Date: Tue, 04 Feb 2003 18:31:58 +0100 From: Dag-Erling Smorgrav <des@ofug.org> To: Shaun Jurrens <shaun@shamz.net> Cc: freebsd-chat@freebsd.org Subject: Re: Project status Message-ID: <xzp4r7kuf3l.fsf@flood.ping.uio.no> In-Reply-To: <20030204151457.GA85804@papagena.rockefeller.edu> (Rahul Siddharthan's message of "Tue, 4 Feb 2003 10:14:58 -0500") References: <20030204151457.GA85804@papagena.rockefeller.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
Rahul Siddharthan <rsidd@online.fr> writes:
> I think it would help if ordinary people like us understood what some
> of the new-fangled features are really about. It's clear that some
> very senior developers don't fully understand, or have doubts about,
> Geom and KSE(*) (to take two examples). The "status reports" describe
> them a bit, but not in nearly enough detail, I feel.
Since you're asking, I'll try to briefly describe some of the new
subsystems we're working on in -CURRENT (for appropriate values of
"we")
- DEVFS is both a framework for creating devices (and allocating
major / minor numbers for them etc) and a file system which exposes
device nodes for these devices and allows standard operations such
as removing a node, creating a symlink to a node, etc. DEVFS also
has a mechanism for specifying the ownership and mode of as-yet
nonexistent devices, to make up for the fact that it doesn't
remember chown / chmod operations across reboots.
Although a very small, very vocal minority is violently opposed to
DEVFS, everybody else seems to agree it's a lifesaver, especially
people who use hot-plug devices a lot (pccard and USB devices)
- devd is a daemon that uses features of DEVFS to detect the creation
or destruction of devices and apply various operations to these
devices (e.g. automatically run dhclient when you insert a NIC into
a pccard slot). I believe it's mostly there to replace pccardd,
which NEWCARD obsoleted.
- GEOM is a framework for applying various transformations to disk
devices, and exporting the transformed devices as new devices. The
heart of GEOM is a method for "tasting" devices to discover what
transformations are applicable and then applying them by creating
instances of one or more GEOM class to the device. This process is
recursive, so a geom (an instance of a GEOM class) can attach to
another geom, etc.
GEOM turns what used to be an ugly ad-hoc jumble of code into a
neat stack or graph of geoms which operate upon one another and
export device nodes to userland. For instance, on a typical i386
system, each disk is "tasted" at boot time and found to contain a
DOS partition table, which causes geoms to be instantiated for each
slice (FreeBSD-speak for partition); then each slice is "tasted"
again, giving rise to other geoms depending on their contents.
GEOM is also exceptionally well documented, both inside and outside
the CVS tree.
- KSE ("kernel scheduled entities") is a threading architecture, both
for in-kernel threads and for mapping userland threads onto kernel
threads (which allows splitting one application across multiple
CPUs; our current thread library doesn't). I don't know of any
substantial opposition against KSE. There are some concerns that
the M-on-N model is not really better than the 1-on-1 model,
especially now that Solaris has switched to the latter, and that
KSE is over-engineered and possibly out of the reach of a team of
(relatively) amateur volunteers. Otherwise, complaints regarding
KSE are mostly about how much time it's taking to implement, and
concerns that the KSE developers aren't subjecting their code to
sufficient testing before committing it. Note that I'm neither
confirming or refuting any of these claims, just reporting what
other people are saying about KSE.
Any more subsystems you're curious about?
> Some time Matt wrote a very detailed and educational article in
> daemonnews about the VM, where even I felt I could understand what was
> going on (an illusion, no doubt).
Other people write articles too. They even get them refereed and
present them at international conferences: USENIX, FREENIX, BSDCon,
BSDCon Europe, NORDU etc. There are also proceedings from various
developer summits, e.g.
<URL:http://www.freebsd.org/events/2002/usenix-devsummit.html>;
Audio recordings of some of these meetings are also available; the
2002 summit is at <URL:http://people.freebsd.org/~joe/summit-2002/>;
> (*)After writing that, I had another look at the project web pages
> cited in the status report, http://www.freebsd.org/kse/ and
> http://people.freebsd.org/~phk/Geom/ . It's not actually so
> uninformative as I thought, but it's still worth filling them out,
> especially Geom.
Poul-Henning has written several papers about Geom; unfortunately,
they're not linked from the Geom page, but you should be able to find
them with Google.
DES
--
Dag-Erling Smorgrav - des@ofug.org
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-chat" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?xzp4r7kuf3l.fsf>