Opened 3 years ago

Closed 3 years ago

#23303 closed defect (fixed)

hs: Explain why we reset the directory connection timestamp client side

Reported by: dgoulet Owned by: dgoulet
Priority: Medium Milestone: Tor: 0.3.2.x-final
Component: Core Tor/Tor Version:
Severity: Normal Keywords: tor-hs, prop224
Cc: mikeperry Actual Points:
Parent ID: #23300 Points:
Reviewer: Sponsor: SponsorR-can


In the v2 HS subsystem we have this in rendclient.c in rend_client_desc_trynow():

      /* restart their timeout values, so they get a fair shake at
       * connecting to the hidden service. */
      base_conn->timestamp_created = now;
      base_conn->timestamp_lastread = now;
      base_conn->timestamp_lastwritten = now;

We've taken the same thing into the v3 client in #17242 but we need a comment that explains *why* we do that because get a fair shake is not very helpful :).

Child Tickets

Change History (4)

comment:1 Changed 3 years ago by dgoulet

Owner: set to dgoulet
Status: newaccepted

comment:2 Changed 3 years ago by mikeperry

The comment and the associated code were Roger's and Nick's from 2005 and 2011, respectively.. But the effect of the lines is that this circuit is now considered fresh by the various timeout things that care. Since this is on an entry conn (and not a circuit), I'm guessing that the applicable timeouts would be stream and general connection timeouts. I've not worked on those. I've only worked on circuit and orconn timeouts.

comment:3 Changed 3 years ago by dgoulet

Status: acceptedneeds_review

Thanks Mike! I think I get it and it makes sense.

Branch: ticket23303_032_01.

comment:4 Changed 3 years ago by nickm

Resolution: fixed
Status: needs_reviewclosed


Note: See TracTickets for help on using tickets.