Add libsetjmp.a/so (#483)

* Add libsetjmp.a/so

Add setjmp/longjump support based on Wasm EH proposal.

It's provided as a separate library (libsetjmp) from libc so that
runtimes w/o EH support can still load libc.so.

To use this setjmp/longjmp implementation, an application should
be compiled with `-mllvm -wasm-enable-sjlj` and linked with `-lsetjmp`.
(You need an LLVM with the change mentioned below.)

Also, you need a runtime with EH support to run such an application.

If you want to use the latest EH instructions, you can use
`binaryen --translate-eh-old-to-new` on your application.

Note: You don't need to translate libsetjmp.a/so to the new EH.
While LLVM currently produces bytecode for an old version of the EH
proposal, luckily for us, the bytecode used in this library (ie. the tag
definition and the "throw" instruction) is compatible with the latest
version of the proposal.

The runtime logic is basically copy-and-paste from:
    https://github.com/yamt/garbage/tree/wasm-sjlj-alt2/wasm/longjmp

The corresponding LLVM change:
    llvm/llvm-project#84137
    (Note: you need this change to build setjmp/longjmp using code.
    otoh, you don't need this to build libsetjmp.)

A similar change for emscripten:
    emscripten-core/emscripten#21502

An older version of this PR, which doesn't require LLVM changes:
    #467

Discussion:
    https://docs.google.com/document/d/1ZvTPT36K5jjiedF8MCXbEmYjULJjI723aOAks1IdLLg/edit

An example to use the latest EH instructions:
```
clang -mllvm -wasm-enable-sjlj -o your_app.wasm your_app.c -lsetjmp
wasm-opt --translate-eh-old-to-new -o your_app.wasm your_app.wasm
toywasm --wasi your_app.wasm
```
Note: use toywasm built with `-DTOYWASM_ENABLE_WASM_EXCEPTION_HANDLING=ON`.

An example to use the older EH instructions, which LLVM currently produces:
```
clang -mllvm -wasm-enable-sjlj -o your_app.wasm your_app.c -lsetjmp
iwasm your_app.wasm
```
Note: use wasm-micro-runtime built with `-DWAMR_BUILD_EXCE_HANDLING=1`.
Note: as of writing this, only the classic interpreter supports EH.

* Make libsetjmp build optional

* CI: Disable libsetjmp for old LLVM

* libc-top-half/musl/include/setjmp.h: fix a rebase botch

Author: yamt

.github/workflows/main.yml

@@ -82,6 +82,12 @@ jobs:
         echo "NM=$CLANG_DIR/llvm-nm" >> $GITHUB_ENV
       if: matrix.os == 'ubuntu-latest'
 
+    - name: Disable libsetjmp for old LLVM
+      shell: bash
+      run: |
+        echo "BUILD_LIBSETJMP=no" >> $GITHUB_ENV
+      if: matrix.clang_version == '10.0.0'
+
     - name: Build libc
       shell: bash
       run: |

Makefile

@@ -21,6 +21,8 @@ WASI_SNAPSHOT ?= p1
 MALLOC_IMPL ?= dlmalloc
 # yes or no
 BUILD_LIBC_TOP_HALF ?= yes
+# yes or no
+BUILD_LIBSETJMP ?= yes
 # The directory where we will store intermediate artifacts.
 OBJDIR ?= build/$(TARGET_TRIPLE)
 # The directory where we store files and tools for generating WASIp2 bindings
@@ -95,6 +97,7 @@ LIBC_BOTTOM_HALF_OMIT_SOURCES := \
 	$(LIBC_BOTTOM_HALF_SOURCES)/poll-wasip2.c
 LIBC_BOTTOM_HALF_ALL_SOURCES := $(filter-out $(LIBC_BOTTOM_HALF_OMIT_SOURCES),$(LIBC_BOTTOM_HALF_ALL_SOURCES))
 # Omit p2-specific headers from include-all.c test.
+# for exception-handling.
 INCLUDE_ALL_CLAUSES := -not -name wasip2.h -not -name descriptor_table.h
 endif
 
@@ -128,6 +131,7 @@ LIBWASI_EMULATED_SIGNAL_MUSL_SOURCES = \
     $(LIBC_TOP_HALF_MUSL_SRC_DIR)/signal/psignal.c \
     $(LIBC_TOP_HALF_MUSL_SRC_DIR)/string/strsignal.c
 LIBDL_SOURCES = $(LIBC_TOP_HALF_MUSL_SRC_DIR)/misc/dl.c
+LIBSETJMP_SOURCES = $(LIBC_TOP_HALF_MUSL_SRC_DIR)/setjmp/wasm32/rt.c
 LIBC_BOTTOM_HALF_CRT_SOURCES = $(wildcard $(LIBC_BOTTOM_HALF_DIR)/crt/*.c)
 LIBC_TOP_HALF_DIR = libc-top-half
 LIBC_TOP_HALF_MUSL_DIR = $(LIBC_TOP_HALF_DIR)/musl
@@ -433,6 +437,7 @@ LIBWASI_EMULATED_GETPID_OBJS = $(call objs,$(LIBWASI_EMULATED_GETPID_SOURCES))
 LIBWASI_EMULATED_SIGNAL_OBJS = $(call objs,$(LIBWASI_EMULATED_SIGNAL_SOURCES))
 LIBWASI_EMULATED_SIGNAL_MUSL_OBJS = $(call objs,$(LIBWASI_EMULATED_SIGNAL_MUSL_SOURCES))
 LIBDL_OBJS = $(call objs,$(LIBDL_SOURCES))
+LIBSETJMP_OBJS = $(call objs,$(LIBSETJMP_SOURCES))
 LIBC_BOTTOM_HALF_CRT_OBJS = $(call objs,$(LIBC_BOTTOM_HALF_CRT_SOURCES))
 
 # These variables describe the locations of various files and
@@ -498,7 +503,6 @@ MUSL_OMIT_HEADERS += \
     "netdb.h" \
     "resolv.h" \
     "pty.h" \
-    "setjmp.h" \
     "ulimit.h" \
     "sys/xattr.h" \
     "wordexp.h" \
@@ -534,6 +538,7 @@ LIBWASI_EMULATED_GETPID_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBWASI_EMULATED_GETP
 LIBWASI_EMULATED_SIGNAL_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBWASI_EMULATED_SIGNAL_OBJS))
 LIBWASI_EMULATED_SIGNAL_MUSL_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBWASI_EMULATED_SIGNAL_MUSL_OBJS))
 LIBDL_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBDL_OBJS))
+LIBSETJMP_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBSETJMP_OBJS))
 BULK_MEMORY_SO_OBJS = $(patsubst %.o,%.pic.o,$(BULK_MEMORY_OBJS))
 DLMALLOC_SO_OBJS = $(patsubst %.o,%.pic.o,$(DLMALLOC_OBJS))
 LIBC_BOTTOM_HALF_ALL_SO_OBJS = $(patsubst %.o,%.pic.o,$(LIBC_BOTTOM_HALF_ALL_OBJS))
@@ -548,6 +553,7 @@ PIC_OBJS = \
 	$(LIBWASI_EMULATED_SIGNAL_SO_OBJS) \
 	$(LIBWASI_EMULATED_SIGNAL_MUSL_SO_OBJS) \
 	$(LIBDL_SO_OBJS) \
+	$(LIBSETJMP_SO_OBJS) \
 	$(BULK_MEMORY_SO_OBJS) \
 	$(DLMALLOC_SO_OBJS) \
 	$(LIBC_BOTTOM_HALF_ALL_SO_OBJS) \
@@ -577,6 +583,8 @@ $(OBJDIR)/libwasi-emulated-signal.so.a: $(LIBWASI_EMULATED_SIGNAL_SO_OBJS) $(LIB
 
 $(OBJDIR)/libdl.so.a: $(LIBDL_SO_OBJS)
 
+$(OBJDIR)/libsetjmp.so.a: $(LIBSETJMP_SO_OBJS)
+
 $(SYSROOT_LIB)/libc.a: $(LIBC_OBJS)
 
 $(SYSROOT_LIB)/libc-printscan-long-double.a: $(MUSL_PRINTSCAN_LONG_DOUBLE_OBJS)
@@ -593,6 +601,8 @@ $(SYSROOT_LIB)/libwasi-emulated-signal.a: $(LIBWASI_EMULATED_SIGNAL_OBJS) $(LIBW
 
 $(SYSROOT_LIB)/libdl.a: $(LIBDL_OBJS)
 
+$(SYSROOT_LIB)/libsetjmp.a: $(LIBSETJMP_OBJS)
+
 %.a:
 	@mkdir -p "$(@D)"
 	# On Windows, the commandline for the ar invocation got too long, so it needs to be split up.
@@ -622,6 +632,9 @@ $(BULK_MEMORY_OBJS) $(BULK_MEMORY_SO_OBJS): CFLAGS += \
 $(BULK_MEMORY_OBJS) $(BULK_MEMORY_SO_OBJS): CFLAGS += \
         -DBULK_MEMORY_THRESHOLD=$(BULK_MEMORY_THRESHOLD)
 
+$(LIBSETJMP_OBJS) $(LIBSETJMP_SO_OBJS): CFLAGS += \
+        -mllvm -wasm-enable-sjlj
+
 $(LIBWASI_EMULATED_SIGNAL_MUSL_OBJS) $(LIBWASI_EMULATED_SIGNAL_MUSL_SO_OBJS): CFLAGS += \
 	    -D_WASI_EMULATED_SIGNAL
 
@@ -665,7 +678,7 @@ startup_files $(LIBC_BOTTOM_HALF_ALL_OBJS) $(LIBC_BOTTOM_HALF_ALL_SO_OBJS): CFLA
     -I$(LIBC_TOP_HALF_MUSL_SRC_DIR)/include \
     -I$(LIBC_TOP_HALF_MUSL_SRC_DIR)/internal
 
-$(LIBC_TOP_HALF_ALL_OBJS) $(LIBC_TOP_HALF_ALL_SO_OBJS) $(MUSL_PRINTSCAN_LONG_DOUBLE_OBJS) $(MUSL_PRINTSCAN_LONG_DOUBLE_SO_OBJS) $(MUSL_PRINTSCAN_NO_FLOATING_POINT_OBJS) $(LIBWASI_EMULATED_SIGNAL_MUSL_OBJS) $(LIBWASI_EMULATED_SIGNAL_MUSL_SO_OBJS) $(LIBDL_OBJS) $(LIBDL_SO_OBJS): CFLAGS += \
+$(LIBC_TOP_HALF_ALL_OBJS) $(LIBC_TOP_HALF_ALL_SO_OBJS) $(MUSL_PRINTSCAN_LONG_DOUBLE_OBJS) $(MUSL_PRINTSCAN_LONG_DOUBLE_SO_OBJS) $(MUSL_PRINTSCAN_NO_FLOATING_POINT_OBJS) $(LIBWASI_EMULATED_SIGNAL_MUSL_OBJS) $(LIBWASI_EMULATED_SIGNAL_MUSL_SO_OBJS) $(LIBDL_OBJS) $(LIBDL_SO_OBJS) $(LIBSETJMP_OBJS) $(LIBSETJMP_SO_OBJS): CFLAGS += \
     -I$(LIBC_TOP_HALF_MUSL_SRC_DIR)/include \
     -I$(LIBC_TOP_HALF_MUSL_SRC_DIR)/internal \
     -I$(LIBC_TOP_HALF_MUSL_DIR)/arch/wasm32 \
@@ -732,11 +745,15 @@ LIBC_SO = \
 	$(SYSROOT_LIB)/libwasi-emulated-getpid.so \
 	$(SYSROOT_LIB)/libwasi-emulated-signal.so \
 	$(SYSROOT_LIB)/libdl.so
+ifeq ($(BUILD_LIBSETJMP),yes)
+LIBC_SO += \
+	$(SYSROOT_LIB)/libsetjmp.so
+endif
 endif
 
 libc_so: include_dirs $(LIBC_SO)
 
-libc: include_dirs \
+STATIC_LIBS = \
     $(SYSROOT_LIB)/libc.a \
     $(SYSROOT_LIB)/libc-printscan-long-double.a \
     $(SYSROOT_LIB)/libc-printscan-no-floating-point.a \
@@ -745,6 +762,12 @@ libc: include_dirs \
     $(SYSROOT_LIB)/libwasi-emulated-getpid.a \
     $(SYSROOT_LIB)/libwasi-emulated-signal.a \
     $(SYSROOT_LIB)/libdl.a
+ifeq ($(BUILD_LIBSETJMP),yes)
+STATIC_LIBS += \
+	$(SYSROOT_LIB)/libsetjmp.a
+endif
+
+libc: include_dirs $(STATIC_LIBS)
 
 finish: startup_files libc
 	#
@@ -804,8 +827,10 @@ check-symbols: startup_files libc
 	#
 	# Generate a test file that includes all public C header files.
 	#
+	# setjmp.h is excluded because it requires a different compiler option
+	#
 	cd "$(SYSROOT_INC)" && \
-	  for header in $$(find . -type f -not -name mman.h -not -name signal.h -not -name times.h -not -name resource.h $(INCLUDE_ALL_CLAUSES) |grep -v /bits/ |grep -v /c++/); do \
+	  for header in $$(find . -type f -not -name mman.h -not -name signal.h -not -name times.h -not -name resource.h -not -name setjmp.h $(INCLUDE_ALL_CLAUSES) |grep -v /bits/ |grep -v /c++/); do \
 	      echo '#include <'$$header'>' | sed 's/\.\///' ; \
 	done |LC_ALL=C sort >$(SYSROOT_SHARE)/include-all.c ; \
 	cd - >/dev/null

libc-top-half/musl/arch/wasm32/bits/setjmp.h

@@ -0,0 +1 @@
+typedef unsigned long __jmp_buf[8];

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

@@ -7,7 +7,12 @@ extern "C" {
 
 #include <features.h>
 
-#ifdef __wasilibc_unmodified_upstream /* WASI has no setjmp */
+#ifndef __wasilibc_unmodified_upstream
+/* WASI has no setjmp */
+#if !defined(__wasm_exception_handling__)
+#error Setjmp/longjmp support requires Exception handling support, which is [not yet standardized](https://github.com/WebAssembly/proposals?tab=readme-ov-file#phase-3---implementation-phase-cg--wg). To enable it, compile with `-mllvm -wasm-enable-sjlj` and use an engine that implements the Exception handling proposal.
+#endif
+#endif
 #include <bits/setjmp.h>
 
 typedef struct __jmp_buf_tag {
@@ -40,9 +45,6 @@ int setjmp (jmp_buf) __setjmp_attr;
 _Noreturn void longjmp (jmp_buf, int);
 
 #define setjmp setjmp
-#else
-#warning setjmp is not yet implemented for WASI
-#endif
 
 #undef __setjmp_attr
 

libc-top-half/musl/src/setjmp/wasm32/rt.c

@@ -0,0 +1,83 @@
+/*
+ * a runtime implementation for
+ * https://github.com/llvm/llvm-project/pull/84137
+ * https://docs.google.com/document/d/1ZvTPT36K5jjiedF8MCXbEmYjULJjI723aOAks1IdLLg/edit
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+
+/*
+ * function prototypes
+ */
+void __wasm_setjmp(void *env, uint32_t label, void *func_invocation_id);
+uint32_t __wasm_setjmp_test(void *env, void *func_invocation_id);
+void __wasm_longjmp(void *env, int val);
+
+/*
+ * jmp_buf should have large enough size and alignment to contain
+ * this structure.
+ */
+struct jmp_buf_impl {
+        void *func_invocation_id;
+        uint32_t label;
+
+        /*
+         * this is a temorary storage used by the communication between
+         * __wasm_sjlj_longjmp and WebAssemblyLowerEmscriptenEHSjL-generated
+         * logic.
+         * ideally, this can be replaced with multivalue.
+         */
+        struct arg {
+                void *env;
+                int val;
+        } arg;
+};
+
+void
+__wasm_setjmp(void *env, uint32_t label, void *func_invocation_id)
+{
+        struct jmp_buf_impl *jb = env;
+        if (label == 0) { /* ABI contract */
+                __builtin_trap();
+        }
+        if (func_invocation_id == NULL) { /* sanity check */
+                __builtin_trap();
+        }
+        jb->func_invocation_id = func_invocation_id;
+        jb->label = label;
+}
+
+uint32_t
+__wasm_setjmp_test(void *env, void *func_invocation_id)
+{
+        struct jmp_buf_impl *jb = env;
+        if (jb->label == 0) { /* ABI contract */
+                __builtin_trap();
+        }
+        if (func_invocation_id == NULL) { /* sanity check */
+                __builtin_trap();
+        }
+        if (jb->func_invocation_id == func_invocation_id) {
+                return jb->label;
+        }
+        return 0;
+}
+
+void
+__wasm_longjmp(void *env, int val)
+{
+        struct jmp_buf_impl *jb = env;
+        struct arg *arg = &jb->arg;
+        /*
+         * C standard says:
+         * The longjmp function cannot cause the setjmp macro to return
+         * the value 0; if val is 0, the setjmp macro returns the value 1.
+         */
+        if (val == 0) {
+                val = 1;
+        }
+        arg->env = env;
+        arg->val = val;
+        __builtin_wasm_throw(1, arg); /* 1 == C_LONGJMP */
+}