[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fwM9zL1npYaUru-hBde_zvyWjDcgQblLzaGClJezrkAY":3},{"item":4,"related":44},{"id":5,"type":6,"title":7,"slug":8,"summary":9,"body":10,"coverUrl":11,"productScreenshots":12,"productLinks":13,"authorName":14,"authorUrl":15,"authorSubject":16,"category":17,"tags":22,"sourceLabel":35,"sourceName":36,"sourceUrl":37,"status":38,"seoTitle":7,"seoDescription":9,"canonicalUrl":35,"isFeatured":39,"sno":40,"sortOrder":41,"publishedAt":42,"updatedAt":43,"createdAt":43},"5879ed15-16b2-4a1c-836d-1c7d4c7b0086","article","seccomp 用户态通知：系统调用怎样交给另一个进程处理？","seccomp-user-notification-syscall-supervision","seccomp 用户态通知可以把系统调用事件交给监督进程，但官方手册明确警告它不适合作为安全策略的唯一基础。本文讲清能力与竞态边界。","系统调用是应用向内核请求服务的入口，例如打开文件、创建进程或改变网络配置。seccomp 可以用过滤器限制进程允许发起的系统调用。Linux 的 seccomp user-space notification（用户态通知）则允许过滤器把某些系统调用事件交给另一个用户态监督程序处理，监督者检查后再决定如何回应。\n\n它有点像大楼前台接到访客请求后先联系值班人员，而不是让访客直接进入后台区域。应用线程发出系统调用，内核过滤器把事件通知给持有监听器的监督进程；监督进程可以读取请求信息并返回处理结果。它为沙箱、兼容层和受控运行环境提供了在用户态插入决策的能力。\n\n但这里有一个特别重要的安全边界：Linux 手册明确提醒，seccomp 通知机制不应被当作构建安全策略的可靠基础，因为监督者观察与实际系统调用执行之间存在竞态风险。通知中的参数可以帮助监督者理解请求，却不能自动冻结所有相关对象状态。比如监督者检查某个文件描述符或路径之后，应用周围的状态可能变化，导致实际执行对象与监督者当时以为的对象不同。\n\n因此，用户态通知更适合协作式控制、仿真、兼容或可接受竞态的策略补充。若目标是强安全隔离，应优先使用适合的内核强制机制、最小权限、命名空间、文件系统隔离等措施，并把通知监督器视为额外层，而非唯一门禁。过滤器本身有安装时机、线程同步和操作模式等要求，监督者还必须处理超时、恶意输入和崩溃等异常。\n\n系统调用代理看起来像是“把内核功能搬到用户态”，其实不是简单转发。内核负责暂停原任务并传递通知，用户态监督进程需要决定如何安全、正确地回应；它可以拒绝或返回结果，但并非所有操作都能无风险地代替原调用执行。设计上要明确谁拥有文件描述符、路径和权限上下文，以及检查结果与执行动作之间是否可能被攻击者改变。\n\n这项能力很强，也容易被误用。理解它的关键不是“能拦截系统调用”，而是先理解通知机制的竞态限制，再决定它适不适合作为安全边界的一部分。\n\n来源：[Linux `seccomp_unotify(2)` 手册](https:\u002F\u002Fman7.org\u002Flinux\u002Fman-pages\u002Fman2\u002Fseccomp_unotify.2.html)","\u002Fuploads\u002F2026-10-06\u002F48e132f5-dc64-4380-9e5e-c8ca8aa55ebf.jpg",[],[],"Foundit","https:\u002F\u002Ffoundit.cn","foundit-ai-editorial",{"id":18,"name":19,"slug":20,"description":21},"6179d3b6-dc34-4483-9ded-3cd9f1b37a47","科普","abbreviation","介绍各领域新兴概念",[23,27,31],{"id":24,"name":25,"slug":26},"7c76bfc2-f80f-4ee0-a95d-27bd8708b434","技术","slug",{"id":28,"name":29,"slug":30},"88d2bc27-0e0f-468a-b907-2991cb97b87b","人工智能","ai",{"id":32,"name":33,"slug":34},"68cedb55-2cac-412f-8f81-fda8c7d686dd","思考","thought",null,"Linux seccomp_unotify(2) Manual","https:\u002F\u002Fman7.org\u002Flinux\u002Fman-pages\u002Fman2\u002Fseccomp_unotify.2.html","published",false,56,0,"2026-10-06T00:00:00.000Z","2026-10-06T07:51:28.799Z",[45,54,62],{"id":46,"type":6,"title":47,"slug":48,"summary":49,"coverUrl":50,"authorName":14,"sno":51,"publishedAt":52,"createdAt":53},"1c047558-44f5-4e84-be6a-5468883d3178","中式巨构：技术终于有能力回应一个古老的梦","chinese-heaven","没有被“发明”，没有被“想出”，它只是被“看见”了。我们做了几千年的梦，终于在2026年，被AI显影在屏幕之上。","\u002Fuploads\u002F2026-08-15\u002Fc987bd06-cca5-4e10-b7a0-1e3213fc78d8.jpg",1,"2026-08-15T00:00:00.000Z","2026-08-15T04:53:56.292Z",{"id":55,"type":6,"title":56,"slug":57,"summary":58,"coverUrl":59,"authorName":14,"sno":51,"publishedAt":60,"createdAt":61},"6fb796fa-60a1-4dba-bd19-82a6cda0fe78","手段与目的：一台没有\"羞耻\"的理性机器","rational-machine","Hugging Face 遭 OpenAI AI 模型自主攻击事件中，真正值得警惕的，不是一台想害我们的机器，而是一台只想完成任务、且对手段毫无羞耻的理性机器。","https:\u002F\u002Foxqtewbrpuiouqqjrvdv.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fpublic-media\u002F2026-07-23\u002Fc90cf6fb-faa4-4454-91b5-d7356c144277.jpg","2026-07-23T00:00:00.000Z","2026-07-23T08:39:37.528Z",{"id":63,"type":6,"title":64,"slug":65,"summary":66,"coverUrl":67,"authorName":14,"sno":68,"publishedAt":69,"createdAt":70},"32fc8411-1570-4736-8e79-b0599be8a962","让 AI 像育种一样进化算法，会发生什么？","alphaevolve-evolutionary-algorithm-discovery","让大模型提出许多程序，再由自动评测器运行、打分、筛选和继续改良，算法发现就像进入了一座高速育种场。本文拆解 AlphaEvolve 的循环，说明它为何适合答案可验证的问题，也解释它不能自动解决哪些开放难题。","\u002Fuploads\u002F2026-09-08\u002Fef0f54b2-ca7d-4d06-8307-80d3e804e6f4.jpg",40,"2026-09-01T00:00:00.000Z","2026-08-14T03:06:10.928Z"]