Build for non-PAE systems

[Alexei Humeniy]

Is there a way to build Chromium OS for processors without PAE? How?

[Grant Grundler]

On Thu, Dec 9, 2010 at 4:29 AM, Alexei Humeniy <ahum...@gmail.com> wrote:
> Is there a way to build Chromium OS for processors without PAE? How?

While I'm no expert on PAE, I believe PAE support is controlled by
this kernel config option:
chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

Modify that to "#CONFIG_X86_PAE is not set" and run the normal tools
inside a chroot to build an x86-generic image/package/etc. See below
for some suggestions.

This really begs the question of why though....Mandeep Baines
explained that PAE has been available since Pentium Pro. Hugh Dickens
pointed out that disabling PAE also disables "NX" (no execute) support
- ie code pushed on the stack could be executed. This is a security
feature. Just FYI - not trying to stop you from making use of
older/less capable HW.

Some detailed steps to do what I'm suggesting (NOT TESTED):
cd /usr/local/src/chromiumos/x86-generic/src/scripts/
./enter_chroot.sh
./cros_workon start chromeos-base/kernel # create local copy of
kernel source
cd ../third_party/kernel/files
fgrep -R PAE chromeos/config/
chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y
vi chromeos/config/i386/config.common.i386 # modify X86_PAE config option

# rebuild everything - there are ways to rebuild just the
kernel...msb promised he'd add a FAQ entry to do that :)
cd ../../../scripts
B="x86-generic"
./build_packages --board=$B --oldchromebinary && ./build_image
--board=$B --noenable_rootfs_verification && ./mod_image_for_test.sh
--board=$B -y && ./image_to_usb.sh --board=$B --to=/dev/sdc -y

hth,
grant

[Olof Johansson]

On Thu, Dec 9, 2010 at 9:12 AM, Grant Grundler <grun...@chromium.org> wrote:

On Thu, Dec 9, 2010 at 4:29 AM, Alexei Humeniy <ahum...@gmail.com> wrote:
> Is there a way to build Chromium OS for processors without PAE? How?

While I'm no expert on PAE, I believe PAE support is controlled by
this kernel config option:
   chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

Modify that to "#CONFIG_X86_PAE is not set" and run the normal tools
inside a chroot to build an x86-generic image/package/etc. See below
for some suggestions.

Part of the issue here is that we have not yet re-split the configs between chromeos and chromiumos. At the end of the chromeos release push we trimmed down the default config to be more in line with what we have on our hardware (Cr-48), with the intention of keeing the chromiumos config for x86-generic more broad so it would boot on more hardware.

Due to some build system issues we had to keep them common, and the re-split hasn't happened yet. I should get to it this week unless someone else wants to beat me to it :)

-Olof

[Alexei Humeniy]

Well, I'm trying to compile for a first gen Pentium M which doesn't
support PAE. When I built one of the earliest sources earlier this
year it worked on that machine.

By the way, the machine I am trying to run it on is a ThinkPad T41
(1.5G RAM, ATI Mobility Radeon 9000 -I don't expect Compiz-fusion to
work on it-, Bluetooth, ThinkPad ABG Wireless Adapeter -works with
ath5k)

~Alexei

On Dec 9, 2:12 pm, Grant Grundler <grund...@chromium.org> wrote:

[Alexei Humeniy]

It now appears to work (It don't ask for PAE anymore) but now it get
stuck on something. I get no display, but system responds to caps
lock. I can even restart it with ctrl+alt+del. In the build machine
(Athlon 64x2, 2GB RAM) it freezes after keyboard lights blinking once.
It only boot on a Centrino Duo machine, but without wlan so I get
stuck on the language/network selection.

~Alexei

On Dec 9, 2:12 pm, Grant Grundler <grund...@chromium.org> wrote:

[AUPhil]

I can confirm that the Dell Latitude D400 will not boot if PAE is
required. This is also a first-gen Pentium M CPU laptop.

Phil

[Alexei Humeniy]

Any news about that re-split?

~Alexei

On Dec 9, 2:40 pm, Olof Johansson <ol...@chromium.org> wrote:
> On Thu, Dec 9, 2010 at 9:12 AM, Grant Grundler <grund...@chromium.org>wrote:

[Mandeep Singh Baines]

Alexei Humeniy (ahum...@gmail.com) wrote:
> Any news about that re-split?
>

I'm not sure we should support non-PAE processors. Without PAE we don't get
NX. We are hoping to rely more and more on NX for security so PAE would
become a minimum requirement. Other than the first generation Pentium M
(Banias) and some second generation Pentium M (Dothan), I don't believe
Intel has released any x86 CPUs without PAE after the 2000.

But nothing should break if we don't have PAE. You'll just have a less
secure system which I wouldn't recommend but it'll work.

Might be reasonable to add something to the ebuild to disable PAE via
a USE flag. I think such a patch might be acceptable.

> --
> Chromium OS Developers mailing list: chromiu...@chromium.org
> View archives, change email options, or unsubscribe:
> http://groups.google.com/a/chromium.org/group/chromium-os-dev?hl=en

[Alexei Humeniy]

But I though cloud computing was all about no third-party applications
running against the machine CPU -excep the browser- so my computer in
theory shouldn't even need to be "secure". Also, it's Linux we're
talking about: There are no viruses for Linux =)

Well, it's case closed then.

~Alexei

On Dec 15, 9:06 pm, Mandeep Singh Baines <m...@chromium.org> wrote:

> Alexei Humeniy (ahume...@gmail.com) wrote:
> > Any news about that re-split?
>
> I'm not sure we should support non-PAE processors. Without PAE we don't get
> NX. We are hoping to rely more and more on NX for security so PAE would
> become a minimum requirement. Other than the first generation Pentium M
> (Banias) and some second generation Pentium M (Dothan), I don't believe
> Intel has released any x86 CPUs without PAE after the 2000.
>
> But nothing should break if we don't have PAE. You'll just have a less
> secure system which I wouldn't recommend but it'll work.
>
> Might be reasonable to add something to the ebuild to disable PAE via
> a USE flag. I think such a patch might be acceptable.> ~Alexei
>

> > On Dec 9, 2:40�pm, Olof Johansson <ol...@chromium.org> wrote:
> > > On Thu, Dec 9, 2010 at 9:12 AM, Grant Grundler <grund...@chromium.org>wrote:
>
> > > > On Thu, Dec 9, 2010 at 4:29 AM, Alexei Humeniy <ahume...@gmail.com> wrote:
> > > > > Is there a way to build Chromium OS for processors without PAE? How?
>
> > > > While I'm no expert on PAE, I believe PAE support is controlled by
> > > > this kernel config option:

> > > > � �chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

>
> > > > Modify that to "#CONFIG_X86_PAE is not set" and run the normal tools
> > > > inside a chroot to build an x86-generic image/package/etc. See below
> > > > for some suggestions.
>
> > > Part of the issue here is that we have not yet re-split the configs between
> > > chromeos and chromiumos. At the end of the chromeos release push we trimmed
> > > down the default config to be more in line with what we have on our hardware
> > > (Cr-48), with the intention of keeing the chromiumos config for x86-generic
> > > more broad so it would boot on more hardware.
>
> > > Due to some build system issues we had to keep them common, and the re-split
> > > hasn't happened yet. I should get to it this week unless someone else wants
> > > to beat me to it :)
>
> > > -Olof
>
> > --

> > Chromium OS Developers mailing list: chromium-os-...@chromium.org

[Chris Masone]

On Thu, Dec 16, 2010 at 8:49 AM, Alexei Humeniy <ahum...@gmail.com> wrote:

But I though cloud computing was all about no third-party applications
running against the machine CPU -excep the browser- so my computer in
theory shouldn't even need to be "secure".

There are still network-facing daemons, and browsers and plugins still have bugs sometimes.

 

Chromium OS Developers mailing list: chromiu...@chromium.org

[gameshints]

I'm trying to get Chromium OS to work on my old Thinkpad X40, because
I figured what a perfect lightweight OS to put on a rather old and
slow computer.

I pulled the source, built, put the disk image on a USB drive. While
it worked in newer computers, I got the following error trying to boot
on my thinkpad:

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

So I came across this forum, made the modifications Grant Grundler
suggested, rebuilt, and now when I try to boot I get this:

init[1]: segfault at 1743e55c ip 009a6627 sp 7fb86f88 error 6 in
libc-2.10.1.so[938000+142000]
Kernel panic - not syncing: Attempted to kill init!

Any suggestions on how to get Chromium OS to work on my old computer?

Here's my CPU info:
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 9
model name : Intel(R) Pentium(R) M processor 1200MHz
stepping : 5
cpu MHz : 600.000
cache size : 1024 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 apic mtrr pge mca
cmov clflush dts acpi mmx fxsr sse sse2 tm pbe up bts est tm2
bogomips : 1195.91
clflush size : 64
cache_alignment : 64
address sizes : 32 bits physical, 32 bits virtual
power management:

Thanks!
Erik W

On Dec 9, 11:12 am, Grant Grundler <grund...@chromium.org> wrote:

[Alexei Humeniy]

I got the same kernel panic message (I have a T41 if you haven't
noticed yet) Well, my friend, I guess we're alone for now because they
said PAE will be mandatory. It's very sad because it would be a nice
retirement for these loyal machines. If you, like me, are looking for
a light and fast-boot OS look for xPud. It works pretty good on these
machines.

~Alexei

[gameshints]

I'd prefer Chrome OS over Firefox OS if somebody could help figure out
what's wrong ;)

Erik

[Mandeep Singh Baines]

gameshints (erw...@gmail.com) wrote:
> I'm trying to get Chromium OS to work on my old Thinkpad X40, because
> I figured what a perfect lightweight OS to put on a rather old and
> slow computer.
>
> I pulled the source, built, put the disk image on a USB drive. While
> it worked in newer computers, I got the following error trying to boot
> on my thinkpad:
>
> This kernel requires the following features not present on the CPU:
> pae
> Unable to boot - please use a kernel appropriate for your CPU.
>
> So I came across this forum, made the modifications Grant Grundler
> suggested, rebuilt, and now when I try to boot I get this:
>
> init[1]: segfault at 1743e55c ip 009a6627 sp 7fb86f88 error 6 in
> libc-2.10.1.so[938000+142000]
> Kernel panic - not syncing: Attempted to kill init!
>
> Any suggestions on how to get Chromium OS to work on my old computer?
>

Try removing the march and mcpu lines from:

/usr/local/portage/chromiumos//chromeos/config/make.conf.x86-target

[gameshints]

@Mandeep: There wasn't a MARCH or MCPU line in that file, but there
was a MARCH_TUNE line and I blanked out the string and rebuilt.
Unfortunately, I'm still getting the same Kernel Panic as I was
getting before.

Erik

On Dec 17, 8:16 pm, Mandeep Singh Baines <m...@chromium.org> wrote:

> > On Dec 9, 11:12�am, Grant Grundler <grund...@chromium.org> wrote:
> > > On Thu, Dec 9, 2010 at 4:29 AM, Alexei Humeniy <ahume...@gmail.com> wrote:
> > > > Is there a way to build Chromium OS for processors without PAE? How?
>
> > > While I'm no expert on PAE, I believe PAE support is controlled by
> > > this kernel config option:

> > > � � chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

>
> > > Modify that to "#CONFIG_X86_PAE is not set" and run the normal tools
> > > inside a chroot to build an x86-generic image/package/etc. See below
> > > for some suggestions.
>
> > > This really begs the question of why though....Mandeep Baines
> > > explained that PAE has been available since Pentium Pro. Hugh Dickens
> > > pointed out that disabling PAE also disables "NX" (no execute) support
> > > - ie code pushed on the stack could be executed. This is a security
> > > feature. Just FYI - not trying to stop you from making use of
> > > older/less capable HW.
>
> > > Some detailed steps to do what I'm suggesting (NOT TESTED):

> > > � � cd /usr/local/src/chromiumos/x86-generic/src/scripts/
> > > � � ./enter_chroot.sh
> > > � � ./cros_workon start chromeos-base/kernel � �# create local copy of
> > > kernel source
> > > � � cd ../third_party/kernel/files
> > > � � fgrep -R PAE chromeos/config/
> > > � � chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y
> > > � � vi chromeos/config/i386/config.common.i386 � �# modify X86_PAE config option
>
> > > � �# rebuild everything - there are ways to rebuild just the

> > > kernel...msb promised he'd add a FAQ entry to do that :)

> > > � � cd ../../../scripts
> > > � � �B="x86-generic"
> > > � � ./build_packages --board=$B --oldchromebinary && ./build_image

> > > --board=$B --noenable_rootfs_verification && ./mod_image_for_test.sh
> > > --board=$B -y && ./image_to_usb.sh --board=$B --to=/dev/sdc -y
>
> > > hth,
> > > grant
>
> > --

> > Chromium OS Developers mailing list: chromium-os-...@chromium.org

[Mandeep Singh Baines]

On Sat, Dec 18, 2010 at 1:15 PM, gameshints <erw...@gmail.com> wrote:

@Mandeep: There wasn't a MARCH or MCPU line in that file, but there
was a MARCH_TUNE line and I blanked out the string and rebuilt.

Hmm, there should have been a mcpu and march:

http://git.chromium.org/gitweb/?p=chromiumos-overlay.git;a=blob;f=chromeos/config/make.conf.x86-target

 

Unfortunately, I'm still getting the same Kernel Panic as I was
getting before.

You'll need to rebuild everything with the new CFLAGS. Easiest way would be to:

1) rm -rf /build/x86-generic

2) ./setup_board --board=x86-generic

3) ./build_package --nousepkg

The key is the --nousepkg. It results in building everything from source.

 

Chromium OS Developers mailing list: chromiu...@chromium.org

[gameshints]

Well, I removed the "-march=atom -mtune=atom" part of MARCH_TUNE then
just re-ran the build_package and build_image. I don't think I want
to re-build everything from source otherwise the changes I made to
remove PAE won't be there. Did I need to re-setup_board first to make
the changes have effect?

Erik

On Dec 21, 6:03 pm, Mandeep Singh Baines <m...@chromium.org> wrote:

> On Sat, Dec 18, 2010 at 1:15 PM, gameshints <erwe...@gmail.com> wrote:
> > @Mandeep: There wasn't a MARCH or MCPU line in that file, but there
> > was a MARCH_TUNE line and I blanked out the string and rebuilt.
>
> Hmm, there should have been a mcpu and march:
>

> http://git.chromium.org/gitweb/?p=chromiumos-overlay.git;a=blob;f=chr...

[Mandeep Singh Baines]

On Tue, Dec 21, 2010 at 4:24 PM, gameshints <erw...@gmail.com> wrote:

Well, I removed the "-march=atom -mtune=atom" part of MARCH_TUNE then
just re-ran the build_package and build_image.  I don't think I want
to re-build everything from source otherwise the changes I made to
remove PAE won't be there.  Did I need to re-setup_board first to make
the changes have effect?

You need to rebuild all target binaries from source with the new CFLAGS to avoid

having binaries with instructions not supported by your machine.

 

Chromium OS Developers mailing list: chromiu...@chromium.org

[gameshints]

Gotcha. Unfortunately after rebuilding everything, I still get the
Kernel Panic error.

Erik

On Dec 21, 6:41 pm, Mandeep Singh Baines <m...@chromium.org> wrote:

[Mandeep Singh Baines]

gameshints (erw...@gmail.com) wrote:
> Gotcha. Unfortunately after rebuilding everything, I still get the
> Kernel Panic error.
>

Can you open a bug and paste a photo of the panic.

> > > > > > > On Dec 9, 11:12�am, Grant Grundler <grund...@chromium.org>

> > > wrote:
> > > > > > > > On Thu, Dec 9, 2010 at 4:29 AM, Alexei Humeniy <
> > > ahume...@gmail.com>
> > > > > wrote:
> > > > > > > > > Is there a way to build Chromium OS for processors without PAE?
> > > > > How?
> >
> > > > > > > > While I'm no expert on PAE, I believe PAE support is controlled
> > > by
> > > > > > > > this kernel config option:

> > > > > > > > � � chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

> >
> > > > > > > > Modify that to "#CONFIG_X86_PAE is not set" and run the normal
> > > tools
> > > > > > > > inside a chroot to build an x86-generic image/package/etc. See
> > > below
> > > > > > > > for some suggestions.
> >
> > > > > > > > This really begs the question of why though....Mandeep Baines
> > > > > > > > explained that PAE has been available since Pentium Pro. Hugh
> > > Dickens
> > > > > > > > pointed out that disabling PAE also disables "NX" (no execute)
> > > > > support
> > > > > > > > - ie code pushed on the stack could be executed. This is a
> > > security
> > > > > > > > feature. Just FYI - not trying to stop you from making use of
> > > > > > > > older/less capable HW.
> >
> > > > > > > > Some detailed steps to do what I'm suggesting (NOT TESTED):

> > > > > > > > � � cd /usr/local/src/chromiumos/x86-generic/src/scripts/
> > > > > > > > � � ./enter_chroot.sh
> > > > > > > > � � ./cros_workon start chromeos-base/kernel � �# create

> > > > > local copy of
> > > > > > > > kernel source

> > > > > > > > � � cd ../third_party/kernel/files
> > > > > > > > � � fgrep -R PAE chromeos/config/
> > > > > > > > � � chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

> > > > > > > > � � vi chromeos/config/i386/config.common.i386 � �#
> > > modify
> > > > > X86_PAE config option
> >
> > > > > > > > � �# rebuild everything - there are ways to rebuild just the

> > > > > > > > kernel...msb promised he'd add a FAQ entry to do that :)

> > > > > > > > � � cd ../../../scripts
> > > > > > > > � � �B="x86-generic"
> > > > > > > > � � ./build_packages --board=$B --oldchromebinary &&

> > > > > ./build_image
> > > > > > > > --board=$B --noenable_rootfs_verification &&
> > > ./mod_image_for_test.sh
> > > > > > > > --board=$B -y && ./image_to_usb.sh --board=$B --to=/dev/sdc -y
> >
> > > > > > > > hth,
> > > > > > > > grant
> >
> > > > > > > --
> > > > > > > Chromium OS Developers mailing list: chromium-os-...@chromium.org
> > > > > > > View archives, change email options, or unsubscribe:
> > > > > > >http://groups.google.com/a/chromium.org/group/chromium-os-dev?hl=en
> >
> > > > > --
> > > > > Chromium OS Developers mailing list: chromium-os-...@chromium.org
> > > > > View archives, change email options, or unsubscribe:
> > > > >http://groups.google.com/a/chromium.org/group/chromium-os-dev?hl=en
> >
> > > --
> > > Chromium OS Developers mailing list: chromium-os-...@chromium.org
> > > View archives, change email options, or unsubscribe:
> > >http://groups.google.com/a/chromium.org/group/chromium-os-dev?hl=en
>
> --

> Chromium OS Developers mailing list: chromiu...@chromium.org

[mmoroca]

I have runned the build_packages over an all new x86-generic board
with the kernel-next option enabled and this is what I got:

Packages failed: dev-lang/v8-5208 ,net-misc/socat-1.7.1.1-r1 ,media-
plugins/o3d-68078 ,net-misc/modemmanager-0.3.997_p2-r14 ,chromeos-base/
firmware-utils-0.0.1-r2 ,chromeos-base/bootstat-0.0.1-r6 ,chromeos-
base/verity-0.0.1-r10 ,sys-apps/flashrom-0.9.3-r1 ,chromeos-base/
libchrome-57098-r3 ,app-crypt/tpm-emulator-0.0.1-r1 ,chromeos-base/
cros_boot_mode-0.0.1-r2

My oldie goldie T41 runs Ubuntu flawlessly and I thought ChromiumOS
would be a good replacement OS... Maybe I'll try again later.

On 23 dic, 02:45, Mandeep Singh Baines <m...@chromium.org> wrote:
> gameshints (erwe...@gmail.com) wrote:
> > Gotcha.  Unfortunately after rebuilding everything, I still get the
> > Kernel Panic error.
>
> Can you open a bug and paste a photo of the panic.
>
>
>
>
>
>
>
> > Erik
>

> > On Dec 21, 6:41�pm, Mandeep Singh Baines <m...@chromium.org> wrote:
> > > On Tue, Dec 21, 2010 at 4:24 PM, gameshints <erwe...@gmail.com> wrote:
> > > > Well, I removed the "-march=atom -mtune=atom" part of MARCH_TUNE then

> > > > just re-ran the build_package and build_image. �I don't think I want

> > > > to re-build everything from source otherwise the changes I made to

> > > > remove PAE won't be there. �Did I need to re-setup_board first to make

> > > > �While

> > > > > > > > it worked in newer computers, I got the following error trying to
> > > > boot
> > > > > > > > on my thinkpad:
>
> > > > > > > > This kernel requires the following features not present on the CPU:
> > > > > > > > pae
> > > > > > > > Unable to boot - please use a kernel appropriate for your CPU.
>
> > > > > > > > So I came across this forum, made the modifications Grant Grundler
> > > > > > > > suggested, rebuilt, and now when I try to boot I get this:
>
> > > > > > > > init[1]: segfault at 1743e55c ip 009a6627 sp 7fb86f88 error 6 in
> > > > > > > > libc-2.10.1.so[938000+142000]
> > > > > > > > Kernel panic - not syncing: Attempted to kill init!
>
> > > > > > > > Any suggestions on how to get Chromium OS to work on my old
> > > > computer?
>
> > > > > > > Try removing the march and mcpu lines from:
>
> > > > > > > /usr/local/portage/chromiumos//chromeos/config/make.conf.x86-target
>
> > > > > > > > Here's my CPU info:

> > > > > > > > processor � � � : 0
> > > > > > > > vendor_id � � � : GenuineIntel
> > > > > > > > cpu family � � �: 6
> > > > > > > > model � � � � � : 9
> > > > > > > > model name � � �: Intel(R) Pentium(R) M processor 1200MHz
> > > > > > > > stepping � � � �: 5
> > > > > > > > cpu MHz � � � � : 600.000
> > > > > > > > cache size � � �: 1024 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 apic mtrr pge mca

> > > > > > > > cmov clflush dts acpi mmx fxsr sse sse2 tm pbe up bts est tm2

> > > > > > > > bogomips � � � �: 1195.91
> > > > > > > > clflush size � �: 64
> > > > > > > > cache_alignment : 64
> > > > > > > > address sizes � : 32 bits physical, 32 bits virtual

> > > > > > > > power management:
>
> > > > > > > > Thanks!
> > > > > > > > Erik W
>

> > > > > > > > On Dec 9, 11:12�am, Grant Grundler <grund...@chromium.org>

> > > > wrote:
> > > > > > > > > On Thu, Dec 9, 2010 at 4:29 AM, Alexei Humeniy <
> > > > ahume...@gmail.com>
> > > > > > wrote:
> > > > > > > > > > Is there a way to build Chromium OS for processors without PAE?
> > > > > > How?
>
> > > > > > > > > While I'm no expert on PAE, I believe PAE support is controlled
> > > > by
> > > > > > > > > this kernel config option:

> > > > > > > > > � � chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

>
> > > > > > > > > Modify that to "#CONFIG_X86_PAE is not set" and run the normal
> > > > tools
> > > > > > > > > inside a chroot to build an x86-generic image/package/etc. See
> > > > below
> > > > > > > > > for some suggestions.
>
> > > > > > > > > This really begs the question of why though....Mandeep Baines
> > > > > > > > > explained that PAE has been available since Pentium Pro. Hugh
> > > > Dickens
> > > > > > > > > pointed out that disabling PAE also disables "NX" (no execute)
> > > > > > support
> > > > > > > > > - ie code pushed on the stack could be executed. This is a
> > > > security
> > > > > > > > > feature. Just FYI - not trying to stop you from making use of
> > > > > > > > > older/less capable HW.
>
> > > > > > > > > Some detailed steps to do what I'm suggesting (NOT TESTED):

> > > > > > > > > � � cd /usr/local/src/chromiumos/x86-generic/src/scripts/
> > > > > > > > > � � ./enter_chroot.sh

> > > > > > > > > � � ./cros_workon start chromeos-base/kernel � �# create

> > > > > > local copy of
> > > > > > > > > kernel source

> > > > > > > > > � � cd ../third_party/kernel/files
> > > > > > > > > � � fgrep -R PAE chromeos/config/
> > > > > > > > > � � chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

> > > > > > > > > � � vi chromeos/config/i386/config.common.i386 � �#
> > > > modify
> > > > > > X86_PAE config option
>
> > > > > > > > > � �# rebuild everything - there are ways to rebuild just the

> > > > > > > > > kernel...msb promised he'd add a FAQ entry to do that :)

> > > > > > > > > � � cd ../../../scripts
> > > > > > > > > � � �B="x86-generic"

> > > > > > > > > � � ./build_packages --board=$B --oldchromebinary &&

[Daniel Leinweber]

Has there been any movement on this solution?  I also have a X40 and would love to get Chromium OS onto it but get the same message.

Daniel

[Mike Frysinger]

you can probably update the kernel config to disable CONFIG_X86_PAE and then build x86-generic yourself.  none of the stock builds will be built this way though.

-mike

--
--
Chromium OS Developers mailing list: chromiu...@chromium.org

[Frans Ojala]

Hi,

I've also been trying to build a ChormiumOS for a non-PAE system without success.

I've tried commenting out the CONFIG_X86_PAE option in the 

/src/third_party/kernel/files/chromeos/config/i386/common.config 

and also setting the option to =n.

Both yielded the same result: my machine (with a Pentium M Dothan, 1st gen) informs me to change to a kernel that does not require PAE.

I've now commented out the pae option and removed "-march=atom -mtune=atom" from MARCH_TUNE (described earlier). I did a new setup and am building from source (--nousepkg). Lets see...

Virta

[Frans Ojala]

Well, this was unexpected. I've been unable to compile the code, could someone help me out a bit?

I ran the build_packages --nousepkg a couple of times, on two of 4 times I had to stop due to boot so I restarted (always from setup_board --force) and two times it failed due to some errors.

I didn't go through the log files of the first fail too much, but what I did pick up was the same git error as on the latest that ran last night. At least the following:

DEBUG: RunGit('/mnt/host/source/chromite/lib', ['var', 'GIT_COMMITTER_IDENT'], **{'print_cmd': False})

deploy_chrome: Unhandled exception:

       //some other files on the traceback

File "/mnt/host/source/chromite/lib/cros_build_lib.py", line 458, in RunCommand

    raise RunCommandError(msg, cmd_result)

chromite.lib.cros_build_lib.RunCommandError: return code: 128

*** Please tell me who you are.

I have set these ofc, so I don't understand why it's whining.. And there were some dangling symlinks, is that normal? Logs attached from the secong fail, which I ran to completion. 

The build slowed down drastically at this point which was where I stopped the first build, stdout:

Pending 50/538, Building 1/1, Retrying 1, [Time 330m15.6s Load 3.76 5.19 10.55]

Still building chromeos-chrome-28.0.1498.1_rc-r1 (239m42.1s). Logs in /tmp/chromeos-chrome-28.0.1498.1_rc-r1-tCMeA0

Any help appreciated, thanks in advance.

Virta

[Mike Frysinger]

try running it manually inside the chroot:

cd /mnt/host/source/chromite/

git var GIT_COMMITTER_IDENT

it should display something like:

Mike Frysinger <vap...@chromium.org> 1367867338 -0400

-mike

[Frans Ojala]

Yep, check out fine. 

I read from this post:

https://groups.google.com/a/chromium.org/forum/?fromgroups=#!topic/chromium-os-dev/b9UNvx-fIZQ

that the daisy and amd86-generics were broken a while ago, I gather it shouldn't have effect on the x86-generic?

Thanks,

Virta

[Mike Frysinger]

that is a diff issue from what you are describing
-mike

--

[Dan Chumsky]

Pentium M processors actually do have PAE support, they just do not have the PAE Flag. 

Is there a way of using something like:  https://help.ubuntu.com/community/Lubuntu-fake-PAE

to trick the install into recognizing  PAE for Pentium M systems.

 

 

On Wednesday, December 15, 2010 7:06:56 PM UTC-5, Mandeep Singh Baines wrote:

Alexei Humeniy (ahum...@gmail.com) wrote:
> Any news about that re-split?
>

I'm not sure we should support non-PAE processors. Without PAE we don't get
NX. We are hoping to rely more and more on NX for security so PAE would
become a minimum requirement. Other than the first generation Pentium M
(Banias) and some second generation Pentium M (Dothan), I don't believe
Intel has released any x86 CPUs without PAE after the 2000.

But nothing should break if we don't have PAE. You'll just have a less
secure system which I wouldn't recommend but it'll work.

Might be reasonable to add something to the ebuild to disable PAE via
a USE flag. I think such a patch might be acceptable.

> ~Alexei
>
> On Dec 9, 2:40�pm, Olof Johansson <ol...@chromium.org> wrote:
> > On Thu, Dec 9, 2010 at 9:12 AM, Grant Grundler <grund...@chromium.org>wrote:
> >
> > > On Thu, Dec 9, 2010 at 4:29 AM, Alexei Humeniy <ahume...@gmail.com> wrote:
> > > > Is there a way to build Chromium OS for processors without PAE? How?
> >
> > > While I'm no expert on PAE, I believe PAE support is controlled by
> > > this kernel config option:

> > > � �chromeos/config/i386/config.common.i386:CONFIG_X86_PAE=y

[Michael K.]

hi i found the following site:
http://chromiumosde.gozaru.jp/

Translated with google translate it says in the introduction:
This custom Chromium OS has the following features.

1. Operation with Pentium M / Celeron M CPU

I also make it work on PC with Pentium M or Celeron M installed in CPU. There is also a track record of running on a CPU that is not compatible with PAE. For now it is running on each of the following Pentium M / Celeron M models.

• ASUS Eee PC 900 (PAE compatible)
• IBM ThinkPad T42 (PAE not supported)
• Panasonic Let's Note CF-W 2 (PAE non-compliant)

However, the performance is as it is. Particularly, since CF - W2 has less memory, it can not be said that response is good. Also, originally Chromium OS does not support enough support for older devices (rather it is truncating more and more). For models not listed here, there is a high possibility that they will not move. I do not expect much, I think if you can try it with a light feeling like money if you move.

..
..
..

maybe someone wants to try it...
at the top is the actual version with the download link to github....

good luck & marry christmas.