackblock_size;
+ __free_stacks (stack_cache_maxsize);
+}
- /* Free the memory associated with the ELF TLS. */
- _dl_deallocate_tls (curr, false);
- /* Remove this block. This should never fail. If it
- does something is really wrong. */
- if (munmap (curr->stackblock, curr->stackblock_size) != 0)
- abort ();
+static int
+internal_function
+change_stack_perm (struct pthread *pd
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ , size_t pagemask
+#endif
+ )
+{
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ void *stack = (pd->stackblock
+ + (((((pd->stackblock_size - pd->guardsize) / 2)
+ & pagemask) + pd->guardsize) & pagemask));
+ size_t len = pd->stackblock + pd->stackblock_size - stack;
+#elif _STACK_GROWS_DOWN
+ void *stack = pd->stackblock + pd->guardsize;
+ size_t len = pd->stackblock_size - pd->guardsize;
+#elif _STACK_GROWS_UP
+ void *stack = pd->stackblock;
+ size_t len = (uintptr_t) pd - pd->guardsize - (uintptr_t) pd->stackblock;
+#else
+# error "Define either _STACK_GROWS_DOWN or _STACK_GROWS_UP"
+#endif
+ if (mprotect (stack, len, PROT_READ | PROT_WRITE | PROT_EXEC) != 0)
+ return errno;
- /* Maybe we have freed enough. */
- if (stack_cache_actsize <= stack_cache_maxsize)
- break;
- }
- }
- }
+ return 0;
}
-
+/* Returns a usable stack for a new thread either by allocating a
+ new stack or reusing a cached stack of sufficient size.
+ ATTR must be non-NULL and point to a valid pthread_attr.
+ PDP must be non-NULL. */
static int
allocate_stack (const struct pthread_attr *attr, struct pthread **pdp,
- void **stack)
+ ALLOCATE_STACK_PARMS)
{
struct pthread *pd;
size_t size;
size_t pagesize_m1 = __getpagesize () - 1;
+ void *stacktop;
- assert (attr != NULL);
assert (powerof2 (pagesize_m1 + 1));
assert (TCB_ALIGNMENT >= STACK_ALIGN);
/* Get the stack size from the attribute if it is set. Otherwise we
use the default we determined at start time. */
- size = attr->stacksize ?: __default_stacksize;
+ if (attr->stacksize != 0)
+ size = attr->stacksize;
+ else
+ {
+ lll_lock (__default_pthread_attr_lock, LLL_PRIVATE);
+ size = __default_pthread_attr.stacksize;
+ lll_unlock (__default_pthread_attr_lock, LLL_PRIVATE);
+ }
/* Get memory for the stack. */
if (__builtin_expect (attr->flags & ATTR_FLAG_STACKADDR, 0))
return EINVAL;
/* Adjust stack size for alignment of the TLS block. */
- adj = ((uintptr_t) attr->stackaddr) & (__static_tls_align - 1);
+#if TLS_TCB_AT_TP
+ adj = ((uintptr_t) attr->stackaddr - TLS_TCB_SIZE)
+ & __static_tls_align_m1;
+ assert (size > adj + TLS_TCB_SIZE);
+#elif TLS_DTV_AT_TP
+ adj = ((uintptr_t) attr->stackaddr - __static_tls_size)
+ & __static_tls_align_m1;
assert (size > adj);
+#endif
/* The user provided some memory. Let's hope it matches the
size... We do not allocate guard pages if the user provided
the stack. It is the user's responsibility to do this if it
is wanted. */
- pd = (struct pthread *) (((uintptr_t) attr->stackaddr - adj)
- & ~(__alignof (struct pthread) - 1)) - 1;
+#if TLS_TCB_AT_TP
+ pd = (struct pthread *) ((uintptr_t) attr->stackaddr
+ - TLS_TCB_SIZE - adj);
+#elif TLS_DTV_AT_TP
+ pd = (struct pthread *) (((uintptr_t) attr->stackaddr
+ - __static_tls_size - adj)
+ - TLS_PRE_TCB_SIZE);
+#endif
/* The user provided stack memory needs to be cleared. */
memset (pd, '\0', sizeof (struct pthread));
/* The first TSD block is included in the TCB. */
pd->specific[0] = pd->specific_1stblock;
-#if LLL_LOCK_INITIALIZER != 0
- /* Initialize the lock. */
- pd->lock = LLL_LOCK_INITIALIZER;
-#endif
-
/* Remember the stack-related values. */
pd->stackblock = (char *) attr->stackaddr - size;
- pd->stackblock_size = size - adj;
+ pd->stackblock_size = size;
/* This is a user-provided stack. It will not be queued in the
stack cache nor will the memory (except the TLS memory) be freed. */
pd->user_stack = true;
/* This is at least the second thread. */
- pd->header.data.multiple_threads = 1;
+ pd->header.multiple_threads = 1;
+#ifndef TLS_MULTIPLE_THREADS_IN_TCB
+ __pthread_multiple_threads = *__libc_multiple_threads_ptr = 1;
+#endif
+
+#ifndef __ASSUME_PRIVATE_FUTEX
+ /* The thread must know when private futexes are supported. */
+ pd->header.private_futex = THREAD_GETMEM (THREAD_SELF,
+ header.private_futex);
+#endif
#ifdef NEED_DL_SYSINFO
/* Copy the sysinfo value from the parent. */
- pd->header.data.sysinfo
- = THREAD_GETMEM (THREAD_SELF, header.data.sysinfo);
+ THREAD_SYSINFO(pd) = THREAD_SELF_SYSINFO;
#endif
+ /* The process ID is also the same as that of the caller. */
+ pd->pid = THREAD_GETMEM (THREAD_SELF, pid);
+
+ /* Don't allow setxid until cloned. */
+ pd->setxid_futex = -1;
+
/* Allocate the DTV for this thread. */
- if (_dl_allocate_tls (pd) == NULL)
- /* Something went wrong. */
- return errno;
+ if (_dl_allocate_tls (TLS_TPADJ (pd)) == NULL)
+ {
+ /* Something went wrong. */
+ assert (errno == ENOMEM);
+ return errno;
+ }
/* Prepare to modify global data. */
- lll_lock (stack_cache_lock);
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
/* And add to the list of stacks in use. */
- list_add (&pd->header.data.list, &__stack_user);
+ list_add (&pd->list, &__stack_user);
- lll_unlock (stack_cache_lock);
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
}
else
{
- /* Allocate some anonymous memory. If possible use the
- cache. */
+ /* Allocate some anonymous memory. If possible use the cache. */
size_t guardsize;
size_t reqsize;
void *mem;
+ const int prot = (PROT_READ | PROT_WRITE
+ | ((GL(dl_stack_flags) & PF_X) ? PROT_EXEC : 0));
#if COLORING_INCREMENT != 0
- /* Add one more page for stack coloring. Don't to it for stacks
+ /* Add one more page for stack coloring. Don't do it for stacks
with 16 times pagesize or larger. This might just cause
unnecessary misalignment. */
if (size <= 16 * pagesize_m1)
#endif
/* Adjust the stack size for alignment. */
- size &= ~(__static_tls_align - 1);
+ size &= ~__static_tls_align_m1;
assert (size != 0);
/* Make sure the size of the stack is enough for the guard and
eventually the thread descriptor. */
guardsize = (attr->guardsize + pagesize_m1) & ~pagesize_m1;
- if (__builtin_expect (size < (guardsize + __static_tls_size
- + MINIMAL_REST_STACK + pagesize_m1 + 1),
+ if (__builtin_expect (size < ((guardsize + __static_tls_size
+ + MINIMAL_REST_STACK + pagesize_m1)
+ & ~pagesize_m1),
0))
/* The stack is too small (or the guard too large). */
return EINVAL;
pd = get_cached_stack (&size, &mem);
if (pd == NULL)
{
- mem = mmap (NULL, size, PROT_READ | PROT_WRITE | PROT_EXEC,
- MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ /* To avoid aliasing effects on a larger scale than pages we
+ adjust the allocated stack size if necessary. This way
+ allocations directly following each other will not have
+ aliasing problems. */
+#if MULTI_PAGE_ALIASING != 0
+ if ((size % MULTI_PAGE_ALIASING) == 0)
+ size += pagesize_m1 + 1;
+#endif
+
+ mem = mmap (NULL, size, prot,
+ MAP_PRIVATE | MAP_ANONYMOUS | MAP_STACK, -1, 0);
if (__builtin_expect (mem == MAP_FAILED, 0))
return errno;
- /* 'size' is guaranteed to be greater than zero. So we can
- never get a NULL pointer back from MMAP. */
+ /* SIZE is guaranteed to be greater than zero.
+ So we can never get a null pointer back from mmap. */
assert (mem != NULL);
#if COLORING_INCREMENT != 0
/* Atomically increment NCREATED. */
- unsigned int ncreated = atomic_exchange_and_add (&nptl_ncreated, 1);
+ unsigned int ncreated = atomic_increment_val (&nptl_ncreated);
/* We chose the offset for coloring by incrementing it for
every new thread by a fixed amount. The offset used
/* Make sure the coloring offsets does not disturb the alignment
of the TCB and static TLS block. */
- if (__builtin_expect ((coloring & (__static_tls_align - 1)) != 0, 0))
- coloring = (((coloring + __static_tls_align - 1)
- & ~(__static_tls_align - 1))
+ if (__builtin_expect ((coloring & __static_tls_align_m1) != 0, 0))
+ coloring = (((coloring + __static_tls_align_m1)
+ & ~(__static_tls_align_m1))
& ~pagesize_m1);
#else
/* Unless specified we do not make any adjustments. */
#endif
/* Place the thread descriptor at the end of the stack. */
+#if TLS_TCB_AT_TP
pd = (struct pthread *) ((char *) mem + size - coloring) - 1;
+#elif TLS_DTV_AT_TP
+ pd = (struct pthread *) ((((uintptr_t) mem + size - coloring
+ - __static_tls_size)
+ & ~__static_tls_align_m1)
+ - TLS_PRE_TCB_SIZE);
+#endif
/* Remember the stack-related values. */
pd->stackblock = mem;
descriptor. */
pd->specific[0] = pd->specific_1stblock;
-#if LLL_LOCK_INITIALIZER != 0
- /* Initialize the lock. */
- pd->lock = LLL_LOCK_INITIALIZER;
+ /* This is at least the second thread. */
+ pd->header.multiple_threads = 1;
+#ifndef TLS_MULTIPLE_THREADS_IN_TCB
+ __pthread_multiple_threads = *__libc_multiple_threads_ptr = 1;
#endif
- /* This is at least the second thread. */
- pd->header.data.multiple_threads = 1;
+#ifndef __ASSUME_PRIVATE_FUTEX
+ /* The thread must know when private futexes are supported. */
+ pd->header.private_futex = THREAD_GETMEM (THREAD_SELF,
+ header.private_futex);
+#endif
#ifdef NEED_DL_SYSINFO
/* Copy the sysinfo value from the parent. */
- pd->header.data.sysinfo
- = THREAD_GETMEM (THREAD_SELF, header.data.sysinfo);
+ THREAD_SYSINFO(pd) = THREAD_SELF_SYSINFO;
#endif
+ /* Don't allow setxid until cloned. */
+ pd->setxid_futex = -1;
+
+ /* The process ID is also the same as that of the caller. */
+ pd->pid = THREAD_GETMEM (THREAD_SELF, pid);
+
/* Allocate the DTV for this thread. */
- if (_dl_allocate_tls (pd) == NULL)
+ if (_dl_allocate_tls (TLS_TPADJ (pd)) == NULL)
{
/* Something went wrong. */
- int err = errno;
+ assert (errno == ENOMEM);
/* Free the stack memory we just allocated. */
(void) munmap (mem, size);
- return err;
+ return errno;
}
/* Prepare to modify global data. */
- lll_lock (stack_cache_lock);
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
/* And add to the list of stacks in use. */
- list_add (&pd->header.data.list, &stack_used);
+ stack_list_add (&pd->list, &stack_used);
+
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
- lll_unlock (stack_cache_lock);
+
+ /* There might have been a race. Another thread might have
+ caused the stacks to get exec permission while this new
+ stack was prepared. Detect if this was possible and
+ change the permission if necessary. */
+ if (__builtin_expect ((GL(dl_stack_flags) & PF_X) != 0
+ && (prot & PROT_EXEC) == 0, 0))
+ {
+ int err = change_stack_perm (pd
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ , ~pagesize_m1
+#endif
+ );
+ if (err != 0)
+ {
+ /* Free the stack memory we just allocated. */
+ (void) munmap (mem, size);
+
+ return err;
+ }
+ }
/* Note that all of the stack and the thread descriptor is
/* Create or resize the guard area if necessary. */
if (__builtin_expect (guardsize > pd->guardsize, 0))
{
- if (mprotect (mem, guardsize, PROT_NONE) != 0)
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ char *guard = mem + (((size - guardsize) / 2) & ~pagesize_m1);
+#elif _STACK_GROWS_DOWN
+ char *guard = mem;
+# elif _STACK_GROWS_UP
+ char *guard = (char *) (((uintptr_t) pd - guardsize) & ~pagesize_m1);
+#endif
+ if (mprotect (guard, guardsize, PROT_NONE) != 0)
{
- int err;
mprot_error:
- err = errno;
-
- lll_lock (stack_cache_lock);
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
/* Remove the thread from the list. */
- list_del (&pd->header.data.list);
+ stack_list_del (&pd->list);
+
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
- lll_unlock (stack_cache_lock);
+ /* Get rid of the TLS block we allocated. */
+ _dl_deallocate_tls (TLS_TPADJ (pd), false);
- /* Free the memory regardless of whether the size of the
- cache is over the limit or not. If this piece of
- memory caused problems we better do not use it
+ /* Free the stack memory regardless of whether the size
+ of the cache is over the limit or not. If this piece
+ of memory caused problems we better do not use it
anymore. Uh, and we ignore possible errors. There
is nothing we could do. */
(void) munmap (mem, size);
- return err;
+ return errno;
}
pd->guardsize = guardsize;
0))
{
/* The old guard area is too large. */
+
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ char *guard = mem + (((size - guardsize) / 2) & ~pagesize_m1);
+ char *oldguard = mem + (((size - pd->guardsize) / 2) & ~pagesize_m1);
+
+ if (oldguard < guard
+ && mprotect (oldguard, guard - oldguard, prot) != 0)
+ goto mprot_error;
+
+ if (mprotect (guard + guardsize,
+ oldguard + pd->guardsize - guard - guardsize,
+ prot) != 0)
+ goto mprot_error;
+#elif _STACK_GROWS_DOWN
if (mprotect ((char *) mem + guardsize, pd->guardsize - guardsize,
- PROT_READ | PROT_WRITE | PROT_EXEC) != 0)
+ prot) != 0)
goto mprot_error;
+#elif _STACK_GROWS_UP
+ if (mprotect ((char *) pd - pd->guardsize,
+ pd->guardsize - guardsize, prot) != 0)
+ goto mprot_error;
+#endif
pd->guardsize = guardsize;
}
+ /* The pthread_getattr_np() calls need to get passed the size
+ requested in the attribute, regardless of how large the
+ actually used guardsize is. */
+ pd->reported_guardsize = guardsize;
}
+ /* Initialize the lock. We have to do this unconditionally since the
+ stillborn thread could be canceled while the lock is taken. */
+ pd->lock = LLL_LOCK_INITIALIZER;
+
+ /* The robust mutex lists also need to be initialized
+ unconditionally because the cleanup for the previous stack owner
+ might have happened in the kernel. */
+ pd->robust_head.futex_offset = (offsetof (pthread_mutex_t, __data.__lock)
+ - offsetof (pthread_mutex_t,
+ __data.__list.__next));
+ pd->robust_head.list_op_pending = NULL;
+#ifdef __PTHREAD_MUTEX_HAVE_PREV
+ pd->robust_prev = &pd->robust_head;
+#endif
+ pd->robust_head.list = &pd->robust_head;
+
/* We place the thread descriptor at the end of the stack. */
*pdp = pd;
#if TLS_TCB_AT_TP
- /* The stack begin before the TCB and the static TLS block. */
- *stack = ((char *) (pd + 1) - __static_tls_size);
-#else
-# error "Implement me"
+ /* The stack begins before the TCB and the static TLS block. */
+ stacktop = ((char *) (pd + 1) - __static_tls_size);
+#elif TLS_DTV_AT_TP
+ stacktop = (char *) (pd - 1);
+#endif
+
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ *stack = pd->stackblock;
+ *stacksize = stacktop - *stack;
+#elif _STACK_GROWS_DOWN
+ *stack = stacktop;
+#elif _STACK_GROWS_UP
+ *stack = pd->stackblock;
+ assert (*stack > 0);
#endif
return 0;
}
-/* This is how the function is called. We do it this way to allow
- other variants of the function to have more parameters. */
-#define ALLOCATE_STACK(attr, pd) allocate_stack (attr, pd, &stackaddr)
-
void
internal_function
__deallocate_stack (struct pthread *pd)
{
- lll_lock (stack_cache_lock);
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
/* Remove the thread from the list of threads with user defined
stacks. */
- list_del (&pd->header.data.list);
+ stack_list_del (&pd->list);
/* Not much to do. Just free the mmap()ed memory. Note that we do
not reset the 'used' flag in the 'tid' field. This is done by
(void) queue_stack (pd);
else
/* Free the memory associated with the ELF TLS. */
- _dl_deallocate_tls (pd, false);
+ _dl_deallocate_tls (TLS_TPADJ (pd), false);
+
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
+}
+
- lll_unlock (stack_cache_lock);
+int
+internal_function
+__make_stacks_executable (void **stack_endp)
+{
+ /* First the main thread's stack. */
+ int err = _dl_make_stack_executable (stack_endp);
+ if (err != 0)
+ return err;
+
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ const size_t pagemask = ~(__getpagesize () - 1);
+#endif
+
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
+
+ list_t *runp;
+ list_for_each (runp, &stack_used)
+ {
+ err = change_stack_perm (list_entry (runp, struct pthread, list)
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ , pagemask
+#endif
+ );
+ if (err != 0)
+ break;
+ }
+
+ /* Also change the permission for the currently unused stacks. This
+ might be wasted time but better spend it here than adding a check
+ in the fast path. */
+ if (err == 0)
+ list_for_each (runp, &stack_cache)
+ {
+ err = change_stack_perm (list_entry (runp, struct pthread, list)
+#ifdef NEED_SEPARATE_REGISTER_STACK
+ , pagemask
+#endif
+ );
+ if (err != 0)
+ break;
+ }
+
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
+
+ return err;
}
{
struct pthread *self = (struct pthread *) THREAD_SELF;
- /* No locking necessary. The caller is the only stack in use. */
+ /* No locking necessary. The caller is the only stack in use. But
+ we have to be aware that we might have interrupted a list
+ operation. */
+
+ if (in_flight_stack != 0)
+ {
+ bool add_p = in_flight_stack & 1;
+ list_t *elem = (list_t *) (in_flight_stack & ~(uintptr_t) 1);
+
+ if (add_p)
+ {
+ /* We always add at the beginning of the list. So in this
+ case we only need to check the beginning of these lists. */
+ int check_list (list_t *l)
+ {
+ if (l->next->prev != l)
+ {
+ assert (l->next->prev == elem);
+
+ elem->next = l->next;
+ elem->prev = l;
+ l->next = elem;
+
+ return 1;
+ }
+
+ return 0;
+ }
+
+ if (check_list (&stack_used) == 0)
+ (void) check_list (&stack_cache);
+ }
+ else
+ {
+ /* We can simply always replay the delete operation. */
+ elem->next->prev = elem->prev;
+ elem->prev->next = elem->next;
+ }
+ }
/* Mark all stacks except the still running one as free. */
list_t *runp;
list_for_each (runp, &stack_used)
{
- struct pthread *curp;
-
- curp = list_entry (runp, struct pthread, header.data.list);
+ struct pthread *curp = list_entry (runp, struct pthread, list);
if (curp != self)
{
/* This marks the stack as free. */
curp->tid = 0;
+ /* The PID field must be initialized for the new process. */
+ curp->pid = self->pid;
+
/* Account for the size of the stack. */
stack_cache_actsize += curp->stackblock_size;
+
+ if (curp->specific_used)
+ {
+ /* Clear the thread-specific data. */
+ memset (curp->specific_1stblock, '\0',
+ sizeof (curp->specific_1stblock));
+
+ curp->specific_used = false;
+
+ for (size_t cnt = 1; cnt < PTHREAD_KEY_1STLEVEL_SIZE; ++cnt)
+ if (curp->specific[cnt] != NULL)
+ {
+ memset (curp->specific[cnt], '\0',
+ sizeof (curp->specific_1stblock));
+
+ /* We have allocated the block which we do not
+ free here so re-set the bit. */
+ curp->specific_used = true;
+ }
+ }
}
}
+ /* Reset the PIDs in any cached stacks. */
+ list_for_each (runp, &stack_cache)
+ {
+ struct pthread *curp = list_entry (runp, struct pthread, list);
+ curp->pid = self->pid;
+ }
+
/* Add the stack of all running threads to the cache. */
list_splice (&stack_used, &stack_cache);
/* Remove the entry for the current thread to from the cache list
and add it to the list of running threads. Which of the two
lists is decided by the user_stack flag. */
- list_del (&self->header.data.list);
+ stack_list_del (&self->list);
/* Re-initialize the lists for all the threads. */
INIT_LIST_HEAD (&stack_used);
INIT_LIST_HEAD (&__stack_user);
if (__builtin_expect (THREAD_GETMEM (self, user_stack), 0))
- list_add (&self->header.data.list, &__stack_user);
+ list_add (&self->list, &__stack_user);
else
- list_add (&self->header.data.list, &stack_used);
+ list_add (&self->list, &stack_used);
/* There is one thread running. */
__nptl_nthreads = 1;
- /* Initialize the lock. */
+ in_flight_stack = 0;
+
+ /* Initialize locks. */
stack_cache_lock = LLL_LOCK_INITIALIZER;
+ __default_pthread_attr_lock = LLL_LOCK_INITIALIZER;
+}
+
+
+#if HP_TIMING_AVAIL
+# undef __find_thread_by_id
+/* Find a thread given the thread ID. */
+attribute_hidden
+struct pthread *
+__find_thread_by_id (pid_t tid)
+{
+ struct pthread *result = NULL;
+
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
+
+ /* Iterate over the list with system-allocated threads first. */
+ list_t *runp;
+ list_for_each (runp, &stack_used)
+ {
+ struct pthread *curp;
+
+ curp = list_entry (runp, struct pthread, list);
+
+ if (curp->tid == tid)
+ {
+ result = curp;
+ goto out;
+ }
+ }
+
+ /* Now the list with threads using user-allocated stacks. */
+ list_for_each (runp, &__stack_user)
+ {
+ struct pthread *curp;
+
+ curp = list_entry (runp, struct pthread, list);
+
+ if (curp->tid == tid)
+ {
+ result = curp;
+ goto out;
+ }
+ }
+
+ out:
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
+
+ return result;
+}
+#endif
+
+
+static void
+internal_function
+setxid_mark_thread (struct xid_command *cmdp, struct pthread *t)
+{
+ int ch;
+
+ /* Wait until this thread is cloned. */
+ if (t->setxid_futex == -1
+ && ! atomic_compare_and_exchange_bool_acq (&t->setxid_futex, -2, -1))
+ do
+ lll_futex_wait (&t->setxid_futex, -2, LLL_PRIVATE);
+ while (t->setxid_futex == -2);
+
+ /* Don't let the thread exit before the setxid handler runs. */
+ t->setxid_futex = 0;
+
+ do
+ {
+ ch = t->cancelhandling;
+
+ /* If the thread is exiting right now, ignore it. */
+ if ((ch & EXITING_BITMASK) != 0)
+ {
+ /* Release the futex if there is no other setxid in
+ progress. */
+ if ((ch & SETXID_BITMASK) == 0)
+ {
+ t->setxid_futex = 1;
+ lll_futex_wake (&t->setxid_futex, 1, LLL_PRIVATE);
+ }
+ return;
+ }
+ }
+ while (atomic_compare_and_exchange_bool_acq (&t->cancelhandling,
+ ch | SETXID_BITMASK, ch));
+}
+
+
+static void
+internal_function
+setxid_unmark_thread (struct xid_command *cmdp, struct pthread *t)
+{
+ int ch;
+
+ do
+ {
+ ch = t->cancelhandling;
+ if ((ch & SETXID_BITMASK) == 0)
+ return;
+ }
+ while (atomic_compare_and_exchange_bool_acq (&t->cancelhandling,
+ ch & ~SETXID_BITMASK, ch));
+
+ /* Release the futex just in case. */
+ t->setxid_futex = 1;
+ lll_futex_wake (&t->setxid_futex, 1, LLL_PRIVATE);
+}
+
+
+static int
+internal_function
+setxid_signal_thread (struct xid_command *cmdp, struct pthread *t)
+{
+ if ((t->cancelhandling & SETXID_BITMASK) == 0)
+ return 0;
+
+ int val;
+ INTERNAL_SYSCALL_DECL (err);
+ val = INTERNAL_SYSCALL (tgkill, err, 3, THREAD_GETMEM (THREAD_SELF, pid),
+ t->tid, SIGSETXID);
+
+ /* If this failed, it must have had not started yet or else exited. */
+ if (!INTERNAL_SYSCALL_ERROR_P (val, err))
+ {
+ atomic_increment (&cmdp->cntr);
+ return 1;
+ }
+ else
+ return 0;
+}
+
+
+int
+attribute_hidden
+__nptl_setxid (struct xid_command *cmdp)
+{
+ int signalled;
+ int result;
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
+
+ __xidcmd = cmdp;
+ cmdp->cntr = 0;
+
+ struct pthread *self = THREAD_SELF;
+
+ /* Iterate over the list with system-allocated threads first. */
+ list_t *runp;
+ list_for_each (runp, &stack_used)
+ {
+ struct pthread *t = list_entry (runp, struct pthread, list);
+ if (t == self)
+ continue;
+
+ setxid_mark_thread (cmdp, t);
+ }
+
+ /* Now the list with threads using user-allocated stacks. */
+ list_for_each (runp, &__stack_user)
+ {
+ struct pthread *t = list_entry (runp, struct pthread, list);
+ if (t == self)
+ continue;
+
+ setxid_mark_thread (cmdp, t);
+ }
+
+ /* Iterate until we don't succeed in signalling anyone. That means
+ we have gotten all running threads, and their children will be
+ automatically correct once started. */
+ do
+ {
+ signalled = 0;
+
+ list_for_each (runp, &stack_used)
+ {
+ struct pthread *t = list_entry (runp, struct pthread, list);
+ if (t == self)
+ continue;
+
+ signalled += setxid_signal_thread (cmdp, t);
+ }
+
+ list_for_each (runp, &__stack_user)
+ {
+ struct pthread *t = list_entry (runp, struct pthread, list);
+ if (t == self)
+ continue;
+
+ signalled += setxid_signal_thread (cmdp, t);
+ }
+
+ int cur = cmdp->cntr;
+ while (cur != 0)
+ {
+ lll_futex_wait (&cmdp->cntr, cur, LLL_PRIVATE);
+ cur = cmdp->cntr;
+ }
+ }
+ while (signalled != 0);
+
+ /* Clean up flags, so that no thread blocks during exit waiting
+ for a signal which will never come. */
+ list_for_each (runp, &stack_used)
+ {
+ struct pthread *t = list_entry (runp, struct pthread, list);
+ if (t == self)
+ continue;
+
+ setxid_unmark_thread (cmdp, t);
+ }
+
+ list_for_each (runp, &__stack_user)
+ {
+ struct pthread *t = list_entry (runp, struct pthread, list);
+ if (t == self)
+ continue;
+
+ setxid_unmark_thread (cmdp, t);
+ }
+
+ /* This must be last, otherwise the current thread might not have
+ permissions to send SIGSETXID syscall to the other threads. */
+ INTERNAL_SYSCALL_DECL (err);
+ result = INTERNAL_SYSCALL_NCS (cmdp->syscall_no, err, 3,
+ cmdp->id[0], cmdp->id[1], cmdp->id[2]);
+ if (INTERNAL_SYSCALL_ERROR_P (result, err))
+ {
+ __set_errno (INTERNAL_SYSCALL_ERRNO (result, err));
+ result = -1;
+ }
+
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
+ return result;
+}
+
+static inline void __attribute__((always_inline))
+init_one_static_tls (struct pthread *curp, struct link_map *map)
+{
+ dtv_t *dtv = GET_DTV (TLS_TPADJ (curp));
+# if TLS_TCB_AT_TP
+ void *dest = (char *) curp - map->l_tls_offset;
+# elif TLS_DTV_AT_TP
+ void *dest = (char *) curp + map->l_tls_offset + TLS_PRE_TCB_SIZE;
+# else
+# error "Either TLS_TCB_AT_TP or TLS_DTV_AT_TP must be defined"
+# endif
+
+ /* Fill in the DTV slot so that a later LD/GD access will find it. */
+ dtv[map->l_tls_modid].pointer.val = dest;
+ dtv[map->l_tls_modid].pointer.is_static = true;
+
+ /* Initialize the memory. */
+ memset (__mempcpy (dest, map->l_tls_initimage, map->l_tls_initimage_size),
+ '\0', map->l_tls_blocksize - map->l_tls_initimage_size);
+}
+
+void
+attribute_hidden
+__pthread_init_static_tls (struct link_map *map)
+{
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
+
+ /* Iterate over the list with system-allocated threads first. */
+ list_t *runp;
+ list_for_each (runp, &stack_used)
+ init_one_static_tls (list_entry (runp, struct pthread, list), map);
+
+ /* Now the list with threads using user-allocated stacks. */
+ list_for_each (runp, &__stack_user)
+ init_one_static_tls (list_entry (runp, struct pthread, list), map);
+
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
+}
+
+
+void
+attribute_hidden
+__wait_lookup_done (void)
+{
+ lll_lock (stack_cache_lock, LLL_PRIVATE);
+
+ struct pthread *self = THREAD_SELF;
+
+ /* Iterate over the list with system-allocated threads first. */
+ list_t *runp;
+ list_for_each (runp, &stack_used)
+ {
+ struct pthread *t = list_entry (runp, struct pthread, list);
+ if (t == self || t->header.gscope_flag == THREAD_GSCOPE_FLAG_UNUSED)
+ continue;
+
+ int *const gscope_flagp = &t->header.gscope_flag;
+
+ /* We have to wait until this thread is done with the global
+ scope. First tell the thread that we are waiting and
+ possibly have to be woken. */
+ if (atomic_compare_and_exchange_bool_acq (gscope_flagp,
+ THREAD_GSCOPE_FLAG_WAIT,
+ THREAD_GSCOPE_FLAG_USED))
+ continue;
+
+ do
+ lll_futex_wait (gscope_flagp, THREAD_GSCOPE_FLAG_WAIT, LLL_PRIVATE);
+ while (*gscope_flagp == THREAD_GSCOPE_FLAG_WAIT);
+ }
+
+ /* Now the list with threads using user-allocated stacks. */
+ list_for_each (runp, &__stack_user)
+ {
+ struct pthread *t = list_entry (runp, struct pthread, list);
+ if (t == self || t->header.gscope_flag == THREAD_GSCOPE_FLAG_UNUSED)
+ continue;
+
+ int *const gscope_flagp = &t->header.gscope_flag;
+
+ /* We have to wait until this thread is done with the global
+ scope. First tell the thread that we are waiting and
+ possibly have to be woken. */
+ if (atomic_compare_and_exchange_bool_acq (gscope_flagp,
+ THREAD_GSCOPE_FLAG_WAIT,
+ THREAD_GSCOPE_FLAG_USED))
+ continue;
+
+ do
+ lll_futex_wait (gscope_flagp, THREAD_GSCOPE_FLAG_WAIT, LLL_PRIVATE);
+ while (*gscope_flagp == THREAD_GSCOPE_FLAG_WAIT);
+ }
+
+ lll_unlock (stack_cache_lock, LLL_PRIVATE);
}
git.samba.org / jlayton / glibc.git / blobdiff