GitLab

Off-by-2 error counting to MAX_REND_FAILURES for onion services:

In can_relaunch_service_rendezvous_point(), we check

  if (circ->build_state->failure_count > MAX_REND_FAILURES ||
      circ->build_state->expiry_time <= time(NULL)) {

to decide whether to abort the relaunch.

But the incrementing of failure_count happens in the relaunch, i.e. after this code. Also, it says ">" rather than ">=".

Yet the definition of MAX_REND_FAILURES is

/** How many times will a hidden service operator attempt to connect to a
 * requested rendezvous point before giving up? */
#define MAX_REND_FAILURES 1

So when the first attempt fails, failure_count is 0, which is not > 1, so we try again. When the second attempt fails, failure_count is 1, which is also not > 1. It's only after the third attempt fails that we decide that MAX_REND_FAILURES has been reached.

This bug affects legacy onion services, and it will also affect nextgen onion services once #24894 (moved) is fixed.