Hi,
Mathieu Othacehe <othacehe <at> gnu.org> skribis:
> Regarding those four, I was able to reproduce the issue this way:
>
> $ guix repl
> (stop-service 'guix-daemon)
> (start-service 'guix-daemon (list (number->string (getpid))))
Or from the shell:
herd stop guix-daemon
herd start guix-daemon $$
I was able to reproduce it using a bare-bones.tmpl VM.
> The latter command hangs and Shepherd becomes unresponsive. I collected
> an (attached) strace dump of Shepherd showing that there is no response
> on the socket when the service is started.
>
> Note that, this works:
>
> $ guix repl
> (stop-service 'guix-daemon)
> (start-service 'guix-daemon)
>
> So the problem could be caused by the "container-excursion*" in the
> "fork+exec-command/container" procedure.
PID 1 gets stuck on read(16, …) forever, after reading the string “2866”
(a PID):
--8<---------------cut here---------------start------------->8---
[pid 2865] clone(child_stack=NULL, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLDstrace: Process 2866 attached
, child_tidptr=0x7fccfbe00a10) = 2866
[pid 2866] set_robust_list(0x7fccfbe00a20, 24) = 0
[pid 2866] close(3) = 0
[pid 2865] write(39, "2866", 4 <unfinished ...>
[pid 2866] close(4 <unfinished ...>
[pid 2865] <... write resumed>) = 4
[pid 2866] <... close resumed>) = 0
[pid 2866] pipe2( <unfinished ...>
[pid 2865] close(39 <unfinished ...>
[pid 2866] <... pipe2 resumed>[3, 4], O_CLOEXEC) = 0
[pid 2865] <... close resumed>) = 0
[pid 2865] exit_group(0) = ?
[pid 2866] rt_sigaction(SIGCHLD, {sa_handler=SIG_DFL, sa_mask=[], sa_flags=SA_RESTORER, sa_restorer=0x7fccfc304d80}, {sa_handler=0x7fccfc427d50, sa_mask=[], sa_flags=SA_RESTORER|SA_NOCLDSTOP, sa_restorer=0x7fccfc304d80}, 8) = 0
[pid 2866] rt_sigaction(SIGINT, {sa_handler=SIG_DFL, sa_mask=[], sa_flags=SA_RESTORER, sa_restorer=0x7fccfc304d80}, {sa_handler=0x7fccfc427d50, sa_mask=[], sa_flags=SA_RESTORER, sa_restorer=0x7fccfc304d80}, 8) = 0
[pid 2866] rt_sigaction(SIGHUP, {sa_handler=SIG_DFL, sa_mask=[], sa_flags=SA_RESTORER, sa_restorer=0x7fccfc304d80}, {sa_handler=0x7fccfc427d50, sa_mask=[], sa_flags=SA_RESTORER, sa_restorer=0x7fccfc304d80}, 8) = 0
[pid 2866] rt_sigaction(SIGTERM, {sa_handler=SIG_DFL, sa_mask=[], sa_flags=SA_RESTORER, sa_restorer=0x7fccfc304d80}, {sa_handler=0x7fccfc427d50, sa_mask=[], sa_flags=SA_RESTORER, sa_restorer=0x7fccfc304d80}, 8) = 0
[pid 2866] rt_sigprocmask(SIG_UNBLOCK, [HUP INT TERM CHLD], [HUP INT TERM CHLD], 8) = 0
[pid 2866] mkdir("/var", 0777) = -1 EEXIST (File exists)
[pid 2866] mkdir("/var/run", 0777) = -1 EEXIST (File exists)
[pid 2865] +++ exited with 0 +++
[pid 1] <... wait4 resumed>[{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 2865
[pid 1] close(39) = 0
[pid 2866] setsid( <unfinished ...>
[pid 1] read(16, <unfinished ...>
[pid 2866] <... setsid resumed>) = 2866
[pid 1] <... read resumed>"2866", 4096) = 4
[pid 2866] chdir("/") = 0
[pid 1] read(16, <unfinished ...>
[pid 2866] prlimit64(0, RLIMIT_NOFILE, NULL, {rlim_cur=1024, rlim_max=4*1024}) = 0
[pid 2866] close(0) = 0
[pid 2866] openat(AT_FDCWD, "/dev/null", O_RDONLY) = 0
[pid 2866] dup2(0, 0) = 0
[pid 2866] close(1) = 0
[pid 2866] close(2) = 0
[pid 2866] openat(AT_FDCWD, "/var/log/guix-daemon.log", O_WRONLY|O_CREAT|O_APPEND, 0640) = 1
[pid 2866] dup2(1, 1) = 1
[pid 2866] dup2(1, 2) = 2
[pid 2866] execve("/gnu/store/bxnkqnpbf4q4z6245b61wgpm8gkr9nj1-guix-1.3.0-29.9e46320/bin/guix-daemon", ["/gnu/store/bxnkqnpbf4q4z6245b61w"..., "--build-users-group", "guixbuild", "--max-silent-time", "0", "--timeout", "0", "--log-compression", "gzip", "--discover=yes", "--substitute-urls", "https://substitutes.nonguix.org "...], 0x7fccf71fa480 /* 3 vars */) = 0
--8<---------------cut here---------------end--------------->8---
This happens because the other end of the file descriptor happens to be
inherited by 2866, which will never close it because it just execs
guix-daemon.
This is fixed by 6abdcef4a68e98f538ab69fde096adc5f5ca4ff4; the log
contains extra details.
Thanks!
Ludo’.
This bug report was last modified 2 years and 295 days ago.