Protected Mode Memory

Index > OS Construction > Protected Mode Memory

Author

Thread

Navic

Joined: 05 Mar 2007
Posts: 16
Location: USA

Navic 05 Mar 2007, 05:07

I have a function that reads a byte from memory at every 1MB interval and stores the number of times this successfully happens, upon error, that value indicates the total megabytes of RAM on the system. This works fine on my DDR machines (P4, Athlon64, Celeron M), but on an older SDRAM machine (AMD K6-III) it returns nothing. Any ideas as to why, and/or how to fix this?

05 Mar 2007, 05:07

tantrikwizard

Joined: 13 Dec 2006
Posts: 142

tantrikwizard 05 Mar 2007, 11:28

Not sure why this is happening. What is the purpose of the code? Is there another way to work the algorithm?

05 Mar 2007, 11:28

Navic

Joined: 05 Mar 2007
Posts: 16
Location: USA

Navic 05 Mar 2007, 12:12

I have figured out that the function does work - there was an error in my printing function which is fixed..... The problem now is the huge output. It will return 100000000h - 24 which just happens to be the beginning of the algorithm I'm using (1MB intervals, at the end subtract 24). I can't tell why the code will loop this many times. Would it be easier if I pasted the code here?

05 Mar 2007, 12:12

vid
Verbosity in development

Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia

vid 05 Mar 2007, 12:56

yes, paste the code please Smile

05 Mar 2007, 12:56

Navic

Joined: 05 Mar 2007
Posts: 16
Location: USA

Navic 05 Mar 2007, 18:20

I just tried this on my Athlon64 2800 and it gave the bogus value, so my two AMD boxes report bogus sizes, and my single Intel box reports the correct amount of RAM.... dunno if that has anything to do with it.

The GDT entry is 4GB limit, R/W. If there is any other known way to determine the total amount of physical RAM in protected mode, it would be much appreciated if that function was shared.

Code:


RAM_Size:
        mov eax,10h
        mov ds,eax
        mov es,eax
        mov eax,1000000
        mov edi,eax
        mov esi,eax
        xor ecx,ecx
        mov eax,0AAh
rep_mem:
        stosb
        lodsb
        cmp eax,0AAh
        jne set_var
        inc ecx
        add esi,999998
        add edi,999998
        jmp rep_mem
set_var:
        lea edi,[mbram]
        mov eax,ecx
        sub eax,24
        stosd

05 Mar 2007, 18:20

DOS386

Joined: 08 Dec 2006
Posts: 1900

DOS386 05 Mar 2007, 20:52

Code:

mov eax,1000000

Code:

add esi,999998

Where do these values come from ??? Question

How many bytes has one MB ? Question

Quote:

at the end subtract 24

Why ?

Code:

lea edi,[mbram]
mov eax,ecx
sub eax,24
stosd

There might be simpler ways to do ...

Finally, you would have to save and compare at least 2 different
values to get sure that RAM exists at that address ...

_________________
Bug Nr.: 12345

Title: Hello World program compiles to 100 KB !!!

Status: Closed: NOT a Bug

05 Mar 2007, 20:52

Japheth

Joined: 26 Oct 2004
Posts: 151

Japheth 05 Mar 2007, 21:11

1.000.000 in decimal is F4240 in hex, which, if the base of your descriptors is 0, is a ROM address and might confuse your algorithm.

05 Mar 2007, 21:11

tantrikwizard

Joined: 13 Dec 2006
Posts: 142

tantrikwizard 05 Mar 2007, 21:26

Navic wrote:

If there is any other known way to determine the total amount of physical RAM in protected mode, it would be much appreciated if that function was shared.

Setup paging and when a page is requested that is out of bounds you will get a page fault. This can be setup in a loop, expanding the limit of a GDT entry until the exception occures.

05 Mar 2007, 21:26

DOS386

Joined: 08 Dec 2006
Posts: 1900

DOS386 05 Mar 2007, 21:48

Quote:

ROM address and might confuse your algorithm

Right ... will return 0 !!! Start testing at $180000 (1.5 MB) and jump in
$80000 (512 KB) steps ... and reject result if it is $180000
(would mean less than 2 MB RAM or algo failure) Idea

And, still, test 2 values per address ...

_________________
Bug Nr.: 12345

Title: Hello World program compiles to 100 KB !!!

Status: Closed: NOT a Bug

05 Mar 2007, 21:48

f0dder

Joined: 19 Feb 2004
Posts: 3175
Location: Denmark

f0dder 05 Mar 2007, 23:39

You shouldn't do memory detection this way... ROM, memory mapped devices, etc. Query the BIOS instead, e820.

05 Mar 2007, 23:39

tantrikwizard

Joined: 13 Dec 2006
Posts: 142

tantrikwizard 06 Mar 2007, 05:48

f0dder wrote:

You shouldn't do memory detection this way... ROM, memory mapped devices, etc. Query the BIOS instead, e820.

Agreed, but any way to perform same through PMode without BIOS? Can't one query the PCI BUS for same reserved addess spaces?

06 Mar 2007, 05:48

Navic

Joined: 05 Mar 2007
Posts: 16
Location: USA

Navic 06 Mar 2007, 07:10

Taking into account most systems that are around today have large amounts of RAM, would testing every 32MBs work? I understand there are I/O, PCI, ROM, etc spaces, but I was hoping to find a test value that would either skip these spaces..... not very probable I guess? My values were based on 1000000 starting, and since I used 0AAh to test, an increment should be 1000000-2 correct?

06 Mar 2007, 07:10

f0dder

Joined: 19 Feb 2004
Posts: 3175
Location: Denmark

f0dder 06 Mar 2007, 10:16

tantrikwizard wrote:

f0dder wrote:
You shouldn't do memory detection this way... ROM, memory mapped devices, etc. Query the BIOS instead, e820.
Agreed, but any way to perform same through PMode without BIOS? Can't one query the PCI BUS for same reserved addess spaces?

Afaik, PCI management is done through BIOS calls? (perhaps only to get a 32bit callable entrypoint - haven't done PCI setup myself). You can either query the e820 memory map before entering pmode, or you can use v86 to thunk down.

Navic: better be safe and do it the e820 way, and falling back to older mechanisms in case bios doesn't support e820. There's a few broken BIOSes that don't support e820 correctly, but IMHO those should just be ignored, unless you want to end up with a bloated "support everything and have no specific goals" kernel.

_________________

- carpe noctem

06 Mar 2007, 10:16

Navic

Joined: 05 Mar 2007
Posts: 16
Location: USA

Navic 06 Mar 2007, 17:37

Well this question will show how much of a newbie I am...... what is the e820 map?

06 Mar 2007, 17:37

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4624
Location: Argentina

LocoDelAssembly 06 Mar 2007, 22:28

http://www.ctyme.com/intr/rb-1741.htm

06 Mar 2007, 22:28

Navic

Joined: 05 Mar 2007
Posts: 16
Location: USA

Navic 07 Mar 2007, 17:59

I've found a good algorithm for physical memory detection:

Code:

;By Jeff Leyda
        xor edx,edx             ;MB counter
        mov eax,0cafef00dh ;test data
        mov edi,100000       ;start
testLoop:
        mov ebx, [edi]       ; save current memory contents
        mov [edi], eax       ; write memory 
        cmp [edi], eax       ; did it stick?
        jnz endoftest
        mov [edi], ebx       ; restore memory
        add edi, 1024*1024   ; next mb
        inc edx              ; update counter
        jmp testLoop
endoftest:
        lea edi,[mbram]
        mov eax,edx
        stosw

This works on all my systems.

07 Mar 2007, 17:59

f0dder

Joined: 19 Feb 2004
Posts: 3175
Location: Denmark

f0dder 07 Mar 2007, 23:13

Again: don't do it this way. Please, think of the kittens!

07 Mar 2007, 23:13

tom tobias

Joined: 09 Sep 2003
Posts: 1320
Location: usa

tom tobias 08 Mar 2007, 09:02

kittens .... Confused

The E820 BIOS method is a real mode solution.
If this were a purely protected mode task, then how will E820 be useful?
Embarassed

08 Mar 2007, 09:02

ilya

Joined: 29 Apr 2007
Posts: 2

ilya 07 May 2007, 13:34

I think it possible to go over all PCI devices and get the BARs largest address + range will be equal to min RAM available.

07 May 2007, 13:34

< Last Thread | Next Thread >

Forum Rules:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum