nodejs shows SIGABRT (Aborted) @ 0 (0) on x64 CentOS about libuv HOT 7 CLOSED

/cc @bnoordhuis

from libuv.

For the record, from the x86_64 log:

syscall_293(0xe6cbbc, 0x80000, 0x65646f, 0xfefefefefefefeff, 0x3, 0x7fff76a041d0, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7) = 0x125

293 is the pipe2() system call and it returns the rather strange error -293 (when interpreting 0x125 as a sign-extended 64 bits integer.)

Seeing how the 32 bits binary works, I wonder if this is simply a kernel bug. With 32 bits, where pipe2 is system call 331, the system call returns the expected ENOSYS error.

@aneek Can you tell me what package that kernel was installed from?

from libuv.

@bnoordhuis If you want to know the base package of the kernel, I really don't know how Amazon EC2 installed that kernel. But, I ran a few commands and the following are the output.

1. uname -r shows "2.6.18-xenU-ec2-v1.4"
2. rpm -q kernel shows
   kernel-2.6.18-164.11.1.el5
   kernel-2.6.18-398.el5
   kernel-2.6.18-400.1.1.el5

If you have any specific commands to check any specific details please let me know and I will give you the findings.

from libuv.

2.6.18-xenU-ec2-v1.4 is pretty old and google suggests there are newer versions. When did you last upgrade?

I took a peek at the 2.6.18 kernel sources and I think the strange return value is a bug that was fixed in torvalds/linux@cc7d479. That's part of the 2.19 kernel but it only affects the return value when tracing system calls so the real error is probably something else. Can you try this:

$ gdb --args /path/to/64/bits/node
> break uv__pipe2
> run
# wait for the breakpoint to hit
> finish
> print errno
# if `print errno` doesn't work:
> print *(int *) __errno_location()

from libuv.

@bnoordhuis I will try this one after this holiday.

from libuv.

Ping

from libuv.

No follow-up, closing.

from libuv.