exec-port stops responding sometimes.

[kape1395]

erlexec sometimes stops to respond. Not sure if it is related to this problem, but sometimes I see segfaults in the syslog (Debian/bullseye, Erlang/OTP 22):

... kernel: [10024684.056118] exec-port[262130]: segfault at 1a ip 00007fb6573ccdb4 sp 00007fff9a7ef338 error 4 in libstdc++.so.6.0.28[7fb6573a7000+dc000]
... kernel: [10024684.056126] Code: 1f 80 00 00 00 00 48 8b 47 18 48 85 c0 74 17 0f 1f 80 00 00 00 00 49 89 c0 48 8b 40 10 48 85 c0 75 f4 4c 89 c0 c3 48 8b 47 08 <48> 3b 78 18 75 1f 66 0f 1f 44 00 00 49 89 c0 48 8b 40 08 4c 39 40

I understand the report is not very informative. I will try to debug it further, but any hints would help.

[saleyn]

Can you upgrade the OTP to 27 and reproduce it there? You may need to compile a non-optimized build in order to see the stacktrace by setting OPTIMIZE=false make.

[kape1395]

Upgrading OTP is not easy there. I will try to run this with OPTIMIZE=false make and valgrind enabled.
Do you think that might be related to the OTP version?
The failures happen now once a day or a week, and I can't find a way yet to reproduce them reliably.

[saleyn]

Not sure. It has been running pretty stably in many uses.

[kape1395]

That's the valgrind's memory detector output.
Unfortunately, I still don't have the Erlexec debug output; it was not written to a file.

==401359== Memcheck, a memory error detector
==401359== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==401359== Using Valgrind-3.16.1 and LibVEX; rerun with -h for copyright info
==401359== Command: /.../lib/erlexec-2.0.6/priv/x86_64-linux-gnu/exec-port
==401359== Parent PID: 401339
==401359==
==403175== Syscall param ioctl(TCSET{S,SW,SF}) points to uninitialised byte(s)
==403175==    at 0x4C77850: tcsetattr (tcsetattr.c:83)
==403175==    by 0x119440: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:568)
==403175==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==403175==    by 0x10D5B9: main (exec.cpp:291)
==403175==  Address 0x1ffefff6f0 is on thread 1's stack
==403175==  in frame #0, created by tcsetattr (tcsetattr.c:45)
==403175==
==403175== Conditional jump or move depends on uninitialised value(s)
==403175==    at 0x1195A3: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:581)
==403175==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==403175==    by 0x10D5B9: main (exec.cpp:291)
==403175==
==404561== Syscall param ioctl(TCSET{S,SW,SF}) points to uninitialised byte(s)
==404561==    at 0x4C77850: tcsetattr (tcsetattr.c:83)
==404561==    by 0x119440: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:568)
==404561==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==404561==    by 0x10D5B9: main (exec.cpp:291)
==404561==  Address 0x1ffefff6f0 is on thread 1's stack
==404561==  in frame #0, created by tcsetattr (tcsetattr.c:45)
==404561==
==404561== Conditional jump or move depends on uninitialised value(s)
==404561==    at 0x1195A3: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:581)
==404561==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==404561==    by 0x10D5B9: main (exec.cpp:291)
==404561==
==404561== Conditional jump or move depends on uninitialised value(s)
==404561==    at 0x1195AC: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:581)
==404561==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==404561==    by 0x10D5B9: main (exec.cpp:291)
==404561==
==404561== Syscall param ioctl(TIOCSWINSZ) points to uninitialised byte(s)
==404561==    at 0x4C78277: ioctl (syscall-template.S:120)
==404561==    by 0x1145C4: ei::set_winsz(int, int, int) (exec_impl.cpp:182)
==404561==    by 0x1195C5: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:582)
==404561==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==404561==    by 0x10D5B9: main (exec.cpp:291)
==404561==  Address 0x1ffefff740 is on thread 1's stack
==404561==  in frame #1, created by ei::set_winsz(int, int, int) (exec_impl.cpp:177)
==404561==
==405903== Syscall param ioctl(TCSET{S,SW,SF}) points to uninitialised byte(s)
==405903==    at 0x4C77850: tcsetattr (tcsetattr.c:83)
==405903==    by 0x119440: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:568)
==405903==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==405903==    by 0x10D5B9: main (exec.cpp:291)
==405903==  Address 0x1ffefff6f0 is on thread 1's stack
==405903==  in frame #0, created by tcsetattr (tcsetattr.c:45)
==405903==
==405903== Conditional jump or move depends on uninitialised value(s)
==405903==    at 0x1195A3: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:581)
==405903==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==405903==    by 0x10D5B9: main (exec.cpp:291)
==405903==
==405903== Conditional jump or move depends on uninitialised value(s)
==405903==    at 0x1195AC: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:581)
==405903==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==405903==    by 0x10D5B9: main (exec.cpp:291)
==405903==
==405903== Syscall param ioctl(TIOCSWINSZ) points to uninitialised byte(s)
==405903==    at 0x4C78277: ioctl (syscall-template.S:120)
==405903==    by 0x1145C4: ei::set_winsz(int, int, int) (exec_impl.cpp:182)
==405903==    by 0x1195C5: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:582)
==405903==    by 0x10DCD0: process_command(bool) (exec.cpp:398)
==405903==    by 0x10D5B9: main (exec.cpp:291)
==405903==  Address 0x1ffefff740 is on thread 1's stack
==405903==  in frame #1, created by ei::set_winsz(int, int, int) (exec_impl.cpp:177)
==405903==

[kape1395]

It looks like it's pty related. Here is a simple case to reproduce:

1> exec:run("/usr/bin/echo asd", [sync, stdout, stderr, pty]).
exec got msg from port: {1,{pid,74130}}
exec added monitor <0.87.0> for OsPid 74130exec got msg from port: {0,{stdout,74130,<<"asd\r\n">>}}
exec got msg from port: {0,{exit_status,74130,0}}
Pid 74130 exited with status: {0,0}
<0.87.0> 74130 got down message ({status,0}) (ismon=true)
{ok,[{stdout,[<<"asd\r\n">>]}]}

Results in valgrind report:

==74130== Syscall param ioctl(TCSET{S,SW,SF}) points to uninitialised byte(s)
==74130==    at 0x4C46D10: tcsetattr (tcsetattr.c:83)
==74130==    by 0x11DC65: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:568)
==74130==    by 0x1121BA: process_command(bool) (exec.cpp:398)
==74130==    by 0x10DBC1: main (exec.cpp:291)
==74130==  Address 0x1ffeffe300 is on thread 1's stack
==74130==  in frame #0, created by tcsetattr (tcsetattr.c:45)
==74130== 
==74130== Conditional jump or move depends on uninitialised value(s)
==74130==    at 0x11CE50: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:581)
==74130==    by 0x1121BA: process_command(bool) (exec.cpp:398)
==74130==    by 0x10DBC1: main (exec.cpp:291)
==74130== 
==74130== Syscall param ioctl(TIOCSWINSZ) points to uninitialised byte(s)
==74130==    at 0x4C4EDED: ioctl (ioctl.c:36)
==74130==    by 0x1151FD: ei::set_winsz(int, int, int) (exec_impl.cpp:182)
==74130==    by 0x11CE58: ei::start_child(ei::CmdOptions&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) (exec_impl.cpp:582)
==74130==    by 0x1121BA: process_command(bool) (exec.cpp:398)
==74130==    by 0x10DBC1: main (exec.cpp:291)
==74130==  Address 0x1ffeffe340 is on thread 1's stack
==74130==  in frame #1, created by ei::set_winsz(int, int, int) (exec_impl.cpp:177)

In my case, it's not critical to use the pty option. It was used accidentally. But I hope that helps.