Re: [PATCH 3/3] Assert notifier_block and notifier_call are not in init section

[Chandra Seetharaman <sekharan@us.ibm.com>]

On Tue, 2006-04-25 at 12:16 -0700, Linus Torvalds wrote:
> 
> On Tue, 25 Apr 2006, Chandra Seetharaman wrote:
> > 
> > Two questions:
> > 1) related to this patch: Do you want me to generate a patch that
> > asserts only notifier calls ?
> 
> I don't really have any strong preferences. It seems a bit strange that 
> we'd do it for notifiers but not for other people. It might be better to 
> try to build it into the build system itself, and get it through the 
> _normal_ "section checking".

I 'll hold off for now then.
> 
> One way to do that would be to make the "register_notifier()" thing just 
> create this dummy asm() that just puts the arguments into a section that 
> doesn't even get loaded, but that cna be checked.
> 
> > 2) Unrelated to this patch: If the _code_ section is never reallocated
> > or reused, what is the purpose of putting _code_ in the init section ?
> > Only to make sure that the init calls are called in order ?
> 
> No, the code section is re-used, it's just never re-used for any other 
> code (since we don't generate code on the fly). So if you pass in a 
> function pointer, you know that if it's in the init section, it means that 
> init-code that was discarded.
> 
> But if you pass in a data pointer, you'll never know if it's a data 
> pointer to the original init-code section, or if it was a data pointer 
> that was just dynamically allocated after the init-code section was freed.

Thanks for the clarification. 
> 
> > PS: I fixed my mailer to put my name. sorry about that.
> 
> Looks good.
> 
> 		Linus
-- 

----------------------------------------------------------------------
    Chandra Seetharaman               | Be careful what you choose....
              - sekharan@us.ibm.com   |      .......you may get it.
----------------------------------------------------------------------