lm_sensors on ThinkPads

Jean Delvare khali at linux-fr.org
Fri Oct 18 13:18:13 CEST 2002

> Output of ibm.pl:
> [root at tester lm_sensors-2.6.5]# ./ibm.pl
> RSD PTR found at 0xF7030.
> OEM string "IBM   ".
> This is an IBM system.

It means that we now have a second method for detecting IBM systems.
This one relies on ACPI instead of DMI, and is a lot faster and easier
than the DMI one.

> I did apply the dmidecode patch, here is the output of dmidecode:
> (...)
> dmi: read: Success
My previous fix was awful wrong. Could you please use the attached
source to rebuild dmidecode? This is my latest version of dmidecode (not
integrated by Alan Cox yet, please do not distribute). Anyway, I feel
like it won't really help. DMI tables are usually present below the 1MB
memory limit and yours doesn't seem to be: 
> DMI table at 0x0FF7C000.
You have 256MB RAM, right? Seems that your DMI table is near the end of
the physical memory.

> sensors-detect still presents the same output as before.
Of course, since nothing had been fixed :(

> I think that's everything.  Please let me know how it goes.
Well, please try the attached dmidecode source. I'd also like a snapshot
of your BIOS area:

dd if=/dev/mem of=bios-pam bs=64k skip=14 count=2
(as root, of course)

And if you don't mind, I'd like a snapshot of your last RAM MB:
dd if=/dev/mem of=lastmb-pam bs=1k skip=261120 count=1024
(still as root)
I guess it will fail getting the whole block, still the exact count will
help, as well as the grabbed data.

Please send the data to me directly, not on the list, since 1* It'll be
rather large 2* It may contain sensible data, it is your RAM after all.

Phil, Mark:

My conclusion anyway is that we can't rely on DMI only for IBM
detection. I plan to add the ACPI way to sensors-detect, before DMI. In
a case where ACPI is sufficent to tell we have an IBM system, dmidecode
wouldn't be called. I could also add preliminary direct (as opposed to
using dmidecode) DMI detection, to prevent from calling dmidecode if no
DMI table is present or table is out of reach (like this seems to be the
case for Pam).

Also, once Alan will have integrated my changes to dmidecode and
released version 1.8, I'll upgrade our old CVS version to this one.

Jean Delvare
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: dmidecode-1.8rc2.c
URL: <http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20021018/8dbe9615/attachment.c>

More information about the lm-sensors mailing list