Porteus Kernel Builder

Here is a place for your projects which are not officially supported by the Porteus Team. For example: your own kernel patched with extra features; desktops not included in the standard ISO like Gnome; base modules that are different than the standard ISO, etc...
User avatar
babam
Samurai
Samurai
Posts: 181
Joined: 16 Nov 2016, 10:30
Distribution: Porteus 3.2.2 Xfce K5.10.5
Location: Rainy city

Porteus Kernel Builder

Post#1096 by babam » 11 Jan 2021, 13:21

@Blaze and @roadie,
What are the benefits of compressed kernel modules?

In Porteus all files including kernel modules are compressed into xzm containers by squashfs-tools (e.g. 000-kernel.xzm).

If kernel modules are compressed then the system will decompress twice when loading a kernel module?
Sorry, my English is bad.

roadie
Full of knowledge
Full of knowledge
Posts: 304
Joined: 02 Jan 2011, 18:41
Distribution: Porteus 5.0-RC1
Location: In a hayfield

Porteus Kernel Builder

Post#1097 by roadie » 11 Jan 2021, 16:33

babam wrote:
11 Jan 2021, 13:21
@Blaze and @roadie,
What are the benefits of compressed kernel modules?

In Porteus all files including kernel modules are compressed into xzm containers by squashfs-tools (e.g. 000-kernel.xzm).

If kernel modules are compressed then the system will decompress twice when loading a kernel module?
Honestly babam, I don't see a lot of benefits with it. I think you're right in that everything is being uncompressed twice, though I don't know how much compression will take place with modules that are already compressed. I'm using an xz compressed version right now, it booted in about the same time as the zstd compressed does, about 12 seconds to the desktop. The DE in use will of course affect that, it's surprising that a light, fast DE can actually take a lot longer to boot than a 'slower', heavier one.

In production use, it quite likely does show a big difference, but for me and a laptop, not so much. Right now, it's a shiny thing.......BTW, kernel 5.10.6 is out and builds nice. It seems to run 'quieter' on my machine, the fan will come on with 5.10.4 and sometimes with 5.10.5, but 5.10.6, it never comes on.

roadie
Full of knowledge
Full of knowledge
Posts: 304
Joined: 02 Jan 2011, 18:41
Distribution: Porteus 5.0-RC1
Location: In a hayfield

Porteus Kernel Builder

Post#1098 by roadie » 11 Jan 2021, 17:44

babam,

Here are times for 5.10.6 uncompressed and zstd compressed, I take the last time before it shows the jump which depends on how fast I run dmesg.


5.10.6-uncompressed with xfce-4.12:

Code: Select all

[    6.338217] logitech-hidpp-device 0003:046D:4013.0005: input,hidraw1: USB HID v1.11 Mouse [Logitech M525] on usb-0000:00:10.0-4/input2:1
[    7.903896] rtw_8822be 0000:01:00.0: start vif dc:f5:05:77:82:51 on port 0
[    7.905238] rtw_8822be 0000:01:00.0: stop vif dc:f5:05:77:82:51 on port 0
[    8.347089] rtw_8822be 0000:01:00.0: start vif be:21:cc:04:c9:f9 on port 0
[    9.789411] random: crng init done
[    9.789415] random: 7 urandom warning(s) missed due to ratelimiting

5.10.6-zstd compressed with xfce-4.12:

Code: Select all

[    6.451255] logitech-hidpp-device 0003:046D:4013.0005: input,hidraw1: USB HID v1.11 Mouse [Logitech M525] on usb-0000:00:10.0-4/input2:1
[    7.824966] rtw_8822be 0000:01:00.0: start vif dc:f5:05:77:82:51 on port 0
[    7.837210] rtw_8822be 0000:01:00.0: stop vif dc:f5:05:77:82:51 on port 0
[    8.280004] rtw_8822be 0000:01:00.0: start vif 52:07:53:9b:1a:e6 on port 0
[    9.730370] random: crng init done
[    9.730374] random: 7 urandom warning(s) missed due to ratelimiting
There's not a lot of difference between them.......I'd like to have a good way to log the times and processes.

roadie
Full of knowledge
Full of knowledge
Posts: 304
Joined: 02 Jan 2011, 18:41
Distribution: Porteus 5.0-RC1
Location: In a hayfield

Porteus Kernel Builder

Post#1099 by roadie » 11 Jan 2021, 19:01

Interesting things when I switch to lxqt-0.16.0.

With 5.10.6 uncompressed:

Code: Select all

[    9.166027] elogind-daemon[1074]: Watching system buttons on /dev/input/event1 (Power Button)
[    9.166140] elogind-daemon[1074]: Watching system buttons on /dev/input/event2 (Sleep Button)
[    9.189924] elogind-daemon[1074]: Watching system buttons on /dev/input/event5 (AT Translated Set 2 keyboard)
[    9.193200] elogind-daemon[1074]: New session c1 of user guest.

With 5.10.6 zstd compressed:

Code: Select all

[   10.933978] elogind-daemon[1081]: Watching system buttons on /dev/input/event1 (Power Button)
[   10.934062] elogind-daemon[1081]: Watching system buttons on /dev/input/event2 (Sleep Button)
[   10.957898] elogind-daemon[1081]: Watching system buttons on /dev/input/event5 (AT Translated Set 2 keyboard)
[   10.960916] elogind-daemon[1081]: New session c1 of user guest.
Notice the big difference, with uncompressed clearly being faster. Both are booted in exactly the same way by editing porteus.conf beforehand so I can just hit Enter,

User avatar
babam
Samurai
Samurai
Posts: 181
Joined: 16 Nov 2016, 10:30
Distribution: Porteus 3.2.2 Xfce K5.10.5
Location: Rainy city

Porteus Kernel Builder

Post#1100 by babam » 12 Jan 2021, 12:11

Thanks roadie, compressed kernel modules don't make the boot process faster but instead make it slow.

What is the size of 000-kernel.xzm (uncompressed kernel modules vs zstd compressed kernel modules)?

Thanks.
Sorry, my English is bad.

roadie
Full of knowledge
Full of knowledge
Posts: 304
Joined: 02 Jan 2011, 18:41
Distribution: Porteus 5.0-RC1
Location: In a hayfield

Porteus Kernel Builder

Post#1101 by roadie » 12 Jan 2021, 17:07

babam,

Here are the sizes. I build for my machine, so small modules. Bigger ones would show the differences better.

000-kernel-5.10.6-aufs-uncomp.xzm = 15920K
000-kernel-5.10.6-aufs-xz.xzm = 15488K
000-kernel-5.10.6-aufs-zstd.xzm = 15824K

maia
Black ninja
Black ninja
Posts: 58
Joined: 08 Aug 2019, 16:56
Distribution: Porteus 5 rc2 < 64b Openbox
Location: Brasil

Porteus Kernel Builder

Post#1102 by maia » 14 Jan 2021, 10:48

J. R. Okajima wrote:
Aufs5.10 branch in Last release was broken and all patch files were
missing. Sorry it was my fault. This release contains real aufs5.10.

J. R. Okajima

- aufs5-linux.git#aufs5.4..aufs5.9
nothing
- aufs5-linux.git#aufs5.10
aufs: version 5.10
- aufs5-linux.git#aufs5.x-rcN
aufs: for linux-v5.11-rc1, dummy def lockdep_is_held()
- aufs5-standalone.git
ditto
- aufs-util.git
nothing
Sorry my English is bad.
My English is very very very bad.
My English is really, g##gle-tr4nsl4t0r

neko
DEV Team
DEV Team
Posts: 1693
Joined: 09 Feb 2013, 09:55
Distribution: APorteus-FVWM-ja-x86_64.iso
Location: japan

Porteus Kernel Builder

Post#1103 by neko » 14 Jan 2021, 12:37

@maia
It has been avoided.
How to avoide this missing.
Please refer to mkKernel-21.01.14-noarch-1.xzm@/usr/local/share/mkKernel/bin/getAUFS.sh
Porteus Kernel Builder (Post by neko #52232)


P.S.
AUFS patch applied to kernel 5.10.7 is fault at patching aufs5-mmap.patch.
Kernel 5.11-rc3 with AUFS patch can be built with updated "Kernel Builder" mkKernel-21.01.14-noarch-1.xzm.


Thanks.

maia
Black ninja
Black ninja
Posts: 58
Joined: 08 Aug 2019, 16:56
Distribution: Porteus 5 rc2 < 64b Openbox
Location: Brasil

Porteus Kernel Builder

Post#1104 by maia » 14 Jan 2021, 12:49

neko wrote:
14 Jan 2021, 12:37
@maia
It has been avoided.
How to avoide this missing.
Please refer to mkKernel-21.01.09-noarch-1.xzm@/usr/local/share/mkKernel/bin/getAUFS.sh
Porteus Kernel Builder (Post by neko #52232)


P.S.
AUFS patch applied to kernel 5.10.7 is fault at patching aufs5-mmap.patch.


Thanks.

Thanks Neko, the first page of the topic went unnoticed.
Sorry my English is bad.
My English is very very very bad.
My English is really, g##gle-tr4nsl4t0r

neko
DEV Team
DEV Team
Posts: 1693
Joined: 09 Feb 2013, 09:55
Distribution: APorteus-FVWM-ja-x86_64.iso
Location: japan

Porteus Kernel Builder

Post#1105 by neko » 17 Jan 2021, 15:29

Kernel 5.10.8 was released.
But patching it with aufs5-mmap.patch of AUFS patch was error.

AUFS VERSION:
aufs5-standalone.git/include/uapi/linux/aufs_type.h:#define AUFS_VERSION "5.x-rcN-20210111"


Thanks.

neko
DEV Team
DEV Team
Posts: 1693
Joined: 09 Feb 2013, 09:55
Distribution: APorteus-FVWM-ja-x86_64.iso
Location: japan

Porteus Kernel Builder

Post#1106 by neko » 18 Jan 2021, 04:36

Kernel 5.11-rc4 has been released.
Patching it with AUFS patch was successful.

AUFS VERSION:
aufs5-standalone.git/include/uapi/linux/aufs_type.h:#define AUFS_VERSION "5.x-rcN-20210111"


Thanks.

peebee
Black ninja
Black ninja
Posts: 38
Joined: 13 Oct 2013, 15:31
Distribution: 4
Location: UK

Porteus Kernel Builder

Post#1107 by peebee » 19 Jan 2021, 07:16

My understanding is that:
aufs-5.10 should be used for kernel 5.10.x
and
aufs-5.x.rcN should be used for kernel 5.11-RCx

If there is a problem with mmap.patch in aufs-5.10 then this should be reported on github - but for me mmap in aufs-5.10 works OK for kernel 5.10.x

neko
DEV Team
DEV Team
Posts: 1693
Joined: 09 Feb 2013, 09:55
Distribution: APorteus-FVWM-ja-x86_64.iso
Location: japan

Porteus Kernel Builder

Post#1108 by neko » 19 Jan 2021, 10:04

Code: Select all

$ xz -dc ../linux-5.10.8.tar.xz | tar -x
$ cd linux-5.10.8
$ cat ../../auf/aufs5-standalone.git/aufs5-mmap.patch | patch -p1
patching file fs/proc/base.c
patching file fs/proc/nommu.c
patching file fs/proc/task_mmu.c
patching file fs/proc/task_nommu.c
patching file include/linux/mm.h
Hunk #1 succeeded at 1712 (offset -3 lines).
patching file include/linux/mm_types.h
Hunk #1 succeeded at 281 (offset 3 lines).
Hunk #2 succeeded at 361 (offset 3 lines).
patching file kernel/fork.c
Hunk #1 succeeded at 554 (offset -1 lines).
patching file mm/Makefile
Hunk #1 FAILED at 52.
1 out of 1 hunk FAILED -- saving rejects to file mm/Makefile.rej
patching file mm/filemap.c
Hunk #1 FAILED at 2989.
1 out of 1 hunk FAILED -- saving rejects to file mm/filemap.c.rej
patching file mm/mmap.c
Hunk #3 FAILED at 1897.
1 out of 8 hunks FAILED -- saving rejects to file mm/mmap.c.rej
patching file mm/nommu.c
patching file mm/prfile.c
$
$ grep VERSION ../../auf/aufs5-standalone.git/include/uapi/linux/aufs_type.h
#define AUFS_VERSION	"5.x-rcN-20210111"
$
Thanks.

peebee
Black ninja
Black ninja
Posts: 38
Joined: 13 Oct 2013, 15:31
Distribution: 4
Location: UK

Porteus Kernel Builder

Post#1109 by peebee » 19 Jan 2021, 15:11

That seems to be kernel 5.10.8 built with aufs-5.x-rcN ??? ..... it should be built with aufs-5.10 .........

I have in /include/uapi/linux/aufs_type.h:
#define AUFS_VERSION "5.10-20210111"

neko
DEV Team
DEV Team
Posts: 1693
Joined: 09 Feb 2013, 09:55
Distribution: APorteus-FVWM-ja-x86_64.iso
Location: japan

Porteus Kernel Builder

Post#1110 by neko » 20 Jan 2021, 03:15

@peebee
Would you tell me how to get aufs-5.10.

Code: Select all

# git clone git://github.com/sfjro/aufs5-standalone.git aufs5-standalone.git
# cd aufs5-standalone.git
# git checkout origin/aufs5.10

Thanks.

Post Reply