You are not logged in.

#1 2018-05-17 10:35:01

eugen-b
Member
Registered: 2017-07-09
Posts: 40

linux-pae

Hello!

I've built linux-pae from AUR https://aur.archlinux.org/cgit/aur.git/ … =linux-pae (adjusting the version to 4.16.8) on my manjaro32 and it runs fine on my machine

~ >>> free
              total        used        free      shared  buff/cache   available
Mem:           7305         343        6235          63         726        6351
Swap:             0           0           0
~ >>> inxi -Fxxxz
System:    Host: manjaro Kernel: 4.16.8-1-pae i686 bits: 32 compiler: gcc v: 8.1.0 Desktop: N/A wm: bspwm
           dm: lightdm 1.26.0,sddm Distro: Manjaro Linux 18.0-alpha-1 Illyria
Machine:   Type: Desktop System: ASUS product: All Series v: N/A serial: N/A
           Mobo: ASUSTeK model: AM1I-A v: Rev X.0x serial: N/A UEFI: American Megatrends v: 0801
           date: 03/10/2016
CPU:       Topology: Quad Core model: AMD Athlon 5370 APU with Radeon R3 bits: 64 type: MCP arch: Jaguar rev: 1
           L2 cache: 2048 KiB
           flags: lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3 svm bogomips: 17601
           Speed: 800 MHz min/max: 800/2200 MHz Core speeds (MHz): 1: 976 2: 1418 3: 1106 4: 1107
Graphics:  Card-1: Advanced Micro Devices [AMD/ATI] Kabini [Radeon HD 8400 / R3 Series] driver: radeon
           v: kernel bus ID: 00:01.0 chip ID: 1002:9830
           Display: x11 server: X.Org 1.19.6 driver: ati,radeon unloaded: fbdev,modesetting,vesa
           resolution: 1920x1080~60Hz
           OpenGL: renderer: AMD KABINI (DRM 2.50.0 / 4.16.8-1-pae LLVM 6.0.0) v: 4.5 Mesa 18.0.3 compat-v: 3.0
           direct render: Yes
(...)
Info:      Processes: 170 Uptime: 42m Memory: 7.13 GiB used: 375.0 MiB (5.1%) Init: systemd v: 238 Compilers:
           gcc: 8.1.0 Shell: zsh v: 5.5.1 running in: server inxi: 3.0.08

The machine now also runs archlinux32.
My only available pure 32bit machine - a VIA Eden thin client https://forum.manjaro.org/t/openchrome- … vx800/9586 - is too much hassle to set up for graphical usage. And my Intel Atom N270 netbook died. That's why I run 32bit Linux system on a 64bit machine as I want to help with testing.

What do you think about offering a precompiled PAE kernel in archlinux32?

It is a matter of adding CONFIG_X86_PAE=y to the kernel config.
A PAE kernel won't boot on some older machines therefore it should not be made default.
But if you are willing to offer an additional kernel it would be great for some users who would be able to utilize the additional RAM.

Thank you for all the work you already do for us!
Eugen

Offline

#2 2018-05-17 15:11:21

andreas_baumann
Administrator
From: Zurich, Switzerland
Registered: 2017-08-10
Posts: 601
Website

Re: linux-pae

Building a 'linux-pae' package by hand and putting it in a private repo is not
a problem. I also agree that PAE should not be the default for the 'linux' package.

We don't have an official place where to put a 'linux-pae' package. As signing is required
one of the Archlinux32 devs must do it (which is not a problem). I would like to
do it automatically, so everytime the 'linux' package is built, there is also the
'linux-pae' package built (but there is no provision for that in the buildmaster, sort
of building a dependend package).

Let me think a little bit more about it. :-)

Offline

#3 2018-05-17 21:04:28

eugen-b
Member
Registered: 2017-07-09
Posts: 40

Re: linux-pae

I'm happy that you show some interest!
FYI, there is a dicussion going on Manjaro Forum.

Offline

#4 2018-05-18 10:24:34

deep42thought
Administrator
From: Jena, Germany
Registered: 2017-06-17
Posts: 378

Re: linux-pae

We can, at some point, introduce an [aur] repository like archlinux arm has. Several thoughts about this:

  • Compilation on i686 is considerably faster than on arm - do we really need this repository?

  • This has great security implications, because aur sources are untrusted: malicious PKGBUILDs may harm the buildmaster and buildslaves, malicious packages may harm build slaves and users (they would be signed by our official package keys which should be trusted by any archlinux32 user.

  • In any case, I'd like to wait with the implementation of [aur] until eschwartz changed the source git layout of archlinux - the current layout would make [aur] fit in nicely via git checkouts from aur.archlinux.org, but I think it's more important, that the upcoming source layout fits nicely.

Regards,
deep42thought

Offline

#5 2018-05-19 21:35:30

eugen-b
Member
Registered: 2017-07-09
Posts: 40

Re: linux-pae

deep42thought wrote:

We can, at some point, introduce an [aur] repository like archlinux arm has. Several thoughts about this:

I don't know for sure, but a PAE kernel doesn't need to be the linux-pae from AUR. You just have to add  CONFIG_X86_PAE=y to the kernel config of the existing kernels. And a CONFIG_LOCALVERSION="-pae" to add pae to the package name.

Offline

#6 2018-05-20 19:13:30

deep42thought
Administrator
From: Jena, Germany
Registered: 2017-06-17
Posts: 378

Re: linux-pae

The cleanest way (= least maintenance burden) besides compiling the aur package would be to add another split target to the linux package, which uses the same config and patches "CONFIG_X86_PAE=y" into that. On the other hand, I think, we should try to keep the patch set minimal (e.g. patches that are needed to run software under i686 which is available upstream) and rather compile some aur package(s).

Offline

Board footer

Powered by FluxBB