Skip to content

Argon2 on x86-64: prove filling columns and block addresses - #480

Closed
reaperhulk wants to merge 4 commits into
argon2-reference-mapfrom
argon2-fill-addresses
Closed

reaperhulk wants to merge 4 commits into
argon2-reference-mapfrom
argon2-fill-addresses

Conversation

@reaperhulk

Copy link
Copy Markdown
Member

Compute the current filling column, its cyclic predecessor, and lane-major block pointers in baseline x86-64 code. The predecessor wraps to the last column at column zero. Reference coordinates may be secret: pointer arithmetic has a fixed trace, and the column branch depends only on public loop coordinates.

Prove the exact natural-number formulas, fault-free execution, memory and permission preservation, enclosing loop register preservation, and MXCSR preservation. Target-independent bounds establish that the current, previous and reference cells are in the matrix; cell_contains covers each complete 1024-byte block. Checked literals support later artifact checks.

This continues the whole-derive implementation above #479. These are inline helpers, with no new API, specification, TCB changes or emitted assembly. The filling operation and complete derive proof remain to be assembled.

Validation: targeted Lean builds of the changed proof modules, standard-axiom and compiler-override audits, specification-origin audit, Lean import and proof-speed checks, and git diff --check. Full checks run in CI.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

1 participant