fix(MMU_BLAKE): add test + debug #1519
Conversation
Signed-off-by: F Bojarski <ceciestunepoubelle@protonmail.ch>
Signed-off-by: F Bojarski <ceciestunepoubelle@protonmail.ch>
letypequividelespoubelles
changed the title
Signed-off-by: F Bojarski <ceciestunepoubelle@protonmail.ch>
Signed-off-by: F Bojarski <ceciestunepoubelle@protonmail.ch>
| final Bytes bytecode = | ||
| BytecodeCompiler.newProgram() | ||
| .push(Bytes.fromHexString("0x0badb077")) // value, some random data to hash | ||
| .push(5) // offset |
There was a problem hiding this comment.
What is the intent here. You are producing this in memory:
<-- 5 bytes --> <-- 32 bytes -->
00 00 00 00 00 | 00 00 00 ... 00 0b ad b0 77
There was a problem hiding this comment.
yes, I just want to have something (ie not 0) in the hash input data, so really don't care of what I put inside as soon as we have a non zero between offset 4 and 212.
| .push(5) // offset | ||
| .op(OpCode.MSTORE) | ||
| .push(10) // value = r * 256 ** 3 for Blake call | ||
| .push(0) // offset |
There was a problem hiding this comment.
What is the intent here. You are producing this in memory:
And now
<-- 5 bytes --> <-- 32 bytes -->
0a 00 00 00 00 | 00 00 00 ... 00 0b ad b0 77
So you have r = 0x 0a 00 00 00 = 10 * 256 ** 3 rounds (and gas cost.)
There was a problem hiding this comment.
yes, I want to produce an OOG (wih the MSTORE8) and just have a non zero input for the hash (same comment as before). It's true we don't care about the input of the hash as we're OOG, but still
| final Bytes bytecode = | ||
| BytecodeCompiler.newProgram() | ||
| .push(Bytes.fromHexString("0x0badb077")) // value, some random data to hash | ||
| .push(5) // offset |
| BytecodeCompiler.newProgram() | ||
| .push(Bytes.fromHexString("0x0badb077")) // value, some random data to hash | ||
| .push(5) // offset | ||
| .op(OpCode.MSTORE) |
There was a problem hiding this comment.
I really don't understand the 5 here. Is it to produce some gibberish value ? Don't get me wrong, that'll do the job, it's just surprising.
There was a problem hiding this comment.
I put a 5 just to be sure it doesn't overlap with the 4 first bytes of the r. But yes the value will be leftPadded in the memory, but I don't care.
| @@ -148,6 +149,7 @@ public MmuData setMicroInstructions(MmuData mmuData) { | |||
| .sourceLimbOffset(sourceLimbOffsetF) | |||
| .sourceByteOffset(sourceByteOffsetF) | |||
| .targetByteOffset((short) (LLARGE - 1)) | |||
| .targetLimbOffset(1) | |||
| .push(2) // value = f for Blake call | ||
| .push(212) // offset | ||
| .op(OpCode.MSTORE8) | ||
| .push(LLARGE * (INDEX_MAX_BLAKE_RESULT + 1)) // return size |
There was a problem hiding this comment.
You actually have BLAKE2f_HASH_OUTPUT_SIZE in the code elsewhere.
There was a problem hiding this comment.
oh yes, forget to make the change everywhere
There was a problem hiding this comment.
done
| @@ -587,7 +591,7 @@ public static MmuCall parameterExtractionForBlake( | |||
| .successBit(blakeSuccess) | |||
| .limb1(blakeR) | |||
| .limb2(blakeF) | |||
| .setBlakeModexp() | |||
| .setBlakeModexp(blakeSuccess) | |||
There was a problem hiding this comment.
this one was easy ;)
| @@ -123,9 +124,9 @@ private void traceParameter(Trace trace, UnsignedByte stamp) { | |||
| } | |||
|
|
|||
| private void traceBlakeResult(Trace trace, UnsignedByte stamp) { | |||
| final Bytes input = computeBlake2fResult(); | |||
| final Bytes hash = computeBlake2fResult(); | |||
There was a problem hiding this comment.
Was that a bug ? or just overloading the word input ?
There was a problem hiding this comment.
just renaming. Everything was named input before (same name for callData, hashInput, and hashOutput ...)
…ithub.com:Consensys/linea-tracer into 1517-wrong-calculation-for-numbermmioinstructions
Signed-off-by: F Bojarski <ceciestunepoubelle@protonmail.ch>
…ithub.com:Consensys/linea-tracer into 1517-wrong-calculation-for-numbermmioinstructions
letypequividelespoubelles
deleted the
1517-wrong-calculation-for-numbermmioinstructions
branch
No description provided.