c - 嵌套线程创建期间的数据竞争-6ren

c - 嵌套线程创建期间的数据竞争

转载作者：太空宇宙更新时间：2023-11-04 08:10:10

在尝试创建嵌套线程时，helgrind 报告了几种不同类型的数据争用。

==4429== Possible data race during write of size 8 at 0x5673830 by thread #13
==4429== Locks held: none
==4429==    at 0x4C379EF: memset (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x5060C85: get_cached_stack (allocatestack.c:250)
==4429==    by 0x5060C85: allocate_stack (allocatestack.c:501)
==4429==    by 0x5060C85: pthread_create@@GLIBC_2.2.5 (pthread_create.c:537)
==4429==    by 0x4C32BF7: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x4022D7: read_group (c_esp.c:318)
==4429==    by 0x4C32DF6: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x50606A9: start_thread (pthread_create.c:333)
==4429==  Address 0x5673830 is 16 bytes inside a block of size 560 alloc'd
==4429==    at 0x4C2EFB5: calloc (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x40134C4: allocate_dtv (dl-tls.c:322)
==4429==    by 0x40134C4: _dl_allocate_tls (dl-tls.c:544)
==4429==    by 0x50610D2: allocate_stack (allocatestack.c:588)
==4429==    by 0x50610D2: pthread_create@@GLIBC_2.2.5 (pthread_create.c:537)
==4429==    by 0x4C32BF7: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x4022D7: read_group (c_esp.c:318)
==4429==    by 0x4C32DF6: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x50606A9: start_thread (pthread_create.c:333)
==4429==  Block was alloc'd by thread #3


==4429== Possible data race during write of size 1 at 0x724368F by thread #13
==4429== Locks held: none
==4429==    at 0x4C3856C: mempcpy (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x40132F6: _dl_allocate_tls_init (dl-tls.c:520)
==4429==    by 0x5060C8D: get_cached_stack (allocatestack.c:253)
==4429==    by 0x5060C8D: allocate_stack (allocatestack.c:501)
==4429==    by 0x5060C8D: pthread_create@@GLIBC_2.2.5 (pthread_create.c:537)
==4429==    by 0x4C32BF7: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x4022D7: read_group (c_esp.c:318)
==4429==    by 0x4C32DF6: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x50606A9: start_thread (pthread_create.c:333)
==4429== 
==4429== This conflicts with a previous write of size 1 by thread #4
==4429== Locks held: none
==4429==    at 0x5060612: start_thread (pthread_create.c:265)
==4429==  Address 0x724368f is in a rw- anonymous segment


==4429== Possible data race during read of size 8 at 0x7243728 by thread #14
==4429== Locks held: none
==4429==    at 0x40178C: read_record (c_esp.c:171)
==4429==    by 0x4C32DF6: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x50606A9: start_thread (pthread_create.c:333)
==4429== 
==4429== This conflicts with a previous write of size 8 by thread #13
==4429== Locks held: none
==4429==    at 0x5060DB7: pthread_create@@GLIBC_2.2.5 (pthread_create.c:589)
==4429==    by 0x4C32BF7: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x4022D7: read_group (c_esp.c:318)
==4429==    by 0x4C32DF6: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==4429==    by 0x50606A9: start_thread (pthread_create.c:333)
==4429==  Address 0x7243728 is in a rw- anonymous segment

提到的行是:

318: pthread_create(pth_array + k, &attrs, read_record, args_array +k);
171: void *read_record(void *_args){

所以看起来在创建线程时存在数据竞争？为“不同分支”中的线程分配的堆栈是否可能重叠？还是我搞砸了什么地方？

我已经尝试减少每个线程的堆栈大小，但没有成功。

根据要求，一个最小的可重现示例:

#include <pthread.h>

void *func_b(void *args){
    return args;
}

void *func_a(void *args){
    pthread_t pth[5];
    int i;
    for(i = 0; i < 5; i++){
        pthread_create(&pth[i], NULL, func_b, NULL);
    }

    for(i = 0; i < 5; i++){
        pthread_join(pth[i], NULL);
    }
    return args;
}

int main(void){
    pthread_t pth[100];
    int i;
    for(i = 0; i < 100; i++){
        pthread_create(&pth[i], NULL, func_a, NULL);
    }

    for(i = 0; i < 100; i++){
        pthread_join(pth[i], NULL);
    }
}

helgrind 的完整输出是 here接下来是所有相关代码。此片段从第 171 行到第 477 行。

void *read_record(void *_args){
    struct _thread_args *args = (struct _thread_args *)_args;
    uint32_t data_size = 0;

    args->item->record->type = calloc(5, sizeof *args->item->record->type);
    memcpy(args->item->record->type, *args->map, 4);
    *args->map += 4;
    memcpy(&data_size, *args->map, 4);
    *args->map += 4;
    memcpy(&args->item->record->flags, *args->map, 4);
    *args->map += 4;
    memcpy(&args->item->record->id, *args->map, 4);
    *args->map += 4;
    memcpy(&args->item->record->revision, *args->map, 4);
    *args->map += 4;
    memcpy(&args->item->record->version, *args->map, 2);
    *args->map += 2;
    memcpy(&args->item->record->unknown, *args->map, 2);
    *args->map += 2;

    args->item->record->misc_data = NULL;
    args->item->record->children = NULL;
    args->item->record->last = NULL;
    // args->item->record->next = NULL;
    // args->item->record->previous = NULL;
    args->item->record->_proxy = NULL;
    args->item->record->compression_level = 0x7f;  // some invalid number, doesn't matter

    if(args->parse_records){
        args->item->record->is_parsed = TRUE;
        if(args->item->record->flags & 0x00040000){
            uLongf uncompressed_size = 0;
            memcpy(&uncompressed_size, *args->map, 4);
            *args->map += 4;

            if(*(*args->map + 1) & (1 << 6)){
                if(*(*args->map + 1) & (1 << 7)){
                    args->item->record->compression_level = Z_BEST_COMPRESSION;  // both hight bits are set  11

                } else{
                    args->item->record->compression_level = Z_BEST_SPEED;  // 01
                }
            } else{
                if(*(*args->map + 1) & (1 << 7)){
                    args->item->record->compression_level = Z_DEFAULT_COMPRESSION;  // 10

                } else{
                    args->item->record->compression_level = Z_NO_COMPRESSION;  // 00
                }
            }

            uint8_t *uncompressed_data = malloc(uncompressed_size * sizeof *uncompressed_data);
            uint8_t *start = uncompressed_data;
            uncompress(uncompressed_data, &uncompressed_size, *args->map, data_size - 4);
            parse_record_data((uint8_t **)&uncompressed_data, uncompressed_size, &args->item->record->children, &args->item->record->last);
            *args->map += data_size - 4;
            free(start);
        } else{
            parse_record_data(args->map, data_size, &args->item->record->children, &args->item->record->last);
        }
    } else{
        args->item->record->is_parsed = FALSE;
        args->item->record->misc_data = malloc(sizeof *args->item->record->misc_data);
        args->item->record->misc_data->data = malloc(data_size * sizeof *args->item->record->misc_data->data);
        args->item->record->misc_data->data_size = data_size;
        memcpy(args->item->record->misc_data->data, *args->map, data_size);
        *args->map += data_size;
    }

    return NULL;
}

void *read_group(void *_args){
    struct _thread_args *args = (struct _thread_args *)_args;
    uint32_t data_size = 0;

    args->item->group->type = "GRUP";
    *args->map += 4;
    memcpy(&data_size, *args->map, 4);
    *args->map += 4;
    memset(args->item->group->label, 0, 5);
    memcpy(args->item->group->label, *args->map, 4);
    *args->map += 4;
    memcpy(&args->item->group->group_type, *args->map, 4);
    *args->map += 4;
    memcpy(&args->item->group->stamp, *args->map, 2);
    *args->map += 2;
    memcpy(&args->item->group->unknown1, *args->map, 2);
    *args->map += 2;
    memcpy(&args->item->group->version, *args->map, 2);
    *args->map += 2;
    memcpy(&args->item->group->unknown2, *args->map, 2);
    *args->map += 2;

    args->item->group->children = NULL;
    args->item->group->last = NULL;
    args->item->group->_proxy = NULL;

    uint8_t *data_start = *args->map;
    int item_count = 0;
    while(*args->map < (data_start + data_size - 24)){
        item_count++;
        char next_type[5] = {0};
        memcpy(next_type, *args->map, 4);
        uint32_t next_data_size = 0;
        memcpy(&next_data_size, *args->map + 4, 4);
        *args->map += next_data_size;
        if(strcmp(next_type, "GRUP") != 0){
            *args->map += 24;
        }
    }
    *args->map = data_start;

    pthread_attr_t  attrs;
    pthread_attr_init(&attrs);
    pthread_attr_setstacksize(&attrs, 1000);

    struct _group_item *current = NULL;
    bool current_is_record = FALSE;
    pthread_t *pth_array = malloc(item_count * sizeof *pth_array);
    uint8_t **map_array = malloc(item_count * sizeof *map_array);
    struct _thread_args *args_array = malloc(item_count * sizeof *args_array);
    int k;
    for(k = 0; k < item_count; k++){
        char next_type[5] = {0};
        memcpy(next_type, *args->map, 4);
        uint32_t new_data_size = 0;
        memcpy(&new_data_size, *args->map + 4, 4);
        *(map_array + k) = *args->map;
        *args->map += new_data_size;

        (*(args_array + k)).map = map_array + k;
        (*(args_array + k)).parse_records = args->parse_records;
        (*(args_array + k)).item = malloc(sizeof *(*(args_array + k)).item);
        if(strcmp(next_type, "GRUP") == 0){
            (*(args_array + k)).is_record = FALSE;
            (*(args_array + k)).item->group = malloc(sizeof *(*(args_array + k)).item->group);
            pthread_create(pth_array + k, &attrs, read_group, args_array + k);
            //read_group(args_array + k);
        } else{
            *args->map += 24;
            (*(args_array + k)).is_record = TRUE;
            (*(args_array + k)).item->record = malloc(sizeof *(*(args_array + k)).item->record);
            pthread_create(pth_array + k, &attrs, read_record, args_array + k);
            //read_record(args_array + k);
        }

        if(args->item->group->children == NULL){
            args->item->group->children = current = (*(args_array + k)).item;
            args->item->group->children_is_record = current_is_record = (*(args_array + k)).is_record;
            if((*(args_array + k)).is_record){
                current->record->previous = NULL;
            } else{
                current->group->previous = NULL;
            }
            continue;
        }

        if((*(args_array + k)).is_record){
            if(current_is_record){
                current->record->next = (*(args_array + k)).item;
                current->record->next_is_record = (*(args_array + k)).is_record;
                (*(args_array + k)).item->record->previous = current;
                (*(args_array + k)).item->record->previous_is_record = current_is_record;
            } else{
                current->group->next = (*(args_array + k)).item;
                current->group->next_is_record = (*(args_array + k)).is_record;
                (*(args_array + k)).item->record->previous = current;
                (*(args_array + k)).item->record->previous_is_record = current_is_record;
            }
        } else{
            if(current_is_record){
                current->record->next = (*(args_array + k)).item;
                current->record->next_is_record = (*(args_array + k)).is_record;
                (*(args_array + k)).item->group->previous = current;
                (*(args_array + k)).item->group->previous_is_record = current_is_record;
            } else{
                current->group->next = (*(args_array + k)).item;
                current->group->next_is_record = (*(args_array + k)).is_record;
                (*(args_array + k)).item->group->previous = current;
                (*(args_array + k)).item->group->previous_is_record = current_is_record;
            }
        }

        current = (*(args_array + k)).item;
        current_is_record = (*(args_array + k)).is_record;
    }

    args->item->group->last = current;
    args->item->group->last_is_record = current_is_record;
    if(current != NULL){
        if(current_is_record){
            current->record->next = NULL;
        } else{
            current->group->next = NULL;
        }
    }

    int i;
    for(i = 0; i < item_count; i++){
       pthread_join(*(pth_array + i), NULL);
    }

    free(pth_array);
    free(map_array);
    free(args_array);

    assert(*args->map == (data_start + data_size - 24));
    return NULL;
}

Plugin *plugin_read(const char *filename, const bool parse_records){
    Plugin *plugin = malloc(sizeof *plugin);
    plugin->children = NULL;
    plugin->last = NULL;
    plugin->_proxy = NULL;

    FILE *fileobject = fopen(filename, "rb");
    if(fileobject == NULL){
        printf("No file found.\n");
        return NULL;
    }
    fseek(fileobject, 0, SEEK_END);
    long int filesize = ftell(fileobject);
    rewind(fileobject);
    uint8_t *map = malloc(filesize * sizeof *map);
    fread(map, filesize, 1, fileobject);
    fclose(fileobject);

    uint8_t *start = map;

    pthread_attr_t  attrs;
    pthread_attr_init(&attrs);
    pthread_attr_setstacksize(&attrs, 1000);

    pthread_t header_pth;
    plugin->header = malloc(sizeof *plugin->header);
    struct _group_item header_item;
    header_item.record = plugin->header;
    uint8_t *header_map = map;
    struct _thread_args header_args;
    header_args.item = &header_item;
    header_args.map = &header_map;
    header_args.parse_records = parse_records;
    pthread_create(&header_pth, &attrs, read_record, &header_args);
    uint32_t header_data_size = 0;
    memcpy(&header_data_size, map + 4, 4);
    map += header_data_size + 24;

    uint8_t *after_header = map;
    int group_count = 0;
    while(map < (start + filesize)){
        group_count++;
        uint32_t next_data_size = 0;
        memcpy(&next_data_size, map + 4, 4);
        map += next_data_size;
    }
    map = after_header;

    pthread_t *pth = malloc(group_count * sizeof *pth);
    uint8_t **map_array = malloc(group_count * sizeof *map_array);
    struct _thread_args *args = malloc(group_count * sizeof *args);
    struct _group_item *current = NULL;
    int k;
    for(k = 0; k < group_count; k++){
        uint32_t new_data_size = 0;
        memcpy(&new_data_size, map + 4, 4);
        *(map_array + k) = map;
        map += new_data_size;

        (*(args + k)).item = malloc(sizeof *(*(args + k)).item);
        (*(args + k)).item->group = malloc(sizeof *(*(args + k)).item->group);
        (*(args + k)).map = map_array + k;
        (*(args + k)).parse_records = parse_records;
        pthread_create(pth + k, &attrs, read_group, args + k);

        if(plugin->children == NULL){
            plugin->children = (*(args + k)).item;
            current = (*(args + k)).item;
            current->group->previous = NULL;
            continue;
        }

        current->group->next = (*(args + k)).item;
        current->group->next_is_record = FALSE;
        (*(args + k)).item->group->previous = current;
        (*(args + k)).item->group->previous_is_record = FALSE;
        current = (*(args + k)).item;
    }
    plugin->last = current;
    plugin->last->group->next = NULL;

    pthread_join(header_pth, NULL);
    int i;
    for(i = 0; i < group_count; i++){
        pthread_join(*(pth + i), NULL);
    }

    assert(map == (start + filesize));

    free(start);
    free(pth);
    free(args);
    free(map_array);

    return plugin;
}

谢谢!

最佳答案

尝试 --sim-hints=no-nptl-pthread-stackcache。

参见 http://www.valgrind.org/docs/manual/manual-core.html#manual-core.rareopts获取更多信息。

关于c - 嵌套线程创建期间的数据竞争，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/40113375/

文章推荐： python - 一行输出字典

文章推荐： java - 从 IType 实例化对象

文章推荐： python - 如何确保字符串的格式正确

文章推荐： javascript - 根据 div 的宽度大小应用不同的颜色？

MSSQL监控数据库的DDL操作(创建，修改，删除存储过程，创建，修改，删除表等)
前言：有时候，一个数据库有多个帐号，包括数据库管理员，开发人员，运维支撑人员等，可能有很多帐号都有比较大的权限，例如DDL操作权限(创建，修改，删除存储过程，创建，修改，删除表等），账户多了，管理
javascript - 使用 Storybook 创建 React App 创建 webpack 问题
所以我用 Create React App 创建并设置了一个大型 React 应用程序。最近我们开始使用 Storybook 来处理和创建组件。它很棒。但是，当我们尝试运行或构建应用程序时，我们不断遇
javascript - 创建 Angular Directive(指令)以重用代码 - 创建 html 时解析错误
遵循我正在创建的控件的代码片段。这个控件用在不同的地方，变量也不同。我正在尝试编写指令来清理代码，但在 {{}} 附近插入值时出现解析错误。刚接触 Angular ，无法确定我错过了什么。请帮忙。
java - 创建 JAX-RS 提供程序以从 InputStream 创建 Java Image
我正在尝试创建一个 image/jpeg jax-rs 提供程序类，它为我的基于 post rest 的 Web 服务创建一个图像。我无法制定请求来测试以下内容，最简单的测试方法是什么？ @POST
c - 当我使用 FILE 创建 txt 文件时，Dev C++ 创建 test.txt
我一直在 Windows 10 的模拟器中练习 c。后来我改用dev C++ IDE。当我在 C 中使用 FILE 时。创建的文件的名称为 test.txt ，而我给出了其他名称。请帮助解决它。下面
ios - 为什么我们不遵循使用 xib 创建 customTableViewCell 的相同过程，就像使用 xib 创建 customView 一样？
当我们创建自定义 View 时，我们将 View 文件的所有者设置为自定义类，并使用 initWithFrame 或 initWithCode 对其进行实例化。当我们创建 customUITable
创建 Pthreads
我正在尝试为函数 * Producer 创建一个线程，但用于创建线程的行显示错误。我为这句话加了星标，但我无法弄清楚它出了什么问题...... #include #include #include
创建、调用JavaScript对象的方法集锦
今天在做项目时，遇到了需要创建JavaScript对象的情况。所以Bing了一篇老外写的关于3种创建JavaScript对象的文章，看后跟着打了一遍代码。感觉方法挺好的，在这里与大家分享一下。 &
python - 创建 StringToSign
我正在阅读将查询字符串传递给 Amazon 的 S3 以进行身份验证的文档，但似乎无法理解 StringToSign 的创建和使用方式。我正在寻找一个具体示例来说明 (1) 如何构造 String
c# - 创建、不等待和确保任务完成的正确方法
前言:我对 C# 中任务的底层实现不太了解，只了解它们的用法。为我在下面屠宰的任何东西道歉: 对于“我怎样才能开始一项任务但不等待它？”这个问题，我找不到一个好的答案。在 C# 中。更具体地说，即使任
linq - 创建 ILookups
我有一个由一些复杂的表达式生成的 ILookup。假设这是按姓氏查找人。 (在我们简单的世界模型中，姓氏在家庭中是唯一的) ILookup families; 现在我有两个对如何构建感兴趣的查询。首
WIX bundle 创建
我试图创建一个 MSI，其中包含和 exe。在 WIX 中使用了捆绑选项。这样做时出错。有人可以帮我解决这个问题。下面是代码: 错误 error LGH
Yii 创建、更新具有不同字段的表单
在 Yii 中，Create 和 Update 通常使用相同的形式。因此，如果我在创建期间有电子邮件、密码、...other_fields...等字段，但我不想在更新期间专门显示电子邮件和密码字段，但
qt - 创建 QModelIndex
上周我一直在努力创建一个给定一行和一列的 QModelIndex。或者，我会满足于在已经存在的 QModelIndex 中更改 row() 的值。任何帮助，将不胜感激。编辑: QModelInd
C: 创建、传递和访问指向常量字符串的常量指针数组
出于某种原因，这不起作用: const char * str_reset_command = "\r\nReset"; const char * str_config_command = "\r\nC
r - 创建 "other"字段
现在，我有以下由 original.df %.% group_by(Category) %.% tally() %.% arrange(desc(n)) 创建的 data.frame。 DF 5),
vim - 创建〜/.vimrc后错误打开文件
在今天之前，我使用/etc/vim/vimrc来配置我的vim设置。今天，我想到了创建.vimrc文件。所以，我用 touch .vimrc cat /etc/vim/vimrc > .vimrc 所
iPhone:创建 MKAnnotation
我可以创建一个 MKAnnotation，还是只读的？我有坐标，但我发现使用 setCooperative 手动创建 MKAnnotation 并不容易。想法？最佳答案 MKAnnotation
iphone - 创建 NSDictionary
在以下代码中，第一个日志语句按预期显示小数，但第二个日志语句记录 NULL。我做错了什么？ NSDictionary *entry = [[NSDictionary alloc] initWithOb
php - 创建/添加多维数组值时的未定义偏移
我正在使用与此类似的代码动态添加到数组； $arrayF[$f+1][$y][$x+1] = $value+1; 但是我在错误报告中收到了这个: undefined offset :1 问题:尝试创

太空宇宙

个人简介

我是一名优秀的程序员,十分优秀！

作者热门文章

滴滴打车优惠券免费领取

全站热门文章

首页

博学

6Ren·AI

商城

c - 嵌套线程创建期间的数据竞争