Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 31 Mar 2015 20:43:26 +0200
From:      Christoph Moench-Tegeder <cmt@burggraben.net>
To:        freebsd-emulation@freebsd.org
Subject:   Re: VirtualBox (sometimes) not starting/hanging after recent openssl updates
Message-ID:  <20150331184325.GA1503@elch.exwg.net>
In-Reply-To: <20150330173027.2207c592@Papi>
References:  <20150330173027.2207c592@Papi>

next in thread | previous in thread | raw e-mail | index | archive | help
## Mario Lobo (lobo@bsd.com.br):

> /usr/local/lib/virtualbox/VBoxRT.so:
> (snip)
>         libcrypto.so.8 => /usr/local/lib/libcrypto.so.8 (0x8026a7000)
>         libcrypto.so.7 => /lib/libcrypto.so.7 (0x80431d000)
>         libcrypt.so.5 => /lib/libcrypt.so.5 (0x804fee000)
> 
> and mine is fine too!

Perhaps you're lucky.
Going slightly deeper into the gory details: virtualbox uses the
environment (VBOX_XPCOM_HOME) to "find itself". Also, here and
in some reports we had the "environment corrupt" message, that's
from libc, indicating that an entry in the environment is malformed
(not NAME=VALUE) - this should (can?) not happen when you use the
environment "normally". I could get around that message (and get
virtualbox working) by making the environment smaller, that is,
removing entries (one does not really need TERMCAP or LC_* for
starting virtualbox...). I pinpointed the location of my problem
by peppering the virtualbox code with assert()s on parts of the
environment... Once I had removed the second (the base system)
libcrypto, the environment (and surrounding memory, as far as I
can tell) does not get corrupted anymore, so I claim causality
on the base of correlation :) (it's completely plausible that
data structure mismatch between the two versions of openssl
may cause out-of-bounds memory access and related trouble).

Regards,
Christoph

-- 
Spare Space



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150331184325.GA1503>