[PATCH] crypto/crc32c.c (was Re: CRC32c warning on sparc64)

[Clay Haapala <chaapala@cisco.com>]

On Thu, 6 May 2004, Meelis Roos uttered the following:
> This is 2.6.6-rc3+BK as of today on a sparc64 (gcc 3.3.3 on Debian):
> 
>   CC [M] crypto/crc32c.o crypto/crc32c.c:89: warning: initialization
>   from incompatible pointer type
> 
> This is because chksum_update uses size_t (64-bit unsigned long on
> sparc64) length argument but dia_update seems to want unsigned int
> as the type of length.
> 
> What is the right fix - change the length in chksum_update() and
> crc32c() to unsigned int?

Well, it's my opinion that using "size_t" is correct usage of type in
this case.  So here's my thinking:

* we leave lib/crc32c() as it is with size_t, as it is meant to be
derived from crc32() and that uses size_t.

* crypto/crc32c is a wrapper for lib/crc32c.  The interface it
presents to the rest of the crypto routines should agree.  My bad.
So, let us let it translate with a cast, as in the patch below.
-- 
Clay Haapala (chaapala@cisco.com) Cisco Systems SRBU +1 763-398-1056
   6450 Wedgwood Rd, Suite 130 Maple Grove MN 55311 PGP: C89240AD
  "Oh, *that* Physics Prize.  Well, I just substituted 'stupidity' for
      'dark matter' in the equations, and it all came together."

--- linux-2.6.6-rc3-bk.orig/crypto/crc32c.c	2004-05-06 14:11:54.000000000 -0500
+++ linux-2.6.6-rc3-bk/crypto/crc32c.c	2004-05-06 14:12:56.000000000 -0500
@@ -56,12 +56,12 @@
 	return 0;
 }
 
-static void chksum_update(void *ctx, const u8 *data, size_t length)
+static void chksum_update(void *ctx, const u8 *data, unsigned int length)
 {
 	struct chksum_ctx *mctx = ctx;
 	u32 mcrc;
 
-	mcrc = crc32c(mctx->crc, data, length);
+	mcrc = crc32c(mctx->crc, data, (size_t)length);
 
 	mctx->crc = mcrc;
 }