Re: [PATCH 2.6.17-rc6 7/9] Remove some of the kmemleak false positives

[Ingo Molnar <mingo@elte.hu>]

* Catalin Marinas <catalin.marinas@gmail.com> wrote:

> My opinion is not to implement the "anywhere inside a block" method as 
> it would increase the risk of false negatives with a little benefit 
> (removing some false positive notifications, probably less than 30).

agreed.

> To the other extreme is Ingo's suggestion of using exact type 
> identification but I don't think this would be acceptable for the 
> kernel as it would to modify all the memory alloc calls in the kernel 
> to either pass an additional parameter (the type id) or another 
> post-allocation call to kmemleak to update the id.

passing in the type ID wouldnt be that bad and it would have other 
advantages as well: for example we could do strict type-checking of 
allocation size versus type-we-use-it-for.

As long as the conversion is gradual i think we could try this. I.e. 
we'd default to 'no ID passed', and in that case we would fall back to 
the size-based method and generate an ID out of the structure size.

> Anyway, the current implementation (I'll update it for 2.6.17) detects 
> real memory leaks. I suspect that a wide range of leaks would be 
> covered if it is used on different platforms and different conditions.

btw., what leaks were found so far? I know about the ACPI one - any 
other ones?

	Ingo