On Wed, Oct 18, 2000 at 01:34:17PM +0200, Christoph Hellwig wrote: > On Wed, Oct 18, 2000 at 01:25:24PM +0200, Alexander S A Kjeldaas wrote: > > > > I agree that we need some changes, but I think going the other way is > > better :-). Keep the functions in crypto.h, but add even more. > > Include the *_cbc and *_ecb versions, and wrap the various > > declarations in #ifdef CONFIG_CIPHER_XXX ... #endif pairs. > > No, this doesn't make sense at all. Adding every function in a whole > class of modules to one header is not a good idea. Because it strikes > against modular additions outside the main (kerneli) tree. > If you really want to make them aviable to the whole kernel add a > separate header for each algorithm, like <crypto/des.h> or > <crypto/blowfish.h>. Anyway you have to export the symbols using > EXPORT_SYMBOL, too. > Although I think using separate header files is ok, I do not see an advantage in this case. Why optimize on an extremely remotely probable case - that the following happens: o Someone wants to develop a cipher _outside_ the kerneli patch tree _using_ the kerneli crypto API. o That some _other_ developer wants to use the aforementioned cipher _without_ using the crypto API, _but insists_ on the functions being available by #including <linux/crypto.h> and nothing else. But this is not a big deal for me. I will accept a patch that does it either way. astor -- Alexander Kjeldaas Mail: astor@xxxxxxx finger astor@xxxxxxxxxxxxxxxxx for OpenPGP key. Linux-crypto: cryptography in and on the Linux system Archive: http://mail.nl.linux.org/linux-crypto/
