X-Git-Url: http://repo.macrolet.net/gitweb/?a=blobdiff_plain;f=src%2Fruntime%2Fthread.c;h=1889213ef3cf0bc72bf9214adcd78e469598a400;hb=f0cb0cf9c0fe1b6fce5d10dbd34a0b7b249c4ae8;hp=3614c80b014777e7a4aaf8a8eebdd35525de7c6a;hpb=0b5119848b6b8713e473fa669356645747e11dbd;p=sbcl.git diff --git a/src/runtime/thread.c b/src/runtime/thread.c index 3614c80..1889213 100644 --- a/src/runtime/thread.c +++ b/src/runtime/thread.c @@ -1,17 +1,38 @@ +/* + * This software is part of the SBCL system. See the README file for + * more information. + * + * This software is derived from the CMU CL system, which was + * written at Carnegie Mellon University and released into the + * public domain. The software is in the public domain and is + * provided with absolutely no warranty. See the COPYING and CREDITS + * files for more information. + */ + +#include "sbcl.h" + #include #include #include +#ifndef LISP_FEATURE_WIN32 #include -#include +#endif +#include "runtime.h" #include #include #include +#ifndef LISP_FEATURE_WIN32 #include +#endif + +#ifdef LISP_FEATURE_MACH_EXCEPTION_HANDLER +#include +#include +#include +#endif -#include "sbcl.h" #include "runtime.h" -#include "validate.h" /* for CONTROL_STACK_SIZE etc */ -#include "alloc.h" +#include "validate.h" /* for BINDING_STACK_SIZE etc */ #include "thread.h" #include "arch.h" #include "target-arch-os.h" @@ -21,41 +42,74 @@ #include "genesis/cons.h" #include "genesis/fdefn.h" #include "interr.h" /* for lose() */ +#include "alloc.h" #include "gc-internal.h" +#include "cpputil.h" +#include "pseudo-atomic.h" +#include "interrupt.h" +#include "lispregs.h" + +#if defined(LISP_FEATURE_WIN32) && defined(LISP_FEATURE_SB_THREAD) +# define IMMEDIATE_POST_MORTEM +#endif + +#if defined(LISP_FEATURE_DARWIN) && defined(LISP_FEATURE_SB_THREAD) +#define DELAY_THREAD_POST_MORTEM 5 +#define LOCK_CREATE_THREAD +#endif -#define ALIEN_STACK_SIZE (1*1024*1024) /* 1Mb size chosen at random */ +#ifdef LISP_FEATURE_FREEBSD +#define CREATE_CLEANUP_THREAD +#define LOCK_CREATE_THREAD +#endif -struct freeable_stack { +#ifdef LISP_FEATURE_SB_THREAD +struct thread_post_mortem { +#ifdef DELAY_THREAD_POST_MORTEM + struct thread_post_mortem *next; +#endif os_thread_t os_thread; - os_vm_address_t stack; + pthread_attr_t *os_attr; + os_vm_address_t os_address; }; -static struct freeable_stack * volatile freeable_stack = 0; +#ifdef DELAY_THREAD_POST_MORTEM +static int pending_thread_post_mortem_count = 0; +pthread_mutex_t thread_post_mortem_lock = PTHREAD_MUTEX_INITIALIZER; +#endif +static struct thread_post_mortem * volatile pending_thread_post_mortem = 0; +#endif -int dynamic_values_bytes=4096*sizeof(lispobj); /* same for all threads */ -struct thread * volatile all_threads; +int dynamic_values_bytes=TLS_SIZE*sizeof(lispobj); /* same for all threads */ +struct thread *all_threads; extern struct interrupt_data * global_interrupt_data; -extern int linux_no_threads_p; #ifdef LISP_FEATURE_SB_THREAD - pthread_mutex_t all_threads_lock = PTHREAD_MUTEX_INITIALIZER; - -#if defined(LISP_FEATURE_X86) || defined(LISP_FEATURE_X86_64) -extern lispobj call_into_lisp_first_time(lispobj fun, lispobj *args, int nargs); +#ifdef LOCK_CREATE_THREAD +static pthread_mutex_t create_thread_lock = PTHREAD_MUTEX_INITIALIZER; +#endif +#ifdef LISP_FEATURE_GCC_TLS +__thread struct thread *current_thread; +#endif +pthread_key_t lisp_thread = 0; #endif +#if defined(LISP_FEATURE_X86) || defined(LISP_FEATURE_X86_64) +extern lispobj call_into_lisp_first_time(lispobj fun, lispobj *args, int nargs) +# ifdef LISP_FEATURE_X86_64 + __attribute__((sysv_abi)) +# endif + ; #endif static void link_thread(struct thread *th) { - th->os_thread=thread_self(); if (all_threads) all_threads->prev=th; th->next=all_threads; th->prev=0; all_threads=th; - protect_control_stack_guard_page(1); } #ifdef LISP_FEATURE_SB_THREAD @@ -69,7 +123,83 @@ unlink_thread(struct thread *th) if (th->next) th->next->prev = th->prev; } -#endif + +#ifndef LISP_FEATURE_SB_SAFEPOINT +/* Only access thread state with blockables blocked. */ +lispobj +thread_state(struct thread *thread) +{ + lispobj state; + sigset_t old; + block_blockable_signals(NULL, &old); + os_sem_wait(thread->state_sem, "thread_state"); + state = thread->state; + os_sem_post(thread->state_sem, "thread_state"); + thread_sigmask(SIG_SETMASK, &old, NULL); + return state; +} + +void +set_thread_state(struct thread *thread, lispobj state) +{ + int i, waitcount = 0; + sigset_t old; + block_blockable_signals(NULL, &old); + os_sem_wait(thread->state_sem, "set_thread_state"); + if (thread->state != state) { + if ((STATE_STOPPED==state) || + (STATE_DEAD==state)) { + waitcount = thread->state_not_running_waitcount; + thread->state_not_running_waitcount = 0; + for (i=0; istate_not_running_sem, "set_thread_state (not running)"); + } + if ((STATE_RUNNING==state) || + (STATE_DEAD==state)) { + waitcount = thread->state_not_stopped_waitcount; + thread->state_not_stopped_waitcount = 0; + for (i=0; istate_not_stopped_sem, "set_thread_state (not stopped)"); + } + thread->state = state; + } + os_sem_post(thread->state_sem, "set_thread_state"); + thread_sigmask(SIG_SETMASK, &old, NULL); +} + +void +wait_for_thread_state_change(struct thread *thread, lispobj state) +{ + sigset_t old; + os_sem_t *wait_sem; + block_blockable_signals(NULL, &old); + start: + os_sem_wait(thread->state_sem, "wait_for_thread_state_change"); + if (thread->state == state) { + switch (state) { + case STATE_RUNNING: + wait_sem = thread->state_not_running_sem; + thread->state_not_running_waitcount++; + break; + case STATE_STOPPED: + wait_sem = thread->state_not_stopped_sem; + thread->state_not_stopped_waitcount++; + break; + default: + lose("Invalid state in wait_for_thread_state_change: "OBJ_FMTX"\n", state); + } + } else { + wait_sem = NULL; + } + os_sem_post(thread->state_sem, "wait_for_thread_state_change"); + if (wait_sem) { + os_sem_wait(wait_sem, "wait_for_thread_state_change"); + goto start; + } + thread_sigmask(SIG_SETMASK, &old, NULL); +} +#endif /* sb-safepoint */ +#endif /* sb-thread */ static int initial_thread_trampoline(struct thread *th) @@ -78,10 +208,28 @@ initial_thread_trampoline(struct thread *th) #if defined(LISP_FEATURE_X86) || defined(LISP_FEATURE_X86_64) lispobj *args = NULL; #endif +#ifdef LISP_FEATURE_SB_THREAD + pthread_setspecific(lisp_thread, (void *)1); +#endif +#if defined(THREADS_USING_GCSIGNAL) && defined(LISP_FEATURE_PPC) + /* SIG_STOP_FOR_GC defaults to blocked on PPC? */ + unblock_gc_signals(0,0); +#endif function = th->no_tls_value_marker; th->no_tls_value_marker = NO_TLS_VALUE_MARKER_WIDETAG; if(arch_os_thread_init(th)==0) return 1; link_thread(th); + th->os_thread=thread_self(); +#ifndef LISP_FEATURE_WIN32 + protect_control_stack_hard_guard_page(1, NULL); +#endif + protect_binding_stack_hard_guard_page(1, NULL); + protect_alien_stack_hard_guard_page(1, NULL); +#ifndef LISP_FEATURE_WIN32 + protect_control_stack_guard_page(1, NULL); +#endif + protect_binding_stack_guard_page(1, NULL); + protect_alien_stack_guard_page(1, NULL); #if defined(LISP_FEATURE_X86) || defined(LISP_FEATURE_X86_64) return call_into_lisp_first_time(function,args,0); @@ -90,93 +238,352 @@ initial_thread_trampoline(struct thread *th) #endif } -#define THREAD_STRUCT_SIZE (THREAD_CONTROL_STACK_SIZE + BINDING_STACK_SIZE + \ - ALIEN_STACK_SIZE + dynamic_values_bytes + \ - 32 * SIGSTKSZ) - #ifdef LISP_FEATURE_SB_THREAD +# if defined(IMMEDIATE_POST_MORTEM) + +/* + * If this feature is set, we are running on a stack managed by the OS, + * and no fancy delays are required for anything. Just do it. + */ +static void +schedule_thread_post_mortem(struct thread *corpse) +{ + pthread_detach(pthread_self()); + gc_assert(!pthread_attr_destroy(corpse->os_attr)); + free(corpse->os_attr); +#if defined(LISP_FEATURE_WIN32) + os_invalidate_free(corpse->os_address, THREAD_STRUCT_SIZE); +#else + os_invalidate(corpse->os_address, THREAD_STRUCT_SIZE); +#endif +} + +# else + +/* THREAD POST MORTEM CLEANUP + * + * Memory allocated for the thread stacks cannot be reclaimed while + * the thread is still alive, so we need a mechanism for post mortem + * cleanups. FIXME: We actually have three, for historical reasons as + * the saying goes. Do we really need three? Nikodemus guesses that + * not anymore, now that we properly call pthread_attr_destroy before + * freeing the stack. */ + +static struct thread_post_mortem * +plan_thread_post_mortem(struct thread *corpse) +{ + if (corpse) { + struct thread_post_mortem *post_mortem = malloc(sizeof(struct thread_post_mortem)); + gc_assert(post_mortem); + post_mortem->os_thread = corpse->os_thread; + post_mortem->os_attr = corpse->os_attr; + post_mortem->os_address = corpse->os_address; +#ifdef DELAY_THREAD_POST_MORTEM + post_mortem->next = NULL; +#endif + return post_mortem; + } else { + /* FIXME: When does this happen? */ + return NULL; + } +} + static void -free_thread_stack_later(struct thread *thread_to_be_cleaned_up) +perform_thread_post_mortem(struct thread_post_mortem *post_mortem) { - struct freeable_stack *new_freeable_stack = 0; - if (thread_to_be_cleaned_up) { - new_freeable_stack = (struct freeable_stack *) - os_validate(0, sizeof(struct freeable_stack)); - new_freeable_stack->os_thread = thread_to_be_cleaned_up->os_thread; - new_freeable_stack->stack = (os_vm_address_t) - thread_to_be_cleaned_up->control_stack_start; +#ifdef CREATE_POST_MORTEM_THREAD + pthread_detach(pthread_self()); +#endif + if (post_mortem) { + gc_assert(!pthread_join(post_mortem->os_thread, NULL)); + gc_assert(!pthread_attr_destroy(post_mortem->os_attr)); + free(post_mortem->os_attr); + os_invalidate(post_mortem->os_address, THREAD_STRUCT_SIZE); + free(post_mortem); } - new_freeable_stack = (struct freeable_stack *) - swap_lispobjs((lispobj *)(void *)&freeable_stack, - (lispobj)new_freeable_stack); - if (new_freeable_stack) { - FSHOW((stderr,"/reaping %lu\n", new_freeable_stack->os_thread)); - /* Under NPTL pthread_join really waits until the thread - * exists and the stack can be safely freed. This is sadly not - * mandated by the pthread spec. */ - gc_assert(pthread_join(new_freeable_stack->os_thread, NULL) == 0); - os_invalidate(new_freeable_stack->stack, THREAD_STRUCT_SIZE); - os_invalidate((os_vm_address_t) new_freeable_stack, - sizeof(struct freeable_stack)); +} + +static void +schedule_thread_post_mortem(struct thread *corpse) +{ + struct thread_post_mortem *post_mortem = NULL; + if (corpse) { + post_mortem = plan_thread_post_mortem(corpse); + +#ifdef DELAY_THREAD_POST_MORTEM + pthread_mutex_lock(&thread_post_mortem_lock); + /* First stick the new post mortem to the end of the queue. */ + if (pending_thread_post_mortem) { + struct thread_post_mortem *next = pending_thread_post_mortem; + while (next->next) { + next = next->next; + } + next->next = post_mortem; + } else { + pending_thread_post_mortem = post_mortem; + } + /* Then, if there are enough things in the queue, clean up one + * from the head -- or increment the count, and null out the + * post_mortem we have. */ + if (pending_thread_post_mortem_count > DELAY_THREAD_POST_MORTEM) { + post_mortem = pending_thread_post_mortem; + pending_thread_post_mortem = post_mortem->next; + } else { + pending_thread_post_mortem_count++; + post_mortem = NULL; + } + pthread_mutex_unlock(&thread_post_mortem_lock); + /* Finally run, the cleanup, if any. */ + perform_thread_post_mortem(post_mortem); +#elif defined(CREATE_POST_MORTEM_THREAD) + gc_assert(!pthread_create(&thread, NULL, perform_thread_post_mortem, post_mortem)); +#else + post_mortem = (struct thread_post_mortem *) + swap_lispobjs((lispobj *)(void *)&pending_thread_post_mortem, + (lispobj)post_mortem); + perform_thread_post_mortem(post_mortem); +#endif } } -/* this is the first thing that runs in the child (which is why the - * silly calling convention). Basically it calls the user's requested - * lisp function after doing arch_os_thread_init and whatever other - * bookkeeping needs to be done - */ -int -new_thread_trampoline(struct thread *th) +# endif /* !IMMEDIATE_POST_MORTEM */ + +/* Note: scribble must be stack-allocated */ +static void +init_new_thread(struct thread *th, init_thread_data *scribble, int guardp) { - lispobj function; - int result; - FSHOW((stderr,"/creating thread %lu\n", thread_self())); - function = th->no_tls_value_marker; - th->no_tls_value_marker = NO_TLS_VALUE_MARKER_WIDETAG; + int lock_ret; + + pthread_setspecific(lisp_thread, (void *)1); if(arch_os_thread_init(th)==0) { /* FIXME: handle error */ lose("arch_os_thread_init failed\n"); } + th->os_thread=thread_self(); + if (guardp) + protect_control_stack_guard_page(1, NULL); + protect_binding_stack_guard_page(1, NULL); + protect_alien_stack_guard_page(1, NULL); /* Since GC can only know about this thread from the all_threads - * list and we're just adding this thread to it there is no danger - * of deadlocking even with SIG_STOP_FOR_GC blocked. */ - pthread_mutex_lock(&all_threads_lock); + * list and we're just adding this thread to it, there is no + * danger of deadlocking even with SIG_STOP_FOR_GC blocked (which + * it is not). */ +#ifdef LISP_FEATURE_SB_SAFEPOINT + *th->csp_around_foreign_call = (lispobj)scribble; +#endif + lock_ret = pthread_mutex_lock(&all_threads_lock); + gc_assert(lock_ret == 0); link_thread(th); - pthread_mutex_unlock(&all_threads_lock); + lock_ret = pthread_mutex_unlock(&all_threads_lock); + gc_assert(lock_ret == 0); - result = funcall0(function); - th->state=STATE_DEAD; + /* Kludge: Changed the order of some steps between the safepoint/ + * non-safepoint versions of this code. Can we unify this more? + */ +#ifdef LISP_FEATURE_SB_SAFEPOINT + gc_state_lock(); + gc_state_wait(GC_NONE); + gc_state_unlock(); + push_gcing_safety(&scribble->safety); +#endif +} + +static void +undo_init_new_thread(struct thread *th, init_thread_data *scribble) +{ + int lock_ret; + + /* Kludge: Changed the order of some steps between the safepoint/ + * non-safepoint versions of this code. Can we unify this more? + */ +#ifdef LISP_FEATURE_SB_SAFEPOINT + block_blockable_signals(0, 0); + gc_alloc_update_page_tables(BOXED_PAGE_FLAG, &th->alloc_region); +#if defined(LISP_FEATURE_SB_SAFEPOINT_STRICTLY) && !defined(LISP_FEATURE_WIN32) + gc_alloc_update_page_tables(BOXED_PAGE_FLAG, &th->sprof_alloc_region); +#endif + pop_gcing_safety(&scribble->safety); + lock_ret = pthread_mutex_lock(&all_threads_lock); + gc_assert(lock_ret == 0); + unlink_thread(th); + lock_ret = pthread_mutex_unlock(&all_threads_lock); + gc_assert(lock_ret == 0); +#else + /* Block GC */ + block_blockable_signals(0, 0); + set_thread_state(th, STATE_DEAD); /* SIG_STOP_FOR_GC is blocked and GC might be waiting for this * thread, but since we are already dead it won't wait long. */ - pthread_mutex_lock(&all_threads_lock); + lock_ret = pthread_mutex_lock(&all_threads_lock); + gc_assert(lock_ret == 0); + + gc_alloc_update_page_tables(BOXED_PAGE_FLAG, &th->alloc_region); +#if defined(LISP_FEATURE_SB_SAFEPOINT_STRICTLY) && !defined(LISP_FEATURE_WIN32) + gc_alloc_update_page_tables(BOXED_PAGE_FLAG, &th->sprof_alloc_region); +#endif unlink_thread(th); pthread_mutex_unlock(&all_threads_lock); + gc_assert(lock_ret == 0); +#endif - gc_alloc_update_page_tables(0, &th->alloc_region); if(th->tls_cookie>=0) arch_os_thread_cleanup(th); +#ifndef LISP_FEATURE_SB_SAFEPOINT + os_sem_destroy(th->state_sem); + os_sem_destroy(th->state_not_running_sem); + os_sem_destroy(th->state_not_stopped_sem); +#endif + +#if defined(LISP_FEATURE_WIN32) + free((os_vm_address_t)th->interrupt_data); +#else os_invalidate((os_vm_address_t)th->interrupt_data, (sizeof (struct interrupt_data))); - free_thread_stack_later(th); +#endif + +#ifdef LISP_FEATURE_MACH_EXCEPTION_HANDLER + mach_lisp_thread_destroy(th); +#endif + +#if defined(LISP_FEATURE_WIN32) + int i; + for (i = 0; i< + (int) (sizeof(th->private_events.events)/ + sizeof(th->private_events.events[0])); ++i) { + CloseHandle(th->private_events.events[i]); + } + TlsSetValue(OUR_TLS_INDEX,NULL); +#endif + + /* Undo the association of the current pthread to its `struct thread', + * such that we can call arch_os_get_current_thread() later in this + * thread and cleanly get back NULL. */ +#ifdef LISP_FEATURE_GCC_TLS + current_thread = NULL; +#else + pthread_setspecific(specials, NULL); +#endif + + schedule_thread_post_mortem(th); +} + +/* this is the first thing that runs in the child (which is why the + * silly calling convention). Basically it calls the user's requested + * lisp function after doing arch_os_thread_init and whatever other + * bookkeeping needs to be done + */ +int +new_thread_trampoline(struct thread *th) +{ + int result; + init_thread_data scribble; + + FSHOW((stderr,"/creating thread %lu\n", thread_self())); + check_deferrables_blocked_or_lose(0); +#ifndef LISP_FEATURE_SB_SAFEPOINT + check_gc_signals_unblocked_or_lose(0); +#endif + + lispobj function = th->no_tls_value_marker; + th->no_tls_value_marker = NO_TLS_VALUE_MARKER_WIDETAG; + init_new_thread(th, &scribble, 1); + result = funcall0(function); + undo_init_new_thread(th, &scribble); + FSHOW((stderr,"/exiting thread %lu\n", thread_self())); return result; } +# ifdef LISP_FEATURE_SB_SAFEPOINT +static struct thread *create_thread_struct(lispobj); + +void +attach_os_thread(init_thread_data *scribble) +{ + os_thread_t os = pthread_self(); + odxprint(misc, "attach_os_thread: attaching to %p", os); + + struct thread *th = create_thread_struct(NIL); + block_deferrable_signals(0, &scribble->oldset); + th->no_tls_value_marker = NO_TLS_VALUE_MARKER_WIDETAG; + /* We don't actually want a pthread_attr here, but rather than add + * `if's to the post-mostem, let's just keep that code happy by + * keeping it initialized: */ + pthread_attr_init(th->os_attr); + +#ifndef LISP_FEATURE_WIN32 + /* On windows, arch_os_thread_init will take care of finding the + * stack. */ + pthread_attr_t attr; + int pthread_getattr_np(pthread_t, pthread_attr_t *); + pthread_getattr_np(os, &attr); + void *stack_addr; + size_t stack_size; + pthread_attr_getstack(&attr, &stack_addr, &stack_size); + th->control_stack_start = stack_addr; + th->control_stack_end = (void *) (((uintptr_t) stack_addr) + stack_size); +#endif + + init_new_thread(th, scribble, 0); + + /* We will be calling into Lisp soon, and the functions being called + * recklessly ignore the comment in target-thread which says that we + * must be careful to not cause GC while initializing a new thread. + * Since we first need to create a fresh thread object, it's really + * tempting to just perform such unsafe allocation though. So let's + * at least try to suppress GC before consing, and hope that it + * works: */ + bind_variable(GC_INHIBIT, T, th); + + uword_t stacksize + = (uword_t) th->control_stack_end - (uword_t) th->control_stack_start; + odxprint(misc, "attach_os_thread: attached %p as %p (0x%lx bytes stack)", + os, th, (long) stacksize); +} + +void +detach_os_thread(init_thread_data *scribble) +{ + struct thread *th = arch_os_get_current_thread(); + odxprint(misc, "detach_os_thread: detaching"); + + undo_init_new_thread(th, scribble); + + odxprint(misc, "deattach_os_thread: detached"); + pthread_setspecific(lisp_thread, (void *)0); + thread_sigmask(SIG_SETMASK, &scribble->oldset, 0); +} +# endif /* safepoint */ + #endif /* LISP_FEATURE_SB_THREAD */ static void free_thread_struct(struct thread *th) { +#if defined(LISP_FEATURE_WIN32) + if (th->interrupt_data) { + os_invalidate_free((os_vm_address_t) th->interrupt_data, + (sizeof (struct interrupt_data))); + } + os_invalidate_free((os_vm_address_t) th->os_address, + THREAD_STRUCT_SIZE); +#else if (th->interrupt_data) os_invalidate((os_vm_address_t) th->interrupt_data, (sizeof (struct interrupt_data))); - os_invalidate((os_vm_address_t) th->control_stack_start, + os_invalidate((os_vm_address_t) th->os_address, THREAD_STRUCT_SIZE); +#endif } +#ifdef LISP_FEATURE_SB_THREAD +/* FIXME: should be MAX_INTERRUPTS -1 ? */ +const unsigned int tls_index_start = + MAX_INTERRUPTS + sizeof(struct thread)/sizeof(lispobj); +#endif + /* this is called from any other thread to create the new one, and * initialize all parts of it that can be initialized from another * thread @@ -187,75 +594,131 @@ create_thread_struct(lispobj initial_function) { union per_thread_data *per_thread; struct thread *th=0; /* subdue gcc */ void *spaces=0; + void *aligned_spaces=0; +#if defined(LISP_FEATURE_SB_THREAD) || defined(LISP_FEATURE_WIN32) + unsigned int i; +#endif - /* may as well allocate all the spaces at once: it saves us from + /* May as well allocate all the spaces at once: it saves us from * having to decide what to do if only some of the allocations - * succeed */ + * succeed. SPACES must be appropriately aligned, since the GC + * expects the control stack to start at a page boundary -- and + * the OS may have even more rigorous requirements. We can't rely + * on the alignment passed from os_validate, since that might + * assume the current (e.g. 4k) pagesize, while we calculate with + * the biggest (e.g. 64k) pagesize allowed by the ABI. */ spaces=os_validate(0, THREAD_STRUCT_SIZE); if(!spaces) - return NULL; - per_thread=(union per_thread_data *) - (spaces+ - THREAD_CONTROL_STACK_SIZE+ + return NULL; + /* Aligning up is safe as THREAD_STRUCT_SIZE has + * THREAD_ALIGNMENT_BYTES padding. */ + aligned_spaces = (void *)((((uword_t)(char *)spaces) + + THREAD_ALIGNMENT_BYTES-1) + &~(uword_t)(THREAD_ALIGNMENT_BYTES-1)); + void* csp_page= + (aligned_spaces+ + thread_control_stack_size+ BINDING_STACK_SIZE+ ALIEN_STACK_SIZE); + per_thread=(union per_thread_data *) + (csp_page + THREAD_CSP_PAGE_SIZE); + struct nonpointer_thread_data *nonpointer_data + = (void *) &per_thread->dynamic_values[TLS_SIZE]; - if(all_threads) { - memcpy(per_thread,arch_os_get_current_thread(), - dynamic_values_bytes); - } else { #ifdef LISP_FEATURE_SB_THREAD - int i; - for(i=0;i<(dynamic_values_bytes/sizeof(lispobj));i++) - per_thread->dynamic_values[i]=NO_TLS_VALUE_MARKER_WIDETAG; + for(i = 0; i < (dynamic_values_bytes / sizeof(lispobj)); i++) + per_thread->dynamic_values[i] = NO_TLS_VALUE_MARKER_WIDETAG; + if (all_threads == 0) { if(SymbolValue(FREE_TLS_INDEX,0)==UNBOUND_MARKER_WIDETAG) { - SetSymbolValue - (FREE_TLS_INDEX, - make_fixnum(MAX_INTERRUPTS+ - sizeof(struct thread)/sizeof(lispobj)), - 0); + SetSymbolValue(FREE_TLS_INDEX,tls_index_start << WORD_SHIFT,0); SetSymbolValue(TLS_INDEX_LOCK,make_fixnum(0),0); } #define STATIC_TLS_INIT(sym,field) \ ((struct symbol *)(sym-OTHER_POINTER_LOWTAG))->tls_index= \ - make_fixnum(THREAD_SLOT_OFFSET_WORDS(field)) + (THREAD_SLOT_OFFSET_WORDS(field) << WORD_SHIFT) STATIC_TLS_INIT(BINDING_STACK_START,binding_stack_start); +#ifdef BINDING_STACK_POINTER STATIC_TLS_INIT(BINDING_STACK_POINTER,binding_stack_pointer); +#endif STATIC_TLS_INIT(CONTROL_STACK_START,control_stack_start); STATIC_TLS_INIT(CONTROL_STACK_END,control_stack_end); +#ifdef ALIEN_STACK STATIC_TLS_INIT(ALIEN_STACK,alien_stack_pointer); +#endif #if defined(LISP_FEATURE_X86) || defined (LISP_FEATURE_X86_64) - STATIC_TLS_INIT(PSEUDO_ATOMIC_ATOMIC,pseudo_atomic_atomic); - STATIC_TLS_INIT(PSEUDO_ATOMIC_INTERRUPTED,pseudo_atomic_interrupted); + STATIC_TLS_INIT(PSEUDO_ATOMIC_BITS,pseudo_atomic_bits); #endif #undef STATIC_TLS_INIT -#endif } +#endif th=&per_thread->thread; - th->control_stack_start = spaces; + th->os_address = spaces; + th->control_stack_start = aligned_spaces; th->binding_stack_start= - (lispobj*)((void*)th->control_stack_start+THREAD_CONTROL_STACK_SIZE); + (lispobj*)((void*)th->control_stack_start+thread_control_stack_size); th->control_stack_end = th->binding_stack_start; + th->control_stack_guard_page_protected = T; th->alien_stack_start= (lispobj*)((void*)th->binding_stack_start+BINDING_STACK_SIZE); - th->binding_stack_pointer=th->binding_stack_start; + set_binding_stack_pointer(th,th->binding_stack_start); th->this=th; th->os_thread=0; + +#ifdef LISP_FEATURE_SB_SAFEPOINT +# ifdef LISP_FEATURE_WIN32 + th->carried_base_pointer = 0; +# endif +# ifdef LISP_FEATURE_C_STACK_IS_CONTROL_STACK + th->pc_around_foreign_call = 0; +# endif + th->csp_around_foreign_call = csp_page; +#endif + +#ifdef LISP_FEATURE_SB_THREAD + /* Contrary to the "allocate all the spaces at once" comment above, + * the os_attr is allocated separately. We cannot put it into the + * nonpointer data, because it's used for post_mortem and freed + * separately */ + th->os_attr=malloc(sizeof(pthread_attr_t)); + th->nonpointer_data = nonpointer_data; +# ifndef LISP_FEATURE_SB_SAFEPOINT + th->state_sem=&nonpointer_data->state_sem; + th->state_not_running_sem=&nonpointer_data->state_not_running_sem; + th->state_not_stopped_sem=&nonpointer_data->state_not_stopped_sem; + os_sem_init(th->state_sem, 1); + os_sem_init(th->state_not_running_sem, 0); + os_sem_init(th->state_not_stopped_sem, 0); +# endif + th->state_not_running_waitcount = 0; + th->state_not_stopped_waitcount = 0; +#endif th->state=STATE_RUNNING; -#ifdef LISP_FEATURE_STACK_GROWS_DOWNWARD_NOT_UPWARD +#ifdef ALIEN_STACK_GROWS_DOWNWARD th->alien_stack_pointer=((void *)th->alien_stack_start + ALIEN_STACK_SIZE-N_WORD_BYTES); #else th->alien_stack_pointer=((void *)th->alien_stack_start); #endif -#if defined(LISP_FEATURE_X86) || defined (LISP_FEATURE_X86_64) - th->pseudo_atomic_interrupted=0; - th->pseudo_atomic_atomic=0; +#if defined(LISP_FEATURE_X86) || defined (LISP_FEATURE_X86_64) || defined(LISP_FEATURE_SB_THREAD) + th->pseudo_atomic_bits=0; #endif #ifdef LISP_FEATURE_GENCGC gc_set_region_empty(&th->alloc_region); +# if defined(LISP_FEATURE_SB_SAFEPOINT_STRICTLY) && !defined(LISP_FEATURE_WIN32) + gc_set_region_empty(&th->sprof_alloc_region); +# endif +#endif +#ifdef LISP_FEATURE_SB_THREAD + /* This parallels the same logic in globals.c for the + * single-threaded foreign_function_call_active, KLUDGE and + * all. */ +#if defined(LISP_FEATURE_X86) || defined(LISP_FEATURE_X86_64) + th->foreign_function_call_active = 0; +#else + th->foreign_function_call_active = 1; +#endif #endif #ifndef LISP_FEATURE_SB_THREAD @@ -269,13 +732,8 @@ create_thread_struct(lispobj initial_function) { SetSymbolValue(CONTROL_STACK_START,(lispobj)th->control_stack_start,th); SetSymbolValue(CONTROL_STACK_END,(lispobj)th->control_stack_end,th); #if defined(LISP_FEATURE_X86) || defined (LISP_FEATURE_X86_64) - SetSymbolValue(BINDING_STACK_POINTER,(lispobj)th->binding_stack_pointer,th); SetSymbolValue(ALIEN_STACK,(lispobj)th->alien_stack_pointer,th); - SetSymbolValue(PSEUDO_ATOMIC_ATOMIC,(lispobj)th->pseudo_atomic_atomic,th); - SetSymbolValue(PSEUDO_ATOMIC_INTERRUPTED,th->pseudo_atomic_interrupted,th); -#else - current_binding_stack_pointer=th->binding_stack_pointer; - current_control_stack_pointer=th->control_stack_start; + SetSymbolValue(PSEUDO_ATOMIC_BITS,(lispobj)th->pseudo_atomic_bits,th); #endif #endif bind_variable(CURRENT_CATCH_BLOCK,make_fixnum(0),th); @@ -283,27 +741,64 @@ create_thread_struct(lispobj initial_function) { bind_variable(FREE_INTERRUPT_CONTEXT_INDEX,make_fixnum(0),th); bind_variable(INTERRUPT_PENDING, NIL,th); bind_variable(INTERRUPTS_ENABLED,T,th); + bind_variable(ALLOW_WITH_INTERRUPTS,T,th); bind_variable(GC_PENDING,NIL,th); + bind_variable(ALLOC_SIGNAL,NIL,th); +#ifdef PINNED_OBJECTS + bind_variable(PINNED_OBJECTS,NIL,th); +#endif #ifdef LISP_FEATURE_SB_THREAD bind_variable(STOP_FOR_GC_PENDING,NIL,th); #endif +#if defined(LISP_FEATURE_SB_SAFEPOINT) + bind_variable(GC_SAFE,NIL,th); + bind_variable(IN_SAFEPOINT,NIL,th); +#endif +#ifdef LISP_FEATURE_SB_THRUPTION + bind_variable(THRUPTION_PENDING,NIL,th); + bind_variable(RESTART_CLUSTERS,NIL,th); +#endif +#ifndef LISP_FEATURE_C_STACK_IS_CONTROL_STACK + access_control_stack_pointer(th)=th->control_stack_start; +#endif +#if defined(LISP_FEATURE_WIN32) + th->interrupt_data = (struct interrupt_data *) + calloc((sizeof (struct interrupt_data)),1); +#else th->interrupt_data = (struct interrupt_data *) os_validate(0,(sizeof (struct interrupt_data))); +#endif if (!th->interrupt_data) { free_thread_struct(th); return 0; } th->interrupt_data->pending_handler = 0; + th->interrupt_data->gc_blocked_deferrables = 0; +#ifdef GENCGC_IS_PRECISE + th->interrupt_data->allocation_trap_context = 0; +#endif th->no_tls_value_marker=initial_function; + +#if defined(LISP_FEATURE_WIN32) + for (i = 0; iprivate_events.events)/ + sizeof(th->private_events.events[0]); ++i) { + th->private_events.events[i] = CreateEvent(NULL,FALSE,FALSE,NULL); + } + th->synchronous_io_handle_and_flag = 0; +#endif + th->stepping = NIL; return th; } void create_initial_thread(lispobj initial_function) { struct thread *th=create_thread_struct(initial_function); +#ifdef LISP_FEATURE_SB_THREAD + pthread_key_create(&lisp_thread, 0); +#endif if(th) { initial_thread_trampoline(th); /* no return */ - } else lose("can't create initial thread"); + } else lose("can't create initial thread\n"); } #ifdef LISP_FEATURE_SB_THREAD @@ -317,75 +812,75 @@ boolean create_os_thread(struct thread *th,os_thread_t *kid_tid) { /* The new thread inherits the restrictive signal mask set here, * and enables signals again when it is set up properly. */ - pthread_attr_t attr; - sigset_t newset,oldset; + sigset_t oldset; boolean r=1; - sigemptyset(&newset); + int retcode = 0, initcode; + + FSHOW_SIGNAL((stderr,"/create_os_thread: creating new thread\n")); + /* Blocking deferrable signals is enough, no need to block * SIG_STOP_FOR_GC because the child process is not linked onto * all_threads until it's ready. */ - sigaddset_deferrable(&newset); - thread_sigmask(SIG_BLOCK, &newset, &oldset); - - if((pthread_attr_init(&attr)) || - (pthread_attr_setstack(&attr,th->control_stack_start, - THREAD_CONTROL_STACK_SIZE-16)) || - (pthread_create - (kid_tid,&attr,(void *(*)(void *))new_thread_trampoline,th))) + block_deferrable_signals(0, &oldset); + +#ifdef LOCK_CREATE_THREAD + retcode = pthread_mutex_lock(&create_thread_lock); + gc_assert(retcode == 0); + FSHOW_SIGNAL((stderr,"/create_os_thread: got lock\n")); +#endif + + if((initcode = pthread_attr_init(th->os_attr)) || + /* call_into_lisp_first_time switches the stack for the initial + * thread. For the others, we use this. */ +#if defined(LISP_FEATURE_WIN32) + (pthread_attr_setstacksize(th->os_attr, thread_control_stack_size)) || +#else +# if defined(LISP_FEATURE_C_STACK_IS_CONTROL_STACK) + (pthread_attr_setstack(th->os_attr,th->control_stack_start, + thread_control_stack_size)) || +# else + (pthread_attr_setstack(th->os_attr,th->alien_stack_start, + ALIEN_STACK_SIZE)) || +# endif +#endif + (retcode = pthread_create + (kid_tid,th->os_attr,(void *(*)(void *))new_thread_trampoline,th))) { + FSHOW_SIGNAL((stderr, "init = %d\n", initcode)); + FSHOW_SIGNAL((stderr, "pthread_create returned %d, errno %d\n", + retcode, errno)); + if(retcode < 0) { + perror("create_os_thread"); + } r=0; + } + +#ifdef LOCK_CREATE_THREAD + retcode = pthread_mutex_unlock(&create_thread_lock); + gc_assert(retcode == 0); + FSHOW_SIGNAL((stderr,"/create_os_thread: released lock\n")); +#endif thread_sigmask(SIG_SETMASK,&oldset,0); return r; } os_thread_t create_thread(lispobj initial_function) { - struct thread *th; - os_thread_t kid_tid; + struct thread *th, *thread = arch_os_get_current_thread(); + os_thread_t kid_tid = 0; - if(linux_no_threads_p) return 0; + /* Must defend against async unwinds. */ + if (SymbolValue(INTERRUPTS_ENABLED, thread) != NIL) + lose("create_thread is not safe when interrupts are enabled.\n"); - th=create_thread_struct(initial_function); - if(th==0) return 0; - - if (create_os_thread(th,&kid_tid)) { - return kid_tid; - } else { + /* Assuming that a fresh thread struct has no lisp objects in it, + * linking it to all_threads can be left to the thread itself + * without fear of gc lossage. initial_function violates this + * assumption and must stay pinned until the child starts up. */ + th = create_thread_struct(initial_function); + if (th && !create_os_thread(th,&kid_tid)) { free_thread_struct(th); - return 0; - } -} - -/* Send the signo to os_thread, retry if the rt signal queue is - * full. */ -static int kill_thread_safely(os_thread_t os_thread, int signo) -{ - int r; - /* The man page does not mention EAGAIN as a valid return value - * for either pthread_kill or kill. But that's theory, this is - * practice. By waiting here we assume that the delivery of this - * signal is not necessary for the delivery of the signals in the - * queue. In other words, we _assume_ there are no deadlocks. */ - while ((r=pthread_kill(os_thread,signo))==EAGAIN) { - /* wait a bit then try again in the hope of the rt signal - * queue not being full */ - FSHOW_SIGNAL((stderr,"/rt signal queue full\n")); - /* FIXME: some kind of backoff (random, exponential) would be - * nice. */ - sleep(1); - } - return r; -} - -int signal_interrupt_thread(os_thread_t os_thread) -{ - int status = kill_thread_safely(os_thread, SIG_INTERRUPT_THREAD); - if (status == 0) { - return 0; - } else if (status == ESRCH) { - return -1; - } else { - lose("cannot send SIG_INTERRUPT_THREAD to thread=%lu: %d, %s", - os_thread, status, strerror(status)); + kid_tid = 0; } + return kid_tid; } /* stopping the world is a two-stage process. From this thread we signal @@ -393,6 +888,10 @@ int signal_interrupt_thread(os_thread_t os_thread) * the usual pseudo-atomic checks (we don't want to stop a thread while * it's in the middle of allocation) then waits for another SIG_STOP_FOR_GC. */ +/* + * (With SB-SAFEPOINT, see the definitions in safepoint.c instead.) + */ +#ifndef LISP_FEATURE_SB_SAFEPOINT /* To avoid deadlocks when gc stops the world all clients of each * mutex must enable or disable SIG_STOP_FOR_GC for the duration of @@ -400,37 +899,51 @@ int signal_interrupt_thread(os_thread_t os_thread) void gc_stop_the_world() { struct thread *p,*th=arch_os_get_current_thread(); - int status; - FSHOW_SIGNAL((stderr,"/gc_stop_the_world:waiting on lock, thread=%lu\n", - th->os_thread)); + int status, lock_ret; +#ifdef LOCK_CREATE_THREAD + /* KLUDGE: Stopping the thread during pthread_create() causes deadlock + * on FreeBSD. */ + FSHOW_SIGNAL((stderr,"/gc_stop_the_world:waiting on create_thread_lock\n")); + lock_ret = pthread_mutex_lock(&create_thread_lock); + gc_assert(lock_ret == 0); + FSHOW_SIGNAL((stderr,"/gc_stop_the_world:got create_thread_lock\n")); +#endif + FSHOW_SIGNAL((stderr,"/gc_stop_the_world:waiting on lock\n")); /* keep threads from starting while the world is stopped. */ - pthread_mutex_lock(&all_threads_lock); \ - FSHOW_SIGNAL((stderr,"/gc_stop_the_world:got lock, thread=%lu\n", - th->os_thread)); + lock_ret = pthread_mutex_lock(&all_threads_lock); \ + gc_assert(lock_ret == 0); + + FSHOW_SIGNAL((stderr,"/gc_stop_the_world:got lock\n")); /* stop all other threads by sending them SIG_STOP_FOR_GC */ for(p=all_threads; p; p=p->next) { - if((p!=th) && ((p->state==STATE_RUNNING))) { - FSHOW_SIGNAL((stderr,"/gc_stop_the_world: suspending %lu\n", + gc_assert(p->os_thread != 0); + FSHOW_SIGNAL((stderr,"/gc_stop_the_world: thread=%lu, state=%x\n", + p->os_thread, thread_state(p))); + if((p!=th) && ((thread_state(p)==STATE_RUNNING))) { + FSHOW_SIGNAL((stderr,"/gc_stop_the_world: suspending thread %lu\n", p->os_thread)); - status=kill_thread_safely(p->os_thread,SIG_STOP_FOR_GC); + /* We already hold all_thread_lock, P can become DEAD but + * cannot exit, ergo it's safe to use pthread_kill. */ + status=pthread_kill(p->os_thread,SIG_STOP_FOR_GC); if (status==ESRCH) { /* This thread has exited. */ - gc_assert(p->state==STATE_DEAD); + gc_assert(thread_state(p)==STATE_DEAD); } else if (status) { - lose("cannot send suspend thread=%lu: %d, %s", + lose("cannot send suspend thread=%lu: %d, %s\n", p->os_thread,status,strerror(status)); } } } FSHOW_SIGNAL((stderr,"/gc_stop_the_world:signals sent\n")); - /* wait for the running threads to stop or finish */ - for(p=all_threads;p;) { - gc_assert(p->os_thread!=0); - if((p==th) || (p->state==STATE_SUSPENDED) || - (p->state==STATE_DEAD)) { - p=p->next; - } else { - sched_yield(); + for(p=all_threads;p;p=p->next) { + if (p!=th) { + FSHOW_SIGNAL + ((stderr, + "/gc_stop_the_world: waiting for thread=%lu: state=%x\n", + p->os_thread, thread_state(p))); + wait_for_thread_state_change(p, STATE_RUNNING); + if (p->state == STATE_RUNNING) + lose("/gc_stop_the_world: unexpected state"); } } FSHOW_SIGNAL((stderr,"/gc_stop_the_world:end\n")); @@ -439,7 +952,7 @@ void gc_stop_the_world() void gc_start_the_world() { struct thread *p,*th=arch_os_get_current_thread(); - int status; + int lock_ret; /* if a resumed thread creates a new thread before we're done with * this loop, the new thread will get consed on the front of * all_threads, but it won't have been stopped so won't need @@ -447,26 +960,144 @@ void gc_start_the_world() FSHOW_SIGNAL((stderr,"/gc_start_the_world:begin\n")); for(p=all_threads;p;p=p->next) { gc_assert(p->os_thread!=0); - if((p!=th) && (p->state!=STATE_DEAD)) { - if(p->state!=STATE_SUSPENDED) { - lose("gc_start_the_world: wrong thread state is %d\n", - fixnum_value(p->state)); - } - FSHOW_SIGNAL((stderr, "/gc_start_the_world: resuming %lu\n", - p->os_thread)); - p->state=STATE_RUNNING; - status=kill_thread_safely(p->os_thread,SIG_STOP_FOR_GC); - if (status) { - lose("cannot resume thread=%lu: %d, %s", - p->os_thread,status,strerror(status)); + if (p!=th) { + lispobj state = thread_state(p); + if (state != STATE_DEAD) { + if(state != STATE_STOPPED) { + lose("gc_start_the_world: wrong thread state is %d\n", + fixnum_value(state)); + } + FSHOW_SIGNAL((stderr, "/gc_start_the_world: resuming %lu\n", + p->os_thread)); + set_thread_state(p, STATE_RUNNING); } } } - /* If we waited here until all threads leave STATE_SUSPENDED, then - * SIG_STOP_FOR_GC wouldn't need to be a rt signal. That has some - * performance implications, but does away with the 'rt signal - * queue full' problem. */ - pthread_mutex_unlock(&all_threads_lock); \ + + lock_ret = pthread_mutex_unlock(&all_threads_lock); + gc_assert(lock_ret == 0); +#ifdef LOCK_CREATE_THREAD + lock_ret = pthread_mutex_unlock(&create_thread_lock); + gc_assert(lock_ret == 0); +#endif + FSHOW_SIGNAL((stderr,"/gc_start_the_world:end\n")); } + +#endif /* !LISP_FEATURE_SB_SAFEPOINT */ +#endif /* !LISP_FEATURE_SB_THREAD */ + +int +thread_yield() +{ +#ifdef LISP_FEATURE_SB_THREAD + return sched_yield(); +#else + return 0; +#endif +} + +int +wake_thread(os_thread_t os_thread) +{ +#if defined(LISP_FEATURE_WIN32) + return kill_safely(os_thread, 1); +#elif !defined(LISP_FEATURE_SB_THRUPTION) + return kill_safely(os_thread, SIGPIPE); +#else + return wake_thread_posix(os_thread); +#endif +} + +/* If the thread id given does not belong to a running thread (it has + * exited or never even existed) pthread_kill _may_ fail with ESRCH, + * but it is also allowed to just segfault, see + * . + * + * Relying on thread ids can easily backfire since ids are recycled + * (NPTL recycles them extremely fast) so a signal can be sent to + * another process if the one it was sent to exited. + * + * For these reasons, we must make sure that the thread is still alive + * when the pthread_kill is called and return if the thread is + * exiting. + * + * Note (DFL, 2011-06-22): At the time of writing, this function is only + * used for INTERRUPT-THREAD, hence the wake_thread special-case for + * Windows is OK. */ +int +kill_safely(os_thread_t os_thread, int signal) +{ + FSHOW_SIGNAL((stderr,"/kill_safely: %lu, %d\n", os_thread, signal)); + { +#ifdef LISP_FEATURE_SB_THREAD + sigset_t oldset; + struct thread *thread; + /* Frequent special case: resignalling to self. The idea is + * that leave_region safepoint will acknowledge the signal, so + * there is no need to take locks, roll thread to safepoint + * etc. */ + /* Kludge (on safepoint builds): At the moment, this isn't just + * an optimization; rather it masks the fact that + * gc_stop_the_world() grabs the all_threads mutex without + * releasing it, and since we're not using recursive pthread + * mutexes, the pthread_mutex_lock() around the all_threads loop + * would go wrong. Why are we running interruptions while + * stopping the world though? Test case is (:ASYNC-UNWIND + * :SPECIALS), especially with s/10/100/ in both loops. */ + if (os_thread == pthread_self()) { + pthread_kill(os_thread, signal); +#ifdef LISP_FEATURE_WIN32 + check_pending_thruptions(NULL); +#endif + return 0; + } + + /* pthread_kill is not async signal safe and we don't want to be + * interrupted while holding the lock. */ + block_deferrable_signals(0, &oldset); + pthread_mutex_lock(&all_threads_lock); + for (thread = all_threads; thread; thread = thread->next) { + if (thread->os_thread == os_thread) { + int status = pthread_kill(os_thread, signal); + if (status) + lose("kill_safely: pthread_kill failed with %d\n", status); +#if defined(LISP_FEATURE_WIN32) && defined(LISP_FEATURE_SB_THRUPTION) + wake_thread_win32(thread); #endif + break; + } + } + pthread_mutex_unlock(&all_threads_lock); + thread_sigmask(SIG_SETMASK,&oldset,0); + if (thread) + return 0; + else + return -1; +#elif defined(LISP_FEATURE_WIN32) + return 0; +#else + int status; + if (os_thread != 0) + lose("kill_safely: who do you want to kill? %d?\n", os_thread); + /* Dubious (as in don't know why it works) workaround for the + * signal sometimes not being generated on darwin. */ +#ifdef LISP_FEATURE_DARWIN + { + sigset_t oldset; + sigprocmask(SIG_BLOCK, &deferrable_sigset, &oldset); + status = raise(signal); + sigprocmask(SIG_SETMASK,&oldset,0); + } +#else + status = raise(signal); +#endif + if (status == 0) { + return 0; + } else { + lose("cannot raise signal %d, %d %s\n", + signal, status, strerror(errno)); + } +#endif + } +}