Muli Ben-Yehuda's journal

We have several cases in syscalltrack, where due to bad design or circumstances, when you change one file (e.g. the syscall stub code template files) you have to change accordingly another file (e.g. the “special syscalls” files). I’m reading an interview with Rusty Russell on kerneltrap, where Rusty describes how he implemented a MODULE_VERSION macro:

“JA: What will the MODULE_VERSION macro do?

Rusty Russell: Oh, that reminds me, I have to test that. Thanks.

OK, it’s sent. The patch allows a module author to add a version tag to their module, which the modern modinfo can extract. This is a one-liner. But I wanted to catch the (common) case where a change has been made to a driver by someone else, and the version number not modified. So I added a step to the build where every module with a version would have a checksum of the source file contents appended. It’s not perfect, but it gives *some* simple protection, and driver authors can always ignore it.

Patch can be found in my patch collection, or once it’s applied, in you will be able to find it here.”

I wonder if we can use something similar in syscalltrack. Even better would be to remove those nasty dependencies.

Shiny!.

“Does your neck hurt from turning between previewer windows and the source too often? This Elisp/LaTeX package will render your displayed LaTeX equations right into the editing window where they belong [in Emacs or XEmacs -ed].”

Reading my email at work, I discovered that Wes Felter of Hack the Planet fame works for IBM, at Austin. I wonder which other IBM bloggers exist out there?

Disclaimer: while I could be construed as an “IBM blogger” in the sense that I work for IBM Research and have a blog (where? you’re reading it!), my words are my own only. I don’t get paid enough to speak for IBM.

If I feel sufficiently bored, I might make a list of IBM bloggers. google is probably a good start.

I just commited a patch to remove support for floating point parameters from the rule engine in syscalltrack. Floating point support was never actually used, and remained there as an artefact of guy’s original vision of sct_rules being useful for other projects besides syscalltrack. I left it in so far, but since no such uses have materialized, and dead code is dangerous code (you try to use it and it blows up in your face due to having bitrotted), it was time to finally remove it. I hope guy won’t mind…

Me, yesterday, in an email to friends and colleagues:

FYI,

My "Design and Implementation of syscalltrack"[0] paper was accepted
to Linux.Conf.Au[1], one of the two top Linux (kernel) conferences in
the world. A victory 'happy happy joy joy' dance will take part as
soon as I finish telling the entire world :-)

[0] http://syscalltrack.sf.net
[1] http://lca2004.linux.org.au/

Cheers,
Muli

----- Forwarded message from lindy (AT) netcraft.com.au -----

Date: Tue, 23 Sep 2003 18:00:10 +0930
From: lindy (AT) netcraft.com.au
User-Agent: Mutt/1.4.1i
To: mulix@mulix.org
Subject: Response to LCA2004 CFP Submission

Hello Muli

Firstly let me thank you once again for your CFP submission for
LCA2004 titled "Design and Implementation of syscalltrack" (paper).

After completing a review of all CFP submissions, the CFP Selection
Team are pleased to be able to offer you an invitation to present at
LCA2004.

[snipped]

----- End forwarded message -----

More on this later.