Lost connection to test machine

132 views
Skip to first unread message

Kenan Abdullahoglu

unread,
Jun 24, 2021, 10:37:36 AM6/24/21
to syzkaller
Dear Dmitry/Friends,

I've just installed syzkaller on my amd64 machine without any problem (syzkaller, kernel , image , ..) I verified the succesful installation by firing up qemu and login into vm via SSH without password as instructed on documentation.

However when I try syz-manager, I get continously "lost connection to test machine".

As instructed, I tried to check the issue with -debug option and what I noticed is "failed to connect to manager ... i/o timeout" message. For your inspection, I collected crash info and config file for syz-manager.However no chance to attach to this message.

Any suggestion?

Regards,
Kenan A.

Dmitry Vyukov

unread,
Jun 24, 2021, 11:15:47 AM6/24/21
to Kenan Abdullahoglu, syzkaller
Hi Kenan,

There is some issue with reverse connectivity from the VM to the host.
It highly depends on VM type and other things, hard to say anything
w/o seeing config/logs. Can you attach them to the email or upload
somewhere (e.g. gist.github.com)?

Kenan Abdullahoglu

unread,
Jun 28, 2021, 8:12:35 AM6/28/21
to syzkaller

Dmitry Vyukov

unread,
Jun 28, 2021, 8:40:57 AM6/28/21
to Kenan Abdullahoglu, syzkaller
On Mon, Jun 28, 2021 at 2:12 PM Kenan Abdullahoglu <kena...@gmail.com> wrote:
>
> Hi Dmitry,
>
> As requested,
>
> https://gist.github.com/kyabd/543170cc6d9d6c7c0b2469766746d8a9
>
> https://gist.github.com/kyabd/343c3982830c82de5544576f41b47679
>
> log file & config file attached.

Hi Kenan,

Try to remove "rpc" from the manager config. Maybe it leads to some
loopback/not loopback or ipv4/6 mismatches.

> Regards,
> Kenan
>
> On Thursday, June 24, 2021 at 6:15:47 PM UTC+3 Dmitry Vyukov wrote:
>>
>> On Thu, Jun 24, 2021 at 4:37 PM Kenan Abdullahoglu <kena...@gmail.com> wrote:
>> >
>> > Dear Dmitry/Friends,
>> >
>> > I've just installed syzkaller on my amd64 machine without any problem (syzkaller, kernel , image , ..) I verified the succesful installation by firing up qemu and login into vm via SSH without password as instructed on documentation.
>> >
>> > However when I try syz-manager, I get continously "lost connection to test machine".
>> >
>> > As instructed, I tried to check the issue with -debug option and what I noticed is "failed to connect to manager ... i/o timeout" message. For your inspection, I collected crash info and config file for syz-manager.However no chance to attach to this message.
>> >
>> > Any suggestion?
>>
>> Hi Kenan,
>>
>> There is some issue with reverse connectivity from the VM to the host.
>> It highly depends on VM type and other things, hard to say anything
>> w/o seeing config/logs. Can you attach them to the email or upload
>> somewhere (e.g. gist.github.com)?
>
> --
> You received this message because you are subscribed to the Google Groups "syzkaller" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to syzkaller+...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/syzkaller/2cf05bc6-fdf7-4e14-93c6-a1335d881c16n%40googlegroups.com.

Kenan Abdullahoglu

unread,
Jun 28, 2021, 9:11:50 AM6/28/21
to syzkaller
Dmitry,

Confige file changed to :

{
    "target": "linux/amd64",
    "http": ":31337",
    "workdir": "/home/ka/workdir",
    "kernel_obj": "/home/ka/linux",
    "image": "/home/ka/image/stretch.img",
    "sshkey": "/home/ka/image/stretch.id_rsa",
    "syzkaller": "/home/ka/go/src/github.com/google/syzkaller",
    "procs": 8,   
    "type": "qemu",
    "vm": {
        "count": 1,
        "kernel": "/home/ka/linux/arch/x86/boot/bzImage",
        "cmdline": "console=ttyS0 root=/dev/sda earlyprintk=serial net.ifnames=0",
        "cpu": 4,
        "mem": 4096
    }
}

Result is same:

ka@syzkaller:~$ syz-manager -config syzkaller.cfg
2021/06/28 16:05:43 loading corpus...
2021/06/28 16:05:43 serving http on http://:31337
2021/06/28 16:05:43 serving rpc on tcp://[::]:46233
2021/06/28 16:05:43 booting test machines...
2021/06/28 16:05:43 wait for the connection from test machine...
2021/06/28 16:08:13 vm-0: crash: lost connection to test machine
2021/06/28 16:09:17 vm-0: crash: lost connection to test machine

Regards,
Kenan

Dmitry Vyukov

unread,
Jun 28, 2021, 9:25:20 AM6/28/21
to Kenan Abdullahoglu, syzkaller
Are you sure it's the same? Maybe it's something different in the log.
If it's the same, I don't have any other ideas. I never saw such an
error before.
> To view this discussion on the web visit https://groups.google.com/d/msgid/syzkaller/2a24f16e-de41-44c1-855a-a2d122179122n%40googlegroups.com.

Kenan Abdullahoglu

unread,
Jun 28, 2021, 9:33:33 AM6/28/21
to syzkaller
Dmitry,

I checked the log, it's same error. I am gonna change host linux mint to debian and try again.Thanks for your time.

Regards,
Kenan

Dmitry Vyukov

unread,
Jun 28, 2021, 9:40:00 AM6/28/21
to Kenan Abdullahoglu, syzkaller
On Mon, Jun 28, 2021 at 3:33 PM Kenan Abdullahoglu <kena...@gmail.com> wrote:
>
> Dmitry,
>
> I checked the log, it's same error. I am gonna change host linux mint to debian and try again.Thanks for your time.

If you want to dig, you may start qemu manually and try to connect
from the VM to the host in the same way (exact qemu/ssh commands
syz-manager use are in the log). That should be a simpler setup to
figure out what changes to args make it work and/or why it's not
working.
> To view this discussion on the web visit https://groups.google.com/d/msgid/syzkaller/e800486a-d328-4b6a-94fc-389797c99bban%40googlegroups.com.

Pavel Skripkin

unread,
Jun 28, 2021, 9:44:49 AM6/28/21
to Kenan Abdullahoglu, syzkaller
Hi, Kenan!

I had similar problem on openSUSE host. The problem was in glibc
version mismatch on host and stretch image. On host I had 2.33, but
streth image had old glibc version. Try to load syz-* binaries via ssh
manualy and to run them on vm.




With regards,
Pavel Skripkin

Kenan Abdullahoglu

unread,
Jun 29, 2021, 5:18:13 AM6/29/21
to syzkaller
Hi Pavel,

Thanks for the explanation. I changed my host from linux mint to ubuntu server, then did everything exactly the same. And the result is; syzkaller works like a charm.

Regards,
Kenan

Pavel Skripkin

unread,
Jun 29, 2021, 8:07:53 AM6/29/21
to Kenan Abdullahoglu, syzkaller
On Tue, 29 Jun 2021 02:18:13 -0700 (PDT)
Kenan Abdullahoglu <kena...@gmail.com> wrote:

> Hi Pavel,
>
> Thanks for the explanation. I changed my host from linux mint to
> ubuntu server, then did everything exactly the same. And the result
> is; syzkaller works like a charm.
>

Glad to help you ;) Alternative solution is to debootstrap debian hirsute image,
it has newest glibc version and it should work with any newest distro
as host, I guess.

You can slightly change create_image.sh script to do it automaticaly,
but you should add some systemd stuff to enable ttyS0 service at least.
With regards,
Pavel Skripkin
Reply all
Reply to author
Forward
0 new messages