Re: [GIT PULL] core kernel fixes

[Ingo Molnar <mingo@elte.hu>]

* Linus Torvalds <torvalds@linux-foundation.org> wrote:

> On Fri, 10 Jul 2009, Ingo Molnar wrote:
> > > 
> > > but I really migth have done soemthing wrong there. It's a 
> > > simple function, but somebody needs to double-check that I 
> > > haven't made it worse.
> > 
> > Looks correct to me.
> 
> Note, I didn't look at how 'end' works, and it really does matter 
> if 'end' is an "inclusive" or "exclusive" end pointer address. So 
> my replacement overlap() function was written more as a conceptual 
> patch - I did not check the exact semantics of the arguments 
> passed in.
> 
> If 'end' is exclusive, then 'b1' should be calculated as 
> 'a1+size-1', because the ranges must have the same rules. And then 
> you should use the 'strict inequality' operators for testing the 
> ranges.

The ranges are inclusive in terms of non-overlap: we can have 
adjacent ranges with b1==a2 or b2==a1 that are still considered 
non-overlapping. Hence the sharp test you used (which is negated) 
looks correct to me.

The end-of-range symbols we use:

        if (overlap(addr, len, _text, _etext) ||
            overlap(addr, len, __start_rodata, __end_rodata))

Are all at the first byte outside of the to-be-avoided range:

        .text : {
                _text = .;      /* Text */
                *(.text)
                *(.text.*)
                _etext = . ;
        }

        ...

        __param : AT(ADDR(__param) - LOAD_OFFSET) {                     \
                VMLINUX_SYMBOL(__start___param) = .;                    \
                *(__param)                                              \
                VMLINUX_SYMBOL(__stop___param) = .;                     \
                . = ALIGN((align));                                     \
                VMLINUX_SYMBOL(__end_rodata) = .;                       \
        }                                                               \

        ...

I think ...

	Ingo