eitan.c/suyu
1
0
Fork 0
forked from suyu/suyu
Code Pull requests Activity

fix clang V2

Browse source
This commit is contained in:
Crimson-Hawk 2024-04-26 17:29:03 +08:00
parent eed403ea0c
commit 7d0b6e781d
Signed by: Crimson-Hawk
GPG key ID: 0804DD39BB9BF5AC
1 changed files with 1265 additions and 1221 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

250
src/common/minicoro.h
View file

@ -50,14 +50,21 @@ to create, resume, yield or destroy a coroutine.
# Caveats # Caveats
- Avoid using coroutines with C++ exceptions, this is not recommended, it may not behave as you expect. - Avoid using coroutines with C++ exceptions, this is not recommended, it may not behave as you
- When using C++ RAII (i.e. destructors) you must resume the coroutine until it dies to properly execute all destructors. expect.
- When using C++ RAII (i.e. destructors) you must resume the coroutine until it dies to properly
execute all destructors.
- Some unsupported sanitizers for C may trigger false warnings when using coroutines. - Some unsupported sanitizers for C may trigger false warnings when using coroutines.
- The `mco_coro` object is not thread safe, you should use a mutex for manipulating it in multithread applications. - The `mco_coro` object is not thread safe, you should use a mutex for manipulating it in
- To use in multithread applications, you must compile with C compiler that supports `thread_local` qualifier. multithread applications.
- Avoid using `thread_local` inside coroutine code, the compiler may cache thread local variables pointers which can be invalid when a coroutine switch threads. - To use in multithread applications, you must compile with C compiler that supports `thread_local`
- Stack space is limited. By default it has 56KB of space, this can be changed on coroutine creation, or by enabling the virtual memory backed allocator to make it 2040KB. qualifier.
- Take care to not cause stack overflows (run out of stack space), otherwise your program may crash or not, the behavior is undefined. - Avoid using `thread_local` inside coroutine code, the compiler may cache thread local variables
pointers which can be invalid when a coroutine switch threads.
- Stack space is limited. By default it has 56KB of space, this can be changed on coroutine
creation, or by enabling the virtual memory backed allocator to make it 2040KB.
- Take care to not cause stack overflows (run out of stack space), otherwise your program may crash
or not, the behavior is undefined.
- On WebAssembly you must compile with Emscripten flag `-s ASYNCIFY=1`. - On WebAssembly you must compile with Emscripten flag `-s ASYNCIFY=1`.
- The WebAssembly Binaryen asyncify method can be used when explicitly enabled, - The WebAssembly Binaryen asyncify method can be used when explicitly enabled,
you may want to do this only to use minicoro with WebAssembly native interpreters you may want to do this only to use minicoro with WebAssembly native interpreters
@ -72,7 +79,8 @@ a coroutine only suspends its execution by explicitly calling a yield function.
You create a coroutine by calling `mco_create`. You create a coroutine by calling `mco_create`.
Its sole argument is a `mco_desc` structure with a description for the coroutine. Its sole argument is a `mco_desc` structure with a description for the coroutine.
The `mco_create` function only creates a new coroutine and returns a handle to it, it does not start the coroutine. The `mco_create` function only creates a new coroutine and returns a handle to it, it does not start
the coroutine.
You execute a coroutine by calling `mco_resume`. You execute a coroutine by calling `mco_resume`.
When calling a resume function the coroutine starts its execution by calling its body function. When calling a resume function the coroutine starts its execution by calling its body function.
@ -81,7 +89,8 @@ After the coroutine starts running, it runs until it terminates or yields.
A coroutine yields by calling `mco_yield`. A coroutine yields by calling `mco_yield`.
When a coroutine yields, the corresponding resume returns immediately, When a coroutine yields, the corresponding resume returns immediately,
even if the yield happens inside nested function calls (that is, not in the main function). even if the yield happens inside nested function calls (that is, not in the main function).
The next time you resume the same coroutine, it continues its execution from the point where it yielded. The next time you resume the same coroutine, it continues its execution from the point where it
yielded.
To associate a persistent value with the coroutine, To associate a persistent value with the coroutine,
you can optionally set `user_data` on its creation and later retrieve with `mco_get_user_data`. you can optionally set `user_data` on its creation and later retrieve with `mco_get_user_data`.
@ -89,7 +98,8 @@ you can optionally set `user_data` on its creation and later retrieve with `mco
To pass values between resume and yield, To pass values between resume and yield,
you can optionally use `mco_push` and `mco_pop` APIs, you can optionally use `mco_push` and `mco_pop` APIs,
they are intended to pass temporary values using a LIFO style buffer. they are intended to pass temporary values using a LIFO style buffer.
The storage system can also be used to send and receive initial values on coroutine creation or before it finishes. The storage system can also be used to send and receive initial values on coroutine creation or
before it finishes.
# Usage # Usage
@ -108,9 +118,9 @@ The following simple example demonstrates on how to use the library:
```c ```c
#define MINICORO_IMPL #define MINICORO_IMPL
#include "minicoro.h"
#include <stdio.h>
#include <assert.h> #include <assert.h>
#include <stdio.h>
#include "minicoro.h"
// Coroutine entry function. // Coroutine entry function.
void coro_entry(mco_coro* co) { void coro_entry(mco_coro* co) {
@ -181,8 +191,9 @@ for example, an application with 100 thousands coroutine with stacks of 56KB wou
as 5GB of memory, however your application may not really full stack usage for every coroutine. as 5GB of memory, however your application may not really full stack usage for every coroutine.
Some developers often prefer stackless coroutines over stackful coroutines Some developers often prefer stackless coroutines over stackful coroutines
because of this problem, stackless memory footprint is low, therefore often considered more lightweight. because of this problem, stackless memory footprint is low, therefore often considered more
However stackless have many other limitations, like you cannot run unconstrained code inside them. lightweight. However stackless have many other limitations, like you cannot run unconstrained code
inside them.
One remedy to the solution is to make stackful coroutines growable, One remedy to the solution is to make stackful coroutines growable,
to only use physical memory on demand when its really needed, to only use physical memory on demand when its really needed,
@ -192,19 +203,21 @@ when supported by the operating system.
The virtual memory backed allocator will reserve virtual memory in the OS for each coroutine stack, The virtual memory backed allocator will reserve virtual memory in the OS for each coroutine stack,
but not trigger real physical memory usage yet. but not trigger real physical memory usage yet.
While the application virtual memory usage will be high, While the application virtual memory usage will be high,
the physical memory usage will be low and actually grow on demand (usually every 4KB chunk in Linux). the physical memory usage will be low and actually grow on demand (usually every 4KB chunk in
Linux).
The virtual memory backed allocator also raises the default stack size to about 2MB, The virtual memory backed allocator also raises the default stack size to about 2MB,
typically the size of extra threads in Linux, typically the size of extra threads in Linux,
so you have more space in your coroutines and the risk of stack overflow is low. so you have more space in your coroutines and the risk of stack overflow is low.
As an example, allocating 100 thousands coroutines with nearly 2MB stack reserved space As an example, allocating 100 thousands coroutines with nearly 2MB stack reserved space
with the virtual memory allocator uses 783MB of physical memory usage, that is about 8KB per coroutine, with the virtual memory allocator uses 783MB of physical memory usage, that is about 8KB per
however the virtual memory usage will be at 98GB. coroutine, however the virtual memory usage will be at 98GB.
It is recommended to enable this option only if you plan to spawn thousands of coroutines It is recommended to enable this option only if you plan to spawn thousands of coroutines
while wanting to have a low memory footprint. while wanting to have a low memory footprint.
Not all environments have an OS with virtual memory support, therefore this option is disabled by default. Not all environments have an OS with virtual memory support, therefore this option is disabled by
default.
This option may add an order of magnitude overhead to `mco_create()`/`mco_destroy()`, This option may add an order of magnitude overhead to `mco_create()`/`mco_destroy()`,
because they will request the OS to manage virtual memory page tables, because they will request the OS to manage virtual memory page tables,
@ -215,17 +228,23 @@ if this is a problem for you, please customize a custom allocator for your own n
The following can be defined to change the library behavior: The following can be defined to change the library behavior:
- `MCO_API` - Public API qualifier. Default is `extern`. - `MCO_API` - Public API qualifier. Default is `extern`.
- `MCO_MIN_STACK_SIZE` - Minimum stack size when creating a coroutine. Default is 32768 (32KB). - `MCO_MIN_STACK_SIZE` - Minimum stack size when creating a coroutine. Default is 32768
(32KB).
- `MCO_DEFAULT_STORAGE_SIZE` - Size of coroutine storage buffer. Default is 1024. - `MCO_DEFAULT_STORAGE_SIZE` - Size of coroutine storage buffer. Default is 1024.
- `MCO_DEFAULT_STACK_SIZE` - Default stack size when creating a coroutine. Default is 57344 (56KB). When `MCO_USE_VMEM_ALLOCATOR` is true the default is 2040KB (nearly 2MB). - `MCO_DEFAULT_STACK_SIZE` - Default stack size when creating a coroutine. Default is 57344
(56KB). When `MCO_USE_VMEM_ALLOCATOR` is true the default is 2040KB (nearly 2MB).
- `MCO_ALLOC` - Default allocation function. Default is `calloc`. - `MCO_ALLOC` - Default allocation function. Default is `calloc`.
- `MCO_DEALLOC` - Default deallocation function. Default is `free`. - `MCO_DEALLOC` - Default deallocation function. Default is `free`.
- `MCO_USE_VMEM_ALLOCATOR` - Use virtual memory backed allocator, improving memory footprint per coroutine. - `MCO_USE_VMEM_ALLOCATOR` - Use virtual memory backed allocator, improving memory footprint per
coroutine.
- `MCO_NO_DEFAULT_ALLOCATOR` - Disable the default allocator using `MCO_ALLOC` and `MCO_DEALLOC`. - `MCO_NO_DEFAULT_ALLOCATOR` - Disable the default allocator using `MCO_ALLOC` and `MCO_DEALLOC`.
- `MCO_ZERO_MEMORY` - Zero memory of stack when poping storage, intended for garbage collected environments. - `MCO_ZERO_MEMORY` - Zero memory of stack when poping storage, intended for garbage
- `MCO_DEBUG` - Enable debug mode, logging any runtime error to stdout. Defined automatically unless `NDEBUG` or `MCO_NO_DEBUG` is defined. collected environments.
- `MCO_DEBUG` - Enable debug mode, logging any runtime error to stdout. Defined
automatically unless `NDEBUG` or `MCO_NO_DEBUG` is defined.
- `MCO_NO_DEBUG` - Disable debug mode. - `MCO_NO_DEBUG` - Disable debug mode.
- `MCO_NO_MULTITHREAD` - Disable multithread usage. Multithread is supported when `thread_local` is supported. - `MCO_NO_MULTITHREAD` - Disable multithread usage. Multithread is supported when
`thread_local` is supported.
- `MCO_USE_ASM` - Force use of assembly context switch implementation. - `MCO_USE_ASM` - Force use of assembly context switch implementation.
- `MCO_USE_UCONTEXT` - Force use of ucontext context switch implementation. - `MCO_USE_UCONTEXT` - Force use of ucontext context switch implementation.
- `MCO_USE_FIBERS` - Force use of fibers context switch implementation. - `MCO_USE_FIBERS` - Force use of fibers context switch implementation.
@ -237,7 +256,6 @@ The following can be defined to change the library behavior:
Your choice of either Public Domain or MIT No Attribution, see end of file. Your choice of either Public Domain or MIT No Attribution, see end of file.
*/ */
#ifndef MINICORO_H #ifndef MINICORO_H
#define MINICORO_H #define MINICORO_H
@ -262,9 +280,11 @@ extern "C" {
/* Coroutine states. */ /* Coroutine states. */
typedef enum mco_state { typedef enum mco_state {
MCO_DEAD = 0, /* The coroutine has finished normally or was uninitialized before finishing. */ MCO_DEAD = 0, /* The coroutine has finished normally or was uninitialized before finishing. */
MCO_NORMAL, /* The coroutine is active but not running (that is, it has resumed another coroutine). */ MCO_NORMAL, /* The coroutine is active but not running (that is, it has resumed another
coroutine). */
MCO_RUNNING, /* The coroutine is active and running. */ MCO_RUNNING, /* The coroutine is active and running. */
MCO_SUSPENDED /* The coroutine is suspended (in a call to yield, or it has not started running yet). */ MCO_SUSPENDED /* The coroutine is suspended (in a call to yield, or it has not started running
yet). */
} mco_state; } mco_state;
/* Coroutine result codes. */ /* Coroutine result codes. */
@ -312,7 +332,8 @@ typedef struct mco_desc {
void* user_data; /* Coroutine user data, can be get with `mco_get_user_data`. */ void* user_data; /* Coroutine user data, can be get with `mco_get_user_data`. */
/* Custom allocation interface. */ /* Custom allocation interface. */
void* (*alloc_cb)(size_t size, void* allocator_data); /* Custom allocation function. */ void* (*alloc_cb)(size_t size, void* allocator_data); /* Custom allocation function. */
void (*dealloc_cb)(void* ptr, size_t size, void* allocator_data); /* Custom deallocation function. */ void (*dealloc_cb)(void* ptr, size_t size,
void* allocator_data); /* Custom deallocation function. */
void* allocator_data; /* User data pointer passed to `alloc`/`dealloc` allocation functions. */ void* allocator_data; /* User data pointer passed to `alloc`/`dealloc` allocation functions. */
size_t storage_size; /* Coroutine storage size, to be used with the storage APIs. */ size_t storage_size; /* Coroutine storage size, to be used with the storage APIs. */
/* These must be initialized only through `mco_init_desc`. */ /* These must be initialized only through `mco_init_desc`. */
@ -321,21 +342,35 @@ typedef struct mco_desc {
} mco_desc; } mco_desc;
/* Coroutine functions. */ /* Coroutine functions. */
MCO_API mco_desc mco_desc_init(void (*func)(mco_coro* co), size_t stack_size); /* Initialize description of a coroutine. When stack size is 0 then MCO_DEFAULT_STACK_SIZE is used. */ MCO_API mco_desc
mco_desc_init(void (*func)(mco_coro* co),
size_t stack_size); /* Initialize description of a coroutine. When stack size is 0
then MCO_DEFAULT_STACK_SIZE is used. */
MCO_API mco_result mco_init(mco_coro* co, mco_desc* desc); /* Initialize the coroutine. */ MCO_API mco_result mco_init(mco_coro* co, mco_desc* desc); /* Initialize the coroutine. */
MCO_API mco_result mco_uninit(mco_coro* co); /* Uninitialize the coroutine, may fail if it's not dead or suspended. */ MCO_API mco_result
MCO_API mco_result mco_create(mco_coro** out_co, mco_desc* desc); /* Allocates and initializes a new coroutine. */ mco_uninit(mco_coro* co); /* Uninitialize the coroutine, may fail if it's not dead or suspended. */
MCO_API mco_result mco_destroy(mco_coro* co); /* Uninitialize and deallocate the coroutine, may fail if it's not dead or suspended. */ MCO_API mco_result mco_create(mco_coro** out_co,
MCO_API mco_result mco_resume(mco_coro* co); /* Starts or continues the execution of the coroutine. */ mco_desc* desc); /* Allocates and initializes a new coroutine. */
MCO_API mco_result mco_destroy(mco_coro* co); /* Uninitialize and deallocate the coroutine, may fail
if it's not dead or suspended. */
MCO_API mco_result
mco_resume(mco_coro* co); /* Starts or continues the execution of the coroutine. */
MCO_API mco_result mco_yield(mco_coro* co); /* Suspends the execution of a coroutine. */ MCO_API mco_result mco_yield(mco_coro* co); /* Suspends the execution of a coroutine. */
MCO_API mco_state mco_status(mco_coro* co); /* Returns the status of the coroutine. */ MCO_API mco_state mco_status(mco_coro* co); /* Returns the status of the coroutine. */
MCO_API void* mco_get_user_data(mco_coro* co); /* Get coroutine user data supplied on coroutine creation. */ MCO_API void* mco_get_user_data(
mco_coro* co); /* Get coroutine user data supplied on coroutine creation. */
/* Storage interface functions, used to pass values between yield and resume. */ /* Storage interface functions, used to pass values between yield and resume. */
MCO_API mco_result mco_push(mco_coro* co, const void* src, size_t len); /* Push bytes to the coroutine storage. Use to send values between yield and resume. */ MCO_API mco_result mco_push(mco_coro* co, const void* src,
MCO_API mco_result mco_pop(mco_coro* co, void* dest, size_t len); /* Pop bytes from the coroutine storage. Use to get values between yield and resume. */ size_t len); /* Push bytes to the coroutine storage. Use to send values
MCO_API mco_result mco_peek(mco_coro* co, void* dest, size_t len); /* Like `mco_pop` but it does not consumes the storage. */ between yield and resume. */
MCO_API size_t mco_get_bytes_stored(mco_coro* co); /* Get the available bytes that can be retrieved with a `mco_pop`. */ MCO_API mco_result mco_pop(mco_coro* co, void* dest,
size_t len); /* Pop bytes from the coroutine storage. Use to get values
between yield and resume. */
MCO_API mco_result mco_peek(mco_coro* co, void* dest,
size_t len); /* Like `mco_pop` but it does not consumes the storage. */
MCO_API size_t mco_get_bytes_stored(
mco_coro* co); /* Get the available bytes that can be retrieved with a `mco_pop`. */
MCO_API size_t mco_get_storage_size(mco_coro* co); /* Get the total storage size. */ MCO_API size_t mco_get_storage_size(mco_coro* co); /* Get the total storage size. */
/* Misc functions. */ /* Misc functions. */
@ -365,7 +400,8 @@ extern "C" {
#ifndef MCO_DEFAULT_STACK_SIZE #ifndef MCO_DEFAULT_STACK_SIZE
/* Use multiples of 64KB minus 8KB, because 8KB is reserved for coroutine internal structures. */ /* Use multiples of 64KB minus 8KB, because 8KB is reserved for coroutine internal structures. */
#ifdef MCO_USE_VMEM_ALLOCATOR #ifdef MCO_USE_VMEM_ALLOCATOR
#define MCO_DEFAULT_STACK_SIZE 2040*1024 /* 2040KB, nearly the same stack size of a thread in x86_64 Linux. */ #define MCO_DEFAULT_STACK_SIZE \
2040 * 1024 /* 2040KB, nearly the same stack size of a thread in x86_64 Linux. */
#else #else
#define MCO_DEFAULT_STACK_SIZE 56 * 1024 /* 56KB */ #define MCO_DEFAULT_STACK_SIZE 56 * 1024 /* 56KB */
#endif #endif
@ -375,7 +411,8 @@ extern "C" {
#define MCO_MAGIC_NUMBER 0x7E3CB1A9 #define MCO_MAGIC_NUMBER 0x7E3CB1A9
/* Detect implementation based on OS, arch and compiler. */ /* Detect implementation based on OS, arch and compiler. */
#if !defined(MCO_USE_UCONTEXT) && !defined(MCO_USE_FIBERS) && !defined(MCO_USE_ASM) && !defined(MCO_USE_ASYNCIFY) #if !defined(MCO_USE_UCONTEXT) && !defined(MCO_USE_FIBERS) && !defined(MCO_USE_ASM) && \
!defined(MCO_USE_ASYNCIFY)
#if defined(_WIN32) #if defined(_WIN32)
#if (defined(__GNUC__) && defined(__x86_64__)) || (defined(_MSC_VER) && defined(_M_X64)) #if (defined(__GNUC__) && defined(__x86_64__)) || (defined(_MSC_VER) && defined(_M_X64))
#define MCO_USE_ASM #define MCO_USE_ASM
@ -390,10 +427,8 @@ extern "C" {
#define MCO_USE_ASYNCIFY #define MCO_USE_ASYNCIFY
#else #else
#if __GNUC__ >= 3 /* Assembly extension supported. */ #if __GNUC__ >= 3 /* Assembly extension supported. */
#if defined(__x86_64__) || \ #if defined(__x86_64__) || defined(__i386) || defined(__i386__) || defined(__ARM_EABI__) || \
defined(__i386) || defined(__i386__) || \ defined(__aarch64__) || defined(__riscv)
defined(__ARM_EABI__) || defined(__aarch64__) || \
defined(__riscv)
#define MCO_USE_ASM #define MCO_USE_ASM
#else #else
#define MCO_USE_UCONTEXT #define MCO_USE_UCONTEXT
@ -436,7 +471,8 @@ extern "C" {
#define MCO_THREAD_LOCAL thread_local #define MCO_THREAD_LOCAL thread_local
#elif __STDC_VERSION__ >= 201112 && !defined(__STDC_NO_THREADS__) #elif __STDC_VERSION__ >= 201112 && !defined(__STDC_NO_THREADS__)
#define MCO_THREAD_LOCAL _Thread_local #define MCO_THREAD_LOCAL _Thread_local
#elif defined(_WIN32) && (defined(_MSC_VER) || defined(__ICL) || defined(__DMC__) || defined(__BORLANDC__)) #elif defined(_WIN32) && \
(defined(_MSC_VER) || defined(__ICL) || defined(__DMC__) || defined(__BORLANDC__))
#define MCO_THREAD_LOCAL __declspec(thread) #define MCO_THREAD_LOCAL __declspec(thread)
#elif defined(__GNUC__) || defined(__SUNPRO_C) || defined(__xlC__) #elif defined(__GNUC__) || defined(__SUNPRO_C) || defined(__xlC__)
#define MCO_THREAD_LOCAL __thread #define MCO_THREAD_LOCAL __thread
@ -456,7 +492,8 @@ extern "C" {
#else #else
#define MCO_FORCE_INLINE inline __attribute__((always_inline)) #define MCO_FORCE_INLINE inline __attribute__((always_inline))
#endif #endif
#elif defined(__BORLANDC__) || defined(__DMC__) || defined(__SC__) || defined(__WATCOMC__) || defined(__LCC__) || defined(__DECC) #elif defined(__BORLANDC__) || defined(__DMC__) || defined(__SC__) || defined(__WATCOMC__) || \
defined(__LCC__) || defined(__DECC)
#define MCO_FORCE_INLINE __inline #define MCO_FORCE_INLINE __inline
#else /* No inline support. */ #else /* No inline support. */
#define MCO_FORCE_INLINE #define MCO_FORCE_INLINE
@ -512,8 +549,8 @@ extern "C" {
#else /* C allocator */ #else /* C allocator */
#ifndef MCO_ALLOC #ifndef MCO_ALLOC
#include <stdlib.h> #include <stdlib.h>
/* We use calloc() so we give a chance for the OS to reserve virtual memory without really using physical memory, /* We use calloc() so we give a chance for the OS to reserve virtual memory without really using
calloc() also has the nice property of initializing the stack to zeros. */ physical memory, calloc() also has the nice property of initializing the stack to zeros. */
#define MCO_ALLOC(size) calloc(1, size) #define MCO_ALLOC(size) calloc(1, size)
#define MCO_DEALLOC(ptr, size) free(ptr) #define MCO_DEALLOC(ptr, size) free(ptr)
#endif #endif
@ -545,7 +582,8 @@ extern "C" {
#endif #endif
#ifdef _MCO_USE_ASAN #ifdef _MCO_USE_ASAN
void __sanitizer_start_switch_fiber(void** fake_stack_save, const void* bottom, size_t size); void __sanitizer_start_switch_fiber(void** fake_stack_save, const void* bottom, size_t size);
void __sanitizer_finish_switch_fiber(void* fake_stack_save, const void **bottom_old, size_t *size_old); void __sanitizer_finish_switch_fiber(void* fake_stack_save, const void** bottom_old,
size_t* size_old);
#endif #endif
#ifdef _MCO_USE_TSAN #ifdef _MCO_USE_TSAN
void* __tsan_get_current_fiber(void); void* __tsan_get_current_fiber(void);
@ -578,7 +616,8 @@ static MCO_FORCE_INLINE void _mco_prepare_jumpin(mco_coro* co) {
if (prev_co) { if (prev_co) {
void* bottom_old = NULL; void* bottom_old = NULL;
size_t size_old = 0; size_t size_old = 0;
__sanitizer_finish_switch_fiber(prev_co->asan_prev_stack, (const void**)&bottom_old, &size_old); __sanitizer_finish_switch_fiber(prev_co->asan_prev_stack, (const void**)&bottom_old,
&size_old);
prev_co->asan_prev_stack = NULL; prev_co->asan_prev_stack = NULL;
} }
__sanitizer_start_switch_fiber(&co->asan_prev_stack, co->stack_base, co->stack_size); __sanitizer_start_switch_fiber(&co->asan_prev_stack, co->stack_base, co->stack_size);
@ -750,9 +789,12 @@ _MCO_ASM_BLOB static unsigned char _mco_switch_code[] = {
}; };
void (*_mco_wrap_main)(void) = (void (*)(void))(void*)_mco_wrap_main_code; void (*_mco_wrap_main)(void) = (void (*)(void))(void*)_mco_wrap_main_code;
void (*_mco_switch)(_mco_ctxbuf* from, _mco_ctxbuf* to) = (void(*)(_mco_ctxbuf* from, _mco_ctxbuf* to))(void*)_mco_switch_code; void (*_mco_switch)(_mco_ctxbuf* from,
_mco_ctxbuf* to) = (void (*)(_mco_ctxbuf* from,
_mco_ctxbuf* to))(void*)_mco_switch_code;
static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base, size_t stack_size) { static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base,
size_t stack_size) {
stack_size = stack_size - 32; /* Reserve 32 bytes for the shadow space. */ stack_size = stack_size - 32; /* Reserve 32 bytes for the shadow space. */
void** stack_high_ptr = (void**)((size_t)stack_base + stack_size - sizeof(size_t)); void** stack_high_ptr = (void**)((size_t)stack_base + stack_size - sizeof(size_t));
stack_high_ptr[0] = (void*)(0xdeaddeaddeaddead); /* Dummy return address. */ stack_high_ptr[0] = (void*)(0xdeaddeaddeaddead); /* Dummy return address. */
@ -776,8 +818,7 @@ typedef struct _mco_ctxbuf {
void _mco_wrap_main(void); void _mco_wrap_main(void);
int _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to); int _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to);
__asm__( __asm__(".text\n"
".text\n"
#ifdef __MACH__ /* Mac OS X assembler */ #ifdef __MACH__ /* Mac OS X assembler */
".globl __mco_wrap_main\n" ".globl __mco_wrap_main\n"
"__mco_wrap_main:\n" "__mco_wrap_main:\n"
@ -794,8 +835,7 @@ __asm__(
#endif #endif
); );
__asm__( __asm__(".text\n"
".text\n"
#ifdef __MACH__ /* Mac OS assembler */ #ifdef __MACH__ /* Mac OS assembler */
".globl __mco_switch\n" ".globl __mco_switch\n"
"__mco_switch:\n" "__mco_switch:\n"
@ -828,8 +868,10 @@ __asm__(
#endif #endif
); );
static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base, size_t stack_size) { static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base,
stack_size = stack_size - 128; /* Reserve 128 bytes for the Red Zone space (System V AMD64 ABI). */ size_t stack_size) {
stack_size =
stack_size - 128; /* Reserve 128 bytes for the Red Zone space (System V AMD64 ABI). */
void** stack_high_ptr = (void**)((size_t)stack_base + stack_size - sizeof(size_t)); void** stack_high_ptr = (void**)((size_t)stack_base + stack_size - sizeof(size_t));
stack_high_ptr[0] = (void*)(0xdeaddeaddeaddead); /* Dummy return address. */ stack_high_ptr[0] = (void*)(0xdeaddeaddeaddead); /* Dummy return address. */
ctx->rip = (void*)(_mco_wrap_main); ctx->rip = (void*)(_mco_wrap_main);
@ -860,19 +902,16 @@ typedef struct _mco_ctxbuf {
void _mco_wrap_main(void); void _mco_wrap_main(void);
int _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to); int _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to);
__asm__( __asm__(".text\n"
".text\n"
".globl _mco_wrap_main\n" ".globl _mco_wrap_main\n"
".type _mco_wrap_main @function\n" ".type _mco_wrap_main @function\n"
".hidden _mco_wrap_main\n" ".hidden _mco_wrap_main\n"
"_mco_wrap_main:\n" "_mco_wrap_main:\n"
" mv a0, s0\n" " mv a0, s0\n"
" jr s1\n" " jr s1\n"
".size _mco_wrap_main, .-_mco_wrap_main\n" ".size _mco_wrap_main, .-_mco_wrap_main\n");
);
__asm__( __asm__(".text\n"
".text\n"
".globl _mco_switch\n" ".globl _mco_switch\n"
".type _mco_switch @function\n" ".type _mco_switch @function\n"
".hidden _mco_switch\n" ".hidden _mco_switch\n"
@ -1029,10 +1068,10 @@ __asm__(
#else #else
#error "Unsupported RISC-V XLEN" #error "Unsupported RISC-V XLEN"
#endif /* __riscv_xlen */ #endif /* __riscv_xlen */
".size _mco_switch, .-_mco_switch\n" ".size _mco_switch, .-_mco_switch\n");
);
static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base, size_t stack_size) { static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base,
size_t stack_size) {
ctx->s[0] = (void*)(co); ctx->s[0] = (void*)(co);
ctx->s[1] = (void*)(_mco_main); ctx->s[1] = (void*)(_mco_main);
ctx->pc = (void*)(_mco_wrap_main); ctx->pc = (void*)(_mco_wrap_main);
@ -1088,7 +1127,8 @@ __asm__(
#endif #endif
); );
static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base, size_t stack_size) { static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base,
size_t stack_size) {
void** stack_high_ptr = (void**)((size_t)stack_base + stack_size - 16 - 1 * sizeof(size_t)); void** stack_high_ptr = (void**)((size_t)stack_base + stack_size - 16 - 1 * sizeof(size_t));
stack_high_ptr[0] = (void*)(0xdeaddead); /* Dummy return address. */ stack_high_ptr[0] = (void*)(0xdeaddead); /* Dummy return address. */
stack_high_ptr[1] = (void*)(co); stack_high_ptr[1] = (void*)(co);
@ -1112,8 +1152,7 @@ typedef struct _mco_ctxbuf {
void _mco_wrap_main(void); void _mco_wrap_main(void);
int _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to); int _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to);
__asm__( __asm__(".text\n"
".text\n"
#ifdef __APPLE__ #ifdef __APPLE__
".globl __mco_switch\n" ".globl __mco_switch\n"
"__mco_switch:\n" "__mco_switch:\n"
@ -1138,8 +1177,7 @@ __asm__(
#endif #endif
); );
__asm__( __asm__(".text\n"
".text\n"
#ifdef __APPLE__ #ifdef __APPLE__
".globl __mco_wrap_main\n" ".globl __mco_wrap_main\n"
"__mco_wrap_main:\n" "__mco_wrap_main:\n"
@ -1158,7 +1196,8 @@ __asm__(
#endif #endif
); );
static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base, size_t stack_size) { static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base,
size_t stack_size) {
ctx->d[0] = (void*)(co); ctx->d[0] = (void*)(co);
ctx->d[1] = (void*)(_mco_main); ctx->d[1] = (void*)(_mco_main);
ctx->d[2] = (void*)(0xdeaddead); /* Dummy return address. */ ctx->d[2] = (void*)(0xdeaddead); /* Dummy return address. */
@ -1179,8 +1218,7 @@ typedef struct _mco_ctxbuf {
void _mco_wrap_main(void); void _mco_wrap_main(void);
int _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to); int _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to);
__asm__( __asm__(".text\n"
".text\n"
#ifdef __APPLE__ #ifdef __APPLE__
".globl __mco_switch\n" ".globl __mco_switch\n"
"__mco_switch:\n" "__mco_switch:\n"
@ -1222,8 +1260,7 @@ __asm__(
#endif #endif
); );
__asm__( __asm__(".text\n"
".text\n"
#ifdef __APPLE__ #ifdef __APPLE__
".globl __mco_wrap_main\n" ".globl __mco_wrap_main\n"
"__mco_wrap_main:\n" "__mco_wrap_main:\n"
@ -1241,7 +1278,8 @@ __asm__(
#endif #endif
); );
static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base, size_t stack_size) { static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base,
size_t stack_size) {
ctx->x[0] = (void*)(co); ctx->x[0] = (void*)(co);
ctx->x[1] = (void*)(_mco_main); ctx->x[1] = (void*)(_mco_main);
ctx->x[2] = (void*)(0xdeaddeaddeaddead); /* Dummy return address. */ ctx->x[2] = (void*)(0xdeaddeaddeaddead); /* Dummy return address. */
@ -1264,7 +1302,8 @@ typedef ucontext_t _mco_ctxbuf;
#if defined(_LP64) || defined(__LP64__) #if defined(_LP64) || defined(__LP64__)
static void _mco_wrap_main(unsigned int lo, unsigned int hi) { static void _mco_wrap_main(unsigned int lo, unsigned int hi) {
mco_coro* co = (mco_coro*)(((size_t)lo) | (((size_t)hi) << 32)); /* Extract coroutine pointer. */ mco_coro* co =
(mco_coro*)(((size_t)lo) | (((size_t)hi) << 32)); /* Extract coroutine pointer. */
_mco_main(co); _mco_main(co);
} }
#else #else
@ -1280,7 +1319,8 @@ static MCO_FORCE_INLINE void _mco_switch(_mco_ctxbuf* from, _mco_ctxbuf* to) {
MCO_ASSERT(res == 0); MCO_ASSERT(res == 0);
} }
static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base, size_t stack_size) { static mco_result _mco_makectx(mco_coro* co, _mco_ctxbuf* ctx, void* stack_base,
size_t stack_size) {
/* Initialize ucontext. */ /* Initialize ucontext. */
if (getcontext(ctx) != 0) { if (getcontext(ctx) != 0) {
MCO_LOG("failed to get ucontext"); MCO_LOG("failed to get ucontext");
@ -1370,8 +1410,7 @@ static void _mco_destroy_context(mco_coro* co) {
static MCO_FORCE_INLINE void _mco_init_desc_sizes(mco_desc* desc, size_t stack_size) { static MCO_FORCE_INLINE void _mco_init_desc_sizes(mco_desc* desc, size_t stack_size) {
desc->coro_size = _mco_align_forward(sizeof(mco_coro), 16) + desc->coro_size = _mco_align_forward(sizeof(mco_coro), 16) +
_mco_align_forward(sizeof(_mco_context), 16) + _mco_align_forward(sizeof(_mco_context), 16) +
_mco_align_forward(desc->storage_size, 16) + _mco_align_forward(desc->storage_size, 16) + stack_size + 16;
stack_size + 16;
desc->stack_size = stack_size; /* This is just a hint, it won't be the real one. */ desc->stack_size = stack_size; /* This is just a hint, it won't be the real one. */
} }
@ -1390,7 +1429,10 @@ typedef struct _mco_context {
static void _mco_jumpin(mco_coro* co) { static void _mco_jumpin(mco_coro* co) {
void* cur_fib = GetCurrentFiber(); void* cur_fib = GetCurrentFiber();
if(!cur_fib || cur_fib == (void*)0x1e00) { /* See http://blogs.msdn.com/oldnewthing/archive/2004/12/31/344799.aspx */ if (!cur_fib ||
cur_fib ==
(void*)0x1e00) { /* See http://blogs.msdn.com/oldnewthing/archive/2004/12/31/344799.aspx
*/
cur_fib = ConvertThreadToFiber(NULL); cur_fib = ConvertThreadToFiber(NULL);
} }
MCO_ASSERT(cur_fib != NULL); MCO_ASSERT(cur_fib != NULL);
@ -1437,7 +1479,8 @@ static mco_result _mco_create_context(mco_coro* co, mco_desc* desc) {
/* Initialize storage. */ /* Initialize storage. */
unsigned char* storage = (unsigned char*)storage_addr; unsigned char* storage = (unsigned char*)storage_addr;
/* Create the fiber. */ /* Create the fiber. */
_mco_fiber* fib = (_mco_fiber*)CreateFiberEx(desc->stack_size, desc->stack_size, FIBER_FLAG_FLOAT_SWITCH, _mco_wrap_main, co); _mco_fiber* fib = (_mco_fiber*)CreateFiberEx(desc->stack_size, desc->stack_size,
FIBER_FLAG_FLOAT_SWITCH, _mco_wrap_main, co);
if (!fib) { if (!fib) {
MCO_LOG("failed to create fiber"); MCO_LOG("failed to create fiber");
return MCO_MAKE_CONTEXT_ERROR; return MCO_MAKE_CONTEXT_ERROR;
@ -1462,8 +1505,7 @@ static void _mco_destroy_context(mco_coro* co) {
static MCO_FORCE_INLINE void _mco_init_desc_sizes(mco_desc* desc, size_t stack_size) { static MCO_FORCE_INLINE void _mco_init_desc_sizes(mco_desc* desc, size_t stack_size) {
desc->coro_size = _mco_align_forward(sizeof(mco_coro), 16) + desc->coro_size = _mco_align_forward(sizeof(mco_coro), 16) +
_mco_align_forward(sizeof(_mco_context), 16) + _mco_align_forward(sizeof(_mco_context), 16) +
_mco_align_forward(desc->storage_size, 16) + _mco_align_forward(desc->storage_size, 16) + 16;
16;
desc->stack_size = stack_size; desc->stack_size = stack_size;
} }
@ -1493,7 +1535,8 @@ static void _mco_jumpin(mco_coro* co) {
emscripten_fiber_t* back_fib = running_fib; emscripten_fiber_t* back_fib = running_fib;
if (!back_fib) { if (!back_fib) {
back_fib = &main_fib; back_fib = &main_fib;
emscripten_fiber_init_from_current_context(back_fib, main_asyncify_stack, MCO_ASYNCFY_STACK_SIZE); emscripten_fiber_init_from_current_context(back_fib, main_asyncify_stack,
MCO_ASYNCFY_STACK_SIZE);
} }
running_fib = &context->fib; running_fib = &context->fib;
context->back_fib = back_fib; context->back_fib = back_fib;
@ -1530,7 +1573,8 @@ static mco_result _mco_create_context(mco_coro* co, mco_desc* desc) {
void* asyncify_stack_base = (void*)asyncify_stack_addr; void* asyncify_stack_base = (void*)asyncify_stack_addr;
size_t asyncify_stack_size = co_addr + desc->coro_size - asyncify_stack_addr; size_t asyncify_stack_size = co_addr + desc->coro_size - asyncify_stack_addr;
/* Create the fiber. */ /* Create the fiber. */
emscripten_fiber_init(&context->fib, _mco_wrap_main, co, stack_base, stack_size, asyncify_stack_base, asyncify_stack_size); emscripten_fiber_init(&context->fib, _mco_wrap_main, co, stack_base, stack_size,
asyncify_stack_base, asyncify_stack_size);
co->context = context; co->context = context;
co->stack_base = stack_base; co->stack_base = stack_base;
co->stack_size = stack_size; co->stack_size = stack_size;
@ -1545,12 +1589,10 @@ static void _mco_destroy_context(mco_coro* co) {
} }
static MCO_FORCE_INLINE void _mco_init_desc_sizes(mco_desc* desc, size_t stack_size) { static MCO_FORCE_INLINE void _mco_init_desc_sizes(mco_desc* desc, size_t stack_size) {
desc->coro_size = _mco_align_forward(sizeof(mco_coro), 16) + desc->coro_size =
_mco_align_forward(sizeof(_mco_context), 16) + _mco_align_forward(sizeof(mco_coro), 16) + _mco_align_forward(sizeof(_mco_context), 16) +
_mco_align_forward(desc->storage_size, 16) + _mco_align_forward(desc->storage_size, 16) + _mco_align_forward(stack_size, 16) +
_mco_align_forward(stack_size, 16) + _mco_align_forward(MCO_ASYNCFY_STACK_SIZE, 16) + 16;
_mco_align_forward(MCO_ASYNCFY_STACK_SIZE, 16) +
16;
desc->stack_size = stack_size; /* This is just a hint, it won't be the real one. */ desc->stack_size = stack_size; /* This is just a hint, it won't be the real one. */
} }
@ -1576,9 +1618,11 @@ typedef struct _mco_context {
_asyncify_stack_region stack_region; _asyncify_stack_region stack_region;
} _mco_context; } _mco_context;
__attribute__((import_module("asyncify"), import_name("start_unwind"))) void _asyncify_start_unwind(void*); __attribute__((import_module("asyncify"), import_name("start_unwind"))) void _asyncify_start_unwind(
void*);
__attribute__((import_module("asyncify"), import_name("stop_unwind"))) void _asyncify_stop_unwind(); __attribute__((import_module("asyncify"), import_name("stop_unwind"))) void _asyncify_stop_unwind();
__attribute__((import_module("asyncify"), import_name("start_rewind"))) void _asyncify_start_rewind(void*); __attribute__((import_module("asyncify"), import_name("start_rewind"))) void _asyncify_start_rewind(
void*);
__attribute__((import_module("asyncify"), import_name("stop_rewind"))) void _asyncify_stop_rewind(); __attribute__((import_module("asyncify"), import_name("stop_rewind"))) void _asyncify_stop_rewind();
MCO_NO_INLINE void _mco_jumpin(mco_coro* co) { MCO_NO_INLINE void _mco_jumpin(mco_coro* co) {
@ -1594,7 +1638,8 @@ MCO_NO_INLINE void _mco_jumpin(mco_coro* co) {
static MCO_NO_INLINE void _mco_finish_jumpout(mco_coro* co, volatile int rewind_id) { static MCO_NO_INLINE void _mco_finish_jumpout(mco_coro* co, volatile int rewind_id) {
_mco_context* context = (_mco_context*)co->context; _mco_context* context = (_mco_context*)co->context;
int next_rewind_id = context->rewind_id + 1; int next_rewind_id = context->rewind_id + 1;
if(rewind_id == next_rewind_id) { /* Begins unwinding the stack (save locals and call stack to rewind later) */ if (rewind_id == next_rewind_id) { /* Begins unwinding the stack (save locals and call stack to
rewind later) */
_mco_prepare_jumpout(co); _mco_prepare_jumpout(co);
context->rewind_id = next_rewind_id; context->rewind_id = next_rewind_id;
_asyncify_start_unwind(&context->stack_region); _asyncify_start_unwind(&context->stack_region);
@ -1645,11 +1690,9 @@ static void _mco_destroy_context(mco_coro* co) {
} }
static MCO_FORCE_INLINE void _mco_init_desc_sizes(mco_desc* desc, size_t stack_size) { static MCO_FORCE_INLINE void _mco_init_desc_sizes(mco_desc* desc, size_t stack_size) {
desc->coro_size = _mco_align_forward(sizeof(mco_coro), 16) + desc->coro_size =
_mco_align_forward(sizeof(_mco_context), 16) + _mco_align_forward(sizeof(mco_coro), 16) + _mco_align_forward(sizeof(_mco_context), 16) +
_mco_align_forward(desc->storage_size, 16) + _mco_align_forward(desc->storage_size, 16) + _mco_align_forward(stack_size, 16) + 16;
_mco_align_forward(stack_size, 16) +
16;
desc->stack_size = stack_size; /* This is just a hint, it won't be the real one. */ desc->stack_size = stack_size; /* This is just a hint, it won't be the real one. */
} }
@ -1823,7 +1866,8 @@ mco_result mco_yield(mco_coro* co) {
size_t stack_addr = (size_t)&dummy; size_t stack_addr = (size_t)&dummy;
size_t stack_min = (size_t)co->stack_base; size_t stack_min = (size_t)co->stack_base;
size_t stack_max = stack_min + co->stack_size; size_t stack_max = stack_min + co->stack_size;
if(co->magic_number != MCO_MAGIC_NUMBER || stack_addr < stack_min || stack_addr > stack_max) { /* Stack overflow. */ if (co->magic_number != MCO_MAGIC_NUMBER || stack_addr < stack_min ||
stack_addr > stack_max) { /* Stack overflow. */
MCO_LOG("coroutine stack overflow, try increasing the stack size"); MCO_LOG("coroutine stack overflow, try increasing the stack size");
return MCO_STACK_OVERFLOW; return MCO_STACK_OVERFLOW;
} }