1 files changed, 278 insertions, 0 deletions
diff --git a/arch/x86/crypto/twofish_glue_3way.c b/arch/x86/crypto/twofish_glue_3way.c
new file mode 100644
index 000000000..768af6075
--- /dev/null
+++ b/arch/x86/crypto/twofish_glue_3way.c
@@ -0,0 +1,278 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Glue Code for 3-way parallel assembler optimized version of Twofish
+ *
+ * Copyright (c) 2011 Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
+ */
+
+#include <asm/crypto/glue_helper.h>
+#include <asm/crypto/twofish.h>
+#include <crypto/algapi.h>
+#include <crypto/b128ops.h>
+#include <crypto/internal/skcipher.h>
+#include <crypto/twofish.h>
+#include <linux/crypto.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/types.h>
+
+EXPORT_SYMBOL_GPL(__twofish_enc_blk_3way);
+EXPORT_SYMBOL_GPL(twofish_dec_blk_3way);
+
+static int twofish_setkey_skcipher(struct crypto_skcipher *tfm,
+				   const u8 *key, unsigned int keylen)
+{
+	return twofish_setkey(&tfm->base, key, keylen);
+}
+
+static inline void twofish_enc_blk_3way(const void *ctx, u8 *dst, const u8 *src)
+{
+	__twofish_enc_blk_3way(ctx, dst, src, false);
+}
+
+static inline void twofish_enc_blk_xor_3way(const void *ctx, u8 *dst,
+					    const u8 *src)
+{
+	__twofish_enc_blk_3way(ctx, dst, src, true);
+}
+
+void twofish_dec_blk_cbc_3way(const void *ctx, u8 *d, const u8 *s)
+{
+	u128 ivs[2];
+	u128 *dst = (u128 *)d;
+	const u128 *src = (const u128 *)s;
+
+	ivs[0] = src[0];
+	ivs[1] = src[1];
+
+	twofish_dec_blk_3way(ctx, (u8 *)dst, (u8 *)src);
+
+	u128_xor(&dst[1], &dst[1], &ivs[0]);
+	u128_xor(&dst[2], &dst[2], &ivs[1]);
+}
+EXPORT_SYMBOL_GPL(twofish_dec_blk_cbc_3way);
+
+void twofish_enc_blk_ctr(const void *ctx, u8 *d, const u8 *s, le128 *iv)
+{
+	be128 ctrblk;
+	u128 *dst = (u128 *)d;
+	const u128 *src = (const u128 *)s;
+
+	if (dst != src)
+		*dst = *src;
+
+	le128_to_be128(&ctrblk, iv);
+	le128_inc(iv);
+
+	twofish_enc_blk(ctx, (u8 *)&ctrblk, (u8 *)&ctrblk);
+	u128_xor(dst, dst, (u128 *)&ctrblk);
+}
+EXPORT_SYMBOL_GPL(twofish_enc_blk_ctr);
+
+void twofish_enc_blk_ctr_3way(const void *ctx, u8 *d, const u8 *s, le128 *iv)
+{
+	be128 ctrblks[3];
+	u128 *dst = (u128 *)d;
+	const u128 *src = (const u128 *)s;
+
+	if (dst != src) {
+		dst[0] = src[0];
+		dst[1] = src[1];
+		dst[2] = src[2];
+	}
+
+	le128_to_be128(&ctrblks[0], iv);
+	le128_inc(iv);
+	le128_to_be128(&ctrblks[1], iv);
+	le128_inc(iv);
+	le128_to_be128(&ctrblks[2], iv);
+	le128_inc(iv);
+
+	twofish_enc_blk_xor_3way(ctx, (u8 *)dst, (u8 *)ctrblks);
+}
+EXPORT_SYMBOL_GPL(twofish_enc_blk_ctr_3way);
+
+static const struct common_glue_ctx twofish_enc = {
+	.num_funcs = 2,
+	.fpu_blocks_limit = -1,
+
+	.funcs = { {
+		.num_blocks = 3,
+		.fn_u = { .ecb = twofish_enc_blk_3way }
+	}, {
+		.num_blocks = 1,
+		.fn_u = { .ecb = twofish_enc_blk }
+	} }
+};
+
+static const struct common_glue_ctx twofish_ctr = {
+	.num_funcs = 2,
+	.fpu_blocks_limit = -1,
+
+	.funcs = { {
+		.num_blocks = 3,
+		.fn_u = { .ctr = twofish_enc_blk_ctr_3way }
+	}, {
+		.num_blocks = 1,
+		.fn_u = { .ctr = twofish_enc_blk_ctr }
+	} }
+};
+
+static const struct common_glue_ctx twofish_dec = {
+	.num_funcs = 2,
+	.fpu_blocks_limit = -1,
+
+	.funcs = { {
+		.num_blocks = 3,
+		.fn_u = { .ecb = twofish_dec_blk_3way }
+	}, {
+		.num_blocks = 1,
+		.fn_u = { .ecb = twofish_dec_blk }
+	} }
+};
+
+static const struct common_glue_ctx twofish_dec_cbc = {
+	.num_funcs = 2,
+	.fpu_blocks_limit = -1,
+
+	.funcs = { {
+		.num_blocks = 3,
+		.fn_u = { .cbc = twofish_dec_blk_cbc_3way }
+	}, {
+		.num_blocks = 1,
+		.fn_u = { .cbc = twofish_dec_blk }
+	} }
+};
+
+static int ecb_encrypt(struct skcipher_request *req)
+{
+	return glue_ecb_req_128bit(&twofish_enc, req);
+}
+
+static int ecb_decrypt(struct skcipher_request *req)
+{
+	return glue_ecb_req_128bit(&twofish_dec, req);
+}
+
+static int cbc_encrypt(struct skcipher_request *req)
+{
+	return glue_cbc_encrypt_req_128bit(twofish_enc_blk, req);
+}
+
+static int cbc_decrypt(struct skcipher_request *req)
+{
+	return glue_cbc_decrypt_req_128bit(&twofish_dec_cbc, req);
+}
+
+static int ctr_crypt(struct skcipher_request *req)
+{
+	return glue_ctr_req_128bit(&twofish_ctr, req);
+}
+
+static struct skcipher_alg tf_skciphers[] = {
+	{
+		.base.cra_name		= "ecb(twofish)",
+		.base.cra_driver_name	= "ecb-twofish-3way",
+		.base.cra_priority	= 300,
+		.base.cra_blocksize	= TF_BLOCK_SIZE,
+		.base.cra_ctxsize	= sizeof(struct twofish_ctx),
+		.base.cra_module	= THIS_MODULE,
+		.min_keysize		= TF_MIN_KEY_SIZE,
+		.max_keysize		= TF_MAX_KEY_SIZE,
+		.setkey			= twofish_setkey_skcipher,
+		.encrypt		= ecb_encrypt,
+		.decrypt		= ecb_decrypt,
+	}, {
+		.base.cra_name		= "cbc(twofish)",
+		.base.cra_driver_name	= "cbc-twofish-3way",
+		.base.cra_priority	= 300,
+		.base.cra_blocksize	= TF_BLOCK_SIZE,
+		.base.cra_ctxsize	= sizeof(struct twofish_ctx),
+		.base.cra_module	= THIS_MODULE,
+		.min_keysize		= TF_MIN_KEY_SIZE,
+		.max_keysize		= TF_MAX_KEY_SIZE,
+		.ivsize			= TF_BLOCK_SIZE,
+		.setkey			= twofish_setkey_skcipher,
+		.encrypt		= cbc_encrypt,
+		.decrypt		= cbc_decrypt,
+	}, {
+		.base.cra_name		= "ctr(twofish)",
+		.base.cra_driver_name	= "ctr-twofish-3way",
+		.base.cra_priority	= 300,
+		.base.cra_blocksize	= 1,
+		.base.cra_ctxsize	= sizeof(struct twofish_ctx),
+		.base.cra_module	= THIS_MODULE,
+		.min_keysize		= TF_MIN_KEY_SIZE,
+		.max_keysize		= TF_MAX_KEY_SIZE,
+		.ivsize			= TF_BLOCK_SIZE,
+		.chunksize		= TF_BLOCK_SIZE,
+		.setkey			= twofish_setkey_skcipher,
+		.encrypt		= ctr_crypt,
+		.decrypt		= ctr_crypt,
+	},
+};
+
+static bool is_blacklisted_cpu(void)
+{
+	if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
+		return false;
+
+	if (boot_cpu_data.x86 == 0x06 &&
+		(boot_cpu_data.x86_model == 0x1c ||
+		 boot_cpu_data.x86_model == 0x26 ||
+		 boot_cpu_data.x86_model == 0x36)) {
+		/*
+		 * On Atom, twofish-3way is slower than original assembler
+		 * implementation. Twofish-3way trades off some performance in
+		 * storing blocks in 64bit registers to allow three blocks to
+		 * be processed parallel. Parallel operation then allows gaining
+		 * more performance than was trade off, on out-of-order CPUs.
+		 * However Atom does not benefit from this parallellism and
+		 * should be blacklisted.
+		 */
+		return true;
+	}
+
+	if (boot_cpu_data.x86 == 0x0f) {
+		/*
+		 * On Pentium 4, twofish-3way is slower than original assembler
+		 * implementation because excessive uses of 64bit rotate and
+		 * left-shifts (which are really slow on P4) needed to store and
+		 * handle 128bit block in two 64bit registers.
+		 */
+		return true;
+	}
+
+	return false;
+}
+
+static int force;
+module_param(force, int, 0);
+MODULE_PARM_DESC(force, "Force module load, ignore CPU blacklist");
+
+static int __init init(void)
+{
+	if (!force && is_blacklisted_cpu()) {
+		printk(KERN_INFO
+			"twofish-x86_64-3way: performance on this CPU "
+			"would be suboptimal: disabling "
+			"twofish-x86_64-3way.\n");
+		return -ENODEV;
+	}
+
+	return crypto_register_skciphers(tf_skciphers,
+					 ARRAY_SIZE(tf_skciphers));
+}
+
+static void __exit fini(void)
+{
+	crypto_unregister_skciphers(tf_skciphers, ARRAY_SIZE(tf_skciphers));
+}
+
+module_init(init);
+module_exit(fini);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("Twofish Cipher Algorithm, 3-way parallel asm optimized");
+MODULE_ALIAS_CRYPTO("twofish");
+MODULE_ALIAS_CRYPTO("twofish-asm");