Opened 17 years ago

Closed 17 years ago

Last modified 16 years ago

#2023 closed task (fixed)

LiveCD is not available

Reported by: alexander@… Owned by: lfs-book@…
Priority: normal Milestone:
Component: Book Version: SVN
Severity: major Keywords:
Cc:

Description

Chapter 1.1 says:

As an alternative to installing a separate distribution onto your machine, you may wish to use the Linux From Scratch LiveCD. The CD works well as a host system, providing all the tools you need to successfully follow the instructions in this book. Additionally, it contains all the source packages, patches and a copy of this book. So once you have the CD, no network connection or additional downloads are necessary. For more information about the LFS LiveCD or to download a copy, visit http://www.linuxfromscratch.org/livecd/.

However, the latest SVN build of the LiveCD is too bad for all owners of Intel motherboards (relevant modules: piix, ata-piix): either does not boot, or does not see SATA hard drives. The problem is in the kernel (caused by not-so-optional switch to libata as an alternative to old IDE drivers) and is not fixable without breaking laptops because they sometimes have disks with host-protected area (apparently not supported by libata). Until this is sorted out, the book should not recommend the Official LFS LiveCD. However, I have no time to fix it and not enough testers with the relevant hardware. I am not even sure if the 6.3 version of the CD will be released at all.

Please comment out the paragraph quoted above.

Attachments (2)

initramfs-add-ide-options.patch (683 bytes ) - added by bryan@linuxfromscratch.org 17 years ago.
Add IDE option support to the initramfs
initramfs-retry-find-cd.patch (1022 bytes ) - added by bryan@linuxfromscratch.org 17 years ago.
Retry finding the LiveCD device a few times

Download all attachments as: .zip

Change History (23)

comment:1 by Ag. Hatzimanikas, 17 years ago

My understanding was that the host system should conform in some minimal versions, as they are described in the Host System Requirements page of the book.

Specifically for the kernel version, it doesn't have to follow LFS's current version, unless there is a special reason. If there is one, please disregard my comment. But if not, I can't see the reason why the Live cd, can't ship with a proved stable version from the 2.6.20 branch.

comment:2 by alexander@…, 17 years ago

Downgrading the kernel won't help against the "LiveCD is completely untested" issue (a.k.a. "only two e-mails of user-initiated feedback in 6 months"). I refuse to maintain it in such conditions.

comment:3 by Ag. Hatzimanikas, 17 years ago

Yes, I can understand your frustration. Then, what you have to do is to post a request in the proper list, for another volunteer to take over the maintainship, if this what you ask for.

The book however, will not have to change at least for quite some time, because even the last release of the Live cd is totally capable to built a proper toolchain. Isn't that right?

By the way. Do we have any stats (I believe we do, but just to be sure) of how many LiveCd downloads (including torrents) we have over the last year or so?

comment:4 by justin, 17 years ago

Sorry I haven't been around much for development, have been very busy with work and school. I do have a normal x86 and x86_64 box to test on if you need my help. But I believe more than just mysql and a couple others need to test, but emails to lfs-support and such to test a new CD might help.

@Ag.Hatzim: It is hard to collect such stats since we have various mirrors which host it and don't get logs from them. The only info I can give is the awstats for my anonymous ftp and http (click Viewed -> Full List):

FTP over HTTP: http://stats.cross-lfs.org/awstats/cgi-bin/awstats.pl?config=httpftp

FTP: http://stats.cross-lfs.org/awstats/cgi-bin/awstats.pl?config=ftp

While some are probably abusive hits or not full downloads, I would guess around 500 downloads/month from my server alone.

comment:5 by Ag. Hatzimanikas, 17 years ago

Oh thanks Justin that was quick.

I can see a very a good number here (at least for my expectations), Alexander.

So perhaps, you don't get any reports because you are doing a good job? I know you do, but just to make it sure for you.

comment:6 by alexander@…, 17 years ago

@Ag.Hatzim: There is one more requirement to the host system, so obvious that it is not mentioned in the book: it should boot and work. The old CD doesn't boot on i965 motherboards with jmicron-based IDE controller. I will post some notes about packages and known bugs on the CD to the livecd list for the new maintainer (maybe you?).

in reply to:  6 comment:7 by Ag. Hatzimanikas, 17 years ago

Replying to alexander@linuxfromscratch.org:

@Ag.Hatzim: There is one more requirement to the host system, so obvious that it is not mentioned in the book: it should boot and work. The old CD doesn't boot on i965 motherboards with jmicron-based IDE controller.

I've spent some time to read the Livecd mailing list, which I should have done before. Apologize.

Reading through this message (1), I realized that there is no way for anyone in the world to properly test all the available hardware *without* testers; and guarantee that the Livecd will be a suitable host system for everyone. So, I now understand that Alexander is totally right and the bug should be considered as a valid one.

Therefore I suggest three solutions:

  1. Abandon the Livecd project and start to recommend another solution, see below.
  1. Change or even eliminate the text in the book to reflect this reality (that our Livecd it's maybe not the best solution available).

For example, Knoppix because of it's widespread (based in debian) it might be another good recommendation; if we have to do a recommendation. Plus. Find a new maintainer (if Alexander doesn't satisfies with the change) that will have the desire to lurk into some dozen (with heavy traffic) mailing lists, plus to have the available skills to maintain almost alone (like Alexander did for months), such a difficult project like the Livecd. Also to find dedicated people with different hardware and the will to test the Livecd regularly to their machines and report back. The last part is critical.

  1. Start to collaborate with an external source (distribution or an individual with resources) outside LFS that will offer a solution (based in LFS), either as a livecd or/and as an installed distribution in the hard disk.
  1. http://linuxfromscratch.org/pipermail/livecd/2007-April/004297.html

comment:8 by manuel@…, 17 years ago

I think that we could add a note bellow that para discussing that the LiveCD might not work on newer hardware configuration and asking users to test it and to report to the livecd mailing list possible issues.

That could allow more testers, reports and, maybe, collaborators.

comment:9 by dnicholson@…, 17 years ago

I think Manuel's suggestion is much better than simply removing the LiveCD from the book. Simply because the LiveCD doesn't work for some hardware should not exclude anyone else from using it.

Long term solution for the LiveCD means someone is gonna have to step up and maintain it. I think the right thing to do is to spam the lists and see if anyone's interested.

comment:10 by alexander@…, 17 years ago

While Manuel's suggestion may work, I don't want to release the current state of the repository even as a -pre2. The following has to be fixed before the pre-release:

  • Extend the initramfs /init script so that options like "ide=nodma ide2=noprobe" (see the full list in linux-2.6.X/Documentation/ide.txt) that applied to non-modular IDE are automatically converted to modprobe.conf statement like this:

options ide_core options="ide=nodma ide2=noprobe"

  • Update the help pages of the boot loader to reflect the "toram", "pata", "load", "blacklist" and "rootdelay" arguments
  • Make the initramfs retry locating the CD sevelal times before dropping to the shell. This will fix the "impossible to boot from USB CD-ROM" issue (rootdelay=20 on the command line is a workaround that works with r1904, but cannot be the default because of much faster-to-detect IDE CD-ROMs).
  • Make brltty actually work with USB devices. Right now, it pokes around non-existing /proc/bus/usb. Debian has a patch that converts it to libusb, and also there is a new untested version upstream.
  • Add net-tools, remove all "ifconfig -> ip" patches.
  • Save and restore sound volume when hibernating/resuming. Also, unmute mixer channels appropriately at startup.

Patches are welcome.

by bryan@linuxfromscratch.org, 17 years ago

Add IDE option support to the initramfs

comment:11 by bryan@linuxfromscratch.org, 17 years ago

I just attached an attempt at a patch for the first item on your list. It'll handle any ide=*, idebus=*, ide[0-3]*, or hd[a-z]=* options, even if not all of them are valid for the driver. (It's fairly dumb code, but it seems to work when running outside the initramfs at least. I haven't tested it inside.)

I'll see what I can do about retrying the CD-locating code, too, after lunch.

by bryan@linuxfromscratch.org, 17 years ago

Retry finding the LiveCD device a few times

comment:12 by bryan@linuxfromscratch.org, 17 years ago

OK, here's some code to retry looking for the CD a few times. I assume that when a slow SCSI or USB controller is the issue, the raw device node does eventually appear via udev, correct? Otherwise this repeated calling of find_cd won't actually do any good.

This code does 5 retries, with a 6-second delay between each one, for a total (maximum) delay of 30 seconds. This can be changed (modify DELAY and/or TRIES). I also slightly modified the final error text ("needs more time to settle" instead of "needs some time to settle").

comment:13 by manuel@…, 17 years ago

Added a note about possible LiveCD issues on newer hardware and asking for testers on r8154

I think that this ticket could be closed, after moving remaining issues to the LiveCD Trac system.

in reply to:  13 ; comment:14 by alexander@…, 17 years ago

Replying to manuel@linuxfromscratch.org:

Added a note about possible LiveCD issues on newer hardware and asking for testers on r8154

This is not only "newer hardware", so a better wording is needed.

There was a complete rewrite of initramfs in r1872, in order to fix the inflexibility (aka deciding for the user what to use - ide or libata) and the bug that the CD refused to boot on some laptops because non-modular SCSI drivers blindly poked random ports. This commit, however, introduced regressions, and they are part of the reason why I say that the CD is not ready for release.

I think that this ticket could be closed, after moving remaining issues to the LiveCD Trac system.

which is useless because there are no co-maintainers.

in reply to:  14 ; comment:15 by manuel@…, 17 years ago

Replying to alexander@linuxfromscratch.org:

This is not only "newer hardware", so a better wording is needed.

IMHO, issues related with newer hardware will be always here.

To discuss the exact issues, or if the ongoing development fixes introduced some regression not solved yet, is outside the scope of the book.

which is useless because there are no co-maintainers.

And that is the goal of the note added to the book: try to recruit more testers and collaborators for the LiveCD.

If we have luck, we might have an usable LiveCD for when LFS-6.3 will be released. If not, we will need to change the text to say that there is no LiveCD for LFS-6,3 due the lack of maintainers.

in reply to:  15 comment:16 by alexander@…, 17 years ago

If we have luck, we might have an usable LiveCD for when LFS-6.3 will be released. If not, we will need to change the text to say that there is no LiveCD for LFS-6,3 due the lack of maintainers.

OK.

comment:17 by alexander@…, 17 years ago

One little problem with the wording (but feel free to ignore it).

The words "newer hardware configurations" sound too much like a valid excuse for the non-working CD. A reader might think: "my hardware is indeed rather new, so the book says it is normal that the CD doesn't work, so I won't report the bug".

comment:18 by alexander@…, 17 years ago

The above patches are applied, and the brltty issue is worked around. So what remains is mostly documentation.

in reply to:  18 comment:19 by Jeremy Huntwork, 17 years ago

Replying to alexander@linuxfromscratch.org:

So what remains is mostly documentation.

More documentation is always good. :)

My intention is to jump back into CD development and help maintain the current CDs. Hopefully, that should help resolve this ticket. I'm refraining from closing this out just yet, mostly to serve as a reminder. Once I've made a bit of progress there, I'll come back and close this up.

As always, more help is always welcome, as is feedback/suggestions for future CDs.

comment:20 by Jeremy Huntwork, 17 years ago

Resolution: fixed
Status: newclosed

Closing this ticket since the current situation seems to be somewhat better than previously.

(Personally, I feel that the wording in the book could be changed somewhat. For example, nearly all my machines at work have SATA drives and none of them have failed to boot or failed to detect the SATA drives. I always boot with minimal options passed to the kernel - only vga=xxx, UTC and TZ. The current message in the book feels overly negative and misleading. But that may just be my perspective. In any case, what is there now is good for addressing the above concerns. We can revisit this more in the near future, especially as we get nearer a 6.3 LiveCD release.)

Alexander, please report any remaining items that you notice as a bug, defect, or needed enhancement as new items to the LiveCD Trac system. That should help us work together toward a release.

comment:21 by Jeremy Huntwork, 16 years ago

Milestone: 6.3

Milestone 6.3 deleted

Note: See TracTickets for help on using tickets.