[lfs-support] compile 4.17.7 32 bit

[Thomas Seeling]

Hallo,


I am compiling the current linux kernels for my 32 and 64 bit machines.
The latest 4.17.7 does not work for me on 32 bit (intel core2 duo) while
it boots fine on the i5 64 bit.
I used the same config as for 4.17.6 which compiles and boots fine on
the 32 bit CPU.
LFS 8.2, gcc 7.3.

Anyone else having problems with the latest kernel under some circumstances?

Tschau...Thomas
--
"Do you wanna be a legend or a footprint on the sands of time?"
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thin

[Thanos Baloukas]

Post by Thomas Seeling
Hallo,
I am compiling the current linux kernels for my 32 and 64 bit machines.
The latest 4.17.7 does not work for me on 32 bit (intel core2 duo) while
it boots fine on the i5 64 bit.
I used the same config as for 4.17.6 which compiles and boots fine on
the 32 bit CPU.
LFS 8.2, gcc 7.3.
Anyone else having problems with the latest kernel under some circumstances?

I'm not sure if intel core2 duo is i386, but you may find this lkml post
interesting:

https://lkml.org/lkml/2018/7/17/434
--
Thanos
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://en.wikipedia.org/wiki/Post

[Bruce Dubbs]

Post by Thanos Baloukas

Post by Thomas Seeling
Hallo,
I am compiling the current linux kernels for my 32 and 64 bit machines.
The latest 4.17.7 does not work for me on 32 bit (intel core2 duo) while
it boots fine on the i5 64 bit.
I used the same config as for 4.17.6 which compiles and boots fine on
the 32 bit CPU.
LFS 8.2, gcc 7.3.
Anyone else having problems with the latest kernel under some
circumstances?

I'm not sure if intel core2 duo is i386, but you may find this lkml post
https://lkml.org/lkml/2018/7/17/434

I used a Core2Duo for about 6 years. It is a native x86_64 system, but
can run 32-bit kernels and thus 32-bit systems.

-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://e

[Ken Moffat]

Post by Thanos Baloukas

Post by Thomas Seeling
Hallo,
I am compiling the current linux kernels for my 32 and 64 bit machines.
The latest 4.17.7 does not work for me on 32 bit (intel core2 duo) while
it boots fine on the i5 64 bit.
I used the same config as for 4.17.6 which compiles and boots fine on
the 32 bit CPU.
LFS 8.2, gcc 7.3.
Anyone else having problems with the latest kernel under some circumstances?

I'm not sure if intel core2 duo is i386, but you may find this lkml post
https://lkml.org/lkml/2018/7/17/434

I used a Core2Duo for about 6 years. It is a native x86_64 system, but can
run 32-bit kernels and thus 32-bit systems.

No idea what the bug on x86_64 is, I'm behind on reading lkml, but I
think it is likely that 'i386' might mean 'on x86 machines running
as 32-bit'.

And apparently there is a one patch revision for 4.17.8 due on
Thursday, no idea if it might help or not:

https://lkml.org/lkml/2018/7/17/443

and

https://lkml.org/lkml/2018/7/17/505

ĸen
--
Keyboard not found, Press F1 to continue
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://en.w

[Ken Moffat]

Post by Ken Moffat

I used a Core2Duo for about 6 years. It is a native x86_64 system, but can
run 32-bit kernels and thus 32-bit systems.

No idea what the bug on x86_64 is, I'm behind on reading lkml, but I
think it is likely that 'i386' might mean 'on x86 machines running
as 32-bit'.
And apparently there is a one patch revision for 4.17.8 due on
https://lkml.org/lkml/2018/7/17/443
and
https://lkml.org/lkml/2018/7/17/505

It *does* affect all 32-bit x86 :

https://lkml.org/lkml/2018/7/16/809

Post by Ken Moffat
total: 171 pass: 159 fail: 12
i386:Broadwell:q35:defconfig:smp:rootfs
i386:Skylake-Client:q35:defconfig:smp:rootfs
i386:SandyBridge:q35:defconfig:smp:rootfs
i386:Haswell:pc:defconfig:smp:rootfs
i386:Nehalem:q35:defconfig:smp:rootfs
i386:phenom:pc:defconfig:smp:rootfs
i386:Opteron_G5:q35:defconfig:smp:initrd
i386:Westmere:q35:defconfig:smp:initrd
i386:core2duo:q35:defconfig:nosmp:rootfs
i386:Conroe:pc:defconfig:nosmp:rootfs
i386:Opteron_G1:pc:defconfig:nosmp:initrd
i386:n270:q35:defconfig:nosmp:rootfs
All 32-bit i386 boot tests crash.

And the reply was that the known fix is not yet in Linus's tree, so
it can't be applied to stable until Linus commits it.

So the one patch in 4.17.8-rc won't help 32-bit.

ĸen
--
Keyboard not found, Press F1 to continue
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://en.wikipedia.org/wi

[Douglas R. Reno]

Post by Ken Moffat

Post by Bruce Dubbs
I used a Core2Duo for about 6 years. It is a native x86_64 system,

but can

Post by Ken Moffat

Post by Bruce Dubbs
run 32-bit kernels and thus 32-bit systems.

No idea what the bug on x86_64 is, I'm behind on reading lkml, but I
think it is likely that 'i386' might mean 'on x86 machines running
as 32-bit'.
And apparently there is a one patch revision for 4.17.8 due on
https://lkml.org/lkml/2018/7/17/443
and
https://lkml.org/lkml/2018/7/17/505

https://lkml.org/lkml/2018/7/16/809

Post by Ken Moffat
total: 171 pass: 159 fail: 12
i386:Broadwell:q35:defconfig:smp:rootfs
i386:Skylake-Client:q35:defconfig:smp:rootfs
i386:SandyBridge:q35:defconfig:smp:rootfs
i386:Haswell:pc:defconfig:smp:rootfs
i386:Nehalem:q35:defconfig:smp:rootfs
i386:phenom:pc:defconfig:smp:rootfs
i386:Opteron_G5:q35:defconfig:smp:initrd
i386:Westmere:q35:defconfig:smp:initrd
i386:core2duo:q35:defconfig:nosmp:rootfs
i386:Conroe:pc:defconfig:nosmp:rootfs
i386:Opteron_G1:pc:defconfig:nosmp:initrd
i386:n270:q35:defconfig:nosmp:rootfs
All 32-bit i386 boot tests crash.

And the reply was that the known fix is not yet in Linus's tree, so
it can't be applied to stable until Linus commits it.
So the one patch in 4.17.8-rc won't help 32-bit.
Äžen
--
Keyboard not found, Press F1 to continue

Guys,

Heads up, this is affecting sporadic x86_64 PCs too. A sizable chunk of my
64bit machines refuse to boot with 4.17.7 and even a few 32bit ones with
4.17.6. There is obviously something wrong here that needs fixing.

[Bruce Dubbs]

Post by Douglas R. Reno
Guys,
Heads up, this is affecting sporadic x86_64 PCs too. A sizable chunk of
my 64bit machines refuse to boot with 4.17.7 and even a few 32bit ones
with 4.17.6. There is obviously something wrong here that needs fixing.

Just to add a few things. I just did a complete x86_64 rebuild with the
most recent kernel (4.17.7), binutils, util-linux, etc.

I did get some new test failures:

075-glibc-2.27:FAIL: nss/tst-nss-files-hosts-multi

105-inetutils-1.9.4:FAIL: ping-localhost.sh

106-perl-5.28.0:FAILED at test 7

138-eudev-3.2.5:FAIL: udev-test.pl

143-vim-8.1:1 FAILED:
143-vim-8.1:TEST FAILURE

I have not yet investigated those.

I also did some experimenting with a separate /usr partition. The
system booted properly. Here is some sizing data:

Filesystem Size Used Avail Use% Mounted on
/dev/root 20G 3.4G 16G 18% /
/dev/sda24 9.8G 806M 8.5G 9% /usr

5.2M bin
15M etc
75M jhalfs
42M lib
4.0K lib64
12K media
4.0K mnt
4.0K opt
24K root
12M sbin
1.7G sources
4.0K srv
1.6G tools
260K var

The bottom line is that after removing jhalfs, tools, and sources, the
size needed for the root partition is only about 150 MB. For
installations that use encrypted partitions, RAID, etc, the root
partition would be to be a bit higher. I'd probably go with 500 MB, but
as a proof of concept, this works nicely.

-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

[Ken Moffat]

Post by Douglas R. Reno

Post by Ken Moffat
And the reply was that the known fix is not yet in Linus's tree, so
it can't be applied to stable until Linus commits it.
So the one patch in 4.17.8-rc won't help 32-bit.

Wrong - the patch _was_ the fix for 32-bit. 4.17.8 is out now.
(I was unwell yesterday, still catching up on lkml.)

But that doesn't alter Douglas's report below.

Post by Douglas R. Reno
Guys,
Heads up, this is affecting sporadic x86_64 PCs too. A sizable chunk of my
64bit machines refuse to boot with 4.17.7 and even a few 32bit ones with
4.17.6. There is obviously something wrong here that needs fixing.

--
Keyboard not found, Press F1 to continue
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://en.wikipedia.org/wi

[Thomas Seeling]

Hallo,

Wrong - the patch _was_ the fix for 32-bit. 4.17.8 is out now.

good news.

4.17.8 in fact works on my 32 bit machine.
I tried out gzip kernel compression instead of xz but that didn't help.


Tschau...Thomas

[Paul Rogers]

I'm not sure I'm clear on this. Yes, Core 2 Duo's run either 64 or 32 bit code. Linus thinks everybody runs 64-bit code. I've gotten the impression from the KPTI thing early in the year that kernel devs aren't spending much time at all on 32-bit code in the kernel. I've been hoping for KPTI mitigation for 32-bit kernels, but mostly have lost hope.

So to be clear, are we talking here about running 4.17 in 32-bit mode, or just on a Core-2 Duo, which LFS would build in 64-bit mode "unless steps were taken" (gmp, et al)?
--
Paul Rogers
***@fastmail.fm
Rogers' Second Law: "Everything you do communicates."
(I do not personally endorse any additions after this line. TANSTAAFL :-)
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://en.wikipedia.org/wik

[Ken Moffat]

Post by Paul Rogers
I'm not sure I'm clear on this. Yes, Core 2 Duo's run either 64 or 32 bit code. Linus thinks everybody runs 64-bit code. I've gotten the impression from the KPTI thing early in the year that kernel devs aren't spending much time at all on 32-bit code in the kernel. I've been hoping for KPTI mitigation for 32-bit kernels, but mostly have lost hope.

Queued in the tip tree this week, should be in 4.19 and then will
presumably get backported to the still-maintained kernels.

But it is expected to have a big performance impact on the old
machines without a later feature (PCID : Process Context Identifier,
according to wikipedia that was introduced with Westmere, the 32nm
die-shrink of Nehalem).

Post by Paul Rogers
So to be clear, are we talking here about running 4.17 in 32-bit mode, or just on a Core-2 Duo, which LFS would build in 64-bit mode "unless steps were taken" (gmp, et al)?

The former - I pasted from a mail showing it was everything 32-bit
in x86 (it used to be called arch/i386, maybe qemu still calls it
that, or maybe the description of the arch has never been updated in
that man's build test infrastructure.

So, inter alia, I think broadwell and opteron were mentioned.

Anyway, fixed in 4.17.8.

ĸen
--
Entropy not found, thump keyboard to continue
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http

[Paul Rogers]

Post by Ken Moffat
Queued in the tip tree this week, should be in 4.19 and then will
presumably get backported to the still-maintained kernels.
But it is expected to have a big performance impact on the old
machines without a later feature (PCID : Process Context Identifier,
according to wikipedia that was introduced with Westmere, the 32nm
die-shrink of Nehalem).

Bummer! My newest is one Bloomfield and one Lynnfield. My "daily driver" is a Conroe Core 2 Duo. But it can't be helped I suppose. Fortunately, I'm not a "gamer".

The trouble with getting a newer i3 is the infrastructure incompatibility. There's no failure short of a direct lightning strike that would take me down for more than a few minutes.

Thanks, Ken. I'll start prowling kernel versions again.
--
Paul Rogers
***@fastmail.fm
Rogers' Second Law: "Everything you do communicates."
(I do not personally endorse any additions after this line. TANSTAAFL :-)
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://en.