Light-Debian-Core-Live-CD-Wheezy + Porteus-Wheezy

For talk and support relating specifically to Puppy derivatives
Message
Author
User avatar
fredx181
Posts: 4448
Joined: Wed 11 Dec 2013, 12:37
Location: holland

#2251 Post by fredx181 »

Hi Toni
Just tested the same code and I can't make it boot
Just in case you didn't see: that code has 'persistence' as boot parameter which is, as you know, for live-boot v3.

Fred

User avatar
saintless
Posts: 3862
Joined: Sat 11 Jun 2011, 13:43
Location: Bulgaria

#2252 Post by saintless »

No, persistence or persistent does not make boot problems for any version of live-boot. I mixed them more than once while testing the first base for Light-Wheezy. Just the save file will not be loaded if the wrong one is used.
The problem is somewhere else.

Toni

mcewanw
Posts: 3169
Joined: Thu 16 Aug 2007, 10:48
Contact:

#2253 Post by mcewanw »

saintless wrote: Hi, William.
The iso size will stay the same. Only if using Fat flash drive it will copy the main module two times on the flash drive.
Ah, okay, I misunderstood. I'm relieved it is that way. It is good to have both options in the one iso, when there is no huge size penalty, of course.
github mcewanw

mcewanw
Posts: 3169
Joined: Thu 16 Aug 2007, 10:48
Contact:

#2254 Post by mcewanw »

saintless wrote: Just tested the same code and I can't make it boot. I get file not found boot message. The strange thing is I can't make it boot now even with live-media-path=/debian-wheezy/live/ and I'm sure I have it working this way before.
Hi Toni, maybe nothing to do with it, but good if you could test on newer machine than your old 128MB grunter. I do most of my testing on my own old machine here, which doesn't boot from usb. However, my installations are all on usb flash because the hard disk is failing on this machine. I manage that because I have grub4dos/menu.list/grldr on hard drive and the menu.lst instructed to find kernel etc on the usb flash (doesn't need a bootable flash of course). However, I do have weird issues 'sometimes' and very unpredicatable where I sometimes on the same usb flash get 'file not found boot message'. Usually after deleting and installing some new stuff. The other thing is I currently can boot debiandog from current usb but no longer Puppy guydog (file not found boot message for that one). I have come to the conclusion that my old machine has these BIOS/ATA limitations (no of heads, cylinders, usb size etc) such that BIOS can sometimes luckily find vmlinuz etc if it happens to be on earlier enough cylinder and low down in the GB. Once vmlinuz etc is booting it takes over from the BIOS so the rest of the system works fine thereafter - that's my intuitive suspicion anyway (I've been messing around with fdisk expert mode to try and prove it one way or the other, but nothing conclusive so far...). There are certainly many limits that could come into play - these old machines are not expecting large hard drives and usb flash view of the chs situation probably even worse for them:

http://www.tldp.org/HOWTO/Large-Disk-HOWTO-4.html

Certainly, in the past, when larger hard discs started coming out, it was important to create a small partition for the bootup files alone (vmlinux etc, which was in /boot usually, so /boot mounted to own small partition back then), in low down cylinders etc - otherwise bios couldn't find vmlinuz.

If I'm right, that would explain much mysterious behaviour when sometimes the same thing seems to work and other times it doesn't. Nothing to do with filesystem formats, in such cases, though might depend sometimes on the boot loader programs (grub/syslinux/wee etc) capabilities also.

I have one 4GB usb flash stick that always refused to boot from my grub4dos setup, but last night after mucking around with fdisk expert mode changing amount of cylinders seen etc, I somehow had debiandog booting from it in a 1 GB ext4 partition. Unfortunately, I wiped it without noting down what I did and I'm been starting from scratch trying to repeat that success ever since! In attempts since I just get the file not found boot message, and in some attempts the machine actually freezes on boot attempt needing a power off before trying again. However I do have debiandog booting from a different flash stick of even greater size (8GB) on this machine (via hard disk grub4dos arrangement). On that one I have an empty first 1GB fat32 partition, followed by a 6GB ext4 partition with debiandog on it and also Puppy guydog (which as I say won't boot, yet debiandog does... all very weird - the same usb flash drives all boot fine on my other newer machine)

Having said all of the above, to be honest, I have no idea what the issue is - maybe nothing to do with BIOS at all, though why I got that 4GB card working only once and why debiandog visible and guydog not on same partition is a real mystery...
github mcewanw

User avatar
saintless
Posts: 3862
Joined: Sat 11 Jun 2011, 13:43
Location: Bulgaria

#2255 Post by saintless »

Hi, William.
mcewanw wrote:Hi Toni, maybe nothing to do with it, but good if you could test on newer machine than your old 128MB grunter.
Thanks, I also suspect this could be the cause of my troubles or the version of Grub I have. I will use grub4dos test on another computer today. I need to be sure it is not a problem caused from the remount /live/image line and one squeeze path_id file added in initrd.img

Thanks to Fred we will not use symlink in the iso and vfat partition issue is solved.

Toni

mcewanw
Posts: 3169
Joined: Thu 16 Aug 2007, 10:48
Contact:

BIOS/ATA limitations workaround found

#2256 Post by mcewanw »

mcewanw wrote:I have come to the conclusion that my old machine has these BIOS/ATA limitations (no of heads, cylinders, usb size etc) such that BIOS can sometimes luckily find vmlinuz etc if it happens to be on earlier enough cylinder and low down in the GB. Once vmlinuz etc is booting it takes over from the BIOS so the rest of the system works fine thereafter - that's my intuitive suspicion anyway
My intuition seems to have proved correct Toni. I now have Debiandog booting from my tricky 4GB usb.
EDIT:

However, I have edited this post and removed the previous txt, which I wrongly thought was the 'solution'.

In case someone else has similar problems with debiandog not booting from usb stick on older machine they might like to try my suggestion in post below (though may or may not work - I only know it works on my old machine):

http://www.murga-linux.com/puppy/viewto ... 156#770156
Last edited by mcewanw on Wed 09 Apr 2014, 04:28, edited 3 times in total.

User avatar
saintless
Posts: 3862
Joined: Sat 11 Jun 2011, 13:43
Location: Bulgaria

Re: BIOS/ATA limitations workaround found

#2257 Post by saintless »

Hi, William.
mcewanw wrote:This could have important implications for debdog-install, if what I have found proves correct, if we want it to work on old machines - should arrange for vmlinuz to be copied first and then a sync before the rest.
... simply renaming it to say 'avmlinuz' would copy it early during installation...
Renaming vmlinuz is not possible in iso install option I think.
There is no need to make debdog-install so complicated since the problems you have on your old hardware are not the same on mine much older hardware. I even doubt someone else will use DebianDog on old hardware like mine.
I have no problem to boot 512Mb and 8Gb flash drive using debdog-install latest version. The copy order of vmlinuz is the same as yours. I have problem to boot if there is 8Mb unalocated space in the beginning of the drive which seems is not a problem for Terry's hardware and I guess for almost any other new hardware.

It is better just to add what we find as solution to not general problems in DebianDog documentation.

Toni

mcewanw
Posts: 3169
Joined: Thu 16 Aug 2007, 10:48
Contact:

Re: BIOS/ATA limitations workaround found

#2258 Post by mcewanw »

saintless wrote: There is no need to make debdog-install so complicated since the problems you have on your old hardware are not the same on mine much older hardware. I even doubt someone else will use DebianDog on old hardware like mine.
I have no problem to boot 512Mb and 8Gb flash drive using debdog-install latest version.
Hi Toni, fair enough, and I wanted to test the mechanism further before even modifying my own copy of debdog-install. Turns out my original 'solution' was wrong anyway (it is complex testing these things so easy to be in error...:-) so I have removed most of the txt from my earlier post above about this (so as not to confuse anyone or give them false hope..!). I have been testing all day to try and find out what is going on.

Note that my 'old' machine turns out to have BIOS from 2003. It is a Pentium M class machine, with 1GB RAM and 1.6GHz Pentium M cpu so a class of machine of which there are many around and all capable of running DebianDog well. Of course, only some such machines will have an equally rubbish BIOS...

Here are the results in case some other user finds them useful:

1. If I use debdog-install to install debiandog onto my 'tricky' 4GB usb, the resulting install seems to ALWAYS SUCCEED when I then boot from that usb on my newer computer.

2. However, even though it will work on newer computer, I can create the installation in such a way that it will ALWAYS FAIL when I try to boot it on my older computer, which is a Fujitsu Siemens Amilo laptop, 1.6GHz CPU with 1GB Ram, whose BIOS turns out to be American Megatrends from year 2003. The boot always fails on this machine if I use debian dog gparted (either alone or via debdog-install) to create the partitions on the usb I am installing to (but will still boot fine on my newer machines). I discovered this in further tests today.

3. I now know how I managed to get it working yesterday (it wasn't what I thought). I forgot that yesterday I had at one stage used Lubuntu fdisk and mkfs.ext4 to re-partition the usb and reformat it. I then took that Lubuntu formatted usb and, using debiandog and debdog-install made an installation. THAT INSTALLATION now booted fine also on my old machine. Further/many repeated tests today led me to realise that the Lubuntu fdisk was somehow adjusting the CHS values automatically such that the no. of heads was displayed as less than 16 after the partition for the installation had been made.

4. I thus tried to use fdisk and mkfs.ext4 on debiandog instead (having first used gparted to make an installation that wouldn't boot on my old machine. However, unlike on Lubuntu, debiandog's fdisk wasn't automatically changing the CHS values automatically. I thus used debiandog's fdisk in 'expert' mode to force the heads to a value less than 16 as follows:

fdisk /dev/sdb
Command: d to delete all the partitions (only was one).
Command: x to enter expert mode.
Command: h to adjust apparent number of heads to 2 (anything under 16 seemed to do the job actually).
Command: r to return to normal menu
Command: n to create a new partition 1 spanning the whole usb. Note that I repeated this experiment with start 2048 sectors (1MiB) and 16384 sectors (8MiB)
Command: w to write the new partition info to the usb and quit fdisk.

EDIT: Note that above can be done from single commandline command using sfdisk commandline options. You can find method in follow-up post below here: http://www.murga-linux.com/puppy/viewto ... 198#770198

5. Running debdog-install with this already partitioned usb, I chose gparted, but ONLY to format the /dev/sdb1 partition (using ext4) - I repeated the experiment using mkfs.ext4 -O ^has_journal on another occasion with the same good results.

6. I then in debdog-install selected Directory 'DebianDog Folder' as /mnt/home/live and pressed install button to copy that to the /dev/sdb1 ext4 partition.

The resulting debiandog usb installation now boots on both my old machine and newer machines. Like I say, I have repeated this experiment many times, first using gparted to partition (which always created an installation that WILL NOT boot on my older machine, but will boot on my newer machine) and then using debdog fdisk in expert mode to change heads to 2 (from whatever they originally were). Sometimes, I used a starting offset of 1 MiB and sometimes 8MiB in fdisk creation of partition - on my machine that made no difference - just the number of heads mattered... Note that I didn't bother changing number of sectors or cylinders using fdisk in that expert mode - just the heads; I left fdisk to decide the sectors and cylinders it liked, and that worked fine.

Though you say your machine won't boot usb with 8MiB offset Toni (and that may well be the case), I would still be grateful if you could try the above method with 8MiB offset just to be sure gparted isn't the culprit really. At least now I don't need to use Lubuntu to fix the usb CHS values, though I don't know why the Lubuntu fdisk is cleverer than the debiandog one (since Lubuntu fdisk automatically adjusted no. of heads to <16 on creating the new partition on this 4GB usb stick).

WOW, that was a lot to write, and maybe makes no sense to anyone else, but it really worked on my old banger machine and repeatably this time (and repeatedly failed if used gparted to partition interestingly enough - but remember - ALWAYS booted fine on newer machine anyway - its just my older machine needed to avoid debiandog gparted for partitioning and needed fdisk to adjust head count value).

Must have installed 100 times today in all these tests! NOTE WELL, that I didn't need to copy vmlinux first afterall; apart from using fdisk as described (and not gparted to partition) the actual installation was just the standard debdog-install one.

EDIT: For some completeness I should mention again that my old Amilo laptop won't itself actually boot from usb (meaning no such option in its BIOS). Instead I have the 'wee' bootloader (grub4dos variant) installed in MBR of Amilo's harddrive and grldr and menu.lst on first partition of same hardrive, which is used to boot vmlinuz1, initrd1.xz, and all the other bits and pieces from the usb where live folder is.
Last edited by mcewanw on Wed 09 Apr 2014, 09:46, edited 3 times in total.
github mcewanw

User avatar
fredx181
Posts: 4448
Joined: Wed 11 Dec 2013, 12:37
Location: holland

#2259 Post by fredx181 »

William wrote:However, unlike on Lubuntu, debiandog's fdisk wasn't automatically changing the CHS values automatically.
I wonder where the fdisk included in DebianDog comes from.
Officially it's part of the gnu-fdisk package but it's not installed.
Looking at the size of fdisk from gnu-fdisk (extracted the .deb) it's different from the included one.

Fred

mcewanw
Posts: 3169
Joined: Thu 16 Aug 2007, 10:48
Contact:

#2260 Post by mcewanw »

fredx181 wrote:
William wrote:However, unlike on Lubuntu, debiandog's fdisk wasn't automatically changing the CHS values automatically.
I wonder where the fdisk included in DebianDog comes from.
Officially it's part of the gnu-fdisk package but it's not installed.
Looking at the size of fdisk from gnu-fdisk (extracted the .deb) it's different from the included one.

Fred
Maybe I should try with gnu-fdisk package installed then in case gparted partitioning problem somehow stems from that?

EDIT: Fred, gnu-fdisk seems to be different from standard fdisk. From apt-cache show gnu-fdisk
Description-en: Linux fdisk replacement based on libparted
GNU fdisk is a replacement to the old Linux fdisk. It provides the same
features as the original fdisk provides plus some interesting ones like:
* partition resizing
* creating filesystem on newly created partitions
* partition integrity checking
* copying/moving partition
Don't think standard fdisk can create filesystems etc

EDIT2: well, just installed/tried gnu-fdisk - seems pretty much like standard fdisk to me, but it is certainly a different one. However the one in Lubuntu is more like the one in standard debiandog install, so think the debiandog existing fdisk is probably correct one (just don't know why Lubuntu fixed heads in a way that my install worked correctly on old system)
Last edited by mcewanw on Wed 09 Apr 2014, 07:23, edited 1 time in total.
github mcewanw

User avatar
saintless
Posts: 3862
Joined: Sat 11 Jun 2011, 13:43
Location: Bulgaria

#2261 Post by saintless »

fredx181 wrote:I wonder where the fdisk included in DebianDog comes from.
Same place where DebianDog comes from, Fred:
http://live.debian.net/cdimage/release/ ... andard.iso

Toni

mcewanw
Posts: 3169
Joined: Thu 16 Aug 2007, 10:48
Contact:

#2262 Post by mcewanw »

fdisk is from the package util-linux. The correct version is already installed in debiandog. Anyway, it's no big deal either way. I'm just posting in case others with old machine have same problem and fix for my old computer may or may not help them. I have no idea why or how my fix works on my machine, but it does, and I'm happy with that :-) debdog-install works fine anyway, without any such fussing around for my newer machines.
github mcewanw

User avatar
fredx181
Posts: 4448
Joined: Wed 11 Dec 2013, 12:37
Location: holland

#2263 Post by fredx181 »

Toni wrote:Same place where DebianDog comes from, Fred:
http://live.debian.net/cdimage/release/ ... andard.iso
Ok, but I ment from which package.
Because of the problems William described I thought for a while something's wrong maybe and I could find only gnu-fdisk when googling.
But as William wrote it's from util-linux, which is installed so it's ok.

Fred

mcewanw
Posts: 3169
Joined: Thu 16 Aug 2007, 10:48
Contact:

#2264 Post by mcewanw »

http://www.murga-linux.com/puppy/viewto ... 156#770156

This is just a follow-up on from my big post linked above for anyone interested. To simplify the process, the way I now get usb flash sticks to install debiandog on one big ext4 partition in such a way that it will boot in both my old and new machine is as follows:

1. I partition and format my usb stick (on /dev/sdc) with the following two commands:

Code: Select all

echo "16384,,L,*" | sfdisk -f -H 16 -uS /dev/sdc
mkfs.ext4 -O ^has_journal /dev/sdc1
2. The above reduces the apparent heads to 16 (-H 16) and formats /dev/sdc1 partition with ext4 fs. Then I simply run debdog-install, click on /dev/sdc1 partition, select debiandog install iso (or my /mnt/home/live folder), and press the install button (I avoid using gparted since that seems to muck up the heads bit; not that there is anything wrong with gparted - it is probably just using the kernel's representation of CHS values).

The result boots on both my old and newer computer and the bootable (*) /dev/sdc1 Linux (L) partition starts on 8MiB boundary (16384 sectors).
Last edited by mcewanw on Wed 09 Apr 2014, 10:33, edited 1 time in total.
github mcewanw

User avatar
fredx181
Posts: 4448
Joined: Wed 11 Dec 2013, 12:37
Location: holland

#2265 Post by fredx181 »

Hi Toni
After some more testing the new setup of DebianDog-PorteusDog I found that the "Always Fresh" entry for porteus-boot in menu.lst doesn't work.
That's because of the 'base_only' parameter is broken.
But found a fix.

Edited my post on previous page:
EDIT2:New revision of DebianDog-PorteusDog-new-setup.tar.gz uploaded:
The base_only boot option didn't work previously, now it does.
Forgot also previously to add the modules, optional and rootcopy folders in /live to the archive.

What you need only in fact is the edited linuxrc so attached here.

Fred
Attachments
linuxrc.zip
linuxrc edited, fixed base_only boot option
(7.03 KiB) Downloaded 245 times

User avatar
saintless
Posts: 3862
Joined: Sat 11 Jun 2011, 13:43
Location: Bulgaria

#2266 Post by saintless »

fredx181 wrote:What you need only in fact is the edited linuxrc so attached here.
Thank you, Fred!

BTW what is yours compress initrd.xz command? Default xz compression is not supported by official Debian kernel.
I use this one which works. It takes long time on my hardware to compress the file.

Code: Select all

find | cpio -H newc -o | xz --check=crc32 --x86 --lzma2 > ../initrd.xz
I think to use the same compression for initrd.img file and it will make it around 2,5Mb smaller. If it works I will add one more initrd.img for live-boot-v3 that uses persistence save file.
The iso size will be 4-5 Mb larger with 3 initrd files but we get all boot methods in one iso: live-boot-3x-persistence, live-boot-2x-persistent and Porteus boot.
This way the user will have a choice to find the best one needed and to remove the other 2 initrd files to save space.

Toni
Last edited by saintless on Wed 09 Apr 2014, 12:50, edited 1 time in total.

User avatar
saintless
Posts: 3862
Joined: Sat 11 Jun 2011, 13:43
Location: Bulgaria

Re: BIOS/ATA limitations workaround found

#2267 Post by saintless »

mcewanw wrote:Though you say your machine won't boot usb with 8MiB offset Toni (and that may well be the case), I would still be grateful if you could try the above method with 8MiB offset just to be sure gparted isn't the culprit really
Hi, William.
I will try this in the next few days and write back.

Toni

User avatar
fredx181
Posts: 4448
Joined: Wed 11 Dec 2013, 12:37
Location: holland

#2268 Post by fredx181 »

Toni wrote:BTW what is yours compress initrd.xz command?
I use this:

Code: Select all

find . -print | cpio -o -H newc 2>/dev/null | xz -f --extreme --check=crc32 > ../initrd.xz
Don't know if it's any better, just found it somewhere and it works.

WOW, Three-in-one!

Fred

User avatar
fredx181
Posts: 4448
Joined: Wed 11 Dec 2013, 12:37
Location: holland

#2269 Post by fredx181 »

Hi All
Because of future setup of DebianDog with changes for porteus-boot here's mod v4 of Terry's magic: debdog-install.

Changes/additions:
- Removed warning in case of FAT fs, not needed anymore with new setup without symlink.
- Added in drive/partition list a column "Label"
- Changed layout a little, mostly larger.
- Added automatic set boot flag on boot partition, using parted.
- Changed menu.lst entry for porteus-boot (without 'from' parameter, NOTE: this will only work on the future DebianDog setup, which will have porteus-boot-setup completely changed)

Attached: debdog-inst-mod4.zip

Fred
Attachments
debdog-inst-mod4.zip
debdog-inst-mod4
(181.91 KiB) Downloaded 184 times

User avatar
saintless
Posts: 3862
Joined: Sat 11 Jun 2011, 13:43
Location: Bulgaria

#2270 Post by saintless »

fredx181 wrote:here's mod v4 of Terry's magic: debdog-install
Thank you, Fred, Terry!

I will start making the last changes for remaster tomorrow and if there is no objections in the next two days I will try to upload the beta iso version.

Toni

Post Reply