Hi,

I would like to report this issue, which I have been experiencing since
I started using SKAS0 mode.

If I run two (or more) UML instances simultaneously, I get random errors
and segmentation fault reports from guest processes during the boot
phase. By "simultaneously" I mean that the UML instances are started a
very short while away from each other, with a command line like the
following:

./linux-2.6.24-rc5 ubda=$HOME/delme1,FedoraCore5-x86-root_fs mem=128M
con=xterm & \
./linux-2.6.24-rc5 ubda=$HOME/delme2,FedoraCore5-x86-root_fs mem=128M
con=xterm &

The errors seem to appear on a completely random basis. I have performed
several experiments by starting two UMLs with the above commands,
waiting for their boot to complete, killing them, removing COW files and
starting them over again.
Excerpts of unexpected guest error messages from different experiments
follow:

> Setting clock : Sun Dec 30 14:40:35 EST 2007 [  OK  ]
> /etc/rc.d/rc.sysinit: line 169:   660 Segmentation fault     
> /sbin/start_udev
> /etc/rc.d/rc.sysinit: line 172:   661 Segmentation fault      $file
> Setting hostname localhost.localdomain:  [  OK  ]

>
> INIT: version 2.86 booting
>                 Welcome to Fedora Core
>                 Press 'I' to enter interactive startup.
> /etc/rc.d/rc.sysinit: line 143:   650 Segmentation fault     
> /sbin/hwclock $CLOCKFLAGS
> Setting clock : Sun Dec 30 14:22:05 EST 2007 [  OK  ]


> Setting clock : Sun Dec 30 14:23:40 EST 2007 [  OK  ]
> cat: /proc/cmdline: No such file or directory
> Setting hostname localhost.localdomain:  [  OK  ]
> raidautorun: unable to autocreate /dev/md0
> modprobe: FATAL: Could not load
> /lib/modules/2.6.24-rc5-g2cc3a8f6-dirty/modules.dep: No such file or
> directory
>
> ERROR: finding sysfs mount point
> ERROR: finding sysfs mount point
> ERROR: finding sysfs mount point

> Setting clock : Sun Dec 30 14:31:38 EST 2007 [  OK  ]
> INIT: Entering runlevel: 3
> Bringing up loopback interface:  [  OK  ]
> Bringing up interface eth0:  Device eth0 does not seem to be present,
> delaying initialization.
> [FAILED]
> touch: cannot touch `/var/lock/subsys/network': Read-only file system
> Starting system logger: 

Apparently, there is no way to systematically reproduce this problem,
apart from attempting to launch multiple UML instances and wait for
errors to come out. If each new UML instance is started after the
previous one has completed the boot phase, the problem never arises.
Also, errors seem to happen in the boot phase: if a UML guest "survives"
the random faults, interactive usage from the shell prompt seems not to
cause any further problem, but I am not sure whether this is a coincidence.

I have been experiencing the very same problem with the following
host/UML kernel combinations:
- 2.6.21-rc4 host kernel, 2.6.22 Fedora kernel compiled as UML
- 2.6.21-rc4 host kernel, 2.6.24-rc5 UML from the official User-Mode
Linux site
- 2.6.21-rc4 host kernel, 2.6.23.1 UML
- 2.6.16.16 host kernel, 2.6.23.1 UML
Host kernels are custom compiled vanilla kernels, and they all run on
x86 32 bit hosts. I have not applied any strange patch or tweak that
might apparently justify this behavior, but I can provide you with the
configuration file if you need it. As an additional bit of information,
I have performed the experiments on hosts running Debian testing and
unstable distributions, but the host kernels were always custom compiled.

By digging around, I have come across just the following references:
- A closed bug, for which the UML kernels I have been using seem to be
already patched:
http://www.mail-archive.com/[EMAIL PROTECTED]/msg28917.html
- Your statement about random process segfaults on x86_64 hosts
- Another fix, which is rather old and is probably marginally related:
http://www.mail-archive.com/[EMAIL PROTECTED]/msg02751.html

I hope this report is useful to you.

Regards,
Massimo.


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
User-mode-linux-user mailing list
User-mode-linux-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user

Reply via email to