The KAOS lab recently bought a fleet of five Inspiron 11 3000 3138[PDF Warning] (Celeron N2815/4GB/500GB)laptops. They’re tiny little machines with 8 hour claimed battery lives, they’re pretty cute and sort of obstinate.
I borrowed one to play with, notes:
Continue reading
Some years ago the KAOS lab bought a number of Netgear WNR3500L routers to use as network glue. A pair of them have been in constant service since 2011 in the lab and machine room, and both of those died in the last two weeks. I’ve finally satisfactorily tracked down the culprit, and figure it’s worth a quick write up.
The WNR3500L is a bit of a false-promise machine; it was sold as an “open router” with excellent community firmware support, but it has some pretty extensive blobs that mean it really only runs versions of DD or Tomato with a small range of 2.6 kernels.
I initially just swapped the first one to die out for a spare (which happened to be flashed with Tomato instead of DD, it was an experiment when they were first set up), and reloaded most of the settings by hand (we failed to save or document many of the settings elsewhere. Our fault.) until I had time to investigate. When the second one died, it became urgent.
Conveniently, the WNR3500L has a 3.3V RS-232-like serial port (115200/8/n/1) under an easily removed panel, retained by one small torx screw/pin thing (Both are out in the picture while I was figuring out which side was which).
This leads to another victory for the Bus Pirate. Any 3.3V compatible UART adapter would work, but the bus pirate has much nicer cables for this sort of thing. In retrospect, those headers are correctly sized to just plug the female harness ends directly, but I used clips out of habit.
It turns out they were booting right up with only warnings, but the configuration was sufficiently garbled that neither the WLAN nor the Ethernet ports were coming up. I blew the first one’s settings away (note: at least on the old version these were running, reboot TWICE after clearing the NVRAM to get back to a consistent state: the first time it writes default values, the second time it boots cleanly). With the software-only nature of the problem established, I took more time to investigate the second one, at which point…
nvram show
[1310 lines excluded]
size: 32782 bytes (-14 left)
Well, There’s Your Problem.
Turns out the tiny little log fields for monthly traffic statistics filled up the nvram and were causing configuration corruption, and thus failures.
There are some forum threads and bug reports on the matter, and it was fixed about two years ago with a test to prevent writing past the end of the file, a default setting to automatically delete traff logs after 12 months, and various other enhancements. Now that we’re up and running again it isn’t desperate, but it looks like there is a DD release from mid-2013 with the fixes applied that still supports these things, and they are probably due to be re-flashed to keep this from happening again.
Lessons:
Other notes:
The Debian ctte init discussion, which has been the drama that keeps on giving, appears to have finally had a decisive vote for systemd [EDIT: I was over-ambitious about decisive. The drama continues.] as the jessie init. Some notes:
/usr (which offend me in a variety of ways) to, giving credit to the merits of the “/usr belongs to the package manager” retcon, /usr/etc the way some of the BSDs use /etc/defaults.Even with all that said, I’m a little surprised OpenRC wasn’t taken more seriously, since it is the conservative incremental progress choice – objectively, it is currently more featured than Upstart by most metrics – but the developers haven’t been politicking the way the Upstart and systemd folks have, and their incidentally combined efforts managed to discredit it. It has better portability (already runs on Hurd and FreeBSD) than the other options, and is one of two with a (admittedly, less featured) cgroups mechanism. The latter one is interesting because via some impressive politicking the cgroups discussion was at some point re-framed from “makes use of cgroups” to “uses cgroups in exactly the same way as systemd.”
I’m curious how the procd/ubus/hotplug2 stack coming out of OpenWRT is going to interact with the rest of the ecosystem, I suspect we’re going to see that stack meet midway with systed/dbus/udev and some of the light distros using it and shims even for PC-like applications, which will likely be the next time things get really interesting, barring a couple skirmishes where user-level software dictates the stack to the metal.
I’ve been terrible about writing things up lately, but want to at least put my semester impressions on the head of the chain of such posts.
One of my overall experiences is that teaching (and also, taking classes outside my typical discipline) has made me much more talky in class.
CS541: Compiler Design/Finkel
It’s a core class in the CS program, but I’m taking it this semester because Dr. Finkel is teaching it, and his classes are always excellent. I believe this leaves only one compilers/programming languages class offered at UK I haven’t taken, and that one is in the math department and entirely not my thing. I’m just shy of qualified to teach this material, so there’s a slightly odd dynamic in which I wait before delivering answers, and Raphi is frequently imposing a cooldown on my responding in class anyway.
We’re building a compiler for a cleaned up, stripped down C-Like language in Java, basically following Crafting A Compiler. I’ve used an old version of Fischer LeBlanc as reference for compiler material and like it better than the others I’ve used, so I figure this should work out well. Thus far, I’m having more trouble getting myself back into the Java mindset (So OOP. Much Type System. Wow.) than with the Compilers material. High expectations based on experience.
CS621: Parallel and Distributed Computing/Zhang
It’s basically a class in MPI. I’m hopeful that having some nice assigned MPI projects to work on will be good for me. There are two problems though: first, I’m afraid it’s going to turn into a linear algebra class that just happens to take place on parallel supercomputers. Worse, the lectures are terrible. The organization is bad. The slide decks are bad. The Englishrish is bad. The quality of response to questions is bad. It may actually be worse than the terrible Differential Equations with Jar Jar Xin class I took as a freshman. However, it is of the form “CS6xx,” and the content is at least in principle something I want, so as long as the projects and grades turn out, I’m happy. Kind of pissed that the crap lecture overlaps the Friday CS BoF and associated conversation.
GS630: Instructional Technology/Rice
I’m having so much fun in here. It basically consists of a bunch of jaded grad students discussing instructional technology for college teaching, largely from experience, lead by a specialist in the area from CELT. Some of the discussions basically go like this. I’m afraid I’m becoming “that guy” in here because I talk disproportionately and occasionally convivially argue with or reference outside material to the instructor, but again, I’m having a ton of fun, and it counts toward my PFF certificate.
GS610: College Teaching Seminar/Worley
I typically don’t post up my beginning of semester impressions post until all my classes have met, but this one doesn’t start until well into February. At some level, I’m pleased that it’s a general college teaching course instead of a department-specific one, I have trouble imagining a full semester in-discipline teaching class that didn’t turn into a myopia reinforcement program. It should be fun, all the GS classes thus far have been.
I’m not teaching this semester, and hopefully will be getting to more research projects with the freed up time.
I noticed while writing the pre- post for the new semester that I lost track of my customary semester review for the past semester while I was moving hosting. Pushing it up now.
Continue reading
Posting up my notes from SC13 is another thing I didn’t get to during the end of the semester. Remedying now.
The main takeaway sequence from conversations on the floor is as such:
The other really exciting thing that it seems AMD is going to make it, and more. Their lean period finished when the payoff on the XBone/PS4 came in, and they have a VERY good plan for the next >2 years. It works with the premise above about single-core/die MIMD performance ending, and points in the HSA direction – this is the crazy parts with MMUs so a CPU and GPU can share memory without skew penalty and such. ARM and partners are also generally pointed that way, and have been for some time, though apparently AMD isn’t getting out of the x86 game, but it does look like they are getting out of the fat core game.
Continue reading
One of my previous posts about my Singer attracted an email conversation with another owner about the swing-arm mechanism on the no. 42 cabinet. Unfortunately, the end-of-semester insanity struck before the matter was settled, and I am still unsatisfied with what I’ve been able to figure out.
Now that I’ve had a bit of time, I pulled apart my mechanism and took photos, shared below. I’m reasonably certain that if the mechanism is complete and correct, the arm will automatically deploy when the leaf is lifted. Unfortunately, I’m also quite sure that the pieces I do have are inadequate to support that functionality, and I can only guess what the other bits might be.
The larger diameter end of the part I do have matches the diameter of the holes in the hinge, and the smaller-diameter end matches the hole through the swing arm and base.
My best guess is that there are several objects similar to the pictured pin, one of which protrudes below the table into the catch hole of the swing arm through the holes in the hinge mechanism, springloaded “up” such that it retracts when the leaf is out, and is depressed by something protruding from the hinge-hole in the leaf pinning the arm when closed. The pictures sent by the other owner show what looks like the end of a similar pin protruding into the bracket, but it does not extend any where near far enough to retain the arm.
Posted partly in the hopes that my pictures will help other folks with their cabinets, but also if anyone with a no. 42, especially if it has a working swing-arm mechanism or parts that are not pictured, sees this I’d love some more information about how they’re supposed to go together.
Post-SC, I sat down to do some deeper reading on the HSA (Heterogeneous System Architecture) stuff. This is AMD/ARM (and many friends)’s plan for the future, and it is pretty fucking exciting (in an obscure technical sort of way).
The best starting point I found is this year old whitepaper [PDF warning]. They’re using slightly odd terminology, the important bits are LCU = Latency Compute Unit = Conventional MIMD CPU Core, TCU = Throughput Compute Unit = Accelerator, typically SIMD-engine-ish like a GPU, HSAIL = HSA Intermediate Language = IR that can be compiled at install/run time to accelerator’s ISAs. The hardware-side implementation details are nowhere to be found, but there are a lot of seriously exciting model-affecting things detailed on the software side. The general model, with things broken into grid, work group, work item, wavefront is FAR more sane than most of the parallel schemes (I’m thinking specifically of the awful CUDA nomenclature). Internally, the exciting stuff includes requiring a limited sort of preemption on the accelerators, a relaxed consistency model across memory shared over a whole system (nice thread-like shared memory), an intermediate low level language/VM for portability, and assurances about barrier capability in the TCU. The actual objects are basically FAT ELFs with a complete copy of the program for the LCU, plus the HSAIL representation for the parts that can be shipped to TCUs. I’m pleased that there seems to be a clever run-time that does a bunch of platform enumeration and controls where parts run in a rule-automated-but-overrideable way.
I had some folks at SC tell me they’d try to get me a more implementation-focused whitepaper on the hardware side at AMD but they weren’t sure if/when details would be clear for distribution. On the software side, the details are in a published draft of the ISA/Model/Compiler Writer’s Guide that I browsed around in a bit and found very enlightening. The reference tool-chain seems to be mostly built on LLVM and OpenCL.
I have some other SC-related thoughts to share, but I want to get them a little bit more refined (and decide which are for public consumption) before I post.
