RE: mem= option for broken bioses

RE: mem= option for broken bioses

[Grover, Andrew]

> From: Pavel Machek [mailto:pavel@ucw.cz] 
> I've seen broken bios that did not mark acpi tables in e820
> tables. This allows user to override it. Please apply,

OK, looks reasonable. Can you also gen up a patch documenting this in
kernel-parameters.txt?

Thanks -- Regards -- Andy

Re: mem= option for broken bioses

[Pavel Machek]

Hi!

> > From: Pavel Machek [mailto:pavel@ucw.cz] 
> > I've seen broken bios that did not mark acpi tables in e820
> > tables. This allows user to override it. Please apply,
> 
> OK, looks reasonable. Can you also gen up a patch documenting this in
> kernel-parameters.txt?

You can, assuming you took the patch ;-).

								Pavel

--- clean/Documentation/kernel-parameters.txt	2003-02-11 17:40:28.000000000 +0100
+++ linux/Documentation/kernel-parameters.txt	2003-02-26 23:43:21.000000000 +0100
@@ -516,6 +516,10 @@
 			[KNL,BOOT] Force usage of a specific region of memory
 			Region of memory to be used, from ss to ss+nn.
 
+	mem=nn[KMG]#ss[KMG]
+			[KNL,BOOT,ACPI] Mark specific memory as ACPI data.
+			Region of memory to be used, from ss to ss+nn.
+
 	mem=nopentium	[BUGS=IA-32] Disable usage of 4MB pages for kernel
 			memory.
 
-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.

Re: [ACPI] Re: mem= option for broken bioses

[Robert Woerle Paceblade/Support]

Pavel Machek schrieb:

>Hi!
>
>  
>
>>>From: Pavel Machek [mailto:pavel@ucw.cz] 
>>>I've seen broken bios that did not mark acpi tables in e820
>>>tables. This allows user to override it. Please apply,
>>>      
>>>
>>OK, looks reasonable. Can you also gen up a patch documenting this in
>>kernel-parameters.txt?
>>    
>>
>
>You can, assuming you took the patch ;-).
>  
>
well how can i find the correct value`s to put in ??

>								Pavel
>
>--- clean/Documentation/kernel-parameters.txt	2003-02-11 17:40:28.000000000 +0100
>+++ linux/Documentation/kernel-parameters.txt	2003-02-26 23:43:21.000000000 +0100
>@@ -516,6 +516,10 @@
> 			[KNL,BOOT] Force usage of a specific region of memory
> 			Region of memory to be used, from ss to ss+nn.
> 
>+	mem=nn[KMG]#ss[KMG]
>+			[KNL,BOOT,ACPI] Mark specific memory as ACPI data.
>+			Region of memory to be used, from ss to ss+nn.
>+
> 	mem=nopentium	[BUGS=IA-32] Disable usage of 4MB pages for kernel
> 			memory.
> 
>  
>

-- 
_____________________________________
*Robert Woerle
**Technical Support | Linux
PaceBlade Technology Europe SA*
phone: +49 89 552 99935
fax: +49 89 552 99910
mobile: +49 179 474 45 27
email: robert@paceblade.com <mailto:robert@paceblade.com>
web: http://www.paceblade.com
_____________________________________

Re: [ACPI] Re: mem= option for broken bioses

[Pavel Machek]

Hi!

> >>OK, looks reasonable. Can you also gen up a patch documenting this in
> >>kernel-parameters.txt?
> >>   
> >>
> >
> >You can, assuming you took the patch ;-).
> > 
> >
> well how can i find the correct value`s to put in ??

Well, similar method to how you use mem=123@456 parameters. You just
guess them. [Given kernel messages, it is actually quite easy.]


> >--- clean/Documentation/kernel-parameters.txt	2003-02-11 
> >17:40:28.000000000 +0100
> >+++ linux/Documentation/kernel-parameters.txt	2003-02-26 
> >23:43:21.000000000 +0100
> >@@ -516,6 +516,10 @@
> >			[KNL,BOOT] Force usage of a specific region of memory
> >			Region of memory to be used, from ss to ss+nn.
> >
> >+	mem=nn[KMG]#ss[KMG]
> >+			[KNL,BOOT,ACPI] Mark specific memory as ACPI data.
> >+			Region of memory to be used, from ss to ss+nn.
> >+
> >	mem=nopentium	[BUGS=IA-32] Disable usage of 4MB pages for kernel
> >			memory.
> >


-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.

Re: [ACPI] Re: mem= option for broken bioses

[Ducrot Bruno]

On Thu, Feb 27, 2003 at 04:19:07PM +0100, Pavel Machek wrote:
> Hi!
> 
> > >>OK, looks reasonable. Can you also gen up a patch documenting this in
> > >>kernel-parameters.txt?
> > >>   
> > >>
> > >
> > >You can, assuming you took the patch ;-).
> > > 
> > >
> > well how can i find the correct value`s to put in ??
> 
> Well, similar method to how you use mem=123@456 parameters. You just
> guess them. [Given kernel messages, it is actually quite easy.]
> 

If I understand you,  you then just have to mem= with the correct
value reported via, for example:

ducrot@novae:~$ dmesg | grep 'ACPI data'
 BIOS-e820: 000000000fef0000 - 000000000feff000 (ACPI data)

Interresting :)

Ok.  Here is actually a method:
A phoenix bios (and you have one) will reserved typically 64ko
for ACPI data below ACPI NVS.

ducrot@novae:~$ dmesg | grep -i 'ACPI NVS'
 BIOS-e820: 000000000feff000 - 000000000ff00000 (ACPI NVS)
ducrot@novae:~$ python
Python 2.1.1+ (#1, Jan  8 2002, 00:37:12) 
[GCC 2.95.4 20011006 (Debian prerelease)] on linux2
Type "copyright", "credits" or "license" for more information.
>>> 0x000000000feff000 - 65536
267317248

If my bios will be incorrect, I will then pass
mem=267317248#65536
and voila.


But big problem though.  It is really really strange that the
BIOS mainteners have broken e820 call, are you sure you have
enabled acpi in BIOS, and/or power management ?

Cheers,

-- 
Ducrot Bruno

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.

Re: [ACPI] Re: mem= option for broken bioses

[Pavel Machek]

Hi!

> > > >>OK, looks reasonable. Can you also gen up a patch documenting this in
> > > >>kernel-parameters.txt?
> > > >>   
> > > >>
> > > >
> > > >You can, assuming you took the patch ;-).
> > > > 
> > > >
> > > well how can i find the correct value`s to put in ??
> > 
> > Well, similar method to how you use mem=123@456 parameters. You just
> > guess them. [Given kernel messages, it is actually quite easy.]
> > 
> 
> If I understand you,  you then just have to mem= with the correct
> value reported via, for example:
> 
> ducrot@novae:~$ dmesg | grep 'ACPI data'
>  BIOS-e820: 000000000fef0000 - 000000000feff000 (ACPI data)

Well, for you bios map is okay and you don't need mem= parameter.

> But big problem though.  It is really really strange that the
> BIOS mainteners have broken e820 call, are you sure you have
> enabled acpi in BIOS, and/or power management ?

Yes, what I needed was bios update. If you downgrade your bios you can
provoke same bug I saw.

-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.

Re: [ACPI] Re: mem= option for broken bioses

[Robert Woerle Paceblade/Support]

Pavel Machek schrieb:

>Hi!
>
>  
>
>>>>OK, looks reasonable. Can you also gen up a patch documenting this in
>>>>kernel-parameters.txt?
>>>>  
>>>>
>>>>        
>>>>
>>>You can, assuming you took the patch ;-).
>>>
>>>
>>>      
>>>
>>well how can i find the correct value`s to put in ??
>>    
>>
>
>Well, similar method to how you use mem=123@456 parameters. You just
>guess them. [Given kernel messages, it is actually quite easy.]
>
>  
>
well .. wow ... what a accurate solution `????
...
..
.
well cant you tell me more ?

>  
>
>>>--- clean/Documentation/kernel-parameters.txt	2003-02-11 
>>>17:40:28.000000000 +0100
>>>+++ linux/Documentation/kernel-parameters.txt	2003-02-26 
>>>23:43:21.000000000 +0100
>>>@@ -516,6 +516,10 @@
>>>			[KNL,BOOT] Force usage of a specific region of memory
>>>			Region of memory to be used, from ss to ss+nn.
>>>
>>>+	mem=nn[KMG]#ss[KMG]
>>>+			[KNL,BOOT,ACPI] Mark specific memory as ACPI data.
>>>+			Region of memory to be used, from ss to ss+nn.
>>>+
>>>	mem=nopentium	[BUGS=IA-32] Disable usage of 4MB pages for kernel
>>>			memory.
>>>
>>>      
>>>
>
>
>  
>

-- 
_____________________________________
*Robert Woerle
**Technical Support | Linux
PaceBlade Technology Europe SA*
phone: +49 89 552 99935
fax: +49 89 552 99910
mobile: +49 179 474 45 27
email: robert@paceblade.com <mailto:robert@paceblade.com>
web: http://www.paceblade.com
_____________________________________

Re: [ACPI] Re: mem= option for broken bioses

[Pavel Machek]

Hi!

> >Well, similar method to how you use mem=123@456 parameters. You just
> >guess them. [Given kernel messages, it is actually quite easy.]
> >
> > 
> >
> well .. wow ... what a accurate solution `????
> ...
> ..
> .
> well cant you tell me more ?

Well, you know where acpi tables start from dmesg. You can only guess
how long tables are, through.
								Pavel
-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.

Re: mem= option for broken bioses

[H. Peter Anvin]

Followup to:  <F760B14C9561B941B89469F59BA3A8471380D7@orsmsx401.jf.intel.com>
By author:    "Grover, Andrew" <andrew.grover@intel.com>
In newsgroup: linux.dev.kernel
>
> > From: Pavel Machek [mailto:pavel@ucw.cz] 
> > I've seen broken bios that did not mark acpi tables in e820
> > tables. This allows user to override it. Please apply,
> 
> OK, looks reasonable. Can you also gen up a patch documenting this in
> kernel-parameters.txt?
> 

This is very much *NOT* reasonable.  In fact, screwing around with the
syntax of the mem= parameter is poison.  I know it has already
happened, and those changes need to be reverted and the new stuff
moved to a different option.

The mem= option is unique in that it is an option that affects both
the boot loader and the kernel.  Therefore, ITS SYNTAX MUST NOT
CHANGE.

	-hpa
-- 
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!
"Unix gives you enough rope to shoot yourself in the foot."
Architectures needed: cris ia64 m68k mips64 ppc ppc64 s390 s390x sh v850 x86-64

Re: mem= option for broken bioses

[Pavel Machek]

Hi!

> > > I've seen broken bios that did not mark acpi tables in e820
> > > tables. This allows user to override it. Please apply,
> > 
> > OK, looks reasonable. Can you also gen up a patch documenting this in
> > kernel-parameters.txt?
> > 
> 
> This is very much *NOT* reasonable.  In fact, screwing around with the
> syntax of the mem= parameter is poison.  I know it has already
> happened, and those changes need to be reverted and the new stuff
> moved to a different option.
> 
> The mem= option is unique in that it is an option that affects both
> the boot loader and the kernel.  Therefore, ITS SYNTAX MUST NOT
> CHANGE.

This should be commented, somewhere.
Why is mem= option used by boot loader?
Does your bootloader really parse stuff
like mem=exactmap?
-- 
				Pavel
Written on sharp zaurus, because my Velo1 broke. If you have Velo you don't need...

Re: mem= option for broken bioses

[H. Peter Anvin]

Pavel Machek wrote:
> 
> This should be commented, somewhere.

It's commented in Documentation/i386/boot.txt as well as the 
kernel-options file.  *ALL* i386 boot loaders need to obey this, 
unfortunately, because of how the initrd protocol was defined.

> Why is mem= option used by boot loader?
> Does your bootloader really parse stuff
> like mem=exactmap?

No, and that's exactly the problem.  A whole bunch of crap was added to 
mem= retroactively (over my explicit objections), and that broke the 
boot protocol.

	-hpa