valgrind 错误和 ucontext。为什么“使用大小为 8 的未初始化值”？

Question

我一直试图理解为什么 valgrind 抱怨这个使用 ucontexts 的小型测试程序“使用未初始化的大小为 8 的值”。 它基本上是一个创建“n_ucs”ucontexts 并在“max_switch”次切换它们的程序。

我了解“警告：客户端切换堆栈？” （这基本上是程序的全部内容），但我无法理解所有“使用大小为 8 的未初始化值”

我想获得一些帮助，了解 Valgrind 错误是误报还是该程序存在根本性错误。 （我在一个使用相同机制的更大程序上看到了很多，但我已将其提炼到最低限度，以便在此处发布）。

感谢任何帮助。

谢谢，

杰克

#include <stdio.h>
#include <stdlib.h>
#include <sys/time.h>
#include <ucontext.h>

#define STACK_SIZE   (8*1024)

int n_ucs = 1;
int max_switchs = 10;
int n_switchs = 0;
int tid = 0;

ucontext_t *ucs;
static ucontext_t engine_uc;

static void func(int arg)
{
    while (n_switchs < max_switchs) {
        int c_tid = tid;
        int n_tid = (tid + 1) % n_ucs;
        n_switchs++;
        tid = n_tid;
        swapcontext(&ucs[c_tid], &ucs[n_tid]);
    }
}

int main(int argc, char **argv)
{
    if (argc > 1)
        n_ucs = atoi(argv[1]);
    if (argc > 2)
        max_switchs = atoi(argv[2]);

    ucs = malloc(sizeof(ucontext_t) * n_ucs);
    int i;
    for (i = 0; i < n_ucs; i++) {
        /* Create initial ucontext_t, including stack */
        getcontext(&ucs[i]);
        ucs[i].uc_stack.ss_sp = malloc(STACK_SIZE);
        ucs[i].uc_stack.ss_size = STACK_SIZE;
        ucs[i].uc_stack.ss_flags = 0;
        ucs[i].uc_link = &engine_uc;
        makecontext(&ucs[i], (void (*)())func, 1, i);
    }

    /* jump to the first uc */
    swapcontext(&engine_uc, &ucs[tid]);

    /* destroy stacks */
    for (i = 0; i < n_ucs; i++)
        free(ucs[i].uc_stack.ss_sp);
    free(ucs);
    return 0;
}

使用 gcc main.c 编译并使用 ./a.out 2 2

运行

gcc -v

使用内置规范。 COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/4.8/lto-wrapper 目标：x86_64-linux-gnu 配置：../src/configure -v --with-pkgversion='Ubuntu 4.8.2-19ubuntu1' --with-bugurl=file:///usr/share/doc/gcc-4.8/README.Bugs --enable-languages=c,c++,java, go,d,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.8 --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without -included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.8 --libdir=/usr/lib --enable-nls --with-sysroot=/ - -enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-gnu-unique-object --disable-libmudflap --enable-plugin --with-system-zlib --禁用浏览器插件 --enable-java-awt=gtk --enable-gtk-cairo --with-java-home=/usr/lib/jvm/java-1.5.0-gcj-4.8-amd64/jre - -enable-java-home --with-jvm-root-dir=/usr/lib/jvm/java-1.5.0-gcj-4.8-amd64 --with-jvm-jar-dir=/usr/lib/jvm -exports/java-1.5.0-gcj-4.8-amd64 --with-arch-directory=amd64 --with-ecj-jar=/usr/share/java/eclipse-ecj.jar --enable-objc-gc --enable-multiarch --disable-werror --with-arch-32=i686 --with-abi=m64 --with-multilib-l ist=m32,m64,mx32 --with-tune=generic --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu 线程模型: posix gcc 版本 4.8.2 (Ubuntu 4.8.2-19ubuntu1)

ldd --version

ldd (Ubuntu EGLIBC 2.19-0ubuntu6.3) 2.19 版权所有 (C) 2014 Free Software Foundation, Inc. 这是免费软件；见 复制条件的来源。没有保修；甚至不为 适销性或特定用途的适用性。作者：罗兰 McGrath 和 Ulrich Drepper。

valgrind --track-origins=yes ./a.out 2 2
==21949== Memcheck, a memory error detector
==21949== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
==21949== Using Valgrind-3.10.0.SVN and LibVEX; rerun with -h for copyright info
==21949== Command: ./a.out 2 2
==21949==
==21949== Warning: client switching stacks?  SP change: 0xffefffdd8 --> 0x51ff7b8
==21949==          to suppress, use: --max-stackframe=68616717856 or greater
==21949== Use of uninitialised value of size 8
==21949==    at 0x400738: func (main.c:25)
==21949==    by 0x4E58EC4: (below main) (libc-start.c:287)
==21949==  Uninitialised value was created by a stack allocation
==21949==    at 0x4E7E445: swapcontext (swapcontext.S:92)
==21949==
==21949== Conditional jump or move depends on uninitialised value(s)
==21949==    at 0x4E807A7: __start_context (__start_context.S:37)
==21949==    by 0x4E58EC4: (below main) (libc-start.c:287)
==21949==  Uninitialised value was created by a stack allocation
==21949==    at 0x4E7E445: swapcontext (swapcontext.S:92)
==21949==
==21949== Syscall param rt_sigprocmask(set) contains uninitialised byte(s)
==21949==    at 0x4E7E0EC: setcontext (setcontext.S:47)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==  Uninitialised value was created by a stack allocation
==21949==    at 0x4E7E445: swapcontext (swapcontext.S:92)
==21949==
==21949== Use of uninitialised value of size 8
==21949==    at 0x4E7E0F5: setcontext (setcontext.S:54)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==  Uninitialised value was created by a stack allocation
==21949==    at 0x4E7E445: swapcontext (swapcontext.S:92)
==21949==
==21949== Use of uninitialised value of size 8
==21949==    at 0x4E7E0FE: setcontext (setcontext.S:56)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==    by 0x4E807AD: __start_context (__start_context.S:39)
==21949==  Uninitialised value was created by a stack allocation
==21949==    at 0x4E7E445: swapcontext (swapcontext.S:92)
==21949==
==21949== Warning: client switching stacks?  SP change: 0x51ff7c0 --> 0xffefffde0
==21949==          to suppress, use: --max-stackframe=68616717856 or greater
==21949==
==21949== HEAP SUMMARY:
==21949==     in use at exit: 0 bytes in 0 blocks
==21949==   total heap usage: 3 allocs, 3 frees, 18,256 bytes allocated
==21949==
==21949== All heap blocks were freed -- no leaks are possible
==21949==
==21949== For counts of detected and suppressed errors, rerun with: -v
==21949== ERROR SUMMARY: 5 errors from 5 contexts (suppressed: 0 from 0)

Answer 1

你必须通知 valgrind 堆栈的变化。在此处阅读示例https://github.com/lu-zero/valgrind/blob/master/memcheck/tests/linux/stack_changes.c

这是正确的代码：

#include <stdio.h>
#include <stdlib.h>
#include <sys/time.h>
#include <ucontext.h>
#include <valgrind/valgrind.h>

#define STACK_SIZE   (8*1024)

int n_ucs = 1;
int max_switchs = 10;
int n_switchs = 0;
int tid = 0;

ucontext_t *ucs;
static ucontext_t engine_uc;

 void func(int arg)
{
    while (n_switchs < max_switchs) {
        int c_tid = tid;
        int n_tid = (tid + 1) % n_ucs;
        n_switchs++;
        tid = n_tid;
        swapcontext(&ucs[c_tid], &ucs[n_tid]);

    }
}

int main(int argc, char **argv)
{
    if (argc > 1)
        n_ucs = atoi(argv[1]);
    if (argc > 2)
        max_switchs = atoi(argv[2]);

    ucs = malloc(sizeof(ucontext_t) * n_ucs);

    //store the VALGRIND_STACK_REGISTER return values
    int* valgrind_ret = malloc(n_ucs*sizeof(int));

    int i;
    for (i = 0; i < n_ucs; i++) {
        /* Create initial ucontext_t, including stack */
        getcontext(&ucs[i]);

        //pass stack to valgrind
        void* mystack = malloc(STACK_SIZE);
        VALGRIND_STACK_REGISTER(mystack, mystack + STACK_SIZE);

        ucs[i].uc_stack.ss_sp = mystack;
        ucs[i].uc_stack.ss_size = STACK_SIZE;
        ucs[i].uc_stack.ss_flags = 0;
        ucs[i].uc_link = &engine_uc;
        makecontext(&ucs[i], (void (*)())func, 1, i);
    }

    /* jump to the first uc */
    swapcontext(&engine_uc, &ucs[tid]);

    /* destroy stacks */
    for (i = 0; i < n_ucs; i++) {
        //valgrind stack deregister 
        VALGRIND_STACK_DEREGISTER(valgrind_ret[i]);

        free(ucs[i].uc_stack.ss_sp);
    }
    free(ucs);
    return 0;
}

Answer 2

我仍然不完全理解为什么 valgrind 会显示这些未初始化的错误完全正确，但我会尽力解释我到目前为止所理解的；

在通过 valgrind 运行和分析程序时，根据来自 swapcontext(3) 和 getcontext(3) 手册页的信息，我认为它未能检测到某些上下文交换（未能从tid 0 到 tid 1 以及从 tid 1 回到 tid 0 的 swapcontext)

如下阅读：谁的堆栈[调用次数]：函数调用

所以，我认为函数调用跟踪是这样的：

main:swapcontext(main, tid 0) ->

main[tid 0's 1st func call]:func() ->

tid 0:swapcontext(tid 0, tid 1) -> {Stack => tiod 0}

tid 1:func() ->

swapcontext(tid 1, tid 0) -> {堆栈 => tiod 1}

tid 0[第二次调用]: func() ->

n_switchs = 2 后立即返回 ->

pop tid 0[第二次调用]：来自 tid 1 堆栈的 func() 堆栈帧 -> {1st 根据 valgrind 的未初始化访问}

tid 0[第二次调用]：func() 完成 -> 检查 uc_link；找到engine_uc （主要上下文）设置在那里 ->

从这里开始对我来说事情变得不清楚，但以下似乎是可能的踪迹：

重置 sigprocmask -> {2nd Uninitialized access} setcontext()s 返回主上下文 -> {3rd Uinitialized access ?} {Stack => 主要}

返回时，[tid 0's 1st call] 的堆栈帧从 main 弹出 堆栈->

main [tid 0's 1st call]:func() 也因为 n_switchs = 而完成 2 -> 检查 uc_link；再次找到 engine_uc -> 重置 sigprocmask -> {不是未初始化的访问？}

返回时，main:swapcontext() 的堆栈帧从 main 中弹出 堆栈->

setcontext()s 返回主上下文 -> {4th Uninitialized access ?} {堆栈 => 主}

我们回到main()，释放东西并退出

一些参考资料：

https://www.gnu.org/software/libc/manual/html_node/System-V-contexts.html http://www.cs.uwm.edu/classes/cs315/Bacon/Lecture/HTML/ch10s07.html

注意：我知道这不是一个完整的答案，但我不想在 cmets 部分发布这么长的解释；因此张贴在这里。