Rework and extend kernel-related documentation

[jeroen@…]

Specifically, discuss and integrate the collaborative work done on: ​http://wiki.linuxfromscratch.org/index.php?pagename=BuildingHostKernel ​http://wiki.linuxfromscratch.org/index.php?pagename=BuildingKernel

[archaic@…]

Is this bug still relevant? Chapter 5 's Host System Requirements seems to fulfill it.

[Matthew Burgess]

I think chapter 5 covers it. I thought we also agreed a while back that configuring a kernel was a pre-req to LFS as well. I can't see this assumption actually written in the book anywhere though.

[archaic@…]

As for putting this assumption in writing, this came from the wiki and seems a decent fit:

Building your own kernel is considered a good testcase to see if you have sufficient knowledge of Linux to get the most out of LFS.

Of course I'd have to change it slightly to fit into the context of the prereqs page. I'd also like to put a decent link in explaining the options, as well as pointing to the Documentation dir. If no link exists, then we'll have to settle with the Documentation dir.

[Matthew Burgess]

Following up from the the thread at ​http://www.linuxfromscratch.org/pipermail/lfs-dev/2005-June/051819.html, I think a script that parses the kernel tarball and extracts the config help text into a pretty HTML page would be a great idea. In addition, I think it'd be useful to go through each of the options and highlight the ones that are particularly useful common needs (e.g. use SCSI-GENERIC for CD Burning, etc.). I'm prepared to tackle this if noone beats me to it.

[archaic@…]

Is this going to happen by 6.1?

[Matthew Burgess]

I very much doubt it.

[zarniwhoop@…]

SCSI_GENERIC is kinda obsolete. For normal IDE CD-Writers CONFIG_BLK_DEV_IDECD is all you need (-dev=/dev/hdc or, I suppose, /dev/cdrom).

[Matthew Burgess]

I was under the impression that SCSI_GENERIC is the preferred interface for things like CD burning and the like.

[zarniwhoop@…]

Matt, that was for 2.4, unless I've totally misunderstood linus' change early in 2.5 to let people pass dev=/hdev/hdc instead of the old 0,0,0 or whatever. Ken

[Matthew Burgess]

See ​http://sg.torque.net/sg/u_index.html:

"In the linux kernel 2.4 series most these utilities must be used with a SCSI generic (sg) driver device name (e.g. /dev/sg0). In the lk 2.6 series almost all of these utilities can be used with the primary device names as well (e.g. /dev/sda, /dev/sdc0, /dev/st0 and /dev/hdd (if it is an ATAPI device)). The SCSI generic (sg) interface still represents a cleaner interface than the primary device names since the drivers behind primary device names have their own policies, may interfere with error processing and in some cases run their own state machines."

I was under the impression that using e.g. /dev/hdc would invoke/be handled by the sg driver if it pointed to an ATAPI device. From the above, I'm not sure that's the case.

[Matthew Burgess]

I don't think this bug is valid any longer. The pre-reqs assume that one can build a kernel. As for writing a script that parses the config entries, etc. that's not what LFS is about and I was stupid for volunteering to write such a thing in the first place. Certainly, if someone does feel the urge to produce such a tool and publicise its results somewhere then we can consider linking to it to help folks out. For things like "How do I do I configure my kernel for [x]" there are howto's and BLFS available.

Now I think about it, if there are any config options that are crucial to a base LFS system working (e.g. netlink, tmpfs, procfs, hotplug(?), etc.) and that aren't turned on by default we'd need to document them.