[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-smp
Subject:    Re: mtrr_message
From:       Richard Gooch <rgooch () atnf ! csiro ! au>
Date:       1999-06-21 22:34:33
[Download RAW message or body]

Klaus Dittrich writes:
> To th problem with mtrr (memory type range register) settings.
> 
> I am not a hardware guru but I have read
> /usr/src/linux/Documentation/mtrr.txt which explains where mttr's are
> good for.
> 
> I have an asus-p2b-ds board with 2 PII/350 on it and also have ..
> 
> mtrr: your CPUs had inconsistent fixed MTRR settings
> mtrr: your CPUs had inconsistent variable MTRR settings
> mtrr: probably your BIOS does not setup all CPUs      

You BIOS is fucked and Linux cleans up after it.

> in my dmesg-output.
> 
> The message appears only when I set the video-memory-cache-mode
> in my bios to uscw. 
> 
> That means "uncacheable, speculative write combining" and "can improve
> display speed  by caching the display data" according to the handbook
> of the motherboard.
> The output of cat /proc/mtrr in this case looks ..
> reg00: base=0x00000000 (   0MB), size= 512MB: write-back, count=1
> reg05: base=0xe3000000 (3632MB), size=  16MB: write-combining, count=1  
> 
> 16MB is the size of the RAM on my MATROX-G200 graphics-card and
> the 512MB ist the size of RAM I have.
> What the 3632MB mean is unclear to me, maybe memory-mapping of
> graphics io-space. 
> 
> I think with uscw the system uses  write-combining transfers to the 16MB
> of the graphics-card and uses the 512MB System-RAM with write-back
> caching.
> 
> When I set video-memory-cache-mode to uc (uncacheable) the mttr-message
> didn't occur and the output of cat /proc/mtrr looks ..
> reg00: base=0x00000000 (   0MB), size= 512MB: write-back, count=1
> 
> This looks to me as there is no write-combining to any memory region 
> enabled which corresponds fine to the BIOS setting.
> 
> In both cases the system runs fine and without any failure in X11.
> But I can't feel any difference in display-speed on my most times
> single-used system.

Use x11perf. The 500x500 SHM image test is good.

> The problem (if there is one) looks as it will have been fixed by
> the mtrr-routine.
> 
> The question is, is the error message right when it means this is a
> failure of the BIOS or not ?
> 
> Can anybody (Richard Gooch ?) explain if this interpretation is ok
> and why the message comes up ?

Yep. Close enough.

				Regards,

					Richard....
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to majordomo@vger.rutgers.edu

[prev in list] [next in list] [prev in thread] [next in thread]