4

When I first open a terminal, or open a new one after not having used one for a while, any kind of tab completion in my home directory (for example, ls and Tab) takes several seconds. I have seen this behavior before when using autofs to mount network drives, but I don't have any in ~/. I do mount NFS shares with systemd's automount, but those are in ~/badabing/, so everything in ~/ itself is just a local file.

In case it is an automount issue, here are the relevant /etc/fstab lines (yes, the server is called "badabing", I named it during a Sopranos binge a few years ago):

badabing:/nfs_shares/music /mnt/badabing/music    nfs4  noauto,x-systemd.automount,x-systemd.device-timeout=5sec,x-systemd.idle-timeout=1min    0 0
badabing:/nfs_shares/series /mnt/badabing/series  nfs4  noauto,x-systemd.automount,x-systemd.device-timeout=5sec,x-systemd.idle-timeout=1min    0 0
badabing:/nfs_shares/movies /mnt/badabing/movies  nfs4  noauto,x-systemd.automount,x-systemd.device-timeout=5sec,x-systemd.idle-timeout=1min    0 0

Then, in ~/badabing I have:

$ ls -l ~/badabing/
total 0
lrwxrwxrwx 1 terdon terdon 32 Jan 10  2016 movies -> /mnt/badabing/nfs_shares/movies/
lrwxrwxrwx 1 terdon terdon 31 Jan 10  2016 music -> /mnt/badabing/nfs_shares/music/
lrwxrwxrwx 1 terdon terdon 31 Jan 10  2016 series -> /mnt/badabing/nfs_shares/series

I'd like to investigate this more. Can I somehow strace tab completion? Is the only way to add echo commands in the various bash completion scripts to see what's hanging? There are quite a few of those so I'd really rather avoid that.

So, what's causing this or, at least, how can I debug it further?

Improve this question
6
  • You can run strace against a given process, and see where the bash process is spending time. Adding echo's to scripts won't show you issues in tab completion. Commented Nov 19, 2016 at 13:36
  • @ThomasDickey I was thinking of the scripts in /usr/share/bash-completion which are run when bash tries to complete something. How would I run strace here though? I can't do strace bash 2>log since that will hang. Ah, I just saw that I can attach strace to an already running process. I'll try that, thanks. Commented Nov 19, 2016 at 14:20
  • That's the -p option of strace. Commented Nov 19, 2016 at 14:22
  • @ThomasDickey yeah, just tried that but it seems to run and exit immediately. How can I keep it running while I test things in the straced bash session? Commented Nov 19, 2016 at 14:22
  • Perhaps you attached to a child process. If you attach to the parent process and use the -f option, that will follow the (new) child processes. Commented Nov 19, 2016 at 14:31

3 Answers 3

Reset to default
4

I would do this:

sudo strace -pXXXX -tfo /tmp/strace.log

where XXXX is the process id of bash. In a quick check of filename completion on one of my NFS-mounted directory trees, it works without problems:

-f

Trace child processes as they are created by currently traced processes as a result of the fork(2) system call.

-t

Prefix each line of the trace with the time of day.

Some people might prefer -r:

-r

Print a relative timestamp upon entry to each system call. This records the time difference between the beginning of successive system calls.

Improve this answer
3
  • strace hangs and can't be closed. Commented Sep 1, 2020 at 18:16
  • My problem is it doesn't hang if I try stracing it. Maybe it's the duration of time that passes until I manage to setup strace. Commented Jul 27, 2023 at 9:00
  • You can get the newest bash PID by running pgrep. That way you can save some setup time sudo strace -p$(pgrep -n [b]ash) -tfo /tmp/strace.log. However the problem persists for me that I can't measure the issue, since if this other bash is open, the auto-completion doesn't hang for me. Commented Jul 27, 2023 at 9:04
2

Run set -x to get a trace of every shell command, even those executed by the completion framework.

Run strace -f -p 1234 from another shell to see a trace of system calls. If you get the error “Operation not permitted”, it may be due to a security restriction that restricts the ptrace system call to children of the tracing process.¹ To disable this restriction, run sysctl kernel.yama.ptrace_scope=0 as root. See https://askubuntu.com/questions/41629/after-upgrade-gdb-wont-attach-to-process for more information. Alternatively, you can still start a shell with strace -T -f -o strace.bash bash (-T adds the time spent in each system call) and do your experiments from there.

¹ This restriction limits the impact of running a malicious application from your account. It's only effective if coupled with many other restrictions (typically effected with SELinux or AppArmor), in particular to limit the restricted process to a small set of files. Even in restricted ptrace mode, the typical use case for ptrace, which is for a debugger to debug a child process, remains allowed.

Improve this answer
1

Maybe this helps:

PS4='Line ${LINENO}: ' bash -x /usr/share/bash-completion/bash_completion ls
Improve this answer
3
  • Its not clear to me as a neophyte what this command does. Could you help by explaining it Commented Jan 2, 2019 at 16:58
  • The command PS4='Line ${LINENO}: ' bash -x XYZ goes line by line through the script XYZ and prints each line with line number. The command includes subscripts and you don’t have to modify the script (like set -X). This is useful if you want to debug scripts. Commented Jan 2, 2019 at 19:43
  • The '-x' or xtrace option tells Bash to print each command it runs as it runs it. The value of the PS4 environment variable is expanded and displayed at the beginning of each line, and $LINENO is set to the number of whatever line it's executing in the script. Commented Sep 3, 2020 at 23:55

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.