[tor-bugs] #31683 [Core Tor/Tor]: md: Bad family line in cached-microdescs.new
Tor Bug Tracker & Wiki
blackhole at torproject.org
Thu Oct 10 14:04:44 UTC 2019
#31683: md: Bad family line in cached-microdescs.new
-----------------------------------+------------------------------------
Reporter: dgoulet | Owner: nickm
Type: defect | Status: assigned
Priority: Medium | Milestone: Tor: 0.4.2.x-final
Component: Core Tor/Tor | Version:
Severity: Normal | Resolution:
Keywords: microdesc, 042-should | Actual Points:
Parent ID: | Points:
Reviewer: | Sponsor:
-----------------------------------+------------------------------------
Comment (by nickm):
Musing aloud:
I'm trying to make sense of this and of #31364, which I strongly suspect
have the same underlying cause.
In both cases, it's microdescriptors, and it's the journal, not the cache.
In both cases, it looks like we have started writing at the wrong point in
the file: here, it seems we've started in the middle of another
microdescriptor, and in #31364, it looks like we've seeked to a point
after the end of the file.
This makes me kind of suspect that our fake O_APPEND implementation in
start_writing_to_file() might have something to do with this. Or maybe
something is doing concurrent modifications on the microdescriptors
journal.
Next to investigate: why do we not observe this with routerdescs? Are we
more tolerant of errors in cached_descriptors.new, or is there an
important difference in the code? If this is a concurrent modification
problem, what could be causing it?
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/31683#comment:4>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online
More information about the tor-bugs
mailing list