Matt Dillon’s updated the DragonFly site diary.
Xin Li pointed out a race condition in FreeBSD that exists in DragonFly too. Matt Dillon, on further inspection, found a deeper problem. I’m pasting his description here because he’s speaking another language – I don’t even recognize the acronyms he’s using.
Alan Cox responded with a link to the CMU Mach algorithm on this page as an example of a correct implementation.
Continue reading “Greek to me”
Several more people have reported successful GENERIC kernel builds – and boots – using gcc3. I wonder how/if this changes performance?
Robert Garrett posted that using gcc3 (set your environment var CCVER to gcc3) to buildworld will now succeed.
There’s been a slight conversation about threading, where Jeroen Ruigrok van der Werven mentioned he would want hybrid, or M:N threading. Miguel Mendez chimed in that 1:1 threading, while not as spiffy, is easier to implement and Sten Spans posted a link to this PDF describing 1:1 threading in Linux. (I assume – haven’t read it yet) Matt Dillon brought up a larger issue: asynchronous syscall messaging support is needed before any of this thread work can be done.
Bill Huey (hui) pointed out that the “RCU” Matt Dillon has been talking about in his recent epiphany is “Read-Copy-Update”, mentioned on this page.
Joerg Sonnenberger warned that his NEWCARD patch will go in Tuesday. Further driver work is needed, but nothing should break outright.
Matt Dillon was answering a question from Bosko Milekic that caused him to have a realization on how to improve token-handling in both single and multiple CPU situations, with apparently significant performance benefits. I’m pasting his post to .kernel here:
“Not entirely. First, please note that the current token implementation does not work very well… it’s a mess to use because cpu #B can steal away cpu #A’s token by sending cpu #A an IPI message requesting the token. This means cpu #A has to be in a critical section to ‘protect’ the token, and must re-acquire the token (in case it was lost) after it blocks.
After thinking about this quite a bit I have come up with a solution that, in a moment of enlightment, would make our tokens operate like self-contained little RCU implementations (superior to Linux’s implementation, in fact!).
I intend to fix our tokens by not allowing an IPI to steal a token until the owning cpu has gone through a thread switch, or by explicit release. This means that, once fixed, our token implementation will allow us to abstract RCU-like operations, which I think is very important. It is not what I originally intended tokens to be but it is turning out to be what they should be.
The nice thing about the revamp I intend to do on the token code is that the RCU abstraction will wind up being a lot better then Linux’s RCU abstraction, because it will be compartmentalized. Only cpus competing for the same token are effected rather then all cpus in the system. We would not have the serialization problem that Linux has.”
Amar Takhar put together a generated-from-SGML version of the DragonFly website, with a few bonus features like a site map.
Hiten Pandya has completed the previously-mentioned To-Do list for contributors; an XML version is available if your browser can handle straight XML.
Hiten Pandya asked that anyone (committers or not) working on a task for DragonFly mail him the following data:
“(1) Name
(2) Owner (name and email address) — can be multiple
(3) Current status (work-in-progress, suspended, done)
(4) Description of the Task (300 chars is good!)
(5) Priority (if any) (If any)”
Send it to hmp@backplane.com with a subject of [TODO ITEM]. Will this show up online? I don’t know.
Robert Garrett <rg70@sbcglobal.net> is working on the aforementioned OSL Beaver Challenge; if you are interested in participating for “Team DragonFly”, contact him
KDE 3.2 is out. It should be possible to compile at least the base parts on DragonFly thanks to Dave Leimbach.
Wouter Clarie posted about the OSU Beaver Lab Challenge, where the OSU Lab wants to benchmark a number of operating systems (Linux and BSD flavors) against each other, with teams for each operating system to tweak as needed. There’s no DragonFly team at this point, though it may not be useful to ‘compete’ before an inital release.
Matt Dillon is doing what he did with binutils-214 – He’s bringing in a prerelease version of gcc 3.3.3 to replace the FreeBSD 5 specific version of gcc 3.2 currently used. This version of gcc (and binutils) are straight from the vendor, which should mean staying in sync with new releases will be much easier.
Note: using gcc3 may/will lead to breakage right now…
As part of the discussion about CVS and similar products, Garance A Drosihn pointed out that he was working on a C++ version of cvsup, meaning that it wouldn’t require Modula-3 installed to build as it does now. He stopped work because these is apparently a C version in the works, which he hopes would be out by the time of FreeBSD 5.3’s release, somewhere in the next year. That’s good news for DragonFly, too, as cvsup is the method for code updates for the foreseeable future.
There was discussion some time ago of using rsync for updates, which is certainly possible… Binary updates would be nice, too.
Pedro F. Giffuni brought up the idea of using subversion instead of cvs for storing DragonFly source. Bakul Shah pointed out that subversion was not yet mature enough for a switch from cvs in his experience, and Matt said we are continuing with cvs.
Thanks to Jeroen Ruigrok, the version of autoconf now in CVS knows what a ‘DragonFly’ system is.
Jeroen Ruigrok posted an interesting link about moving machine state between machine.
Matt Dillon’s planning to work on AMD64 support for February. He listed these steps:
“* build support and cross compilation work
* kernel build
* boot 64-bit kernel almost to single user
* 32 bit userland support
* boot kernel to single user
* basic device driver and filesystem testing
* boot kernel to multi user (fully working system at this point)
* everyone w/ 64 bit boxes start banging on it, fixing additional
device drivers, get 64 bit buildworlds working, and so forth.”