--- /dev/null
+/* sha256-asm.S */
+/*
+ This file is part of the AVR-Crypto-Lib.
+ Copyright (C) 2008 Daniel Otte (daniel.otte@rub.de)
+
+ 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 3 of the License, or
+ (at your option) 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, see <http://www.gnu.org/licenses/>.
+*/
+/*
+ * Author: Daniel Otte
+ *
+ * License: GPLv3 or later
+*/
+; sha-256 implementation in assembler
+SHA256_BLOCK_BITS = 512
+SHA256_HASH_BITS = 256
+
+.macro precall
+ /* push r18 - r27, r30 - r31*/
+ push r0
+ push r1
+ push r18
+ push r19
+ push r20
+ push r21
+ push r22
+ push r23
+ push r24
+ push r25
+ push r26
+ push r27
+ push r30
+ push r31
+ clr r1
+.endm
+
+.macro postcall
+ pop r31
+ pop r30
+ pop r27
+ pop r26
+ pop r25
+ pop r24
+ pop r23
+ pop r22
+ pop r21
+ pop r20
+ pop r19
+ pop r18
+ pop r1
+ pop r0
+.endm
+
+
+.macro hexdump length
+ push r27
+ push r26
+ ldi r25, '\r'
+ mov r24, r25
+ call uart_putc
+ ldi r25, '\n'
+ mov r24, r25
+ call uart_putc
+ pop r26
+ pop r27
+ movw r24, r26
+.if \length > 16
+ ldi r22, lo8(16)
+ ldi r23, hi8(16)
+ push r27
+ push r26
+ call uart_hexdump
+ pop r26
+ pop r27
+ adiw r26, 16
+ hexdump \length-16
+.else
+ ldi r22, lo8(\length)
+ ldi r23, hi8(\length)
+ call uart_hexdump
+.endif
+.endm
+
+/* X points to Block */
+.macro dbg_hexdump length
+ precall
+ hexdump \length
+ postcall
+.endm
+
+.section .text
+
+SPL = 0x3D
+SPH = 0x3E
+SREG = 0x3F
+
+
+;
+;sha256_ctx_t is:
+;
+; [h0][h1][h2][h3][h4][h5][h6][h7][length]
+; hn is 32 bit large, length is 64 bit large
+
+;###########################################################
+
+.global sha256_ctx2hash
+; === sha256_ctx2hash ===
+; this function converts a state into a normal hash (bytestring)
+; param1: the 16-bit destination pointer
+; given in r25,r24 (r25 is most significant)
+; param2: the 16-bit pointer to sha256_ctx structure
+; given in r23,r22
+sha256_ctx2hash:
+ movw r26, r22
+ movw r30, r24
+ ldi r21, 8
+ sbiw r26, 4
+1:
+ ldi r20, 4
+ adiw r26, 8
+2:
+ ld r0, -X
+ st Z+, r0
+ dec r20
+ brne 2b
+
+ dec r21
+ brne 1b
+
+ ret
+
+;###########################################################
+
+.global sha256
+; === sha256 ===
+; this function calculates SHA-256 hashes from messages in RAM
+; param1: the 16-bit hash destination pointer
+; given in r25,r24 (r25 is most significant)
+; param2: the 16-bit pointer to message
+; given in r23,r22
+; param3: 32-bit length value (length of message in bits)
+; given in r21,r20,r19,r18
+sha256:
+sha256_prolog:
+ push r8
+ push r9
+ push r10
+ push r11
+ push r12
+ push r13
+ push r16
+ push r17
+ in r16, SPL
+ in r17, SPH
+ subi r16, 8*4+8
+ sbci r17, 0
+ in r0, SREG
+ cli
+ out SPL, r16
+ out SPH, r17
+ out SREG, r0
+
+ push r25
+ push r24
+ inc r16
+ adc r17, r1
+
+ movw r8, r18 /* backup of length*/
+ movw r10, r20
+
+ movw r12, r22 /* backup pf msg-ptr */
+
+ movw r24, r16
+ rcall sha256_init
+ /* if length >= 512 */
+1:
+ tst r11
+ brne 4f
+ tst r10
+ brne 4f
+ mov r19, r9
+ cpi r19, 0x02
+ brlo 4f
+
+ movw r24, r16
+ movw r22, r12
+ rcall sha256_nextBlock
+ ldi r19, 0x64
+ add r22, r19
+ adc r23, r1
+ /* length -= 512 */
+ ldi r19, 0x02
+ sub r9, r19
+ sbc r10, r1
+ sbc r11, r1
+ rjmp 1b
+
+4:
+ movw r24, r16
+ movw r22, r12
+ movw r20, r8
+ rcall sha256_lastBlock
+
+ pop r24
+ pop r25
+ movw r22, r16
+ rcall sha256_ctx2hash
+
+sha256_epilog:
+ in r30, SPL
+ in r31, SPH
+ adiw r30, 8*4+8
+ in r0, SREG
+ cli
+ out SPL, r30
+ out SPH, r31
+ out SREG, r0
+ pop r17
+ pop r16
+ pop r13
+ pop r12
+ pop r11
+ pop r10
+ pop r9
+ pop r8
+ ret
+
+;###########################################################
+
+
+; block MUST NOT be larger than 64 bytes
+
+.global sha256_lastBlock
+; === sha256_lastBlock ===
+; this function does padding & Co. for calculating SHA-256 hashes
+; param1: the 16-bit pointer to sha256_ctx structure
+; given in r25,r24 (r25 is most significant)
+; param2: an 16-bit pointer to 64 byte block to hash
+; given in r23,r22
+; param3: an 16-bit integer specifing length of block in bits
+; given in r21,r20
+sha256_lastBlock_localSpace = (SHA256_BLOCK_BITS/8+1)
+
+
+sha256_lastBlock:
+ cpi r21, 0x02
+ brlo sha256_lastBlock_prolog
+ push r25
+ push r24
+ push r23
+ push r22
+ push r21
+ push r20
+ rcall sha256_nextBlock
+ pop r20
+ pop r21
+ pop r22
+ pop r23
+ pop r24
+ pop r25
+ subi r21, 0x02
+ subi r23, -2
+ rjmp sha256_lastBlock
+sha256_lastBlock_prolog:
+ /* allocate space on stack */
+ in r30, SPL
+ in r31, SPH
+ in r1, SREG
+ subi r30, lo8(64)
+ sbci r31, hi8(64)
+ cli
+ out SPL, r30
+ out SPH, r31
+ out SREG,r1
+
+ adiw r30, 1 /* SP points to next free byte on stack */
+ mov r18, r20 /* r20 = LSB(length) */
+ lsr r18
+ lsr r18
+ lsr r18
+ bst r21, 0 /* may be we should explain this ... */
+ bld r18, 5 /* now: r18 == length/8 (aka. length in bytes) */
+
+
+ movw r26, r22 /* X points to begin of msg */
+ tst r18
+ breq sha256_lastBlock_post_copy
+ mov r1, r18
+sha256_lastBlock_copy_loop:
+ ld r0, X+
+ st Z+, r0
+ dec r1
+ brne sha256_lastBlock_copy_loop
+sha256_lastBlock_post_copy:
+sha256_lastBlock_insert_stuffing_bit:
+ ldi r19, 0x80
+ mov r0,r19
+ ldi r19, 0x07
+ and r19, r20 /* if we are in bitmode */
+ breq 2f /* no bitmode */
+1:
+ lsr r0
+ dec r19
+ brne 1b
+ ld r19, X
+/* maybe we should do some ANDing here, just for safety */
+ or r0, r19
+2:
+ st Z+, r0
+ inc r18
+
+/* checking stuff here */
+ cpi r18, 64-8+1
+ brsh 0f
+ rjmp sha256_lastBlock_insert_zeros
+0:
+ /* oh shit, we landed here */
+ /* first we have to fill it up with zeros */
+ ldi r19, 64
+ sub r19, r18
+ breq 2f
+1:
+ st Z+, r1
+ dec r19
+ brne 1b
+2:
+ sbiw r30, 63
+ sbiw r30, 1
+ movw r22, r30
+
+ push r31
+ push r30
+ push r25
+ push r24
+ push r21
+ push r20
+ rcall sha256_nextBlock
+ pop r20
+ pop r21
+ pop r24
+ pop r25
+ pop r30
+ pop r31
+
+ /* now we should subtract 512 from length */
+ movw r26, r24
+ adiw r26, 4*8+1 /* we can skip the lowest byte */
+ ld r19, X
+ subi r19, hi8(512)
+ st X+, r19
+ ldi r18, 6
+1:
+ ld r19, X
+ sbci r19, 0
+ st X+, r19
+ dec r18
+ brne 1b
+
+; clr r18 /* not neccessary ;-) */
+ /* reset Z pointer to begin of block */
+
+sha256_lastBlock_insert_zeros:
+ ldi r19, 64-8
+ sub r19, r18
+ breq sha256_lastBlock_insert_length
+ clr r1
+1:
+ st Z+, r1 /* r1 is still zero */
+ dec r19
+ brne 1b
+
+; rjmp sha256_lastBlock_epilog
+sha256_lastBlock_insert_length:
+ movw r26, r24 /* X points to state */
+ adiw r26, 8*4 /* X points to (state.length) */
+ adiw r30, 8 /* Z points one after the last byte of block */
+ ld r0, X+
+ add r0, r20
+ st -Z, r0
+ ld r0, X+
+ adc r0, r21
+ st -Z, r0
+ ldi r19, 6
+1:
+ ld r0, X+
+ adc r0, r1
+ st -Z, r0
+ dec r19
+ brne 1b
+
+ sbiw r30, 64-8
+ movw r22, r30
+ rcall sha256_nextBlock
+
+sha256_lastBlock_epilog:
+ in r30, SPL
+ in r31, SPH
+ in r1, SREG
+ adiw r30, 63 ; lo8(64)
+ adiw r30, 1 ; hi8(64)
+ cli
+ out SPL, r30
+ out SPH, r31
+ out SREG,r1
+ clr r1
+ clr r0
+ ret
+
+/**/
+;###########################################################
+
+.global sha256_nextBlock
+; === sha256_nextBlock ===
+; this is the core function for calculating SHA-256 hashes
+; param1: the 16-bit pointer to sha256_ctx structure
+; given in r25,r24 (r25 is most significant)
+; param2: an 16-bit pointer to 64 byte block to hash
+; given in r23,r22
+sha256_nextBlock_localSpace = (64+8)*4 ; 64 32-bit values for w array and 8 32-bit values for a array (total 288 byte)
+
+Bck1 = 12
+Bck2 = 13
+Bck3 = 14
+Bck4 = 15
+Func1 = 22
+Func2 = 23
+Func3 = 24
+Func4 = 25
+Accu1 = 16
+Accu2 = 17
+Accu3 = 18
+Accu4 = 19
+XAccu1 = 8
+XAccu2 = 9
+XAccu3 = 10
+XAccu4 = 11
+T1 = 4
+T2 = 5
+T3 = 6
+T4 = 7
+LoopC = 1
+/* byteorder: high number <--> high significance */
+sha256_nextBlock:
+ ; initial, let's make some space ready for local vars
+ push r4 /* replace push & pop by mem ops? */
+ push r5
+ push r6
+ push r7
+ push r8
+ push r9
+ push r10
+ push r11
+ push r12
+ push r13
+ push r14
+ push r15
+ push r16
+ push r17
+ push r28
+ push r29
+ in r20, SPL
+ in r21, SPH
+ movw r18, r20 ;backup SP
+; movw r26, r20 ; X points to free space on stack
+ movw r30, r22 ; Z points to message
+ subi r20, lo8(sha256_nextBlock_localSpace) ;sbiw can do only up to 63
+ sbci r21, hi8(sha256_nextBlock_localSpace)
+ movw r26, r20 ; X points to free space on stack
+ in r0, SREG
+ cli ; we want to be uninterrupted while updating SP
+ out SPL, r20
+ out SPH, r21
+ out SREG, r0
+ push r18
+ push r19
+ push r24
+ push r25 /* param1 will be needed later */
+ ; now we fill the w array with message (think about endianess)
+ adiw r26, 1 ; X++
+ ldi r20, 16
+sha256_nextBlock_wcpyloop:
+ ld r23, Z+
+ ld r22, Z+
+ ld r19, Z+
+ ld r18, Z+
+ st X+, r18
+ st X+, r19
+ st X+, r22
+ st X+, r23
+ dec r20
+ brne sha256_nextBlock_wcpyloop
+/* for (i=16; i<64; ++i){
+ w[i] = SIGMA_b(w[i-2]) + w[i-7] + SIGMA_a(w[i-15]) + w[i-16];
+ } */
+ /* r25,r24,r23,r24 (r21,r20) are function values
+ r19,r18,r17,r16 are the accumulator
+ r15,r14,r13,rBck1 are backup1
+ r11,r10,r9 ,r8 are xor accu
+ r1 is round counter */
+
+ ldi r20, 64-16
+ mov LoopC, r20
+sha256_nextBlock_wcalcloop:
+ movw r30, r26 ; cp X to Z
+ sbiw r30, 63
+ sbiw r30, 1 ; substract 64 = 16*4
+ ld Accu1, Z+
+ ld Accu2, Z+
+ ld Accu3, Z+
+ ld Accu4, Z+ /* w[i] = w[i-16] */
+ ld Bck1, Z+
+ ld Bck2, Z+
+ ld Bck3, Z+
+ ld Bck4, Z+ /* backup = w[i-15] */
+ /* now sigma 0 */
+ mov Func1, Bck2
+ mov Func2, Bck3
+ mov Func3, Bck4
+ mov Func4, Bck1 /* prerotated by 8 */
+ ldi r20, 1
+ rcall bitrotl
+ movw XAccu1, Func1
+ movw XAccu3, Func3 /* store ROTR(w[i-15],7) in xor accu */
+ movw Func1, Bck3
+ movw Func3, Bck1 /* prerotated by 16 */
+ ldi r20, 2
+ rcall bitrotr
+ eor XAccu1, Func1 /* xor ROTR(w[i-15], 18)*/
+ eor XAccu2, Func2
+ eor XAccu3, Func3
+ eor XAccu4, Func4
+ ldi Func2, 3 /* now shr3 */ /*we can destroy backup now*/
+sigma0_shr:
+ lsr Bck4
+ ror Bck3
+ ror Bck2
+ ror Bck1
+ dec Func2
+ brne sigma0_shr
+ eor XAccu1, Bck1
+ eor XAccu2, Bck2
+ eor XAccu3, Bck3
+ eor XAccu4, Bck4 /* xor SHR(w[i-15], 3)*/ /* xor accu == sigma1(w[i-15]) */
+ add Accu1, XAccu1
+ adc Accu2, XAccu2
+ adc Accu3, XAccu3
+ adc Accu4, XAccu4 /* finished with sigma0 */
+ ldd Func1, Z+7*4 /* now accu += w[i-7] */
+ ldd Func2, Z+7*4+1
+ ldd Func3, Z+7*4+2
+ ldd Func4, Z+7*4+3
+ add Accu1, Func1
+ adc Accu2, Func2
+ adc Accu3, Func3
+ adc Accu4, Func4
+ ldd Bck1, Z+12*4 /* now backup = w[i-2]*/
+ ldd Bck2, Z+12*4+1
+ ldd Bck3, Z+12*4+2
+ ldd Bck4, Z+12*4+3
+ /* now sigma 1 */
+ movw Func1, Bck3
+ movw Func3, Bck1 /* prerotated by 16 */
+ ldi r20, 1
+ rcall bitrotr
+ movw XAccu3, Func3
+ movw XAccu1, Func1 /* store in ROTR(w[i-2], 17) xor accu */
+; movw Func1, Bck3
+; movw Func3, Bck1 /* prerotated by 16 */
+ ldi r20, 2
+ rcall bitrotr
+ eor XAccu1, Func1 /* xor ROTR(w[i-2], 19)*/
+ eor XAccu2, Func2
+ eor XAccu3, Func3
+ eor XAccu4, Func4
+ ldi Func2, 2 /* now shr10 (dirty trick, skipping a byte) */ /*we can destroy backup now*/
+sigma1_shr:
+ lsr Bck4
+ ror Bck3
+ ror Bck2
+ dec Func2
+ brne sigma1_shr
+ eor XAccu1, Bck2
+ eor XAccu2, Bck3
+ eor XAccu3, Bck4 /* xor SHR(w[i-2], 10)*/ /* xor accu == sigma1(w[i-15]) */
+ add Accu1, XAccu1
+ adc Accu2, XAccu2
+ adc Accu3, XAccu3
+ adc Accu4, XAccu4 /* finished with sigma0 */
+ /* now let's store the shit */
+ st X+, Accu1
+ st X+, Accu2
+ st X+, Accu3
+ st X+, Accu4
+ dec LoopC
+ breq 3f ; skip if zero
+ rjmp sha256_nextBlock_wcalcloop
+3:
+ /* we are finished with w array X points one byte post w */
+/* init a array */
+ pop r31
+ pop r30
+ push r30
+ push r31
+ ldi r25, 8*4 /* 8 32-bit values to copy from ctx to a array */
+init_a_array:
+ ld r1, Z+
+ st X+, r1
+ dec r25
+ brne init_a_array
+
+/* now the real fun begins */
+/* for (i=0; i<64; ++i){
+ t1 = a[7] + SIGMA1(a[4]) + CH(a[4],a[5],a[6]) + k[i] + w[i];
+ t2 = SIGMA0(a[0]) + MAJ(a[0],a[1],a[2]);
+ memmove(&(a[1]), &(a[0]), 7*4); // a[7]=a[6]; a[6]=a[5]; a[5]=a[4]; a[4]=a[3]; a[3]=a[2]; a[2]=a[1]; a[1]=a[0];
+ a[4] += t1;
+ a[0] = t1 + t2;
+ } */
+ /* Y points to a[0], Z ('cause lpm wants it) points to k[i], X points to w[i] */
+ sbiw r26, 8*4 /* X still points at a[7]+1*/
+ movw r28, r26
+ ldi r30, lo8(sha256_kv)
+ ldi r31, hi8(sha256_kv)
+ dec r27 /* X - (64*4 == 256) */
+ ldi r25, 64
+ mov LoopC, r25
+sha256_main_loop:
+ /* now calculate t1 */
+ /*CH(x,y,z) = (x&y)^((~x)&z)*/
+ ldd T1, Y+5*4
+ ldd T2, Y+5*4+1
+ ldd T3, Y+5*4+2
+ ldd T4, Y+5*4+3 /* y in T */
+ ldd Func1, Y+4*4
+ ldd Func2, Y+4*4+1
+ ldd Func3, Y+4*4+2
+ ldd Func4, Y+4*4+3 /* x in Func */
+ ldd Bck1, Y+6*4
+ ldd Bck2, Y+6*4+1
+ ldd Bck3, Y+6*4+2
+ ldd Bck4, Y+6*4+3 /* z in Bck */
+ and T1, Func1
+ and T2, Func2
+ and T3, Func3
+ and T4, Func4
+ com Func1
+ com Func2
+ com Func3
+ com Func4
+ and Bck1, Func1
+ and Bck2, Func2
+ and Bck3, Func3
+ and Bck4, Func4
+ eor T1, Bck1
+ eor T2, Bck2
+ eor T3, Bck3
+ eor T4, Bck4 /* done, CH(x,y,z) is in T */
+ /* now SIGMA1(a[4]) */
+ ldd Bck4, Y+4*4 /* think about using it from Func reg above*/
+ ldd Bck1, Y+4*4+1
+ ldd Bck2, Y+4*4+2
+ ldd Bck3, Y+4*4+3 /* load prerotate by 8-bit */
+ movw Func1, Bck1
+ movw Func3, Bck3
+ ldi r20, 2
+ rcall bitrotl /* rotr(x,6) */
+ movw XAccu1, Func1
+ movw XAccu3, Func3
+ movw Func1, Bck1
+ movw Func3, Bck3
+ ldi r20, 3
+ rcall bitrotr /* rotr(x,11) */
+ eor XAccu1, Func1
+ eor XAccu2, Func2
+ eor XAccu3, Func3
+ eor XAccu4, Func4
+ movw Func1, Bck3 /* this prerotates furteh 16 bits*/
+ movw Func3, Bck1 /* so we have now prerotated by 24 bits*/
+ ldi r20, 1
+ rcall bitrotr /* rotr(x,11) */
+ eor XAccu1, Func1
+ eor XAccu2, Func2
+ eor XAccu3, Func3
+ eor XAccu4, Func4 /* finished with SIGMA1, add it to T */
+ add T1, XAccu1
+ adc T2, XAccu2
+ adc T3, XAccu3
+ adc T4, XAccu4
+ /* now we've to add a[7], w[i] and k[i] */
+ ldd XAccu1, Y+4*7
+ ldd XAccu2, Y+4*7+1
+ ldd XAccu3, Y+4*7+2
+ ldd XAccu4, Y+4*7+3
+ add T1, XAccu1
+ adc T2, XAccu2
+ adc T3, XAccu3
+ adc T4, XAccu4 /* add a[7] */
+ ld XAccu1, X+
+ ld XAccu2, X+
+ ld XAccu3, X+
+ ld XAccu4, X+
+ add T1, XAccu1
+ adc T2, XAccu2
+ adc T3, XAccu3
+ adc T4, XAccu4 /* add w[i] */
+ lpm XAccu1, Z+
+ lpm XAccu2, Z+
+ lpm XAccu3, Z+
+ lpm XAccu4, Z+
+ add T1, XAccu1
+ adc T2, XAccu2
+ adc T3, XAccu3
+ adc T4, XAccu4 /* add k[i] */ /* finished with t1 */
+ /*now t2 = SIGMA0(a[0]) + MAJ(a[0],a[1],a[2]) */ /*i did to much x86 asm, i always see 4 32bit regs*/
+ /* starting with MAJ(x,y,z) */
+ ldd Func1, Y+4*0+0
+ ldd Func2, Y+4*0+1
+ ldd Func3, Y+4*0+2
+ ldd Func4, Y+4*0+3 /* load x=a[0] */
+ ldd XAccu1, Y+4*1+0
+ ldd XAccu2, Y+4*1+1
+ ldd XAccu3, Y+4*1+2
+ ldd XAccu4, Y+4*1+3 /* load y=a[1] */
+ and XAccu1, Func1
+ and XAccu2, Func2
+ and XAccu3, Func3
+ and XAccu4, Func4 /* XAccu == (x & y) */
+ ldd Bck1, Y+4*2+0
+ ldd Bck2, Y+4*2+1
+ ldd Bck3, Y+4*2+2
+ ldd Bck4, Y+4*2+3 /* load z=a[2] */
+ and Func1, Bck1
+ and Func2, Bck2
+ and Func3, Bck3
+ and Func4, Bck4
+ eor XAccu1, Func1
+ eor XAccu2, Func2
+ eor XAccu3, Func3
+ eor XAccu4, Func4 /* XAccu == (x & y) ^ (x & z) */
+ ldd Func1, Y+4*1+0
+ ldd Func2, Y+4*1+1
+ ldd Func3, Y+4*1+2
+ ldd Func4, Y+4*1+3 /* load y=a[1] */
+ and Func1, Bck1
+ and Func2, Bck2
+ and Func3, Bck3
+ and Func4, Bck4
+ eor XAccu1, Func1
+ eor XAccu2, Func2
+ eor XAccu3, Func3
+ eor XAccu4, Func4 /* XAccu == Maj(x,y,z) == (x & y) ^ (x & z) ^ (y & z) */
+ /* SIGMA0(a[0]) */
+ ldd Bck1, Y+4*0+0 /* we should combine this with above */
+ ldd Bck2, Y+4*0+1
+ ldd Bck3, Y+4*0+2
+ ldd Bck4, Y+4*0+3
+ movw Func1, Bck1
+ movw Func3, Bck3
+ ldi r20, 2
+ rcall bitrotr
+ movw Accu1, Func1
+ movw Accu3, Func3 /* Accu = shr(a[0], 2) */
+ movw Func1, Bck3
+ movw Func3, Bck1 /* prerotate by 16 bits */
+ ldi r20, 3
+ rcall bitrotl
+ eor Accu1, Func1
+ eor Accu2, Func2
+ eor Accu3, Func3
+ eor Accu4, Func4 /* Accu ^= shr(a[0], 13) */
+ mov Func1, Bck4
+ mov Func2, Bck1
+ mov Func3, Bck2
+ mov Func4, Bck3 /* prerotate by 24 bits */
+ ldi r20, 2
+ rcall bitrotl
+ eor Accu1, Func1
+ eor Accu2, Func2
+ eor Accu3, Func3
+ eor Accu4, Func4 /* Accu ^= shr(a[0], 22) */
+ add Accu1, XAccu1 /* add previous result (MAJ)*/
+ adc Accu2, XAccu2
+ adc Accu3, XAccu3
+ adc Accu4, XAccu4
+ /* now we are finished with the computing stuff (t1 in T, t2 in Accu)*/
+ /* a[7]=a[6]; a[6]=a[5]; a[5]=a[4]; a[4]=a[3]; a[3]=a[2]; a[2]=a[1]; a[1]=a[0]; */
+
+ ldi r21, 7*4
+ adiw r28, 7*4
+a_shift_loop:
+ ld r25, -Y /* warning: this is PREdecrement */
+ std Y+4, r25
+ dec r21
+ brne a_shift_loop
+
+ ldd Bck1, Y+4*4+0
+ ldd Bck2, Y+4*4+1
+ ldd Bck3, Y+4*4+2
+ ldd Bck4, Y+4*4+3
+ add Bck1, T1
+ adc Bck2, T2
+ adc Bck3, T3
+ adc Bck4, T4
+ std Y+4*4+0, Bck1
+ std Y+4*4+1, Bck2
+ std Y+4*4+2, Bck3
+ std Y+4*4+3, Bck4
+ add Accu1, T1
+ adc Accu2, T2
+ adc Accu3, T3
+ adc Accu4, T4
+ std Y+4*0+0, Accu1
+ std Y+4*0+1, Accu2
+ std Y+4*0+2, Accu3
+ std Y+4*0+3, Accu4 /* a array updated */
+
+
+ dec LoopC
+ breq update_state
+ rjmp sha256_main_loop ;brne sha256_main_loop
+update_state:
+ /* update state */
+ /* pointers to state should still exist on the stack ;-) */
+ pop r31
+ pop r30
+ ldi r21, 8
+update_state_loop:
+ ldd Accu1, Z+0
+ ldd Accu2, Z+1
+ ldd Accu3, Z+2
+ ldd Accu4, Z+3
+ ld Func1, Y+
+ ld Func2, Y+
+ ld Func3, Y+
+ ld Func4, Y+
+ add Accu1, Func1
+ adc Accu2, Func2
+ adc Accu3, Func3
+ adc Accu4, Func4
+ st Z+, Accu1
+ st Z+, Accu2
+ st Z+, Accu3
+ st Z+, Accu4
+ dec r21
+ brne update_state_loop
+ /* now we just have to update the length */
+ adiw r30, 1 /* since we add 512, we can simply skip the LSB */
+ ldi r21, 2
+ ldi r22, 6
+ ld r20, Z
+ add r20, r21
+ st Z+, r20
+ clr r21
+sha256_nextBlock_fix_length:
+ brcc sha256_nextBlock_epilog
+ ld r20, Z
+ adc r20, r21
+ st Z+, r20
+ dec r22
+ brne sha256_nextBlock_fix_length
+
+; EPILOG
+sha256_nextBlock_epilog:
+/* now we should clean up the stack */
+
+ pop r21
+ pop r20
+ in r0, SREG
+ cli ; we want to be uninterrupted while updating SP
+ out SPL, r20
+ out SPH, r21
+ out SREG, r0
+
+ clr r1
+ pop r29
+ pop r28
+ pop r17
+ pop r16
+ pop r15
+ pop r14
+ pop r13
+ pop r12
+ pop r11
+ pop r10
+ pop r9
+ pop r8
+ pop r7
+ pop r6
+ pop r5
+ pop r4
+ ret
+
+sha256_kv: ; round-key-vector stored in ProgMem
+.word 0x2f98, 0x428a, 0x4491, 0x7137, 0xfbcf, 0xb5c0, 0xdba5, 0xe9b5, 0xc25b, 0x3956, 0x11f1, 0x59f1, 0x82a4, 0x923f, 0x5ed5, 0xab1c
+.word 0xaa98, 0xd807, 0x5b01, 0x1283, 0x85be, 0x2431, 0x7dc3, 0x550c, 0x5d74, 0x72be, 0xb1fe, 0x80de, 0x06a7, 0x9bdc, 0xf174, 0xc19b
+.word 0x69c1, 0xe49b, 0x4786, 0xefbe, 0x9dc6, 0x0fc1, 0xa1cc, 0x240c, 0x2c6f, 0x2de9, 0x84aa, 0x4a74, 0xa9dc, 0x5cb0, 0x88da, 0x76f9
+.word 0x5152, 0x983e, 0xc66d, 0xa831, 0x27c8, 0xb003, 0x7fc7, 0xbf59, 0x0bf3, 0xc6e0, 0x9147, 0xd5a7, 0x6351, 0x06ca, 0x2967, 0x1429
+.word 0x0a85, 0x27b7, 0x2138, 0x2e1b, 0x6dfc, 0x4d2c, 0x0d13, 0x5338, 0x7354, 0x650a, 0x0abb, 0x766a, 0xc92e, 0x81c2, 0x2c85, 0x9272
+.word 0xe8a1, 0xa2bf, 0x664b, 0xa81a, 0x8b70, 0xc24b, 0x51a3, 0xc76c, 0xe819, 0xd192, 0x0624, 0xd699, 0x3585, 0xf40e, 0xa070, 0x106a
+.word 0xc116, 0x19a4, 0x6c08, 0x1e37, 0x774c, 0x2748, 0xbcb5, 0x34b0, 0x0cb3, 0x391c, 0xaa4a, 0x4ed8, 0xca4f, 0x5b9c, 0x6ff3, 0x682e
+.word 0x82ee, 0x748f, 0x636f, 0x78a5, 0x7814, 0x84c8, 0x0208, 0x8cc7, 0xfffa, 0x90be, 0x6ceb, 0xa450, 0xa3f7, 0xbef9, 0x78f2, 0xc671
+
+
+;###########################################################
+
+.global sha256_init
+;uint32_t sha256_init_vector[]={
+; 0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A,
+; 0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19 };
+;
+;void sha256_init(sha256_ctx_t *state){
+; state->length=0;
+; memcpy(state->h, sha256_init_vector, 8*4);
+;}
+; param1: (r23,r24) 16-bit pointer to sha256_ctx_t struct in ram
+; modifys: Z(r30,r31), Func1, r22
+sha256_init:
+ movw r26, r24 ; (24,25) --> (26,27) load X with param1
+ ldi r30, lo8((sha256_init_vector))
+ ldi r31, hi8((sha256_init_vector))
+ ldi r22, 32+8
+sha256_init_vloop:
+ lpm r23, Z+
+ st X+, r23
+ dec r22
+ brne sha256_init_vloop
+ ret
+
+sha256_init_vector:
+.word 0xE667, 0x6A09
+.word 0xAE85, 0xBB67
+.word 0xF372, 0x3C6E
+.word 0xF53A, 0xA54F
+.word 0x527F, 0x510E
+.word 0x688C, 0x9B05
+.word 0xD9AB, 0x1F83
+.word 0xCD19, 0x5BE0
+.word 0x0000, 0x0000
+.word 0x0000, 0x0000
+
+;###########################################################
+
+.global rotl32
+; === ROTL32 ===
+; function that rotates a 32 bit word to the left
+; param1: the 32-bit word to rotate
+; given in r25,r24,r23,r22 (r25 is most significant)
+; param2: an 8-bit value telling how often to rotate
+; given in r20
+; modifys: r21, r22
+rotl32:
+ cpi r20, 8
+ brlo bitrotl
+ mov r21, r25
+ mov r25, r24
+ mov r24, r23
+ mov r23, r22
+ mov r22, r21
+ subi r20, 8
+ rjmp rotl32
+bitrotl:
+ clr r21
+ clc
+bitrotl_loop:
+ tst r20
+ breq fixrotl
+ rol r22
+ rol r23
+ rol r24
+ rol r25
+ rol r21
+ dec r20
+ rjmp bitrotl_loop
+fixrotl:
+ or r22, r21
+ ret
+
+
+;###########################################################
+
+.global rotr32
+; === ROTR32 ===
+; function that rotates a 32 bit word to the right
+; param1: the 32-bit word to rotate
+; given in r25,r24,r23,22 (r25 is most significant)
+; param2: an 8-bit value telling how often to rotate
+; given in r20
+; modifys: r21, r22
+rotr32:
+ cpi r20, 8
+ brlo bitrotr
+ mov r21, r22
+ mov r22, r23
+ mov r23, r24
+ mov r24, r25
+ mov r25, r21
+ subi r20, 8
+ rjmp rotr32
+bitrotr:
+ clr r21
+ clc
+bitrotr_loop:
+ tst r20
+ breq fixrotr
+ ror r25
+ ror r24
+ ror r23
+ ror r22
+ ror r21
+ dec r20
+ rjmp bitrotr_loop
+fixrotr:
+ or r25, r21
+ ret
+
+
+;###########################################################
+
+.global change_endian32
+; === change_endian32 ===
+; function that changes the endianess of a 32-bit word
+; param1: the 32-bit word
+; given in r25,r24,r23,22 (r25 is most significant)
+; modifys: r21, r22
+change_endian32:
+ movw r20, r22 ; (r22,r23) --> (r20,r21)
+ mov r22, r25
+ mov r23, r24
+ mov r24, r21
+ mov r25, r20
+ ret
+