diff --git a/.travis.yml b/.travis.yml
index 1062d99..cdb79d1 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -80,6 +80,58 @@
         - sleep 5
         - scripts/windows_msbuild.bat v141 # Visual Studio 2017
 
+    - name: full configuration on arm64
+      os: linux
+      dist: focal
+      arch: arm64
+      addons:
+        apt:
+          packages:
+          - gcc
+      script:
+        # Do a manual build+test sequence rather than using all.sh, because
+        # there's no all.sh component that does what we want. We should set
+        # CFLAGS for arm64 host CC.
+        - scripts/config.py full
+        - scripts/config.py unset MBEDTLS_SHA512_USE_A64_CRYPTO_IF_PRESENT
+        - scripts/config.py unset MBEDTLS_SHA512_USE_A64_CRYPTO_ONLY
+        - scripts/config.py unset MBEDTLS_SHA256_USE_A64_CRYPTO_IF_PRESENT
+        - scripts/config.py unset MBEDTLS_SHA256_USE_A64_CRYPTO_ONLY
+        - make generated_files
+        - make CFLAGS='-march=armv8-a+crypto -O3 -Werror -fsanitize=address,undefined -fno-sanitize-recover=all' LDFLAGS='-Werror -fsanitize=address,undefined -fno-sanitize-recover=all'
+        - make test
+        - programs/test/selftest
+        - tests/scripts/test_psa_constant_names.py
+        # Modern OpenSSL does not support fixed ECDH or null ciphers.
+        - tests/compat.sh -p OpenSSL -e 'NULL\|ECDH_'
+        - tests/scripts/travis-log-failure.sh
+        - tests/context-info.sh
+
+    - name: full configuration(GnuTLS compat tests) on arm64
+      os: linux
+      dist: focal
+      arch: arm64
+      addons:
+        apt:
+          packages:
+          - clang
+          - gnutls-bin
+      script:
+        # Do a manual build+test sequence rather than using all.sh, because
+        # there's no all.sh component that does what we want. We should set
+        # CFLAGS for arm64 host CC.
+        - scripts/config.py full
+        - scripts/config.py unset MBEDTLS_SHA512_USE_A64_CRYPTO_IF_PRESENT
+        - scripts/config.py unset MBEDTLS_SHA512_USE_A64_CRYPTO_ONLY
+        - scripts/config.py unset MBEDTLS_SHA256_USE_A64_CRYPTO_IF_PRESENT
+        - scripts/config.py unset MBEDTLS_SHA256_USE_A64_CRYPTO_ONLY
+        - make generated_files
+        - make CC=clang CFLAGS='-march=armv8-a+crypto -O3 -Werror -fsanitize=address,undefined -fno-sanitize-recover=all' LDFLAGS='-Werror -fsanitize=address,undefined -fno-sanitize-recover=all'
+        # GnuTLS supports CAMELLIA but compat.sh doesn't properly enable it.
+        - tests/compat.sh -p GnuTLS -e 'CAMELLIA'
+        - tests/scripts/travis-log-failure.sh
+        - tests/context-info.sh
+
 after_failure:
 - tests/scripts/travis-log-failure.sh
 
diff --git a/include/mbedtls/check_config.h b/include/mbedtls/check_config.h
index 1efabdc..ac374d2 100644
--- a/include/mbedtls/check_config.h
+++ b/include/mbedtls/check_config.h
@@ -70,6 +70,10 @@
 #error "MBEDTLS_AESNI_C defined, but not all prerequisites"
 #endif
 
+#if defined(MBEDTLS_AESCE_C) && !defined(MBEDTLS_HAVE_ASM)
+#error "MBEDTLS_AESCE_C defined, but not all prerequisites"
+#endif
+
 #if defined(MBEDTLS_CTR_DRBG_C) && !defined(MBEDTLS_AES_C)
 #error "MBEDTLS_CTR_DRBG_C defined, but not all prerequisites"
 #endif
diff --git a/include/mbedtls/mbedtls_config.h b/include/mbedtls/mbedtls_config.h
index b874995..5aff9c5 100644
--- a/include/mbedtls/mbedtls_config.h
+++ b/include/mbedtls/mbedtls_config.h
@@ -2066,6 +2066,34 @@
 #define MBEDTLS_AESNI_C
 
 /**
+ * \def MBEDTLS_AESCE_C
+ *
+ * Enable AES crypto extension support on Arm64.
+ *
+ * Module:  library/aesce.c
+ * Caller:  library/aes.c
+ *
+ * Requires: MBEDTLS_HAVE_ASM, MBEDTLS_AES_C
+ *
+ * \note The code uses Neon intrinsics, so \c CFLAGS must be set to a minimum
+ * of \c -march=armv8-a+crypto .
+ *
+ * \warning If the target architecture is set to something that includes the
+ *          SHA3 feature (e.g. `-march=armv8.2-a+sha3`), for example because
+ *          `MBEDTLS_SHA512_USE_A64_CRYPTO_IF_PRESENT` is desired, compilers
+ *          generate code for `MBEDTLS_AESCE_C` that includes instructions
+ *          only present with the (optional) SHA3 feature. This will lead to an
+ *          undefined instruction exception if the code is run on a CPU without
+ *          that feature.
+ *
+ * \warning Runtime detection only works on linux. For non-linux operation
+ *          system, crypto extension MUST be supported by CPU.
+ *
+ * This module adds support for the AES crypto instructions on Arm64
+ */
+#define MBEDTLS_AESCE_C
+
+/**
  * \def MBEDTLS_AES_C
  *
  * Enable the AES block cipher.
diff --git a/library/CMakeLists.txt b/library/CMakeLists.txt
index c9714bb..bef2e1c 100644
--- a/library/CMakeLists.txt
+++ b/library/CMakeLists.txt
@@ -13,6 +13,7 @@
 set(src_crypto
     aes.c
     aesni.c
+    aesce.c
     aria.c
     asn1parse.c
     asn1write.c
diff --git a/library/Makefile b/library/Makefile
index dd16d06..ed5e1e1 100644
--- a/library/Makefile
+++ b/library/Makefile
@@ -78,6 +78,7 @@
 OBJS_CRYPTO= \
 	     aes.o \
 	     aesni.o \
+	     aesce.o \
 	     aria.o \
 	     asn1parse.o \
 	     asn1write.o \
diff --git a/library/aes.c b/library/aes.c
index 566e747..64392fc 100644
--- a/library/aes.c
+++ b/library/aes.c
@@ -39,6 +39,9 @@
 #if defined(MBEDTLS_AESNI_C)
 #include "aesni.h"
 #endif
+#if defined(MBEDTLS_AESCE_C)
+#include "aesce.h"
+#endif
 
 #include "mbedtls/platform.h"
 
@@ -544,6 +547,12 @@
     }
 #endif
 
+#if defined(MBEDTLS_AESCE_C) && defined(MBEDTLS_HAVE_ARM64)
+    if (mbedtls_aesce_has_support()) {
+        return mbedtls_aesce_setkey_enc((unsigned char *) RK, key, keybits);
+    }
+#endif
+
     for (i = 0; i < (keybits >> 5); i++) {
         RK[i] = MBEDTLS_GET_UINT32_LE(key, i << 2);
     }
@@ -652,6 +661,16 @@
     }
 #endif
 
+#if defined(MBEDTLS_AESCE_C) && defined(MBEDTLS_HAVE_ARM64)
+    if (mbedtls_aesce_has_support()) {
+        mbedtls_aesce_inverse_key(
+            (unsigned char *) RK,
+            (const unsigned char *) (cty.buf + cty.rk_offset),
+            ctx->nr);
+        goto exit;
+    }
+#endif
+
     SK = cty.buf + cty.rk_offset + cty.nr * 4;
 
     *RK++ = *SK++;
@@ -944,6 +963,12 @@
     }
 #endif
 
+#if defined(MBEDTLS_AESCE_C) && defined(MBEDTLS_HAVE_ARM64)
+    if (mbedtls_aesce_has_support()) {
+        return mbedtls_aesce_crypt_ecb(ctx, mode, input, output);
+    }
+#endif
+
 #if defined(MBEDTLS_PADLOCK_C) && defined(MBEDTLS_HAVE_X86)
     if (aes_padlock_ace > 0) {
         if (mbedtls_padlock_xcryptecb(ctx, mode, input, output) == 0) {
diff --git a/library/aesce.c b/library/aesce.c
new file mode 100644
index 0000000..ee0c8e1
--- /dev/null
+++ b/library/aesce.c
@@ -0,0 +1,257 @@
+/*
+ *  Arm64 crypto extension support functions
+ *
+ *  Copyright The Mbed TLS Contributors
+ *  SPDX-License-Identifier: Apache-2.0
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License"); you may
+ *  not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *  http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
+ *  WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#include <string.h>
+#include "common.h"
+
+#if defined(MBEDTLS_AESCE_C)
+
+#include "aesce.h"
+
+#if defined(MBEDTLS_HAVE_ARM64)
+
+#if defined(__clang__)
+#   if __clang_major__ < 4
+#       error "A more recent Clang is required for MBEDTLS_AESCE_C"
+#   endif
+#elif defined(__GNUC__)
+#   if __GNUC__ < 6
+#       error "A more recent GCC is required for MBEDTLS_AESCE_C"
+#   endif
+#else
+#    error "Only GCC and Clang supported for MBEDTLS_AESCE_C"
+#endif
+
+#if !defined(__ARM_FEATURE_CRYPTO)
+#   error "`crypto` feature moddifier MUST be enabled for MBEDTLS_AESCE_C."
+#   error "Typical option for GCC and Clang is `-march=armv8-a+crypto`."
+#endif /* !__ARM_FEATURE_CRYPTO */
+
+#include <arm_neon.h>
+
+#if defined(__linux__)
+#include <asm/hwcap.h>
+#include <sys/auxv.h>
+#endif
+
+/*
+ * AES instruction support detection routine
+ */
+int mbedtls_aesce_has_support(void)
+{
+#if defined(__linux__)
+    unsigned long auxval = getauxval(AT_HWCAP);
+    return (auxval & (HWCAP_ASIMD | HWCAP_AES)) ==
+           (HWCAP_ASIMD | HWCAP_AES);
+#else
+    /* Assume AES instructions are supported. */
+    return 1;
+#endif
+}
+
+static uint8x16_t aesce_encrypt_block(uint8x16_t block,
+                                      unsigned char *keys,
+                                      int rounds)
+{
+    for (int i = 0; i < rounds - 1; i++) {
+        /* AES AddRoundKey, SubBytes, ShiftRows (in this order).
+         * AddRoundKey adds the round key for the previous round. */
+        block = vaeseq_u8(block, vld1q_u8(keys + i * 16));
+        /* AES mix columns */
+        block = vaesmcq_u8(block);
+    }
+
+    /* AES AddRoundKey for the previous round.
+     * SubBytes, ShiftRows for the final round.  */
+    block = vaeseq_u8(block, vld1q_u8(keys + (rounds -1) * 16));
+
+    /* Final round: no MixColumns */
+
+    /* Final AddRoundKey */
+    block = veorq_u8(block, vld1q_u8(keys + rounds  * 16));
+
+    return block;
+}
+
+static uint8x16_t aesce_decrypt_block(uint8x16_t block,
+                                      unsigned char *keys,
+                                      int rounds)
+{
+
+    for (int i = 0; i < rounds - 1; i++) {
+        /* AES AddRoundKey, SubBytes, ShiftRows */
+        block = vaesdq_u8(block, vld1q_u8(keys + i * 16));
+        /* AES inverse MixColumns for the next round.
+         *
+         * This means that we switch the order of the inverse AddRoundKey and
+         * inverse MixColumns operations. We have to do this as AddRoundKey is
+         * done in an atomic instruction together with the inverses of SubBytes
+         * and ShiftRows.
+         *
+         * It works because MixColumns is a linear operation over GF(2^8) and
+         * AddRoundKey is an exclusive or, which is equivalent to addition over
+         * GF(2^8). (The inverse of MixColumns needs to be applied to the
+         * affected round keys separately which has been done when the
+         * decryption round keys were calculated.) */
+        block = vaesimcq_u8(block);
+    }
+
+    /* The inverses of AES AddRoundKey, SubBytes, ShiftRows finishing up the
+     * last full round. */
+    block = vaesdq_u8(block, vld1q_u8(keys + (rounds - 1) * 16));
+
+    /* Inverse AddRoundKey for inverting the initial round key addition. */
+    block = veorq_u8(block, vld1q_u8(keys + rounds * 16));
+
+    return block;
+}
+
+/*
+ * AES-ECB block en(de)cryption
+ */
+int mbedtls_aesce_crypt_ecb(mbedtls_aes_context *ctx,
+                            int mode,
+                            const unsigned char input[16],
+                            unsigned char output[16])
+{
+    uint8x16_t block = vld1q_u8(&input[0]);
+    unsigned char *keys = (unsigned char *) (ctx->buf + ctx->rk_offset);
+
+    if (mode == MBEDTLS_AES_ENCRYPT) {
+        block = aesce_encrypt_block(block, keys, ctx->nr);
+    } else {
+        block = aesce_decrypt_block(block, keys, ctx->nr);
+    }
+    vst1q_u8(&output[0], block);
+
+    return 0;
+}
+
+/*
+ * Compute decryption round keys from encryption round keys
+ */
+void mbedtls_aesce_inverse_key(unsigned char *invkey,
+                               const unsigned char *fwdkey,
+                               int nr)
+{
+    int i, j;
+    j = nr;
+    vst1q_u8(invkey, vld1q_u8(fwdkey + j * 16));
+    for (i = 1, j--; j > 0; i++, j--) {
+        vst1q_u8(invkey + i * 16,
+                 vaesimcq_u8(vld1q_u8(fwdkey + j * 16)));
+    }
+    vst1q_u8(invkey + i * 16, vld1q_u8(fwdkey + j * 16));
+
+}
+
+static inline uint32_t aes_rot_word(uint32_t word)
+{
+    return (word << (32 - 8)) | (word >> 8);
+}
+
+static inline uint32_t aes_sub_word(uint32_t in)
+{
+    uint8x16_t v = vreinterpretq_u8_u32(vdupq_n_u32(in));
+    uint8x16_t zero = vdupq_n_u8(0);
+
+    /* vaeseq_u8 does both SubBytes and ShiftRows. Taking the first row yields
+     * the correct result as ShiftRows doesn't change the first row. */
+    v = vaeseq_u8(zero, v);
+    return vgetq_lane_u32(vreinterpretq_u32_u8(v), 0);
+}
+
+/*
+ * Key expansion function
+ */
+static void aesce_setkey_enc(unsigned char *rk,
+                             const unsigned char *key,
+                             const size_t key_bit_length)
+{
+    static uint8_t const rcon[] = { 0x01, 0x02, 0x04, 0x08, 0x10,
+                                    0x20, 0x40, 0x80, 0x1b, 0x36 };
+    /* See https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.197.pdf
+     *   - Section 5, Nr = Nk + 6
+     *   - Section 5.2, the key expansion size is Nb*(Nr+1)
+     */
+    const uint32_t key_len_in_words = key_bit_length / 32;  /* Nk */
+    const size_t round_key_len_in_words = 4;                /* Nb */
+    const size_t round_keys_needed = key_len_in_words + 6;  /* Nr */
+    const size_t key_expansion_size_in_words =
+        round_key_len_in_words * (round_keys_needed + 1);   /* Nb*(Nr+1) */
+    const uint32_t *rko_end = (uint32_t *) rk + key_expansion_size_in_words;
+
+    memcpy(rk, key, key_len_in_words * 4);
+
+    for (uint32_t *rki = (uint32_t *) rk;
+         rki + key_len_in_words < rko_end;
+         rki += key_len_in_words) {
+
+        size_t iteration = (rki - (uint32_t *) rk) / key_len_in_words;
+        uint32_t *rko;
+        rko = rki + key_len_in_words;
+        rko[0] = aes_rot_word(aes_sub_word(rki[key_len_in_words - 1]));
+        rko[0] ^= rcon[iteration] ^ rki[0];
+        rko[1] = rko[0] ^ rki[1];
+        rko[2] = rko[1] ^ rki[2];
+        rko[3] = rko[2] ^ rki[3];
+        if (rko + key_len_in_words > rko_end) {
+            /* Do not write overflow words.*/
+            continue;
+        }
+        switch (key_bit_length) {
+            case 128:
+                break;
+            case 192:
+                rko[4] = rko[3] ^ rki[4];
+                rko[5] = rko[4] ^ rki[5];
+                break;
+            case 256:
+                rko[4] = aes_sub_word(rko[3]) ^ rki[4];
+                rko[5] = rko[4] ^ rki[5];
+                rko[6] = rko[5] ^ rki[6];
+                rko[7] = rko[6] ^ rki[7];
+                break;
+        }
+    }
+}
+
+/*
+ * Key expansion, wrapper
+ */
+int mbedtls_aesce_setkey_enc(unsigned char *rk,
+                             const unsigned char *key,
+                             size_t bits)
+{
+    switch (bits) {
+        case 128:
+        case 192:
+        case 256:
+            aesce_setkey_enc(rk, key, bits);
+            break;
+        default:
+            return MBEDTLS_ERR_AES_INVALID_KEY_LENGTH;
+    }
+
+    return 0;
+}
+
+#endif /* MBEDTLS_HAVE_ARM64 */
+
+#endif /* MBEDTLS_AESCE_C */
diff --git a/library/aesce.h b/library/aesce.h
new file mode 100644
index 0000000..da42446
--- /dev/null
+++ b/library/aesce.h
@@ -0,0 +1,98 @@
+/**
+ * \file aesce.h
+ *
+ * \brief AES-CE for hardware AES acceleration on ARMv8 processors with crypto
+ *        extension.
+ *
+ * \warning These functions are only for internal use by other library
+ *          functions; you must not call them directly.
+ */
+/*
+ *  Copyright The Mbed TLS Contributors
+ *  SPDX-License-Identifier: Apache-2.0
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License"); you may
+ *  not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *  http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
+ *  WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+#ifndef MBEDTLS_AESCE_H
+#define MBEDTLS_AESCE_H
+
+#include "mbedtls/build_info.h"
+
+#include "mbedtls/aes.h"
+
+
+#if defined(MBEDTLS_HAVE_ASM) && defined(__GNUC__) && \
+    defined(__aarch64__) && !defined(MBEDTLS_HAVE_ARM64)
+#define MBEDTLS_HAVE_ARM64
+#endif
+
+#if defined(MBEDTLS_HAVE_ARM64)
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * \brief          Internal function to detect the crypto extension in CPUs.
+ *
+ * \return         1 if CPU has support for the feature, 0 otherwise
+ */
+int mbedtls_aesce_has_support(void);
+
+/**
+ * \brief          Internal AES-ECB block encryption and decryption
+ *
+ * \param ctx      AES context
+ * \param mode     MBEDTLS_AES_ENCRYPT or MBEDTLS_AES_DECRYPT
+ * \param input    16-byte input block
+ * \param output   16-byte output block
+ *
+ * \return         0 on success (cannot fail)
+ */
+int mbedtls_aesce_crypt_ecb(mbedtls_aes_context *ctx,
+                            int mode,
+                            const unsigned char input[16],
+                            unsigned char output[16]);
+
+/**
+ * \brief           Internal round key inversion. This function computes
+ *                  decryption round keys from the encryption round keys.
+ *
+ * \param invkey    Round keys for the equivalent inverse cipher
+ * \param fwdkey    Original round keys (for encryption)
+ * \param nr        Number of rounds (that is, number of round keys minus one)
+ */
+void mbedtls_aesce_inverse_key(unsigned char *invkey,
+                               const unsigned char *fwdkey,
+                               int nr);
+
+/**
+ * \brief           Internal key expansion for encryption
+ *
+ * \param rk        Destination buffer where the round keys are written
+ * \param key       Encryption key
+ * \param bits      Key size in bits (must be 128, 192 or 256)
+ *
+ * \return          0 if successful, or MBEDTLS_ERR_AES_INVALID_KEY_LENGTH
+ */
+int mbedtls_aesce_setkey_enc(unsigned char *rk,
+                             const unsigned char *key,
+                             size_t bits);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* MBEDTLS_HAVE_ARM64 */
+
+#endif /* MBEDTLS_AESCE_H */
diff --git a/tests/scripts/all.sh b/tests/scripts/all.sh
index 7d91fa2..353ec69 100755
--- a/tests/scripts/all.sh
+++ b/tests/scripts/all.sh
@@ -1972,6 +1972,7 @@
     # aesni.c and padlock.c reference mbedtls_aes_context fields directly.
     scripts/config.py unset MBEDTLS_AESNI_C
     scripts/config.py unset MBEDTLS_PADLOCK_C
+    scripts/config.py unset MBEDTLS_AESCE_C
     # MBEDTLS_ECP_RESTARTABLE is documented as incompatible.
     scripts/config.py unset MBEDTLS_ECP_RESTARTABLE
     # You can only have one threading implementation: alt or pthread, not both.
@@ -3341,6 +3342,7 @@
     scripts/config.py unset MBEDTLS_HAVE_ASM
     scripts/config.py unset MBEDTLS_AESNI_C
     scripts/config.py unset MBEDTLS_PADLOCK_C
+    scripts/config.py unset MBEDTLS_AESCE_C
     make CC=gcc CFLAGS='-Werror -Wall -Wextra -DMBEDTLS_HAVE_INT32'
 
     msg "test: gcc, force 32-bit bignum limbs"
@@ -3352,6 +3354,7 @@
     scripts/config.py unset MBEDTLS_HAVE_ASM
     scripts/config.py unset MBEDTLS_AESNI_C
     scripts/config.py unset MBEDTLS_PADLOCK_C
+    scripts/config.py unset MBEDTLS_AESCE_C
     make CC=gcc CFLAGS='-Werror -Wall -Wextra -DMBEDTLS_HAVE_INT64'
 
     msg "test: gcc, force 64-bit bignum limbs"