diff --git a/musl_src.gni b/musl_src.gni
index 4357fdecd52d74b8339e2fbe7c3961c66e8e0fa6..bdeea8630189eda0fd1581afc3b29c99f6dfa7c2 100644
--- a/musl_src.gni
+++ b/musl_src.gni
@@ -1854,6 +1854,7 @@ musl_src_porting_file = [
   "src/exit/atexit.c",
   "crt/arm/crti.s",
   "crt/aarch64/crti.s",
+  "src/thread/pthread_create.c",
 ]
 
 musl_inc_hook_files = [
diff --git a/musl_template.gni b/musl_template.gni
index bbed76b8e6811d564cb94d87d0883fe12dcd02aa..566a08667d860de81318b90e0e0d37b348729637 100644
--- a/musl_template.gni
+++ b/musl_template.gni
@@ -163,6 +163,8 @@ template("musl_libs") {
       cflags_auto += [ "-DHOOK_ENABLE" ]
     }
 
+    cflags_auto += [ "-DRESERVE_SIGNAL_STACK" ]
+
     cflags_c99fse = [
       "-std=c99",
       "-nostdinc",
diff --git a/porting/linux/user/src/internal/pthread_impl.h b/porting/linux/user/src/internal/pthread_impl.h
index 816cfa7dbe99e847e552c7fce604c677fda1b9ca..3c280d502498515a505beb5047ce9ae268a377a2 100644
--- a/porting/linux/user/src/internal/pthread_impl.h
+++ b/porting/linux/user/src/internal/pthread_impl.h
@@ -48,6 +48,9 @@ struct pthread {
 	volatile int killlock[1];
 	char *dlerror_buf;
 	void *stdio_locks;
+#ifdef RESERVE_SIGNAL_STACK
+	void *signal_stack;
+#endif
 
 	/* Part 3 -- the positions of these fields relative to
 	 * the end of the structure is external and internal ABI. */
diff --git a/porting/linux/user/src/thread/pthread_create.c b/porting/linux/user/src/thread/pthread_create.c
new file mode 100644
index 0000000000000000000000000000000000000000..998d1993202ce090d9584801b8b1f594b7b91ee4
--- /dev/null
+++ b/porting/linux/user/src/thread/pthread_create.c
@@ -0,0 +1,476 @@
+#define _GNU_SOURCE
+#include "pthread_impl.h"
+#include "stdio_impl.h"
+#include "libc.h"
+#include "lock.h"
+#include <sys/mman.h>
+#include <string.h>
+#include <stddef.h>
+#include <stdarg.h>
+
+void log_print(const char* info,...)
+{
+    va_list ap;
+    va_start(ap, info);
+    vfprintf(stdout,info, ap);
+    va_end(ap);
+}
+
+#ifdef RESERVE_SIGNAL_STACK
+#include <sys/prctl.h>
+#if defined (__LF64__)
+#define RESERVE_SIGNAL_STACK_SIZE (32 * 1024)
+#else
+#define RESERVE_SIGNAL_STACK_SIZE (20 * 1024)
+#endif
+static void __pthread_reserve_signal_stack()
+{
+	void* stack = mmap(NULL, RESERVE_SIGNAL_STACK_SIZE, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1 , 0);
+	if (stack != MAP_FAILED) {
+		if (mprotect(stack, __default_guardsize, PROT_NONE) == -1) {
+			munmap(stack, RESERVE_SIGNAL_STACK_SIZE);
+			return;
+		}
+	}
+
+	stack_t signal_stack;
+	signal_stack.ss_sp = (uint8_t*)stack + __default_guardsize;
+	signal_stack.ss_size = RESERVE_SIGNAL_STACK_SIZE - __default_guardsize;
+	signal_stack.ss_flags = 0;
+	sigaltstack(&signal_stack, NULL);
+
+	pthread_t self = __pthread_self();
+	self->signal_stack = stack;
+	prctl(PR_SET_VMA, PR_SET_VMA_ANON_NAME, signal_stack.ss_sp, signal_stack.ss_size, "signal_stack:musl");
+	return;
+}
+
+static void __pthread_release_signal_stack()
+{
+	pthread_t self = __pthread_self();
+	if (self->signal_stack == NULL) {
+		return;
+	}
+
+	stack_t signal_stack;
+	memset(&signal_stack, 0, sizeof(signal_stack));
+	signal_stack.ss_flags = SS_DISABLE;
+	sigaltstack(&signal_stack, NULL);
+	munmap(self->signal_stack, RESERVE_SIGNAL_STACK_SIZE);
+	self->signal_stack = NULL;
+}
+#endif
+
+static void dummy_0()
+{
+}
+weak_alias(dummy_0, __acquire_ptc);
+weak_alias(dummy_0, __release_ptc);
+weak_alias(dummy_0, __pthread_tsd_run_dtors);
+weak_alias(dummy_0, __do_orphaned_stdio_locks);
+weak_alias(dummy_0, __dl_thread_cleanup);
+weak_alias(dummy_0, __membarrier_init);
+
+static int tl_lock_count;
+static int tl_lock_waiters;
+
+void __tl_lock(void)
+{
+	int tid = __pthread_self()->tid;
+	int val = __thread_list_lock;
+	if (val == tid) {
+		tl_lock_count++;
+		return;
+	}
+	while ((val = a_cas(&__thread_list_lock, 0, tid)))
+		__wait(&__thread_list_lock, &tl_lock_waiters, val, 0);
+}
+
+void __tl_unlock(void)
+{
+	if (tl_lock_count) {
+		tl_lock_count--;
+		return;
+	}
+	a_store(&__thread_list_lock, 0);
+	if (tl_lock_waiters) __wake(&__thread_list_lock, 1, 0);
+}
+
+void __tl_sync(pthread_t td)
+{
+	a_barrier();
+	int val = __thread_list_lock;
+	if (!val) return;
+	__wait(&__thread_list_lock, &tl_lock_waiters, val, 0);
+	if (tl_lock_waiters) __wake(&__thread_list_lock, 1, 0);
+}
+
+_Noreturn void __pthread_exit(void *result)
+{
+	pthread_t self = __pthread_self();
+	sigset_t set;
+
+	self->canceldisable = 1;
+	self->cancelasync = 0;
+	self->result = result;
+
+	while (self->cancelbuf) {
+		void (*f)(void *) = self->cancelbuf->__f;
+		void *x = self->cancelbuf->__x;
+		self->cancelbuf = self->cancelbuf->__next;
+		f(x);
+	}
+
+	__pthread_tsd_run_dtors();
+
+	/* Access to target the exiting thread with syscalls that use
+	 * its kernel tid is controlled by killlock. For detached threads,
+	 * any use past this point would have undefined behavior, but for
+	 * joinable threads it's a valid usage that must be handled. */
+	LOCK(self->killlock);
+
+	/* The thread list lock must be AS-safe, and thus requires
+	 * application signals to be blocked before it can be taken. */
+	__block_app_sigs(&set);
+	__tl_lock();
+
+#ifdef RESERVE_SIGNAL_STACK
+	__pthread_release_signal_stack();
+#endif
+	/* If this is the only thread in the list, don't proceed with
+	 * termination of the thread, but restore the previous lock and
+	 * signal state to prepare for exit to call atexit handlers. */
+	if (self->next == self) {
+		__tl_unlock();
+		__restore_sigs(&set);
+		UNLOCK(self->killlock);
+		exit(0);
+	}
+
+	/* At this point we are committed to thread termination. Unlink
+	 * the thread from the list. This change will not be visible
+	 * until the lock is released, which only happens after SYS_exit
+	 * has been called, via the exit futex address pointing at the lock. */
+	libc.threads_minus_1--;
+	self->next->prev = self->prev;
+	self->prev->next = self->next;
+	self->prev = self->next = self;
+
+	/* Process robust list in userspace to handle non-pshared mutexes
+	 * and the detached thread case where the robust list head will
+	 * be invalid when the kernel would process it. */
+	__vm_lock();
+	volatile void *volatile *rp;
+	while ((rp=self->robust_list.head) && rp != &self->robust_list.head) {
+		pthread_mutex_t *m = (void *)((char *)rp
+			- offsetof(pthread_mutex_t, _m_next));
+		int waiters = m->_m_waiters;
+		int priv = (m->_m_type & 128) ^ 128;
+		self->robust_list.pending = rp;
+		self->robust_list.head = *rp;
+		int cont = a_swap(&m->_m_lock, 0x40000000);
+		self->robust_list.pending = 0;
+		if (cont < 0 || waiters)
+			__wake(&m->_m_lock, 1, priv);
+	}
+	__vm_unlock();
+
+	__do_orphaned_stdio_locks();
+	__dl_thread_cleanup();
+
+	/* This atomic potentially competes with a concurrent pthread_detach
+	 * call; the loser is responsible for freeing thread resources. */
+	int state = a_cas(&self->detach_state, DT_JOINABLE, DT_EXITING);
+
+	if (state==DT_DETACHED && self->map_base) {
+		/* Detached threads must block even implementation-internal
+		 * signals, since they will not have a stack in their last
+		 * moments of existence. */
+		__block_all_sigs(&set);
+
+		/* Robust list will no longer be valid, and was already
+		 * processed above, so unregister it with the kernel. */
+		if (self->robust_list.off)
+			__syscall(SYS_set_robust_list, 0, 3*sizeof(long));
+
+		/* Since __unmapself bypasses the normal munmap code path,
+		 * explicitly wait for vmlock holders first. */
+		__vm_wait();
+
+		/* The following call unmaps the thread's stack mapping
+		 * and then exits without touching the stack. */
+		__unmapself(self->map_base, self->map_size);
+	}
+
+	/* Wake any joiner. */
+	__wake(&self->detach_state, 1, 1);
+
+	/* After the kernel thread exits, its tid may be reused. Clear it
+	 * to prevent inadvertent use and inform functions that would use
+	 * it that it's no longer available. */
+	self->tid = 0;
+	UNLOCK(self->killlock);
+
+	for (;;) __syscall(SYS_exit, 0);
+}
+
+void __do_cleanup_push(struct __ptcb *cb)
+{
+	struct pthread *self = __pthread_self();
+	cb->__next = self->cancelbuf;
+	self->cancelbuf = cb;
+}
+
+void __do_cleanup_pop(struct __ptcb *cb)
+{
+	__pthread_self()->cancelbuf = cb->__next;
+}
+
+struct start_args {
+	void *(*start_func)(void *);
+	void *start_arg;
+	volatile int control;
+	unsigned long sig_mask[_NSIG/8/sizeof(long)];
+};
+
+static int start(void *p)
+{
+	struct start_args *args = p;
+	int state = args->control;
+	if (state) {
+		if (a_cas(&args->control, 1, 2)==1)
+			__wait(&args->control, 0, 2, 1);
+		if (args->control) {
+			__syscall(SYS_set_tid_address, &args->control);
+			for (;;) __syscall(SYS_exit, 0);
+		}
+	}
+	__syscall(SYS_rt_sigprocmask, SIG_SETMASK, &args->sig_mask, 0, _NSIG/8);
+#ifdef RESERVE_SIGNAL_STACK
+	__pthread_reserve_signal_stack();
+#endif
+	__pthread_exit(args->start_func(args->start_arg));
+	return 0;
+}
+
+static int start_c11(void *p)
+{
+#ifdef RESERVE_SIGNAL_STACK
+	__pthread_reserve_signal_stack();
+#endif
+	struct start_args *args = p;
+	int (*start)(void*) = (int(*)(void*)) args->start_func;
+	__pthread_exit((void *)(uintptr_t)start(args->start_arg));
+	return 0;
+}
+
+#define ROUND(x) (((x)+PAGE_SIZE-1)&-PAGE_SIZE)
+
+/* pthread_key_create.c overrides this */
+static volatile size_t dummy = 0;
+weak_alias(dummy, __pthread_tsd_size);
+static void *dummy_tsd[1] = { 0 };
+weak_alias(dummy_tsd, __pthread_tsd_main);
+
+static FILE *volatile dummy_file = 0;
+weak_alias(dummy_file, __stdin_used);
+weak_alias(dummy_file, __stdout_used);
+weak_alias(dummy_file, __stderr_used);
+
+static void init_file_lock(FILE *f)
+{
+	if (f && f->lock<0) f->lock = 0;
+}
+
+int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict attrp, void *(*entry)(void *), void *restrict arg)
+{
+	int ret, c11 = (attrp == __ATTRP_C11_THREAD);
+	size_t size, guard;
+	struct pthread *self, *new;
+	unsigned char *map = 0, *stack = 0, *tsd = 0, *stack_limit;
+	unsigned flags = CLONE_VM | CLONE_FS | CLONE_FILES | CLONE_SIGHAND
+		| CLONE_THREAD | CLONE_SYSVSEM | CLONE_SETTLS
+		| CLONE_PARENT_SETTID | CLONE_CHILD_CLEARTID | CLONE_DETACHED;
+	pthread_attr_t attr = { 0 };
+	sigset_t set;
+
+	if (!libc.can_do_threads) return ENOSYS;
+	self = __pthread_self();
+	if (!libc.threaded) {
+		for (FILE *f=*__ofl_lock(); f; f=f->next)
+			init_file_lock(f);
+		__ofl_unlock();
+		init_file_lock(__stdin_used);
+		init_file_lock(__stdout_used);
+		init_file_lock(__stderr_used);
+		__syscall(SYS_rt_sigprocmask, SIG_UNBLOCK, SIGPT_SET, 0, _NSIG/8);
+		self->tsd = (void **)__pthread_tsd_main;
+		__membarrier_init();
+		libc.threaded = 1;
+	}
+	if (attrp && !c11) attr = *attrp;
+
+	__acquire_ptc();
+	if (!attrp || c11) {
+		attr._a_stacksize = __default_stacksize;
+		attr._a_guardsize = __default_guardsize;
+	}
+
+	if (attr._a_stackaddr) {
+		size_t need = libc.tls_size + __pthread_tsd_size;
+		size = attr._a_stacksize;
+		stack = (void *)(attr._a_stackaddr & -16);
+		stack_limit = (void *)(attr._a_stackaddr - size);
+		/* Use application-provided stack for TLS only when
+		 * it does not take more than ~12% or 2k of the
+		 * application's stack space. */
+		if (need < size/8 && need < 2048) {
+			tsd = stack - __pthread_tsd_size;
+			stack = tsd - libc.tls_size;
+			memset(stack, 0, need);
+		} else {
+			size = ROUND(need);
+		}
+		guard = 0;
+	} else {
+		guard = ROUND(attr._a_guardsize);
+		size = guard + ROUND(attr._a_stacksize
+			+ libc.tls_size +  __pthread_tsd_size);
+	}
+
+	if (!tsd) {
+		if (guard) {
+			map = __mmap(0, size, PROT_NONE, MAP_PRIVATE|MAP_ANON, -1, 0);
+			if (map == MAP_FAILED) goto fail;
+			if (__mprotect(map+guard, size-guard, PROT_READ|PROT_WRITE)
+			    && errno != ENOSYS) {
+				__munmap(map, size);
+				goto fail;
+			}
+		} else {
+			map = __mmap(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANON, -1, 0);
+			if (map == MAP_FAILED) goto fail;
+		}
+		tsd = map + size - __pthread_tsd_size;
+		if (!stack) {
+			stack = tsd - libc.tls_size;
+			stack_limit = map + guard;
+		}
+	}
+
+	new = __copy_tls(tsd - libc.tls_size);
+	new->map_base = map;
+	new->map_size = size;
+	new->stack = stack;
+	new->stack_size = stack - stack_limit;
+	new->guard_size = guard;
+	new->self = new;
+	new->tsd = (void *)tsd;
+	new->locale = &libc.global_locale;
+	if (attr._a_detach) {
+		new->detach_state = DT_DETACHED;
+	} else {
+		new->detach_state = DT_JOINABLE;
+	}
+	new->robust_list.head = &new->robust_list.head;
+	new->CANARY = self->CANARY;
+	new->sysinfo = self->sysinfo;
+
+	/* Setup argument structure for the new thread on its stack.
+	 * It's safe to access from the caller only until the thread
+	 * list is unlocked. */
+	stack -= (uintptr_t)stack % sizeof(uintptr_t);
+	stack -= sizeof(struct start_args);
+	struct start_args *args = (void *)stack;
+	args->start_func = entry;
+	args->start_arg = arg;
+	args->control = attr._a_sched ? 1 : 0;
+
+	/* Application signals (but not the synccall signal) must be
+	 * blocked before the thread list lock can be taken, to ensure
+	 * that the lock is AS-safe. */
+	__block_app_sigs(&set);
+
+	/* Ensure SIGCANCEL is unblocked in new thread. This requires
+	 * working with a copy of the set so we can restore the
+	 * original mask in the calling thread. */
+	memcpy(&args->sig_mask, &set, sizeof args->sig_mask);
+	args->sig_mask[(SIGCANCEL-1)/8/sizeof(long)] &=
+		~(1UL<<((SIGCANCEL-1)%(8*sizeof(long))));
+
+	__tl_lock();
+	libc.threads_minus_1++;
+	ret = __clone((c11 ? start_c11 : start), stack, flags, args, &new->tid, TP_ADJ(new), &__thread_list_lock);
+
+	/* All clone failures translate to EAGAIN. If explicit scheduling
+	 * was requested, attempt it before unlocking the thread list so
+	 * that the failed thread is never exposed and so that we can
+	 * clean up all transient resource usage before returning. */
+	if (ret < 0) {
+		ret = -EAGAIN;
+	} else if (attr._a_sched) {
+		ret = __syscall(SYS_sched_setscheduler,
+			new->tid, attr._a_policy, &attr._a_prio);
+		if (a_swap(&args->control, ret ? 3 : 0)==2)
+			__wake(&args->control, 1, 1);
+		if (ret)
+			__wait(&args->control, 0, 3, 0);
+	}
+
+	if (ret >= 0) {
+		new->next = self->next;
+		new->prev = self;
+		new->next->prev = new;
+		new->prev->next = new;
+	} else {
+		libc.threads_minus_1--;
+	}
+	__tl_unlock();
+	__restore_sigs(&set);
+	__release_ptc();
+
+	if (ret < 0) {
+		if (map) __munmap(map, size);
+		return -ret;
+	}
+
+	*res = new;
+	return 0;
+fail:
+	__release_ptc();
+	return EAGAIN;
+}
+
+weak_alias(__pthread_exit, pthread_exit);
+weak_alias(__pthread_create, pthread_create);
+
+struct pthread* __pthread_list_find(pthread_t thread_id, const char* info)
+{
+    struct pthread *thread = (struct pthread *)thread_id; 
+    if (NULL == thread) {
+        log_print("invalid pthread_t (0) passed to %s\n", info);
+        return NULL;
+    }
+
+    struct pthread *self = __pthread_self();
+    if (thread == self) {
+        return thread;
+    }
+    struct pthread *t = self;
+    t = t->next ;
+    while (t != self) {
+        if (t == thread) return thread;
+        t = t->next ;
+    }
+    log_print("invalid pthread_t %p passed to %s\n", thread, info); 
+    return NULL;
+}
+
+pid_t __pthread_gettid_np(pthread_t t)
+{
+    __tl_lock();
+    struct pthread* thread = __pthread_list_find(t, "pthread_gettid_np");
+    __tl_unlock();
+    return thread ? thread->tid : -1;
+}
+weak_alias(__pthread_gettid_np, pthread_gettid_np);
\ No newline at end of file