- tor-dev - lists.torproject.org

Re: [tor-dev] gettimeofday() Syscall Issues
by teor 02 Jan '15

02 Jan '15

> From: Yawning Angel <yawning(a)schwanenlied.me> > Subject: Re: [tor-dev] gettimeofday() Syscall Issues > > On Thu, 01 Jan 2015 23:42:42 -0500 > Libertas <libertas(a)mykolab.com> wrote: > >> The first two account for the bulk of the calls, as they are in the >> core data relaying logic. >> >> Ultimately, the problem seems to be that the caching is very weak. At >> most, only half of the calls to tor_gettimeofday_cached_monotonic() >> use the cache. It appears in the vomiting print statements that >> loading a single simple HTML page >> (http://www.openbsd.org/faq/ports/guide.html to be exact) will cause >>> 30 gettimeofday() syscalls. You can imagine how that would >>> accumulate for an exit carrying 800 KB/s if the caching >> doesn't improve much with additional circuits. > > So while optimization is cool and all, I'm not seeing why this > specifically is the underlying issue. > > Each cell can contain 498 bytes of user payload. Looking at things > simplistically this is 800 KiB/s -> 1644 cells/sec, leaving you with > approximately 608 microseconds of processing time per cell. > > On my i5-4250U box, gettimeofday() takes 22 ns on Linux, and 2441 ns on > FreeBSD. I'm not sure how accurate the FreeBSD results are as it was > in a VirtualBox VM (getpid() on the same VM takes 124 ns). If someone > has a OpenBSD box they should benchmark gettimeofday() and see how long > the call takes. > > Taking the FreeBSD case (since we know that tor works fine on Linux), a IPredator has complained that tor on Linux spends too much time calling time() when pushing 500Mbit/s, which is an issue for them under 3.x series kernels, but not kernel 2.6. https://ipredator.se/guide/torserver#performance > single gettimeofday() call takes approximately, 0.39% of the per-cell > processing budget. > > For reference (assuming gettimeofday() in *BSD really is this shit > performance wise), 7000 calls to gettimeofday() is 17.09 ms worth of > calls. > > The clock code in tor does need love, so I wouldn't object to cleanup, > but I'm not sure it's in the state where it's causing the massive > performance degradation that you are seeing. > Yawning/Libertas, I just reviewed my profiling of an exit relay running chutney verify with 200MB of random data. This is on OS X 10.9.5 with tor 0.2.6.2-alpha-dev running the chutney basic-min network. The three leaf functions that take the most time in the call graph are: * channel_timestamp_recv * channel_timestamp_active * time Each of these functions takes around 16% of the execution time, the next nearest function is sha1_block_data_order_avx on 4%. While I understand that OS X, BSD, and Linux syscalls aren't necessarily identical, we now have results for the following platforms suggesting that calling time() too often has a performance impact: * Linux kernel 3.x * OpenBSD * OS X 10.9 My results suggest a maximum performance improvement of 15% on OS X if we reduced the calls to time() to a reasonable number per second. teor pgp 0xABFED1AC hkp://pgp.mit.edu/ https://gist.github.com/teor2345/d033b8ce0a99adbc89c5 http://0bin.net/paste/Mu92kPyphK0bqmbA#Zvt3gzMrSCAwDN6GKsUk7Q8G-eG+Y+BLpe7w…

2 2

Deadline Approaching: EECEA2015 (Electrical, Electronics, Computer Engineering and their Applications)
by Jackie Blanco 02 Jan '15

02 Jan '15

The Second International Conference on Electrical, Electronics, Computer Engineering and their Applications (EECEA2015) World Trade Center, Manila Philippines February 11, 2015 University of Perpetual Help System Dalta, Las Piñas - Manila, Philippines February 12-14, 2015 http://sdiwc.net/conferences/eecea2015/ All registered papers will be included in SDIWC Digital Library. =========================================================== The conference aims to enable researchers build connections between different digital applications. The event will be held over four days, with presentations delivered by researchers from the international community, including presentations from keynote speakers and state-of-the-art lectures. Trainings will be conducted by companies namely Microchip, Dassault Systèmes and TechSource in parallel with the paper presentations where delegates are encouraged to participate. RESEARCH TOPICS ARE NOT LIMITED TO: * Electronics Engineering * Electrical Engineering * Computer Engineering SUBMISSION GUIDELINES: - Researchers are encouraged to submit their work electronically. - Full paper must be submitted (Abstracts are not acceptable). - Submitted paper should not exceed 15 pages, including illustrations. All papers must be without page numbers. - Papers should be submitted electronically as pdf format without author(s) name. - Paper submission link: http://sdiwc.net/conferences/eecea2015/openconf/openconf.php IMPORTANT DATES: Submission Deadline: January 12, 2015 Notification of Acceptance: 2-3 weeks from the date of submission Camera Ready Submission: January 30, 2015 Registration Deadline: January 30, 2015 Seminar Date: February 11, 2015 Conference Dates: February 12-14, 2015

1 0

Re: [tor-dev] Tor BSD underperformance (was [Tor-BSD] Recognizing Randomness Exhaustion)
by teor 01 Jan '15

01 Jan '15

Libertas, As I wrote to you earlier today, IPredator on Linux complained of similar issues with excessive calls to time() under the Linux 3.x kernel series, but not 2.6. Let's track these issues in https://trac.torproject.org/projects/tor/ticket/14056 as they appear to be quite similar. I'd like to get a sense of how many calls per second this represents. (400,000 would seem to be 100 to 1000 per second, unless you were debugging for a long while.) I'd also like to know which function(s)/call stack(s) these calls are being made from. I wonder if TIME_IS_FAST is being defined as 1 in any of these Linux or BSD builds. This If TIME_IS_FAST is not defined (the default), we could change these calls to approx_time(), which is much faster. And in that case there should only be around 1 call to time() each second, which is clearly not the behaviour that either Libertas or IPredator are seeing. Libertas, can you search your tor binary or tor debug symbols for the function "approx_time"? grep will do this for you, or strings, or perhaps gdb if you feel like doing it the hard way. I'll add these questions to trac ticket #14056 as well. Libertas wrote: > I just ran ktrace/kdump (used for observing system calls) on the Tor > process of my exit node, which relays about 800 KB/s. It listed >400,000 > calls to gettimeofday(). The list was swamped with them. > > I think I remember reading somewhere that that sort of system call is > way slower in OpenBSD than Linux. Could this be related to the issue? > I've found a lot of similarly mysterious slowdowns related to *BSD > gettimeofday() on other projects' bug trackers, but nothing definitive. > > As you've likely noticed, although I started this discussion I'm very > new to system-level performance tuning. Let me know if I'm not making > sense, or if there's something else I should be focusing on. > >> <snip>

2 4

log files on windows
by Tony Willingham 01 Jan '15

01 Jan '15

Is there a trick to writing to log files in windows 7? The sample torrc file has this: Log debug file @LOCALSTATEDIR@/log/tor/debug.log I changed mine to this with no success: Log debug file C:\tor\tor_debug.log

1 0

Trac spam filter
by David Fifield 31 Dec '14

31 Dec '14

The Trac spam filter just caused me to lose a comment full of carefully researched external links. It's not the first time, and it's really frustrating. Can we whitelist usernames or something? Submission rejected as potential spam Maximum number of external links per post exceeded I tried to solve the recaptcha five times (they were really hard) and when I finally solved it, the comment wasn't posted anyway. David Fifield

1 0

Internet-wide scanning for bridges
by Vlad Tsyrklevich 31 Dec '14

31 Dec '14

Hello, while taking a looking at TorCloud I it used static OR and obfs2+obfs3 ports making them trivial to discover by scanning EC2's IP address space. This led me to consider the more general attack of internet-wide scanning to discover Tor bridges. Most Tor relays run their ORPorts on port 443 or 9001 so I began investigating there. I began by looking at Project Sonar's port 443 SSL certificate database looking for certificates matching the tor certificate pattern. In talking with Eric Wustrow about ZMap, I discovered that he and the other ZMap authors had already published about this attack in their original paper. Eric mentioned that in his discussion with Roger that the response to the attack was going to be the soon-to-be-deployed obfuscated transports which were intended to be run on random high ports; however, the eventual roll out of obfsproxy did not change the fact that bridge's ORPorts continue to be publicly accessible. While I was analyzing Internet-wide survey-data for port 443 and 9001 (helpfully provided by Project Sonar/H D Moore) I found the Onionoo data set. I was able to quantify the impact of the attack using the Onionoo data and verify it using the ZMap scans. There are 4267 bridges, of them 1819 serve their ORPort on port 443 and 383 serve on port 9001. That's 52% of tor bridges. There are 1926 pluggable-transports enabled bridges, 316 with ORPort 443 and 33 with ORPort 9001. That's 18% of Tor bridges. 203 (64%) of PT-enabled bridges with ORPort 443 were TorCloud. I was able to approximately verify these figures using the Internet-wide scans. By grabbing bridge's certificates and calculating their hashed fingerprint I realized I was also discovering a fair amount of private bridges not included in the Onionoo data set. The results aren't too awful, just under 20% of PT-enabled bridges are affected. It's unclear to me if the low number of PT-enabled bridges (and corresponding high number of vulnerable non-PT-enabled bridges) is a historical artifact or a consequence of out-of-date documentation like https://www.torproject.org/docs/bridges.html.en not setting up pluggable transports. Eliminating the ORPort entirely for PT-enabled bridges seems like the ultimate solution to this attack, but the implications of such a change are unclear to me. Another change to mitigate this issue is changing the behavior of ports specified as 'auto' and then changing defaults and documentation to use auto by default. Currently specifying 'auto' in your torrc lets the OS decide the port; however, if tor were to generate a random port on first use that it continued to use across restarts it would it possible to use this for ORPorts and pluggable transports by default (as opposed to the current situation where for bridges you want ORPort auto but constant PT ports and a constant ORPort for relays which is confusing.) I've attached a patch to warn bridge operators running with ORPort set to 443 or 9001 as a stop-gap measure. - Vladislav

8 12

stem on windows 7
by Tony Willingham 31 Dec '14

31 Dec '14

Any suggestions on how to deal with stem on Windows 7 getting stuck at: [notice] Bootstrapped 45%: Asking for relay descriptors. Occasionally it will make it to: [notice] Bootstrapped 50%: Asking for relay descriptors. If I run Vidalia in client mode only, it works fine. I see in the documentation it says: The timeout argument does not work on Windows, and relies on the global state of the signal module. I have no idea what the "global state of the signal module" means. Python 2.7 Thanks Tony

3 3

Torsocks compile errors
by grarpamp 30 Dec '14

30 Dec '14

On gcc 4.2.1 freebsd 8 (thus not really expecting a fix)... commit ab81544478eb9a0db821fd33684cc2f6aac88c7c syscall.c:213: warning: implicit declaration of function 'handle_mmap' syscall.c:283: warning: cast from pointer to integer of different size ../../src/lib/.libs/libtorsocks.so: undefined reference to `handle_mmap' *** Error code 1

1 0

Hidden service policies
by Mike Hearn 30 Dec '14

30 Dec '14

Hello, As we know, hidden services can be useful for all kinds of legitimate things (Pond's usage is particularly interesting), however they do also sometimes get used by botnets and other problematic things. Tor provides exit policies to let exit relay operators restrict traffic they consider to be unwanted or abusive. In this way a kind of international group consensus emerges about what is and is not acceptable usage of Tor. For instance, SMTP out is widely restricted. Has there been any discussion of implementing similar controls for hidden services, where relays would refuse to act as introduction points for hidden services that match certain criteria e.g. have a particular key, or whose key appears in a list downloaded occasionally via Tor itself. In this way relay operators could avoid their resources being used for establishing communication with botnet CnC servers. Obviously such a scheme would require a protocol and client upgrade to avoid nodes building circuits to relays that then refuse to introduce. The downside is additional complexity. The upside is potentially recruiting new relay operators.

6 11

ScrambleSuit's replay protection incomplete
by Philipp Winter 28 Dec '14

28 Dec '14

In short: The implementation of ScrambleSuit's replay protection is incomplete which means that an active adversary can circumvent it. All the credit for this discovery goes to Lasse Øverlier. ScrambleSuit uses Uniform Diffie-Hellman as one of its authentication mechanisms. To defend against replay attacks, a sever caches the HMAC of a client's authentication message. The attack works as follows. In the first step, an active adversary (e.g., a censor trying to detect ScrambleSuit) observes a client authenticate successfully towards a ScrambleSuit server and captures the server's Uniform Diffie-Hellman response. In the second step, the adversary replays the captured response to the very same server. Since the server did not cache the HMAC of its own response, it will interpret the replayed data as legitimate authentication message of a new client and respond with an authentication response. The adversary now successfully tricked the server into responding despite not knowing the shared secret. This creates a noteworthy distinguisher which can help identifying ScrambleSuit. Luckily, it's easy to fix this problem. Introducing message types would be one option but it would break backwards compatibility. The easiest fix which retains backwards compatibility is to make the server also cache its own HMACs which are part of the response to a client's authentication message. The downside is that it doubles the size of the replay table but that's tolerable. Note that obfs4 is not affected by this problem because a client's and a server's authentication message are different. Cheers, Philipp

1 0