[Android 17] Early observations: JIT thread pool crash on attach and JNI transition frame aborts
还没有人认领这个 Issue。
评估
- 难度
- 5/5
- 预计耗时
- 一周以上
- 新手友好度
- 12/100
- Issue 类型
- 缺陷
- 描述清晰度
- 需要澄清
- 活跃度
- 活跃
- 技术栈
- android, java, javascript
调研方向
The payload names no file in this repository; the crashes are in ART internals (art::jit::Jit::CreateThreadPool) and in the JNI transition check. Start by reproducing the Java.use hooks on android.webkit.WebView $init and android.os.Binder attachInterface on an API 37 device or emulator, then gather the tombstone. Done would mean a hook pattern that survives on API 37 with a stated cause, which needs Android runtime knowledge and a maintainer decision on scope.
由索引模型根据 Issue 内容生成。
描述
Ran into a couple of quirks while testing Frida 17.21.0 on Android 17 (API 37, build CE2A.260420.019). Opening this up to share logs and reproduction details in case they're useful as A17 support gets fleshed out.
1. SIGSEGV in art::jit::Jit::CreateThreadPool during attach
- Behavior: When attaching
frida-serverearly during process launch, ART intermittently segfaults insideart::jit::Jit::CreateThreadPool. - Observations: Appears to be a race between ptrace attachment and JIT thread pool initialization. Disabling JIT before launching or attaching avoids the crash on our end:
Not sure if there's a cleaner synchronization hook or if this is primarily visible on emulator timing, but wanted to mention it.adb shell setprop dalvik.vm.usejit false
2. Runtime abort: JNI transition frame error
- Behavior: Hooking native constructor stubs or Binder entrypoints causes ART to abort the process:
art E ... JNI transition frame error - Reproduction cases we saw:
- Replacing
$initon native classes:
Java.use('android.webkit.WebView').$init.implementation = ... - Hooking Binder interface registration:
Java.use('android.os.Binder').attachInterface.implementation = ...
- Replacing
- Observations: Looks like A17 might have tightened up stack validation or frame accounting around native-to-managed transitions. Replacing these native stubs with JS trampolines seems to trigger that transition check.
- Current workaround for us: Steered our scripts away from hooking
$initandBinder.attachInterfacedirectly on API 37, targeting higher-level callers or concrete factory methods instead.
Happy to pull more tombstones or test proposed patches if anyone is digging into A17 support.
- 主要语言
- JavaScript
- 星标
- 413
- 派生
- 174
- PR 合并指标
- 30 天内没有已合并 PR
环境准备
这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
frida/frida-java-bridge 的其他 Issue
-
Types is missing `deoptimizeMethod`可能已有人在做 @Electric1447 于 180 天前认领。 未关闭
难度 1/5 1 小时以内 新手友好度 65/100
frida/frida-java-bridge#386 ·
-
[v7.0.11+] Process crash on accessing object["<init>"] with JVMTI available可能已有人在做 关联的 PR 仍在进行中或已合并。 未关闭
难度 2/5 1-3 小时 新手友好度 62/100
frida/frida-java-bridge#384 · 2 条评论 · 2 个 reaction ·
-
难度 3/5 1-2 天 新手友好度 42/100
frida/frida-java-bridge#409 ·
-
ensureClassInitialized() no longer initializes on ART since 7983e815 — hooks on uninitialized classes install successfully but never dispatch可能已有人在做 关联的 PR 仍在进行中或已合并。 未关闭
难度 4/5 3-5 天 新手友好度 52/100
frida/frida-java-bridge#408 ·
-
难度 5/5 一周以上 新手友好度 20/100
frida/frida-java-bridge#404 ·
查看 frida/frida-java-bridge 的全部 Issue
相似的 Issue
-
bug/1-unconfirmed kind/bug
难度 2/5 1-3 小时 新手友好度 78/100
维护者通常 1 天内回复
-
难度 1/5 1 小时以内 新手友好度 74/100
521xueweihan/HelloGitHub#4010 · 2 条评论 ·
-
难度 2/5 1-3 小时 新手友好度 67/100
维护者通常 1 天内回复
-
BUG
难度 2/5 1-3 小时 新手友好度 74/100
维护者通常 1 天内回复
-
[BUG]: filesystem-edit-file converts a CRLF file's line endings to LF on every edit可能已有人在做 @MohammadHijjawi97 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 82/100
Mintplex-Labs/anything-llm#6755 ·
维护者通常 1 天内回复