Reverse Engineering IRIX Multithreading For NetBSD

Anonymous Coward writes "Onlamp.com publishes the sixth paper of Emmanuel Dreyfus's series on NetBSD's IRIX binary compatibility implementation. This time, this is about reverse engineering IRIX multithreading and the odd virtual memory features involved with it. It's an adventure at kernel and userland boundaries, with a debugger as the sole weapon. A must read!"

IRIX?

[cpeterso]

ok, so SGI is moving from IRIX to Windows and Linux. So how many IRIX users will be transitioning from IRIX to NetBSD?

Re:IRIX?

[Anonymous Coward]

ok, so cpeterso hasn't been paying attention for the past couple years. Windows is no more at SGI, and Linux is only on Itanium Origins / Altix at the moment, and no indication (at the moment) that it's going anywhere else. Non-hobbyist (except perhaps the most quirky commercial-type) IRIX users will not be transitioning to NetBSD until NetBSD has the graphics and scaleability capabilities that they require.

Re:IRIX?

[zzyp]

What's the point of doing IRIX compat anyway? Not many IRIX users, and is it useful?

Re:IRIX?

[fyonn]

remember this isn't the commercial software world. no-one's been tasked with providing irix compatibility. someone's doing it because they want it and thats reason enough, isn't it?

dave

Re:IRIX?

[Anonymous Coward]

I think most of Slashdot contributors miss the article point: it is not an advertising for running NetBSD instead of IRIX, it is a technical paper on the actual implementation and the reverse engineering techniques used to reveal IRIX undocumented secrets.

IRIX binary compatibility is not sexy, indeed. But IMHO, the tricks exposed in the paper are quite interesting. There is not that much documentation on kernel programming and reverse engineering available around there.

support "legacy" users (Re:IRIX?)

[Anonymous Coward]

SGI may be moving, but what about all its IRIX customers. Some may be happy running what they have now. This way even if the OS is no longer supported by SGI, the users can still run their apps on NetBSD.

Doubtful IMHO

[BuBu_]

I doubt very many will. Don't forget, just because it has the bin compat. doesn't mean that SGI won't be porting all of their good solid IRIX apps. over to their Linux distros. I'd say that if anything, most IRIX guys are going to be heading to Linux in the future, if SGI is definte on phasing out IRIX (ie I haven't read up on it at all). NetBSD's bin compat. with IRIX is going to be quite useful because hopefully it will let some other people cut their teeth on IRIX apps. that might not have access to a

Eeer, Should of been a reply to cpeterso's post.

[BuBu_]

topic.

Re:Doubtful IMHO

[LizardKing]

hopefully it will let some other people cut their teeth on IRIX apps. that might not have access to an actual IRIX machine.

Note that this effectively maps IRIX system calls to NetBSD ones, but you're still going to have to run this on a suitable MIPS machine.

Chris

Some people are missing the point.

[LizardKing]

Slashdot's usual bunch of Linux fanboys are missing the point of these articles. So SGI might move to Linux and might drop IRIX. They might port there apps to Linux (their developers are certainly experienced enough). However, SGI's future operating system strategy has little bearing on NetBSD.

NetBSD has a strong following in the academic world (I'm talking about researchers and postgrads here, not undergrads running Linux file sharing apps in their dorm). The BSD license, along with clear and well documented source make Net an ideal choice for academic work. However, a lot of cutting edge work takes place outside academia, often in companies like Sun and SGI. Often this work is not publicly documented.

Reverse engineering things like IRIX's kernel can give valuable insights into advancements made by SGI. These can then suggest new avenues for research that may have been overlooked otherwise.

So those questioning the utility of IRIX binary compatability, are missing the primary motivation for such work. Of course someone may find the ability to run IRIX apps useful one day - after all, who would have thought that NetBSD's emulation of a niche operating system like Linux would have proved so useful ;-).

Chris

Re:Some people are missing the point.

[Anonymous Coward]

Indeed reverse engineering IRIX gave NetBSD folks some ideas. The libc provided signal trampoline described in part 4 [example.http] is a very good example.

It removes one of the needs of an executable stack. Running with a non executable stack would be a great step forward: stack overflows exploits would not work anymore.

Re:Some people are missing the point.

[jo42]

:s/fanboys/fagboys/g