Unable to start Ubuntu 12.04 live CD with syslinux loader on Pentium M x86 Laptop due to bug in PAE kernel, initramfs or syslinux

[BubbaJ]

Ubuntu 12.04 doesn't start from Desktop CD or USB with syslinux boot loader on Pentium M 1.6Ghz or faster Pentium M CPU - displays error message about missing PAE feature in CPU, but *the same* *Ubuntu 12.04* Desktop CD/LiveUSB starts fine on *the same CPU* (and same PAE kernel) if GRUB boot loader is used, for example when WUBI or LiveUSB with GRUB boot loader, like Multisystem (http://liveusb.info/dotclear/index.php?pages/install ) is used!

The error message is:
"This kernel requires the following features not present on the CPU: pae.
Unable to boot - please use a kernel appropriate for you CPU."

THIS IS AN IMPORTANT REGRESSION! People are able to install and successfully use Ubuntu 12.04 on such pretty new hardware, like IBM Thinkpad T42 laptop with Pentium M 1700Mhz processor, but the bug in syslinux (or something related) forbids Ubuntu 12.04 installation.
This bug is reproducible on lots of computers, there are several log files and /proc/cpuinfo file attached to this bugreport, AFAIK it's enough to reopen this bug.

---
ApportVersion: 1.23-0ubuntu4
Architecture: i386
CurrentDmesg: Error: command ['sh', '-c', 'dmesg | comm -13 --nocheck-order /var/log/dmesg -'] failed with exit code 1: comm: /var/log/dmesg: Permission denied
MachineType: IBM 2373PPU
dmi.bios.date: 06/18/2007
dmi.bios.vendor: IBM
dmi.bios.version: 1RETDRWW (3.23 )
dmi.board.name: 2373PPU
dmi.board.vendor: IBM
dmi.board.version: Not Available
dmi.chassis.asset.tag: No Asset Information
dmi.chassis.type: 10
dmi.chassis.vendor: IBM
dmi.chassis.version: Not Available
dmi.modalias: dmi:bvnIBM:bvr1RETDRWW(3.23):bd06/18/2007:svnIBM:pn2373PPU:pvrThinkPadT42:rvnIBM:rn2373PPU:rvrNotAvailable:cvnIBM:ct10:cvrNotAvailable:
dmi.product.name: 2373PPU
dmi.product.version: ThinkPad T42
dmi.sys.vendor: IBM

[Brad Figg (brad-figg)]

This bug is missing log files that will aid in diagnosing the problem. From a terminal window please run:

apport-collect 930447

and then change the status of the bug to 'Confirmed'.

If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

[Brad Figg (brad-figg)]

Thank you for taking the time to file a bug report on this issue.

However, given the number of bugs that the Kernel Team receives during any development cycle it is impossible for us to review them all. Therefore, we occasionally resort to using automated bots to request further testing. This is such a request.

We have noted that there is a newer version of the development kernel than the one you last tested when this issue was found. Please test again with the newer kernel and indicate in the bug if this issue still exists or not.

You can update to the latest development kernel by simply running the following commands in a terminal window:

    sudo apt-get update
    sudo apt-get upgrade

If the bug still exists, change the bug status from Incomplete to Confirmed. If the bug no longer exists, change the bug status from Incomplete to Fix Released.

If you want this bot to quit automatically requesting kernel tests, add a tag named: bot-stop-nagging.

 Thank you for your help, we really do appreciate it.

[Erick Brunzell (lbsolost)]

I changed the title/description of this bug report based on knowledge gathered here:

https://answers.launchpad.net/ubuntu/+source/linux-meta/+question/187166

and also at bug 897786.

While I personally have no hardware that won't run the pae kernel a few people have reported the inability to install the i386 Precise Alpha 2 image due to the change to pae, generally with an error like "kernel requires feature pae and won't boot".

Sadly most people that are effected by a bug just complain and won't take the time to actually report it. In this case BubbaJ was the first person I managed to convince to report it so you could collect a hardware profile.

You might also find this info helpful:

http://ubuntuforums.org/showpost.php?p=11661116&postcount=1

I've since learned that Lubuntu will not be rebuilding with the non-pae kernel, so those wanting a fresh install of Precise will need to either use the non-pae mini.iso (which can be troublesome) or upgrade from Oneiric.

I find this problematic and not truly in the spirit of "supporting non-pae throughout the Precise life cycle, therefore providing another 5 years of support"!

[Erick Brunzell (lbsolost)]

BubbaJ,

I'm subscribed now. If running:

apport-collect 930447

fails to collect a hardware profile for your machine I'll provide you with another method of doing so.

[jerrylamos (jerrylamos)]

IBM Thinkpad T40 has Pentium M, pae is not listed in the flags.

Some precise pangolin daily builds complain and won't boot example 20120204.

lubuntu precise pangolin 20120207 did boot, shows generic-pae files in /boot, pae-y shows up when asked,
but in spite of all that is running fine. For an Alpha 2 of course.
Now this is a 2003 design with 1.5 gHz Pentium M and 768 MB of storage....

I tried an even older IBM Thinkpad R31 with Intel Celeron, however it does list pae in the flags and does run 20120207 lubuntu precise daily build...512 MB storage, 1 gHz, noticeably jerky on flash video because of old intel graphics processor won't run Compiz since Intrepid no concern to me because Unity-2D and also lubuntu don't use compiz.

?

Jerry

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

apport information

[BubbaJ (azstuenthome)]

I've posted the apport-collect data as requested. Please note that I ran the command while logged into the laptop using the existing 11.10 install. I am not able to run the command in the 12.04 environment because I am not able to boot from the 12.04 live usb iso. Hence the need for the bug report.

[Jan Claeys (janc)]

Seems like our locoteam has at least 5 or 6 Thinkpad R51 with a 1.7 GHz Pentium M affected by this issue. They come with ATI graphics capable of running Unity 3D (and Compiz) on open source drivers, so I find it a bit surprising that this harware is now suddenly considered "obsolete"…

[Ursache Dogariu Daniel (danniel)]

I have the same problem with an Asus EeePC 701 (which has a Celeron M processor).

[Alan Pope 🍺🐧🐱 🦄 (popey)]

I have just been hit by this on a Thinkpad X40. Filed bug 930778 for it.

[BubbaJ (azstuenthome)]

I thought I would add some additional context to the issue. In my day job, I work for a US company with near $8B in annual revenue. We probably still have hundreds of Thinkpads still in the field. While they may be close to end of life for us, they are likely to become beginning of life for others. Not everyone can afford to replace hardware that still has plenty of useful life. The main reason I switched to Ubuntu in the first place was to improve the performance (compared to WinXP Pro) on the same hardware. I have not been disappointed, until now. If one of the goals of 12.04 is to establish a long-term stable platform for large corporations to adopt Ubuntu, it would seem like a smart idea not to alienate those organizations with large numbers of Pentium M laptops still in the field.

[Nils (nils-fishtown)]

My HP Desktop has the same problem with processor Intel® Core™2 Duo CPU E8400 @ 3.00GHz × 2 and Intel® Q45/Q43 x86/MMX/SSE2

[jerrylamos (jerrylamos)]

pae kernel running fine on my non-pae flag IBM Thinkpad T40 1.5 gHz Pentium M with Lubuntu precise pangolin 20120207 daily build install, and as daily updated since. I haven't tried Live or install since Feb 7 so I don't know if the

-in my case, totally unnecessary check for pae-

is back into the daily builds or not.

jerry@ThinkPad-T40:~$ ls /boot
abi-3.2.0-15-generic-pae grub memtest86+.bin System.map-3.2.0-15-generic-pae
config-3.2.0-15-generic-pae initrd.img-3.2.0-15-generic-pae memtest86+_multiboot.bin vmlinuz-3.2.0-15-generic-pae

jerry@ThinkPad-T40:~$ uname -a
Linux ThinkPad-T40 3.2.0-15-generic-pae #24-Ubuntu SMP Tue Feb 7 23:30:35 UTC 2012 i686 i686 i386 GNU/Linux

flags: fpu vme de pse tsc msr mce cx8 mtrr pge mca cmov clflush dts acpi mmx fxsr sse sse2 tm pbe up bts est tm2

jerry@ThinkPad-T40:~$ grep PAE /boot/config-3.2.0-15-generic-pae
CONFIG_X86_PAE=y

No, it's ATI doesn't do unity-3D which is a non-problem to me since I even run -2D on my notebook, netbook, tower that will run -3D. I don't like big fuzzy out of focus graphics. Now when HUD comes around I'll test that.

My experience I found a couple precise pangolin daily builds that checked for pae flag and refused to boot.

Perhaps the 20120207 doesn't check for the flag - so it goes on and runs live CD and installs and runs fine??

Did development test precise pangolin generic-pae would really fail?? Works fine on my T40.

"unable to boot" only becaue of the test for the pae flag?? Any way to convince development into changing the test to a

"warning, pae flag not present, may not run, do you want to continue Y/n"?? I'd say Y. Alpha/Beta/RC testing is risky anyway.

Anything I can test or report on for anyone?? (I'm not a developer, I don't do kernel "make".)

Jerry

[Joseph Salisbury (jsalisbury)]

12.04 is the last release that will support non-PAE. The default (and boot) kernel for 12.04 32 bit is now the PAE kernel, so there is no easy way to install a non-PAE kernel.

An upgrade path from prior releases is the only supported option at this time.

[Erick Brunzell (lbsolost)]

Adding just a bit to what Joseph Salisbury said, another option is the non-pae mini.iso:

http://www.us.archive.ubuntu.com/ubuntu/dists/precise/main/installer-i386/current/images/netboot/non-pae/

I'm performing some testing of the same and tracking the results here:

http://ubuntuforums.org/showpost.php?p=11684152&postcount=1

[BubbaJ (azstuenthome)]

So I guess we have a definition of what "support" means for non-PAE kernel. It sounds like it means that you are on your own for completing a fresh install of 12.04 from the image file on the download page. New users will be totally lost as they'll likely never figure out that they have to somehow find the 11.10 image, install that, then perform an upgrade to 12.04. Or somehow find the non-PAE mini.iso and build a system from that. Just disappointing as it doesn't live up to the easy to install system of the people mission.

[jerrylamos (jerrylamos)]

mini.iso, pae version, booted O.K. on my IBM Thinkpad T40 which does not have the pae flag. To do this, I added the following to the file /etc/grub.d/40_custom being careful to leave what was already in the file

menuentry "pangolin mini" {
set isofile="/boot/iso/minipae.iso"
loopback loop (hd0,1)$isofile
linux (loop)/linux boot=/ iso-scan/filename=$isofile noprompt noeject
initrd (loop)/initrd.gz
}

As you see above, I have the .iso in a folder /boot/iso. Your option. sudo update-grub and reboot the mini.

mini booted O.K. directly from the hard drive, no problem with the missing flag.

As it installed, messages were 3.2.0-15-generic-pae however after booting it up O.K. the boot and /proc/version is 3.2.0.15-generic which is running fine.

Well, I'm used to a whole lubuntu (or unity) setup so there's a lot missing.

What's running on the T40 now is (Lucid Lynx of course), three precise pangolins:

One updated from Oneiric, non-pae

One new install from 20120207 which is generic-pae not supposed to run but does. My opinion, the test for the flag is missing/doesn't work so it goes on and boots and runs O.K.

One mini install from the pae mini but did result in non-pae generic. A whole lot of accesories & tools missing but did just get synaptic and even firefox up and running.

Oh, they're all lubuntu's. I've got unity on other test pc's.

Having fun,

Jerry

[Erick Brunzell (lbsolost)]

BubbaJ,

It's never hard to find supported Ubuntu images:

http://cdimage.ubuntu.com/

For pure Ubuntu click on releases/:

http://cdimage.ubuntu.com/releases/

It's as handy as having pockets, and easy to remember "cdimage-dot-ubuntu" ;^)

[Erick Brunzell (lbsolost)]

I just thought to add one more thing. There will actually be two official upgrade paths to Precise (12.04), you'll be able to upgrade from either Lucid (10.04) or Oneiric (11.10).

I happened to think of this because I'm testing potential 10.04.4 i386 images right now, the official release date for 10.04.4 is February 16th.

[jerrylamos (jerrylamos)]

Has anyone who had failures to boot precise pangolin in the past because of pae tried lately? Here's my results today:

jerry@ThinkPad-T40:~$ cat /proc/version
Linux version 3.2.0-16-generic-pae (buildd@zirconium) (gcc version 4.6.2 (Ubuntu/Linaro 4.6.2-14ubuntu1) ) #25-Ubuntu SMP Tue Feb 14 04:00:45 UTC 2012

Installed from
http://cdimage.ubuntu.com/lubuntu/da...sktop-i386.iso

precise-desktop-i386.iso 14-Feb-2012 17:25 554M Desktop CD for PC (Intel x86) computers (standard download)

jerry@ThinkPad-T40:~$ grep PAE /boot/config-3.2.0-16-generic-pae
CONFIG_X86_PAE=y

model name : Intel(R) Pentium(R) M processor 1500MHz
flags : fpu vme de pse tsc msr mce cx8 mtrr pge mca cmov clflush dts acpi mmx fxsr sse sse2 tm pbe up bts est tm2

synaptic crash on reloading after adding partners to repositories, launchpad bug report started.

Otherwise running fine, removed chromium, added firefox & flash, ... usual install goodies.

Jerry

[eexpress (eexpress)]

same here.

i do hard to upgrade frome 10.04 to 12.04. and system now dead.

after down iso. and occurred this.

[eexpress (eexpress)]

Asus EeePC 701

[Henry Wertz (hwertz)]

     I know that this is a bug report and not a forum. But my 2 cents... First, it's a damn shame the Pentium M shipped without PAE, otherwise this would not really be as much of an issue. I don't use my Pentium M any more but it was a great machine, and I am planning to give it to someone to use shortly. Other than lack of PAE it's a significant upgrade for them...

     1) Regarding the argument that people should just upgrade hardware... this isn't the Linux way. Linux (and UNIX before that in general...) let people keep using their system until it was either too slow, or they couldn't cram enough RAM into it, to run current software. Not forced obsolescence. I would say, if anything, those running 4GB+ of RAM should be encouraged to get a 64-bit system, rather than expecting people who have already shown they are not interested in the upgrade treadmill (by having old non-PAE systems) to replace a working system.

     2) 2 constructive solutions. I would ship the CD with the non-PAE kernel, then have a little something similar to the "Additional Drivers" app that runs on first boot and presents the option of installing a PAE kernel on PAE-capable systems. (As opposed to the current situation where I'm sure some users miss out on PAE's benefits since they have to know to choose it in the package manager, this would allow a 1-click choice.)

      Barring that, I do urge Canonical et al. to at least have an alternate non-PAE build available (at least once it's out of alpha), perhaps just note that Pentium M users should select this build. I'd guess there are more Ubuntu users with Pentium Ms than users of the Ubuntu ARM builds (as much as I'd love to have an ARM notebook, I don't...)

[Erick Brunzell (lbsolost)]

Henry,

I subscribed you in hopes you'll read this. There should be three methods to install Ubuntu i386 with a non-pae kernel:

(1) Upgrade from Oneiric.

(2) Upgrade from Lucid.

(3) Install using the non-pae mini.iso:

http://www.us.archive.ubuntu.com/ubuntu/dists/precise/main/installer-i386/current/images/netboot/non-pae/

I've been testing and I can assure you that all three are possible, but we're still in Alpha stage so things are buggy. You'll notice that "eexpress" doesn't mention what failed in the Lucid -> Precise upgrade.

I'll be testing the Lucid -> Precise upgrade path in about one week, I do know there are a few problems, but some are getting fixed almost daily ............ really ;^)

And using the mini.iso is not bad at all.

[Peter Hurley (phurley)]

@Erick Brunzell

Option (2) doesn't work. I just tried to upgrade from completely stock, just-installed, just-formatted Lucid install - install completes but no kernel is installed :(

Bug report here: https://bugs.launchpad.net/ubuntu/+source/linux-meta/+bug/941146

[Erick Brunzell (lbsolost)]

@ Peter Hurley and all concerned,

I'm not sure what could have gone wrong. I have no hardware that actually requires a non-pae kernel but I performed upgrade tests from desktop versions of both Oneiric and Lucid today and both remained non-pae. I simply used the command "update-manager -d -c" in both instances.

The Lucid was a fresh install of 10.04.4 and naturally began with an insignificant number of updates to Lucid itself before proceeding with the dist-upgrade, but all went well afterward. I did however file a bug report about wording of one of the dialogue prompts ....... minor, but I'm a whiner.

The Oneiric install was naturally a different story, still a fresh install but it wanted 341 updates before proceeding with the upgrade, so I edited the software sources to require only "important security updates" and reloaded which brought the requirement down to only 81 updates. Then a restart was required, following that I once again ran "update-manager -d -c" and the dist-upgrade completed just fine.

Note: Once Precise Beta 1 is released, or actually after each milestone release, you'd then want to drop the "-c" and rather just use the command "update-manager -d", but right now that would only give me Alpha 2 which was quite buggy for me, and it would of course require a sled load of updates.

I've never done any server work so I'm not sure how the procedure for the server edition might vary.

Testing is still ongoing with the non-pae mini.iso as I've requested some changes to 'tasksel'. I may not get around to retesting the mini.iso for nearly a week though.

I hope this might be helpful to someone.

[ImConfused (ed-dori-robinov)]

There should be a place to poll the people who have CPUs needing the non PAE support. (I do). I will not be able to upgrade after my operating system becomes unsupported.

[David Henningsson (diwic)]

I just created a custom Live-CD, based on 12.04 beta1, with the only difference that it boots and installs a non-pae kernel instead of the pae one.

It's available here for those who want it:
http://people.canonical.com/~diwic/12.04-nonpae/

Note that this is work made by somebody who is not a LiveCD expert by any means, never made a custom LiveCD before, and merely followed the instructions at https://help.ubuntu.com/community/LiveCDCustomization - you have been warned. :-)

[Erick Brunzell (lbsolost)]

David Henningsson,

You did a great job! I have no effected hardware but I tried that iso image on a hardware rebuild with Win XP that I use for testing Wubi and everything seemed to work great.

Now that I know installation works properly I'll try to set up a partition on another test box so I can more closely monitor kernel upgrades to be sure that non-pae remains non-pae.

Many, many thanks. IMHO this should provide a decent work-around for those Pentium M/Thinkpad users that lack a wired connection.

I'll try to follow up testing future images if you produce them and I asked for more testing on the forums:

http://ubuntuforums.org/showthread.php?t=1937376

Your efforts are appreciated.

[David Henningsson (diwic)]

@Erick, thanks. I've been using the installation for a few days (as installed by the Live-CD I made), and noticed that you might want to remove the linux-headers-generic-pae package - while having it there is unlikely to cause any problems, it's just a minor annoyance to have something installed that you don't need.

Also let me restate that even though I work for Canonical, this image is as unofficial and unsupported as it possibly can be - I did it just because I have one of those laptops, I like it, and I want it to stay alive for five more years or so. :-)

[JohnWashington (ubuntu-johnwash)]

Erick, I agree David has done a great job, but does it really provide a decent workaround for the real world? In the real world 99 out of 100 people trying out Linux (often people who are doing so because a proprietary OS has become too slow and bloated, so they've been encouraged to give Linux a try) those people download the standard CD. If it doesn't work, there's little chance these people will be aware of what's wrong and how to get a workaround.

#39 Henry really hits the nail on the head. Isn't the second part of his post the best way forward, for everyone?

[BubbaJ (azstuenthome)]

David, thanks for taking the initiative to build the Live CD. It proves the old adage, "necessity is the mother of invention". I took a look at the LiveCD Customization page and it looked a little daunting to me. If you were able to follow it without prior experience and were able to get it to work, imagine what someone who had experience in building it could achieve.

I guess it begs the question, how much of an inconvenience would it be for the Ubuntu team to build two versions of the i386 CD (pae and non-pae), and have both available for download from the Ubuntu daily build download page. Is it really that difficult, or does it really create that much of an extra burden to support?

[Erick Brunzell (lbsolost)]

@ David, regarding the 'linux-headers' issue someone testing Lubuntu even filed bug 937248, and if you look at an Ubuntu daily i386 manifest it shows the following:

linux-generic-pae 3.2.0.18.20
linux-headers-3.2.0-18 3.2.0-18.28
linux-headers-3.2.0-18-generic-pae 3.2.0-18.28
linux-headers-generic-pae 3.2.0.18.20
linux-image-3.2.0-18-generic-pae 3.2.0-18.28
linux-image-generic-pae 3.2.0.18.20
linux-libc-dev 3.2.0-18.28

Again not a problem for me but the Lubuntu devs are always concerned about each and every wasted byte of disc space ;^)

[Erick Brunzell (lbsolost)]

I'm really sticking my neck out here, I subscribed Colin Watson which is totally rude and I hope he'll forgive me :^)

I was so clueless about the change to the pae enabled kernel that I had to ask a question at the forums. All of my hardware works great with "pae" and I wouldn't even have noticed except for the new suffix in grub. But that forum post became a bit of a lightning rod so I followed it up with two questions at Ubuntu answers. So even though I'm not directly effected I've now thrown myself into the fray, testing the non-pae mini.iso, upgrade options, etc.

So my question to Colin is; according to the minutes of the Ubuntu Technical Board meeting on Dec 12, 2011 you said, "Switch precise over to PAE kernel by default on i386; we retain the option to revert if it causes too much fallout (Colin)".

Now obviously I want to know if we've yet seen enough "fallout"?

I think we must seriously consider that question based on the following facts:

#1: Both pae and non-pae kernel versions are going to be supported throughout the 5 year Precise life cycle anyway, and those who desire a PAE-enabled kernel can easily install it post-installation.

#2: The only official methods to install non-pae Precise are:

(a) Install Lucid (10.04.4), install all available updates, then upgrade to Precise. Do note however this is not likely to work for Lubuntu.

(b) Install Oneiric (11.10), install all available updates, then upgrade to Precise. This should work for all official derivatives.

(c) Use the non-pae mini.iso, aka; netboot image, and then install the desired DE.

#3: I have a 5000kbps+ wired connection and any of the three above options are time consuming. Depending on which DE I use a mini.iso install takes about 1 hour and 20 minutes to 1 hour and 45 minutes to complete.

Either of the install + update + dist-upgrade options are even more time consuming, Oneiric more so since Lucid just had a point-release (10.04.4).

#4: More important than just time required is the fact that most, if not all, of those reporting the inability to install Precise due to the lack of pae support are laptop users many of whom lack access to a wired connection which is almost a requirement for a netboot/mini.iso install, and certainly preferable for dist-upgrades.

#5: It's nearly impossible to determine how many users are effected at this point but based on PM's, bug reports, mailing lists, etc. I'd put it at no less than 2 dozen .......... and remember we're just in Beta 1.

#6: David Henningsson has already produced the first "hybrid" Ubuntu image, and Greg Faith has reported on the Lubuntu mailing list that he's managed to create an Lubuntu hybrid image. While I applaud their efforts, how much difficulty does this present for the devs when it comes to squashing unrelated bugs? I'd think we can be sure that others will create their own hybrid images for other DE's and such.

#7: I understand that the Precise default download will be 64 bit anyway, and most hardware capable of 4GB+ RAM will run a 64 bit system so why default to the PAE kernel on i386? Isn't this a bit like rubbing salt in a wound?

#8: We're running out of time to revert ........ maybe we've al...

[Erick Brunzell (lbsolost)]

I just wanted to correct any FUD I may have started. First of all Jeremy Bicha tells me that I'm mistaken about the 64 bit version of Ubuntu being the default download, but more importantly, while some have reported the inability to install the mini.iso due to the lack of wireless support (eg: bug 945053) at least one person found a solution for themselves:

http://ubuntuforums.org/showpost.php?p=11754050&postcount=20

I'm also working on a method to install using netboot, but doing so from a live CD using a chroot. I don't quite have it worked out yet but I think it should be feasible using the initrd.gz and linux files from "ubuntu-installer" listed with the netboot images. If not it's great fun trying anyway ;^)

[Luca Lesinigo (luca404)]

I'll add my experience, hope this helps. Ultra short version: add many Mini-ITX EPIA boards to the list of affected systems, and possibly remove the Asus EeePC 701 from the same list.

The Via EPIA EK-10000G board with a Nehemiah 1GHz CPU does not have PAE, but having two ethernet ports and onboard Serial-ATA it's still a useful system. Of course it fails booting from the current 12.04 standard netboot image, but boots and installs just fine from the current 12.04 non-pae netboot image. The real problem on this board is that with the default vga=something command line option (from the netboot image) you won't see the kernel dmesg, so you won't see the "This kernel requires the following features not present on the CPU: pae" message either and will just keep staring at a blank screen. Since this is affected by the VGA, I cannot confirm if a different monitor would work better (but I can confirm that the non-pae image will show up the installer fine with the same vga= option on my system).
I sill have around some slightly older EPIA M10000 boards that should use the same CPU and be affected by the same non-pae issue.

@lbsolost: I can confirm that the non-pae netboot image (item #2.C of your list ;) works perfectly on my non-PAE EPIA EK-10000G (I choose, as usual, the Ubuntu Server environment, and it installed and rebooted correctly on the non-pae kernel image). I use that as a low-wattage server and it's great to be able to run 12.04 LTS on the system. I guess its real usefulness will come to an end before Precise Pangolin end of life :)

@danniel and @eexpress
My Asus EeePC 701 (the first model, with 4GB onboard SSD and Celeron-M 900MHz underclocked to ~600MHz) does have PAE, and is currently running Precise installed from the standard netboot image (not the non-pae one). It is actually running the non-pae kernel but I can't remember if the installer choose it automatically or if it let me choose. You should see the 'pae' flag in /proc/cpuinfo even if you're running a non-pae kernel.
Anyway, I manually installed the PAE kernel and can confirm it can boot just fine too.
IIRC, the CPU inside the 701 should be a Celeron M ULV 353: http://ark.intel.com/products/27157

[Erick Brunzell (lbsolost)]

I've been holding off on updating this as far as installation options until Beta 2 because it obviously takes time for any change to be completed. But we do now actually have two new options for installing a non-pae Precise distro.

Both Lubuntu and Xubuntu are going non-pae, please see bug 955009 and bug 958866, but also the new alternate images are not yet working (re: bug 961218). So please just be patient.

I think this will create a more than acceptable balance for non-pae options in Precise.

[BubbaJ (azstuenthome)]

While it is interesting that there is a method to install a non-pae Precise distro (aka Lubuntu and Xubuntu), this bug was filed to report the fact that it is not possible to install "Ubuntu 12.04". I am personally not interested in installing other Precise variants as it would mean that my family, who currently use "Ubuntu 11.10", would have to learn to use a different desktop UI and base applications. I do not view installing other Ubuntu variants as a viable resolution to this bug. We should probably be consistent to communicate only alternatives that are related to "Ubuntu" specifically.

[Greg Faith (gregfaith)]

Here is a thought you install Lubuntu desktop i386 from live CD or USB then
install ubuntu-desktop ..
You will get the non-pae and then get the Ubuntu Desktop your family uses.
nm_geo Greg

On Thu, Mar 22, 2012 at 6:37 PM, BubbaJ <email address hidden> wrote:

> While it is interesting that there is a method to install a non-pae
> Precise distro (aka Lubuntu and Xubuntu), this bug was filed to report
> the fact that it is not possible to install "Ubuntu 12.04". I am
> personally not interested in installing other Precise variants as it
> would mean that my family, who currently use "Ubuntu 11.10", would have
> to learn to use a different desktop UI and base applications. I do not
> view installing other Ubuntu variants as a viable resolution to this
> bug. We should probably be consistent to communicate only alternatives
> that are related to "Ubuntu" specifically.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/930447
>
> Title:
> Unable to Install Ubuntu 12.04 on Pentium M x86 Laptop due to PAE
> kernel
>
> Status in “linux” package in Ubuntu:
> Won't Fix
>
> Bug description:
> I recently downloaded and built a live usb of the latest 12.04 daily
> build iso for i386 hardware using the following link address:
> http://cdimage.ubuntu.com/daily-live/current/precise-desktop-i386.iso.
> I was not able to boot from the live usb due to the incompatibility of
> the linux kernel on the 12.04 iso and my Pentium M laptop hardware.
>
> The error message I received is as follows: "This kernel requires the
> following features not present on the CPU: pae. Unable to boot -
> please use a kernel appropriate for you CPU." I expected that I would
> be able to install Ubuntu 12.04 on my IBM Thinkpad T42 laptop that has
> a Pentium M processor, 1700Mhz processor. The root cause of the issue
> appears to be that the default kernel bundled in the current daily
> build i386 iso is the generic-pae kernel. It's my understanding that
> the generic-pae kernel cannot be installed on cpu hardware that does
> not support pae.
>
> I have been using Ubuntu since 10.10. I have been able to install 10.10,
> 11.04, and 11.10 on my Thinkpad without issue. My understanding of the
> December Tech Board kernel meeting is that the non-pae kernel will be
> supported until 12.10. If that's the case, there should be a version of the
> Ubuntu 12.04 iso available for download that can be loaded on non-pae
> hardware.
> ---
> AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.24.
> ApportVersion: 1.23-0ubuntu4
> Architecture: i386
> AudioDevicesInUse:
> USER PID ACCESS COMMAND
> /dev/snd/controlC0: debbie 9252 F.... pulseaudio
> Card0.Amixer.info:
> Card hw:0 'I82801DBICH4'/'Intel 82801DB-ICH4 with AD1981B at irq 11'
> Mixer name : 'Analog Devices AD1981B'
> Components : 'AC97a:41445374'
> Controls : 26
> Simple ctrls : 18
> Card29.Amixer.info:
> Card hw:29 'ThinkPadEC'/'ThinkPad Console Audio Control at EC reg 0x30,
> fw 1RHT71WW-3.04'
> Mixer name : 'ThinkPad EC 1RHT71WW-3.04'
> Components : ''
> Controls ...

[JohnWashington (ubuntu-johnwash)]

@ Greg Faith, #55
Yours is a thought, and it's a helpful thought, but it's a thought directed at helping someone who knows what they're doing and is looking for a workaround. Ubuntu aims at inclusivity, geeks who know where to find answers can use workarounds, but for every geek there are many non-geeks who try this alternative to Windows they've heard about and will bail out in disgust at the first hurdle.

@Erick, #53
Will the ubuntu installer just fall over, or will it detect the lack of PAE and give sensible advice, such as pointers to the variants you mention?

My view is unchanged (#47): #39 Henry really hits the nail on the head. Isn't the second part of his post the best way forward, for everyone?

[Erick Brunzell (lbsolost)]

@JohnWashington,

I don't disagree with anything you've said here but I'm just another Ubuntu end user that performs testing and tries to offer work-arounds when possible. Neither Greg or I have any real influence over any decision made by Ubuntu/Canonical but we owe a debt of gratitude to Colin Watson for just keeping the non-pae kernel alive in 12.04.

Once Precise is actually released I'll work on the simplest possible "how-to keep or install non-pae" thread at the forums .............. that's the best I can do. At this point I doubt that any additional complaining will be useful.

[JohnWashington (ubuntu-johnwash)]

@Erick
I understand and I thoroughly agree about gratitude to Colin. I hope my comments aren't going to be taken in any negative way, they're definitely not offered in that spirit. It's just that I know how easy it is for people busy at the coal face to be unaware of what a mountain Joe Public has to climb, and imagine that posting brief instructions on Alpha Centauri "just change the flurgle to a fuffle and blurge the crungebluck to solve this little inconvenience" will get everyone sorted!

Anyhow, time for a happy Easter trying beta 2! :)

[Nick Lowe (nick-int-r)]

The devil is in the detail with regards to the Pentium M.

Some support NX and PAE, others do not:

Those released in 2005 with a 533MHz FSB have PAE and NX support. (Dothan on the Sonoma Platform)
Those released prior with a 400MHz FSB lack PAE and NX support. (Banias and Dothan on the Carmel Platform)

Thanks,

Nick

[Nick Lowe (nick-int-r)]

See http://en.wikipedia.org/wiki/List_of_Intel_Pentium_M_microprocessors

[Elbarbudo (patricearnal)]

The status "won't fix " of this bug is for me the end of Ubuntu which I used from version 7.10 until 11.10.

I warmly recomended Ubuntu to my friends and people in my association.

These people can't afford brand new computers with bells and whistle, every 2 years as Microsoft and other vendors believe, so they mostly use repackaged old machines.

These people are all newbies in computers and the ease of use of Ubuntu is a big plus.

From the beta test I ran on my relatively recent machine, I believe that the Unity version shipped suits best the needs of the newbees, using a small set of applications (mail, Internet, photos and music), but if it's unusable on these older machines, I will need to find other distributions, which makes me rather sad, according the high level of satisfaction I got with Ubuntu.

[Erick Brunzell (lbsolost)]

I typed up a few preliminary notes regarding non-pae options here:

http://ubuntuforums.org/showpost.php?p=11847791&postcount=1

That of course will need to be moved and updated shortly after Precise' final release.

[Christiansen (happylinux)]

What a decision to make, leaving older but fine and working hardware out of acceptable simple install options - couldn't it had waited until pre-precise releases ?.

Unfortunately I've only first become hit by this after the release, as I've been using (testing) Precise on this hardware type (non-PAE capable Intel) since before the beta1 releases and just been upgrading without noticing this change. This leaves me with a several boxes, for which I had decided that Kubuntu Precise should be the final upgrade, while it works like an absolutely charm (best compared to several previous releases including Lucid) and would become an LTS.

Sorry for the late comment, but I had to make a statement - just in hope of a simpler CD/USB ISO install option would come around as more people now probably gets hit by this decision, when trying to install the final Precise release.

[EricDHH (ericdhh)]

Thinkpad T42 ready for trashbin? Needs the non PAE kernel too

[BubbaJ (azstuenthome)]

Don't lose hope. I'm the one that filed the initial bug report. I was able to boot from a Live CD image on the hard drive using Grub 2 (http://ubuntuforums.org/showthread.php?t=1549847&highlight=boot+grub+iso) and the final pae flavor ISO saved to the hard drive. I have no idea why the pae version ISO boots from the hard drive using Grub, but doesn't boot from a USB stick. Anyone else like to explain that I'd be interested in knowing.

Today I tried a clean install using the mini.iso method (http://archive.ubuntu.com/ubuntu/dists/precise/main/installer-i386/current/images/netboot/non-pae/). Just downloaded and burned the mini.iso (non-pae flavor) to a CD, booted using the CD and clicked on "Install" from the menu. Just make sure you have the PC is connected to an internet connection so that the packages can be downloaded. There are quite a few questions to answer, but nothing that wasn't that straightforward. It took close to an hour by the time all the packages were downloaded and installed. But if you figure the time to download the CD image ISO, burn time to a disk, and install time from the disk, it's not that much different overall.

I ended up with a fully functional 12.04 generic non-pae install on my Thinkpad T42. So far, Precise working fine.

[Jane Atkinson (irihapeti)]

You can also install a command-line system using the Xubuntu alternate CD, which gives a non-pae kernel. This doesn't need an internet connection, by the way.

Then reboot, login and enter the commands:

sudo apt-get update
sudo apt-get install ubuntu-desktop

to give a standard Ubuntu install with non-pae kernel. This stage will need an internet connection (or access to an apt-cacher-ng server), of course.

(I figure that there ought to be some way to use the Ubuntu alternate CD for this stage, but haven't worked that out yet. That would mean needing an internet connection only for the final updates, which might be useful for people whose connections are a bit uncertain or slow.)

Though I haven't tried it, I assume that one could instead do:

sudo apt-get install kubuntu-desktop

[Luiz Angelo Daros de Luca (luizluca)]

C'mon guys... EeePC 700 is just 5 years old and have no PAE. All EeePC 700 and most of 900(newer than 700) are Celerom M, with no PAE.

Please, it is not that much difficult to keep non-pae in build tree and a alternative non-pae livecd.

I know lots of people that deals with old hardware that are suffering from the lack of non-pae in ubuntu. LTS is the ideal version for digital inclusion center that runs with old and donated hardware.

[dalmolin (dalmolin-e-cology)]

As an owner of a rock solid, reliable Thinkpad T42p that is my everyday business computer and far from obsolete + over 14 years of commitment and experience in open source I am absolutely baffled by the path Canonical/Ubuntu has chosen to deal with non-pae hardware. Frankly my jaw dropped when I tried to launch the live CD and it didn't work. I have spent several hours searching for workarounds etc. so that I can give Precise/Unity a whirl on my laptop and frankly while there are some quasi-viable options I feel that Canonical/Ubuntu have really dropped the ball on this issue... especially given Mark Shuttleworth's goal of having 10 times the current installed base. I have to echo what others have said a) the early obsoleting of non-pae hardware is not consistent with Ubuntu's or Linux philosophy b) option 2 in post 39 makes sense c) there is a far larger installed base and potential installed base that will be affected by this. What worries me more than my own circumstances is that this particular decision represents the a trend to being out of touch with the customer.... one would expect this from Redmond and not from a Linux distribution... and especially not from Canonical/Ubuntu. Canonical/Ubuntu do the right thing... make it easy to help you sell the next generation of Ubuntu... and give non-pae machines the right to die a natural death.

[Luiz Angelo Daros de Luca (luizluca)]

Hello guys,

I replaced the kernel in Livecd for i586 with the one that supports non-pae processor. It installs and update normally. Just download and be happy.

http://dl.dropbox.com/u/1286502/ubuntu-12.04-desktop-i386-nonpae.iso
MD5SUM 7435d1d1740dcf16fbbba2e746de4de1
SHA1SUM 4fb527b2e2c6a00499677102df36ec466e79c373
SHA256SUM a8e4427889fa6786e4cfe82ad578fcbf833fe66146a180f90abcc5caa6fe6dbb

[bobbyjean58 (bobbyjean58)]

Thanks Luiz,
Downloading as I type this. :). Looking forward to trying trying it.

Thanks for your efforts. Many will appreciate this I'm sure.

Again, thanks.

I will report back on my Sony S360 nonpae install experience.

Later...

[JohnWashington (ubuntu-johnwash)]

Thanks Luiz. This kind of helpful support is one of the things I love about the Linux community. Yet I can't help thinking that another of the things that I love about most Linux distros is the added confidence one gets when installing software from known repositories, a big contrast with Windows, where one goes to arbitrary sites for applications, hoping there's no malware included.

It would not be difficult for an evil-minded person to generate an unofficial .iso, include a trojan, and post a link to it in forums or on Launchpad.

This solution needs to be officially provided by Canonical. And prominently featured on the download page.

[bobbyjean58 (bobbyjean58)]

Luiz,
Your ubuntu 12.04 nonpae iso works great. Very impressive. Installed on my Sony S360 nonpae 1.7 ghz laptop with no problems at all. This gives new life to my old laptop.

Thanks, dude. :).

Later....

[B. Cavin Lang (cavinlang)]

Thanks Luiz.
Your ubuntu 12.04 nonpae iso works great! On my Dell Latitude D600 with Pentium M CPU.

Thanks for the great work...

-- Cavin

[Veleno (veleno)]

I firmly second #68 and #71 comments, I'm in the same situation of dalmolin.
Indeed the upgrade process works fine, I'm running 12.04 upgraded from 11.10, but a fresh install is always better.

[David Henningsson (diwic)]

Hi,

I made a nonpae iso for beta 1 (see comment #44) and was planning to do the same for the final release today, but now I see Luiz has already made one.
I haven't tested Luiz's iso, but is there any point in me essentially redoing the same thing he has already done and putting it up on people.canonical.com? It would still be unofficial.

[Luiz Angelo Daros de Luca (luizluca)]

For those who like my LiveCD, I'm very happy that you liked it.

@David Henningsson (diwic),

Your work with beta1 inspired what I did. Thanks. If you managed to do it with beta1, it would certainly work with the final version. It was just a matter of redoing your job. As the ISO did not appeared until yesterday, I did it myself.

@JohnWashington,

I agree that it would be very easy for me to add a malware. However, it would not be too much difficult to check the integrity of my iso. Just diffs the contents of the iso, squashfs and initrd with the official iso. The binary files that might differ are from official ubuntu packages, which can be downloaded and extracted in order to compare with the one that is inside my livecd. This is what is is expected to be modified:

1) in squasfs, the new kernel (and removal of the pae one) and some apt/dpkg transaction files
2) the initrd.lz was generated automatically when I installed the kernel (but in gz format).
3) In isofs, I replaced kernel and initrd with the ones that were present in /boot after installing the new kernel. The initrd was uncompressed and recompressed using lzma. I also needed to change some control files (md5sum, filesystem.size) and the livecd identity.
4) I might have missed something :-)

I really really would appreciate that Canonical would provide a official LiveCD, even just coping mine and validating or doing it from scratch themselves.

[David Henningsson (diwic)]

@Luiz:
The stuff I had to pay a little extra attention to when I made the CD, was that
 1) For some reason grub complanied when I changed kernel inside the chroot. I worked around this by temporarily renaming /usr/sbin/update-grub and then renaming it back.
 2) Change the installed header files from pae to non-pae (I didn't discover this until later, so my beta1 iso contains this error)

An official LiveCD is unlikely to happen. Also, for 12.10, I think the plan is to completely drop the non-pae kernel from the archive. This will likely be decided upon at UDS, in this session: https://blueprints.launchpad.net/ubuntu/+spec/hardware-q-kernel-version-and-flavors

[Jaap Woldringh (jjhwoldringh)]

I am downloading the non-pae .iso, mentioned in # 69, by luizluca, thanks to him for that :) Hope it works.

I want to react to this bug report, without having read all comments (sorry for that) to cast another vote for non pae kernels. My Toshiba A50 laptop has an Intel Pentium M 1.6GHz processor, and is still a very fine laptop, which it would be very much a pity to have to ditch.

I guess there must be quite a number of 32 bit i686 systems around in the world (also in less prosperous places) , that have less than 3 GiB RAM, and which do not need pae at all, which would also slow them down by up to 10%.

Jaap Woldringh

[Luiz Angelo Daros de Luca (luizluca)]

Hello guys,

Dropbox disabled my iso link due to overtraffic. :-)

@David Henningsson (diwic), can you please provide a 12.04 ISO with the non-pae kernel or publish mine in people.canonical.com? About your comments, I had no problem with grub. I just installed the package after mounting /proc /sys /dev and removed the pae version. I didn't undestand the "installed header files". Which "installed header files" are you talking about?

[JohnWashington (ubuntu-johnwash)]

I've just had my second encounter in a week with a Windows user trying out Ubuntu on an older laptop. This one was a Sony Vaio PCG-Z1RSP. I don't know that many folk, if I meet 2 instances in a week, how many millions are stumped by this? Workarounds, forum posts, unofficial CD images... these do NOT solve the issue for first timers. You have one shot, if that fails, they throw away the 'useless' laptop and return to the cosy world of Microsoft.

It's such a slap in the face, announcing a plan that non-pae support will be kept until 12.10, yet denying access to an official live CD that supports non-pae. How many laptops will be needlessly wasted? How many potential converts to Linux will be lost?

I don't know why I'm even writing this, anyone who has the power to change this policy has seen 79 comments and muttered TL;DR.

[bobbyjean58 (bobbyjean58)]

Luiz,
Your Dropbox account is once again active. :).

I also noticed that David Henningsson has created an Ubuntu 12.04 nonpae Final iso, as well....

http://people.canonical.com/~diwic/12.04-nonpae/ubuntu-12.04-desktop-i386-nonpae.iso

Hope it's alright that I posted a link, David.

Thanks for both of your efforts.

Best regards,

Bob Good

[Luiz Angelo Daros de Luca (luizluca)]

Yes, it blocks for a day when the traffic exceeds 10GB/day. So, I have 14 downloads per day :-)

If David version is good, I have no problem to recommend it. At least, people will not have to doubt about the "integrity" of my ISO :-) I would also double check the ISO contents from a stranger (me, in this case) if I wanted to install it in some important computer.

[adonet (jeroen-adolfse)]

I tried to install Ubuntu 12.04 using the Xubuntu alternate CD and installing a textbased installation. After that I did
sudo apt-get update
sudo apt-get install ubuntu-desktop
and a reboot.

And it gave me an Ubuntu 12.04 without pae kernel. But the network manager didn't work. Both WIFI and Wired didn't work. I had to install wicd to get an wired internet connection. The WIFI adapter isn't seen by this xubuntu non pae kernel. Even
sudo lshw
doesn't show the pcmcia wifi adapter. (and it's working in Windows XP though)

Now I'm reinstalling ubuntu 10.04 LTS which was all working fine.

Later on I will try the iso made by Luiz. Thanks for that Luiz.

[David Henningsson (diwic)]

Yep, as already noted, here it is:
http://people.canonical.com/~diwic/12.04-nonpae/ubuntu-12.04-desktop-i386-nonpae.iso

I've only tested the new iso by booting it, as I had no spare partitions on the relevant machine to test installing it as well. Please test installing from it.

And as with the previous one, this is unofficial, made by someone who is not an LiveCD expert (but by someone who just happens to have one of these computers), and I promise no support.

$ md5sum ubuntu-12.04-desktop-i386-nonpae.iso
e857a7ccde6d6a85b2080d285ee48023 ubuntu-12.04-desktop-i386-nonpae.iso
$ sha256sum ubuntu-12.04-desktop-i386-nonpae.iso
047ca71b469a23dce844ba1571a6eae4c64e7e24527449f94d0fb6402676ced1 ubuntu-12.04-desktop-i386-nonpae.iso

[David Henningsson (diwic)]

@Luiz:
The headers are in the linux-headers-generic-pae package and should be replaced with linux-headers-generic package. Not sure why they are needed at all though, but I assume there is a good reason or they wouldn't have been there.

I didn't do the uncompress/recompress of initrd. I wonder if that's why your image is smaller than mine.

Somebody claimed that non-pae was faster or more power saving than pae on computers that can run both (and have less than 4 GB of RAM). From what I've seen, it's the contrary: using pae when available is insignificantly faster or no difference for most workloads, but I don't think there have been enough testing across different hardware (e g non-intel processors) to give a serious verdict either way.

And a final note. I said that whether to completely drop the non-pae kernel for 12.10 was to be decided at UDS next week. It looks like it's already a done deal to do so, judging from this email:
https://lists.ubuntu.com/archives/ubuntu-devel/2012-May/035176.html

[Luiz Angelo Daros de Luca (luizluca)]

@David,

I replaced all -pae packages. So, I guess I also got the kernel headers replaced.
Recompressing the initrd can save some bytes and save me from changing isolinux conf.

I tested multiple squash options but the best was the one that is in wiki (https://help.ubuntu.com/community/LiveCDCustomization):

  sudo mksquashfs edit extract-cd/casper/filesystem.squashfs -b 1048576

The original ISO was very close to the 700MB CD limit. So, I tried hard to get, at least, the previous size.

I tested in a vbox machine and, as the kernel was always nonpae, during and after the installation, it considered it good to go (even running on a pae machine). You can test this way. After that I tested in a real machine but if you got some problem, it might appear even in pae processors.

BTW, bandwitdh limit of my dropbox is over again :-) It says it blocks for 3 days but it was back in less than that. Sorry, I'm a free dropbox user.

[Ian_Wilson (iwilson-mweb)]

I downloaded David Henningsson's Ubuntu 12.04 nonpae iso, but it ended up 748 138, just too much to burn onto a CD. Does anybody have any ideas?
I have 10 Thinkpads with no pae for re-use by under priligeged persons, with possibly more to come, and need Ubuntu urgently

[Ian_Wilson (iwilson-mweb)]

Found free app at www.unetbootin.sourceforge.net which sucessfully created a boot flash drive.

[Luiz Angelo Daros de Luca (luizluca)]

@Ian_Wilson (iwilson-mweb),

If you trust me, you can use mine ISO. It works and it is even smaller than the original pae version.
My dropbox just can't stand too much traffic but I can provide some downloads per day.

[Luiz Angelo Daros de Luca (luizluca)]

I published my ISO with torrent. I might not have a broad uplink but I will let torrent open if someone connects to it.

magnet:?xt=urn:btih:KKUJ7U5Z6XH3RECGK77DLCA32OOFVQVO

[oliver (oliver-schinagl)]

What I can't understand, is why ist he LiveCD broken, but the actual kernel works just fine.

I still am running 10.10 on my T42 (2.1 Ghz Pentium-M sans PAE or NX). I wanted to do a fresh install due to the fact that I have an PATA SSD in it, that does not support the trim command, so a full security clear is required to reset the drive to full performance. Something that I've accepted to live with for the life of this unit. And doing so every 2 years hasn't been so bad for me ;)

Anyhow, I've compiled the 2.6.38.8 kernel on it and ENABLED PAE. Also installing the native 2.6.38.8-PAE kernel on this machine boots just fine! Yes, a kernel that has built in support for PAE works normally, or so it seems. The 'must have pae option' warning (which comes from the kernel itself, cpu.c) does not show. Taking this further, I booted (on a nother box) the LiveCD, added deb-src to the sources file and followed the kernel howto which installed gcc and the whole shebang into my livecd enviroment. I then compiled the 3.2 kernel twice. Once with the config from my current 2.6.38.8 kernel, saying no to all new things and ENABELING PAE. Obviously also this booted just fine. Then using the 3.2 config supplied in /boot in the live enviroment, with PAE also enabled, booted JUST FINE.

So please, what happened here, that the LiveCD kernel that gets booted, does NOT work, yet any kernel I have tried using PAE works just fine on the pentium-m.

P.S. I used to have the same 1.7GHz pentium-m the original bug reporter has and know for a fact, that PAE works just fine there aswell.

Find attached, dmesg.log and config.log, which are | grep -i pae for 2.6.38.8 config, and dmesg_3.log, config_3.log with the 3.2 config. See this as proof that a PAE kernel does work on the pentium-m. I won't speak for pre-pentium-pro cpu's that actually fully lack pae hardware support.

[oliver (oliver-schinagl)]

Note, that I can supply full logs if anybody actually wants to see what else is in there ;) This was purely to show my claims.

[Mantas Kriaučiūnas (mantas)]

This is really bug somewhere, maybe in initramfs or syslinux - Ubuntu 12.04 doesn't start from CD/USB with error message about missing PAE feature in CPU, but works fine with the same CPU and same kernel after installation. Also it seems Ubuntu 12.04 starts fine when the same installation disk is started from GRUB boot loaded, for example when WUBI is used!

I can confirm that oliver (oliver-schinagl) is telling true - yesterday I tried to start Ubuntu 12.04 CD on Fujitsu laptop with Pentium(R) M processor 1.60GHz (model: 13, stepping: 6, cache: 2MB) and got error message "following features not present on the CPU: pae. Unable to boot ..." But the same kernel works fine with the same CPU after installation - it's very easy to check this:
1. I attached hard disk from this laptop with Pentium-M 1.6GHz CPU to another computer with newer CPU and installed Ubuntu 12.04 with PAE kernel in laptop's hard disk.
2. Then I put hard disk back to my laptop with Pentium-M 1.6GHz CPU (without PAE flag, see attached /proc/cpuinfo file) and Ubuntu 12.04 with PAE kernel works fine on laptop, where LiveCD refuses to start! I even installed all updates, including newer PAE kernel and still working without any bugs!

Unity 2D works fine on this computer, Gnome Classic works fine too, I'm using computer for 2 days without any problems, so, PAE kernel works fine and bug exists only on LiveCD/LiveUSB.
Maybe somewhere LiveCD checks for CPU flags and stops to boot if pae flag is missing?

Ubuntu developers, please reopen this bug and fix it - this is an important *regression* for lots of people. If you need any info just tell us, I know several computers where Ubuntu 12.04 doesn't boot from CD because of PAE error, but the same PAE kernel works fine after installation...

I'm attaching /proc/cpuinfo file and uname -a output on the same computer is:

ubuntu@ubuntu:~$ uname -a
Linux ubuntu 3.2.0-24-generic-pae #39-Ubuntu SMP Mon May 21 18:54:21 UTC 2012 i686 i686 i386 GNU/Linux

[oliver (oliver-schinagl)]

The actual error message is from the kernel.

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=arch/x86/boot/cpu.c;h=6ec6bb6e9957a0319259f966c0ac7b2ca4ff69a0;hb=HEAD#l53

The thing I fail to understand, is how this message is triggerd from the LiveCD, but not from the actual kernel. I built the kernel that is in /boot/config-3.2* and could not get this message to trigger. I wouldn't be supprised that the actual booted kernel on the liveCD (/casper/vmlinuz I think it was? (don't have it handy right now) is different from the kernel in the boot partition which is in either the squashfs image or in initrd.gz; If this is the case, the livecd on USB is very easily fixed, by replacing the kernel on it.

[Micah Gersten (micahg)]

The answer is to use a Lubuntu or Xubuntu live image. See the second bullet point here: https://wiki.ubuntu.com/PrecisePangolin/ReleaseNotes/UbuntuDesktop#System_requirements

[Mantas Kriaučiūnas (mantas)]

dmesg output from computer with Pentium M 1.6 Ghz CPU, where Ubuntu 12.04 doesn't start from Desktop CD with syslinux boot loader - displays error message about missing PAE feature in CPU, but *the same* *Ubuntu 12.04* Desktop CD starts fine on *the same CPU* when GRUB boot loader is used!
Next attachments will be /proc/cpuinfo , kern.log ans syslog

[Christiansen (happylinux)]

From what I understand Oliver is saying in #97, the problem is not that a non-pae kernel should be supportet, but that the LiveCD does prevent installation on boxes that run the -pae kerne just fine.

I've successfully installed and used the -pae enabled kernel on a ThinkPad R51, on which I receive the the error message about the the CPU not supporting PAE when running the LiveCD.
A clean installation was first carried out using the the non-pae mini.iso (http://archive.ubuntu.com/ubuntu/dists/precise/main/installer-i386/current/images/netboot/non-pae/mini.iso) to perform a network installation. After the installation and a full online-update was done, the PAE enabled kernel (linux-image-3.2.0-24-generic-pae_3.2.0-24.39_i386) was installed. A reboot and a the non-PAE kernel ( linux-image-3.2.0-24-generic_3.2.0-24.39_i386) was removed.

I've tried (like #96 suggested) to exchange the kernel and initrd in /casper on the LiveCD from USB (by startup-disk creator), with the working -pae kernel from harddisk. It bootet from the USB, but the error message about missing PAE capability was still thrown, why I believe it must be somthing in the squashfs image then.

So the conclusion must be, like that of that the #91, that there is a bug in the LiveCD image, that prevent it from running the PAE enabled kernel even on some PAE capable hardware.

[Mantas Kriaučiūnas (mantas)]

Christiansen (happylinux), problem is not in the squashfs image, but in syslinux boot loader or something related - the same Ubuntu 12.04 squashfs image starts and installs fine if GRUB boot loader is used for startup. Try Multisystem (http://liveusb.info/dotclear/index.php?pages/install ) LiveUSB creator (choose GRUB boot loader) instead of Ubuntu's startup-disk creator (which uses syslinux) and you will see the difference :)

[Christiansen (happylinux)]

My comment was expected to become #98 as an comment on what Micah Gersten (micahg) #97 wrote. Further more I just wanted to comment the statement oliver (oliver-schinagl) made in #96 about it was either the squash image or the initrd.gz (initrd.lz I reckon). I know the first line mixes #97 and Oliver in an unintended way though. But anyway I was to long to commit my comment and you came first...

That said, your probably right about syslinux, as I've just managed to put Grub2 on my USB stik too, and your absolutely right it now boots and install Kubuntu from exactly the same files that throw the error message with syslinux as bootloader.

[oliver (oliver-schinagl)]

I've also made a new USB stick, using grub2 and loopbacking the ISO.

Here is what I did to an empty formatted 2GB USB stick using the LiveDVD iso (liveCD was same result) and mounted at /mnt.

grub-install --no-floppy --root-directory=/mnt /dev/sdb

grub-mkconfig > /mnt/boot/grub/grub.cfg

Replace all menuentries with the following one.

menuentry "TITLE" {
  iso_path=/ubuntu-12.04-dvd-i386.iso
  export iso_path
  search --set --file $iso_path
  loopback loop $iso_path
  root=(loop)
  configfile /grub/loopback.cfg
  loopback --delete loop
}

Not sure if i copied loopback.cfg into the grub dir, or if it reads it directly from the iso, but i belive it does read it directly from the iso.

Boot.

This boots the iso, as one would expect, except for using syslinux. So the kernel and initrd from the iso are used, the disk is booted and the os installed.

So yes, this looks very much like a bug and should be re-opened/fixed.
P.S. I did not try burning the iso to disk, I used the usbcreator and grub only.

So at the moment, I'm quite happy to be running gnome-session-fallback (hate unity) on stock 12.04 (for now) WITH!! PAE kernel.

[Ian (iktyler007)]

Hi. Total newbie but willing to help. I'm getting the same issue, i.e. using a 12.4 live CD but getting the message:

"This kernel requires the following features not present on the CPU: pae.
Unable to boot - please use a kernel appropriate for you CPU.

Using a Dell Latitude D600. If log files would help, just let me know what you want and how to retrieve the data.

[oliver (oliver-schinagl)]

#106, If you just want to get ubuntu booting from a USB stick, try to follow my really short instructions above :) Feel free to ask anything (on here on in PM) if you have questions about getting it to work.

[sanmiguel9 (againsttcpa84)]

Based on oliver's instructions (thx a lot!! safed my a.. at work today) I successfully installed Ubuntu 12.04 on a Pentium M laptop. The official ubuntu-12.04-desktop-i386.iso file always triggered the PAE-missing-error and failed to install.

As I am a newbie, it took me a while to figure out how exactly it needs to be done, and besides there is a small error in the menuentry of oliver's post. So here it is again for other beginners :-)

1.) Download the official ubuntu-12.04-desktop-i386.iso and use the "Startup Disk Creator" to make a LiveUSB installation on your favorite empty USB key (FYI, I used the option "Discarded on shutdown, unless you safe them elsewhere", just in case...)
2.) Copy the ubuntu-12.04-desktop-i386.iso to the USB key (not in a folder!)
3.) Unmount your USB key. I assume your USB key is the device /dev/sdb. Then remount it to /mnt (e.g. with "sudo mount /dev/sdb1 /mnt" or similar)
4.) sudo grub-install --no-floppy --root-directory=/mnt /dev/sdb
5.) sudo grub-mkconfig |sudo tee /mnt/boot/grub/grub.cfg
6.) sudo gedit /mnt/boot/grub/grub.cfg and look for the "menuentry{ ... }" sections. Replace all of them with one single entry

menuentry "PAE kernel for Pentium M" {
   iso_path=/ubuntu-12.04-desktop-i386.iso
   export iso_path
   search --set --file $iso_path
   loopback loop $iso_path
   root=(loop)
   configfile /boot/grub/loopback.cfg
   loopback --delete loop
 }

7.) reboot and make sure your computer boots from the USB key
8.) send oliver a postcard :-)

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Nick_Hill (nick-nickhill)]

Including a PAE only kernel with the default ISO significantly and unnecessarily limits scope for installation of 12.04, and going forward, Ubuntu generally.

Many good spec machines don't have PAE. A notable example being Centrino laptops.

Within weeks (or days) of a distribution being released, a new kernel is generally available for update. Perhaps the sensible approach would be for the package manager to install the latest PAE kernel on first update IF the CPU supports PAE.

[oliver (oliver-schinagl)]

#108

You are correct, configfile needs 'boot' in it. This is because the file in the ISO is found 'boot/grub/loopback.cfg'

configfile /boot/grub/loopback.cfg

You do not have to use the disk-creator on the USB stick at all!

grub-setup creates /boot and initializes the MBR. grub then chain-loads the iso. That should be it :)

#110
The problem appears to be that syslinux does something that triggers the error. The PAE enabled kernel runs just fine on a pentium-m based system (with less then 4GB of ram is my guess).

[Mantas Kriaučiūnas (mantas)]

Syslinux contains some code (modules), that checks the CPU flags, like PAE support or for 64-bitness and loads the appropriate label in a pxelinux.cfg file. See
http://git.kernel.org/?p=boot/syslinux/syslinux.git;f=com32/modules/ifcpu.c;hb=HEAD
http://syslinux.org/wiki/index.php/Ifcpu64.c32

It seems Ubuntu doesn't use this syslinux feature, could Ubuntu developers tell us where is the problem? Why PAE kernel works fine when GRUB is used, but reports an error about missing PAE when syslinux is used?
Maybe syslinux boot loader forces some CPUs not to work with PAE kernel (maybe some CPUs can work either in PAE mode or not PAE mode?) or maybe syslinux reports to the PAE kernel, that CPU doesn't support PAE?

In any case this syslinux bug should be fixed ASAP - lots of people now can't even start Ubuntu 12.04, while all other versions and most other Ubuntu-based distros (like LinuxMint 13) work fine.

[Duncan J Murray (duncan-mostest)]

Also affects my thinkpad T40 centrino which I use as my main computer (not because I'm stingy, but because it works great with 10.04! - why upgrade?). Really great to see some valid workarounds above. Many thanks, I'll give them a go when 10.04 support runs out next year.

[Jerry Dunmire (jerry-dunmire)]

I'm having a problem with installs from the non-pae ISO (thanks David and Luiz): on a clean install the symlink for /etc/resolv.conf is not created. This results in DNS lookup problems (in my case systems in the local private domain were not resolved) and is probably the issue reported as Bug #1000244. I am reporting it here because the non-pae ISO is not an official ISO.

The work-around/fix is to run dpkg-reconfigure resolvconf and allow it to create the necessary links.

[Luiz Angelo Daros de Luca (luizluca)]

hum... strange this resolv.conf problem. Maybe is is related to some post-release update. I faced no problem like this.

Anyway, it is good to hear that everything is working for you.

[Jerry Dunmire (jerry-dunmire)]

I have installed the ISO three times on two different systems (Sony Viao VGN-S150 and Dell D600) and even before updating packages, the /etc/resolv.conf link is missing. Note that unless you look for the link, or are on a private network with a local DNS server, I don't think the problem will be noticed.

The other thing that might be different about my installation is that I am doing a clean install, but with a manually configured disk layout. Doesn't seem like that is related, but ...

Given Bug #1000244, the problem is probably not non-PAE related, but I don't have a system with a PAE CPU to test.

[Veleno (veleno)]

#112
mantas you're right.
I can cope with obsolescence, but if a system is indeed capable of running PAE kernel flawlessly even if it's non-PAE it should be allowed to.
Both with upgarde process or a fresh install using mini.iso, I've switched afterwards to the PAE kernel on Pentium M 1.6 HGz laptop (ICH4 i855), and everithing works.

[Brad C (8hcf0rc5t)]

@Nick_Hall: +1

I just downloaded a 12.04 LTS iso, downloaded Pendrivelinux's USB creator, and tried to install it on my Dell Inspiron 8600, only to get:

"This kernel requires the following features not present on the CPU: pae"

Arggh.

So a short side note on the Downloads page:

- [If possible to provide for most machines, something like:] Try this to see if you have 'pae' enabled on your processor: 1., 2., 3.
- [Then--or at least, if first '-' isn't possible:] If you don't have 'pae' on your processor, please choose this alternate 'non-pae' .iso image: <image link>

Would be nice.

My $0.02 worth.

~ Brad

[Brad C (8hcf0rc5t)]

Oops, sorry Nick: I meant @Nick_Hill (not Hall)! No way to edit these comments either, it seems...

[Rhys (rimmington)]

I can confirm this issue on a Thinkpad R51 laptop, and that following sanmiguel9/oliver's instructions allows proper booting and installation.

[psamuel (persaudsamuel)]

Descarga la iso de ubuntu nuevamente, quizas sea un problema de la imagen, o verificar si la iso esta buena, tambien puedes hacer un una actualizacion a desarrollo para ver si esto se mejora, ejecuta en la terminal
    sudo apt-get update
    sudo apt-get upgrade

[Dominic Raferd (dominic-timedicer)]

+1 for Luiz's iso image which has worked perfectly for me on HP Pavilion ZE4900. I first tried using the netboot mini.iso but this was complicated (even thought it is stated to be 'non-pae' you have to select advanced mode or it uses pae kernel) and very very slow for downloading. And as others have noted David Henningsson's iso does not fit on a CD.

Luiz's iso just works - thanks!

I guess that there is no point updating the kernel when advised from update manager because all updated kernels will be pae only? And maybe automatic security updates needs to be turned off to prevent this happening automatically.

[Jane Atkinson (irihapeti)]

I don't see any reason why the kernel should update to pae, unless you happen to have a pae metapackage installed (such as linux-generic-image-pae), which is rather unlikely.

I have a non-pae install that I made just after Precise was released, and it's had all the updates without any change of kernel type.

This particular install was done with a Xubuntu alternate CD to get a command-line system, and then "sudo apt-get install ubuntu-desktop" to get the complete system. I doubt though that this would behave any differently post-install from Luiz's image.

[Luiz Angelo Daros de Luca (luizluca)]

@Dominic, You can update your system normally as ubuntu still maintains nonpae kernel on their repo. They only stopped to generated a livecd with this kernel (the gap that my livecd fills). As 12.10 might not contain nonpae kernel, you might need to keep 12.04 forever but let's see the options when the final release happens.

@Jane, the xubuntu installation plus the apt-get install ubuntu-desktop would get very similar results. I just think that this is not an option for systems that have restrictions on storage space (as my eeepc with 4GB of SSD). For anyone with plenty of space and time/bandwidth to download ubuntu-desktop packages over the internet, the xubuntu step is good.

[Jane Atkinson (irihapeti)]

@Luiz
I agree with you on all points, particularly as I have an EeePC (900) myself and have to watch what I install to the 4GB SSD. The Xubuntu-based install was actually to an external HDD, and done more as an experiment to see what options would work. The point I wanted to make above was about upgrading the kernel using the normal updates, and from that point of view, I think my experience is relevant.

Definitely, it is better to have a dedicated nonpae live CD, and thanks for making that available.

[oliver (oliver-schinagl)]

The question still remains, will this bug be actually reopened and fixed for 12.10 for Pentium-M based systems, where PAE kernels just work and syslinux screws things up? Or should a new bugreport be created.

[ImConfused (ed-dori-robinov)]

Linux Mint Maya 13 32bit is based on xubuntu 12.04. It does not require the PAE in the CPU. By accident I downloaded and installed it. I thought I was getting the Linux Mint Debian Edition (not related to Ubuntu). To my surprise I am finding it works rather well on my Pentium M computer. Also it is supposed to be supported till 2017. The Xubuntu 12.04 unlike Ubuntu 12.04 from what I think I read was supported for only 18 months.

I gave up on the hoop jumping with the mini.iso before this out of frustration.

[Erick Brunzell (lbsolost)]

"The Xubuntu 12.04 unlike Ubuntu 12.04 from what I think I read was supported for only 18 months."

Actually Xubuntu 12.04 is supported for 3 years, likewise with Ubuntu Studio 12.04.

Ubuntu, Edubuntu, and Kubuntu 12.04 are all supported for 5 years.

Lubuntu 12.04 is only supported for 18 months.

Mythbuntu 12.04 is less clear, quote from the release notes; "Mythbuntu is now supported on an LTS cycle. This means you should continue to have updates available all the way through the next LTS, currently slated to be 14.04."

The Xubuntu and Lubuntu 12.04 live and alternate images are non-PAE, and an alternate iso can also be used to perform a CLI install quite similar to the mini.iso.

[Erick Brunzell (lbsolost)]

@ Oliver,

In reply to comment #126 you might try subscribing the syslinux maintainer:

https://launchpad.net/syslinux

Since I'm not directly effected by the PAE issue I'm largely bowing out to focus on other issues.

[Dominic Raferd (dominic-timedicer)]

@Luiz and @Jane:

thanks for that. Yes I have found that I can use update manager to update the kernel and it works fine. I am now using 3.2.0-29-generic. I think/presume that the 'generic' kernels are always non-pae?

I suppose they took the non-pae kernel out of the LiveCd to save space? But they could keep it on the LiveDVD surely, and change the message on the LiveCD on finding non-pae system so that it directs user to try with the LiveDVD? This seems to me a much better approach than just dropping non-pae support in 12.10.

[Veleno (veleno)]

I agree with oliver in comment #125
There is hardware capable of running pae kernel flawlessly without having pae CPU flag.
Given that non pae hardware incapable of running pae kernel is the intended/effective/right target of this decision, all of those non pae hardware capable of running pae kernel are just to be considered some sort of "collateral damage"?
If this is it, it's a bit disappointing.
Kudos to Luiz and others which have contributed with their home brewed working ISOs.

[Jeff Powell (jeffp-bangtherockstogether)]

This was a surprise. I have hit several problems with Ubuntu of late - mostly related to the underdeveloped open source NVIDIA drivers - but to see 12.04 not even boot from the CD on my Dell 8600 Laptop is just pathetic. I am downloading Luiz's image now and will try that (rumor has it I can get the proprietary NVIDIA drivers back, which appear to be the only things that work for me).

Should that image not work, though, I am afraid I may have made my last Ubuntu installation. Don't ask me what I will move to, but the issues are piling up.

You cannot discard support for perfectly good machines like this. If you do, I suspect a lot of the community will move on. Don't go there.

[saejin (roger-burney)]

I must agree with Jeff on #132. I have been using Ubuntu since (I think) 6.04 I have current generation hardware without the PAE issue.
I faithfully upgraded with each release, and worked hard to work around the NVidia garbage for old KVM switches that do not allow the driver to see the hardware. 12.04 has been the first release that has consistently frozen, which seems related to the amount of video displayed. OK I thought, This has had to many generations of upgrade. Time to start clean.
I downloaded the 64 bit live disk. IT WOULD NOT BOOT! (Did it multiple times with multiple systems always doing the checksums) So I have a system that runs on 12.04 (and freezes) and I can not do a clean install.

So I tried Fidora 19 (heritic) and it seems just fine..... hummm what to do?
Please fix the live disk problem!

[Hendrik Lönngren (hendrik0)]

I found one thing that doesn't work when running the PAE kernel on my Pentium M based laptop: To be able to use wireless, I need the acerhk module, which relies on BIOS calls. For me, it only works on the non-PAE kernel, with no differences otherwise (see bug #555828).

[Jeff Powell (jeffp-bangtherockstogether)]

This is almost certainly the wrong place to post this, but...

So far it appears that Linux Mint 13 Xfce runs just fine on my ancient, NVIDIA equipped, non-PAE laptop. There are no issues with focus follows mouse - something that appears to be a serious problem with Unity - and it has a long term support plan (5 years).

Sorry, Ubuntu, but I think you've lost me now that it appears I have a real alternative that hasn't lost its way.

[phreich (phreich)]

Same issue for a Dell Inspirion 700M with 1.7ghz Pentium M processor. This is an early "ultrabook" 12" sub-compact notebook computer, built in 2004.

Ubuntu needs to continue to support the thousands of useable older notebook computers that have Pentium M processors in them. Forcing an install of an older version of Ubuntu, and then upgrading is not a user-friendly solution. I suggest checking the amount of memory during the install, and default to non-pae if 3gb or less memory installed, or prompting a question during the install asking if the system can support more than 4gb of memory....

[oliver (oliver-schinagl)]

It's not a PAE kernel bug, it's a syslinux bug, or atleast how syslinux is setup for 12.04. To bad no dev has commented about including a possible fix. I intend to do a clean-install from the iso for 12.10 and hope it actually is resolved. Otherwise, i'msure it's still installable via the grub-trick.

[ImConfused (ed-dori-robinov)]

My computer is PAE deficient. Because of Linux Mint 13 Maya 32bit I am able to use the Ubuntu XFCE 12.04 on my machine. My machine was originally designed for Windows XP. Due to certain websites and a few other reasons every once in a while I am foreced into Windows XP which is a very painful experience. I personally have found Ubuntu as an operating system that keeps my old computer out of the landfill. It has better programs than Windows XP and it run circles around it as far as what it can do under Windows. Windows is necessary for certain functions. Some US Government websites are dependent on Windows. Some websites are totally designed around Internet Explorer. If you want to access those sites you have to have Windows on the computer. Internet Explorer under Linux isn't the greatest experience to deal with. Also being a much smaller piece of the computer pie Ubuntu is less attractive for virus designers than Windows. Canonical LTD is pushing non PAE enabled computers off to the side by ending support for a sizable number of computers that have not died a natural hardware death. Making an operating system that will only run well on the new computers is a Windows / Intel marketing game. It the type of practice that should not be adopted by Linux developers. My 2 cents.

[oliver (oliver-schinagl)]

I don't think that was their soul purpose and intention and don't think this is the best place to vent ;)

Ubuntu 12.04 works just fine, they have non-PAE kernels, it's just the liveCD that lacks this functionality. Also again, it's a 'bug' in syslinux.

That said, I'm sure Ubuntu is aware that their OS should ideally be installed on a lot of old PC's, going to countries not much different then South Africa? If just a Ubuntu-dev would comment here and investigate, it would make things much better.

[Jonathon Conte (thesicktwist)]

I'm running a VIA EPIA board sans PAE that is about 3-4 years old. I am a longtime Ubuntu user who has been running Ubuntu on various machines since Warty. I am extremely disappointed that Tim Gardner describes Ubuntu users such as myself as "simply out of luck (and no longer supported)" when it comes to removing support for relatively modern, non-PAE hardware. Way to alienate your users.

[Douglas Brown (drphysic)]

As a 64-bit ubuntu user, I was unaware of this issue until I tired to test lubuntu on my 2 older laptops: Dell 600m, Fujitsu P-Lifebook. I downloaded the Alternate 386 iso file last night and discovered to my great surprise that this version also requires the pae capability to either perform an install or a disk check. For a version "designed for low RAM systems" with less than 700MB, it is incredible that it also fell into the pae trap. At this point, claims that lubuntu is "non-pae" seem somewhat hollow at best.

For those of us on the outside, the decision to drop support for non-pae systems for 12.10 and beyond seems cavalier in a way that I never detected before in ubuntu development, over a period of 7 years.

[oliver (oliver-schinagl)]

What stings is, that the entire distro just WORKS WITH PAE enabled on pentium-M processors and it simple is a bug in syslinux that has been ignored ever since this thread was started. And since this bug is 'closed' as 'won't fix' it won't get noticed either. It also has been marked as a bug on syslinux where it is 'undecided'.

[Christopher Grayce (cgrayce)]

This is really disappointing. This problem has been around for about a year, IIRC, and seems to affect quite a lot of people with older hardware. If I wanted an OS that only worked on the latest and best hardware, I'd buy a new machine and install Windows 7, which is actually a pretty good OS. I've been using Linux since 1994 and it's unfortunate how far from the spirit of the original Canonical seems to have strayed in recent years. I'm going to give Mint a shot for my Inspiron 8600.

[oliver (oliver-schinagl)]

Christpher, I think that Girish Sanenhalli is going to fix this in syslinux soon. So it seems like that atleast in 13.04, pentium-m's may work as expected again.

[bryncoles (brunomatti)]

Just found this bug report and wanted to chime in. I have a Packard Bell r-series easy note with an Intel Celeron M processor. It is in fine working order, except for the lack of PAE support. It cannot boot Ubuntu 12:04 or any *buntu family 12:10.

Shame to have the machine obsoleted for me...

[Dominic Raferd (dominic-timedicer)]

@bryncoles: you can install Ubuntu 12.04 using Luiz's image which is still available here - dl.dropbox.com/u/1286502/ubuntu-12.04-desktop-i386-nonpae.iso

Its unofficial so all the usual caveats apply, but it worked great for me and for others. Afterwards you can do any normal updates (including kernel etc) - they all work fine.

[JackT (anonymous352)]

@brycoles: I'm pretty sure 12.04 Lubuntu installs on a non-PAE machine (I have it running on an ancient Sony Vaio laptop which is non-PAE)...

[Bernd Kreuss (prof7bit)]

The affected Pentium-M CPUs of these not so old Laptops all have phyical address space of 36 bits but for some reason Intel decided to not set the PAE flag (maybe its a bug in these series of CPU itself?) These were very popular CPUs at that time and they were built into a lot of Laptops, including many of the legendary ThinkPad models of that time, many of which are still in use today.

I can confirm that the PAE-Kernel will boot and run on these machines without any problems.

There are currently 2 Problems that make it extremely hard for people to use these kernels:
(1) the syslinux bootloader will refuse to boot from the install CD (although it could if it would at least try)
(2) the install scripts in the kernel .deb packages will simply grep -q ' pae ' /proc/cpuinfo and fail with error

The problem in (2) can be hacked around by the desperate user by simply bind-mounting a faked /proc/cpuinfo that contains this flag and then one can happily upgrade to 12.10 and install and use these new pae kernels without the slightest problem. It should be fixed by making the check in the install script in these kernel .deb packages more elaborate, maybe instead of grepping for pae it should grep for "36 bits physical" or these few very popular CPUs should simply be whitelisted.

The problem in (1) should be fixed in the way syslinux detects this (maybe counting the physical address bits instead of looking for the flag or simply whitelisting these few CPUs that actually have PAE and just not announce it in their flags would be a good idea?)

It wouldn't be such a problem if it were only old and exotic hardware but this bug also affects all the popular ThinkPads that are not so old and still in use today! These devices are for example still used today in the international space station (ISS) (they actually are the *only* Laptops officially certified for space travel and Ubuntu does not want to support them anymore) Richard Stallman (GNU) has been seen with ThinkPads on multiple occasions and even Shuttleworth (Canonical) himself has been photographed on board of the ISS using a ThinkPad. You cannot simply ignore this entire class of wonderful and useful Devices.

https://www.google.de/search?q=iss%20thinkpad

[oliver (oliver-schinagl)]

I haven't tried any of the alpha's of 13.04 but I would not be supprised if this bug report goes completly ignored. So far, no dev. has even confirmed or denied any of this.

[mtu (mtu)]

I can confirm this problem for a Thinkpad X31. This is a crazy display of arrogance and ignorance on Canonical's part, but alas, it is not the first.

[JohnWashington (ubuntu-johnwash)]

@mtu: "This is a crazy display of arrogance and ignorance on Canonical's part..."

You might think that. I might think that. And certainly I think this has the potential to affect many disadvantaged people around the world, trying to reuse old but serviceable hardware.

But 'mtu', look at the numbers. There are bugs where 1,000+ people have marked them as being affected. This bug has been marked by 72.

Find 1,000 friends with Thinkpads and you might get noticed! Better still, find someone with the skills to fix this.

As you'll see from my earlier comments, I too am most unhappy with the policy, but I'm trying to be realistic and pragmatic and look at alternatives such as Debian.

[oliver (oliver-schinagl)]

Debian may be very well equally hit. Anyway, maybe someone needs to make a new bug report, link it to this one and maybe that gets the attention to the official dev's... As it stands now, this is simply being ignored or just not noticed.

[Felix Moreno (felix-justdust)]

It also happen in Ubuntu 12.10, same problem, not interested in fix it? 75 people same problem, other bugs with just 5 people and is fixed...

[Luiz Angelo Daros de Luca (luizluca)]

@felix-justdust,

12.04 does not have a livecd with nonpae kernel but it has the kernel package in repo. So, who uses it gets the updates.
12.10 does not include the nonpae kernel anymore. If anyone wants to fix 12.04 the same way, he must also recompile the kernel
for nonpae machines and maintain it whenever canonical release a new kernel.

Sorry,

[ImConfused (ed-dori-robinov)]

72 people reported this problem on 12.10? The fact that Ubuntu has dropped support for non PAE systems after 12.04 indicates that these 72 people didn't know that fact before they tried to install it with no success. The number who didn't even try to use 12.10 (like myself) has to be much larger since they already knew that there was no support. It is a waste of time to even attempt to use 12.10 because of that. Personally I am using Linux Mint Maya 13 which is compatible with and based on Ubuntu 12.04. I didn't have to jump through hoops to install it either like I experienced with Ubuntu 12.04 unsuccessfully.

[oliver (oliver-schinagl)]

Remember, this is not a problem with ubuntu, but with syslinux. The penitum-M works just fine with a PAE kernel. Syslinux checks for the pae flag in the CPU (it does lack that) and thus fails to boot.

Hence why you can use grub instead of syslinux to boot and install it just fine (i wrote about it above somewhere). You could of course remove the check in syslinux (there's a bug report attached to this for that) and use that instead, but using grub (on a USB stick anyway) is just really easy.

[Chris Bainbridge (chris-bainbridge)]

Two possible workarounds:

* Use a non-PAE kernel - see "How to install Ubuntu 12.10 on non-PAE CPU" at http://www.webupd8.org/2012/11/how-to-install-ubuntu-1210-on-non-pae.html which explains how to modify the iso and links to .deb packages at http://bazaar.launchpad.net/~webtom/+junk/linux-image-i386-non-pae/files

* Use the PAE kernel with "fake-pae" - it adds pae to /proc/cpuflags so the check in the Ubuntu package will pass - see https://launchpad.net/~prof7bit/+archive/fake-pae and the associated discussion thread at http://ubuntuforums.org/showthread.php?t=2113826 (this option works surprisingly well - it appears that the PAE kernel runs fine on Pentium M CPUs, which is surprising because the kernel config help states "If you say 64GB here, then the kernel will not boot on CPUs that don't support PAE!")

[oliver (oliver-schinagl)]

fake-pae. That the single most awesome comment so far ;) It's a workaround, to work around an issue with the CPU. While without special whitelists.

I personally think, the CPU supports PAE just fine, but it was 'experimental' maybe at the time or a 'luxury' function so the CPU simple didn't advertise it.

Anyhow, while this fake-pae bit sounds like it's 'the' solution once the system is running; We still need sys-linux to play well. So this bug report is still relevant to that regard.

[oliver (oliver-schinagl)]

New release, same old bugs.

Comment #105 and #108 are still 100% accurate. You can have the iso in any dir you want btw, just adjust the menu entry accordingly. You can even have multiple ISO's this way.

Right now I have a 2GB USB stick with 32 bit and 64 bit Ubuntu Gnome ISO's. If my stick where bigger, I could put several other flavors on their.

Adjusting grub.cfg of course is requried for this, as explained in #105. One menuentry per iso. You do loose the graphics boot menu this way however, syslinux is responsible for that. Grub gives you only the textual menu's. No issue really though. After 2 seconds the GUI does kick in as normal and you have the regular experience.

[sudodus (nio-wiklund)]

Now, that we have launched Raring, it's time to start thinking about the
future of the Ubuntu family.

One thing that I hope we can achieve, is to make the next version able
to cooperate with Pentium M, that has no pae flag, but still the ability
to run i686-pae kernels of *ubuntu.

I have verified that it works for me in an IBM Thinkpad T42 to use the
fake-pae method for Pentium M CPUs by 7bit according to this link

http://ubuntuforums.org/showthread.php?t=2113826

This is good, but it would be great if this feature could be included in
*ubuntu, so that people can install a new version directly, and avoid
installing 12.04, add the fake-pae and upgrade to newer versions.

I'm prepared to use my Thinkpad for testing, but I don't know enough to
find and change the code.

Maybe I can even try to compile the crucial package(s), if someone can
hold my hands until I get the tools installed to do it (compilers,
makefiles, source code packages etc.)

The fake-pae fix in the ubuntuforums thread is not for everyone,
but I think it would be fairly easy to enter some logic to make a
difference between the Pentium M CPUs, that support pae, and the other
non-pae CPUs. I can make such logic statements, but I need help to find
where to put them.

By the way, I think most if not all of the computers without pae, that are useful for
anything but 'retro playing', are laptops with Pentium M. And, I think
the objective of this task is not to stuff those old computers with more than 2
GB of RAM to exploit the pae feature, but to be able to run a new
*ubuntu version at all with the memory available, often around 1 GB.

[Chris Bainbridge (chris-bainbridge)]

The kernel team have made it clear that pentium-m is no longer supported on Ubuntu because it is old (5+ years). I don't mean to sound harsh or troll-like, but at this point the debate is over, so if you really want to install the latest then I suggest you try Fedora, Debian, or any other distro that supports new releases on older hardware. It really is easier than messing about with fake pae etc.

[oliver (oliver-schinagl)]

Then the Ubuntu kernel team is just retardedly stupid. And I will make it harsh.

There is no SINGLE technical reason to not support it. The kernel runs PAE kernels just fine. There is a bug maybe you can call it, that does not tell you that it IS pae (36 bit addressing) capable. It just lags the PAE flag. THe sheer amount of users coming here and to the forum thread alone should give you an idea that it is still relevant.

That said, they deliberatly broke the Pentium-M by checking for the PAE flag. Granted this should be a valid check, and the Pentium-M messes that check up. There's your bug.

Also, this bug is about SYSLINUX. NOT the kernel (it is now too though). SYSLINUX here checks on boot if you have the PAE flag and refuses to boot. It REFUSES, not because its technically not possible, no, because it was decided so, because of a silicon bug if you will.

Recently the kernel team decided to have a PAE check int he package. 'Only install if you have the PAE flag'. Now in the kernel package, this may make a little more sense, there's PAE and non-PAE kernels available, and adding a non-PAE ppa is easy. Syslinux can't be easily replaced, well you can grub chainload your ISO and all is well.

So if this truely is about support, then stop supporting the 32 bit architecture. Period. But with that even being the recommended architecture for must, I don't see this happening soon. This again, has nothing to do with the Pentium-M not being supported, but about the CPU not properly announcing PAE.

[oliver (oliver-schinagl)]

P.S. I didn't mean to sound to offensive, but no developer has responded with a very valid technical response. The only answers we have gotten 'its no longer supported, cus I said so'.

[JohnWashington (ubuntu-johnwash)]

Oliver, I too am frustrated by this situation. Following your logic though, perhaps the people you should be labelling as 'retardedly stupid' are the guys in Intel who failed to make the chip describe its capabilities accurately. Or IBM and the various other manufacturers who chose to use this chip (though I don't know whether the oversight was known at the time).

The solution is clear. Find someone who can fix the code. The code is GPL v2, it's not locked away in some proprietary vault (isn't that one of the reasons you and I like using open source?). I imagine you can fetch the source with apt-get, or if that doesn't work, go to syslinux.org where there's a web interface to the git tree.

Insulting people does communicate how strongly you feel about something. But does it encourage them to act as your willing servants?

[Chris Bainbridge (chris-bainbridge)]

> That said, they deliberatly broke the Pentium-M by checking for the PAE flag.

And this breaks upgrades, hopefully this will be fixed so upgrades get blocked before systems get broken, see bug #1160346 ("do-release-upgrade from Ubuntu 12.04 on Pentium-M fails, breaks system without any warning")

> P.S. I didn't mean to sound to offensive, but no developer has responded with a
> very valid technical response. The only answers we have gotten 'its no longer
> supported, cus I said so'.

The technical discussion was at http://ubuntu.5.x6.nabble.com/Dropping-i386-non-PAE-as-a-supported-kernel-flavour-in-Precise-Pangolin-td731288i20.html - the conclusion / ultimate reason for dropping was:

"Dropping this flavour saves 5 minutes per build on a 4-way 80 thread
server, which for some of the team can add up to quite a bit of time
over the course of a day. Its one less variant that needs to be tested
in Q/A, and its one less flavour we have to mess with in our meta and
LBM packages. "

Also noted in same discussions on lubuntu list was that Ubuntu only targets a 3 year hardware support window, so any hardware older than 3 years can be dropped. Ubuntu is not a distribution for old hardware.

Whether it was a good decision or not is a moot point now, it is done, and for a Pentium-M user it is easier to choose a distribution that still supports that CPU rather than try to install a distribution that does not.

> So if this truely is about support, then stop supporting the 32 bit architecture.

I am sure it is being considered for the same reasons, Apple already dropped 32-bit, and Windows 8 will be the last 32-bit release.

[sudodus (nio-wiklund)]

My suggestion is to include Pentium M in the CPUs available for the 32-bit pae version of *ubuntu. This can be done with a few logical statements at the right place in the code, and it will hardly change the time to create the *ubuntu daily build iso files.

I am not pleading for reviving a non-pae version.

[oliver (oliver-schinagl)]

I'm supprised that some people still don't understand what this is about (nofi).

This is about syslinux adapting:

if (!((flags & PAE) && (cpuid == Pentium-M)) {

[oliver (oliver-schinagl)]

printf("CPU not supported\n");
} else {
boot()
}

It should be that simple. The Pentium-M is a supported CPU, but lacks the PAE flag. That's it.

We don't pleed for support for non-PAE kernels, non-PAE builds, not wasting 5 minutes of build time. Just extend the if (flags & PAE) check to also check if it is a Pentium-M.

Yes the pentium-M is a rare beast and old. Yes intel should be slapped for having this bug (PAE wasn't that common then?). I don't think IBM can be hold accountable, they choose the best CPU for that time. Desktop CPU's where to power hungry, and besides not having the FLAG, the CPU is still quite good.

p.s. sorry for the wrong enter there :)

[Stan Osborne (stan-ana)]

Thanks for taking a year or so to let me know. I understand
you are overworked closing out trival bug reports. That you
are able to finaly respond, is impressive.

Soon after I discovered 12.04 stopped supporting PAE, I found
a build with non-PAE support. The 13 year old IBM laptop now
uses 12.04 just fine. The CPU may be unsupported, but automatic
update works fine. Thanks for continuing to support the un-
supported CPU.

Prior to 12.04, non-PAE CPU's were supported, so to me it makes
little sense to suddenly stop this support. It probably is taking
more effort to turn it off and deal with all the complaints, than
it would to leave in the non-PAE support. Or does Ubuntu help
hardware vendors force customers to upgrade CPUs?

FWIW, from talking with other Ubuntu users, they too would like
to see an error message when attempting to install a PAE only
distribution on a non-PAE machine.

Stan

[oliver (oliver-schinagl)]

Stan, just an FYI, in case you haven't read any of the previous posts. The Pentium-M supports PAE just fine, it just lies about not having it. Not sure what your 13 year old IBM uses however. My IBM is 9 years old (doing the math on that surprises me immensely, not realizing it was that old already, but its still going strong!).

If you DO have a Pentium-M, there are instructions here how to use the ISO via grub on a USB stick. Its quite painless.

If however you have an older Pentium, that predates the Pentium M, it is still possible it features PAE! It was first introduced on the Pentium Pro (e.g. the first i686 if memory serves right, which PRE-dates the Pentium III of which the Pentium-M is based on). Though the Pentium-M seems to start in 2003, mean, leaving me to wonder what ancient CPU you have ;) Though even the Pentium III should support PAE, which does date 13 years back.

RE-reading the wikipedia article (which I admit I haven't done in a while) it claims the following.

"PAE is supported by Intel Pentium Pro and later Pentium-series processors except most 400 MHz-bus versions of the Pentium M."

I hope that line makes it clear why this is causing an issue for us Pentium-M users. That the Ubuntu team decided to no longer support non-PAE hardware makes perfect sense, it predates the Pentium-Pro. So we're talking CPU's with less then 150 MHz to put it in context (yes MHz means nothing, but 150 MHz should tell you something). Very valid reasoning. But whoever made the decision overlooked the Pentium-M, and no none of the Ubuntu folk even look at this thread to see why this is a problem.

[Christiansen (happylinux)]

oliver, I agree with you entirely on your previous comments. The ThinkPad series were rather expensive hardware, and maybe on that account rather long lasting hardware. To this date I'm servicing 5 of these older ThinkPads running 12.04 *buntu, installed using the USB GRUB method described above in this bug.
In later *buntus the kernel packages have been deliberately crippled by developers, so it isn't quiet as easy to work around, while you now have to fiddle with the kernel package for installation, and every time a subsequent kernel upgrade happens.

I completely agree with the Ubuntu decision to let NON-PAE hardware go, but can't understand why hardware that works absolutely perfect with even the latest (kernel-fixed) versions of *buntu is left out this way. The change in the pre 12.04 kernel packages placing an OBSTRUCTION, is absolutely misplaced. The syslinux bug on the other hand demands bug-fixing work to be done, and we must accept if neither Canonical or the syslinux developer will spend time on that.

But what I absolutely don't understand is, that the time used to make the *buntu kernel packages OBSTRUCTION, wasn't spent on fixing syslinux bug instead.
A lot of people (ThinkPad among others) would probably have benefitted from this, and even those who don't find this bug rapport, and probably don't know what hit them when trying to install *buntu on their great hardware. Instead they may try another newer bad operating system from around 2007, and find that it just installs and works on their hardware - thus linux or *buntu is just a pain to use, which is hard for us who value open source and linux to argue against. So everybody, except you know who, looses out.

[ldlandis (ldlandis)]

Hi,

My machines (IBM ThinkPad T41, Type 2373-9EU) all claim to
be Pentium-M and do NOT support PAE according to the Ubuntu
boot message. I do have 2GB (the max without PAE) of DRAM
installed.

How much simpler is life with PAE-only support?

(Since that is the path that was taken, I am hoping that the
answer is very trivial to quantify since the work was done
and the code base reflects that change).

I understand moving ahead, but, to me, this fairly capable 32-bit
machine should still be supportable by Ubuntu. I have not tried
using CentOS as my desktop yet, but maybe that is an option if
I can't get the next LTS (14.04) Ubuntu to boot straight up.

Cheers,
  --ldl

ldl@boGus:~$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 12.04.2 LTS
Release: 12.04
Codename: precise

On Fri, May 3, 2013 at 3:35 PM, Christiansen <email address hidden>wrote:

> oliver, I agree with you entirely on your previous comments. The ThinkPad
> series were rather expensive hardware, and maybe on that account rather
> long lasting hardware. To this date I'm servicing 5 of these older
> ThinkPads running 12.04 *buntu, installed using the USB GRUB method
> described above in this bug.
> In later *buntus the kernel packages have been deliberately crippled by
> developers, so it isn't quiet as easy to work around, while you now have to
> fiddle with the kernel package for installation, and every time a
> subsequent kernel upgrade happens.
>
> I completely agree with the Ubuntu decision to let NON-PAE hardware go,
> but can't understand why hardware that works absolutely perfect with
> even the latest (kernel-fixed) versions of *buntu is left out this way.
> The change in the pre 12.04 kernel packages placing an OBSTRUCTION, is
> absolutely misplaced. The syslinux bug on the other hand demands bug-
> fixing work to be done, and we must accept if neither Canonical or the
> syslinux developer will spend time on that.
>
> But what I absolutely don't understand is, that the time used to make the
> *buntu kernel packages OBSTRUCTION, wasn't spent on fixing syslinux bug
> instead.
> A lot of people (ThinkPad among others) would probably have benefitted
> from this, and even those who don't find this bug rapport, and probably
> don't know what hit them when trying to install *buntu on their great
> hardware. Instead they may try another newer bad operating system from
> around 2007, and find that it just installs and works on their hardware -
> thus linux or *buntu is just a pain to use, which is hard for us who value
> open source and linux to argue against. So everybody, except you know who,
> looses out.
>
> --
> You received this bug notification because you are subscribed to a
> duplicate bug report (1029041).
> https://bugs.launchpad.net/bugs/930447
>
> Title:
> Unable to start Ubuntu 12.04 live CD with syslinux loader on Pentium M
> x86 Laptop due to bug in PAE kernel, initramfs or syslinux
>
> Status in “linux” package in Ubuntu:
> Won't Fix
> Status in “syslinux” package in Ubuntu:
> Confirmed
> Status in Baltix GNU/Linux:
> Triaged
>
> Bug description:
> Ubuntu 12.04 doesn't start from Desktop CD or USB wit...

[oliver (oliver-schinagl)]

#171 actually if you search for the 'fake-pae' package (and disable it for startup imo) it re-mounts /proc/cpuinfo and modified to list the PAE flag, so kernel upgrades can be performend normally.

I don't want it at start up, as I try to have an unmodified system for as long as possible. So that I can spot other PAE issues quickly.

[oliver (oliver-schinagl)]

#172 I think the intention was to drop 32-non-pae builds, E.g. everything prior to the Pentium-Pro.

Someone overlooked the Pentium-M, and have not yet noticed this little pentium-m Fallout.

[sudodus (nio-wiklund)]

 Hello Pentium M users,

It would be nice if you post a reply in this new thread at the Ubuntu Forums

"What OS are you running with Pentium M and what do you want next?"
http://ubuntuforums.org/showthread.php?t=2143297

and describe your experiences with Pentium M.

Best regards
sudodus

[Harry French (htfrench)]

On 5/9/2013 2:47 AM, sudodus wrote:
> Hello Pentium M users,
>
> It would be nice if you post a reply in this new thread at the Ubuntu
> Forums
>
> "What OS are you running with Pentium M and what do you want next?"
> http://ubuntuforums.org/showthread.php?t=2143297
>
> and describe your experiences with Pentium M.
>
> Best regards
> sudodus
>
I have 6 Dell Latitude D800 notebooks.
They work fine with several of the older distros.
I like them because 3 have 1920x1200 resolution which is no longer
available. Also, they are tough.
The old distros are also are needed because the "nouveau" open NVIDIA
drivers do not correctly identify the video cards.
I have GoForce FX4200 32MB and 64MB cards, also GoForce FX5200 32MB and
64MB, and FX5265 128MB.
The 32MB cards, which are more common, are recognized as 64MB.
I have WXGA, WSXGA+, and WUXGA screens, if some one wants to help with
the "nouveau" problems.
I have put the newer model Pentium M in one machine.
The machine claims it is not supported, but it seems to work.

[sudodus (nio-wiklund)]

On 2013-05-09 23:08, Harry French wrote:
> On 5/9/2013 2:47 AM, sudodus wrote:
>> Hello Pentium M users,
>>
>> It would be nice if you post a reply in this new thread at the Ubuntu
>> Forums
>>
>> "What OS are you running with Pentium M and what do you want next?"
>> http://ubuntuforums.org/showthread.php?t=2143297
>>
>> and describe your experiences with Pentium M.
>>
>> Best regards
>> sudodus
>>
> I have 6 Dell Latitude D800 notebooks.
> They work fine with several of the older distros.
> I like them because 3 have 1920x1200 resolution which is no longer
> available. Also, they are tough.
> The old distros are also are needed because the "nouveau" open NVIDIA
> drivers do not correctly identify the video cards.
> I have GoForce FX4200 32MB and 64MB cards, also GoForce FX5200 32MB and
> 64MB, and FX5265 128MB.
> The 32MB cards, which are more common, are recognized as 64MB.
> I have WXGA, WSXGA+, and WUXGA screens, if some one wants to help with
> the "nouveau" problems.
> I have put the newer model Pentium M in one machine.
> The machine claims it is not supported, but it seems to work.
>
Hi Harry,

I see that you have more than one issue with your Dell Latitude D800
notebooks, and I can understand that you might hesitate to try my
Lubuntu-fake-pae because of expected problems with the nvidia graphics.
The only thing I can say is that I have a few older nvidia chips that
work with the present version of nouveau (but these are is desktop
computers, not in any computer with Pentium M CPU).

- Which distro and version are you running right now, and what about the
end of life of that version?

- Is there any bug report about your issue with nouveau?

- Have you tried any proprietary nvidia driver?

Best regards
sudodus

[sudodus (nio-wiklund)]

Hi again, Pentium M and Celeron M users.

Lubuntu-fake-PAE offers methods to make Lubuntu 13.04 work with Pentium M and Celeron M. It it more polished now: the instructions are easier to follow, now that we have received feedback from several people. Have a look at this wiki page

https://help.ubuntu.com/community/Lubuntu-fake-PAE

I hope it can help you

Best regards
sudodus

[Chris Bainbridge (chris-bainbridge)]

sudodus, your description around the 36-bit physical addressing seems a little bit confused: "If you don't get 36 bits with a PAE kernel, your CPU has no PAE capability." - the Linux documentation states that if you can boot a PAE kernel, then your CPU has PAE. If you don't have PAE, then the kernel won't even boot.

If you think that there are some non-PAE Pentium M models (which I have not seen reports of anywhere) and want to identify them then report the family, model and stepping, you can see these in dmesg ("CPU:" line) But at the moment I suspect every Pentium M can do PAE (have you seen any reports otherwise?)

[sudodus (nio-wiklund)]

On 2013-05-19 14:15, Chris Bainbridge wrote:
> sudodus, your description around the 36-bit physical addressing seems a
> little bit confused: "If you don't get 36 bits with a PAE kernel, your
> CPU has no PAE capability." - the Linux documentation states that if you
> can boot a PAE kernel, then your CPU has PAE. If you don't have PAE,
> then the kernel won't even boot.
>
> If you think that there are some non-PAE Pentium M models (which I have
> not seen reports of anywhere) and want to identify them then report the
> family, model and stepping, you can see these in dmesg ("CPU:" line)
> But at the moment I suspect every Pentium M can do PAE (have you seen
> any reports otherwise?)
>
You are probably right about that, but I'm not sure about the oldest
versions of Pentium M and Celeron M. The reason why I want people to
check the address size is the we are working around the check, that
should make it impossible to boot a non-pae CPU with a pae kernel.

I also want to warn people with old non-pae CPUs (that are not Pentium M
or Celeron M), that fake-PAE will not help their pae kernels to work.

     -o-

No, I have not seen any such reports, 'just trying to play safe'.

[Chris Bainbridge (chris-bainbridge)]

> The reason why I want people to
> check the address size is the we are working around the check, that
> should make it impossible to boot a non-pae CPU with a pae kernel.

Fake -PAE bypasses the check to stop dpkg installing a PAE kernel. But, even after that, the PAE kernel won't boot on a non-PAE CPU. PAE is not something that gets enabled (or disabled) at runtime, it is actually a replacement of functions in the kernel source (#ifdef .. #endif etc.), so if you have a PAE kernel it will not boot on non-PAE. So it makes no sense to ask people to check the address size on a CPU that has already booted a PAE kernel, since the fact that it successfully booted a PAE kernel means it must support PAE. (afaik - this is from the documentation, I do not have a non-Pentium-M non-PAE system to test with)

[sudodus (nio-wiklund)]

On 2013-05-19 19:22, Chris Bainbridge wrote:
>> The reason why I want people to
>> check the address size is the we are working around the check, that
>> should make it impossible to boot a non-pae CPU with a pae kernel.
>
> Fake -PAE bypasses the check to stop dpkg installing a PAE kernel. But,
> even after that, the PAE kernel won't boot on a non-PAE CPU. PAE is not
> something that gets enabled (or disabled) at runtime, it is actually a
> replacement of functions in the kernel source (#ifdef .. #endif etc.),
> so if you have a PAE kernel it will not boot on non-PAE. So it makes no
> sense to ask people to check the address size on a CPU that has already
> booted a PAE kernel, since the fact that it successfully booted a PAE
> kernel means it must support PAE. (afaik - this is from the
> documentation, I do not have a non-Pentium-M non-PAE system to test
> with)
>
I see your point. I will look for some links, that support what you say,
and change the text accordingly. If you have a link available, please
send it to me :-)

Can you explain why the non-pae ubuntu code reports 32 bits physical
memory, while the pae code reports 36 bits? Some people say it should
report the same. But there is a difference on the computers I have tested.

-o-

If it is the general truth that non-pae CPUs won't boot at all, it is
even harder to understand why the Pentium M CPUs are excluded by the
Ubuntu developers. Why are those tests implemented (in syslinux), or why
are they shutting down, and not only printing the message?

[Greg Weber (gregweber)]

I have 2 or three pentium m systems to test on. may i help out with testing
sodudus? (I have ibm thinkpad x41, t41, and maybe a t40 that only works on
power cord/battery one or other)

On Sun, May 19, 2013 at 12:51 PM, sudodus <email address hidden> wrote:

> On 2013-05-19 19:22, Chris Bainbridge wrote:
> >> The reason why I want people to
> >> check the address size is the we are working around the check, that
> >> should make it impossible to boot a non-pae CPU with a pae kernel.
> >
> > Fake -PAE bypasses the check to stop dpkg installing a PAE kernel. But,
> > even after that, the PAE kernel won't boot on a non-PAE CPU. PAE is not
> > something that gets enabled (or disabled) at runtime, it is actually a
> > replacement of functions in the kernel source (#ifdef .. #endif etc.),
> > so if you have a PAE kernel it will not boot on non-PAE. So it makes no
> > sense to ask people to check the address size on a CPU that has already
> > booted a PAE kernel, since the fact that it successfully booted a PAE
> > kernel means it must support PAE. (afaik - this is from the
> > documentation, I do not have a non-Pentium-M non-PAE system to test
> > with)
> >
> I see your point. I will look for some links, that support what you say,
> and change the text accordingly. If you have a link available, please
> send it to me :-)
>
> Can you explain why the non-pae ubuntu code reports 32 bits physical
> memory, while the pae code reports 36 bits? Some people say it should
> report the same. But there is a difference on the computers I have tested.
>
> -o-
>
> If it is the general truth that non-pae CPUs won't boot at all, it is
> even harder to understand why the Pentium M CPUs are excluded by the
> Ubuntu developers. Why are those tests implemented (in syslinux), or why
> are they shutting down, and not only printing the message?
>
> --
> You received this bug notification because you are subscribed to a
> duplicate bug report (1058748).
> https://bugs.launchpad.net/bugs/930447
>
> Title:
> Unable to start Ubuntu 12.04 live CD with syslinux loader on Pentium M
> x86 Laptop due to bug in PAE kernel, initramfs or syslinux
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/930447/+subscriptions
>

[sudodus (nio-wiklund)]

Hi Greg,

You are very welcome to test fake-PAE :-)

You find the information at this link

https://help.ubuntu.com/community/Lubuntu-fake-PAE

Look at it all, then decide which way you want to go

https://help.ubuntu.com/community/grub-n-iso
or
https://help.ubuntu.com/community/InstalledSystemFakePAE
or mörgæs's way
https://help.ubuntu.com/community/PAE

If you need more details to make it work, I suggest that we exchange
information directly (for example via private email), because those
details are probably not relevant for the bug report.

I have a Thinkpad T42 with a Pentium M 1.7 GHz myself, so it will be
valuable to get test results from your x41, t41, and last but not least t40.

And in the end, we can publish your results in the list of tested CPUs
(in the wiki) and more verbose at the bug report.

Best regards
sudodus

On 2013-05-20 04:23, Greg Weber wrote:
> I have 2 or three pentium m systems to test on. may i help out with testing
> sodudus? (I have ibm thinkpad x41, t41, and maybe a t40 that only works on
> power cord/battery one or other)
>
>
> On Sun, May 19, 2013 at 12:51 PM, sudodus <email address hidden> wrote:
>
>> On 2013-05-19 19:22, Chris Bainbridge wrote:
>>>> The reason why I want people to
>>>> check the address size is the we are working around the check, that
>>>> should make it impossible to boot a non-pae CPU with a pae kernel.
>>>
>>> Fake -PAE bypasses the check to stop dpkg installing a PAE kernel. But,
>>> even after that, the PAE kernel won't boot on a non-PAE CPU. PAE is not
>>> something that gets enabled (or disabled) at runtime, it is actually a
>>> replacement of functions in the kernel source (#ifdef .. #endif etc.),
>>> so if you have a PAE kernel it will not boot on non-PAE. So it makes no
>>> sense to ask people to check the address size on a CPU that has already
>>> booted a PAE kernel, since the fact that it successfully booted a PAE
>>> kernel means it must support PAE. (afaik - this is from the
>>> documentation, I do not have a non-Pentium-M non-PAE system to test
>>> with)
>>>
>> I see your point. I will look for some links, that support what you say,
>> and change the text accordingly. If you have a link available, please
>> send it to me :-)
>>
>> Can you explain why the non-pae ubuntu code reports 32 bits physical
>> memory, while the pae code reports 36 bits? Some people say it should
>> report the same. But there is a difference on the computers I have tested.
>>
>> -o-
>>
>> If it is the general truth that non-pae CPUs won't boot at all, it is
>> even harder to understand why the Pentium M CPUs are excluded by the
>> Ubuntu developers. Why are those tests implemented (in syslinux), or why
>> are they shutting down, and not only printing the message?
>>
>> --
>> You received this bug notification because you are subscribed to a
>> duplicate bug report (1058748).
>> https://bugs.launchpad.net/bugs/930447
>>
>> Title:
>> Unable to start Ubuntu 12.04 live CD with syslinux loader on Pentium M
>> x86 Laptop due to bug in PAE kernel, initramfs or syslinux
>>
>> To manage notifications about this bug go to:
>> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/930447/+subscriptions
>>...

[sudodus (nio-wiklund)]

Sounds great :-)

Where can the fix be found? I mean, are there iso files to be downloaded yet? In that case from where?

Best regards
sudodus

[sudodus (nio-wiklund)]

Has this bug really been fixed and released? If so, where is the code?

[Shahar Or (mightyiam)]

I sent a message to piotr asking him about the changes. My guess is that this is a mistake.

[Luiz Angelo Daros de Luca (luizluca)]

As my dropbox is continuously breaking the download limit, I'm also providing an google drive link for the 12.04 iso with non-pae kernel

https://docs.google.com/file/d/0B54xzz44RpW6RkVLc3d4WEdpMEU/edit?usp=sharing

[sudodus (nio-wiklund)]

There is updated information at this link

https://help.ubuntu.com/community/Lubuntu-fake-PAE

Our test results so far for Pentium M and Celeron M CPUs suitable for fakePAE, 'No PAE flags but 36 bit physical memory address size':

---- CPU name --------- ---- CPUID Output of 'cpuid|grep ^00000001' --
Celeron M 1200 Mhz 00000001 00000695 00000812 00000000 a7e9f9bf
Celeron M 1.40 GHz
Pentium M 1.50 GHz 00000001 000006d6 00000816 00000180 afe9f9bf
Pentium M 1600 MHz 00000001 00000695 00000816 00000180 a7e9f9bf
Pentium M 1.70 GHz 00000001 000006d6 00000816 00000180 afe9f9bf

[Ubuntu QA Website (ubuntuqa)]

This bug has been reported on the Ubuntu ISO testing tracker.

A list of all reports related to this bug can be found here:
http://iso.qa.ubuntu.com/qatracker/reports/bugs/930447

[sudodus (nio-wiklund)]

This bug is reported to be fixed in syslinux, but it has not yet trickled down to our iso file. I hope it is true, and that we will soon have it in Saucy. Or is it a fake fix relying on fake-PAE ;-)

[Shahar Or (mightyiam)]

The status of this bug in syslinux might be wrong. I didn't receive a response from piotr.

[papukaija (papukaija)]

Could someone please reopen the syslinux and baltix tasks? Thanks.

[Karl Haines (karlmatthewhaines)]

I currently use a Dell Latitude D500 laptop, equipped with a Pentium M processor. I was able to install 13.04 by creating a custom USB startup disk and following directions from http://www.webupd8.org/2012/05/how-to-install-ubuntu-1204-on-non-pae.html and using newer non-PAE kernels from http://bazzar.launchpad.net/~webtom that matched 13.04. I should have created an ISO for everyone who didn't want to go through the process theirself, but hind-sight is always 20/20!

I was very surprised to come accross this thread and see how many other users were having to deal with this issue! This hardware is not trash yet, it actually runs Ubuntu 13.04 quite nicely, however I am planning on doing a downgrade and repurposing this machine as my daily user (which is how i ended up here, downloading Luiz's ISO), thanks Luiz!

I am a programmer by profession and a Linux lover by heart. If there is anything I can do to help with this situation, I would be glad to. I'm sure that if a few of us team up and make a solid effort, we can get something done about this issue.

Anyone else interested in getting this issue taken care of, please message me ASAP!

[sudodus (nio-wiklund)]

Hi Karl,

There is also the fake-PAE method

https://help.ubuntu.com/community/Lubuntu-fake-PAE

-o-

We are many people who would be happy, if you can and want to program the changes necessary to treat Pentium M and Celeron M CPUs with PAE capability but no PAE flag in a separate way than the very old CPUs without PAE capability. I mean so that the standard Ubuntu flavour iso files will make good CD/DVD/USB boot drives for Pentium M and Celeron M computer.

I think there are several people in the Lubuntu community, who are willing to help testing and spread the word. Finally we must convince the decision-makers of Ubuntu to include your bug-fixes into the Ubuntu code. I hope we can help each other in this task.

[Chris Bainbridge (chris-bainbridge)]

Karl, the problem is political, not technical: Ubuntu kernel team not does not want to support pentium m. The CPU PAE flag check in the kernel package would be trivial to remove.

[oliver (oliver-schinagl)]

As Chris said, it is a purely political. The changes are trivial. Syslinux would need to remove the PAE check, or ideally, add the exclusion of the Pentium-M. Syslinux can be bypassed by using grub and chainloading the iso (described in a few comments above) bypasses the boot issue.

As for kernel upgrades, yes, fake-pae solves that part. They are both hacks, but work quite well, but the same 'fix' would apply there aswell. If PAE && !Pentium-M ();

Anyay, even if you have these patches, you'd have to convince the Ubuntu council as they said 'no pae' and forgot about Pentium-M (PAE is from the pentium-pro area, so understandable to ignore CPU's before Pentium Pro.

[JohnWashington (ubuntu-johnwash)]

#194 "I'm sure that if a few of us team up and make a solid effort, we can get something done about this issue."

In your dreams, Karl. Once upon a time I believed that "Ubuntu" meant "humanity to others" and that Shuttleworth, coming from the African continent, would be concerned about inclusivity and giving third world countries the best chance to reuse old hardware. But Canonical's policies show this isn't so, if it ever was. Your best chance is to look at other distros, there are plenty that still address this.

Unfortunately, no other desktop distro for the first time user has as big a market share, so potential Linux converts fall at the first hurdle and likely return to proprietary systems.

[Shahar Or (mightyiam)]

Hey,

Why don't you just make a patch, assume good faith on the part of the maintainers and you don't have to explain the whole thing. Just write something like "Hey, you forgot Pentium M processors that actually do support PAE, although they don't declare it".

They will merge.

[Chris Bainbridge (chris-bainbridge)]

A patch effort isn't necessary. The CPU check is one line of code in kernel package preinst. Just remove that line and it goes back to working.

[oliver (oliver-schinagl)]

Chris, I think Shahar's idea is to submit the patch and hopefully gets merged so that regular users don't have to remove anything to get everything working.

But yeah, why not, submit a patch and see what happens. I am curious.

[Shahar Or (mightyiam)]

Exactly :-)

Sent from mobile נשלח מנייד
On Jul 18, 2013 3:41 PM, "oliver" <email address hidden> wrote:

> Chris, I think Shahar's idea is to submit the patch and hopefully gets
> merged so that regular users don't have to remove anything to get
> everything working.
>
> But yeah, why not, submit a patch and see what happens. I am curious.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/930447
>
> Title:
> Unable to start Ubuntu 12.04 live CD with syslinux loader on Pentium M
> x86 Laptop due to bug in PAE kernel, initramfs or syslinux
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/930447/+subscriptions
>

[sudodus (nio-wiklund)]

Two tests were added to

https://help.ubuntu.com/community/Lubuntu-fake-PAE

See this link for details

http://ubuntuforums.org/showthread.php?t=2113826&page=3&p=12729858#post12729858

For the first time a Pentium M CPU was found not suitable for fakePAE.

Our test results so far for Pentium M and Celeron M CPUs suitable for fakePAE, 'No PAE flags but 36 bit physical memory address size':

         ---- CPU name ---- -- CPUID Output of 'cpuid|grep ^00000001' --
Lowest: Celeron M 1200 Mhz 00000001 00000695 00000812 00000000 a7e9f9bf
         Celeron M 1300 MHz 00000001 00000695 00000812 00000000 a7e9f9bf
         Celeron M 1.40 GHz
         Pentium M 1.50 GHz 00000001 000006d6 00000816 00000180 afe9f9bf
         Pentium M 1600 MHz 00000001 00000695 00000816 00000180 a7e9f9bf
Highest: Pentium M 1.70 GHz 00000001 000006d6 00000816 00000180 afe9f9bf

Our test results so far for Pentium M and Celeron M CPUs not suitable for fakePAE, 'No PAE flags and only 32 bit physical memory address size':

         ---- CPU name ---- -- CPUID Output of 'cpuid|grep ^00000001' --
Lowest:
         Pentium M 1200 Mhz 00000001 00000695 00000816 00000180 a7e9fbbf
Highest:

[Shahar Or (mightyiam)]

Thank you for running the tests.

[Jörg Hartmann (jhartmann)]

The patch "Mageia 3" works just fine. ;-)

[Flames_in_Paradise (ellisistfroh-deactivatedaccount)]

On 10.04 LTS my little Dell D600I could use PAE-Kernels with no problems. (36-bits) . As per today it runs with the Mainline Kernel 3.10.7-031007-generic - lawlessly. Thanks to Prof7bits "Fake-PAE"

And this confirms comment #179:
... the Linux documentation states that if you can boot a PAE kernel, then your CPU has PAE. If you don't have PAE, then the kernel won't even boot.

Simple as that. Don't argue wt dev's why priorize the non-PAE over the PAE. see also https://en.wikipedia.org/wiki/Physical_Address_Extension#Linux . It's an non-arguable decision, even though anyone wt. a RAM > 4GiB should decide for a 64-Bit system. :-)

If the Kernel naming would clearly state it's a PAE, since the other Kernel's part was removed, we would probably not need a preinst. script to drag people away Ubuntu, even from other flavours. Like Elementary-OS

Well, if Intel thought they shouldn't equipe some Pentium M with a PAE flag (even though it's capable) they deserve a bug-report. And should publish a new updated BIOS. They always had open ears to Linux users.

 It's the old linux-users desease working rather on the effects than on the impact.

Nowadays I need to edit & manipulate the downloaded CD for an installation? That's possible & ends up with ISO-start in GRUB2. This is Not o.k.! Pls support bug#1214462 – " Kernel Description doesn't explain its a PAE-Kernel _ since 3.4.0.1.1 _ Linus would not smile _. "

[oliver (oliver-schinagl)]

So lets say, we get intel to release a microcode update for the Pentium-M, which has been discontinued for years. Lets assume they'd do that. While those could be loaded runtime I suppose, for bootable installation medium, it needs to be patched into the bios. So now you not only have to pray to the intel-gods to release a fix, but all those notebook manufactures aswell, for all Pentium-M modules, for those 3 ubuntu users left. Never gonna happen.

What's the quick and easy fix? Allow pentium-m when checking for PAE.

[Andrea Corbellini (andrea.corbellini)]

The bug status was changed either by mistake or by a vandal.

[sudodus (nio-wiklund)]

to everybody subscribing for this bug report,

If you know *which file contains the test for the PAE flag* (the source code of syslinux I suppose) and *how to get it*, please post the information in a new comment here.

Best regards
sudodus

[roland aus köln (devzero-c)]

Hi there,
here is another user of Lubuntu-"this is good for your old hardware"-Distro and it DOES NOT INSTALL on my Pentium 1,5Ghz Notebook.

I spent hours on this today to make it work and my dsl line was glowing becaus of this, too - as fetching 12.04 for installation and then doing online upgrade 12.04->12.10->13.04->13.10 needed some hours and also some gigabytes download.

After reading trough all the ifs and buts i cannot believe how arrogant developers must be for not caring about issues with older hardware.

I`m really pissed. it`s not the first time that i come across some "naah, that`s old. buy a new hardware"-bug. linux always breaks sooner or later on a hardware where it worked without a problem before. and the bugs causing it get never fixed.

This is planned obsolescense, as you, developers who decided to drop non-pae kernels and who decided to drop support for Pentium M processor are responsible for thousands of working notebooks being put to trash because they can`t be used with a recent and supported OS !!!

So - all the fuzz is just all about a bug in syslinux ?

HPA does not want to fix it?

Can someone point me where he is telling about that? If he does not want to fix it, some other may fix it for him.

[roland aus köln (devzero-c)]

sudodus, it`s already being mentioned here

[roland aus köln (devzero-c)]

i did a comparison of the installer`s kernel .config with the kernel .config of my working system - and they appear nearly identical:

3c3
< # Linux/i386 3.11.0-12-generic Kernel Configuration
---
> # Linux/i386 3.11.0-13-generic Kernel Configuration
68c68
< CONFIG_VERSION_SIGNATURE="Ubuntu 3.11.0-12.19-generic 3.11.3"
---
> CONFIG_VERSION_SIGNATURE="Ubuntu 3.11.0-13.20-generic 3.11.6"
5435c5435
< CONFIG_USB_OTG=y
---
> # CONFIG_USB_OTG is not set
5703d5702
< CONFIG_USB_ZERO_HNPTEST=y
6717a6717
> CONFIG_DM_RAID45=m

so i would confirm the suspicion that the problem must be related to the bootloader somehow.

but the question is if syslinux makes the kernel fail boot or if grub has something "inside" which avoids the kernel fail to boot.

if PAE cpu flag is not available in some processor`s CPUID, i wonder why the kernel can boot. does linux kernel not check for cpu flags and does it detect pae capability differently (i.e. more reliable?)

[roland aus köln (devzero-c)]

i have copied the installer disks kernel and initrd to my working system and it (indeed) boots via grub.

so it`s either syslinux which makes it fail boot or grub has something alongside wich makes it work "magically" (as the kernel checks for PAE i wonder, why it boots at all)

[sudodus (nio-wiklund)]

On 2013-11-24 11:13, roland aus köln wrote:
> i did a comparison of the installer`s kernel .config with the kernel
> .config of my working system - and they appear nearly identical:
>
> 3c3
> < # Linux/i386 3.11.0-12-generic Kernel Configuration
> ---
>> # Linux/i386 3.11.0-13-generic Kernel Configuration
> 68c68
> < CONFIG_VERSION_SIGNATURE="Ubuntu 3.11.0-12.19-generic 3.11.3"
> ---
>> CONFIG_VERSION_SIGNATURE="Ubuntu 3.11.0-13.20-generic 3.11.6"
> 5435c5435
> < CONFIG_USB_OTG=y
> ---
>> # CONFIG_USB_OTG is not set
> 5703d5702
> < CONFIG_USB_ZERO_HNPTEST=y
> 6717a6717
>> CONFIG_DM_RAID45=m
>
> so i would confirm the suspicion that the problem must be related to the
> bootloader somehow.
>
> but the question is if syslinux makes the kernel fail boot or if grub
> has something "inside" which avoids the kernel fail to boot.
>
> if PAE cpu flag is not available in some processor`s CPUID, i wonder why
> the kernel can boot. does linux kernel not check for cpu flags and does
> it detect pae capability differently (i.e. more reliable?)
>
The flag is only an indicator, that is used by other programs to decide
what to do. Most Celeron M and Pentium M computers have PAE capability,
at least they can manage to run PAE kernels without problems. But there
is no PAE flag.

There are at least two checks,

- in syslinux that prevents installation if there is no PAE flag

- in the system for upgrading, that prevents upgrading to a new kernel.
I guess more precisely it prevents upgrading to a PAE kernel, if there
is no PAE flag.

But there is no check in grub, so that way we can boot the kernel.

-o-

If you have the time and insight to fix these issues, we are many people
who would be happy :-)

Best regards
Nio

[roland aus köln (devzero-c)]

syslinux CAN check if there is PAE, but it looks that ubuntu makes no use of that.

the message comes from the linux kernel, so the kernel checks for PAE and does not find PAE. this is somewhat expected, when the cpu lies about it.

but the question is, why booting via grub makes things work. so i`m not sure if syslinux is to blame. think we need to find out how pae detection in linux kernel works and why it passes when booted via grub

i found that ubuntu is using an older syslinux version, so i made a bootable usb-stick with a recent syslinux version and directly booted via syslinux.cfg:

DEFAULT test
LABEL test
KERNEL /casper/vmlinuz
APPEND initrd=/casper/initrd.lz

but the problem remains:

SYSLINUX 6.02 EDD 2013-10-13 Copyright (C) 1994-2013 H. Peter Anvin et al
Loading /casper/vmlinuz... ok
Loading /casper/initrd.lz... ok
This kernel requires the following features not present on the CPU:
pae
Unable to boot - please use a kernel appropriate for your CPU.

[roland aus köln (devzero-c)]

one interesting finding:

i tried grub4dos (0.95) to boot /casper/vmlinuz+initrd.lz from USB and it also does NOT boot , but also bails out with the "missing pae" message.

so, apparently it`s NOT a syslinux triggered problem.

maybe this has something to do with realmode ?

think we need to find out why normal grub _does_ work.

[sudodus (nio-wiklund)]

On 2013-11-24 14:54, roland aus köln wrote:
> one interesting finding:
>
> i tried grub4dos (0.95) to boot /casper/vmlinuz+initrd.lz from USB and
> it also does NOT boot , but also bails out with the "missing pae"
> message.
>
> so, apparently it`s NOT a syslinux triggered problem.
>
> maybe this has something to do with realmode ?
>
> think we need to find out why normal grub _does_ work.

This is getting really interesting. My old 'truth', that the first
problem is in syslinux is proven false :-D

[roland aus köln (devzero-c)]

interesting - yes :)

please everybody excuse my harsh words in a previous comment. this is a really obscure issue....

[roland aus köln (devzero-c)]

so here we go:

i had some very short conversation with hpa (thank you for that valuable input!) and he is telling, that grub2 bypasses standard linux kernel initialization code . he thinks this is not good at all and wonders, what makes grub2 think it`s smart enough to do so.

by piece of luck, grub2 "circumvents" the pae problem with cpu`s not correctly reporting it - and that made it possible to provide installation-quirks like being described on the ubuntu wiki.

from my limited point of view, skipping standard initialization code from the linux kernel does not appear to be smart to me too, because it made finding the real problem so extremly difficult.

i guess, if grub2 (and so the kernel) would just behave the standard way, maybe we already would have an appropriate quirk in the linux initialization code, or at least a hint that there are cpu`s which have pae but do not report about - and a kernel param to force skipping the pae check.

you can easily verify that hpa is correct - just edit the grub commandline and replace "linux /boot/vmlinz....and "initrd /boot/initrd..." with "linux16 /boot/vmlinuz..." and "initrd16 /boot/inititrd..." and - voila, the problem also happens on bootable systems.

i think this is stuff for being further discussed on linux-kernel and grub mailing lists...

[roland aus köln (devzero-c)]

now i also understand, why setting vga textmodes on linux kernel does not work anymore with ubuntu. i have come to ubuntu very recently (i`t not the only distro i used) and always wondered about that:

http://savannah.gnu.org/bugs/?34610

[roland aus köln (devzero-c)]

reference for lkml discussion:

http://marc.info/?l=linux-kernel&m=138531102626148&w=2

[sudodus (nio-wiklund)]

Now the snowball is rolling and it can take several directions.

1. The direction we want, that there will be bug-fixes, so that Celeron M and Pentium M CPUs will be accepted for PAE kernels from the very beginning of the installation :-)

2. A direction so that nothing happens to the Ubuntu code, maybe the most probable direction :-|

3. A direction I fear, that grub2 and/or the kernel module will close the loop-hole for fake-PAE. In that case we have to rely on flavours and respins built on Ubuntu 12.04 LTS lasting until April 2017, but without some modern features. Or tell the users to use another distro :-(

I hope it will roll in a favourable direction.

[roland aus köln (devzero-c)]

reference for grub mailinglist discussion (unfortunately, neither lkml nor grub-ml post had been commented yet):

http://lists.gnu.org/archive/html/help-grub/2013-11/msg00031.html

[roland aus köln (devzero-c)]

I've been tinkering around a little bit this evening....

To: <email address hidden>
Cc: "H. Peter Anvin" <email address hidden>
Subject: [PATCH] - add "fixpae" bootparam to fix/workaround #930447

Hello kernel-team,

regarding https://bugs.launchpad.net/ubuntu/+source/syslinux/+bug/930447 , i have developed a simple patch to (hopefully) provide an easy and elegant workaround/fix for the issue. I assume some hundred users (if not some thousand, as entries in bugreports are only the tip of an iceberg) are affected by this issue.

Please take a review and feel free to merge.

As defaulting to PAE kernels is specific for some distros only, i think this is no patch for mainline (yet).

I did not often do patches, nor am i a good programmer - so sorry if the patch or the submission style is not perfect.

regards
Roland

---
This patch adds a kernel bootparam "fixpae". If integrated into the kernel, this should make life more easy, as installation quirks like fake-pae package, upgrade path from ubuntu 12.04 onwards or specially crafted installation media are not needed anymore. Affected users also get a hint that their issue may be workarounded with this param. Just add the bootparam to the kernel commandline and you`re done.

       fixpae [x86] Workaround for a nasty PAE issue with older CPU`s
                       like Pentium M, as they may report PAE incapability
                       although they support it. This bootparam adds a fake
                       pae entry to the flags section in /proc/cpuinfo and
                       skips the validate_cpu() routine in arch/x86/boot/main.c
                       This is necesssary as major distros nowadays only ship
                       PAE Kernels for x86 and there is no easy workaround.

signed-off-by: Roland Kletzing <email address hidden>
---

diff -uprN linux-source-3.11.0.orig/arch/x86/boot/main.c linux-source-3.11.0/arch/x86/boot/main.c
--- linux-source-3.11.0.orig/arch/x86/boot/main.c 2013-09-02 22:46:10.000000000 +0200
+++ linux-source-3.11.0/arch/x86/boot/main.c 2013-11-26 20:56:01.175269235 +0100
@@ -146,9 +146,17 @@ void main(void)

        /* Make sure we have all the proper CPU support */
        if (validate_cpu()) {
- puts("Unable to boot - please use a kernel appropriate "
- "for your CPU.\n");
- die();
+ if(cmdline_find_option_bool("fixpae"))
+ puts("fixpae bootparam active. Skipping CPU "
+ "validation and continuing Kernel execution\n");
+ else {
+ puts("Unable to boot - please use a kernel appropriate "
+ "for your CPU.\n"
+ "Hint: CPU`s like Pentium M may incorrectly report "
+ "on PAE incapability, so you may try booting with "
+ "fixpae bootparam as a workaround.\n");
+ die();
+ }
        }

        /* Tell the BIOS what CPU mode we intend to run in. */
diff -uprN linux-source-3.11.0.orig/arch/x86/kernel/cpu/proc.c linux-source-3.11.0/arch/x86/kernel/cpu/proc.c
--- linu...

[sudodus (nio-wiklund)]

Great job Roland :-)

[Mantas Kriaučiūnas (mantas)]

It seems, that Roland's (devzero-c) patch should be improved - see this email from kernel developer HPA:

H. Peter Anvin <email address hidden> wrote:
> The right way to do this is to have the option, if it is found, simply set the bit in the CPU array.
> However, it is important to verify that we are on the affected type of CPU, specifically the affected Pentium Ms.

[roland aus köln (devzero-c)]

i don`t have the skills for this and i don`t understand, why hpa forwarded discussion to lkml at this early stage, as it is a patch primary meant for the ubuntu kernel to adress an ubuntu specific problem.

sure it is not perfect and not the optimal way to proceed, but it is a solution to a problem.

we wouldn`t have this problem and we wouldn`t have this patch if there wasn`t the discision to drop support for non-pae kernels.

so hopefully the ubuntu kernel maintainers will give a comment on this patch or pick it up to make a version which fulfils kernel code-quality/style or patch-submission requirements.

this patch is rather unintrusive and it easily solves a problem for many users. i spend many hours on debugging this issue and developing this patch, so for now please understand that i`m done with it.

lkml reference: http://marc.info/?l=linux-kernel&m=138550884425543&w=2

[sudodus (nio-wiklund)]

2013-11-27 11:24, Mantas Kriaučiūnas skrev:
> It seems, that Roland's (devzero-c) patch should be improved - see this
> email from kernel developer HPA:
>
> H. Peter Anvin <email address hidden> wrote:
>> The right way to do this is to have the option, if it is found, simply set the bit in the CPU array.
>> However, it is important to verify that we are on the affected type of CPU, specifically the affected Pentium Ms.
>

I would like to add Celeron M

'the affected Celeron M and Pentium M CPUs'

See for example this link

https://help.ubuntu.com/community/Lubuntu-fake-PAE#Test

[Dimitri John Ledkov (xnox)]

On 27 November 2013 10:42, roland aus köln <email address hidden> wrote:
> i don`t have the skills for this and i don`t understand, why hpa
> forwarded discussion to lkml at this early stage, as it is a patch
> primary meant for the ubuntu kernel to adress an ubuntu specific
> problem.
>

I understand your concern. The fact that certain CPUs are buggy and do
not declare PAE capability is a hardware bug which cannot be rectified
and if quirks are needed on the kernel side, the right way is to
essentially white-list / set PAE options of the affected CPUs.

ubuntu kernels do not typically apply original patches, mostly only
cherrypicks from upstream trees/vendors.

> sure it is not perfect and not the optimal way to proceed, but it is a
> solution to a problem.
>
> we wouldn`t have this problem and we wouldn`t have this patch if there
> wasn`t the discision to drop support for non-pae kernels.
>
> so hopefully the ubuntu kernel maintainers will give a comment on this
> patch or pick it up to make a version which fulfils kernel code-
> quality/style or patch-submission requirements.
>
> this patch is rather unintrusive and it easily solves a problem for many
> users. i spend many hours on debugging this issue and developing this
> patch, so for now please understand that i`m done with it.
>
> lkml reference: http://marc.info/?l=linux-kernel&m=138550884425543&w=2
>

I understand your concerns, and I'll talk to kernel team, to see if we
can get this properly worked out and upstreamed.

Regards,

Dmitrijs.

[Chris Bainbridge (chris-bainbridge)]

Actually the problem is not Ubuntu specific as Debian also includes a pae flags check in the 686 kernel package. Having a detection patch in the upstream kernel would also fix the bug in Debian, and would be better for users than having to manually alter the kernel command line.

[roland aus köln (devzero-c)]

>as Debian also includes a pae flags check in the 686 kernel package

the pae flags check is in the x86 bootcode of the kernel and apparently it depends on the bootloader if it`s being executed or not.

only grub2 seems to skips that code which does vga initialization, cpu-checking and other steps like edd ( https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/arch/x86/boot/edd.c?id=e479c8306f898fcdb9b36179071eae6338a17364 ) , which makes all the work pretty pointless.

i wonder, why skipping that code is possible at all and nobody complains (besides that silent complain of hpa)

for details, see
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/x86/boot/main.c (end of the file)
and
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/x86/boot/cpu.c
and
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/x86/boot/cpucheck.c

btw - thanks Dimitrijs for the positive response.

[Chris Bainbridge (chris-bainbridge)]

>>as Debian also includes a pae flags check in the 686 kernel package

>the pae flags check is in the x86 bootcode of the kernel and apparently it depends on the bootloader if it`s being executed or not.

I meant the check in the linux-image-* package preinst, the check prevents the package from being installed on systems where pae is not in cpu flags. I thought that Debian had the same check, but in fact it doesn't now - either I remembered incorrectly or the check was removed at some point.

[Andy Whitcroft (apw)]

For clarity the error referred to in this bug is emitted by the kernel from its 16bit entry point. The same check is not made when booted via its 32bit entry point. This is arguably a bug, the kernel should check and refuse to run when required features are not present. In this case the main issue is that some PAE capable CPUs do not correctly list this in their CPUID capability bits. The logical fix is to fake the PAE bit for those cpus, through a whitelist.

[sudodus (nio-wiklund)]

There is a start for such a whitelist in comment #203 :-)

If *you* have a Celeron M or Pentium M different from those in that list, please supply your result of the PAE test described in

https://help.ubuntu.com/community/Lubuntu-fake-PAE

[Dimitri John Ledkov (xnox)]

On 29 November 2013 16:22, Andy Whitcroft <email address hidden> wrote:
> For clarity the error referred to in this bug is emitted by the kernel
> from its 16bit entry point. The same check is not made when booted via
> its 32bit entry point. This is arguably a bug, the kernel should check
> and refuse to run when required features are not present. In this case
> the main issue is that some PAE capable CPUs do not correctly list this
> in their CPUID capability bits. The logical fix is to fake the PAE bit
> for those cpus, through a whitelist.
>

A whitelist of some CPUIDs is collected here:

https://help.ubuntu.com/community/Lubuntu-fake-PAE#Test

(first table).

There might be more.

Regards,

Dmitrijs.

[roland aus köln (devzero-c)]

here is amore comprehensive database containing cpuid information. http://www.cpu-world.com/cgi-bin/CPUID.pl?s=21
unfortunately i have no clue how to make a whitelist from that.

[roland aus köln (devzero-c)]

hopefully, this should (hopefully) be a comprehensive list of all affected CPU`s:

Intel(R) Celeron(R) M processor 1400MHz 00000695 00000812 00000000 A7E9F9BF
Intel(R) Celeron(R) M processor 1400Mhz 000006D6 00000812 00000000 AFE9F9BF
Intel(R) Celeron(R) M processor 1500MHz 00000695 00000812 00000000 A7E9F9BF
Intel(R) Pentium(R) M processor 1300MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1400MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1500MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1600MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1600MHz 000006D6 00000816 00000180 AFE9FBBF
Intel(R) Pentium(R) M processor 1700MHz 000006D6 00000816 00000180 AFE9F9BF
Intel(R) Pentium(R) M processor 1800MHz 000006D6 00000816 00000180 AFE9F9BF
Intel(R) Pentium(R) M processor 2000MHz 000006D6 00000816 00000180 AFE9F9BF

the numbers after the cpu type are CPUID 0000_0001 (like on https://help.ubuntu.com/community/Lubuntu-fake-PAE#Test )

[sudodus (nio-wiklund)]

Hi Roland,

Nice list :-) How did you compile it? Are all those CPUs tested, or are you adding some of the CPUs versions via logical conclusions?

Best regards
Nio

[roland aus köln (devzero-c)]

it`s based on information from cpu-world.com, they have a large "real world" cpuid database (user submitted data).

it`s ineed not tested, but based on the infos at https://help.ubuntu.com/community/Lubuntu-fake-PAE#Test i would logically assume, that all of these behave identically. they all have no PAE flag shown and are the only "non pre-dinosaur-age" cpu`s missing that feature.

i`m curious about the 1,2 ghz one listed, as this is not in the database at all.

oh, and i see there is one of the 1600Mhz Pentium M with AFE9FBBF, so this one could fail to work, too.

[oliver (oliver-schinagl)]

Not seeing my 2100 MHz Pentium-M (100 MHz FSB f i'm not mistaken) I'll try to run the test sometime soon.

[Julius Schwartzenberg (jschwart)]

With Ubuntu 13.10 it has become a mess. I tried to install it 4 times on my notebook now. It would go through the complete installation process and only at the end suddenly the kernel failed to install. I thought I may have connectivity problems at first. Only after manually doing a dpkg -i on the kernel deb through a root shell, I saw the error message about non-PAE.

I hope at least some warning can be added about this, because it's stupid to go through the whole process and discover at the end it's not going to work.

[roland aus köln (devzero-c)]

is your cpu on the list above (see 5 posts before) ?

[oliver (oliver-schinagl)]

Here is my full cpuinfo, Pentium-M 2.1 GHz not having the PAE flag, happily running a PAE kernel.

processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 13
model name : Intel(R) Pentium(R) M processor 2.10GHz
stepping : 6
microcode : 0x18
cpu MHz : 2100.000
cache size : 2048 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu vme de pse tsc msr mce cx8 sep mtrr pge mca cmov clflush dts acpi mmx fxsr sse sse2 ss tm pbe bts est tm2
bogomips : 4186.17
clflush size : 64
cache_alignment : 64
address sizes : 36 bits physical, 32 bits virtual
power management:

[sudodus (nio-wiklund)]

Hi Oliver,

That is the fastest Pentium M without a PAE flag reported yet in this project :-) Please post also the short output of the following command line,

cpuid|grep ^00000001

Best regards
Nio

[oliver (oliver-schinagl)]

So that's where those lines are from.

00000001 000006d6 00000816 00000180 afe9f9bf

And yeah, I bought the fastest CPU at the time that would fit in this laptop. I think the faster cores in the same series are all 133 MHz FSB and actually do report PAE if I'm not mistaken, only the 100 MHz parts such as this where wrong.

[sudodus (nio-wiklund)]

I think so too [that the faster ones with 133 MHz FSB (533 MHz transfer rate) actually do report PAE].

Thank you for that cpuid line! I'll upload it into the list at

https://help.ubuntu.com/community/Lubuntu-fake-PAE#Test

[roland aus köln (devzero-c)]

adding to the list:

Intel(R) Celeron(R) M processor 1400MHz 00000695 00000812 00000000 A7E9F9BF
Intel(R) Celeron(R) M processor 1400Mhz 000006D6 00000812 00000000 AFE9F9BF
Intel(R) Celeron(R) M processor 1500MHz 00000695 00000812 00000000 A7E9F9BF
Intel(R) Pentium(R) M processor 1300MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1400MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1500MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1600MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1600MHz 000006D6 00000816 00000180 AFE9FBBF
Intel(R) Pentium(R) M processor 1700MHz 000006D6 00000816 00000180 AFE9F9BF
Intel(R) Pentium(R) M processor 1800MHz 000006D6 00000816 00000180 AFE9F9BF
Intel(R) Pentium(R) M processor 2000MHz 000006D6 00000816 00000180 AFE9F9BF
Intel(R) Pentium(R) M processor 2100Mhz 000006D6 00000816 00000180 AFE9F9BF

ubuntu team - are, are you watching bugticket ?

[sudodus (nio-wiklund)]

Roland,

1. Thanks for keeping this issue alive :-)

2. We have evidence that these two CPUs also have PAE capability but no PAE flag:

Celeron M 1200 Mhz 00000001 00000695 00000812 00000000 a7e9f9bf
Celeron M 1300 MHz 00000001 00000695 00000812 00000000 a7e9f9bf

Best regards
Nio

[roland aus köln (devzero-c)]

So here we go....

Intel(R) Celeron(R) M processor 1200Mhz 00000695 00000812 00000000 A7E9F9BF
Intel(R) Celeron(R) M processor 1300MHz 00000695 00000812 00000000 A7E9F9BF
Intel(R) Celeron(R) M processor 1400MHz 00000695 00000812 00000000 A7E9F9BF
Intel(R) Celeron(R) M processor 1400Mhz 000006D6 00000812 00000000 AFE9F9BF
Intel(R) Celeron(R) M processor 1500MHz 00000695 00000812 00000000 A7E9F9BF
Intel(R) Pentium(R) M processor 1300MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1400MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1500MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1600MHz 00000695 00000816 00000180 A7E9F9BF
Intel(R) Pentium(R) M processor 1600MHz 000006D6 00000816 00000180 AFE9FBBF
Intel(R) Pentium(R) M processor 1700MHz 000006D6 00000816 00000180 AFE9F9BF
Intel(R) Pentium(R) M processor 1800MHz 000006D6 00000816 00000180 AFE9F9BF
Intel(R) Pentium(R) M processor 2000MHz 000006D6 00000816 00000180 AFE9F9BF
Intel(R) Pentium(R) M processor 2100Mhz 000006D6 00000816 00000180 AFE9F9BF

[Flames_in_Paradise (ellisistfroh-deactivatedaccount)]

It's also interesting for Ubuntu Kylin:

https://wiki.ubuntu.com/Ubuntu%20Kylin/1404-alpha-1-ReleaseNote#System_Requirements

;-)

[roland aus köln (devzero-c)]

No further progress ?

ubuntu/lubuntu/xubunto 14.04 & co will soon get off the starting blocks - i heard it will also be a LTS.

And all of those still being shipped with a known problem affecting thousands of users ?

Please, Ubuntu team put an hour effort into this, this would save much hassle and thousands of "workaround-hours" for your user community !

[roland aus köln (devzero-c)]

more than one and a half months gone by and nothing happened.

14.04 will be out soon, and i swear it will still have that pentium M issue.

i don`t know what you 101 guys will do, but i have one task on my todo:

wiping all computers in our fablab and replacing lubuntu with a linux distro w/o planned obsolescence inside.

[roland aus köln (devzero-c)]

bugticket closed, as ubuntu team is ignorant and will not fix this. sad but true.

[Stan Osborne (stan-ana)]

Roland,

Thanks for trying.

In 2012 an Ubuntu developer made a private version of 12.04 with PAE
support, so I am still using Ubuntu with my old laptop.

The Ubuntu team has people who could fix this the right way, but as you
said, they are lazy.

Stan

On Wed, February 12, 2014 4:39 pm, roland aus köln wrote:
> bugticket closed, as ubuntu team is ignorant and will not fix this. sad
> but true.
>
> --
> You received this bug notification because you are subscribed to a
> duplicate bug report (993939).
> https://bugs.launchpad.net/bugs/930447
>
> Title:
> Unable to start Ubuntu 12.04 live CD with syslinux loader on Pentium M
> x86 Laptop due to bug in PAE kernel, initramfs or syslinux
>
> Status in “linux” package in Ubuntu:
> Confirmed
> Status in “syslinux” package in Ubuntu:
> Invalid
> Status in Baltix GNU/Linux:
> Triaged
>
> Bug description:
> Ubuntu 12.04 doesn't start from Desktop CD or USB with syslinux boot
> loader on Pentium M 1.6Ghz or faster Pentium M CPU - displays error
> message about missing PAE feature in CPU, but *the same* *Ubuntu
> 12.04* Desktop CD/LiveUSB starts fine on *the same CPU* (and same PAE
> kernel) if GRUB boot loader is used, for example when WUBI or LiveUSB
> with GRUB boot loader, like Multisystem
> (http://liveusb.info/dotclear/index.php?pages/install ) is used!
>
> The error message is:
> "This kernel requires the following features not present on the CPU:
> pae.
> Unable to boot - please use a kernel appropriate for you CPU."
>
> THIS IS AN IMPORTANT REGRESSION! People are able to install and
> successfully use Ubuntu 12.04 on such pretty new hardware, like IBM
> Thinkpad T42 laptop with Pentium M 1700Mhz processor, but the bug in
> syslinux (or something related) forbids Ubuntu 12.04 installation.
> This bug is reproducible on lots of computers, there are several log
> files and /proc/cpuinfo file attached to this bugreport, AFAIK it's
> enough to reopen this bug.
>
> ---
> ApportVersion: 1.23-0ubuntu4
> Architecture: i386
> CurrentDmesg: Error: command ['sh', '-c', 'dmesg | comm -13
> --nocheck-order /var/log/dmesg -'] failed with exit code 1: comm:
> /var/log/dmesg: Permission denied
> MachineType: IBM 2373PPU
> dmi.bios.date: 06/18/2007
> dmi.bios.vendor: IBM
> dmi.bios.version: 1RETDRWW (3.23 )
> dmi.board.name: 2373PPU
> dmi.board.vendor: IBM
> dmi.board.version: Not Available
> dmi.chassis.asset.tag: No Asset Information
> dmi.chassis.type: 10
> dmi.chassis.vendor: IBM
> dmi.chassis.version: Not Available
> dmi.modalias:
> dmi:bvnIBM:bvr1RETDRWW(3.23):bd06/18/2007:svnIBM:pn2373PPU:pvrThinkPadT42:rvnIBM:rn2373PPU:rvrNotAvailable:cvnIBM:ct10:cvrNotAvailable:
> dmi.product.name: 2373PPU
> dmi.product.version: ThinkPad T42
> dmi.sys.vendor: IBM
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/930447/+subscriptions
>

[roland aus köln (devzero-c)]

they are probably putting the focus on the wrong things and/or are understaffed.

i will not use ubuntu anymore because of this, because ubuntu does not satisfy my expectation of a healty user and developer community.

[Elbarbudo (patricearnal)]

Ubuntu is more and more microsoft-like :

What is good for UBUNTU/Window is good for customers, no matter what they say.

This is the main problem of hegemonic situations.

By the way, LxLe hich is based on Lubuntu, runs perfectly from an Usb key on these old laptops.
And it is light enough (not like the latest Ubuntu versions) to run fluidly and quickly

[Flames_in_Paradise (ellisistfroh-deactivatedaccount)]

hi there!

reverted bug-status back to confirmed. There are other bug-statii like "opinion, incomplete, fix released, fix commited"

Non of these reflect the situation, neither. The bug-report is valid since a lot of users have confirmed it's existance.

May be "one-hundred-papercuts" is the right place if you feel a bug doesn't get enough attention foralongtime.

Another valuable thought is that one cannot blame the ubuntu-people solely for not getting this right -

Intel should include the regarded little Pentium-M-boxes in the package 'intel-microcode' - probabely they'll lend us an open ear with the forthcoming end of W$ XP. BUT they won't take any action w/o a valuable bug-report.

@ roland aus koeln: Keep in mind: a bug report disappears from list once it's marked as invalid. But is also appears in trusty

[roland aus köln (devzero-c)]

>May be "one-hundred-papercuts" is the right place if you feel a bug doesn't get enough attention foralongtime.

that`s a pointer, thanks.

>Another valuable thought is that one cannot blame the ubuntu-people solely for not getting this right

yes, for sure intel also is to blame, but when the ubuntu team decided to drop non-pae kernel ( http://ubuntu.5.x6.nabble.com/Dropping-i386-non-PAE-as-a-supported-kernel-flavour-in-Precise-Pangolin-td366373.html ) they should better care for the consequences instead of ignoring the problems of hundreds of users. and , they cannot say that nobody complained - there had been complaints already at time of the announcement.

i would have no problem when ubuntu was simply ubuntu desktop, but there is ubuntu server, there is ltsp, there is lubuntu and there is xubuntu, and these are all suitable or even meant for old hardware. hardware you cannot use with these ubuntu derivates anymore. and there are many more ubuntu based distros - get a clue at: http://distrowatch.com/search.php?basedon=Ubuntu

so, deciding to drop non-pae ubuntu team also decided killing old hardware, as users are more likely to switch to new hardware than switching to a distro like debian.

>@ roland aus koeln: Keep in mind: a bug report disappears from list once it's marked as invalid. But is also appears in trusty

i neither maintain this bug nor can i remember touching this bug`s status information, i´m not even sure if i can change a bug`s status at all.

what i really no not understand is why adding my rather non-intrusive patch to the ubuntu kernel as a workaround to the "drop-non-pae"-decision is such a no-go and why kernel devs telling they would probably have a look and probably make a better version of it and then i never ever get any response again. to me this looks like they neither put attention on this issue nor do they appreciate/esteem work/fixes from the user community, as the work i put into this is for nothing and just a waste of time.

[roland aus köln (devzero-c)]

>May be "one-hundred-papercuts" is the right place if
>you feel a bug doesn't get enough attention foralongtime.

kernel specific bugs are excluded - see https://wiki.ubuntu.com/One%20Hundred%20Papercuts/Papercuts

[Chris Bainbridge (chris-bainbridge)]

LKML thread: https://lkml.org/lkml/2014/2/25/18

[roland aus köln (devzero-c)]

good news, thanks !

[sudodus (nio-wiklund)]

+1
(good news, thanks)

[roland aus köln (devzero-c)]

dear ubuntu kernel team, please consider adding the patch from https://lkml.org/lkml/2014/2/26/192 to the ubuntu kernel, so the pae issue will be fixed with 14.04

[Luke (luk-gunner)]

I am not a dev, not a skilled programmer, but only a loyal and satisfied ubuntu user since I switched from Suse to Dapper Drake 6.06 on my deprecated Pentium M.
This old and poor notebook have seen also Windows XP, Windows Vista and now it can also run Windows 7 with not much pretends. I decided to give him new life upgrading from 11.10 and make Xubuntu 13.10 the only OS installed in my notebook and guess what?? I cannot install it. Why?? PAE Kernel!! I had to install 12.04, install fake-pae and the make 3 (THREE) dist upgrades! (after days of research)
The system goes smooth and fast as it has never been, so what's the problem of adding those few lines of code in linux kernel?
Why force users that have a PAE capable cpu (unlikely not advertised) to make tricks and 3-4-5 etc distribution upgrades just to be able to install the latest distribution that could also run with no problems at all?
I understand that old hardware cannot been supported forever...but this is not the case for Pentium M.
Should I have stayed with Windows 7 or even XP? This will be the scenario the first time I will be in need of formatting my notebook (don't have the will/the time to make 4-5-... dist upgrade from 12.04 everytime).
I really thanks all the devs, but the decision to drop off pae without thinking about consequencies, is a little bit stupid.
I decided to write this comment just to point out that LIKE ME, there are thousands (not hundreds) of desktop users like me that decided to return to run Windows on their systems after that decision and they didn't come to launchpad to make you know.
Just search "Ubuntu non PAE" on google...

@Roland
Thanks for your efforts!

[roland aus köln (devzero-c)]

Luke, thanks for letting us know and for the credits.

What you write is exactly the reason why i spent my free time on help fixing this and it`s also the reason, why i was constantly nagging here - and why i was (or still am, as it`s not fixed yet) very dissapointed, too.

I hope that ubuntu will put the fix in very soon, as release date for 14.04 is just some weeks ahead....

[Luke (luk-gunner)]

Thanks again Roland for spending your time into this...I read the mails on LKML, and since I don't know how it works (I am only a simple user) I am asking...have we got more chances now to see this problem fixed somewhere in the future or am I hoping for something that could never arrive?

[sudodus (nio-wiklund)]

'...Why?? PAE Kernel!! I had to install 12.04, install fake-pae and the make 3 (THREE) dist upgrades! (after days of research). Why force users that have a PAE capable cpu (unlikely not advertised) to make tricks and 3-4-5 etc distribution upgrades just to be able to install the latest distribution that could also run with no problems at all? ...'

If you had searched a little bit more, or asked at the Ubuntu Forums, maybe you had found

https://help.ubuntu.com/community/Lubuntu-fake-PAE

and you had been able to install Lubuntu 13.10 directly without all those upgrading steps.

[Luke (luk-gunner)]

I know about lubuntu fake pae, and upgrading from 12.04 is one of the options. By the way, this is not the point.. And what about if someone needs Xubuntu or something else running out of the box? And when 14.04 will be released, there will be a lubuntu fake pae version? These are all workarounds with advantages and disadvantages, while the problem could be easily fixed by forcing Pae flag on Pentium M.

[sudodus (nio-wiklund)]

All the main Ubuntu flavours of version 13.10 are available as tarballs to be installed with the One Button Installer,

ubuntuforums.org/showthread.php?t=2172971

and this works with Pentium M and Celeron M CPUs. Experimental Trusty tarballs are already uploaded,

http://ubuntuforums.org/showthread.php?t=2172971&p=12931564#post12931564

There are community efforts to make a non-pae kernel for Trusty Tahr. See this link

http://ubuntuforums.org/showthread.php?t=2172971&p=12940628#post12940628

-o-

But the code snipped published by Chris Bainbridge (in comment #260) makes me hope that we will soon get a main-stream ubuntu kernel that will include Pentium M and Celeron M CPUs, so that we can focus our community efforts on other problems.

[roland aus köln (devzero-c)]

i read "experimental" in those links - and i think there are people who dislike installing their operating system in a re-packaged way from a community based, untrusted source. nothing against you, nio, you appear trustworthy - but i`m one of those.

if all the community efforts had been put into the kernel fix instead of fake pae or building alternative installation methods , then we would not sit here and discuss. it may be a logical step, that these workaround exist, but it`s always better to adress the root cause instead of tempering with a symptom.

it was good work to provide those workarounds, but it would be even better work to integrate the kernel fix from upstream now and help bugging the ubuntu devs to do so.

while writing this, i bet the kernel based fix had probably alrady being done if the community workarounds never existed, because while they existed the ubuntu devs could lean back and tell the user: Pentium M? naah, that`s old, watch out for fake-pae and alternative installer disks as a workaround. that probably HELPED a LOT of users but it did not help fixing the root cause, because many users never needed to complain because of that. linux people like tinkering...oh yes.

[sudodus (nio-wiklund)]

There were *very* strong signals that the board, deciding about the main-stream ubuntu edition, had decided to stop supporting non-pae CPUs including Pentium M amd Celeron M, and that the decision was not going to be changed.

You think that fake-PAE has been counter-productive, that it has done more harm than good. Maybe you are right, maybe not. I still think it was right to offer it as a solution to users, who would otherwise not use these computers with modern Ubuntu based versions (because I thought that the decision was not going to be changed). However, I'm hoping it will change now, that there will be Trusty kernels, that can be booted and upgraded with Pentium M amd Celeron M.

I can understand that you and many other people do not trust minor distros or the organisations behind them or community re-spins. But who can we trust? I think the [big] companies are at least as likely to create and hide back-doors for themselves or some national intelligence service organisation.

[oliver (oliver-schinagl)]

I'm reprinting what I wrote in comment #105 because it is still so very easy to install PAE, though it does require a little effort.

1) Download ISO of ubuntu flavor you wish to use, for example, xubuntu. I wil use the name xubuntu.iso, for convenience.
2) partition and format USB stick, in this example, /dev/sdb and /dev/sdb1. The USBStick has to be larger then the iso. Use mkvfat for formatting of /dev/sdb1.
3) Eject and insert the USB to have the automounter mount the USB drive. Make note of the location.
4) Use grub to install grub to the USB stick
      grub-install --no-floppy --root-directory=/media/ubuntu/2341-af31/ (obviously this will have to match the mounted path)
5) grub-mkconfig > /meda/ubuntu/2341-af31/boot/grub/grub.cfg

Open grub.cfg in your favorite editor and remove all the menuentry sections to replace it with the following.

menuentry "Xubuntu (32bit)" {
  iso_path=/xubuntu-13.10-dvd-i386.iso
  export iso_path
  search --set --file $iso_path
  loopback loop $iso_path
  root=(loop)
  configfile /grub/loopback.cfg
  loopback --delete loop
}

Save, unmount and reboot from the USB stick.

What will happen is, the USB stick will be booted as usual by grub, grub will then do a loopback mount of the iso, as is (so no modification to it required). The beauty of this all is, you can add MORE iso's and just add menu entries for them. So xubuntu 32bit, 64bit, gnomebuntu, you name it, you can add it. The only requirement is that the iso does actually support grub booting (e.g. has the /grub/loopback.cfg). Not all iso's have this!

After the installer has finished, it is required however to install fake-pae to make updates function properly, so to that extend, this bugreport is still relevant.

[roland aus köln (devzero-c)]

please let`s stop discussion as any further discussion does not add anything valuable to the solution.

i think 272 comments (=thousands of lines of text) is enough discussion for a problem which can be entirely fixed for everyone by some kernel developer adding 22 lines of code to the linux kernel.

[Chris Bainbridge (chris-bainbridge)]

There's a new pair of patches: https://lkml.org/lkml/2014/2/28/258 and https://lkml.org/lkml/2014/2/26/394

It's a small patch but test reports to LKML are still welcome (you don't have to be subscribed to the list to respond, just use the correct subject line).

[roland aus köln (devzero-c)]

i just tested your patch on ubuntu 13.10 with kernel from 14.04 repository (complete package build)

works like a charm!

i would recommend adding the newly introduced param to to Documentation/kernel-parameters.txt , see my patch at

https://bugs.launchpad.net/baltix/+bug/930447/comments/224

Thanks for your work !

[Andy Whitcroft (apw)]

@Chris -- I have taken the liberty of cleaning up the commentary on those two patches and applying them to a 14.04 kernel for testing. You might want to do something similar (feel free to steal the text I used and clean it up) and then submit the patches together to upstream for consideration; as things have been left I doubt they will be applied as the diff and commentary were separate. Also could you test these kernels and let me know if they work at all, I do not have affected h/w. We can then consider this for the 14.04 kernel. Kernels are at the URL below:

    http://people.canonical.com/~apw/lp930447-trusty/

[roland aus köln (devzero-c)]

works for me.

[sudodus (nio-wiklund)]

Works in my IBM Thinkpad T42 with Pentium M without a PAE flag but with PAE capability.

But I have only booted via grub, because I cannot remaster an iso file. Did you boot from syslinux, Roland?

[Chris Bainbridge (chris-bainbridge)]

lkml: "grub is jumping to the 32-bit entry point and skipping the entire real mode setup code. Bad grub."

You don't need to build a syslinux bootable iso. You can boot the kernel from Grub in 16-bit mode by using the linux16 command http://www.gnu.org/software/grub/manual/html_node/linux16.html and initrd16 https://www.gnu.org/software/grub/manual/html_node/initrd16.html

According to the discussion on LKML, linux16 is now the default on Fedora, and should be used by other distributions.

(If you still want to build a custom syslinux bootable system, just for fun, see the instructions at http://willhaley.com/blog/create-a-custom-debian-live-environment/ - it's about 25 lines of code in total, and not difficult to follow)

[sudodus (nio-wiklund)]

ERROR: PAE is disabled on this Pentium M
(PAE can potentially be enabled with kernel parameter
"forcepae" - this is unsupported, may cause unknown
problems, and will taint the kernel)
This kernel requires the following features not present on the CPU:
pae
Unable to boot - please use a kernel appropriate for your CPU

-o-

So I used the following modified lines in grub.cfg

 linux16 /boot/vmlinuz-3.13.0-14-generic root=UUID="string" ro quiet splash forcepae $vt_handoff
 initrd16 /boot/initrd.img-3.13.0-14-generic

and it boots and runs :-)

WARNING: Forcing PAE in CPU flags
...

[sudodus (nio-wiklund)]

Thanks Chris :-)

[Shahar Or (mightyiam)]

Thank you Brian, Chris and everyone sudodus and everyone involved.

נשלח מנייד Sent from mobile
On Mar 3, 2014 8:55 PM, "sudodus" <email address hidden> wrote:

> ERROR: PAE is disabled on this Pentium M
> (PAE can potentially be enabled with kernel parameter
> "forcepae" - this is unsupported, may cause unknown
> problems, and will taint the kernel)
> This kernel requires the following features not present on the CPU:
> pae
> Unable to boot - please use a kernel appropriate for your CPU
>
> -o-
>
> So I used the following modified lines in grub.cfg
>
> linux16 /boot/vmlinuz-3.13.0-14-generic root=UUID="string" ro
> quiet splash forcepae $vt_handoff
> initrd16 /boot/initrd.img-3.13.0-14-generic
>
> and it boots and runs :-)
>
> WARNING: Forcing PAE in CPU flags
> ...
>
>
> ** Attachment added: "computer-info.zip"
>
> https://bugs.launchpad.net/baltix/+bug/930447/+attachment/4005394/+files/computer-info.zip
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/930447
>
> Title:
> Unable to start Ubuntu 12.04 live CD with syslinux loader on Pentium M
> x86 Laptop due to bug in PAE kernel, initramfs or syslinux
>
> Status in “linux” package in Ubuntu:
> Confirmed
> Status in “syslinux” package in Ubuntu:
> Confirmed
> Status in Baltix GNU/Linux:
> Triaged
>
> Bug description:
> Ubuntu 12.04 doesn't start from Desktop CD or USB with syslinux boot
> loader on Pentium M 1.6Ghz or faster Pentium M CPU - displays error
> message about missing PAE feature in CPU, but *the same* *Ubuntu
> 12.04* Desktop CD/LiveUSB starts fine on *the same CPU* (and same PAE
> kernel) if GRUB boot loader is used, for example when WUBI or LiveUSB
> with GRUB boot loader, like Multisystem
> (http://liveusb.info/dotclear/index.php?pages/install ) is used!
>
> The error message is:
> "This kernel requires the following features not present on the CPU: pae.
> Unable to boot - please use a kernel appropriate for you CPU."
>
> THIS IS AN IMPORTANT REGRESSION! People are able to install and
> successfully use Ubuntu 12.04 on such pretty new hardware, like IBM
> Thinkpad T42 laptop with Pentium M 1700Mhz processor, but the bug in
> syslinux (or something related) forbids Ubuntu 12.04 installation.
> This bug is reproducible on lots of computers, there are several log
> files and /proc/cpuinfo file attached to this bugreport, AFAIK it's enough
> to reopen this bug.
>
> ---
> ApportVersion: 1.23-0ubuntu4
> Architecture: i386
> CurrentDmesg: Error: command ['sh', '-c', 'dmesg | comm -13
> --nocheck-order /var/log/dmesg -'] failed with exit code 1: comm:
> /var/log/dmesg: Permission denied
> MachineType: IBM 2373PPU
> dmi.bios.date: 06/18/2007
> dmi.bios.vendor: IBM
> dmi.bios.version: 1RETDRWW (3.23 )
> dmi.board.name: 2373PPU
> dmi.board.vendor: IBM
> dmi.board.version: Not Available
> dmi.chassis.asset.tag: No Asset Information
> dmi.chassis.type: 10
> dmi.chassis.vendor: IBM
> dmi.chassis.version: Not Available
> dmi.modalias:
> dmi:bvnIBM:bvr1RETDRWW(3.23):bd06/18/2007:svnIBM:pn2373PPU:pvrThinkPadT42:rvnIBM:rn2373PPU...

[roland aus köln (devzero-c)]

>According to the discussion on LKML, linux16 is now the default on Fedora, and should be used by other distributions.

i installed fedora20 for a test and so did hpa, and for both of us, the result is different from what is being told:

http://marc.info/?l=linux-kernel&m=139380554401194&w=2

i do not see any linux16 entry in grub.cfg

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux - 3.13.0-16.36

---------------
linux (3.13.0-16.36) trusty; urgency=low

  [ Tim Gardner ]

  * Release Tracking Bug
    - LP: #1287903

  [ Andy Whitcroft ]

  * Revert "[Config] lowlatency -- turn CONFIG_IRQ_FORCED_THREADING_DEFAULT
    off temporarily"

  [ Chris Bainbridge ]

  * SAUCE: x86: set Pentium M as PAE capable
    - LP: #930447

  [ Dave Jones ]

  * SAUCE: taint: repurpose TAINT_UNSAFE_SMP to TAINT_CPU_OUT_OF_SPEC
    - LP: #930447

  [ Paolo Pisati ]

  * [Config] SND_DAVINCI_SOC && SND_AM33XX_SOC_EVM =m
  * [Config] armhf: DRM_TILCDC=m

  [ Philippe Bergheaud ]

  * SAUCE: powerpc: fix xmon disassembler for little-endian
    - LP: #1286255

  [ Tim Gardner ]

  * [Config] CONFIG_MICROCODE_EARLY=y
  * [Config] CONFIG_R8821AE=m
  * [Config] Add some virtio drivers to -virtual
    - LP: #1287401
  * [Config] inclusion-list: vesafb and virtio_balloon are built-in
  * SAUCE: vmwgfx: Expose U32_MAX

  [ Upstream Kernel Changes ]

  * Revert "drm/vmwgfx: Fix regression caused by "drm/ttm: make ttm
    reservation calls behave like reservation calls""
  * Revert "drm/vmwgfx: Fix the driver for large dma addresses"
  * usb: ehci: fix deadlock when threadirqs option is used
    - LP: #1274987, #1279081
  * Staging: rtl8812ae: Add Realtek 8821 PCI WIFI driver
    - LP: #1287298
  * intel_pstate: Remove periodic P state boost
  * intel_pstate: Add trace point to report internal state.
  * intel_pstate: Take core C0 time into account for core busy calculation
  * intel_pstate: Use LFM bus ratio as min ratio/P state
  * intel_pstate: Add support for Baytrail turbo P states
  * intel_pstate: Change busy calculation to use fixed point math.
  * PM / hibernate: Fix restore hang in freeze_processes()
  * ipmi: remove deprecated IRQF_DISABLED
  * ipmi: use USEC_PER_SEC instead of 1000000 for more meaningful
  * ipmi: fix timeout calculation when bmc is disconnected
  * ipmi: Cleanup error return
  * ipmi: Add missing rv in ipmi_parisc_probe()
  * drm/ttm: ttm object security fixes for render nodes
  * drivers: gpu: Mark functions as static in vmwgfx_kms.c
  * drivers: gpu: Mark functions as static in vmwgfx_buffer.c
  * drivers: gpu: Mark functions as static in vmwgfx_fence.c
  * drm/vmwgfx: Fix the driver for large dma addresses
  * drm/vmwgfx: Update the svga3d register header file for new device
    version
  * drm/vmwgfx: Update the driver user-space interface for guest-backed
    objects
  * drm/vmwgfx: Replace vram_size with prim_bb_mem for calculation of max
    resolution
  * drm/vmwgfx: Update the svga register definition
  * drm/vmwgfx: Adapt capability reporting to new hardware version
  * drm/vmwgfx: Add MOB management
  * drm/vmwgfx: Hook up MOBs to TTM as a separate memory type
  * drm/vmwgfx: Read bounding box memory from the appropriate register
  * drm/vmwgfx: Add the possibility to validate a buffer as a MOB
  * drm/vmwgfx: Hook up guest-backed queries
  * drm/vmwgfx: Detach backing store from its resources when it is evicted
  * drm/vmwgfx: Hook up guest-backed contexts
  * drm/vmwgfx: Hook up guest-backed surfaces
  * drm/vmwgfx: Add guest-backed shaders
  * drm/vmwgfx: Validate guest...

[roland aus köln (devzero-c)]

this is really good news, thanks!

[roland aus köln (devzero-c)]

apparently, the trusty-tahr daily build cd-images already contain the fixed kernel, so if you have an affected system and want to try installing (l/x)ubuntu you can get them here:

http://cdimage.ubuntu.com/lubuntu/daily-live/current/
http://cdimage.ubuntu.com/xubuntu/daily-live/current/
http://cdimage.ubuntu.com/daily-live/current/

[sudodus (nio-wiklund)]

Now it is time for me to say thank you :-)

First of all, *I want to thank everybody contributing* to solving this bug, or complex of bugs, concerning Pentium M and Celeron M CPUs and new Ubuntu based kernels.

Then I want to mention roland aus köln (devzero-c), who continued pushing this issue when we had given up (either completely or like me, resorting to work-arounds).

And I want to thank Chris Bainbridge, who was able to do the actual bug-fixes that made the ship turn around. I don't know if you are a developer or a very skilled user, but you made a great difference.

Finally I want to thank you Chris for sharing the link to instructions, that actually got me over the threshold to build a custom syslinux bootable system. I'm working right now on a debian system, that should be able to install from a CD/DVD or USB drive to very different systems (including very old and limited systems) by flashing compressed image files (img.xz files) or iso files and growing the file system when there is extra space available.

I show the link again for everybody else, who might be interested. It works for me with a couple of minor tweaks, the main one that I needed to import a gpg key separately.

Best regards
sudodus alias Nio

----- Comment #279 -----
Chris Bainbridge (chris-bainbridge) wrote on 2014-03-03: #279

lkml: "grub is jumping to the 32-bit entry point and skipping the entire real mode setup code. Bad grub."

You don't need to build a syslinux bootable iso. You can boot the kernel from Grub in 16-bit mode by using the linux16 command http://www.gnu.org/software/grub/manual/html_node/linux16.html and initrd16 https://www.gnu.org/software/grub/manual/html_node/initrd16.html

According to the discussion on LKML, linux16 is now the default on Fedora, and should be used by other distributions.

(If you still want to build a custom syslinux bootable system, just for fun, see the instructions at http://willhaley.com/blog/create-a-custom-debian-live-environment/ - it's about 25 lines of code in total, and not difficult to follow)
-----

[roland aus köln (devzero-c)]

one thing to add - on lubuntu iso i just tested, entering a kernel param at boot/installation-time does not seem to find it`s way into grub.cfg - so it needs to be re-added after installation.

since i´m not a long-time ubuntu user, i`m unsure if this a bug or by design....

[Luke (luk-gunner)]

Yeah!!!! Thank you devs, thank you Roland, thank you Chris and all of you.....
Ok, now I'm gonna try and download the iso...

[Christiansen (happylinux)]

This is absolutly amasing news guyes, had lost any hope that this status on this bug would ever change. Thankyou to the most persistent advocats arguing this bug should be solved, and not least to Chris and others involved in figuring out howto...

After reciving these great news, the first thing I did was to download the Kubuntu Trusty daily image (2014.03.06) from cdimages.ubunut.com. Used a 12.04.4 box an flashed the Trusty ISO to an USB stick, using the "Start Disk Creator" a so often before. Booting my good old ThinkPad R51 from the USB stick, just to eksperience the "your CPU i not supported ... no PAE". Restarded the R51 and added the kernel "forcepae" parameter to the syslinux parameter line.
Things looked excactly the same as so often before, installing any sort of 12.04 ubuntu flavors. Back to the other ubuntu box, re-fdisk and format the USB, "Start Disk Creator" and then ever so often used grub2 installation to USB when creating 12.04.x USBs for older hardware:

 sudo grub-install --recheck --boot-directory /media/usb-stick/boot /dev/sdX
 nano /media/usb-stick/boot/grub.cfg

menuentry 'Start Kubuntu' --class ubuntu --class gnu-linux --class gnu --class os {
 recordfail
 insmod gzio
 insmod part_msdos
 insmod ext2
 set root='(hd0,msdos1)'
 echo 'Indlaeser Linux 3.13.0-16-generic-pae ...'
 linux /casper/vmlinuz file=/cdrom/preseed/kubuntu.seed boot=casper maybe-ubiquity quiet splash --
 echo 'Indlaeser startramdisk ...'
 initrd /casper/initrd.lz
}

Returned the stick to the good old R51, at it just booted from the USB like a charm - just amasing. So right now upgrades from Precise to Trusty should work correct ?, but the old syslinux "original" of this bug still persist right ?.

So new users, and users unaware of this bug, will still experienc problems booting CDs or USB sticks created from ISOs, which with all the good work and all the posting above seems almost unbearable to me...

[roland aus köln (devzero-c)]

>Restarded the R51 and added the kernel "forcepae" parameter to the syslinux parameter line.
>Things looked excactly the same as so often before, installing any sort of 12.04 ubuntu flavors.
>but the old syslinux "original" of this bug still persist right ?.

No, it shouldn`t.

If booted via syslinux and adding forcepae as a bootparam, it should boot - regardless if being botted via grub 16-bit mode or via syslinux or whatever.

could you please re-check if you typed it correctly?

does your method of creating a bootable iso really use the kernel from the recent cd-image, or maybe there`s an older kernel in place?

if that fails for you and if you are that your cpu does have pae, please post the contents of /proc/cpuinfo.

you may attach an usb cd-rom to your r51 and try to directly boot from that.

i´m curious what`s going wrong here.

[Christiansen (happylinux)]

@roland #291

Booting Kubuntu Trusty from either USB or DVD (Kubuntu > 700MB):
Daily ISO 2014.03.06 => Kernel: Linux kubuntu 3.13.0-15-generic #35-Ubuntu SMP = not working
Daily ISO 2014.03.07 => Kernel: Linux kubuntu 3.13.0-16-generic #36-Ubuntu SMP = working (without grub BL)

I'm very sorry for this mistake, reported 1 day to soon then...

Just downloaded the Kubuntu Trusty 2014.03.07, and it boots just fine with syslinux + forcepae (with the 3.13.0-16-generic #36 kernel). Even though booting (and installing) theese old boxes from buildin IDE-DVD takes forever compared to USB stik (less than 2 minutes boot).

Thanks for the very fast response, and sorry for not dobbelt checking the kernel version. We (users) are no even precentet with instructions on how to boot these Pentium boxes, if omitting the forcepae parameter - very polished indeed.

[roland aus köln (devzero-c)]

oh, apparently kubuntu daily-build was a little bit "behind", i did only check ubuntu and lubuntu and did not expect, that kubuntu iso apparently did not have the most recent kernel.

[Chris Bainbridge (chris-bainbridge)]

Good news, the forcepae patch got added to the upstream kernel source: http://git.kernel.org/cgit/linux/kernel/git/tip/tip.git/commit/?id=69f2366c9456d0ce784cf5aba87ee77eeadc1d5e

[sudodus (nio-wiklund)]

:-)

[Colin Allinson (l-colin)]

This may not have been a burning issue in the past but, with Win XP going unsupported, there are many Celeron/Pentium M processor systems (concurrent with Win XP). that are now looking for easy solutions. I am in exactly that position and have never touched any flavour of Linux before. I needed something reasonably well packaged without a huge learning curve up front.

I tried various other flavours of Linux (including Lubuntu 12.04) and none of them really floated my boat but then I found the nonpae 12.04 desktop build mentioned by Dave Henningsson in reply #84 and that is exactly what an ignorant exile from Win XP was looking for. Easy to install, well packaged, no learning of command line fiddles and easy to use for a Linux first timer. Thanks Dave.

If it was that easy without searching then I guess more ex XP users would migrate rather than give up and go away.

Incidentally - Dave said he hadn't fully tested this build - I have given it quite a bending and have found no problems so far.

[Luiz Angelo Daros de Luca (luizluca)]

@l-colin,

14.04 works flawlessly in Celeron/Pentium M processor systems. You just need to add forcepae do the kernel command line (at boot)
On a livecd, press anykey as soon as it loads (the first purple screen). Use the F6 option to add the forcepae. You'll have something like:

 "...initrd.lz quiet splash -- forcepae"

The only problem is that 14.04 is not released yet. It might be released on April 17th. However, you can download the current development version at: http://cdimage.ubuntu.com/daily-live/current/trusty-desktop-i386.iso

If you update packages after the 14.04 release date, it will update them to match the stable version. It will be some dozens of MB but, besides that, no greater side effect.

[Stan Osborne (stan-ana)]

a todos: Gracias por la ayuda!

Stan

On Tue, April 8, 2014 1:03 pm, Luiz Angelo Daros de Luca wrote:
> @l-colin,
>
> 14.04 works flawlessly in Celeron/Pentium M processor systems. You just
> need to add forcepae do the kernel command line (at boot)
> On a livecd, press anykey as soon as it loads (the first purple screen).
> Use the F6 option to add the forcepae. You'll have something like:
>
> "...initrd.lz quiet splash -- forcepae"
>
>

[Colin Allinson (l-colin)]

Hi Luis,

Now that I have got into it I might do this if I feel the need to upgrade to the latest & greatest version - so, thanks for the howto.

Unfortunately I am afraid that many disenfranchised Win XP users will not fiddle around like this (or even read this thread to find how to).

It's a pity because, once I got past the pae issue I found something far more mature and developed than I ever suspected.

I know there is probably some good reason but it would be nice if the next version could automatically add the 'forcepae' if it found the appropriate conditions to do so.

>@l-colin,

>14.04 works flawlessly in Celeron/Pentium M processor systems. You just need to add forcepae do the kernel command line (at boot)
>On a livecd, press anykey as soon as it loads (the first purple screen). Use the F6 option to add the forcepae. You'll have something like:

> "...initrd.lz quiet splash -- forcepae"

[Minar (idiotenkonto)]

Hey guys,
So, I was looking for Ubuntu because Win Xp is not supported anymore. I carefully readed everything here, but i have no clue how i could fix my problem.

Can please someone explain what i have to do, in a language that a "noob" like me can understand? I never did anything with linux.

[Pengfei Lin (pengfei-lin)]

Ｉtried the <ubuntu-14.04-desktop-i386.iso> on my Lenovo E600A with [--forcepae] parameter.
I seemed successfully installed.
But the desktop showed me slowly just as a slow-motion video :-(
Maybe the forcepae caused that problem.
Hope David Henningsson can provide us another <ubuntu-14.04-desktop-i386-nonpae.iso>.

[roland aus köln (devzero-c)]

please try lubuntu/xubunu on such old hardware

[Chris Bainbridge (chris-bainbridge)]

> tried the <ubuntu-14.04-desktop-i386.iso> on my Lenovo E600A with [--forcepae] parameter.
> I seemed successfully installed.
> But the desktop showed me slowly just as a slow-motion video :-(
> Maybe the forcepae caused that problem.

I doubt it, a more likely cause is that you don't have 3D acceleration enabled. Try running '/usr/lib/nux/unity_support_test -p' or 'glxinfo|grep OpenGL' to check and install the proprietary Nvidia drivers.

[sudodus (nio-wiklund)]

The vast majority of Pentium M and Celeron M CPUs are suitable for fakepae or forcepae and can work with PAE kernels. But some of these processors need a non-pae kernel.

We found a Pentium M 1.70 GHz which does not work with forcepae. It gives the following output from

cpuid|grep ^00000001

00000001 000006d6 00000816 00000180 afe9fbbf

So I have updated the wiki page

https://help.ubuntu.com/community/Lubuntu-fake-PAE

See the details at

https://help.ubuntu.com/community/Lubuntu-fake-PAE#Test

Now I think there is even more reason to continue the work with a non-pae kernel for Lubuntu and ToriOS. See this link

https://help.ubuntu.com/community/9w

[Chris Bainbridge (chris-bainbridge)]

> We found a Pentium M 1.70 GHz which does not work with forcepae. It gives the following output from
>
> cpuid|grep ^00000001
>
> 00000001 000006d6 00000816 00000180 afe9fbbf

What happens when you try to boot with forcepae? Do you get an error message? Black screen?

In the wiki you have mentioned "afe9fbbf for both of the CPUs that lack PAE capability" but the 1.2GHz that was reported to not work has cpuid "a7e9fbbf" and the 1.7Ghz "afe9fbbf" which are different (a7,af)

Compiling a non-PAE kernel is trivial - you just need to `apt-get source linux-image`, set CONFIG_X86_PAE=y in .config then `dpkg-buildpackage -b -uc`. You could easily automate this from a 32-bit 14.04 daily cron job that checks whether there is a new kernel version, and if so rebuild the kernel and pushes the result to a PPA (which would have to be enabled by default in Lubuntu or whatever).

[sudodus (nio-wiklund)]

Hi Chris,

1. It is not my computer. I have one with a very similar Pentium M CPU which works with forcepae. I have asked the same questions but not yet received a full answer from the owner. I will ask again ...

2. Thanks for finding the difference between the cpuid flags "a7e9fbbf" and "afe9fbbf" (a7,af). The wiki page is corrected now.

3. In post #279 you told us about Will Haley's instructions to make a syslinux system. Starting from those instructions I made a system that can install an Ubuntu minimal text screen system with a non-pae kernel compiled by Phill Whiteside. This is described in the following link

https://help.ubuntu.com/community/9w

It is very easy to expand that minimal Ubuntu system into Lubuntu or Xubuntu (or an Ubuntu Server).

If I understand it correctly, Phill intends to maintain the non-pae kernel. Anyway, thanks for the tips about making and keeping a non-pae kernel up to date :-)

[sudodus (nio-wiklund)]

Sorry for the confusion,

*I asked*

> I have a couple of questions about your Sony Vaio with Pentium M:
>
> What happens when you try to boot a PAE kernel with forcepae? Do you get
> an error message? Black screen?
>
> Are you sure it is not 'only' a problem with some other hardware, for
> example graphics or wifi? You could try to boot with the boot option
> 'text' and switch off the wifi if possible.

*The owner replied*

> Most of my 'nonpae' observations have come inside VirtualBox where I
> noticed the 'pae/nx' checkbox is greyed out and every distro. I try that
> is pae only, informs me that I have a cpu that doesn't support the pae
> option. Trying 'forcepae' using a recent kernel eventually ended with a
> blank black screen.
>
> I tried to use a Gparted live CD recently and that gave me a similar
> message when it attempted to boot.
>
> Yesterday I squeezed some unused space out of my HDD using Gparted and
> created an empty new extra linux partition (about 9 GiB) and I'll try a
> 'real metal' install of a pae kernel sometime later today.

*I replied*

> You can test that it is not problems with the graphics chip using the
> boot options 'forcepae text' in a 'standard Lubuntu 14.04 LTS installer'.

*And the owner solved the problem*

You were quite correct to keep asking questions.

Using the 'forcepae' parameter I have now successfully installed Lubuntu 14.04 on 'baremetal'.

I'm happy with that as you can imagine. There are many ways that I tried before that failed but
obviously there is nothing that replaces a 'baremetal' install. My first clue that it may work was
when I used 'forcepae' to boot from the LiveCD that I made and that worked.

Apologies for causing you extra work and innocently misleading your research. All's well that
ends well, as the saying goes.