openssl-1_1/openssl-1_1-Fix-AES-GCM-on-Power-8-CPUs.patch

209 lines
7.9 KiB
Diff

From 9ab6b64ac856157a31a54c0d12207c2338bfa8e2 Mon Sep 17 00:00:00 2001
From: Tomas Mraz <tomas@openssl.org>
Date: Fri, 9 Sep 2022 14:46:24 +0200
Subject: [PATCH] Fix AES-GCM on Power 8 CPUs
Properly fallback to the default implementation on CPUs
missing necessary instructions.
Fixes #19163
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19182)
---
crypto/evp/e_aes.c | 146 ++++++++++++++++++++++++++---------------------------
1 file changed, 74 insertions(+), 72 deletions(-)
--- a/crypto/evp/e_aes.c
+++ b/crypto/evp/e_aes.c
@@ -181,30 +181,16 @@ static void ctr64_inc(unsigned char *cou
# define PPC_AES_GCM_CAPABLE (OPENSSL_ppccap_P & PPC_MADD300)
# define AES_GCM_ENC_BYTES 128
# define AES_GCM_DEC_BYTES 128
-# if PPC_AES_GCM_CAPABLE
size_t ppc_aes_gcm_encrypt(const unsigned char *in, unsigned char *out,
size_t len, const void *key, unsigned char ivec[16],
u64 *Xi);
size_t ppc_aes_gcm_decrypt(const unsigned char *in, unsigned char *out,
size_t len, const void *key, unsigned char ivec[16],
u64 *Xi);
-size_t ppc_aes_gcm_encrypt_wrap(const unsigned char *in, unsigned char *out,
- size_t len, const void *key,
- unsigned char ivec[16], u64 *Xi);
-size_t ppc_aes_gcm_decrypt_wrap(const unsigned char *in, unsigned char *out,
- size_t len, const void *key,
- unsigned char ivec[16], u64 *Xi);
-# define AES_gcm_encrypt ppc_aes_gcm_encrypt_wrap
-# define AES_gcm_decrypt ppc_aes_gcm_decrypt_wrap
-# define AES_GCM_ASM(gctx) ((gctx)->ctr==aes_p8_ctr32_encrypt_blocks && \
- (gctx)->gcm.ghash==gcm_ghash_p8)
+# define AES_GCM_ASM_PPC(gctx) ((gctx)->ctr==aes_p8_ctr32_encrypt_blocks && \
+ (gctx)->gcm.ghash==gcm_ghash_p8)
void gcm_ghash_p8(u64 Xi[2],const u128 Htable[16],const u8 *inp, size_t len);
-extern size_t ppc_aes_gcm_encrypt(const unsigned char *in, unsigned char *out, size_t len,
- const void *key, unsigned char ivec[16], u64 *Xi);
-extern size_t ppc_aes_gcm_decrypt(const unsigned char *in, unsigned char *out, size_t len,
- const void *key, unsigned char ivec[16], u64 *Xi);
-
static inline u32 UTO32(unsigned char *buf)
{
return ((u32) buf[0] << 24) | ((u32) buf[1] << 16) | ((u32) buf[2] << 8) | ((u32) buf[3]);
@@ -223,62 +209,6 @@ static inline u32 add32TOU(unsigned char
return r;
}
-static size_t aes_p10_gcm_crypt(const unsigned char *in, unsigned char *out, size_t len,
- const void *key, unsigned char ivec[16], u64 *Xi, int encrypt)
-{
- int s = 0;
- int ndone = 0;
- int ctr_reset = 0;
- u64 blocks_unused;
- u64 nb = len / 16;
- u64 next_ctr = 0;
- unsigned char ctr_saved[12];
-
- memcpy(ctr_saved, ivec, 12);
-
- while (nb) {
- blocks_unused = (u64) 0xffffffffU + 1 - (u64) UTO32 (ivec + 12);
- if (nb > blocks_unused) {
- len = blocks_unused * 16;
- nb -= blocks_unused;
- next_ctr = blocks_unused;
- ctr_reset = 1;
- } else {
- len = nb * 16;
- next_ctr = nb;
- nb = 0;
- }
-
- s = encrypt ? ppc_aes_gcm_encrypt(in, out, len, key, ivec, Xi)
- : ppc_aes_gcm_decrypt(in, out, len, key, ivec, Xi);
-
- /* add counter to ivec */
- add32TOU(ivec + 12, (u32) next_ctr);
- if (ctr_reset) {
- ctr_reset = 0;
- in += len;
- out += len;
- }
- memcpy(ivec, ctr_saved, 12);
- ndone += s;
- }
-
- return ndone;
-}
-
-size_t ppc_aes_gcm_encrypt_wrap(const unsigned char *in, unsigned char *out, size_t len,
- const void *key, unsigned char ivec[16], u64 *Xi)
-{
- return aes_p10_gcm_crypt(in, out, len, key, ivec, Xi, 1);
-}
-
-size_t ppc_aes_gcm_decrypt_wrap(const unsigned char *in, unsigned char *out, size_t len,
- const void *key, unsigned char ivec[16], u64 *Xi)
-{
- return aes_p10_gcm_crypt(in, out, len, key, ivec, Xi, 0);
-}
-
-# endif
#endif
#if defined(OPENSSL_CPUID_OBJ) && ( \
@@ -3294,6 +3224,51 @@ static int aes_gcm_tls_cipher(EVP_CIPHER
return rv;
}
+#if defined(OPENSSL_CPUID_OBJ) && (defined(__powerpc__) || defined(__ppc__) || defined(_ARCH_PPC))
+static size_t ppc_aes_gcm_crypt(const unsigned char *in, unsigned char *out, size_t len,
+ const void *key, unsigned char ivec[16], u64 *Xi, int encrypt)
+{
+ int s = 0;
+ int ndone = 0;
+ int ctr_reset = 0;
+ u64 blocks_unused;
+ u64 nb = len / 16;
+ u64 next_ctr = 0;
+ unsigned char ctr_saved[12];
+
+ memcpy(ctr_saved, ivec, 12);
+
+ while (nb) {
+ blocks_unused = (u64) 0xffffffffU + 1 - (u64) UTO32 (ivec + 12);
+ if (nb > blocks_unused) {
+ len = blocks_unused * 16;
+ nb -= blocks_unused;
+ next_ctr = blocks_unused;
+ ctr_reset = 1;
+ } else {
+ len = nb * 16;
+ next_ctr = nb;
+ nb = 0;
+ }
+
+ s = encrypt ? ppc_aes_gcm_encrypt(in, out, len, key, ivec, Xi)
+ : ppc_aes_gcm_decrypt(in, out, len, key, ivec, Xi);
+
+ /* add counter to ivec */
+ add32TOU(ivec + 12, (u32) next_ctr);
+ if (ctr_reset) {
+ ctr_reset = 0;
+ in += len;
+ out += len;
+ }
+ memcpy(ivec, ctr_saved, 12);
+ ndone += s;
+ }
+
+ return ndone;
+}
+#endif
+
static int aes_gcm_cipher(EVP_CIPHER_CTX *ctx, unsigned char *out,
const unsigned char *in, size_t len)
{
@@ -3325,6 +3300,20 @@ static int aes_gcm_cipher(EVP_CIPHER_CTX
out + res, len - res,
gctx->gcm.key, gctx->gcm.Yi.c,
gctx->gcm.Xi.u);
+
+ gctx->gcm.len.u[1] += bulk;
+ bulk += res;
+ }
+#elif defined(AES_GCM_ASM_PPC) && defined(OPENSSL_CPUID_OBJ) && (defined(__powerpc__) || defined(__ppc__) || defined(_ARCH_PPC))
+ if (PPC_AES_GCM_CAPABLE && len >= AES_GCM_ENC_BYTES && AES_GCM_ASM_PPC(gctx)) {
+ size_t res = (16 - gctx->gcm.mres) % 16;
+
+ if (CRYPTO_gcm128_encrypt(&gctx->gcm, in, out, res))
+ return -1;
+
+ bulk = ppc_aes_gcm_crypt(in + res, out + res, len - res,
+ gctx->gcm.key,
+ gctx->gcm.Yi.c, gctx->gcm.Xi.u, 1);
gctx->gcm.len.u[1] += bulk;
bulk += res;
}
@@ -3372,6 +3361,19 @@ static int aes_gcm_cipher(EVP_CIPHER_CTX
gctx->gcm.len.u[1] += bulk;
bulk += res;
}
+#elif defined(AES_GCM_ASM_PPC) && defined(OPENSSL_CPUID_OBJ) && (defined(__powerpc__) || defined(__ppc__) || defined(_ARCH_PPC))
+ if (PPC_AES_GCM_CAPABLE && len >= AES_GCM_DEC_BYTES && AES_GCM_ASM_PPC(gctx)) {
+ size_t res = (16 - gctx->gcm.mres) % 16;
+
+ if (CRYPTO_gcm128_decrypt(&gctx->gcm, in, out, res))
+ return -1;
+
+ bulk = ppc_aes_gcm_crypt(in + res, out + res, len - res,
+ gctx->gcm.key,
+ gctx->gcm.Yi.c, gctx->gcm.Xi.u, 0);
+ gctx->gcm.len.u[1] += bulk;
+ bulk += res;
+ }
#endif
if (CRYPTO_gcm128_decrypt_ctr32(&gctx->gcm,
in + bulk,