Re: tor 0.1.2.17 server died after only a few hours

[Scott Bennett <bennett@xxxxxxxxxx>]

     On Thu, 6 Sep 2007 12:32:17 -0400 Nick Mathewson <nickm@xxxxxxxxxxxxx>
wrote:

>On Wed, Sep 05, 2007 at 02:28:27AM -0500, Scott Bennett wrote:
>> >     I wrote:
>> >>     I switched from 0.1.2.16 to 0.1.2.17 this morning.  About 5 hrs. 2=
>0-odd
>> >>minutes later, tor wrote the following message to the notice log and ex=
>ited:
>> >>
>> >>Sep 04 14:06:29.140 [err] connection.c:2387: assert_connection_ok: Asse=
>rtion connection_is_writing(conn) || conn->wants_to_write || (conn->type =
>=3D=3D CONN_TYPE_DIR && TO_DIR_CONN(conn)->is_blocked_on_or_conn) failed; a=
>borting.
>> >>
>> >>Does anyone know what caused this?  I can still switch back to 0.1.2.16=
> if
>> >>need be.
>> >
>> >     It just did it again after being up only a few minutes.  I've now r=
>everted
>> >to 0.1.2.16 and will stay with that until the next release comes out.  S=
>igh.
>>=20
>>      I can't believe this!  My 0.1.2.16 server just did the same thing!  =
>And
>> that is after it ran fine for quite some time before trying 0.1.2.17 (thi=
>s is
>> the first time 0.1.2.16 has acted up).  Here's the message:
>>=20
>> Sep 05 02:09:33.960 [err] connection.c:2391: assert_connection_ok:
>> Assertion connection_is_writing(conn) || conn->wants_to_write ||
>> (conn->type =3D=3D CONN_TYPE_DIR &&
>> TO_DIR_CONN(conn)->is_blocked_on_or_conn) failed; aborting.
>
>On further consideration, this looks exactly like Bug 406 on the bug
>tracker.
>
>  ( http://bugs.noreply.org/flyspray/index.php?do=3Ddetails&id=3D406 )
>
>This is a tricky one; something to do with the way we handle "linked"[*]
>directory connections on 0.1.2.x servers seems to be broken; I'll try
>to track it down.  (It seems like we're mis-managing the
>blocked_on_or_conn field somehow; see bug report for more details.)

     Thanks very much, Nick.  A core file is still available if you
need it, but now that 0.1.2.16 has blown up, the file will be from that
version rather than the newer one.  I'll hang onto it for a week or two
in case you want it.
>
>As a workaround, I believe that this bug doesn't appear in the
>0.2.0.x-alpha series so far; if you want to try one of those, it may

     That's interesting.  Would a comparison of the differences between
0.2.* and 0.1.2.1[67] in that section of code reveal any clues then?

>work better for you.  On the other hand, it might do worse; it _is_ an
>alpha after all. :/

     Actually, it's a beta, regardless of how it's labeled.  Alphas are
only tested in house.  Betas are released to the brave and/or foolish in
the outside user community.  Either way, your point is well taken and is
how I tend to see that sort of thing, too.  It's a gamble.  I spend enough
time babysitting the interaction of tor and my flaky-by-ISP's-design ADSL
link that I've been avoiding the "development" releases so far.  However,
if the "stable" releases are going to be acting up like this, it may not
make a difference, i.e., similar time commitments for the two lines of
released versions may be required.
>
>[*] (We used to call these internal directory connections "bridged"
>  instead of "linked", but we changed the word when we introduced
>  "bridge servers" in 0.2.0.x, because we wanted to avoid confusing
>  people.)
>
     Ah.  Thanks for that tidbit. :-)


                                  Scott Bennett, Comm. ASMELG, CFIAG
**********************************************************************
* Internet:       bennett at cs.niu.edu                              *
*--------------------------------------------------------------------*
* "A well regulated and disciplined militia, is at all times a good  *
* objection to the introduction of that bane of all free governments *
* -- a standing army."                                               *
*    -- Gov. John Hancock, New York Journal, 28 January 1790         *
**********************************************************************