--- libcoro/coro.c 2010/04/02 20:21:21 1.52 +++ libcoro/coro.c 2012/12/07 14:21:09 1.66 @@ -1,16 +1,16 @@ /* - * Copyright (c) 2001-2009 Marc Alexander Lehmann - * + * Copyright (c) 2001-2011 Marc Alexander Lehmann + * * Redistribution and use in source and binary forms, with or without modifica- * tion, are permitted provided that the following conditions are met: - * + * * 1. Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. - * + * * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. - * + * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER- * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO @@ -40,6 +40,7 @@ #include "coro.h" +#include #include /*****************************************************************************/ @@ -88,6 +89,10 @@ coro_transfer (new_coro, create_coro); +#if __GCC_HAVE_DWARF2_CFI_ASM && __amd64 + asm (".cfi_undefined rip"); +#endif + func ((void *)arg); /* the new coro returned. bad. just abort() for now */ @@ -112,71 +117,132 @@ # if CORO_ASM + #if _WIN32 || __CYGWIN__ + #define CORO_WIN_TIB 1 + #endif + asm ( - ".text\n" - ".globl coro_transfer\n" - ".type coro_transfer, @function\n" + "\t.text\n" + #if _WIN32 || __CYGWIN__ + "\t.globl _coro_transfer\n" + "_coro_transfer:\n" + #else + "\t.globl coro_transfer\n" "coro_transfer:\n" + #endif /* windows, of course, gives a shit on the amd64 ABI and uses different registers */ /* http://blogs.msdn.com/freik/archive/2005/03/17/398200.aspx */ #if __amd64 - #define NUM_SAVED 6 - "\tpush %rbp\n" - "\tpush %rbx\n" - "\tpush %r12\n" - "\tpush %r13\n" - "\tpush %r14\n" - "\tpush %r15\n" - #if CORO_WIN_TIB - "\tpush %gs:0x0\n" - "\tpush %gs:0x8\n" - "\tpush %gs:0xc\n" - #endif - "\tmov %rsp, (%rdi)\n" - "\tmov (%rsi), %rsp\n" - #if CORO_WIN_TIB - "\tpop %gs:0xc\n" - "\tpop %gs:0x8\n" - "\tpop %gs:0x0\n" + + #if _WIN32 || __CYGWIN__ + #define NUM_SAVED 29 + "\tsubq $168, %rsp\t" /* one dummy qword to improve alignment */ + "\tmovaps %xmm6, (%rsp)\n" + "\tmovaps %xmm7, 16(%rsp)\n" + "\tmovaps %xmm8, 32(%rsp)\n" + "\tmovaps %xmm9, 48(%rsp)\n" + "\tmovaps %xmm10, 64(%rsp)\n" + "\tmovaps %xmm11, 80(%rsp)\n" + "\tmovaps %xmm12, 96(%rsp)\n" + "\tmovaps %xmm13, 112(%rsp)\n" + "\tmovaps %xmm14, 128(%rsp)\n" + "\tmovaps %xmm15, 144(%rsp)\n" + "\tpushq %rsi\n" + "\tpushq %rdi\n" + "\tpushq %rbp\n" + "\tpushq %rbx\n" + "\tpushq %r12\n" + "\tpushq %r13\n" + "\tpushq %r14\n" + "\tpushq %r15\n" + #if CORO_WIN_TIB + "\tpushq %fs:0x0\n" + "\tpushq %fs:0x8\n" + "\tpushq %fs:0xc\n" + #endif + "\tmovq %rsp, (%rcx)\n" + "\tmovq (%rdx), %rsp\n" + #if CORO_WIN_TIB + "\tpopq %fs:0xc\n" + "\tpopq %fs:0x8\n" + "\tpopq %fs:0x0\n" + #endif + "\tpopq %r15\n" + "\tpopq %r14\n" + "\tpopq %r13\n" + "\tpopq %r12\n" + "\tpopq %rbx\n" + "\tpopq %rbp\n" + "\tpopq %rdi\n" + "\tpopq %rsi\n" + "\tmovaps (%rsp), %xmm6\n" + "\tmovaps 16(%rsp), %xmm7\n" + "\tmovaps 32(%rsp), %xmm8\n" + "\tmovaps 48(%rsp), %xmm9\n" + "\tmovaps 64(%rsp), %xmm10\n" + "\tmovaps 80(%rsp), %xmm11\n" + "\tmovaps 96(%rsp), %xmm12\n" + "\tmovaps 112(%rsp), %xmm13\n" + "\tmovaps 128(%rsp), %xmm14\n" + "\tmovaps 144(%rsp), %xmm15\n" + "\taddq $168, %rsp\n" + #else + #define NUM_SAVED 6 + "\tpushq %rbp\n" + "\tpushq %rbx\n" + "\tpushq %r12\n" + "\tpushq %r13\n" + "\tpushq %r14\n" + "\tpushq %r15\n" + "\tmovq %rsp, (%rdi)\n" + "\tmovq (%rsi), %rsp\n" + "\tpopq %r15\n" + "\tpopq %r14\n" + "\tpopq %r13\n" + "\tpopq %r12\n" + "\tpopq %rbx\n" + "\tpopq %rbp\n" #endif - "\tpop %r15\n" - "\tpop %r14\n" - "\tpop %r13\n" - "\tpop %r12\n" - "\tpop %rbx\n" - "\tpop %rbp\n" + "\tpopq %rcx\n" + "\tjmpq *%rcx\n" + #elif __i386 + #define NUM_SAVED 4 - "\tpush %ebp\n" - "\tpush %ebx\n" - "\tpush %esi\n" - "\tpush %edi\n" + "\tpushl %ebp\n" + "\tpushl %ebx\n" + "\tpushl %esi\n" + "\tpushl %edi\n" #if CORO_WIN_TIB - "\tpush %fs:0\n" - "\tpush %fs:4\n" - "\tpush %fs:8\n" + #undef NUM_SAVED + #define NUM_SAVED 7 + "\tpushl %fs:0\n" + "\tpushl %fs:4\n" + "\tpushl %fs:8\n" #endif - "\tmov %esp, (%eax)\n" - "\tmov (%edx), %esp\n" + "\tmovl %esp, (%eax)\n" + "\tmovl (%edx), %esp\n" #if CORO_WIN_TIB - "\tpop %fs:8\n" - "\tpop %fs:4\n" - "\tpop %fs:0\n" + "\tpopl %fs:8\n" + "\tpopl %fs:4\n" + "\tpopl %fs:0\n" #endif - "\tpop %edi\n" - "\tpop %esi\n" - "\tpop %ebx\n" - "\tpop %ebp\n" + "\tpopl %edi\n" + "\tpopl %esi\n" + "\tpopl %ebx\n" + "\tpopl %ebp\n" + "\tpopl %ecx\n" + "\tjmpl *%ecx\n" + #else #error unsupported architecture #endif - "\tret\n" ); # endif void -coro_create (coro_context *ctx, coro_func coro, void *arg, void *sptr, long ssize) +coro_create (coro_context *ctx, coro_func coro, void *arg, void *sptr, size_t ssize) { coro_context nctx; # if CORO_SJLJ @@ -212,8 +278,8 @@ } /* set the new stack */ - nstk.ss_sp = STACK_ADJUST_PTR (sptr,ssize); /* yes, some platforms (IRIX) get this wrong. */ - nstk.ss_size = STACK_ADJUST_SIZE (sptr,ssize); + nstk.ss_sp = STACK_ADJUST_PTR (sptr, ssize); /* yes, some platforms (IRIX) get this wrong. */ + nstk.ss_size = STACK_ADJUST_SIZE (sptr, ssize); nstk.ss_flags = 0; if (sigaltstack (&nstk, &ostk) < 0) @@ -253,16 +319,16 @@ #elif __CYGWIN__ && __x86_64 ctx->env[7] = (long) coro_init; ctx->env[6] = (long) ((char *)sptr + ssize) - sizeof (long); - #elif defined(__MINGW32__) + #elif defined __MINGW32__ ctx->env[5] = (long) coro_init; ctx->env[4] = (long) ((char *)sptr + ssize) - sizeof (long); - #elif defined(_M_IX86) + #elif defined _M_IX86 ((_JUMP_BUFFER *)&ctx->env)->Eip = (long) coro_init; ((_JUMP_BUFFER *)&ctx->env)->Esp = (long) STACK_ADJUST_PTR (sptr, ssize) - sizeof (long); - #elif defined(_M_AMD64) + #elif defined _M_AMD64 ((_JUMP_BUFFER *)&ctx->env)->Rip = (__int64) coro_init; ((_JUMP_BUFFER *)&ctx->env)->Rsp = (__int64) STACK_ADJUST_PTR (sptr, ssize) - sizeof (__int64); - #elif defined(_M_IA64) + #elif defined _M_IA64 ((_JUMP_BUFFER *)&ctx->env)->StIIP = (__int64) coro_init; ((_JUMP_BUFFER *)&ctx->env)->IntSp = (__int64) STACK_ADJUST_PTR (sptr, ssize) - sizeof (__int64); #else @@ -307,6 +373,7 @@ #endif ctx->sp -= NUM_SAVED; + memset (ctx->sp, 0, sizeof (*ctx->sp) * NUM_SAVED); # elif CORO_UCONTEXT @@ -377,7 +444,7 @@ } void -coro_create (coro_context *ctx, coro_func coro, void *arg, void *sptr, long ssize) +coro_create (coro_context *ctx, coro_func coro, void *arg, void *sptr, size_t ssize) { static coro_context nctx; static int once; @@ -404,7 +471,15 @@ args.main = &nctx; pthread_attr_init (&attr); +#if __UCLIBC__ + /* exists, but is borked */ + /*pthread_attr_setstacksize (&attr, (size_t)ssize);*/ +#elif __CYGWIN__ + /* POSIX, not here */ + pthread_attr_setstacksize (&attr, (size_t)ssize); +#else pthread_attr_setstack (&attr, sptr, (size_t)ssize); +#endif pthread_attr_setscope (&attr, PTHREAD_SCOPE_PROCESS); pthread_create (&ctx->id, &attr, coro_init, &args); @@ -428,7 +503,205 @@ pthread_cond_destroy (&ctx->cv); } +/*****************************************************************************/ +/* fiber backend */ +/*****************************************************************************/ +#elif CORO_FIBER + +#define WIN32_LEAN_AND_MEAN +#if _WIN32_WINNT < 0x0400 + #undef _WIN32_WINNT + #define _WIN32_WINNT 0x0400 +#endif +#include + +VOID CALLBACK +coro_init (PVOID arg) +{ + coro_context *ctx = (coro_context *)arg; + + ctx->coro (ctx->arg); +} + +void +coro_transfer (coro_context *prev, coro_context *next) +{ + if (!prev->fiber) + { + prev->fiber = GetCurrentFiber (); + + if (prev->fiber == 0 || prev->fiber == (void *)0x1e00) + prev->fiber = ConvertThreadToFiber (0); + } + + SwitchToFiber (next->fiber); +} + +void +coro_create (coro_context *ctx, coro_func coro, void *arg, void *sptr, size_t ssize) +{ + ctx->fiber = 0; + ctx->coro = coro; + ctx->arg = arg; + + if (!coro) + return; + + ctx->fiber = CreateFiber (ssize, coro_init, ctx); +} + +void +coro_destroy (coro_context *ctx) +{ + DeleteFiber (ctx->fiber); +} + #else -# error unsupported backend + #error unsupported backend +#endif + +/*****************************************************************************/ +/* stack management */ +/*****************************************************************************/ +#if CORO_STACKALLOC + +#include + +#ifndef _WIN32 +# include +#endif + +#if CORO_USE_VALGRIND +# include +#endif + +#if _POSIX_MAPPED_FILES +# include +# define CORO_MMAP 1 +# ifndef MAP_ANONYMOUS +# ifdef MAP_ANON +# define MAP_ANONYMOUS MAP_ANON +# else +# undef CORO_MMAP +# endif +# endif +# include +#else +# undef CORO_MMAP +#endif + +#if _POSIX_MEMORY_PROTECTION +# ifndef CORO_GUARDPAGES +# define CORO_GUARDPAGES 4 +# endif +#else +# undef CORO_GUARDPAGES +#endif + +#if !CORO_MMAP +# undef CORO_GUARDPAGES +#endif + +#if !__i386 && !__x86_64 && !__powerpc && !__m68k && !__alpha && !__mips && !__sparc64 +# undef CORO_GUARDPAGES +#endif + +#ifndef CORO_GUARDPAGES +# define CORO_GUARDPAGES 0 +#endif + +#if !PAGESIZE + #if !CORO_MMAP + #define PAGESIZE 4096 + #else + static size_t + coro_pagesize (void) + { + static size_t pagesize; + + if (!pagesize) + pagesize = sysconf (_SC_PAGESIZE); + + return pagesize; + } + + #define PAGESIZE coro_pagesize () + #endif +#endif + +int +coro_stack_alloc (struct coro_stack *stack, unsigned int size) +{ + if (!size) + size = 256 * 1024; + + stack->sptr = 0; + stack->ssze = ((size_t)size * sizeof (void *) + PAGESIZE - 1) / PAGESIZE * PAGESIZE; + +#if CORO_FIBER + + stack->sptr = (void *)stack; + return 1; + +#else + + size_t ssze = stack->ssze + CORO_GUARDPAGES * PAGESIZE; + void *base; + + #if CORO_MMAP + /* mmap supposedly does allocate-on-write for us */ + base = mmap (0, ssze, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + + if (base == (void *)-1) + { + /* some systems don't let us have executable heap */ + /* we assume they won't need executable stack in that case */ + base = mmap (0, ssze, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + + if (base == (void *)-1) + return 0; + } + + #if CORO_GUARDPAGES + mprotect (base, CORO_GUARDPAGES * PAGESIZE, PROT_NONE); + #endif + + base = (void*)((char *)base + CORO_GUARDPAGES * PAGESIZE); + #else + base = malloc (ssze); + if (!base) + return 0; + #endif + + #if CORO_USE_VALGRIND + stack->valgrind_id = VALGRIND_STACK_REGISTER ((char *)base, (char *)ssze - CORO_GUARDPAGES * PAGESIZE); + #endif + + stack->sptr = base; + return 1; + +#endif +} + +void +coro_stack_free (struct coro_stack *stack) +{ +#if CORO_FIBER + /* nop */ +#else + #if CORO_USE_VALGRIND + VALGRIND_STACK_DEREGISTER (stack->valgrind_id); + #endif + + #if CORO_MMAP + if (stack->sptr) + munmap ((void*)((char *)stack->sptr - CORO_GUARDPAGES * PAGESIZE), + stack->ssze + CORO_GUARDPAGES * PAGESIZE); + #else + free (stack->sptr); + #endif +#endif +} + #endif