[patch] match limit with inverse [!]

[Jerome de Vivie]

Harald Welte a écrit :

> the problem is, that we compare the aligned size. the size of the structure
> is aligned to cache line bondaries, which are architecture dependent.  So
> unless the size of the new structure doesn't differ enough from the old size,
> nobody will notice.

I don't know where are those sizes is the kernel tree. If this potential
bug can't be checked, i don't see a clean way to solve this
compatibility problem.

j.