Implement a stub pthreads library for THREAD_MODEL=single (#518)

~~This patch series first starts with a number of commits stubbing out
functions in the existing `THREAD_model=posix` code. According to "The
Open Group Base Specifications Issue 7, 2018 edition", there are a
number of mandatory functions which have not been provided. There are
also some optional functions that have been partially provided in a
not-useful way (e.g. get but no set function). For these, I have chosen
to clean them up and remove the get functions for consistency.~~ EDIT:
These have been split off into separate PRs and merged.

The remainder of the patches then build up a stub implementation of
pthreads for `THREAD_MODEL=single`. I have done my best to try to make
sure that all functions are as conforming as possible (under the
assumption that another thread cannot ever be launched). This means that
objects such as mutexes and rwlocks actually do update their state and
will correctly fail when locks cannot be acquired.

When an inevitable deadlock occurs, I have chosen to return EDEADLK when
it has been explicitly listed as a permissible return value, and to
invoke `__builtin_trap` otherwise.

I have tested this by rebuilding libc++ with threads enabled and then
smoke-testing Clang/LLVM-on-WASI to make sure that it can compile a
simple program. I have not run any more-extensive conformance testing.

Fixes #501

Makefile

@@ -76,6 +76,7 @@ DLMALLOC_SOURCES = $(DLMALLOC_SRC_DIR)/dlmalloc.c
 DLMALLOC_INC = $(DLMALLOC_DIR)/include
 EMMALLOC_DIR = emmalloc
 EMMALLOC_SOURCES = $(EMMALLOC_DIR)/emmalloc.c
+STUB_PTHREADS_DIR = stub-pthreads
 LIBC_BOTTOM_HALF_DIR = libc-bottom-half
 LIBC_BOTTOM_HALF_CLOUDLIBC_SRC = $(LIBC_BOTTOM_HALF_DIR)/cloudlibc/src
 LIBC_BOTTOM_HALF_CLOUDLIBC_SRC_INC = $(LIBC_BOTTOM_HALF_CLOUDLIBC_SRC)/include
@@ -142,6 +143,8 @@ LIBWASI_EMULATED_SIGNAL_SOURCES = \
 LIBWASI_EMULATED_SIGNAL_MUSL_SOURCES = \
     $(LIBC_TOP_HALF_MUSL_SRC_DIR)/signal/psignal.c \
     $(LIBC_TOP_HALF_MUSL_SRC_DIR)/string/strsignal.c
+LIBWASI_EMULATED_PTHREAD_SOURCES = \
+    $(STUB_PTHREADS_DIR)/stub-pthreads-emulated.c
 LIBDL_SOURCES = $(LIBC_TOP_HALF_MUSL_SRC_DIR)/misc/dl.c
 LIBSETJMP_SOURCES = $(LIBC_TOP_HALF_MUSL_SRC_DIR)/setjmp/wasm32/rt.c
 LIBC_BOTTOM_HALF_CRT_SOURCES = $(wildcard $(LIBC_BOTTOM_HALF_DIR)/crt/*.c)
@@ -274,7 +277,53 @@ LIBC_TOP_HALF_MUSL_SOURCES += \
     )
 endif
 
+# pthreads functions (possibly stub) for either thread model
+LIBC_TOP_HALF_MUSL_SOURCES += \
+    $(addprefix $(LIBC_TOP_HALF_MUSL_SRC_DIR)/, \
+        thread/default_attr.c \
+        thread/pthread_attr_destroy.c \
+        thread/pthread_attr_get.c \
+        thread/pthread_attr_init.c \
+        thread/pthread_attr_setdetachstate.c \
+        thread/pthread_attr_setguardsize.c \
+        thread/pthread_attr_setschedparam.c \
+        thread/pthread_attr_setstack.c \
+        thread/pthread_attr_setstacksize.c \
+        thread/pthread_barrierattr_destroy.c \
+        thread/pthread_barrierattr_init.c \
+        thread/pthread_barrierattr_setpshared.c \
+        thread/pthread_cancel.c \
+        thread/pthread_cleanup_push.c \
+        thread/pthread_condattr_destroy.c \
+        thread/pthread_condattr_init.c \
+        thread/pthread_condattr_setclock.c \
+        thread/pthread_condattr_setpshared.c \
+        thread/pthread_equal.c \
+        thread/pthread_getspecific.c \
+        thread/pthread_key_create.c \
+        thread/pthread_mutex_destroy.c \
+        thread/pthread_mutex_init.c \
+        thread/pthread_mutexattr_destroy.c \
+        thread/pthread_mutexattr_init.c \
+        thread/pthread_mutexattr_setprotocol.c \
+        thread/pthread_mutexattr_setpshared.c \
+        thread/pthread_mutexattr_setrobust.c \
+        thread/pthread_mutexattr_settype.c \
+        thread/pthread_rwlock_destroy.c \
+        thread/pthread_rwlock_init.c \
+        thread/pthread_rwlockattr_destroy.c \
+        thread/pthread_rwlockattr_init.c \
+        thread/pthread_rwlockattr_setpshared.c \
+        thread/pthread_self.c \
+        thread/pthread_setcancelstate.c \
+        thread/pthread_setcanceltype.c \
+        thread/pthread_setspecific.c \
+        thread/pthread_spin_destroy.c \
+        thread/pthread_spin_init.c \
+        thread/pthread_testcancel.c \
+    )
 ifeq ($(THREAD_MODEL), posix)
+# pthreads functions needed for actual thread support
 LIBC_TOP_HALF_MUSL_SOURCES += \
     $(addprefix $(LIBC_TOP_HALF_MUSL_SRC_DIR)/, \
         env/__init_tls.c \
@@ -285,57 +334,26 @@ LIBC_TOP_HALF_MUSL_SOURCES += \
         thread/__lock.c \
         thread/__wait.c \
         thread/__timedwait.c \
-        thread/default_attr.c \
-        thread/pthread_attr_destroy.c \
-        thread/pthread_attr_get.c \
-        thread/pthread_attr_init.c \
-        thread/pthread_attr_setdetachstate.c \
-        thread/pthread_attr_setguardsize.c \
-        thread/pthread_attr_setstack.c \
-        thread/pthread_attr_setstacksize.c \
-        thread/pthread_attr_setschedparam.c \
         thread/pthread_barrier_destroy.c \
         thread/pthread_barrier_init.c \
         thread/pthread_barrier_wait.c \
-        thread/pthread_barrierattr_destroy.c \
-        thread/pthread_barrierattr_init.c \
-        thread/pthread_barrierattr_setpshared.c \
-        thread/pthread_cleanup_push.c \
-        thread/pthread_cancel.c \
         thread/pthread_cond_broadcast.c \
         thread/pthread_cond_destroy.c \
         thread/pthread_cond_init.c \
         thread/pthread_cond_signal.c \
         thread/pthread_cond_timedwait.c \
         thread/pthread_cond_wait.c \
-        thread/pthread_condattr_destroy.c \
-        thread/pthread_condattr_init.c \
-        thread/pthread_condattr_setclock.c \
-        thread/pthread_condattr_setpshared.c \
         thread/pthread_create.c \
         thread/pthread_detach.c \
-        thread/pthread_equal.c \
         thread/pthread_getattr_np.c \
-        thread/pthread_getspecific.c \
         thread/pthread_join.c \
-        thread/pthread_key_create.c \
         thread/pthread_mutex_consistent.c \
-        thread/pthread_mutex_destroy.c \
-        thread/pthread_mutex_init.c \
         thread/pthread_mutex_getprioceiling.c \
         thread/pthread_mutex_lock.c \
         thread/pthread_mutex_timedlock.c \
         thread/pthread_mutex_trylock.c \
         thread/pthread_mutex_unlock.c \
-        thread/pthread_mutexattr_destroy.c \
-        thread/pthread_mutexattr_init.c \
-        thread/pthread_mutexattr_setprotocol.c \
-        thread/pthread_mutexattr_setpshared.c \
-        thread/pthread_mutexattr_setrobust.c \
-        thread/pthread_mutexattr_settype.c \
         thread/pthread_once.c \
-        thread/pthread_rwlock_destroy.c \
-        thread/pthread_rwlock_init.c \
         thread/pthread_rwlock_rdlock.c \
         thread/pthread_rwlock_timedrdlock.c \
         thread/pthread_rwlock_timedwrlock.c \
@@ -343,19 +361,9 @@ LIBC_TOP_HALF_MUSL_SOURCES += \
         thread/pthread_rwlock_trywrlock.c \
         thread/pthread_rwlock_unlock.c \
         thread/pthread_rwlock_wrlock.c \
-        thread/pthread_rwlockattr_destroy.c \
-        thread/pthread_rwlockattr_init.c \
-        thread/pthread_rwlockattr_setpshared.c \
-        thread/pthread_setcancelstate.c \
-        thread/pthread_setcanceltype.c \
-        thread/pthread_setspecific.c \
-        thread/pthread_self.c \
-        thread/pthread_spin_destroy.c \
-        thread/pthread_spin_init.c \
         thread/pthread_spin_lock.c \
         thread/pthread_spin_trylock.c \
         thread/pthread_spin_unlock.c \
-        thread/pthread_testcancel.c \
         thread/sem_destroy.c \
         thread/sem_getvalue.c \
         thread/sem_init.c \
@@ -366,6 +374,16 @@ LIBC_TOP_HALF_MUSL_SOURCES += \
         thread/wasm32/wasi_thread_start.s \
     )
 endif
+ifeq ($(THREAD_MODEL), single)
+# pthreads stubs for single-threaded environment
+LIBC_TOP_HALF_MUSL_SOURCES += \
+    $(STUB_PTHREADS_DIR)/barrier.c \
+    $(STUB_PTHREADS_DIR)/condvar.c \
+    $(STUB_PTHREADS_DIR)/mutex.c \
+    $(STUB_PTHREADS_DIR)/rwlock.c \
+    $(STUB_PTHREADS_DIR)/spinlock.c \
+    $(STUB_PTHREADS_DIR)/stub-pthreads-good.c
+endif
 
 MUSL_PRINTSCAN_SOURCES = \
     $(LIBC_TOP_HALF_MUSL_SRC_DIR)/internal/floatscan.c \
@@ -418,10 +436,10 @@ ifeq ($(THREAD_MODEL), posix)
 # https://reviews.llvm.org/D130053).
 CFLAGS += -mthread-model posix -pthread -ftls-model=local-exec
 ASMFLAGS += -matomics
+endif
 
 # Include cloudlib's directory to access the structure definition of clockid_t
 CFLAGS += -I$(LIBC_BOTTOM_HALF_CLOUDLIBC_SRC)
-endif
 
 ifneq ($(LTO),no)
 ifeq ($(LTO),full)
@@ -490,6 +508,7 @@ LIBWASI_EMULATED_PROCESS_CLOCKS_OBJS = $(call objs,$(LIBWASI_EMULATED_PROCESS_CL
 LIBWASI_EMULATED_GETPID_OBJS = $(call objs,$(LIBWASI_EMULATED_GETPID_SOURCES))
 LIBWASI_EMULATED_SIGNAL_OBJS = $(call objs,$(LIBWASI_EMULATED_SIGNAL_SOURCES))
 LIBWASI_EMULATED_SIGNAL_MUSL_OBJS = $(call objs,$(LIBWASI_EMULATED_SIGNAL_MUSL_SOURCES))
+LIBWASI_EMULATED_PTHREAD_OBJS = $(call objs,$(LIBWASI_EMULATED_PTHREAD_SOURCES))
 LIBDL_OBJS = $(call objs,$(LIBDL_SOURCES))
 LIBSETJMP_OBJS = $(call objs,$(LIBSETJMP_SOURCES))
 LIBC_BOTTOM_HALF_CRT_OBJS = $(call objs,$(LIBC_BOTTOM_HALF_CRT_SOURCES))
@@ -593,6 +612,7 @@ LIBWASI_EMULATED_PROCESS_CLOCKS_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBWASI_EMULA
 LIBWASI_EMULATED_GETPID_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBWASI_EMULATED_GETPID_OBJS))
 LIBWASI_EMULATED_SIGNAL_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBWASI_EMULATED_SIGNAL_OBJS))
 LIBWASI_EMULATED_SIGNAL_MUSL_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBWASI_EMULATED_SIGNAL_MUSL_OBJS))
+LIBWASI_EMULATED_PTHREAD_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBWASI_EMULATED_PTHREAD_OBJS))
 LIBDL_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBDL_OBJS))
 LIBSETJMP_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBSETJMP_OBJS))
 BULK_MEMORY_SO_OBJS = $(patsubst %.o,%.pic.o,$(BULK_MEMORY_OBJS))
@@ -608,6 +628,7 @@ PIC_OBJS = \
 	$(LIBWASI_EMULATED_GETPID_SO_OBJS) \
 	$(LIBWASI_EMULATED_SIGNAL_SO_OBJS) \
 	$(LIBWASI_EMULATED_SIGNAL_MUSL_SO_OBJS) \
+	$(LIBWASI_EMULATED_PTHREAD_SO_OBJS) \
 	$(LIBDL_SO_OBJS) \
 	$(LIBSETJMP_SO_OBJS) \
 	$(BULK_MEMORY_SO_OBJS) \
@@ -649,6 +670,8 @@ $(OBJDIR)/libwasi-emulated-getpid.so.a: $(LIBWASI_EMULATED_GETPID_SO_OBJS)
 
 $(OBJDIR)/libwasi-emulated-signal.so.a: $(LIBWASI_EMULATED_SIGNAL_SO_OBJS) $(LIBWASI_EMULATED_SIGNAL_MUSL_SO_OBJS)
 
+$(OBJDIR)/libwasi-emulated-pthread.so.a: $(LIBWASI_EMULATED_PTHREAD_SO_OBJS)
+
 $(OBJDIR)/libdl.so.a: $(LIBDL_SO_OBJS)
 
 $(OBJDIR)/libsetjmp.so.a: $(LIBSETJMP_SO_OBJS)
@@ -667,6 +690,8 @@ $(SYSROOT_LIB)/libwasi-emulated-getpid.a: $(LIBWASI_EMULATED_GETPID_OBJS)
 
 $(SYSROOT_LIB)/libwasi-emulated-signal.a: $(LIBWASI_EMULATED_SIGNAL_OBJS) $(LIBWASI_EMULATED_SIGNAL_MUSL_OBJS)
 
+$(SYSROOT_LIB)/libwasi-emulated-pthread.a: $(LIBWASI_EMULATED_PTHREAD_OBJS)
+
 $(SYSROOT_LIB)/libdl.a: $(LIBDL_OBJS)
 
 $(SYSROOT_LIB)/libsetjmp.a: $(LIBSETJMP_OBJS)
@@ -769,6 +794,12 @@ $(FTS_OBJS): CFLAGS += \
 $(LIBWASI_EMULATED_PROCESS_CLOCKS_OBJS) $(LIBWASI_EMULATED_PROCESS_CLOCKS_SO_OBJS): CFLAGS += \
     -I$(LIBC_BOTTOM_HALF_CLOUDLIBC_SRC)
 
+$(LIBWASI_EMULATED_PTHREAD_OBJS) $(LIBWASI_EMULATED_PTHREAD_SO_OBJS): CFLAGS += \
+    -I$(LIBC_TOP_HALF_MUSL_SRC_DIR)/include \
+    -I$(LIBC_TOP_HALF_MUSL_SRC_DIR)/internal \
+    -I$(LIBC_TOP_HALF_MUSL_DIR)/arch/wasm32 \
+    -D_WASI_EMULATED_PTHREAD
+
 # emmalloc uses a lot of pointer type-punning, which is UB under strict aliasing,
 # and this was found to have real miscompilations in wasi-libc#421.
 $(EMMALLOC_OBJS): CFLAGS += \
@@ -820,6 +851,7 @@ LIBC_SO = \
 	$(SYSROOT_LIB)/libwasi-emulated-process-clocks.so \
 	$(SYSROOT_LIB)/libwasi-emulated-getpid.so \
 	$(SYSROOT_LIB)/libwasi-emulated-signal.so \
+	$(SYSROOT_LIB)/libwasi-emulated-pthread.so \
 	$(SYSROOT_LIB)/libdl.so
 ifeq ($(BUILD_LIBSETJMP),yes)
 LIBC_SO += \
@@ -838,6 +870,10 @@ STATIC_LIBS = \
     $(SYSROOT_LIB)/libwasi-emulated-getpid.a \
     $(SYSROOT_LIB)/libwasi-emulated-signal.a \
     $(SYSROOT_LIB)/libdl.a
+ifneq ($(THREAD_MODEL), posix)
+    STATIC_LIBS += \
+        $(SYSROOT_LIB)/libwasi-emulated-pthread.a
+endif
 ifeq ($(BUILD_LIBSETJMP),yes)
 STATIC_LIBS += \
 	$(SYSROOT_LIB)/libsetjmp.a

expected/wasm32-wasip1/defined-symbols.txt

@@ -11,6 +11,7 @@ __EINVAL
 __ENOMEM
 __SIG_ERR
 __SIG_IGN
+__acquire_ptc
 __asctime_r
 __assert_fail
 __c_dot_utf8
@@ -34,7 +35,11 @@ __ctype_tolower_loc
 __ctype_toupper_loc
 __cxa_atexit
 __cxa_finalize
+__default_guardsize
+__default_stacksize
 __des_setkey
+__do_cleanup_pop
+__do_cleanup_push
 __do_des
 __duplocale
 __env_rm_add
@@ -175,10 +180,34 @@ __pow_log_data
 __powf_log2_data
 __progname
 __progname_full
+__pthread_cond_timedwait
+__pthread_create
+__pthread_detach
+__pthread_join
+__pthread_key_create
+__pthread_key_delete
+__pthread_mutex_consistent
+__pthread_mutex_lock
+__pthread_mutex_timedlock
+__pthread_mutex_trylock
+__pthread_mutex_unlock
+__pthread_rwlock_rdlock
+__pthread_rwlock_timedrdlock
+__pthread_rwlock_timedwrlock
+__pthread_rwlock_tryrdlock
+__pthread_rwlock_trywrlock
+__pthread_rwlock_unlock
+__pthread_rwlock_wrlock
+__pthread_setcancelstate
+__pthread_testcancel
+__pthread_tsd_main
+__pthread_tsd_run_dtors
+__pthread_tsd_size
 __putenv
 __qsort_r
 __rand48_step
 __reallocarray
+__release_ptc
 __rem_pio2
 __rem_pio2_large
 __rem_pio2f
@@ -234,6 +263,9 @@ __sysv_signal
 __tan
 __tandf
 __tanl
+__testcancel
+__tl_lock
+__tl_unlock
 __tm_to_secs
 __tm_to_tzname
 __tolower_l
@@ -330,6 +362,7 @@ __wasilibc_nocwd_symlinkat
 __wasilibc_nocwd_utimensat
 __wasilibc_open_nomode
 __wasilibc_populate_preopens
+__wasilibc_pthread_self
 __wasilibc_register_preopened_fd
 __wasilibc_rename_newat
 __wasilibc_rename_oldat
@@ -352,6 +385,8 @@ _environ
 _exit
 _flushlbf
 _initialize
+_pthread_cleanup_pop
+_pthread_cleanup_push
 _start
 a64l
 abort
@@ -921,6 +956,89 @@ program_invocation_name
 program_invocation_short_name
 pselect
 psignal
+pthread_attr_destroy
+pthread_attr_getdetachstate
+pthread_attr_getguardsize
+pthread_attr_getschedparam
+pthread_attr_getstack
+pthread_attr_getstacksize
+pthread_attr_init
+pthread_attr_setdetachstate
+pthread_attr_setguardsize
+pthread_attr_setschedparam
+pthread_attr_setstack
+pthread_attr_setstacksize
+pthread_barrier_destroy
+pthread_barrier_init
+pthread_barrier_wait
+pthread_barrierattr_destroy
+pthread_barrierattr_getpshared
+pthread_barrierattr_init
+pthread_barrierattr_setpshared
+pthread_cancel
+pthread_cond_broadcast
+pthread_cond_destroy
+pthread_cond_init
+pthread_cond_signal
+pthread_cond_timedwait
+pthread_cond_wait
+pthread_condattr_destroy
+pthread_condattr_getclock
+pthread_condattr_getpshared
+pthread_condattr_init
+pthread_condattr_setclock
+pthread_condattr_setpshared
+pthread_create
+pthread_detach
+pthread_equal
+pthread_exit
+pthread_getspecific
+pthread_join
+pthread_key_create
+pthread_key_delete
+pthread_mutex_consistent
+pthread_mutex_destroy
+pthread_mutex_init
+pthread_mutex_lock
+pthread_mutex_timedlock
+pthread_mutex_trylock
+pthread_mutex_unlock
+pthread_mutexattr_destroy
+pthread_mutexattr_getprotocol
+pthread_mutexattr_getpshared
+pthread_mutexattr_getrobust
+pthread_mutexattr_gettype
+pthread_mutexattr_init
+pthread_mutexattr_setprotocol
+pthread_mutexattr_setpshared
+pthread_mutexattr_setrobust
+pthread_mutexattr_settype
+pthread_once
+pthread_rwlock_destroy
+pthread_rwlock_init
+pthread_rwlock_rdlock
+pthread_rwlock_timedrdlock
+pthread_rwlock_timedwrlock
+pthread_rwlock_tryrdlock
+pthread_rwlock_trywrlock
+pthread_rwlock_unlock
+pthread_rwlock_wrlock
+pthread_rwlockattr_destroy
+pthread_rwlockattr_getpshared
+pthread_rwlockattr_init
+pthread_rwlockattr_setpshared
+pthread_self
+pthread_setcancelstate
+pthread_setcanceltype
+pthread_setspecific
+pthread_spin_destroy
+pthread_spin_init
+pthread_spin_lock
+pthread_spin_trylock
+pthread_spin_unlock
+pthread_testcancel
+pthread_timedjoin_np
+pthread_tryjoin_np
 putc
 putc_unlocked
 putchar
@@ -1102,6 +1220,8 @@ tfind
 tgamma
 tgammaf
 tgammal
+thrd_current
+thrd_equal
 thrd_sleep
 time
 timegm
@@ -1123,6 +1243,7 @@ truncate
 truncf
 truncl
 tsearch
+tss_get
 twalk
 uname
 ungetc

expected/wasm32-wasip1/predefined-macros.txt

@@ -1463,8 +1463,10 @@
 #define PTHREAD_COND_INITIALIZER {{{0}}}
 #define PTHREAD_CREATE_DETACHED 1
 #define PTHREAD_CREATE_JOINABLE 0
+#define PTHREAD_DESTRUCTOR_ITERATIONS 4
 #define PTHREAD_EXPLICIT_SCHED 1
 #define PTHREAD_INHERIT_SCHED 0
+#define PTHREAD_KEYS_MAX 128
 #define PTHREAD_MUTEX_DEFAULT 0
 #define PTHREAD_MUTEX_ERRORCHECK 2
 #define PTHREAD_MUTEX_INITIALIZER {{{0}}}
@@ -1482,6 +1484,7 @@
 #define PTHREAD_RWLOCK_INITIALIZER {{{0}}}
 #define PTHREAD_SCOPE_PROCESS 1
 #define PTHREAD_SCOPE_SYSTEM 0
+#define PTHREAD_STACK_MIN 2048
 #define PTRBITS (sizeof(char *) * 8)
 #define PTRDIFF_MAX INT32_MAX
 #define PTRDIFF_MIN INT32_MIN
@@ -3386,7 +3389,12 @@
 #define preadv64 preadv
 #define pthread_cleanup_pop(r) _pthread_cleanup_pop(&__cb, (r)); } while(0)
 #define pthread_cleanup_push(f,x) do { struct __ptcb __cb; _pthread_cleanup_push(&__cb, f, x);
+#define pthread_create(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_detach(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
 #define pthread_equal(x,y) ((x)==(y))
+#define pthread_join(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_timedjoin_np(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_tryjoin_np(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
 #define pwrite64 pwrite
 #define pwritev64 pwritev
 #define readdir64 readdir

expected/wasm32-wasip2/defined-symbols.txt

@@ -12,6 +12,7 @@ __EINVAL
 __ENOMEM
 __SIG_ERR
 __SIG_IGN
+__acquire_ptc
 __asctime_r
 __assert_fail
 __c_dot_utf8
@@ -37,7 +38,11 @@ __ctype_tolower_loc
 __ctype_toupper_loc
 __cxa_atexit
 __cxa_finalize
+__default_guardsize
+__default_stacksize
 __des_setkey
+__do_cleanup_pop
+__do_cleanup_push
 __do_des
 __duplocale
 __env_rm_add
@@ -178,10 +183,34 @@ __pow_log_data
 __powf_log2_data
 __progname
 __progname_full
+__pthread_cond_timedwait
+__pthread_create
+__pthread_detach
+__pthread_join
+__pthread_key_create
+__pthread_key_delete
+__pthread_mutex_consistent
+__pthread_mutex_lock
+__pthread_mutex_timedlock
+__pthread_mutex_trylock
+__pthread_mutex_unlock
+__pthread_rwlock_rdlock
+__pthread_rwlock_timedrdlock
+__pthread_rwlock_timedwrlock
+__pthread_rwlock_tryrdlock
+__pthread_rwlock_trywrlock
+__pthread_rwlock_unlock
+__pthread_rwlock_wrlock
+__pthread_setcancelstate
+__pthread_testcancel
+__pthread_tsd_main
+__pthread_tsd_run_dtors
+__pthread_tsd_size
 __putenv
 __qsort_r
 __rand48_step
 __reallocarray
+__release_ptc
 __rem_pio2
 __rem_pio2_large
 __rem_pio2f
@@ -237,6 +266,9 @@ __sysv_signal
 __tan
 __tandf
 __tanl
+__testcancel
+__tl_lock
+__tl_unlock
 __tm_to_secs
 __tm_to_tzname
 __tolower_l
@@ -346,6 +378,7 @@ __wasilibc_nocwd_symlinkat
 __wasilibc_nocwd_utimensat
 __wasilibc_open_nomode
 __wasilibc_populate_preopens
+__wasilibc_pthread_self
 __wasilibc_register_preopened_fd
 __wasilibc_rename_newat
 __wasilibc_rename_oldat
@@ -368,6 +401,8 @@ _environ
 _exit
 _flushlbf
 _initialize
+_pthread_cleanup_pop
+_pthread_cleanup_push
 _start
 a64l
 abort
@@ -1054,6 +1089,89 @@ program_invocation_name
 program_invocation_short_name
 pselect
 psignal
+pthread_attr_destroy
+pthread_attr_getdetachstate
+pthread_attr_getguardsize
+pthread_attr_getschedparam
+pthread_attr_getstack
+pthread_attr_getstacksize
+pthread_attr_init
+pthread_attr_setdetachstate
+pthread_attr_setguardsize
+pthread_attr_setschedparam
+pthread_attr_setstack
+pthread_attr_setstacksize
+pthread_barrier_destroy
+pthread_barrier_init
+pthread_barrier_wait
+pthread_barrierattr_destroy
+pthread_barrierattr_getpshared
+pthread_barrierattr_init
+pthread_barrierattr_setpshared
+pthread_cancel
+pthread_cond_broadcast
+pthread_cond_destroy
+pthread_cond_init
+pthread_cond_signal
+pthread_cond_timedwait
+pthread_cond_wait
+pthread_condattr_destroy
+pthread_condattr_getclock
+pthread_condattr_getpshared
+pthread_condattr_init
+pthread_condattr_setclock
+pthread_condattr_setpshared
+pthread_create
+pthread_detach
+pthread_equal
+pthread_exit
+pthread_getspecific
+pthread_join
+pthread_key_create
+pthread_key_delete
+pthread_mutex_consistent
+pthread_mutex_destroy
+pthread_mutex_init
+pthread_mutex_lock
+pthread_mutex_timedlock
+pthread_mutex_trylock
+pthread_mutex_unlock
+pthread_mutexattr_destroy
+pthread_mutexattr_getprotocol
+pthread_mutexattr_getpshared
+pthread_mutexattr_getrobust
+pthread_mutexattr_gettype
+pthread_mutexattr_init
+pthread_mutexattr_setprotocol
+pthread_mutexattr_setpshared
+pthread_mutexattr_setrobust
+pthread_mutexattr_settype
+pthread_once
+pthread_rwlock_destroy
+pthread_rwlock_init
+pthread_rwlock_rdlock
+pthread_rwlock_timedrdlock
+pthread_rwlock_timedwrlock
+pthread_rwlock_tryrdlock
+pthread_rwlock_trywrlock
+pthread_rwlock_unlock
+pthread_rwlock_wrlock
+pthread_rwlockattr_destroy
+pthread_rwlockattr_getpshared
+pthread_rwlockattr_init
+pthread_rwlockattr_setpshared
+pthread_self
+pthread_setcancelstate
+pthread_setcanceltype
+pthread_setspecific
+pthread_spin_destroy
+pthread_spin_init
+pthread_spin_lock
+pthread_spin_trylock
+pthread_spin_unlock
+pthread_testcancel
+pthread_timedjoin_np
+pthread_tryjoin_np
 putc
 putc_unlocked
 putchar
@@ -1335,6 +1453,8 @@ tfind
 tgamma
 tgammaf
 tgammal
+thrd_current
+thrd_equal
 thrd_sleep
 time
 timegm
@@ -1356,6 +1476,7 @@ truncate
 truncf
 truncl
 tsearch
+tss_get
 twalk
 udp_accept
 udp_bind

expected/wasm32-wasip2/predefined-macros.txt

@@ -1594,8 +1594,10 @@
 #define PTHREAD_COND_INITIALIZER {{{0}}}
 #define PTHREAD_CREATE_DETACHED 1
 #define PTHREAD_CREATE_JOINABLE 0
+#define PTHREAD_DESTRUCTOR_ITERATIONS 4
 #define PTHREAD_EXPLICIT_SCHED 1
 #define PTHREAD_INHERIT_SCHED 0
+#define PTHREAD_KEYS_MAX 128
 #define PTHREAD_MUTEX_DEFAULT 0
 #define PTHREAD_MUTEX_ERRORCHECK 2
 #define PTHREAD_MUTEX_INITIALIZER {{{0}}}
@@ -1613,6 +1615,7 @@
 #define PTHREAD_RWLOCK_INITIALIZER {{{0}}}
 #define PTHREAD_SCOPE_PROCESS 1
 #define PTHREAD_SCOPE_SYSTEM 0
+#define PTHREAD_STACK_MIN 2048
 #define PTRBITS (sizeof(char *) * 8)
 #define PTRDIFF_MAX INT32_MAX
 #define PTRDIFF_MIN INT32_MIN
@@ -3541,7 +3544,12 @@
 #define preadv64 preadv
 #define pthread_cleanup_pop(r) _pthread_cleanup_pop(&__cb, (r)); } while(0)
 #define pthread_cleanup_push(f,x) do { struct __ptcb __cb; _pthread_cleanup_push(&__cb, f, x);
+#define pthread_create(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_detach(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
 #define pthread_equal(x,y) ((x)==(y))
+#define pthread_join(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_timedjoin_np(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_tryjoin_np(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; to enable stub functions which always fail, compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
 #define pwrite64 pwrite
 #define pwritev64 pwritev
 #define readdir64 readdir

libc-top-half/musl/include/limits.h

@@ -65,11 +65,9 @@
 
 /* Implementation choices... */
 
-#if defined(__wasilibc_unmodified_upstream) || defined(_REENTRANT)
 #define PTHREAD_KEYS_MAX 128
 #define PTHREAD_STACK_MIN 2048
 #define PTHREAD_DESTRUCTOR_ITERATIONS 4
-#endif
 #if defined(__wasilibc_unmodified_upstream) || defined(_REENTRANT)
 #define SEM_VALUE_MAX 0x7fffffff
 #define SEM_NSEMS_MAX 256

libc-top-half/musl/include/pthread.h

@@ -77,12 +77,29 @@ extern "C" {
 #define PTHREAD_NULL ((pthread_t)0)
 
 
+#ifdef __wasilibc_unmodified_upstream
 int pthread_create(pthread_t *__restrict, const pthread_attr_t *__restrict, void *(*)(void *), void *__restrict);
 int pthread_detach(pthread_t);
-#ifdef __wasilibc_unmodified_upstream
 _Noreturn void pthread_exit(void *);
-#endif
 int pthread_join(pthread_t, void **);
+#else
+#if defined(_WASI_EMULATED_PTHREAD) || defined(_REENTRANT)
+int pthread_create(pthread_t *__restrict, const pthread_attr_t *__restrict, void *(*)(void *), void *__restrict);
+int pthread_detach(pthread_t);
+int pthread_join(pthread_t, void **);
+#else
+#include <assert.h>
+#define pthread_create(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; \
+to enable stub functions which always fail, \
+compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_detach(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; \
+to enable stub functions which always fail, \
+compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_join(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; \
+to enable stub functions which always fail, \
+compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#endif
+#endif
 
 #ifdef __GNUC__
 __attribute__((const))
@@ -232,8 +249,17 @@ int pthread_setname_np(pthread_t, const char *);
 int pthread_getname_np(pthread_t, char *, size_t);
 int pthread_getattr_default_np(pthread_attr_t *);
 int pthread_setattr_default_np(const pthread_attr_t *);
+#if defined(__wasilibc_unmodified_upstream) || defined(_WASI_EMULATED_PTHREAD) || defined(_REENTRANT)
 int pthread_tryjoin_np(pthread_t, void **);
 int pthread_timedjoin_np(pthread_t, void **, const struct timespec *);
+#else
+#define pthread_tryjoin_np(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; \
+to enable stub functions which always fail, \
+compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#define pthread_timedjoin_np(...) ({ _Static_assert(0, "This mode of WASI does not have threads enabled; \
+to enable stub functions which always fail, \
+compile with -D_WASI_EMULATED_PTHREAD and link with -lwasi-emulated-pthread"); 0;})
+#endif
 #endif
 
 #if _REDIR_TIME64

libc-top-half/musl/src/internal/pthread_impl.h

@@ -186,8 +186,10 @@ static inline void __wake(volatile void *addr, int cnt, int priv)
 	__syscall(SYS_futex, addr, FUTEX_WAKE|priv, cnt) != -ENOSYS ||
 	__syscall(SYS_futex, addr, FUTEX_WAKE, cnt);
 #else
+#ifdef _REENTRANT
 	__builtin_wasm_memory_atomic_notify((int*)addr, cnt);
 #endif
+#endif
 }
 static inline void __futexwait(volatile void *addr, int val, int priv)
 {

libc-top-half/musl/src/thread/pthread_self.c

@@ -1,8 +1,7 @@
 #include "pthread_impl.h"
 #include <threads.h>
 
-#if !defined(__wasilibc_unmodified_upstream) && defined(__wasm__) &&           \
+#if !defined(__wasilibc_unmodified_upstream) && defined(__wasm__)
-    defined(_REENTRANT)
 _Thread_local struct pthread __wasilibc_pthread_self;
 #endif
 

stub-pthreads/README.md

@@ -0,0 +1,16 @@
+# Single-threaded pthreads stubs
+
+The goal of these files is to provide stub implementations of pthreads
+functions for `THREAD_MODEL=single`. This implementation should _always_
+follow the strict letter of the POSIX specifications. This means that
+"doing nothing", "always succeeding", etc. are not proper implementations
+-- these stubs aim for higher conformance than that.
+
+The code that is "more" aligned with the spirit of the POSIX specifications
+ ends up compiled into libc itself. This primarily consists of synchronization
+  primitives (which are implemented to actually track state).
+
+The code that is "less" aligned with the spirit of the specifications
+(e.g. by "rules-lawyering" and always failing) are built into a library
+`wasi-emulated-pthread.a`. The distinction is ultimately made by vibes and a
+judgement call, not formal criteria.

stub-pthreads/barrier.c

@@ -0,0 +1,23 @@
+#include "pthread_impl.h"
+/*
+	Note on PTHREAD_PROCESS_SHARED:
+	Because WASM doesn't have virtual memory nor subprocesses,
+	there isn't any way for the same synchronization object to have multiple mappings.
+	Thus we can completely ignore it here.
+*/
+
+int pthread_barrier_init(pthread_barrier_t *restrict b, const pthread_barrierattr_t *restrict a, unsigned count)
+{
+	if (count-1 > INT_MAX-1) return EINVAL;
+	*b = (pthread_barrier_t){ ._b_limit = count-1 };
+	return 0;
+}
+int pthread_barrier_destroy(pthread_barrier_t *b)
+{
+	return 0;
+}
+int pthread_barrier_wait(pthread_barrier_t *b)
+{
+	if (!b->_b_limit) return PTHREAD_BARRIER_SERIAL_THREAD;
+	__builtin_trap();
+}

stub-pthreads/condvar.c

@@ -0,0 +1,36 @@
+#include "pthread_impl.h"
+#include <time.h>
+
+int pthread_cond_init(pthread_cond_t *restrict c, const pthread_condattr_t *restrict a)
+{
+	return 0;
+}
+int pthread_cond_destroy(pthread_cond_t *c)
+{
+	return 0;
+}
+int pthread_cond_broadcast(pthread_cond_t *c)
+{
+	return 0;
+}
+int pthread_cond_signal(pthread_cond_t *c)
+{
+	return 0;
+}
+int pthread_cond_wait(pthread_cond_t *restrict c, pthread_mutex_t *restrict m)
+{
+	/* Because there is no other thread that can signal us, this is a deadlock immediately.
+	The other possible choice is to return immediately (spurious wakeup), but that is likely to
+	just result in the program spinning forever on a condition that cannot become true. */
+	__builtin_trap();
+}
+int __pthread_cond_timedwait(pthread_cond_t *restrict c, pthread_mutex_t *restrict m, const struct timespec *restrict ts)
+{
+	/* Error check mutexes must detect if they're not locked (UB for others) */
+	if (!m->_m_count) return EPERM;
+	int ret = clock_nanosleep(CLOCK_REALTIME, TIMER_ABSTIME, ts, 0);
+	if (ret == 0) return ETIMEDOUT;
+	if (ret != EINTR) return ret;
+	return 0;
+}
+weak_alias(__pthread_cond_timedwait, pthread_cond_timedwait);

stub-pthreads/mutex.c

@@ -0,0 +1,67 @@
+#include "pthread_impl.h"
+/*
+	Musl mutex (FYI)
+
+	_m_type:
+	b[7]	- process shared
+	b[3]	- priority inherit
+	b[2] 	- robust
+	b[1:0] 	- type
+		0 - normal
+		1 - recursive
+		2 - errorcheck
+
+	_m_lock:
+	b[30]	- owner dead, if robust
+	b[29:0]	- tid, if not normal
+	b[4]	- locked?, if normal
+*/
+
+int __pthread_mutex_consistent(pthread_mutex_t *m)
+{
+	/* cannot be a robust mutex, as they're entirely unsupported in WASI */
+	return EINVAL;
+}
+weak_alias(__pthread_mutex_consistent, pthread_mutex_consistent);
+
+int __pthread_mutex_lock(pthread_mutex_t *m)
+{
+	if (m->_m_type&3 != PTHREAD_MUTEX_RECURSIVE) {
+		if (m->_m_count) return EDEADLK;
+		m->_m_count = 1;
+	} else {
+		if ((unsigned)m->_m_count >= INT_MAX) return EAGAIN;
+		m->_m_count++;
+	}
+	return 0;
+}
+weak_alias(__pthread_mutex_lock, pthread_mutex_lock);
+
+int __pthread_mutex_trylock(pthread_mutex_t *m)
+{
+	if (m->_m_type&3 != PTHREAD_MUTEX_RECURSIVE) {
+		if (m->_m_count) return EBUSY;
+		m->_m_count = 1;
+	} else {
+		if ((unsigned)m->_m_count >= INT_MAX) return EAGAIN;
+		m->_m_count++;
+	}
+	return 0;
+}
+weak_alias(__pthread_mutex_trylock, pthread_mutex_trylock);
+
+int __pthread_mutex_unlock(pthread_mutex_t *m)
+{
+	if (!m->_m_count) return EPERM;
+	m->_m_count--;
+	return 0;
+}
+weak_alias(__pthread_mutex_unlock, pthread_mutex_unlock);
+
+int __pthread_mutex_timedlock(pthread_mutex_t *restrict m, const struct timespec *restrict at)
+{
+	/* "The pthread_mutex_timedlock() function may fail if: A deadlock condition was detected." */
+	/* This means that we don't have to wait and then return timeout, we can just detect deadlock. */
+	return pthread_mutex_lock(m);
+}
+weak_alias(__pthread_mutex_timedlock, pthread_mutex_timedlock);

stub-pthreads/rwlock.c

@@ -0,0 +1,60 @@
+#include "pthread_impl.h"
+/* Musl uses bit31 to mark "has waiters", bit[30:0] all 1s to indicate writer */
+
+/* These functions have the __ prefix to help stub out thread-specific data */
+
+int __pthread_rwlock_rdlock(pthread_rwlock_t *rw)
+{
+	if (rw->_rw_lock == 0x7fffffff) return EDEADLK;
+	if (rw->_rw_lock == 0x7ffffffe) return EAGAIN;
+	rw->_rw_lock++;
+	return 0;
+}
+weak_alias(__pthread_rwlock_rdlock, pthread_rwlock_rdlock);
+
+int __pthread_rwlock_timedrdlock(pthread_rwlock_t *restrict rw, const struct timespec *restrict at)
+{
+	return pthread_rwlock_rdlock(rw);
+}
+weak_alias(__pthread_rwlock_timedrdlock, pthread_rwlock_timedrdlock);
+
+int __pthread_rwlock_tryrdlock(pthread_rwlock_t *rw)
+{
+	if (rw->_rw_lock == 0x7fffffff) return EBUSY;
+	if (rw->_rw_lock == 0x7ffffffe) return EAGAIN;
+	rw->_rw_lock++;
+	return 0;
+}
+weak_alias(__pthread_rwlock_tryrdlock, pthread_rwlock_tryrdlock);
+
+int __pthread_rwlock_wrlock(pthread_rwlock_t *rw)
+{
+	if (rw->_rw_lock) return EDEADLK;
+	rw->_rw_lock = 0x7fffffff;
+	return 0;
+}
+weak_alias(__pthread_rwlock_wrlock, pthread_rwlock_wrlock);
+
+int __pthread_rwlock_timedwrlock(pthread_rwlock_t *restrict rw, const struct timespec *restrict at)
+{
+	return pthread_rwlock_wrlock(rw);
+}
+weak_alias(__pthread_rwlock_timedwrlock, pthread_rwlock_timedwrlock);
+
+int __pthread_rwlock_trywrlock(pthread_rwlock_t *rw)
+{
+	if (rw->_rw_lock) return EBUSY;
+	rw->_rw_lock = 0x7fffffff;
+	return 0;
+}
+weak_alias(__pthread_rwlock_trywrlock, pthread_rwlock_trywrlock);
+
+int __pthread_rwlock_unlock(pthread_rwlock_t *rw)
+{
+	if (rw->_rw_lock == 0x7fffffff)
+		rw->_rw_lock = 0;
+	else
+		rw->_rw_lock--;
+	return 0;
+}
+weak_alias(__pthread_rwlock_unlock, pthread_rwlock_unlock);

stub-pthreads/spinlock.c

@@ -0,0 +1,21 @@
+
+#include "pthread_impl.h"
+/* The only reason why we have to do anything is trylock */
+
+int pthread_spin_lock(pthread_spinlock_t *s)
+{
+	if (*s) return EDEADLK;
+	*s = 1;
+	return 0;
+}
+int pthread_spin_trylock(pthread_spinlock_t *s)
+{
+	if (*s) return EBUSY;
+	*s = 1;
+	return 0;
+}
+int pthread_spin_unlock(pthread_spinlock_t *s)
+{
+	*s = 0;
+	return 0;
+}

stub-pthreads/stub-pthreads-emulated.c

@@ -0,0 +1,40 @@
+// This file is linked into wasi-emulated-pthread
+
+#include "pthread_impl.h"
+
+int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict attrp, void *(*entry)(void *), void *restrict arg)
+{
+	/*
+		"The system lacked the necessary resources to create another thread,
+		or the system-imposed limit on the total number of threads in a process
+		{PTHREAD_THREADS_MAX} would be exceeded."
+	*/
+	return EAGAIN;
+}
+weak_alias(__pthread_create, pthread_create);
+int __pthread_detach(pthread_t t)
+{
+	/*
+		If we are the only thread, when we exit the whole process exits.
+		So the storage will be reclaimed no matter what.
+	*/
+	return 0;
+}
+weak_alias(__pthread_detach, pthread_detach);
+int __pthread_join(pthread_t t, void **res)
+{
+	/*
+		"The behavior is undefined if the value specified by the thread argument
+		to pthread_join() refers to the calling thread."
+	*/
+	return 0;
+}
+weak_alias(__pthread_join, pthread_join);
+int pthread_tryjoin_np(pthread_t t, void **res)
+{
+	return 0;
+}
+int pthread_timedjoin_np(pthread_t t, void **res, const struct timespec *at)
+{
+	return 0;
+}

stub-pthreads/stub-pthreads-good.c

@@ -0,0 +1,43 @@
+// This file is linked into libc
+
+#include "pthread_impl.h"
+
+static void dummy_0()
+{
+}
+weak_alias(dummy_0, __acquire_ptc);
+weak_alias(dummy_0, __release_ptc);
+weak_alias(dummy_0, __pthread_tsd_run_dtors);
+
+int pthread_once(pthread_once_t *control, void (*init)(void))
+{
+	if (!*control) {
+		init();
+		*control = 1;
+	}
+	return 0;
+}
+
+_Noreturn void pthread_exit(void *result)
+{
+	/*
+		We are the only thread, so when we exit the whole process exits.
+		But we still have to run cancellation handlers...
+	*/
+	pthread_t self = __pthread_self();
+
+	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();
+
+	exit(0);
+}