Difference between revisions of "83Plus:BCALLs:8018"
m (→Example) |
(→Comments: a slightly troubling bug) |
||
| Line 22: | Line 22: | ||
== Comments == | == Comments == | ||
This B_CALL "completes" a hash started by [[83Plus:BCALLs:808D|MD5Init]] and [[83Plus:BCALLs:8090|MD5Update]]. | This B_CALL "completes" a hash started by [[83Plus:BCALLs:808D|MD5Init]] and [[83Plus:BCALLs:8090|MD5Update]]. | ||
| + | |||
| + | '''Warning:''' Early versions of this B_CALL contain a bug, such that computing the hash of a string with length congruent to 55 mod 64 gives an incorrect result. (Rather than adding only one byte of padding, it adds a full 65 bytes, the last of which is E5 rather than zero.) | ||
| + | |||
| + | This is most likely not a security hole if you are simply using MD5 as a black-box hash function, and you don't assume the result on one calculator will match that on another. | ||
| + | |||
| + | If you need the actual, RFC-compliant MD5 hash, you can add the padding and length bytes yourself using MD5Update. (Leaving off the padding and length altogether would be a significant security hole; and in any case, MD5Update only updates the hash every 64 bytes.) See below for a simple implementation of this. | ||
| + | |||
| + | Boot code versions known to contain this bug: | ||
| + | * TI-73 1.3005 | ||
| + | * TI-83 Plus BE 1.00, 1.01 | ||
| + | |||
| + | Boot code versions known not to contain this bug: | ||
| + | * TI-83 Plus SE 1.00 | ||
| + | * TI-84 Plus BE/SE 1.00, 1.02 | ||
| + | |||
| + | === Replacement code === | ||
| + | (uses OP1) | ||
| + | MD5Final: | ||
| + | ld hl,[[83Plus:RAM:8269|MD5Length]] | ||
| + | rst 20h | ||
| + | |||
| + | ld hl,MD5Final_Padding | ||
| + | ld bc,1 | ||
| + | B_CALL [[83Plus:BCALLs:8090|MD5Update]] | ||
| + | MD5Final_PadLoop: | ||
| + | ld a,(MD5Length) | ||
| + | cp 0c0h | ||
| + | jr z,MD5Final_24mod32 | ||
| + | MD5Final_ContinuePadding: | ||
| + | ld hl,MD5Final_Padding+1 | ||
| + | ld bc,1 | ||
| + | B_CALL MD5Update | ||
| + | jr MD5Final_PadLoop | ||
| + | MD5Final_24mod32: | ||
| + | ld a,(MD5Length+1) | ||
| + | and 1 | ||
| + | jr z,MD5Final_ContinuePadding | ||
| + | |||
| + | ld hl,OP1 | ||
| + | ld bc,8 | ||
| + | B_CALL MD5Update | ||
| + | ret | ||
| + | |||
| + | MD5Final_Padding: db 80h, 00h | ||
== Example == | == Example == | ||
Latest revision as of 14:12, 24 July 2007
Synopsis
Official Name: MD5Final
Other Name: FinishMD5
BCALL Address: 8018
Finishes an MD5 hash by adding padding and length bits.
Inputs
- 8269: (8 bytes) Holds the length in bits of the data hashed so far.
- 8292: (16 bytes) Current MD5 hash
Outputs
- 8292: (16 bytes) Completed MD5 hash
Destroys
- 8259: (16 bytes) Alternate registers (roughly equivalent to the AA, BB, CC, and DD registers specified in the MD5 standard.)
- 83A5: (64 bytes) Buffer holding data to be hashed.
Comments
This B_CALL "completes" a hash started by MD5Init and MD5Update.
Warning: Early versions of this B_CALL contain a bug, such that computing the hash of a string with length congruent to 55 mod 64 gives an incorrect result. (Rather than adding only one byte of padding, it adds a full 65 bytes, the last of which is E5 rather than zero.)
This is most likely not a security hole if you are simply using MD5 as a black-box hash function, and you don't assume the result on one calculator will match that on another.
If you need the actual, RFC-compliant MD5 hash, you can add the padding and length bytes yourself using MD5Update. (Leaving off the padding and length altogether would be a significant security hole; and in any case, MD5Update only updates the hash every 64 bytes.) See below for a simple implementation of this.
Boot code versions known to contain this bug:
- TI-73 1.3005
- TI-83 Plus BE 1.00, 1.01
Boot code versions known not to contain this bug:
- TI-83 Plus SE 1.00
- TI-84 Plus BE/SE 1.00, 1.02
Replacement code
(uses OP1)
MD5Final:
ld hl,MD5Length
rst 20h
ld hl,MD5Final_Padding
ld bc,1
B_CALL MD5Update
MD5Final_PadLoop:
ld a,(MD5Length)
cp 0c0h
jr z,MD5Final_24mod32
MD5Final_ContinuePadding:
ld hl,MD5Final_Padding+1
ld bc,1
B_CALL MD5Update
jr MD5Final_PadLoop
MD5Final_24mod32:
ld a,(MD5Length+1)
and 1
jr z,MD5Final_ContinuePadding
ld hl,OP1
ld bc,8
B_CALL MD5Update
ret
MD5Final_Padding: db 80h, 00h
Example
B_CALL MD5Init ld bc,14 ld hl,String B_CALL MD5Update B_CALL MD5Final ; should give: f96b697d7cb7938d525a2f31aaf161d0 String: .db "message digest"