On Wed, Sep 10, 2014 at 04:26:46PM +0200, Jiri Kosina wrote: > On Wed, 10 Sep 2014, Theodore Ts'o wrote: > > > I'd much rather depending on better testing and static checkers to fix > > them, since kfree *is* a hot path. > > BTW if we stretch this argument a little bit more, we should also kill the > ZERO_OR_NULL_PTR() check from kfree() and make it callers responsibility > to perform the checking only if applicable ... we are currently doing a > lot of pointless checking in cases where caller would be able to guarantee > that the pointer is going to be non-NULL. What you're saying is that we should remove the ZERO_SIZE_PTR completely. ZERO_SIZE_PTR is a very useful idiom and also it's too late to remove it because everything depends on it. Returning ZERO_SIZE_PTR is not an error. Callers shouldn't test for it. It works like this: 1) User space says "copy zero items to somewhere." 2) The kernel says "here is a zero size pointer" 3) We do some stuff like: copy_from_user(zero_pointer, src, 0) or: for (i = 0; i < 0; i++) 4) The caller frees the ZERO_SIZE_PTR. 5) We return success. If we get rid of it then we're start returning -ENOMEM all over the place and that breaks userspace. Or we introduce zero as a special case for every kmalloc. You would think there would be a lot of bugs with ZERO_SIZE_POINTERs but they seem fairly rare to me. There are some where we allocate a zero length string and then put a NUL terminator at the end. regards, dan carpenter -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>