core/vm: implement EIP-3540: EVM Object Format (EOF) v1 #22958

gumb0 · 2021-05-27T16:37:40Z

This is PoC for EIP-3540 implementation.

Some consensus tests are at ethereum/tests#847

core/vm/eof.go

core/vm/evm.go

axic · 2021-11-25T00:49:37Z

core/vm/contract.go

+		return 0
+	}
+	// For EOF contracts running out of code section means abort execution with failure
+	return 0xfe


Nice trick.

However isn't there an enum for opcodes? If so, it would make sense using it for both 0 and 0xfe so it is more apparent to the reader.

Hm, the current 0 returned is more in line with the concept of over-reading just returns zeroes (in case of codecopy/calldatacopy), and on that vein 0xfe is indeed a "hack".

However isn't there an enum for opcodes? If so, it would make sense using it for both 0 and 0xfe so it is more apparent to the reader.

No enum for 0xfe yet 😃 Also this function returns byte that's why it used 0 I think.
However it doesn't make sense to return byte because it's always casted to OpCode in GetOp above.
I guess I'll submit a separate small cleanup PR for this.

For anyone interested, see #23974 and #24017.

axic · 2021-11-29T13:43:40Z

core/vm/contract.go

 		return OpCode(c.Code[n])
 	}

-	return STOP
+	// For legacy contracts running out of code section means STOP
+	if c.header.codeSize == 0 {


I'd say a helper for this would be nice, otherwise just leave a comment stating that based EIP-3540 codeSize == 0 is invalid, i.e. it must be legacy.

Update: a helper would definitely be nice since this check is used in several places.

holiman · 2022-09-26T16:45:47Z

core/vm/analysis.go

-		op := OpCode(code[pc])
+func codeBitmapInternal(c *Contract, bits bitvec) bitvec {
+	for pc := uint64(0); pc < c.CodeSize(); {
+		op := OpCode(c.Code[c.CodeBeginOffset()+pc])


We def want to be sure that c.CodeBeginOffset() is not actually called every iteration, but only once. Perhaps better to move that out of the loop, to be sure in the future too?

holiman · 2022-10-19T08:34:58Z

core/vm/contract.go

+	return c.header.CodeEndOffset()
+}
+
+func (c *Contract) CodeSize() uint64 {


Docstring. This is actually not just a nit-pick, one pretty important thing to mention here is whether CodeSize returns the size of header+code, the entire container, or something else

holiman · 2022-10-19T08:41:08Z

core/vm/contract.go

 		return false
 	}
 	// Only JUMPDESTs allowed for destinations
-	if OpCode(c.Code[udest]) != JUMPDEST {
+	if OpCode(c.Code[c.CodeBeginOffset()+udest]) != JUMPDEST {


Maybe something like this, and internalize the call to CodeBeginOffset?

Suggested change

if OpCode(c.Code[c.CodeBeginOffset()+udest]) != JUMPDEST {

if c.OpCodeAt(udest) != JUMPDEST {

I went with @axic's suggestion below and used existing GetOp for this.

holiman · 2022-10-19T08:45:06Z

core/vm/evm.go

+		code := evm.StateDB.GetCode(addrCopy)
+
+		var header EOF1Header
+		if hasEIP3540(&evm.Config) && hasEOFMagic(code) {


This looks a bit odd to me. Shouldn't you instead do something like

if evm.chainRules.IsEip3540 && hasEOFMagic(code) {

?

This was implemented to make EIP-3540 (and EIP-3670) only activatable as an "extra" EIP (list of these is in evm.Config.ExtraEips)

If you think it's okay now to make it part of Shanghai implementation, I'll be happy to change it to evm.chainRules.IsShanghai, this would actually simplify some things I believe.

holiman · 2022-10-19T08:46:39Z

core/vm/interpreter.go

@@ -184,7 +184,7 @@ func (in *EVMInterpreter) Run(contract *Contract, input []byte, readOnly bool) (
 		}
 		// Get the operation from the jump table and validate the stack to ensure there are
 		// enough stack items available to perform the operation.
-		op = contract.GetOp(pc)
+		op = contract.GetOp(contract.CodeBeginOffset() + pc)


Could also use that contract.OpCodeAt(udest) I suggested above

Could also just change GetOp to the new semantics, couldn't it? It seems to be the only place it is being used at (and has internal length checks).

gumb0 · 2022-11-08T18:03:50Z

d5cf7a2 is a refactoring to save code section bounds instead of EOF1Header in Contract to have fewer calculations and branching on the hot paths in interpreter loop.

gumb0 · 2022-11-08T18:35:18Z

Renamed Contract.Code to Contract.Container to use terminology from the EIP and disambiguate from mentions of code section.

gumb0 · 2022-11-08T18:36:06Z

core/vm/contract.go

 	jumpdests map[common.Hash]bitvec // Aggregated result of JUMPDEST analysis.
 	analysis  bitvec                 // Locally cached result of JUMPDEST analysis

-	Code     []byte
-	CodeHash common.Hash
-	CodeAddr *common.Address


CodeAddr was not used.

gumb0 · 2022-12-29T11:59:43Z

Outdated and superseded by #26133

gumb0 changed the title ~~Eip 3540~~ Implement EIP-3540: EVM Object Format (EOF) v1 May 27, 2021

gumb0 mentioned this pull request May 27, 2021

Tests for EIP-3540: EVM Object Format (EOF) v1 ethereum/tests#847

Merged

gumb0 changed the title ~~Implement EIP-3540: EVM Object Format (EOF) v1~~ core/vm: implement EIP-3540: EVM Object Format (EOF) v1 May 27, 2021

gumb0 force-pushed the eip-3540 branch 2 times, most recently from 036cbe7 to 0695098 Compare May 27, 2021 17:02

axic mentioned this pull request May 28, 2021

Proposal to add EIP-3540 to Shanghai ethereum/pm#329

Closed

gumb0 force-pushed the eip-3540 branch from d601654 to 1197b69 Compare July 27, 2021 10:04

gumb0 force-pushed the eip-3540 branch 2 times, most recently from b7a2fad to 679ba38 Compare August 18, 2021 17:00

gumb0 force-pushed the eip-3540 branch from 679ba38 to d0127f6 Compare August 19, 2021 15:48

gumb0 force-pushed the eip-3540 branch 2 times, most recently from 69b1645 to 5e765d7 Compare September 28, 2021 18:29

gumb0 force-pushed the eip-3540 branch 2 times, most recently from eff336d to 10d7532 Compare October 14, 2021 14:21

jwasinger reviewed Oct 14, 2021

View reviewed changes

core/vm/eof.go Outdated Show resolved Hide resolved

gumb0 force-pushed the eip-3540 branch from 10d7532 to b42507c Compare October 14, 2021 19:44

gumb0 commented Nov 9, 2021

View reviewed changes

core/vm/evm.go Outdated Show resolved Hide resolved

gumb0 commented Nov 9, 2021

View reviewed changes

core/vm/evm.go Outdated Show resolved Hide resolved

gumb0 force-pushed the eip-3540 branch 4 times, most recently from 77fa3e7 to 35e7209 Compare November 24, 2021 12:14

axic reviewed Nov 25, 2021

View reviewed changes

gumb0 force-pushed the eip-3540 branch from 35e7209 to a7470f6 Compare November 26, 2021 12:38

axic reviewed Nov 29, 2021

View reviewed changes

gumb0 force-pushed the eip-3540 branch 2 times, most recently from b5c39ef to 2b311d7 Compare November 29, 2021 18:50

gumb0 mentioned this pull request Nov 30, 2021

core/vm: Make INVALID a defined opcode #24017

Merged

gumb0 force-pushed the eip-3540 branch from 2b311d7 to 89d74fb Compare December 9, 2021 13:52

gumb0 mentioned this pull request Dec 9, 2021

core/vm: implement EIP-3670: EOF - Code Validation #24090

Closed

holiman added the maybe shanghai label Sep 26, 2022

holiman reviewed Sep 26, 2022

View reviewed changes

gumb0 force-pushed the eip-3540 branch 2 times, most recently from 71fb6e7 to b5013e1 Compare September 29, 2022 10:24

holiman reviewed Oct 19, 2022

View reviewed changes

axic force-pushed the eip-3540 branch from b5013e1 to c1a0d94 Compare October 31, 2022 16:46

gumb0 and others added 8 commits November 7, 2022 17:07

core/vm: Add EIP-3540 to supported extra EIPs

1fa49ea

core/vm: Add validation of deployed code format according to EIP-3540

00b1256

core/vm: Add validation of init code according to EIP-3540

8710a78

core/vm: EOF1 validation unit tests

bbe63e3

core/vm: Make PC relative to code section for EOF1 code

0ed7b04

core/vm: Limit JUMPDEST analysis only to code section of EOF

422b0ea

Add docstring to CodeSize

933c304

Make GetOp take offset inside code and use it for validJumpdest

4913c1d

gumb0 force-pushed the eip-3540 branch from c1a0d94 to 4913c1d Compare November 8, 2022 01:48

gumb0 commented Nov 8, 2022

View reviewed changes

gumb0 force-pushed the eip-3540 branch 3 times, most recently from 1770a8e to 55cbab6 Compare November 8, 2022 19:05

gumb0 added 2 commits November 8, 2022 15:49

Save code section bounds in Contract instead of EOF header

d5cf7a2

core/vm: Rename Contract.Code => Contract.Container

8adc30c

gumb0 force-pushed the eip-3540 branch from 55cbab6 to 8adc30c Compare November 8, 2022 21:50

gumb0 marked this pull request as ready for review November 8, 2022 22:59

gumb0 requested review from s1na, karalabe and rjl493456442 as code owners November 8, 2022 22:59

gumb0 closed this Dec 29, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

core/vm: implement EIP-3540: EVM Object Format (EOF) v1 #22958

core/vm: implement EIP-3540: EVM Object Format (EOF) v1 #22958

gumb0 commented May 27, 2021 •

edited

Loading

axic Nov 25, 2021

axic Nov 25, 2021 •

edited

Loading

gumb0 Nov 25, 2021

axic Nov 30, 2021

axic Nov 29, 2021 •

edited

Loading

holiman Sep 26, 2022

gumb0 Sep 28, 2022

holiman Oct 19, 2022

axic Oct 31, 2022

holiman Oct 19, 2022

gumb0 Nov 8, 2022

holiman Oct 19, 2022

gumb0 Nov 7, 2022

holiman Oct 19, 2022

axic Oct 31, 2022

gumb0 commented Nov 8, 2022 •

edited

Loading

gumb0 commented Nov 8, 2022

gumb0 Nov 8, 2022 •

edited

Loading

gumb0 commented Dec 29, 2022

	if OpCode(c.Code[c.CodeBeginOffset()+udest]) != JUMPDEST {
	if c.OpCodeAt(udest) != JUMPDEST {

core/vm: implement EIP-3540: EVM Object Format (EOF) v1 #22958

core/vm: implement EIP-3540: EVM Object Format (EOF) v1 #22958

Conversation

gumb0 commented May 27, 2021 • edited Loading

Choose a reason for hiding this comment

axic Nov 25, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

axic Nov 29, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gumb0 commented Nov 8, 2022 • edited Loading

gumb0 commented Nov 8, 2022

gumb0 Nov 8, 2022 • edited Loading

Choose a reason for hiding this comment

gumb0 commented Dec 29, 2022

gumb0 commented May 27, 2021 •

edited

Loading

axic Nov 25, 2021 •

edited

Loading

axic Nov 29, 2021 •

edited

Loading

gumb0 commented Nov 8, 2022 •

edited

Loading

gumb0 Nov 8, 2022 •

edited

Loading