#20991 closed defect (fixed)

hs: Move end_sig_fields field of ESTABLISH_INTRO v3 cell before sig_len

Reported by: dgoulet Owned by: dgoulet
Priority: Medium Milestone: Tor: 0.3.0.x-final
Component: Core Tor/Tor Version:
Severity: Normal Keywords: tor-hs, prop224, review-group-14
Cc: Actual Points:
Parent ID: #17241 Points:
Reviewer: Sponsor: SponsorR-must

Description

There's a spec bug that we need to fix, which goes along with a corresponding trunnel bug: The @ptr end_sig_fields marker in hs_cell_establish_intro needs to be before sig_len. Otherwise we can't handle future signature schemes where sig_len is not determined until the signature is generated.

Child Tickets

Change History (8)

comment:1 Changed 23 months ago by dgoulet

Summary: hs: Move sig_len field of ESTABLISH_INTRO v3 cell after sighs: Move end_sig_fields field of ESTABLISH_INTRO v3 cell before sig_len

comment:2 Changed 23 months ago by dgoulet

This ticket will also upgrade the HS trunnel definition to 1.5.1.

comment:3 Changed 23 months ago by dgoulet

Owner: set to dgoulet
Status: newaccepted

comment:4 Changed 23 months ago by teor

Does this mean that the signature length is unauthenticated?
This could be problematic - are we sure there are no security implications?

comment:5 Changed 23 months ago by nickm

The signature length IMO _needs_ to be unauthenticated, or we'll never be able to support another signature method with variable-length keys.

Imagine if we were doing this with RSA instead of Ed25519: If we had to authenticate the signature length, we couldn't actually sign anything reliably, since we wouldn't know the signature length for sure until we had made the signature.

comment:6 Changed 23 months ago by dgoulet

Status: acceptedneeds_review

See branch bug20991_030_01

comment:7 Changed 22 months ago by nickm

Keywords: review-group-14 added

comment:8 Changed 22 months ago by nickm

Resolution: fixed
Status: needs_reviewclosed

Looks good. Merged this!

Note: See TracTickets for help on using tickets.