Re: Using getpid() often, another way? [was Re: clone() <-> getpid() bug in 2.6?]

[Russell Leighton <russ@elegant-software.com>]

David Mosberger wrote:

>>>>>>On Mon, 7 Jun 2004 15:00:09 +0100, Christoph Hellwig <hch@infradead.org> said:
>>>>>>            
>>>>>>
>
>  Christoph> On Mon, Jun 07, 2004 at 02:48:31PM +0100, Sean Neakums
>  Christoph> wrote:
>  >> > for example ia64 doesn't have it.
>
>  >> Then what is the sys_clone2 implementation in
>  >> arch/is64/kernel/entry.S for?
>
>  Christoph> It's clone with a slightly different calling convention.
>
>Note that the only difference is that the stack-area is expressed as a
>range (starting-address + size), rather than a direct stack-pointer
>value.  IMHO, it was a mistake to not do it that way right from the
>beginning (consider that different arches grow stacks in different
>directions, for example).
>
>	
>
So Ia64 does have it..that's good. Does glibc wrap it?

I agree with the above...could glibc's clone() should have a size added? 
Then the arch specific stack issues
could be hidden.

BTW, does gcc have a built-in #define like __STACK_GROWSUP__ that would 
allow one to deal with the missing size parameter
when you called clone() by adjusting what you passed with and #ifdef?.