[crypto] Replace AES implementation

Replace the AES implementation from AXTLS with a dedicated iPXE
implementation which is slightly smaller and around 1000% faster.
This implementation has been verified using the existing self-tests
based on the NIST AES test vectors.

Signed-off-by: Michael Brown <mcb30@ipxe.org>

5 files changed, 836 insertions, 635 deletions

diff --git a/src/crypto/aes.c b/src/crypto/aes.c
new file mode 100644
index 00000000..1605d2ee
--- /dev/null
+++ b/src/crypto/aes.c
@@ -0,0 +1,804 @@
+/*
+ * Copyright (C) 2015 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * AES algorithm
+ *
+ */
+
+#include <stdint.h>
+#include <string.h>
+#include <errno.h>
+#include <assert.h>
+#include <byteswap.h>
+#include <ipxe/rotate.h>
+#include <ipxe/crypto.h>
+#include <ipxe/ecb.h>
+#include <ipxe/cbc.h>
+#include <ipxe/aes.h>
+
+/** AES strides
+ *
+ * These are the strides (modulo 16) used to walk through the AES
+ * input state bytes in order of byte position after [Inv]ShiftRows.
+ */
+enum aes_stride {
+	/** Input stride for ShiftRows
+	 *
+	 *    0 4 8 c
+	 *     \ \ \
+	 *    1 5 9 d
+	 *     \ \ \
+	 *    2 6 a e
+	 *     \ \ \
+	 *    3 7 b f
+	 */
+	AES_STRIDE_SHIFTROWS = +5,
+	/** Input stride for InvShiftRows
+	 *
+	 *    0 4 8 c
+	 *     / / /
+	 *    1 5 9 d
+	 *     / / /
+	 *    2 6 a e
+	 *     / / /
+	 *    3 7 b f
+	 */
+	AES_STRIDE_INVSHIFTROWS = -3,
+};
+
+/** A single AES lookup table entry
+ *
+ * This represents the product (in the Galois field GF(2^8)) of an
+ * eight-byte vector multiplier with a single scalar multiplicand.
+ *
+ * The vector multipliers used for AES will be {1,1,1,3,2,1,1,3} for
+ * MixColumns and {1,9,13,11,14,9,13,11} for InvMixColumns.  This
+ * allows for the result of multiplying any single column of the
+ * [Inv]MixColumns matrix by a scalar value to be obtained simply by
+ * extracting the relevant four-byte subset from the lookup table
+ * entry.
+ *
+ * For example, to find the result of multiplying the second column of
+ * the MixColumns matrix by the scalar value 0x80:
+ *
+ * MixColumns column[0]: {			      2,    1,    1,    3 }
+ * MixColumns column[1]: {			3,    2,    1,    1	  }
+ * MixColumns column[2]: {		  1,    3,    2,    1		  }
+ * MixColumns column[3]: {	    1,    1,    3,    2			  }
+ * Vector multiplier:	 {    1,    1,    1,    3,    2,    1,    1,    3 }
+ * Scalar multiplicand:	   0x80
+ * Lookup table entry:	 { 0x80, 0x80, 0x80, 0x9b, 0x1b, 0x80, 0x80, 0x9b }
+ *
+ * The second column of the MixColumns matrix is {3,2,1,1}.  The
+ * product of this column with the scalar value 0x80 can be obtained
+ * by extracting the relevant four-byte subset of the lookup table
+ * entry:
+ *
+ * MixColumns column[1]: {			3,    2,    1,    1	  }
+ * Vector multiplier:	 {    1,    1,    1,    3,    2,    1,    1,    3 }
+ * Lookup table entry:	 { 0x80, 0x80, 0x80, 0x9b, 0x1b, 0x80, 0x80, 0x9b }
+ * Product:		 {		     0x9b, 0x1b, 0x80, 0x80	  }
+ *
+ * The column lookups require only seven bytes of the eight-byte
+ * entry: the remaining (first) byte is used to hold the scalar
+ * multiplicand itself (i.e. the first byte of the vector multiplier
+ * is always chosen to be 1).
+ */
+union aes_table_entry {
+	/** Viewed as an array of bytes */
+	uint8_t byte[8];
+} __attribute__ (( packed ));
+
+/** An AES lookup table
+ *
+ * This represents the products (in the Galois field GF(2^8)) of a
+ * constant eight-byte vector multiplier with all possible 256 scalar
+ * multiplicands.
+ *
+ * The entries are indexed by the AES [Inv]SubBytes S-box output
+ * values (denoted S(N)).  This allows for the result of multiplying
+ * any single column of the [Inv]MixColumns matrix by S(N) to be
+ * obtained simply by extracting the relevant four-byte subset from
+ * the Nth table entry.  For example:
+ *
+ * Input byte (N):	   0x3a
+ * SubBytes output S(N):   0x80
+ * MixColumns column[1]: {			3,    2,    1,    1	  }
+ * Vector multiplier:	 {    1,    1,    1,    3,    2,    1,    1,    3 }
+ * Table entry[0x3a]:	 { 0x80, 0x80, 0x80, 0x9b, 0x1b, 0x80, 0x80, 0x9b }
+ * Product:		 {		     0x9b, 0x1b, 0x80, 0x80	  }
+ *
+ * Since the first byte of the eight-byte vector multiplier is always
+ * chosen to be 1, the value of S(N) may be lookup up by extracting
+ * the first byte of the Nth table entry.
+ */
+struct aes_table {
+	/** Table entries, indexed by S(N) */
+	union aes_table_entry entry[256];
+} __attribute__ (( aligned ( 8 ) ));
+
+/** AES MixColumns lookup table */
+static struct aes_table aes_mixcolumns;
+
+/** AES InvMixColumns lookup table */
+static struct aes_table aes_invmixcolumns;
+
+/**
+ * Multiply [Inv]MixColumns matrix column by scalar multiplicand
+ *
+ * @v entry		AES lookup table entry for scalar multiplicand
+ * @v column		[Inv]MixColumns matrix column index
+ * @ret product		Product of matrix column with scalar multiplicand
+ */
+static inline __attribute__ (( always_inline )) uint32_t
+aes_entry_column ( const union aes_table_entry *entry, unsigned int column ) {
+	const uint8_t *first __attribute__ (( may_alias ));
+
+	/* Locate start of relevant four-byte subset */
+	first = &entry->byte[ 4 - column ];
+
+	/* Extract this four-byte subset */
+	return ( *( ( uint32_t * ) first ) );
+}
+
+/**
+ * Multiply [Inv]MixColumns matrix column by S-boxed input byte
+ *
+ * @v table		AES lookup table
+ * @v stride		AES row shift stride
+ * @v in		AES input state
+ * @v offset		Output byte offset (after [Inv]ShiftRows)
+ * @ret product		Product of matrix column with S(input byte)
+ *
+ * Note that the specified offset is not the offset of the input byte;
+ * it is the offset of the output byte which corresponds to the input
+ * byte.  This output byte offset is used to calculate both the input
+ * byte offset and to select the appropriate matric column.
+ *
+ * With a compile-time constant offset, this function will optimise
+ * down to a single "movzbl" (to extract the input byte) and will
+ * generate a single x86 memory reference expression which can then be
+ * used directly within a single "xorl" instruction.
+ */
+static inline __attribute__ (( always_inline )) uint32_t
+aes_column ( const struct aes_table *table, size_t stride,
+	     const union aes_matrix *in, size_t offset ) {
+	const union aes_table_entry *entry;
+	unsigned int byte;
+
+	/* Extract input byte corresponding to this output byte offset
+	 * (i.e. perform [Inv]ShiftRows).
+	 */
+	byte = in->byte[ ( stride * offset ) & 0xf ];
+
+	/* Locate lookup table entry for this input byte (i.e. perform
+	 * [Inv]SubBytes).
+	 */
+	entry = &table->entry[byte];
+
+	/* Multiply appropriate matrix column by this input byte
+	 * (i.e. perform [Inv]MixColumns).
+	 */
+	return aes_entry_column ( entry, ( offset & 0x3 ) );
+}
+
+/**
+ * Calculate intermediate round output column
+ *
+ * @v table		AES lookup table
+ * @v stride		AES row shift stride
+ * @v in		AES input state
+ * @v key		AES round key
+ * @v column		Column index
+ * @ret output		Output column value
+ */
+static inline __attribute__ (( always_inline )) uint32_t
+aes_output ( const struct aes_table *table, size_t stride,
+	     const union aes_matrix *in, const union aes_matrix *key,
+	     unsigned int column ) {
+	size_t offset = ( column * 4 );
+
+	/* Perform [Inv]ShiftRows, [Inv]SubBytes, [Inv]MixColumns, and
+	 * AddRoundKey for this column.  The loop is unrolled to allow
+	 * for the required compile-time constant optimisations.
+	 */
+	return ( aes_column ( table, stride, in, ( offset + 0 ) ) ^
+		 aes_column ( table, stride, in, ( offset + 1 ) ) ^
+		 aes_column ( table, stride, in, ( offset + 2 ) ) ^
+		 aes_column ( table, stride, in, ( offset + 3 ) ) ^
+		 key->column[column] );
+}
+
+/**
+ * Perform a single intermediate round
+ *
+ * @v table		AES lookup table
+ * @v stride		AES row shift stride
+ * @v in		AES input state
+ * @v out		AES output state
+ * @v key		AES round key
+ */
+static inline __attribute__ (( always_inline )) void
+aes_round ( const struct aes_table *table, size_t stride,
+	    const union aes_matrix *in, union aes_matrix *out,
+	    const union aes_matrix *key ) {
+
+	/* Perform [Inv]ShiftRows, [Inv]SubBytes, [Inv]MixColumns, and
+	 * AddRoundKey for all columns.  The loop is unrolled to allow
+	 * for the required compile-time constant optimisations.
+	 */
+	out->column[0] = aes_output ( table, stride, in, key, 0 );
+	out->column[1] = aes_output ( table, stride, in, key, 1 );
+	out->column[2] = aes_output ( table, stride, in, key, 2 );
+	out->column[3] = aes_output ( table, stride, in, key, 3 );
+}
+
+/**
+ * Perform encryption intermediate rounds
+ *
+ * @v in		AES input state
+ * @v out		AES output state
+ * @v key		Round keys
+ * @v rounds		Number of rounds (must be odd)
+ *
+ * This function is deliberately marked as non-inlinable to ensure
+ * maximal availability of registers for GCC's register allocator,
+ * which has a tendency to otherwise spill performance-critical
+ * registers to the stack.
+ */
+static __attribute__ (( noinline )) void
+aes_encrypt_rounds ( union aes_matrix *in, union aes_matrix *out,
+		     const union aes_matrix *key, unsigned int rounds ) {
+	union aes_matrix *tmp;
+
+	/* Perform intermediate rounds */
+	do {
+		/* Perform one intermediate round */
+		aes_round ( &aes_mixcolumns, AES_STRIDE_SHIFTROWS,
+			    in, out, key++ );
+
+		/* Swap input and output states for next round */
+		tmp = in;
+		in = out;
+		out = tmp;
+
+	} while ( --rounds );
+}
+
+/**
+ * Perform decryption intermediate rounds
+ *
+ * @v in		AES input state
+ * @v out		AES output state
+ * @v key		Round keys
+ * @v rounds		Number of rounds (must be odd)
+ *
+ * As with aes_encrypt_rounds(), this function is deliberately marked
+ * as non-inlinable.
+ *
+ * This function could potentially use the same binary code as is used
+ * for encryption.  To compensate for the difference between ShiftRows
+ * and InvShiftRows, half of the input byte offsets would have to be
+ * modifiable at runtime (half by an offset of +4/-4, half by an
+ * offset of -4/+4 for ShiftRows/InvShiftRows).  This can be
+ * accomplished in x86 assembly within the number of available
+ * registers, but GCC's register allocator struggles to do so,
+ * resulting in a significant performance decrease due to registers
+ * being spilled to the stack.  We therefore use two separate but very
+ * similar binary functions based on the same C source.
+ */
+static __attribute__ (( noinline )) void
+aes_decrypt_rounds ( union aes_matrix *in, union aes_matrix *out,
+		     const union aes_matrix *key, unsigned int rounds ) {
+	union aes_matrix *tmp;
+
+	/* Perform intermediate rounds */
+	do {
+		/* Perform one intermediate round */
+		aes_round ( &aes_invmixcolumns, AES_STRIDE_INVSHIFTROWS,
+			    in, out, key++ );
+
+		/* Swap input and output states for next round */
+		tmp = in;
+		in = out;
+		out = tmp;
+
+	} while ( --rounds );
+}
+
+/**
+ * Perform standalone AddRoundKey
+ *
+ * @v state		AES state
+ * @v key		AES round key
+ */
+static inline __attribute__ (( always_inline )) void
+aes_addroundkey ( union aes_matrix *state, const union aes_matrix *key ) {
+
+	state->column[0] ^= key->column[0];
+	state->column[1] ^= key->column[1];
+	state->column[2] ^= key->column[2];
+	state->column[3] ^= key->column[3];
+}
+
+/**
+ * Perform final round
+ *
+ * @v table		AES lookup table
+ * @v stride		AES row shift stride
+ * @v in		AES input state
+ * @v out		AES output state
+ * @v key		AES round key
+ */
+static void aes_final ( const struct aes_table *table, size_t stride,
+			const union aes_matrix *in, union aes_matrix *out,
+			const union aes_matrix *key ) {
+	const union aes_table_entry *entry;
+	unsigned int byte;
+	size_t out_offset;
+	size_t in_offset;
+
+	/* Perform [Inv]ShiftRows and [Inv]SubBytes */
+	for ( out_offset = 0, in_offset = 0 ; out_offset < 16 ;
+	      out_offset++, in_offset = ( ( in_offset + stride ) & 0xf ) ) {
+
+		/* Extract input byte (i.e. perform [Inv]ShiftRows) */
+		byte = in->byte[in_offset];
+
+		/* Locate lookup table entry for this input byte
+		 * (i.e. perform [Inv]SubBytes).
+		 */
+		entry = &table->entry[byte];
+
+		/* Store output byte */
+		out->byte[out_offset] = entry->byte[0];
+	}
+
+	/* Perform AddRoundKey */
+	aes_addroundkey ( out, key );
+}
+
+/**
+ * Encrypt data
+ *
+ * @v ctx		Context
+ * @v src		Data to encrypt
+ * @v dst		Buffer for encrypted data
+ * @v len		Length of data
+ */
+static void aes_encrypt ( void *ctx, const void *src, void *dst, size_t len ) {
+	struct aes_context *aes = ctx;
+	union aes_matrix buffer[2];
+	union aes_matrix *in = &buffer[0];
+	union aes_matrix *out = &buffer[1];
+	unsigned int rounds = aes->rounds;
+
+	/* Sanity check */
+	assert ( len == sizeof ( *in ) );
+
+	/* Initialise input state */
+	memcpy ( in, src, sizeof ( *in ) );
+
+	/* Perform initial round (AddRoundKey) */
+	aes_addroundkey ( in, &aes->encrypt.key[0] );
+
+	/* Perform intermediate rounds (ShiftRows, SubBytes,
+	 * MixColumns, AddRoundKey).
+	 */
+	aes_encrypt_rounds ( in, out, &aes->encrypt.key[1], ( rounds - 2 ) );
+	in = out;
+
+	/* Perform final round (ShiftRows, SubBytes, AddRoundKey) */
+	out = dst;
+	aes_final ( &aes_mixcolumns, AES_STRIDE_SHIFTROWS, in, out,
+		    &aes->encrypt.key[ rounds - 1 ] );
+}
+
+/**
+ * Decrypt data
+ *
+ * @v ctx		Context
+ * @v src		Data to decrypt
+ * @v dst		Buffer for decrypted data
+ * @v len		Length of data
+ */
+static void aes_decrypt ( void *ctx, const void *src, void *dst, size_t len ) {
+	struct aes_context *aes = ctx;
+	union aes_matrix buffer[2];
+	union aes_matrix *in = &buffer[0];
+	union aes_matrix *out = &buffer[1];
+	unsigned int rounds = aes->rounds;
+
+	/* Sanity check */
+	assert ( len == sizeof ( *in ) );
+
+	/* Initialise input state */
+	memcpy ( in, src, sizeof ( *in ) );
+
+	/* Perform initial round (AddRoundKey) */
+	aes_addroundkey ( in, &aes->decrypt.key[0] );
+
+	/* Perform intermediate rounds (InvShiftRows, InvSubBytes,
+	 * InvMixColumns, AddRoundKey).
+	 */
+	aes_decrypt_rounds ( in, out, &aes->decrypt.key[1], ( rounds - 2 ) );
+	in = out;
+
+	/* Perform final round (InvShiftRows, InvSubBytes, AddRoundKey) */
+	out = dst;
+	aes_final ( &aes_invmixcolumns, AES_STRIDE_INVSHIFTROWS, in, out,
+		    &aes->decrypt.key[ rounds - 1 ] );
+}
+
+/**
+ * Multiply a polynomial by (x) modulo (x^8 + x^4 + x^3 + x^2 + 1) in GF(2^8)
+ *
+ * @v poly		Polynomial to be multiplied
+ * @ret result		Result
+ */
+static __attribute__ (( const )) unsigned int aes_double ( unsigned int poly ) {
+
+	/* Multiply polynomial by (x), placing the resulting x^8
+	 * coefficient in the LSB (i.e. rotate byte left by one).
+	 */
+	poly = rol8 ( poly, 1 );
+
+	/* If coefficient of x^8 (in LSB) is non-zero, then reduce by
+	 * subtracting (x^8 + x^4 + x^3 + x^2 + 1) in GF(2^8).
+	 */
+	if ( poly & 0x01 ) {
+		poly ^= 0x01; /* Subtract x^8 (currently in LSB) */
+		poly ^= 0x1b; /* Subtract (x^4 + x^3 + x^2 + 1) */
+	}
+
+	return poly;
+}
+
+/**
+ * Fill in MixColumns lookup table entry
+ *
+ * @v entry		AES lookup table entry for scalar multiplicand
+ *
+ * The MixColumns lookup table vector multiplier is {1,1,1,3,2,1,1,3}.
+ */
+static void aes_mixcolumns_entry ( union aes_table_entry *entry ) {
+	unsigned int scalar_x_1;
+	unsigned int scalar_x;
+	unsigned int scalar;
+
+	/* Retrieve scalar multiplicand */
+	scalar = entry->byte[0];
+	entry->byte[1] = scalar;
+	entry->byte[2] = scalar;
+	entry->byte[5] = scalar;
+	entry->byte[6] = scalar;
+
+	/* Calculate scalar multiplied by (x) */
+	scalar_x = aes_double ( scalar );
+	entry->byte[4] = scalar_x;
+
+	/* Calculate scalar multiplied by (x + 1) */
+	scalar_x_1 = ( scalar_x ^ scalar );
+	entry->byte[3] = scalar_x_1;
+	entry->byte[7] = scalar_x_1;
+}
+
+/**
+ * Fill in InvMixColumns lookup table entry
+ *
+ * @v entry		AES lookup table entry for scalar multiplicand
+ *
+ * The InvMixColumns lookup table vector multiplier is {1,9,13,11,14,9,13,11}.
+ */
+static void aes_invmixcolumns_entry ( union aes_table_entry *entry ) {
+	unsigned int scalar_x3_x2_x;
+	unsigned int scalar_x3_x2_1;
+	unsigned int scalar_x3_x2;
+	unsigned int scalar_x3_x_1;
+	unsigned int scalar_x3_1;
+	unsigned int scalar_x3;
+	unsigned int scalar_x2;
+	unsigned int scalar_x;
+	unsigned int scalar;
+
+	/* Retrieve scalar multiplicand */
+	scalar = entry->byte[0];
+
+	/* Calculate scalar multiplied by (x) */
+	scalar_x = aes_double ( scalar );
+
+	/* Calculate scalar multiplied by (x^2) */
+	scalar_x2 = aes_double ( scalar_x );
+
+	/* Calculate scalar multiplied by (x^3) */
+	scalar_x3 = aes_double ( scalar_x2 );
+
+	/* Calculate scalar multiplied by (x^3 + 1) */
+	scalar_x3_1 = ( scalar_x3 ^ scalar );
+	entry->byte[1] = scalar_x3_1;
+	entry->byte[5] = scalar_x3_1;
+
+	/* Calculate scalar multiplied by (x^3 + x + 1) */
+	scalar_x3_x_1 = ( scalar_x3_1 ^ scalar_x );
+	entry->byte[3] = scalar_x3_x_1;
+	entry->byte[7] = scalar_x3_x_1;
+
+	/* Calculate scalar multiplied by (x^3 + x^2) */
+	scalar_x3_x2 = ( scalar_x3 ^ scalar_x2 );
+
+	/* Calculate scalar multiplied by (x^3 + x^2 + 1) */
+	scalar_x3_x2_1 = ( scalar_x3_x2 ^ scalar );
+	entry->byte[2] = scalar_x3_x2_1;
+	entry->byte[6] = scalar_x3_x2_1;
+
+	/* Calculate scalar multiplied by (x^3 + x^2 + x) */
+	scalar_x3_x2_x = ( scalar_x3_x2 ^ scalar_x );
+	entry->byte[4] = scalar_x3_x2_x;
+}
+
+/**
+ * Generate AES lookup tables
+ *
+ */
+static void aes_generate ( void ) {
+	union aes_table_entry *entry;
+	union aes_table_entry *inventry;
+	unsigned int poly = 0x01;
+	unsigned int invpoly = 0x01;
+	unsigned int transformed;
+	unsigned int i;
+
+	/* Iterate over non-zero values of GF(2^8) using generator (x + 1) */
+	do {
+
+		/* Multiply polynomial by (x + 1) */
+		poly ^= aes_double ( poly );
+
+		/* Divide inverse polynomial by (x + 1).  This code
+		 * fragment is taken directly from the Wikipedia page
+		 * on the Rijndael S-box.  An explanation of why it
+		 * works would be greatly appreciated.
+		 */
+		invpoly ^= ( invpoly << 1 );
+		invpoly ^= ( invpoly << 2 );
+		invpoly ^= ( invpoly << 4 );
+		if ( invpoly & 0x80 )
+			invpoly ^= 0x09;
+		invpoly &= 0xff;
+
+		/* Apply affine transformation */
+		transformed = ( 0x63 ^ invpoly ^ rol8 ( invpoly, 1 ) ^
+				rol8 ( invpoly, 2 ) ^ rol8 ( invpoly, 3 ) ^
+				rol8 ( invpoly, 4 ) );
+
+		/* Populate S-box (within MixColumns lookup table) */
+		aes_mixcolumns.entry[poly].byte[0] = transformed;
+
+	} while ( poly != 0x01 );
+
+	/* Populate zeroth S-box entry (which has no inverse) */
+	aes_mixcolumns.entry[0].byte[0] = 0x63;
+
+	/* Fill in MixColumns and InvMixColumns lookup tables */
+	for ( i = 0 ; i < 256 ; i++ ) {
+
+		/* Fill in MixColumns lookup table entry */
+		entry = &aes_mixcolumns.entry[i];
+		aes_mixcolumns_entry ( entry );
+
+		/* Populate inverse S-box (within InvMixColumns lookup table) */
+		inventry = &aes_invmixcolumns.entry[ entry->byte[0] ];
+		inventry->byte[0] = i;
+
+		/* Fill in InvMixColumns lookup table entry */
+		aes_invmixcolumns_entry ( inventry );
+	}
+}
+
+/**
+ * Rotate key column
+ *
+ * @v column		Key column
+ * @ret column		Updated key column
+ */
+static inline __attribute__ (( always_inline )) uint32_t
+aes_key_rotate ( uint32_t column ) {
+
+	return ( ( __BYTE_ORDER == __LITTLE_ENDIAN ) ?
+		 ror32 ( column, 8 ) : rol32 ( column, 8 ) );
+}
+
+/**
+ * Apply S-box to key column
+ *
+ * @v column		Key column
+ * @ret column		Updated key column
+ */
+static uint32_t aes_key_sbox ( uint32_t column ) {
+	unsigned int i;
+	uint8_t byte;
+
+	for ( i = 0 ; i < 4 ; i++ ) {
+		byte = ( column & 0xff );
+		byte = aes_mixcolumns.entry[byte].byte[0];
+		column = ( ( column & ~0xff ) | byte );
+		column = rol32 ( column, 8 );
+	}
+	return column;
+}
+
+/**
+ * Apply schedule round constant to key column
+ *
+ * @v column		Key column
+ * @v rcon		Round constant
+ * @ret column		Updated key column
+ */
+static inline __attribute__ (( always_inline )) uint32_t
+aes_key_rcon ( uint32_t column, unsigned int rcon ) {
+
+	return ( ( __BYTE_ORDER == __LITTLE_ENDIAN ) ?
+		 ( column ^ rcon ) : ( column ^ ( rcon << 24 ) ) );
+}
+
+/**
+ * Set key
+ *
+ * @v ctx		Context
+ * @v key		Key
+ * @v keylen		Key length
+ * @ret rc		Return status code
+ */
+static int aes_setkey ( void *ctx, const void *key, size_t keylen ) {
+	struct aes_context *aes = ctx;
+	union aes_matrix *enc;
+	union aes_matrix *dec;
+	union aes_matrix temp;
+	union aes_matrix zero;
+	unsigned int rcon = 0x01;
+	unsigned int rounds;
+	size_t offset = 0;
+	uint32_t *prev;
+	uint32_t *next;
+	uint32_t *end;
+	uint32_t tmp;
+
+	/* Generate lookup tables, if not already done */
+	if ( ! aes_mixcolumns.entry[0].byte[0] )
+		aes_generate();
+
+	/* Validate key length and calculate number of intermediate rounds */
+	switch ( keylen ) {
+	case ( 128 / 8 ) :
+		rounds = 11;
+		break;
+	case ( 192 / 8 ) :
+		rounds = 13;
+		break;
+	case ( 256 / 8 ) :
+		rounds = 15;
+		break;
+	default:
+		DBGC ( aes, "AES %p unsupported key length (%zd bits)\n",
+		       aes, ( keylen * 8 ) );
+		return -EINVAL;
+	}
+	aes->rounds = rounds;
+	enc = aes->encrypt.key;
+	end = enc[rounds].column;
+
+	/* Copy raw key */
+	memcpy ( enc, key, keylen );
+	prev = enc->column;
+	next = ( ( ( void * ) prev ) + keylen );
+	tmp = next[-1];
+
+	/* Construct expanded key */
+	while ( next < end ) {
+
+		/* If this is the first column of an expanded key
+		 * block, or the middle column of an AES-256 key
+		 * block, then apply the S-box.
+		 */
+		if ( ( offset == 0 ) || ( ( offset | keylen ) == 48 ) )
+			tmp = aes_key_sbox ( tmp );
+
+		/* If this is the first column of an expanded key
+		 * block then rotate and apply the round constant.
+		 */
+		if ( offset == 0 ) {
+			tmp = aes_key_rotate ( tmp );
+			tmp = aes_key_rcon ( tmp, rcon );
+			rcon = aes_double ( rcon );
+		}
+
+		/* XOR with previous key column */
+		tmp ^= *prev;
+
+		/* Store column */
+		*next = tmp;
+
+		/* Move to next column */
+		offset += sizeof ( *next );
+		if ( offset == keylen )
+			offset = 0;
+		next++;
+		prev++;
+	}
+	DBGC2 ( aes, "AES %p expanded %zd-bit key:\n", aes, ( keylen * 8 ) );
+	DBGC2_HDA ( aes, 0, &aes->encrypt, ( rounds * sizeof ( *enc ) ) );
+
+	/* Convert to decryption key */
+	memset ( &zero, 0, sizeof ( zero ) );
+	dec = &aes->decrypt.key[ rounds - 1 ];
+	memcpy ( dec--, enc++, sizeof ( *dec ) );
+	while ( dec > aes->decrypt.key ) {
+		/* Perform InvMixColumns (by reusing the encryption
+		 * final-round code to perform ShiftRows+SubBytes and
+		 * reusing the decryption intermediate-round code to
+		 * perform InvShiftRows+InvSubBytes+InvMixColumns, all
+		 * with a zero encryption key).
+		 */
+		aes_final ( &aes_mixcolumns, AES_STRIDE_SHIFTROWS,
+			    enc++, &temp, &zero );
+		aes_decrypt_rounds ( &temp, dec--, &zero, 1 );
+	}
+	memcpy ( dec--, enc++, sizeof ( *dec ) );
+	DBGC2 ( aes, "AES %p inverted %zd-bit key:\n", aes, ( keylen * 8 ) );
+	DBGC2_HDA ( aes, 0, &aes->decrypt, ( rounds * sizeof ( *dec ) ) );
+
+	return 0;
+}
+
+/**
+ * Set initialisation vector
+ *
+ * @v ctx		Context
+ * @v iv		Initialisation vector
+ */
+static void aes_setiv ( void *ctx __unused, const void *iv __unused ) {
+	/* Nothing to do */
+}
+
+/** Basic AES algorithm */
+struct cipher_algorithm aes_algorithm = {
+	.name = "aes",
+	.ctxsize = sizeof ( struct aes_context ),
+	.blocksize = AES_BLOCKSIZE,
+	.setkey = aes_setkey,
+	.setiv = aes_setiv,
+	.encrypt = aes_encrypt,
+	.decrypt = aes_decrypt,
+};
+
+/* AES in Electronic Codebook mode */
+ECB_CIPHER ( aes_ecb, aes_ecb_algorithm,
+	     aes_algorithm, struct aes_context, AES_BLOCKSIZE );
+
+/* AES in Cipher Block Chaining mode */
+CBC_CIPHER ( aes_cbc, aes_cbc_algorithm,
+	     aes_algorithm, struct aes_context, AES_BLOCKSIZE );
diff --git a/src/crypto/axtls/aes.c b/src/crypto/axtls/aes.c
deleted file mode 100644
index bd99a709..00000000
--- a/src/crypto/axtls/aes.c
+++ /dev/null
@@ -1,457 +0,0 @@
-/*
- * Copyright (c) 2007, Cameron Rich
- *
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *
- * * Redistributions of source code must retain the above copyright notice,
- *   this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright notice,
- *   this list of conditions and the following disclaimer in the documentation
- *   and/or other materials provided with the distribution.
- * * Neither the name of the axTLS project nor the names of its contributors
- *   may be used to endorse or promote products derived from this software
- *   without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
- * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
- * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
- * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
- * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
- * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
- * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-/**
- * AES implementation - this is a small code version. There are much faster
- * versions around but they are much larger in size (i.e. they use large 
- * submix tables).
- */
-
-#include <string.h>
-#include "os_port.h"
-#include "crypto.h"
-
-/* all commented out in skeleton mode */
-#ifndef CONFIG_SSL_SKELETON_MODE
-
-#define rot1(x) (((x) << 24) | ((x) >> 8))
-#define rot2(x) (((x) << 16) | ((x) >> 16))
-#define rot3(x) (((x) <<  8) | ((x) >> 24))
-
-/* 
- * This cute trick does 4 'mul by two' at once.  Stolen from
- * Dr B. R. Gladman <brg@gladman.uk.net> but I'm sure the u-(u>>7) is
- * a standard graphics trick
- * The key to this is that we need to xor with 0x1b if the top bit is set.
- * a 1xxx xxxx   0xxx 0xxx First we mask the 7bit,
- * b 1000 0000   0000 0000 then we shift right by 7 putting the 7bit in 0bit,
- * c 0000 0001   0000 0000 we then subtract (c) from (b)
- * d 0111 1111   0000 0000 and now we and with our mask
- * e 0001 1011   0000 0000
- */
-#define mt  0x80808080
-#define ml  0x7f7f7f7f
-#define mh  0xfefefefe
-#define mm  0x1b1b1b1b
-#define mul2(x,t)	((t)=((x)&mt), \
-			((((x)+(x))&mh)^(((t)-((t)>>7))&mm)))
-
-#define inv_mix_col(x,f2,f4,f8,f9) (\
-			(f2)=mul2(x,f2), \
-			(f4)=mul2(f2,f4), \
-			(f8)=mul2(f4,f8), \
-			(f9)=(x)^(f8), \
-			(f8)=((f2)^(f4)^(f8)), \
-			(f2)^=(f9), \
-			(f4)^=(f9), \
-			(f8)^=rot3(f2), \
-			(f8)^=rot2(f4), \
-			(f8)^rot1(f9))
-
-/*
- * AES S-box
- */
-static const uint8_t aes_sbox[256] =
-{
-	0x63,0x7C,0x77,0x7B,0xF2,0x6B,0x6F,0xC5,
-	0x30,0x01,0x67,0x2B,0xFE,0xD7,0xAB,0x76,
-	0xCA,0x82,0xC9,0x7D,0xFA,0x59,0x47,0xF0,
-	0xAD,0xD4,0xA2,0xAF,0x9C,0xA4,0x72,0xC0,
-	0xB7,0xFD,0x93,0x26,0x36,0x3F,0xF7,0xCC,
-	0x34,0xA5,0xE5,0xF1,0x71,0xD8,0x31,0x15,
-	0x04,0xC7,0x23,0xC3,0x18,0x96,0x05,0x9A,
-	0x07,0x12,0x80,0xE2,0xEB,0x27,0xB2,0x75,
-	0x09,0x83,0x2C,0x1A,0x1B,0x6E,0x5A,0xA0,
-	0x52,0x3B,0xD6,0xB3,0x29,0xE3,0x2F,0x84,
-	0x53,0xD1,0x00,0xED,0x20,0xFC,0xB1,0x5B,
-	0x6A,0xCB,0xBE,0x39,0x4A,0x4C,0x58,0xCF,
-	0xD0,0xEF,0xAA,0xFB,0x43,0x4D,0x33,0x85,
-	0x45,0xF9,0x02,0x7F,0x50,0x3C,0x9F,0xA8,
-	0x51,0xA3,0x40,0x8F,0x92,0x9D,0x38,0xF5,
-	0xBC,0xB6,0xDA,0x21,0x10,0xFF,0xF3,0xD2,
-	0xCD,0x0C,0x13,0xEC,0x5F,0x97,0x44,0x17,
-	0xC4,0xA7,0x7E,0x3D,0x64,0x5D,0x19,0x73,
-	0x60,0x81,0x4F,0xDC,0x22,0x2A,0x90,0x88,
-	0x46,0xEE,0xB8,0x14,0xDE,0x5E,0x0B,0xDB,
-	0xE0,0x32,0x3A,0x0A,0x49,0x06,0x24,0x5C,
-	0xC2,0xD3,0xAC,0x62,0x91,0x95,0xE4,0x79,
-	0xE7,0xC8,0x37,0x6D,0x8D,0xD5,0x4E,0xA9,
-	0x6C,0x56,0xF4,0xEA,0x65,0x7A,0xAE,0x08,
-	0xBA,0x78,0x25,0x2E,0x1C,0xA6,0xB4,0xC6,
-	0xE8,0xDD,0x74,0x1F,0x4B,0xBD,0x8B,0x8A,
-	0x70,0x3E,0xB5,0x66,0x48,0x03,0xF6,0x0E,
-	0x61,0x35,0x57,0xB9,0x86,0xC1,0x1D,0x9E,
-	0xE1,0xF8,0x98,0x11,0x69,0xD9,0x8E,0x94,
-	0x9B,0x1E,0x87,0xE9,0xCE,0x55,0x28,0xDF,
-	0x8C,0xA1,0x89,0x0D,0xBF,0xE6,0x42,0x68,
-	0x41,0x99,0x2D,0x0F,0xB0,0x54,0xBB,0x16,
-};
-
-/*
- * AES is-box
- */
-static const uint8_t aes_isbox[256] = 
-{
-    0x52,0x09,0x6a,0xd5,0x30,0x36,0xa5,0x38,
-    0xbf,0x40,0xa3,0x9e,0x81,0xf3,0xd7,0xfb,
-    0x7c,0xe3,0x39,0x82,0x9b,0x2f,0xff,0x87,
-    0x34,0x8e,0x43,0x44,0xc4,0xde,0xe9,0xcb,
-    0x54,0x7b,0x94,0x32,0xa6,0xc2,0x23,0x3d,
-    0xee,0x4c,0x95,0x0b,0x42,0xfa,0xc3,0x4e,
-    0x08,0x2e,0xa1,0x66,0x28,0xd9,0x24,0xb2,
-    0x76,0x5b,0xa2,0x49,0x6d,0x8b,0xd1,0x25,
-    0x72,0xf8,0xf6,0x64,0x86,0x68,0x98,0x16,
-    0xd4,0xa4,0x5c,0xcc,0x5d,0x65,0xb6,0x92,
-    0x6c,0x70,0x48,0x50,0xfd,0xed,0xb9,0xda,
-    0x5e,0x15,0x46,0x57,0xa7,0x8d,0x9d,0x84,
-    0x90,0xd8,0xab,0x00,0x8c,0xbc,0xd3,0x0a,
-    0xf7,0xe4,0x58,0x05,0xb8,0xb3,0x45,0x06,
-    0xd0,0x2c,0x1e,0x8f,0xca,0x3f,0x0f,0x02,
-    0xc1,0xaf,0xbd,0x03,0x01,0x13,0x8a,0x6b,
-    0x3a,0x91,0x11,0x41,0x4f,0x67,0xdc,0xea,
-    0x97,0xf2,0xcf,0xce,0xf0,0xb4,0xe6,0x73,
-    0x96,0xac,0x74,0x22,0xe7,0xad,0x35,0x85,
-    0xe2,0xf9,0x37,0xe8,0x1c,0x75,0xdf,0x6e,
-    0x47,0xf1,0x1a,0x71,0x1d,0x29,0xc5,0x89,
-    0x6f,0xb7,0x62,0x0e,0xaa,0x18,0xbe,0x1b,
-    0xfc,0x56,0x3e,0x4b,0xc6,0xd2,0x79,0x20,
-    0x9a,0xdb,0xc0,0xfe,0x78,0xcd,0x5a,0xf4,
-    0x1f,0xdd,0xa8,0x33,0x88,0x07,0xc7,0x31,
-    0xb1,0x12,0x10,0x59,0x27,0x80,0xec,0x5f,
-    0x60,0x51,0x7f,0xa9,0x19,0xb5,0x4a,0x0d,
-    0x2d,0xe5,0x7a,0x9f,0x93,0xc9,0x9c,0xef,
-    0xa0,0xe0,0x3b,0x4d,0xae,0x2a,0xf5,0xb0,
-    0xc8,0xeb,0xbb,0x3c,0x83,0x53,0x99,0x61,
-    0x17,0x2b,0x04,0x7e,0xba,0x77,0xd6,0x26,
-    0xe1,0x69,0x14,0x63,0x55,0x21,0x0c,0x7d
-};
-
-static const unsigned char Rcon[30]=
-{
-	0x01,0x02,0x04,0x08,0x10,0x20,0x40,0x80,
-	0x1b,0x36,0x6c,0xd8,0xab,0x4d,0x9a,0x2f,
-	0x5e,0xbc,0x63,0xc6,0x97,0x35,0x6a,0xd4,
-	0xb3,0x7d,0xfa,0xef,0xc5,0x91,
-};
-
-/* ----- static functions ----- */
-static void AES_encrypt(const AES_CTX *ctx, uint32_t *data);
-static void AES_decrypt(const AES_CTX *ctx, uint32_t *data);
-
-/* Perform doubling in Galois Field GF(2^8) using the irreducible polynomial
-   x^8+x^4+x^3+x+1 */
-static unsigned char AES_xtime(uint32_t x)
-{
-	return (x&0x80) ? (x<<1)^0x1b : x<<1;
-}
-
-/**
- * Set up AES with the key/iv and cipher size.
- */
-void AES_set_key(AES_CTX *ctx, const uint8_t *key, 
-        const uint8_t *iv, AES_MODE mode)
-{
-    int i, ii;
-    uint32_t *W, tmp, tmp2;
-    const unsigned char *ip;
-    int words;
-
-    switch (mode)
-    {
-        case AES_MODE_128:
-            i = 10;
-            words = 4;
-            break;
-
-        case AES_MODE_256:
-            i = 14;
-            words = 8;
-            break;
-
-        default:        /* fail silently */
-            return;
-    }
-
-    ctx->rounds = i;
-    ctx->key_size = words;
-    W = ctx->ks;
-    for (i = 0; i < words; i+=2)
-    {
-        W[i+0]=	((uint32_t)key[ 0]<<24)|
-            ((uint32_t)key[ 1]<<16)|
-            ((uint32_t)key[ 2]<< 8)|
-            ((uint32_t)key[ 3]    );
-        W[i+1]=	((uint32_t)key[ 4]<<24)|
-            ((uint32_t)key[ 5]<<16)|
-            ((uint32_t)key[ 6]<< 8)|
-            ((uint32_t)key[ 7]    );
-        key += 8;
-    }
-
-    ip = Rcon;
-    ii = 4 * (ctx->rounds+1);
-    for (i = words; i<ii; i++)
-    {
-        tmp = W[i-1];
-
-        if ((i % words) == 0)
-        {
-            tmp2 =(uint32_t)aes_sbox[(tmp    )&0xff]<< 8;
-            tmp2|=(uint32_t)aes_sbox[(tmp>> 8)&0xff]<<16;
-            tmp2|=(uint32_t)aes_sbox[(tmp>>16)&0xff]<<24;
-            tmp2|=(uint32_t)aes_sbox[(tmp>>24)     ];
-            tmp=tmp2^(((unsigned int)*ip)<<24);
-            ip++;
-        }
-
-        if ((words == 8) && ((i % words) == 4))
-        {
-            tmp2 =(uint32_t)aes_sbox[(tmp    )&0xff]    ;
-            tmp2|=(uint32_t)aes_sbox[(tmp>> 8)&0xff]<< 8;
-            tmp2|=(uint32_t)aes_sbox[(tmp>>16)&0xff]<<16;
-            tmp2|=(uint32_t)aes_sbox[(tmp>>24)     ]<<24;
-            tmp=tmp2;
-        }
-
-        W[i]=W[i-words]^tmp;
-    }
-
-    /* copy the iv across */
-    memcpy(ctx->iv, iv, 16);
-}
-
-/**
- * Change a key for decryption.
- */
-void AES_convert_key(AES_CTX *ctx)
-{
-    int i;
-    uint32_t *k,w,t1,t2,t3,t4;
-
-    k = ctx->ks;
-    k += 4;
-
-    for (i= ctx->rounds*4; i > 4; i--)
-    {
-        w= *k;
-        w = inv_mix_col(w,t1,t2,t3,t4);
-        *k++ =w;
-    }
-}
-
-/**
- * Encrypt a byte sequence (with a block size 16) using the AES cipher.
- */
-void AES_cbc_encrypt(AES_CTX *ctx, const uint8_t *msg, uint8_t *out, int length)
-{
-    int i;
-    uint32_t tin[4], tout[4], iv[4];
-
-    memcpy(iv, ctx->iv, AES_IV_SIZE);
-    for (i = 0; i < 4; i++)
-        tout[i] = ntohl(iv[i]);
-
-    for (length -= AES_BLOCKSIZE; length >= 0; length -= AES_BLOCKSIZE)
-    {
-        uint32_t msg_32[4];
-        uint32_t out_32[4];
-        memcpy(msg_32, msg, AES_BLOCKSIZE);
-        msg += AES_BLOCKSIZE;
-
-        for (i = 0; i < 4; i++)
-            tin[i] = ntohl(msg_32[i])^tout[i];
-
-        AES_encrypt(ctx, tin);
-
-        for (i = 0; i < 4; i++)
-        {
-            tout[i] = tin[i];
-            out_32[i] = htonl(tout[i]);
-        }
-
-        memcpy(out, out_32, AES_BLOCKSIZE);
-        out += AES_BLOCKSIZE;
-    }
-
-    for (i = 0; i < 4; i++)
-        iv[i] = htonl(tout[i]);
-    memcpy(ctx->iv, iv, AES_IV_SIZE);
-}
-
-/**
- * Decrypt a byte sequence (with a block size 16) using the AES cipher.
- */
-void AES_cbc_decrypt(AES_CTX *ctx, const uint8_t *msg, uint8_t *out, int length)
-{
-    int i;
-    uint32_t tin[4], xor[4], tout[4], data[4], iv[4];
-
-    memcpy(iv, ctx->iv, AES_IV_SIZE);
-    for (i = 0; i < 4; i++)
-        xor[i] = ntohl(iv[i]);
-
-    for (length -= 16; length >= 0; length -= 16)
-    {
-        uint32_t msg_32[4];
-        uint32_t out_32[4];
-        memcpy(msg_32, msg, AES_BLOCKSIZE);
-        msg += AES_BLOCKSIZE;
-
-        for (i = 0; i < 4; i++)
-        {
-            tin[i] = ntohl(msg_32[i]);
-            data[i] = tin[i];
-        }
-
-        AES_decrypt(ctx, data);
-
-        for (i = 0; i < 4; i++)
-        {
-            tout[i] = data[i]^xor[i];
-            xor[i] = tin[i];
-            out_32[i] = htonl(tout[i]);
-        }
-
-        memcpy(out, out_32, AES_BLOCKSIZE);
-        out += AES_BLOCKSIZE;
-    }
-
-    for (i = 0; i < 4; i++)
-        iv[i] = htonl(xor[i]);
-    memcpy(ctx->iv, iv, AES_IV_SIZE);
-}
-
-/**
- * Encrypt a single block (16 bytes) of data
- */
-static void AES_encrypt(const AES_CTX *ctx, uint32_t *data)
-{
-    /* To make this code smaller, generate the sbox entries on the fly.
-     * This will have a really heavy effect upon performance.
-     */
-    uint32_t tmp[4];
-    uint32_t tmp1, old_a0, a0, a1, a2, a3, row;
-    int curr_rnd;
-    int rounds = ctx->rounds; 
-    const uint32_t *k = ctx->ks;
-
-    /* Pre-round key addition */
-    for (row = 0; row < 4; row++)
-        data[row] ^= *(k++);
-
-    /* Encrypt one block. */
-    for (curr_rnd = 0; curr_rnd < rounds; curr_rnd++)
-    {
-        /* Perform ByteSub and ShiftRow operations together */
-        for (row = 0; row < 4; row++)
-        {
-            a0 = (uint32_t)aes_sbox[(data[row%4]>>24)&0xFF];
-            a1 = (uint32_t)aes_sbox[(data[(row+1)%4]>>16)&0xFF];
-            a2 = (uint32_t)aes_sbox[(data[(row+2)%4]>>8)&0xFF]; 
-            a3 = (uint32_t)aes_sbox[(data[(row+3)%4])&0xFF];
-
-            /* Perform MixColumn iff not last round */
-            if (curr_rnd < (rounds - 1))
-            {
-                tmp1 = a0 ^ a1 ^ a2 ^ a3;
-                old_a0 = a0;
-                a0 ^= tmp1 ^ AES_xtime(a0 ^ a1);
-                a1 ^= tmp1 ^ AES_xtime(a1 ^ a2);
-                a2 ^= tmp1 ^ AES_xtime(a2 ^ a3);
-                a3 ^= tmp1 ^ AES_xtime(a3 ^ old_a0);
-            }
-
-            tmp[row] = ((a0 << 24) | (a1 << 16) | (a2 << 8) | a3);
-        }
-
-        /* KeyAddition - note that it is vital that this loop is separate from
-           the MixColumn operation, which must be atomic...*/ 
-        for (row = 0; row < 4; row++)
-            data[row] = tmp[row] ^ *(k++);
-    }
-}
-
-/**
- * Decrypt a single block (16 bytes) of data
- */
-static void AES_decrypt(const AES_CTX *ctx, uint32_t *data)
-{ 
-    uint32_t tmp[4];
-    uint32_t xt0,xt1,xt2,xt3,xt4,xt5,xt6;
-    uint32_t a0, a1, a2, a3, row;
-    int curr_rnd;
-    int rounds = ctx->rounds;
-    const uint32_t *k = ctx->ks + ((rounds+1)*4);
-
-    /* pre-round key addition */
-    for (row=4; row > 0;row--)
-        data[row-1] ^= *(--k);
-
-    /* Decrypt one block */
-    for (curr_rnd = 0; curr_rnd < rounds; curr_rnd++)
-    {
-        /* Perform ByteSub and ShiftRow operations together */
-        for (row = 4; row > 0; row--)
-        {
-            a0 = aes_isbox[(data[(row+3)%4]>>24)&0xFF];
-            a1 = aes_isbox[(data[(row+2)%4]>>16)&0xFF];
-            a2 = aes_isbox[(data[(row+1)%4]>>8)&0xFF];
-            a3 = aes_isbox[(data[row%4])&0xFF];
-
-            /* Perform MixColumn iff not last round */
-            if (curr_rnd<(rounds-1))
-            {
-                /* The MDS cofefficients (0x09, 0x0B, 0x0D, 0x0E)
-                   are quite large compared to encryption; this 
-                   operation slows decryption down noticeably. */
-                xt0 = AES_xtime(a0^a1);
-                xt1 = AES_xtime(a1^a2);
-                xt2 = AES_xtime(a2^a3);
-                xt3 = AES_xtime(a3^a0);
-                xt4 = AES_xtime(xt0^xt1);
-                xt5 = AES_xtime(xt1^xt2);
-                xt6 = AES_xtime(xt4^xt5);
-
-                xt0 ^= a1^a2^a3^xt4^xt6;
-                xt1 ^= a0^a2^a3^xt5^xt6;
-                xt2 ^= a0^a1^a3^xt4^xt6;
-                xt3 ^= a0^a1^a2^xt5^xt6;
-                tmp[row-1] = ((xt0<<24)|(xt1<<16)|(xt2<<8)|xt3);
-            }
-            else
-                tmp[row-1] = ((a0<<24)|(a1<<16)|(a2<<8)|a3);
-        }
-
-        for (row = 4; row > 0; row--)
-            data[row-1] = tmp[row-1] ^ *(--k);
-    }
-}
-
-#endif
diff --git a/src/crypto/axtls_aes.c b/src/crypto/axtls_aes.c
deleted file mode 100644
index 9d6b6459..00000000
--- a/src/crypto/axtls_aes.c
+++ /dev/null
@@ -1,165 +0,0 @@
-/*
- * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License as
- * published by the Free Software Foundation; either version 2 of the
- * License, or any later version.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
- * 02110-1301, USA.
- */
-
-FILE_LICENCE ( GPL2_OR_LATER );
-
-#include <string.h>
-#include <errno.h>
-#include <assert.h>
-#include <byteswap.h>
-#include <ipxe/crypto.h>
-#include <ipxe/ecb.h>
-#include <ipxe/cbc.h>
-#include <ipxe/aes.h>
-#include "crypto/axtls/crypto.h"
-
-/** @file
- *
- * AES algorithm
- *
- */
-
-/**
- * Set key
- *
- * @v ctx		Context
- * @v key		Key
- * @v keylen		Key length
- * @ret rc		Return status code
- */
-static int aes_setkey ( void *ctx, const void *key, size_t keylen ) {
-	struct aes_context *aes_ctx = ctx;
-	AES_MODE mode;
-	void *iv;
-
-	switch ( keylen ) {
-	case ( 128 / 8 ):
-		mode = AES_MODE_128;
-		break;
-	case ( 256 / 8 ):
-		mode = AES_MODE_256;
-		break;
-	default:
-		return -EINVAL;
-	}
-
-	/* IV is not a relevant concept at this stage; use a dummy
-	 * value that will have no side-effects.
-	 */
-	iv = &aes_ctx->axtls_ctx.iv;
-
-	AES_set_key ( &aes_ctx->axtls_ctx, key, iv, mode );
-
-	aes_ctx->decrypting = 0;
-
-	return 0;
-}
-
-/**
- * Set initialisation vector
- *
- * @v ctx		Context
- * @v iv		Initialisation vector
- */
-static void aes_setiv ( void *ctx __unused, const void *iv __unused ) {
-	/* Nothing to do */
-}
-
-/**
- * Call AXTLS' AES_encrypt() or AES_decrypt() functions
- *
- * @v axtls_ctx		AXTLS AES context
- * @v src		Data to process
- * @v dst		Buffer for output
- * @v func		AXTLS AES function to call
- */
-static void aes_call_axtls ( AES_CTX *axtls_ctx, const void *src, void *dst,
-			     void ( * func ) ( const AES_CTX *axtls_ctx,
-					       uint32_t *data ) ){
-	const uint32_t *srcl = src;
-	uint32_t *dstl = dst;
-	unsigned int i;
-
-	/* AXTLS' AES_encrypt() and AES_decrypt() functions both
-	 * expect to deal with an array of four dwords in host-endian
-	 * order.
-	 */
-	for ( i = 0 ; i < 4 ; i++ )
-		dstl[i] = ntohl ( srcl[i] );
-	func ( axtls_ctx, dstl );
-	for ( i = 0 ; i < 4 ; i++ )
-		dstl[i] = htonl ( dstl[i] );
-}
-
-/**
- * Encrypt data
- *
- * @v ctx		Context
- * @v src		Data to encrypt
- * @v dst		Buffer for encrypted data
- * @v len		Length of data
- */
-static void aes_encrypt ( void *ctx, const void *src, void *dst,
-			  size_t len ) {
-	struct aes_context *aes_ctx = ctx;
-
-	assert ( len == AES_BLOCKSIZE );
-	if ( aes_ctx->decrypting )
-		assert ( 0 );
-	aes_call_axtls ( &aes_ctx->axtls_ctx, src, dst, axtls_aes_encrypt );
-}
-
-/**
- * Decrypt data
- *
- * @v ctx		Context
- * @v src		Data to decrypt
- * @v dst		Buffer for decrypted data
- * @v len		Length of data
- */
-static void aes_decrypt ( void *ctx, const void *src, void *dst,
-			  size_t len ) {
-	struct aes_context *aes_ctx = ctx;
-
-	assert ( len == AES_BLOCKSIZE );
-	if ( ! aes_ctx->decrypting ) {
-		AES_convert_key ( &aes_ctx->axtls_ctx );
-		aes_ctx->decrypting = 1;
-	}
-	aes_call_axtls ( &aes_ctx->axtls_ctx, src, dst, axtls_aes_decrypt );
-}
-
-/** Basic AES algorithm */
-struct cipher_algorithm aes_algorithm = {
-	.name = "aes",
-	.ctxsize = sizeof ( struct aes_context ),
-	.blocksize = AES_BLOCKSIZE,
-	.setkey = aes_setkey,
-	.setiv = aes_setiv,
-	.encrypt = aes_encrypt,
-	.decrypt = aes_decrypt,
-};
-
-/* AES in Electronic Codebook mode */
-ECB_CIPHER ( aes_ecb, aes_ecb_algorithm,
-	     aes_algorithm, struct aes_context, AES_BLOCKSIZE );
-
-/* AES in Cipher Block Chaining mode */
-CBC_CIPHER ( aes_cbc, aes_cbc_algorithm,
-	     aes_algorithm, struct aes_context, AES_BLOCKSIZE );
diff --git a/src/include/ipxe/aes.h b/src/include/ipxe/aes.h
index 91e7483a..0432e43e 100644
--- a/src/include/ipxe/aes.h
+++ b/src/include/ipxe/aes.h
@@ -1,30 +1,49 @@
 #ifndef _IPXE_AES_H
 #define _IPXE_AES_H
 
-FILE_LICENCE ( GPL2_OR_LATER );
+/** @file
+ *
+ * AES algorithm
+ *
+ */
 
-struct cipher_algorithm;
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 
-/** Basic AES blocksize */
+#include <ipxe/crypto.h>
+
+/** AES blocksize */
 #define AES_BLOCKSIZE 16
 
-#include "crypto/axtls/crypto.h"
+/** Maximum number of AES rounds */
+#define AES_MAX_ROUNDS 15
+
+/** AES matrix */
+union aes_matrix {
+	/** Viewed as an array of bytes */
+	uint8_t byte[16];
+	/** Viewed as an array of four-byte columns */
+	uint32_t column[4];
+} __attribute__ (( packed ));
+
+/** AES round keys */
+struct aes_round_keys {
+	/** Round keys */
+	union aes_matrix key[AES_MAX_ROUNDS];
+};
 
 /** AES context */
 struct aes_context {
-	/** AES context for AXTLS */
-	AES_CTX axtls_ctx;
-	/** Cipher is being used for decrypting */
-	int decrypting;
+	/** Encryption keys */
+	struct aes_round_keys encrypt;
+	/** Decryption keys */
+	struct aes_round_keys decrypt;
+	/** Number of rounds */
+	unsigned int rounds;
 };
 
 /** AES context size */
 #define AES_CTX_SIZE sizeof ( struct aes_context )
 
-/* AXTLS functions */
-extern void axtls_aes_encrypt ( const AES_CTX *ctx, uint32_t *data );
-extern void axtls_aes_decrypt ( const AES_CTX *ctx, uint32_t *data );
-
 extern struct cipher_algorithm aes_algorithm;
 extern struct cipher_algorithm aes_ecb_algorithm;
 extern struct cipher_algorithm aes_cbc_algorithm;
diff --git a/src/include/ipxe/errfile.h b/src/include/ipxe/errfile.h
index 6553cd87..de50fab9 100644
--- a/src/include/ipxe/errfile.h
+++ b/src/include/ipxe/errfile.h
@@ -264,7 +264,7 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 #define ERRFILE_imgmgmt		      ( ERRFILE_OTHER | 0x00050000 )
 #define ERRFILE_pxe_tftp	      ( ERRFILE_OTHER | 0x00060000 )
 #define ERRFILE_pxe_udp		      ( ERRFILE_OTHER | 0x00070000 )
-#define ERRFILE_axtls_aes	      ( ERRFILE_OTHER | 0x00080000 )
+#define ERRFILE_aes	 	      ( ERRFILE_OTHER | 0x00080000 )
 #define ERRFILE_cipher		      ( ERRFILE_OTHER | 0x00090000 )
 #define ERRFILE_image_cmd	      ( ERRFILE_OTHER | 0x000a0000 )
 #define ERRFILE_uri_test	      ( ERRFILE_OTHER | 0x000b0000 )