The most bizarre thing has happened with the AGPGART code.
So, it's been working fine for months. Until recently, when
I disabled the 4g/4g patch in the Fedora kernel (currently
still in 'updates-testing'), and people started noticing X
random oopses and lockups on logging out of Gnome.
This led to much head-scratching, and various theories.
Today Alan did an audit of the agpgart code, and opened
up a real can of worms. There are a number of PCI posting
bugs, CPU cache flushing bugs, missing TLB flushes,
and god knows what next. The more I look into it, the
uglier it gets. It's a miracle things work at all.
Worms. Everywhere.

I recently learned that my
book
is used in the operating system design
course at
Columbia (that is the Ivy League, to you). Happy to hear that!

The other day, I was debugging a hard lock in my experimental rewrite of
inotify's locking. I finally got some sort of back trace out of the kernel
before it choked, so I wrote down the trace, register dump, etc. on my
whiteboard. I spent the next day or so studying my whiteboard, tracking down
the problem:

Went into the office to do battle with the build system.
Came off worse for wear. After much 'it works' 'no it doesnt'
to'ing and fro'ing, things seem in better state.
I think I broke a record. Yesterdays kernel build took
26 hours to complete.

The other day, I read the latest figures on the dollar's performance: Over
the last three years, it has fallen some 30-odd percent against the euro
and 20-something percent against the yen.

While a lot of people are pulling for a weaker dollar, the
world needs to get serious about the current global financial system before
we all regret it. To be fair, the blame for the dollar's current state is
largely--although not entirely--domestic. Our deficit is huge
and reckless and expected to get larger. With that, the dollar will only
weaken and those willing to finance our debt will only dwindle (or demand
higher returns). Our current fiscal policies suck and absolutely harm the
dollar's reserve-currency status. We need to stop blaming Europe and Japan's
slow growth and
starting listening to Greenspan
and begin a change in US policy toward a stronger dollar, beginning with a
reduction in the deficit.

The other day, I read the latest figures on the dollar's performance: Over
the last three years, it has fallen some 30-odd percent against the euro
and 20-something percent against the yen.

While a lot of people are pulling for a weaker dollar, the
world needs to get serious about the current global financial system before
we all regret it. To be fair, the blame for the dollar's current state is
largely--although not entirely--domestic. Our deficit is huge
and reckless and expected to get larger. With that, the dollar will only
weaken and those willing to finance our debt will only dwindle (or demand
higher returns). Our current fiscal policies suck and absolutely harm the
dollar's reserve-currency status. We need to stop blaming Europe and Japan's
slow growth and
starting listening to Greenspan
and begin a change in US policy toward a stronger dollar, beginning with a
reduction in the deficit.

Got a little backlogged after yesterdays fun. Decided to stay
home and try to dig my way out of the patchmountain I was
buried under. Got most of the way through it by the end of
the day.

Introduction
This article covers some recent changes in the SELinux kernel code including a performance patch from Kaigai Kohei of NEC and related updates to the selinuxfs API. Currently, these changes are waiting in the -mm tree for merging into Linus' kernel in 2.6.10, and are also available in the development Fedora and RHEL4 Beta kernels. This article should be useful to sysadmins who are looking to understand SELinux performance and generally also to curious people.

Went in early to attend a meeting hosted by the desktop folks.
After a quick intro by Havoc, we found out that the rest of the
day-long meeting was going to take place at jrb's place.
Drove there, did stuff, had fun, found out I suck at super bomberman.
When we were leaving, it had been raining quite a bit, and the cold
wind had turned it all into ice everywhere. Fun.

The people shipping our belongings seem to think our stuff is
in New York again for some reason. It's starting to get silly.

Up early to go to the dentist to finally get the swab taken
out of my dry socket. Everything aparently all in order now,
though it still aches like hell. Took some painkillers,
and got a warm gel-pack to put on my face for a while.

Spent the rest of the day catching up with mail & patches.

I really only have one irrational fear in life, which is being
sequestered on a jury for a long period of time. I have no idea why I fear
this, I am sure the odds are smaller than plenty of other things I should worry
about, but the thought pains me nonetheless. I do not like the idea of not
having control of what I do, spending months on some media circus of a trial.
I mean, I have code to hack and books to write. Deep down, I know that
ultimately this fear is retarded, like so much of my life.

A 'not doing much' day, mostly due to feeling like
crap from the moment I suddenly woke up and needed to
run to the bathroom. Hopefully I don't have what Paul
had last week. Being ill right now would suck.

Did some experimenting with gcc's -mtune option.
In particular, I wanted to see how well code scheduled
for pentium4 ran on non-p4 systems. I was quite surprised
that my mobile athlon-xp ate it up pretty well,
despite doing no alignment at all. (Due to the P4's
trace cache removes the decoding stage from the pipeline
for frequently executed code (such as that found in
inner loops of most benchmark type programs), loop

Pushed out a new kernel for rawhide, and also FC-3 to
testing. Biggest change here being the introduction of
-hugemem kernels, which mean that the default kernels
no longer have the 4g/4g address space split, but default
to the same as upstream (3g userspace/1g kernel space).
There have been a number of reasons for this change,
which mean we can close a bunch of bugs in bugzilla.

Went back to the dentist to get swab removed from my gum.
Ended up getting it replaced, and asked to come back
yet again next Monday.

Lots of discussion at the office about the various
tracking bugs in bugzilla.

Food in the evening with sct, agk and Arjan.
Various discussion about lots of the current doom-of-the-moment
items.

Victoria bought me 3 cacti. They're called 'neon cacti'
and for good reason. They're really cool colours.