Crash on node:10-alpine docker image on some systems

[orgads]

Problem description

The grpc node module crashes on a particular system. I tried to reproduce on another machine and failed, but on this machine it reliably reproduces.

Reproduction steps

docker run -it node:10-alpine sh
mkdir a
cd a
npm init -y
export NODE_TLS_REJECT_UNAUTHORIZED=0
npm i grpc
cd node_modules/grpc/src/node/extension_binary/node-v64-linux-x64-musl
node grpc_node.node
Segmentation fault (core dumped)

The same steps work with node:12-alpine.

Environment

• OS name, version and architecture: Host: Linux Ubuntu 18.04.2 amd64, Docker: 18.09.7
• Kernel version: 5.0.0-37-generic
• Node version: 10.18.0
• Node installation method: Docker
• If applicable, compiler version: Downloaded the pre-built binary
• Package name and version: [email protected]

Additional context

(gdb) run grpc_node.node
Starting program: /usr/local/bin/node grpc_node.node
warning: Error disabling address space randomization: Operation not permitted
[New LWP 39]
[New LWP 40]
[New LWP 41]
[New LWP 42]
[New LWP 43]
[New LWP 44]

Thread 7 "node" received signal SIG34, Real-time event 34.
[Switching to LWP 44]
__cp_end () at src/thread/x86_64/syscall_cp.s:29
29      src/thread/x86_64/syscall_cp.s: No such file or directory.
(gdb) bt
#0  __cp_end () at src/thread/x86_64/syscall_cp.s:29
#1  0x00007f64cf437895 in __syscall_cp_c (nr=202, u=<optimized out>, v=<optimized out>, w=<optimized out>, x=<optimized out>, y=<optimized out>, z=0)
    at src/thread/pthread_cancel.c:33
#2  0x00007f64cf436e9f in __futex4_cp (to=0x0, val=-1, op=<optimized out>,
    addr=0x555f5e7a4720 <node::inspector::(anonymous namespace)::start_io_thread_semaphore>) at src/thread/__timedwait.c:52
#3  __timedwait_cp (addr=addr@entry=0x555f5e7a4720 <node::inspector::(anonymous namespace)::start_io_thread_semaphore>, val=val@entry=-1, clk=clk@entry=0,
    at=at@entry=0x0, priv=<optimized out>) at src/thread/__timedwait.c:52
#4  0x00007f64cf439fe1 in sem_timedwait (sem=sem@entry=0x555f5e7a4720 <node::inspector::(anonymous namespace)::start_io_thread_semaphore>, at=at@entry=0x0)
    at src/thread/sem_timedwait.c:23
#5  0x00007f64cf43a055 in sem_wait (sem=sem@entry=0x555f5e7a4720 <node::inspector::(anonymous namespace)::start_io_thread_semaphore>) at src/thread/sem_wait.c:5
#6  0x0000555f5ce88cd2 in uv__sem_wait (sem=0x555f5e7a4720 <node::inspector::(anonymous namespace)::start_io_thread_semaphore>)
    at ../deps/uv/src/unix/thread.c:604
#7  uv_sem_wait (sem=0x555f5e7a4720 <node::inspector::(anonymous namespace)::start_io_thread_semaphore>) at ../deps/uv/src/unix/thread.c:660
#8  0x0000555f5cdf94b0 in node::inspector::(anonymous namespace)::StartIoThreadMain(void*) ()
#9  0x00007f64cf43833e in start (p=0x7f64cca2aab0) at src/thread/pthread_create.c:192
#10 0x00007f64cf43a447 in __clone () at src/thread/x86_64/clone.s:22

[lxzlovelyidiot]

The same on centos 3.10.0-957.21.3.el7.x86_64 #1 SMP Tue Jun 18 16:35:19 UTC 2019 x86_64 Linux with alpine:3.10 node-10.16.3

[murgatroid99]

I can't reproduce this failure on either of the listed Alpine docker images.

[vdeturckheim]

I have a similar issue with another native addon on alpine linux with node 10. But so far I can only reproduce it in AWS codebuild which runs Ubuntu 14 with am Amazon build of Linux Kernel.

Out of curiosity, @orgads is the machine where it crashes on AWS?

[orgads]

IIRC it was on Azure.

[vdeturckheim]

Weird, I might have a tempoary fix (worked for me duting my tests) but I am not clear regarding the implications of it: adding --cap-add=SYS_PTRACE --security-opt seccomp=unconfined flags to docker run

[vdeturckheim]

I am coming back to this, we identified an issue with a syscall made during the dlopen sequence on Alpine - it is not allowed on docker. My collegue made a PR on Moby to fix it and it has been released in docker engine 20.10. I expect this should fix the issue.