VirtualBox > 5.0.26 causes data corruption errors in Linux guests

[cbj4074]

An increasing number of PHP web developers are using Composer (a PHP dependency manager) and have come to rely on it heavily.

Many of those developers are also using VirtualBox, usually with Ubuntu as the guest OS, and oftentimes with a networked filesystem configuration that mounts a Windows or Mac OS volume to the guest (Ubuntu) filesystem (usually with CIFS or NFS).

In versions of VirtualBox > 5.0.26, some change seems to have been introduced that causes incessant failures with Composer when it attempts to download compressed archives and expand them with PHP's zlib_decode() function.

After much frustration (I upgraded VirtualBox to an affected version a few days ago), I found this thread, in which several people confirm that downgrading VirtualBox to 5.0.26 solves the problem:

https://github.com/composer/composer/issues/5814

I just tried this and can confirm that indeed it solves the problem.

Given that people are experiencing the exact same problem on both Mac OS and Windows hosts, it seems that the problem may be guest-specific (and perhaps even specific to particular OS running on the guest).

Does anyone know what might have changed in these recent version of VirtualBox that could be causing this odd network data corruption?

Guest OS details:

# uname -a
Linux ubuntu-vm.example.com 4.4.0-45-generic #66-Ubuntu SMP Wed Oct 19 14:12:37 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

# lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 16.04.1 LTS
Release: 16.04
Codename: xenial

Thanks in advance for any tips!

[socratis]

Do you have a simple case scenario to showcase this, without having to install Vagrant, Composer and all the other layers? Because I'll tell you that the typical answer here would be that anything that's based on Vagrant is not supported, because they change too many things. So, if you can demonstrate your problem with VirtualBox only, you'd have a better case.

[cbj4074]

Thanks for taking a look, socratis.

I'm not using Vagrant (though others might be); just "vanilla" VirtualBox.

As yet, I do not have a simple, reproducible test-case, but I will work to establish one. As you suggested, there are many layers, so it may take a while!

Thanks again.

[magnetik]

We ran quite a few tests in the bug comment on github
It seems that the error is something related to the connection behind reseted for some reason.

(You can find the strace on the gist in the comments, I cannot post links until tomorrow)

[socratis]

@magnetik
It didn't hit me at first, because I didn't realize there was networking involved, I thought the problem was on a standalone system. But after reading the comments in the github (thanks), I realized that the "smell" was familiar. And then that both 5.0.28 and 5.1.8 being similarly broken.

Let me take a wild guess: you're using NAT for networking. If so, please take a look at viewtopic.php?f=1&t=80310

There is a test build for 5.1.8, but none as of this writing for 5.0.26. Please, whoever is using the 5.1.x series, try to update to this one and see if it fixes it. And let your fellow posters in github know about it...

[magnetik]

The guest has three network cards : NAT, bridge and private network.

Running latest 5.1 test build available (111724), I still face the exact same issue.

Windows 10 host, Ubuntu 16.04 guest.

Should I give you the tcpdump reports ?

[socratis]

"Too bad" it's not the NAT issue, because that would have been easier...

I, personally, wouldn't have a clue what to do with the tcpdump reports, but I'm sure that someone could. Mind you, we're mostly users here just trying to see if we can help you faster, before escalating this to the bugtracker. Several of the developers hang in here as well, so, they will definitely notice the problem. They have an Oracle "badge" under their name on the right (where mine says Site Moderator for example).

[magnetik]

Issue is fixed with test build 111846.

Thanks

[socratis]

Yes, I noticed you reply on the ticket as well. Thanks you!

[cbj4074]

@magnetik Thank you for doing all the heavy-lifting here! Your efforts are much appreciated!