module choices affecting base kernel size???

module choices affecting base kernel size???

[Tom Oehser]

Changing all =m to =n in my config makes a 4K difference in the kernel size.

But, the kernel compiled with them off still seems to load the modules fine.

Why is there such a size difference in the static part and what do I risk if
I mix a kernel compiled with the modules off with said modules?  What stuff
is actually being compiled into the static part because of the modules I
choose?  Doesn't it defeat the point of modules to have dependencies or even
effects on the static part caused by the choice of module compilation?


-Tom

Re: module choices affecting base kernel size???

[Keith Owens]

On Thu, 2 May 2002 10:49:41 -0400 (EDT), 
Tom Oehser <tom@toms.net> wrote:
>Changing all =m to =n in my config makes a 4K difference in the kernel size.
>
>But, the kernel compiled with them off still seems to load the modules fine.
>
>Why is there such a size difference in the static part and what do I risk if
>I mix a kernel compiled with the modules off with said modules?  What stuff
>is actually being compiled into the static part because of the modules I
>choose?  Doesn't it defeat the point of modules to have dependencies or even
>effects on the static part caused by the choice of module compilation?

The majority of modules have no effect on kernel size but some modules
require base kernel code as well.  This is typically common code or low
level setup functions.  You will find that those modules will not load
now or will break.

Re: module choices affecting base kernel size???

[Martin Dalecki]

Uz.ytkownik Tom Oehser napisa?:
>>Tom Oehser <tom@toms.net> wrote:
>>
>>>Changing all =m to =n in my config makes a 4K difference in the kernel size.
>>
> 
> On Fri, 3 May 2002, Keith Owens wrote:
> 
> 
>>The majority of modules have no effect on kernel size but some modules
>>require base kernel code as well.  This is typically common code or low
>>level setup functions.  You will find that those modules will not load
>>now or will break.
> 
> 
> Great.  I must have missed the list of exactly *which* modules do this...
> 
> Any ideas on a reasonable way of how to identify them?

Please grep for EXPORT_SYMBOL().

Re: module choices affecting base kernel size???

[Tom Oehser]

> Tom Oehser <tom@toms.net> wrote:
> >Changing all =m to =n in my config makes a 4K difference in the kernel size.

On Fri, 3 May 2002, Keith Owens wrote:

> The majority of modules have no effect on kernel size but some modules
> require base kernel code as well.  This is typically common code or low
> level setup functions.  You will find that those modules will not load
> now or will break.

Great.  I must have missed the list of exactly *which* modules do this...

Any ideas on a reasonable way of how to identify them?

-Tom

Re: module choices affecting base kernel size???

[Keith Owens]

On Thu, 2 May 2002 20:57:40 -0400 (EDT), 
Tom Oehser <tom@toms.net> wrote:
>
>> Tom Oehser <tom@toms.net> wrote:
>> >Changing all =m to =n in my config makes a 4K difference in the kernel size.
>
>On Fri, 3 May 2002, Keith Owens wrote:
>
>> The majority of modules have no effect on kernel size but some modules
>> require base kernel code as well.  This is typically common code or low
>> level setup functions.  You will find that those modules will not load
>> now or will break.
>
>Great.  I must have missed the list of exactly *which* modules do this...
>
>Any ideas on a reasonable way of how to identify them?

Not easy.  Some of it is in the Makefiles, some of it is in the code.
Looking for conditional EXPORT_SYMBOLS() will find some of the extra
space, I don't know of any way of finding all of the extra code/data,
apart from compiling and measuring.

Re: module choices affecting base kernel size???

[David Woodhouse]

tom@toms.net said:
>  Any ideas on a reasonable way of how to identify them? 

A recursive grep for #if.*CONFIG_.*MODULE should find the majority of
offending code. As Keith says, some of it is in the Makefiles, but the 
majority is in the code.

--
dwmw2