/proc/cpuinfo incomplete for AMD 386DX 40?

/proc/cpuinfo incomplete for AMD 386DX 40?

[John O'Donnell]

Please CC my address as I do not belong to the list.  TIA

I am just curious.

I have a file/web server that I updated with Slackware 8.1.  It runs
virtual web hosts, DynDNS, PPPoE, NFS, NAT and some other minor services.
This is an AMD 386 40Mhz with a 387 co-processor add-on and 32MB RAM.

It is functioning fine but I was poking around /proc/cpuinfo and it
says:

root@juanisan:/proc# cat cpuinfo
processor 
: 0
vendor_id 
: unknown
cpu family	: 3
model 
	: 0
model name	: 386
stepping 
: unknown
fdiv_bug 
: no
hlt_bug 
	: no
f00f_bug 
: no
coma_bug 
: no
fpu 
	: yes
fpu_exception 
: no
cpuid level	: -1
wp 
	: no
flags 
	:
bogomips 
: 5.17

Is there any harm in Linux not identifying stuff like the manufacturer.
I dont know if the i386 supports any extensions that would show up in
the flags field.  Think the bogomips is right?!?

Hey it runs GREAT and I love this old box I bought piece by piece when
it was new :-)  Like I said ... I was just curious.

I have not met a box Linux hasn't liked yet :-)
BTW it took half a day to compile 2.4.18 but It chugged away without a hitch!
Linux juanisan 2.4.18 #1 Wed Jun 12 19:29:12 EDT 2002 i386 unknown

TIA
Johnny O

-- 
=== Never ask a geek why, just nod your head and slowly back away.===
+==============================+====================================+
| John O'Donnell               |                                    |
|  (Sr. Systems Engineer,      | http://johnnyo.home.mindspring.com |
|  Net Admin, Webmaster, etc.) |   E-Mail: johnnyo@mindspring.com   |
+==============================+====================================+

Re: /proc/cpuinfo incomplete for AMD 386DX 40?

[Zwane Mwaikambo]

On Thu, 27 Jun 2002, John O'Donnell wrote:

> This is an AMD 386 40Mhz with a 387 co-processor add-on and 32MB RAM.

ahhh =), i have a 386/40 w/ 4M RAM which erm does my bidding.

Here is what mine looks like;

processor       : 0
vendor_id       : unknown
cpu family      : 3
model           : 0
model name      : 386
stepping        : unknown
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : no
fpu_exception   : no
cpuid level     : -1
wp              : no
flags           :
bogomips        : 7.93

> Is there any harm in Linux not identifying stuff like the manufacturer.
> I dont know if the i386 supports any extensions that would show up in
> the flags field.  Think the bogomips is right?!?

The flags field is stuff deduced from doing cpuid calls, so nothing there. 
The vendor might be a little difficult and might require depending on 
quirks of specific cpu models (i'm not 100% sure) therefore it would be a 
waste of memory to do.

> I have not met a box Linux hasn't liked yet :-)
> BTW it took half a day to compile 2.4.18 but It chugged away without a hitch!
> Linux juanisan 2.4.18 #1 Wed Jun 12 19:29:12 EDT 2002 i386 unknown

My you're patient, i build mine locally and netboot, 4M of RAM (no 
harddisk for swap) is too little to even attempt a kernel build i think.

Cheers,
	Zwane

-- 
http://function.linuxpower.ca
		

Re: /proc/cpuinfo incomplete for AMD 386DX 40?

[Sean Neakums]

commence  Zwane Mwaikambo quotation:

> On Thu, 27 Jun 2002, John O'Donnell wrote:
>> BTW it took half a day to compile 2.4.18 but It chugged away
>> without a hitch!
>> Linux juanisan 2.4.18 #1 Wed Jun 12 19:29:12 EDT 2002 i386 unknown
>
> My you're patient, i build mine locally and netboot, 4M of RAM (no
> harddisk for swap) is too little to even attempt a kernel build i
> think.

Since we're all boasting about our pathetic hardware, I'd like to note
that I built 1.2.8 on my 2M 386/16 some years ago.  It took nine days.
(There was a whole lot of paging going on.)

-- 
 /////////////////  |                  | The spark of a pin
<sneakums@zork.net> |  (require 'gnu)  | dropping, falling feather-like.
 \\\\\\\\\\\\\\\\\  |                  | There is too much noise.

Re: /proc/cpuinfo incomplete for AMD 386DX 40?

[Willy Tarreau]

Hello!


> > bogomips        : 5.17
> bogomips        : 7.93

I remember having had this last rate on my Am386DX/40 too, when the cache
was enabled on the mainboard. If I disabled it, it dropped to about 5.2,
which might explain differences noticed here. So check if you have some
cache on your motherboard, and if it's enabled in your bios setup.  And
don't trust these boards with fake plastic chips labelled "write back"
with no other vendor name, and for which the bios reported "Write Back
cache ON" instead of a size.

> > Is there any harm in Linux not identifying stuff like the manufacturer.
> > I dont know if the i386 supports any extensions that would show up in
> > the flags field.  Think the bogomips is right?!?
> 
> The flags field is stuff deduced from doing cpuid calls, so nothing there. 
> The vendor might be a little difficult and might require depending on 
> quirks of specific cpu models (i'm not 100% sure) therefore it would be a 
> waste of memory to do.

CPUID was introduced in latest Intel's 486, when there was a lot of relabelling
of cheaper AMDs to Intel equivalents with higher frequencies (eg: Amd486-50 ->
i486-66). AMD took the step too at the time they were sending their new
DX4/write-back core, IIRC. But I've never seen a 386 with a CPUID instruction,
and trust me, I've searched for many ways to differenciate Intel's from AMD's.
Even the register values after reset were the same as intel's. And they were
very hard to catch because you had to reset the CPU and bypass the bios to
get the values, then restore all its context. The only noticeable difference
I found was that they didn't prefetch instructions the same way, and when you
disabled the external cache, you could notice a different pipeline stall depending
on instruction alignment.

So no reliable means to do what you want without opening the case, IMHO.

Cheers,
Willy

Re: /proc/cpuinfo incomplete for AMD 386DX 40?

[John O'Donnell]

OK - Thank you all - I feel better now.  :-)
I tried fooling with the cache to no avail.  still 5.17
But I dont care - Linux humms on this minor beast of a system.
I was just looking for some clarification and I got it.
Now I can put the 386 back in the closet and let it fun a few years more! :-)
Thanks again!
Johnny O

Willy Tarreau wrote:
> Hello!
> 
>>>bogomips        : 5.17
>>
>>bogomips        : 7.93
> 
> I remember having had this last rate on my Am386DX/40 too, when the cache
> was enabled on the mainboard. If I disabled it, it dropped to about 5.2,
> which might explain differences noticed here. So check if you have some
> cache on your motherboard, and if it's enabled in your bios setup.  And
> don't trust these boards with fake plastic chips labelled "write back"
> with no other vendor name, and for which the bios reported "Write Back
> cache ON" instead of a size.
> 
> 
>>>Is there any harm in Linux not identifying stuff like the manufacturer.
>>>I dont know if the i386 supports any extensions that would show up in
>>>the flags field.  Think the bogomips is right?!?
>>
>>The flags field is stuff deduced from doing cpuid calls, so nothing there. 
>>The vendor might be a little difficult and might require depending on 
>>quirks of specific cpu models (i'm not 100% sure) therefore it would be a 
>>waste of memory to do.
> 
> 
> CPUID was introduced in latest Intel's 486, when there was a lot of relabelling
> of cheaper AMDs to Intel equivalents with higher frequencies (eg: Amd486-50 ->
> i486-66). AMD took the step too at the time they were sending their new
> DX4/write-back core, IIRC. But I've never seen a 386 with a CPUID instruction,
> and trust me, I've searched for many ways to differenciate Intel's from AMD's.
> Even the register values after reset were the same as intel's. And they were
> very hard to catch because you had to reset the CPU and bypass the bios to
> get the values, then restore all its context. The only noticeable difference
> I found was that they didn't prefetch instructions the same way, and when you
> disabled the external cache, you could notice a different pipeline stall depending
> on instruction alignment.
> 
> So no reliable means to do what you want without opening the case, IMHO.
> 
> Cheers,
> Willy
> 
> 



-- 
=== Never ask a geek why, just nod your head and slowly back away.===
+==============================+====================================+
| John O'Donnell               |                                    |
|  (Sr. Systems Engineer,      | http://johnnyo.home.mindspring.com |
|  Net Admin, Webmaster, etc.) |   E-Mail: johnnyo@mindspring.com   |
+==============================+====================================+