It reports failures by returning NULL; that's not a good idea for
something that is supposed to allocate a crypto-algorithm-specific
object, since anything that does *NOT* need any allocations at all
can't just return NULL.
This "(void *)1" is basically "let me invent some non-NULL pointer,
I won't be using it at all". Yes, it's a kludge; there are more idiomatic
ways, but that really ought to be discussed with maintainers, especially
since there are only two places using that sucker in the first place,
one of them being in staging...
In any case, that (void *)1 is _not_ something you want to express as
ERR_CAST() (or ERR_PTR(), for that matter); that would only make the damn
thing harder for readers.
Speaking of making things harder for readers, near the 4th of these callers
there's something weird:
ops = lib80211_get_crypto_ops(alg);
if (!ops) {
char tempbuf[100];
memset(tempbuf, 0x00, 100);
sprintf(tempbuf, "%s", module);
request_module("%s", tempbuf);
ops = lib80211_get_crypto_ops(alg);
}
What the hell is going on there? module is one of the "rtllib_crypt_wep",
"rtllib_crypt_tkip" and "rtllib_crypt_ccmp"... Looks like a very odd
obfuscation; other callers also come with calls of request_module(),
but those don't do that insane dance...
