crypto: gf128mul - switch gf128mul_x_ble to le128

Currently, gf128mul_x_ble works with pointers to be128, even though it
actually interprets the words as little-endian. Consequently, it uses
cpu_to_le64/le64_to_cpu on fields of type __be64, which is incorrect.

This patch fixes that by changing the function to accept pointers to
le128 and updating all users accordingly.

Signed-off-by: Ondrej Mosnacek <[e-mail]>
Reviewd-by: Eric Biggers <[e-mail]>
Signed-off-by: Herbert Xu <[e-mail]>

arch/x86/crypto/camellia_glue.c

···
 		       struct scatterlist *src, unsigned int nbytes)
 {
 	struct camellia_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
-	be128 buf[2 * 4];
+	le128 buf[2 * 4];
 	struct xts_crypt_req req = {
 		.tbuf = buf,
 		.tbuflen = sizeof(buf),
···
 		       struct scatterlist *src, unsigned int nbytes)
 {
 	struct camellia_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
-	be128 buf[2 * 4];
+	le128 buf[2 * 4];
 	struct xts_crypt_req req = {
 		.tbuf = buf,
 		.tbuflen = sizeof(buf),

arch/x86/crypto/serpent_sse2_glue.c

···
 		       struct scatterlist *src, unsigned int nbytes)
 {
 	struct serpent_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
-	be128 buf[SERPENT_PARALLEL_BLOCKS];
+	le128 buf[SERPENT_PARALLEL_BLOCKS];
 	struct crypt_priv crypt_ctx = {
 		.ctx = &ctx->crypt_ctx,
 		.fpu_enabled = false,
···
 		       struct scatterlist *src, unsigned int nbytes)
 {
 	struct serpent_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
-	be128 buf[SERPENT_PARALLEL_BLOCKS];
+	le128 buf[SERPENT_PARALLEL_BLOCKS];
 	struct crypt_priv crypt_ctx = {
 		.ctx = &ctx->crypt_ctx,
 		.fpu_enabled = false,

arch/x86/crypto/twofish_glue_3way.c

···
 		       struct scatterlist *src, unsigned int nbytes)
 {
 	struct twofish_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
-	be128 buf[3];
+	le128 buf[3];
 	struct xts_crypt_req req = {
 		.tbuf = buf,
 		.tbuflen = sizeof(buf),
···
 		       struct scatterlist *src, unsigned int nbytes)
 {
 	struct twofish_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
-	be128 buf[3];
+	le128 buf[3];
 	struct xts_crypt_req req = {
 		.tbuf = buf,
 		.tbuflen = sizeof(buf),

crypto/xts.c

···
 };
 
 struct rctx {
-	be128 buf[XTS_BUFFER_SIZE / sizeof(be128)];
+	le128 buf[XTS_BUFFER_SIZE / sizeof(le128)];
 
-	be128 t;
+	le128 t;
 
-	be128 *ext;
+	le128 *ext;
 
 	struct scatterlist srcbuf[2];
 	struct scatterlist dstbuf[2];
···
 static int post_crypt(struct skcipher_request *req)
 {
 	struct rctx *rctx = skcipher_request_ctx(req);
-	be128 *buf = rctx->ext ?: rctx->buf;
+	le128 *buf = rctx->ext ?: rctx->buf;
 	struct skcipher_request *subreq;
 	const int bs = XTS_BLOCK_SIZE;
 	struct skcipher_walk w;
···
 
 	while (w.nbytes) {
 		unsigned int avail = w.nbytes;
-		be128 *wdst;
+		le128 *wdst;
 
 		wdst = w.dst.virt.addr;
 
 		do {
-			be128_xor(wdst, buf++, wdst);
+			le128_xor(wdst, buf++, wdst);
 			wdst++;
 		} while ((avail -= bs) >= bs);
 
···
 static int pre_crypt(struct skcipher_request *req)
 {
 	struct rctx *rctx = skcipher_request_ctx(req);
-	be128 *buf = rctx->ext ?: rctx->buf;
+	le128 *buf = rctx->ext ?: rctx->buf;
 	struct skcipher_request *subreq;
 	const int bs = XTS_BLOCK_SIZE;
 	struct skcipher_walk w;
···
 
 	while (w.nbytes) {
 		unsigned int avail = w.nbytes;
-		be128 *wsrc;
-		be128 *wdst;
+		le128 *wsrc;
+		le128 *wdst;
 
 		wsrc = w.src.virt.addr;
 		wdst = w.dst.virt.addr;
 
 		do {
 			*buf++ = rctx->t;
-			be128_xor(wdst++, &rctx->t, wsrc++);
+			le128_xor(wdst++, &rctx->t, wsrc++);
 			gf128mul_x_ble(&rctx->t, &rctx->t);
 		} while ((avail -= bs) >= bs);
 
···
 	const unsigned int max_blks = req->tbuflen / bsize;
 	struct blkcipher_walk walk;
 	unsigned int nblocks;
-	be128 *src, *dst, *t;
-	be128 *t_buf = req->tbuf;
+	le128 *src, *dst, *t;
+	le128 *t_buf = req->tbuf;
 	int err, i;
 
 	BUG_ON(max_blks < 1);
···
 		return err;
 
 	nblocks = min(nbytes / bsize, max_blks);
-	src = (be128 *)walk.src.virt.addr;
-	dst = (be128 *)walk.dst.virt.addr;
+	src = (le128 *)walk.src.virt.addr;
+	dst = (le128 *)walk.dst.virt.addr;
 
 	/* calculate first value of T */
 	req->tweak_fn(req->tweak_ctx, (u8 *)&t_buf[0], walk.iv);
···
 				t = &t_buf[i];
 
 				/* PP <- T xor P */
-				be128_xor(dst + i, t, src + i);
+				le128_xor(dst + i, t, src + i);
 			}
 
 			/* CC <- E(Key2,PP) */
···
 
 			/* C <- T xor CC */
 			for (i = 0; i < nblocks; i++)
-				be128_xor(dst + i, dst + i, &t_buf[i]);
+				le128_xor(dst + i, dst + i, &t_buf[i]);
 
 			src += nblocks;
 			dst += nblocks;
···
 			nblocks = min(nbytes / bsize, max_blks);
 		} while (nblocks > 0);
 
-		*(be128 *)walk.iv = *t;
+		*(le128 *)walk.iv = *t;
 
 		err = blkcipher_walk_done(desc, &walk, nbytes);
 		nbytes = walk.nbytes;
···
 			break;
 
 		nblocks = min(nbytes / bsize, max_blks);
-		src = (be128 *)walk.src.virt.addr;
-		dst = (be128 *)walk.dst.virt.addr;
+		src = (le128 *)walk.src.virt.addr;
+		dst = (le128 *)walk.dst.virt.addr;
 	}
 
 	return err;

include/crypto/gf128mul.h

···
 }
 
 /* needed by XTS */
-static inline void gf128mul_x_ble(be128 *r, const be128 *x)
+static inline void gf128mul_x_ble(le128 *r, const le128 *x)
 {
 	u64 a = le64_to_cpu(x->a);
 	u64 b = le64_to_cpu(x->b);
 
 	/* equivalent to gf128mul_table_be[b >> 63] (see crypto/gf128mul.c): */
-	u64 _tt = gf128mul_mask_from_bit(b, 63) & 0x87;
+	u64 _tt = gf128mul_mask_from_bit(a, 63) & 0x87;
 
-	r->a = cpu_to_le64((a << 1) ^ _tt);
-	r->b = cpu_to_le64((b << 1) | (a >> 63));
+	r->a = cpu_to_le64((a << 1) | (b >> 63));
+	r->b = cpu_to_le64((b << 1) ^ _tt);
 }
 
 /* 4k table optimization */

include/crypto/xts.h

···
 #define XTS_BLOCK_SIZE 16
 
 struct xts_crypt_req {
-	be128 *tbuf;
+	le128 *tbuf;
 	unsigned int tbuflen;
 
 	void *tweak_ctx;