Debugging

Using the GNU Debugger

If you think you found a problem in OPS or Nanos first thing is to verify where the bug lies. Generally, if you turn on the -d flag and study the bottom you can tell if it is in user or kernel:

1 general protection fault in user mode, rip 0x13783afe7

Clearly in this example we see a GPF in user.

At this moment, we do not have interactive debugging support with ops, but we plan on integrating it soon. For now if you wish to have symbol access within your user program the following conditions must be met:

  • Ensure your program is statically linked.

  • Ensure you have debugging symbols to begin with:

You can do this with c by doing the following:

Example

For this example will examine a segfault (that we purposely injected):

#include <stdio.h>
#include <stdlib.h>

void mybad() {
  int x = 1;
  char *stuff = "asdf";

  printf("about to die\n");
  *(int*)0 = 0;
}

int main(void) {
  mybad();
  printf("should not get here\n");

  return 0;
}

We compile with debugging symbols and link statically:

cc main.c -static -g -o main

First (since we are missing interactive debug support in ops) you need to modify ops to manually add the noaslr flag in lepton/image.go:

m.AddDebugFlag("noaslr", 't')

This is important because otherwise we randomize the location of the .text and other parts of your program.

Next, we'll run without accel support:

ops run --accel=false main

Then we let it crash.

Now let's manually start qemu with gdb support: (Not all of this is necessary but definitely ensure your 'drive file' line matches where your disk image is)

Also the '-s -S' starts the remote gdb debugger:

qemu-system-x86_64 \
-device pcie-root-port,port=0x10,chassis=1,id=pci.1,bus=pcie.0,multifunction=on,addr=0x3 \
-device pcie-root-port,port=0x11,chassis=2,id=pci.2,bus=pcie.0,addr=0x3.0x1 \
-device pcie-root-port,port=0x12,chassis=3,id=pci.3,bus=pcie.0,addr=0x3.0x2 \
-device virtio-scsi-pci,bus=pci.2,addr=0x0,id=scsi0 \
-device scsi-hd,bus=scsi0.0,drive=hd0 -no-reboot -cpu max -machine q35 \
-device isa-debug-exit -m 2G \
-drive file=/home/eyberg/.ops/images/main.img,format=raw,if=none,id=hd0 \
-device virtio-net,bus=pci.3,addr=0x0,netdev=n0 \
-netdev user,id=n0,hostfwd=tcp::8080-:8080,hostfwd=tcp::9090-:9090,hostfwd=udp::5309-:5309 \
-display none -serial stdio -s -S

This will pause waiting on a gdb to attach to it.

In another window, we'll launch gdb pointing it at whatever kernel you are using:

gdb ~/.ops/0.1.27/kernel.img

We'll connect to qemu by specifying the remote:

(gdb) target remote localhost:1234
Remote debugging using localhost:1234
0x000000000000fff0 in ?? ()

Great - now we are connected.

Now we load the symbols for our program:

(gdb) symbol-file main
Load new symbol table from "main"? (y or n) y
Reading symbols from main...

You can see the source now:

(gdb) list
1       #include <stdio.h>
2       #include <stdlib.h>
3
4       void mybad() {
5         int x = 1;
6         char *stuff = "asdf";
7
8         printf("about to die\n");
9        *(int*)0 = 0;

Let's set a breakpoint for 'mybad':

(gdb) b mybad
Breakpoint 1 at 0x401d35: file main.c, line 4.

Now let's continue:

(gdb) c
Continuing.

Breakpoint 1, mybad () at main.c:4
4       void mybad() {

You should see in your other window (where you started qemu) that the program starts:

assigned: 10.0.2.15

Now if we single step through the program we can print out various locals:

4       void mybad() {
(gdb) step
5         int x = 1;
(gdb) step
6         char *stuff = "asdf";
(gdb) step
8         printf("about to die\n");
(gdb) p x
$1 = 1
(gdb) p stuff
$2 = 0x495004 "asdf"

This should get you further down the path when you find various bugs in your program. Hope this helps.

If your program isn't even starting there might be an issue with OPS.

However, if you open an issue in Nanos please provide the following:

Your OPS profile:

➜  ~  ops profile
Ops version: 0.1.9
Nanos version: 0.1.25
Qemu version: 4.1.90
Arch: darwin

Nightly vs Master?

Does this work on the nightly build? Running '-n' will run ops with whatever was in the master branch last night.

Reproducible steps

GUI Debugger in VSCode

This part of the debugging guide shows how to use the GNU Debugger (GDB) in combination with VSCode to get better visualization of the debugging process. It requires that the Native Debug extension is installed.

Prerequisite

Launch the application in debug mode with ops:

$ ops run -d main
booting ~/.ops/images/main.img ...
You have disabled hardware acceleration

Waiting for gdb connection. Connect to qemu through "(gdb) target remote localhost:1234"
See further instructions in https://nanovms.gitbook.io/ops/debugging

Attach the Debugger

  1. Click the Run icon on the left sidebar (alternatively use Ctrl+Shift+D) and then create a launch.json file.

  2. Select GDB as the environment. This will create an autogenerated launch.json file.

  3. Replace the contents of the launch.json file with following:

{
    // Use IntelliSense to learn about possible attributes.
    // Hover to view descriptions of existing attributes.
    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
    "version": "0.2.0",
    "configurations": [
        {
            "name": "Debug",
            "type": "gdb",
            "request": "attach",
            "executable": "${workspaceFolder}/main",
            "target": "localhost:1234",
            "remote": true,
            "cwd": "${workspaceRoot}",
            "valuesFormatting": "parseText"
        }
    ]
}

  1. Set a Breakpoint in the source file (main.c) and start the debugging session from the Run on the left sidebar (alternatively use Ctrl+Shift+D) and click on the > Debug icon.

It is now possible to use the debugging palette to debug the application code.

Dump

Ops provides a tool that allow you to inspect image crash logs and image manifests. The dump tool binary is inside the ops version directory (~/.ops/<ops_version>/dump). Make sure you use the dump tool of the same ops version you used to build the image you want to analyze.

If the application crashes the unikernel writes the error stack to a log file before exiting. You are able to see the log content if you run the command dump -l <image_path>.

$ dump -l ~/.ops/main.img
detected filesystem at 0x11600
klog offset: 0x10600
klog size: 4096 bytes
boot id: 0
exit code: 1

en1: assigned 10.0.2.15
2021/03/22 11:35:35 Failed

The image manifest has details about the image like:

  • files and their paths in the filesystem;

  • environment variable values, including nanos base image version used;

  • program and arguments to run on initialization;

  • static IP, gateway and netmask to configure the network interface;

  • etc. You can look into your image manifest by using the command dump <image_path>.

$ dump ~/.ops/main.img
detected filesystem at 0xc11600
Label:
UUID: ad63ee2d-58d1-336f-7484-9fc81f3bc835
metadata (
program:/main
arguments:(0:main)
environment:(PWD:/ NANOS_VERSION:0.1.32 USER:root)
children:(
    .:
    proc:(children:(
        .:
        ..:
        self:(mtime:6942441066663300181 atime:6942441066663300181 children:(
            .:
            ..:
            exe:(mtime:6942441066669261459 atime:6942441066669261459 linktarget:/main)
        ))
        sys:(children:(
            .:
            ..:
            kernel:(children:(
                .:
                ..:
                hostname:(extents:(0:(allocated:1 length:1 offset:5668)) filelength:7)
            ))
        ))
    ))
    etc:(children:(
        ssl:(children:(
            .:
            ..:
            certs:(children:(
                ca-certificates.crt:(extents:(0:(allocated:406 length:406 offset:1078)) filelength:207436)
                .:
                ..:
            ))
        ))
        passwd:(extents:(0:(allocated:1 length:1 offset:1485)) filelength:33)
        .:
        ..:
        resolv.conf:(extents:(0:(allocated:1 length:1 offset:1484)) filelength:18)
    ))
    ..:
    lib:(children:(
        .:
        ..:
        x86_64-linux-gnu:(children:(
            libnss_dns.so.2:(extents:(0:(allocated:53 length:53 offset:1025)) filelength:26936)
            .:
            ..:
        ))
    ))
    main:(extents:(0:(allocated:4182 length:4182 offset:1486)) filelength:2140732)
)
booted:)

FUSE

Nanos has a FUSE driver which allows us to mount the TFS image used on the host filesystem. This makes it easy for rapid development, ease of debugging and other interesting tools.

If you'd like to read more check out https://nanovms.com/dev/tutorials/nanos-unikernel-has-fuse-driver-for-tfs .

To try it out on Linux:

sudo apt-get install libfuse-dev

To try it out on Mac:

brew install macfuse

Then you simply create a mount point and mount your desired image:

mkdir -p /tmp/mnt
~/.ops/0.1.34/tfs-fuse /tmp/mnt ~/.ops/images/myimg.img

Debug Logging

You can use the 'net console' logging feature if you'd like to log without printing via serial/vga. Simply specify it via:

{
  "ManifestPassthrough": {
    "consoles": ["+net"]
  }
}

Then you can capture it via netcat:

nc -l -u 4444

The default ip/port that Nanos will ship the logging to is 10.0.2.2 (assuming user-mode) and port of 4444.

You can adjust these via the following config though:

netconsole_ip

netconsole_port

Syscall Execution Time

OPS/Nanos has the ability to trace the number of syscall executions and timing information much like strace -c.

Here is a short c example:

#include <stdio.h>

int main() {
  printf("hello!\n");
}

The output:

eyberg@box:~/z$ ops run --syscall-summary main
booting /home/eyberg/.ops/images/main.img ...
en1: assigned 10.0.2.15
hello!

% time      seconds   usecs/call        calls       errors syscall
------ ------------ ------------ ------------------------------------------
 48.53     0.000644          215            3              brk
 25.16     0.000334          334            1              read
 16.05     0.000213          213            1              write
  3.31     0.000044            6            7              mmap
  1.73     0.000023            6            4              pread64
  1.58     0.000021            4            5            4 openat
  0.97     0.000013           13            1              close
  0.90     0.000012            3            4              mprotect
  0.67     0.000009            9            1              uname
  0.37     0.000005            5            1            1 access
  0.22     0.000003            2            2              fstat
  0.22     0.000003            1            3            3 stat
  0.22     0.000003            2            2            1 arch_prctl
------ ------------ ------------ ------------------------------------------
100.00     0.001327                        35            9 total
exit status 1

Core Dumps

Nanos supports core dumps. By default they are turned off and enabled if specifying a > 0 'coredumplimit' config variable. Ensure that the volume size is large enough to contain the core dump as well however.

Locally you can do something like this:

#include <stdio.h>
#include <stdlib.h>

int main() {
  printf("yo\n");
  abort();

  return 0;
}

with a config like:

{
  "BaseVolumeSz": "200m",
  "ManifestPassthrough": {
    "coredumplimit": "150m"
  }
}

Your size will vary with the size and scope of your application.

ops run -c config.json main

You can now see a core has been generated:

➜  ops image tree main
/
|   proc
|   |   sys
|   |   |   kernel
|   |   |   |   hostname
|   |   self
|   |   |   exe -> /main
|   lib
|   |   x86_64-linux-gnu
|   |   |   libnss_dns.so.2
|   etc
|   |   passwd
|   |   ssl
|   |   |   certs
|   |   |   |   ca-certificates.crt
|   |   resolv.conf
|   main
|   coredumps
|   |   core

Now you can copy it out:

ops image cp main coredumps/core .

And quickly view a backtrace:

➜ gdb -ex bt -ex quit main core
GNU gdb (Ubuntu 11.1-0ubuntu2) 11.1
Copyright (C) 2021 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Type "show copying" and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu".
Type "show configuration" for configuration details.
For bug reporting instructions, please see:
<https://www.gnu.org/software/gdb/bugs/>.
Find the GDB manual and other documentation resources online at:
    <http://www.gnu.org/software/gdb/documentation/>.
For help, type "help".
Type "apropos word" to search for commands related to "word"...
Reading symbols from main...
[New LWP 2]

Core was generated by `main'.
Program terminated with signal SIGABRT, Aborted.
#0  __pthread_kill_implementation (no_tid=0, signo=6, threadid=761004893632) at pthread_kill.c:44
44      pthread_kill.c: No such file or directory.
#0  __pthread_kill_implementation (no_tid=0, signo=6, threadid=761004893632) at pthread_kill.c:44
#1  __pthread_kill_internal (signo=6, threadid=761004893632) at pthread_kill.c:80
#2  __GI___pthread_kill (threadid=761004893632, signo=signo@entry=6) at pthread_kill.c:91
#3  0x0000007a1af59476 in __GI_raise (sig=sig@entry=6) at ../sysdeps/posix/raise.c:26
#4  0x0000007a1af3f7b7 in __GI_abort () at abort.c:79
#5  0x000000000056d185 in main () at main.c:6

Tracing

If you are using '--trace' you can set an optional 'notrace' variable to exclude certain output such as:

"notrace": ["open", "close"]

You can also do the opposite such as:

"tracelist": ["open", "close"]

You can also trace groups of related syscalls:

"tracelist": ["%file"]

The supported groups are:

  • %file

  • %desc

  • %network

  • %process

  • %signal

  • %memory

PreviousIntegrationsNextCrossBuilding

Last updated