Re: [PATCH] kmemleak: Only use GFP_KERNEL|GFP_ATOMIC for the internal allocations

[Catalin Marinas <catalin.marinas@arm.com>]

On Wed, 2009-06-17 at 16:01 +0300, Pekka Enberg wrote:
> On Wed, Jun 17, 2009 at 3:52 PM, Catalin Marinas<catalin.marinas@arm.com> wrote:
> > Kmemleak allocates memory for pointer tracking and it tries to avoid
> > using GFP_ATOMIC if the caller doesn't require it. However other gfp
> > flags may be passed by the caller which aren't required by kmemleak.
> > This patch filters the gfp flags so that only GFP_KERNEL | GFP_ATOMIC
> > are used.
> >
> > Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
> 
> Is this really safe? What if we're in a middle of a filesystem
> operation that uses GFP_NOFAIL and all of a sudden kmemleak allocation
> fails and causes a panic?

A kmemleak_panic() call only disables the kmemleak but doesn't stop the
kernel. The __GFP_NOFAIL allocation had already happened and the pointer
returned to the caller.

If we pass the __GFP_NOFAIL flag via the kmemleak's alloc, we get the
warning reported by Ingo (though it's not clear to me why this happens
as a kmemleak_object cache allocation is 192 bytes on a 32 bit machine).

Are there (common) situations where we expect kmemleak's alloc to fail
without __GFP_NOFAIL?

-- 
Catalin