cannon: Implement thread-safe ll and sc operations

[mbaxter]

Description

Modify the behavior of load linked word (ll) and store conditional word (sc) operations in MTCannon. These opcodes provide the primitives used to implement atomic read-modify-write (RMW) operations. Because MTCannon can context-switch between the ll and sc calls, we need track whether the memory location reserved via the ll call is modified before sc is called.

This PR adds several new global fields to the MT State to track the memory location reserved via ll. Only a single memory reservation can be active at a given time - a new reservation will clear any previous reservation. If the reserved memory location is modified, the memory reservation is cleared. When an sc operation is executed, it will now only succeed if the memory reservation is intact, otherwise sc returns a 0 for failure.

Tests

Added additional differential tests to check ll and sc behavior. Ran the new ll/sc tests against develop to validate that the behavior is unchanged for the single-threaded implementation. Also added tests to validate that the memory reservation is cleared as expected when memory is modified.

Metadata

Fixes #11757

[codecov]

Codecov Report

Attention: Patch coverage is 84.52381% with 26 lines in your changes missing coverage. Please review.

Project coverage is 79.42%. Comparing base (c8d6dbb) to head (4666ffc).
Report is 7 commits behind head on develop.

Files with missing lines	Patch %	Lines
cannon/mipsevm/multithreaded/state.go	47.05%	12 Missing and 6 partials ⚠️
cannon/mipsevm/multithreaded/mips.go	91.66%	4 Missing and 1 partial ⚠️
cannon/mipsevm/singlethreaded/mips.go	92.30%	1 Missing and 1 partial ⚠️
...non/mipsevm/multithreaded/testutil/expectations.go	93.33%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop   #11906      +/-   ##
===========================================
- Coverage    79.52%   79.42%   -0.10%     
===========================================
  Files           41       41              
  Lines         3287     3412     +125     
===========================================
+ Hits          2614     2710      +96     
- Misses         511      532      +21     
- Partials       162      170       +8     

Flag	Coverage Δ
cannon-go-tests	79.42% <84.52%> (-0.10%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
cannon/mipsevm/exec/mips_instructions.go	97.89% <100.00%> (+0.03%)	⬆️
cannon/mipsevm/exec/mips_syscalls.go	93.75% <100.00%> (+0.08%)	⬆️
cannon/mipsevm/multithreaded/testutil/mutators.go	71.42% <100.00%> (-5.32%)	⬇️
...non/mipsevm/multithreaded/testutil/expectations.go	97.22% <93.33%> (+0.12%)	⬆️
cannon/mipsevm/singlethreaded/mips.go	93.24% <92.30%> (-0.76%)	⬇️
cannon/mipsevm/multithreaded/mips.go	95.67% <91.66%> (-0.80%)	⬇️
cannon/mipsevm/multithreaded/state.go	53.67% <47.05%> (-2.78%)	⬇️

[Inphi]

The ll/sc changes to MIPS.sol looks good to me. We should be doubly sure we're not introducing any new behavior to that contract. So it'll be good to differentially fuzz both singlethreaded VMs for ll/sc to confirm that we're not introducing new behavior.

One way to do this is to save the bytecode of the current MIPS.sol implementation in develop. And write a test that differentially fuzzes ll/sc inputs on that old bytecode against the new singlethreaded Go VM.

[mbaxter]

We should be doubly sure we're not introducing any new behavior

To double-check, I ran the new ll/sc tests against develop - they are passing there without the restructuring introduced in this PR.

[Inphi]

LGTM!