On 05/23/2012 06:48 PM, Christoph Lameter wrote:

On Wed, 23 May 2012, James Bottomley wrote:

quoted

quoted

quoted

So, why not simply patch slab to rely on the string lifetime being the
cache lifetime (or beyond) and therefore not having it take a copy?

Well thats they way it was for a long time. There must be some reason that
someone started to add this copying business....  Pekka?

The question is less why we added, but rather why we're keeping.

Of course reasoning about why it was added helps (so let's try to 
determine that), but so far the only reasonably strong argument in favor 
of keeping it was robustness.

But given that a lot of systems still uses SLAB, and we have no record 
of bugs due to dangling name pointers, this might very well be 
overzealousness on our part.