[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#506345: marked as done (linux-image-2.6-486: Facility to allow falsified information in /proc/cpuinfo)



Your message dated Thu, 20 Nov 2008 22:32:12 +0100
with message-id <20081120213212.GA8074@wavehammer.waldi.eu.org>
and subject line Re: Bug#506345: linux-image-2.6-486: Facility to allow falsified information in /proc/cpuinfo
has caused the Debian Bug report #506345,
regarding linux-image-2.6-486: Facility to allow falsified information in /proc/cpuinfo
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner@bugs.debian.org
immediately.)


-- 
506345: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=506345
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems
--- Begin Message ---
Package: linux-image-2.6-486
Version: 2.6.26+16
Severity: wishlist


It would be useul to have the facility to provide a generic entry in 
/proc/cpuinfo rather than revealing processor specific information.

This could possibly be done via a boot option, such as

cpuinfo=486 or cpuinfo=pentium

This would enable the computer to act as a generic base level computer 
for a specific architecture, providing backward compatibility with 
lower level machines. This would enable a machine to appear as a generic 
486 machine or as a generic Pentium machine, without revealing the machine is
actually a higher level machine 686 machine (or using an IBM compatible 
processor, supplied by an alternative manufacturer such as a Cyrix 686, 
or an AMD K6 or AMD K7.)

One of the problems that I am encountering is that build and runtime 
systems appear to be taking information from /proc/cpuinfo, and this is 
influencing compiler or program behaviour.

All of my machines are using 486 (IA-32) compatible processors. However, 
not all machines are true 486 machines. I have 586, 686 and K7 series 
processors on some machines, and the processors are made by various 
manufacturers (such as Intel, Cyrix and AMD.)

I require all machines to behave with "lowest common denominator" 
behaviour, and for all compiled binaries to be generic enough to move 
across machines. I am finding that distribution provided binaries, and 
third party provided build scripts are detecting that the machine doing 
the build is a 686, and is utilizing features for that architecture, 
even though the target architecture is generic 486.

I know that a workaround for this would be to modify all of the build 
scripts in every single package, and find a way of rebuilding the entire
Debian distribution from source, but I feel that a falsified /proc/cpuinfo
would be much less of a headache.

I propose that the boot time switch cpuinfo=pentium provides the following 
falsified information in /proc/cpuinfo:

vendor_id       : GenuineIntel
cpu family      : 5
model           : 2
model name      : Pentium 75 - 200

Maybe the vendor_id, model and model name could be simply be set to 
"Unknown" or "Generic" if the boot switch is used. I am not sure what 
effect this would have, but hopefully it would mean generic behaviour.

-- System Information:
Debian Release: lenny/sid
  APT prefers testing
  APT policy: (500, 'testing')
Architecture: i386 (i686)
                    ____      Damn!! This bit is wrong!

Kernel: Linux 2.6.26-1-486
Locale: LANG=en_GB, LC_CTYPE=en_GB (charmap=ISO-8859-1)
Shell: /bin/sh linked to /bin/dash

Versions of packages linux-image-2.6-486 depends on:
ii  linux-image-2.6.26-1-486      2.6.26-8   Linux 2.6.26 image on x86

linux-image-2.6-486 recommends no packages.

linux-image-2.6-486 suggests no packages.

-- no debconf information



--- End Message ---
--- Begin Message ---
On Thu, Nov 20, 2008 at 07:39:50PM +0000, Mark Hobley wrote:
> It would be useul to have the facility to provide a generic entry in 
> /proc/cpuinfo rather than revealing processor specific information.

This patch would have to come the way via upstream. Also the kernel
always provide anything to do that:

$ sudo mount -t proc none proc
$ cat proc/cpuinfo | head -n 2
processor       : 0
vendor_id       : AuthenticAMD
$ cat cpuinfo 
fake
$ sudo mount --bind cpuinfo proc/cpuinfo
$ cat proc/cpuinfo
fake

> One of the problems that I am encountering is that build and runtime 
> systems appear to be taking information from /proc/cpuinfo, and this is 
> influencing compiler or program behaviour.

This behaviour is not allowed for Debian packages.

Because it does not fix anything and it is already possible, I close this bug.

Bastian

-- 
Schshschshchsch.
		-- The Gorn, "Arena", stardate 3046.2


--- End Message ---

Reply to: