PocketHandyBox (Debian 12 "Bookworm" with NVIDIA drivers OOTB)

[gumanzoy]

Debian 12 moved on to GRUB2 v2.12. This GRUB2 release supports boot-use of NTFS compressed partitions.

Any chance the work you are doing would embrace this upgrade for Debian v12s going forward?

I primary use Ventoy.
Ventoy based on GRUB 2.04

There is no problem to build PocketHandyBox .iso with GRUB2 v2.12
But i have no idea about use NTFS compressed partitions in Linux.

I think, better to avoid NTFS at all, use exfat or ext4 for removable drives.

[gumanzoy]

Major changes for initrd and folder structure

GitHub Commit f3d3845

Major changes for initrd and folder structure:

Add special extension for kernel modules .squashfs - .kmodsfs
File name should contain kernel release `uname -r`
Don't need to use extramod=/live/dk5 and noload=k-6.1 for boot old kernel anymore

Add arch i686 detection. If boot with i686 kernel - FOLDER=live32
Move i686 data from /live32/live to /live32
Don't need to use from=/live32 anymore

Add special $CFGDEV/live/firmware folder common for x86_64 and i686
Don't need to use extramod=/live/firmware anymore

Move nvidia driver modules from /live/optional to special folder /live/nvidia
Cheatcode nvidia_detect replaced with nvidia=detect
Select version to load: nvidia=nv390 or nvidia=nv340

/opt/initramfs/linuxrc updated

grub.cfg and extgrub.cfg.example updated

Need to modify Porteus Initrd to combine 32+64bit verions in one .iso file. To make possible to use from=/file.iso and from=/live32 together.

Done. Don't need to use from=/live32 anymore

This will be enabled in next v0006 phb_test .iso
I will do 2 versions: first 64bit only. and second 64+32 combined in one .iso

---

This also revert old behavior to extramod=
With this update extramod= will load at the end and can overwrite files of all previous loaded modules.

It change order of modules load. Before extramod= loads at the end, now its load after base modules.
Now order:
/live/*.squashfs
extramod
/live/modules/*.squashfs
/live/optional/
That makes extramod= modules can not overwrite files of modules from /live/modules/*.squashfs and /live/optional/*.
Before:
/live/*.squashfs
/live/modules/*.squashfs
/live/optional/
extramod

Maybe it is important, maybe not.

[Clarity]

Hello @gumanzoy

... i have no idea about use NTFS compressed partitions in Linux. ...

I have used NTFS partitions for years in Linux without issue; same as any other supported filesystems during system operations. No special commands, no special exemptions, nothing. It just works like the other formats. I have, in the past, created partitions that were compressed without issue by formatting with the compress option in Linux. It allows at least a 40% size reduction for files stored. Thus doubling capacity.

The issue is not a Linux issue. GRUB2 only recently has been upgraded in v2.12 to support use with NTFS compressed partitions.

I offer this post merely as info on "why" I had been looking forward to this support. I'm sure it will become standard in forum distros soon as most of the many Linux distros have moved to it in their latest versions.

[gumanzoy]

@Clarity

I have used NTFS partitions for years in Linux without issue.

Hi

I never use NTFS to store files in my own Linux PC's.
But i frequently use Linux to copy other user's data (for example in case of re-install Windows).
Most time it is work ok.

But if i try to write raw image of hard drive ~ 250GB to NTFS partition under Linux - it fails.
Refuse to continue to write to img file after copying ~ 200Gb maybe.

... i have no idea about use NTFS compressed partitions in Linux. ...

I have, in the past, created partitions that were compressed without issue by formating with the compress options in Linux.. It allows at least a 40% size reduction for files stored.

I am curious, what filesystem you talking about? "compress options in Linux" ?

---

I cannot seem to find the steps to move the audio from the PC's internal speaker to the TV over HDMI. HELP!

Have you try phb_test_x64_12_25_0005_vtgrub2.iso ?

New alsaselect tool works for you or not?

Not need to unmute S/PDIF in alsamixer and edit .asoundrc anymore (i hope, needs testing with different hardware).
Just press button and speaker-test run automatically. Listen have sounds or not, and if not - press another button.

[Clarity]

Hello @gumanzoy

"What filesystem?" ... you ask.

Its the Linux format command has an options: mkfs.ntfs ... -C ... IIRC. Seems to do the same as what is done when running Windows 'diskmgr' to format a partition as NTFS compressed. ... IIRC

mkfs.ntfs -C /dev/sdXn

[gumanzoy]

mkfs.ntfs -C /dev/sdXn

I never listen about somebody use it in Linux.
Have you try it?
What driver you use fuse ntfs-3g or kernel 5.15+ ntfs3 ?

Anyway i think use it in Linux to store data - bad idea.

In Windows 10+ have other feature Compact OS and ntfs-3g can not read files compressed by Compact OS (usually in C:\Windows directory)

[Clarity]

No problem. Dont want anything distasteful in your distro offering.

I just booted PHB v0005 ISO file to desktop via Ventoy. Testing now. It takes me awhile as I have a few admin tools I use to do prelimary reviews that I install via synaptic that are missing. Same as before; namely one of the

• terminal 'fetch' programs (neo/fast),

• tealdeer (tldr replacement),

• btop,

• hardinfo2

• chromium

Also, I notice that SAMBA support is not present that I use to pull info from this test PC from time to time from my primary PC. Thus a bit of a pain to get it setup via package manager properly and is the primary reasons that most of the forum ISOs have it prepackaged. Like many forum users, it a stable and well understood subsystem.

So far system, the system is 'smooth as butter'.

Request: If in the future you do consider adding either fastfetch/neofetch, would you update their default config to show the driver in use on the desktop, please. I'm sure you know the benefit when we sent screen-shots.

Next: I will follow-up on your request for audio tests.

[Clarity]

First, I forgot to mention that it appears that NO boot-menu changes were done on this COPY2RAM boot. And the system seemingly posted boot-time messages that it found a driver for this adapter.

The Sound Card Selector you have works wonderfully!

I will modify or re-write sound-card-selector to make settings applied immediately when card selected.

Question: Will this be called "Sound Card Selector" and found in the taskbar menu named that way, in the future?

Screenshot_2025-04-21_10-38-07.png (25.3 KiB) Viewed 381 times

Next: This may be a item for consideration by yourself, @fredx181 or @rcrsn51 about systems with modern adapters with HDMI ports. For the nvidia adapter should the system setup audio over the HDMI port OOTB, or is manual the best approach in OOTB pristine starts? Reason: for new users of the DOG distros which have nvidia support built-in to have audio working upon desktop. Or, maybe a message should be posted OOTB to help users until an auto-HDMI-audio setup approach is crafted.

Screenshot_2025-04-21_11-09-22.png (177.2 KiB) Viewed 377 times

[Clarity]

@gumanzoy if there are other tests you want to see on this bare-metal use before you go to PHB v006, post and I will help where I can.

[gumanzoy]

Question: Will this be called "Sound Card Selector" and found in the taskbar menu named that way, in the future?

I call it alsaselect. I will not change the name.
"Sound Card Selector" is another tool by @rcrsn51

For the nvidia adapter should the system setup audio over the HDMI port OOTB, or is manual the best approach in OOTB pristine starts?

I will keep analog speakers by default. Primary needs for laptop, because all laptops have built-in analog speakers.

[gumanzoy]

@fredx181

Hi

I have more progress with re-writing initrd.
There is a overview of changes i made.

Use singe binary /bin/busybox
From fresh busybox-static .deb package i build.

Add busybox-static deb package re-build with enabled: blkid eject lspci
busybox-static_1.37.0-4phb_amd64.deb busybox-static_1.37.0-4phb_i386.deb

Not need /cr-initrd/initramfs/lib/ anymore.
But cryptsetup support disabled. Maybe this is disadvantage.

Actual versions of scripts:
/usr/local/sbin/mkinitrd /opt/initramfs/finit /opt/initramfs/linuxrc
And /opt/initramfs/modlist trimmed (remove unused modules)

Major changes for initrd and folder structure: viewtopic.php?p=147260#p147260
grub.cfg and extgrub.cfg.example updated.

GitHub Commit bf642db Unload unused kernel/drivers/mmc (cardreader) modules
Now try to unload unused ata_ and mmc kernel modules before unmount devices. To keep modules for existing drives loaded.

Re-write code to mount -o move or umount devices.
Depending on changes= used or not, and copy2ram / noauto cheatcodes.

Code: Select all

# Create /mnt/home symlink
if [ "$BOOTDEV" ] && [ -d "$BOOTDEV" ]; then
ln -snf $BOOTDEV /union/mnt/home
elif [ -d "$DATADEV" ]; then
ln -snf $DATADEV /union/mnt/home
fi

# Create /mnt/saved symlink
if [ -d /memory/images/changes-exit ]; then
ln -snf /mnt/live/memory/images/changes-exit /union/mnt/saved
else
ln -snf /mnt/live/memory/changes /union/mnt/saved
fi

MNTPOINTS=`grep /mnt/ /etc/fstab | cut -d" " -f2`
[ -d /mnt/nfs ] && MNTPOINTS="/mnt/nfs $MNTPOINTS"
[ -d /mnt/isoloop ] && MNTPOINTS="/mnt/isoloop $MNTPOINTS"
if [ $CHNDEV ] && [ -d $CHNDEV ]; then
MNTPOINTS="$CHNDEV $MNTPOINTS"
fi

# Move mountpoints /mnt/XX to /union/mnt/XX
for MNTDEV in $MNTPOINTS
do mkdir -p /union"$MNTDEV"
if [ "$MNTDEV" = "$CHNDEV" ]; then
	if [ "$CHNEXIT" ] && param copy2ram ; then
		mount -o move "$MNTDEV" /union"$MNTDEV" # && umount /union"$MNTDEV" 2>/dev/null
		continue
	else
		mount -o move "$MNTDEV" /union"$MNTDEV"
		continue # keep CHNDEV mounted
	fi
elif [ "$MNTDEV" = "$DATADEV" -o "$MNTDEV" = "$BOOTDEV" ]; then
	if param noauto && param copy2ram ; then
		umount "$MNTDEV" 2>/dev/null || mount -o move "$MNTDEV" /union"$MNTDEV"
	else
		mount -o move "$MNTDEV" /union"$MNTDEV"
	fi
	continue
fi
param noauto && umount "$MNTDEV" 2>/dev/null || mount -o move "$MNTDEV" /union"$MNTDEV"
done

# Remove unused /mnt/XX dir's
rmdir /mnt/* 2>/dev/null

And also symlinks /mnt/home and /mnt/saved creation moved to linuxrc, /usr/local/bin/mountlink removed.

This in GitHub Commit 157cf4b Update opt/initramfs scripts

And as you can see i update messages with color variables instead of directly writing codes.

Code: Select all

echo "${YELLOW}from= cheatcode is incorrect, press enter to search through all devices${RST}"

Next GitHub Commit aa5542e Update initramfs/linuxrc (re-write chroot part)

Re-write last part of linuxrc - pivot_root, exec chroot
Make it more compact. Support only merged-usr but possible to change it slightly to support non-merged-usr too.

Code: Select all

echo "${i}changing root directory" 
if [ -x /union/usr/sbin/init ]; then
	cp -f /union/usr/sbin/init /bin
	INIT=init
elif [ -x /union/usr/lib/systemd/systemd ]; then
	cp -f /union/usr/lib/systemd/systemd /bin
	INIT=systemd
fi

if [ -x /bin/$INIT ]; then
	mkdir -p /union/mnt/live
	echo "${BOLD}live system is ready now - starting Debian${RST}"
	pivot_root /union /union/mnt/live
	exec /mnt/live/bin/chroot . /mnt/live/bin/$INIT "$@" <dev/console >dev/console 2>&1
else
	echo "${RED}ERROR Something went wrong. Press Ctrl+Alt+Del to reboot.${RST}"
	sh
fi

Last, huge part GitHub Commit 5507eda Upd opt/initramfs scripts (re-write changes= code)

Re-write changes= and pfull= code
De-duplicate it. Make it more easy to understand (for me ).

This will be enabled in next v0006 phb_test .iso
I will do 2 versions: first 64bit only. and second 64+32 combined in one .iso
I will build and upload it soon to make new initrd available for testing.

Later i will modify fstab() function in finit to mount all partitions read-only, and then re-mount rw if changes= used, or noauto cheatcode not used.

Changes for ntfs ro driver in my To Do list for next test .iso:

Changes (with ntfs) currently not tested. Maybe this option can work after do some more modifications to linuxrc.
PocketHandyBox 64bit Kernel 5.10 Copy2RAM changes=EXIT:/phb/changes.dat

[gumanzoy]

v0006 Pre release version .iso:
github.com/PocketHandyBox/pocket-handybox/releases

md5sum:
5c9f7780afbcc8f71684f424744399e6 pocket-handybox_12_25_0006_vtgrub2.iso
706b6fd301d92d3908d0f0487da66a7e pocket-handybox_x64_12_25_0006_vtgrub2.iso

64+32bit combined in one pocket-handybox_12_25_0006_vtgrub2.iso

DogLinux repository disabled by OOTB. Uncomment lines if needed.
/etc/apt/sources.list.d/doglinux.list

Includes changes to initrd described in previous post.

Plus GitHub Commit bc9a21d Upd linuxrc and finit (add support to comma separated values)

Add support to both semicolon and comma separated values
Now can use load=module1,module2 or load=module1;module2

Added tigervncserver 1.15.0

To get remote access to PC with PocketHandyBox
There 3 options to run:

• Use vncpasswd to create /root/.config/tigervnc/passwd

• Run with cheatcode vnc=password

• Run with cheatcode vnc to connect without password

[gumanzoy]

@gumanzoy if there are other tests you want to see on this bare-metal use before you go to PHB v006, post and I will help where I can.

@Clarity @fredx181

Hi

Now feel free to test PHB v0006.
Check boot from different flash drives (formatted to NTFS/exFAT/FAT32/Ext4) on different PC's with changes= and without it.

Maybe somebody can try pfull= in PHB ?

[fredx181]

Hi @gumanzoy
Tested quickly booting the v006 ISO (x64) from Ventoy and all seems fine, setup changes too, fast booting, btw!

Then tried a manual frugal install booting from HDD (as I often use, and to possibly continue testing pfull=..) (I copied "live" in folder "phb"), grub entry:

Code: Select all

menuentry "PHB" {
insmod part_msdos 
insmod ext2
search --no-floppy --fs-uuid --set=root 34b10665-11c5-493c-9bff-f4f1e7db86f2
linux /phb/live/vmlinuz1 noauto from=/phb lang=en_US mitigations=off nvme_core.default_ps_max_latency_us=5500 nvidia=detect changes=/phb/live 
initrd /phb/live/initrd1.xz
}

But that failed to boot, with message that /live/01-filesystem.squashfs couldn't be found.
Any solution ?, e.g. modify my grub entry ?
edit: also tried with from=UUID:34b10665-11c5-493c-9bff-f4f1e7db86f2:/phb , same result, not booting.

[gumanzoy]

But that failed to boot, with message that /live/01-filesystem.squashfs couldn't be found.
Any solution ?

Yes. I try, and confirm - no boot.
But i add / to from=/phb/ and it boot fine.
Except squashfs from /phb/live/firmware not loaded.

Better to just rename live to phb and add folder=phb

Code: Select all

menuentry "PHB" {
insmod part_msdos 
insmod ext2
search --no-floppy --fs-uuid --set=root 34b10665-11c5-493c-9bff-f4f1e7db86f2
linux /phb/vmlinuz1 noauto from=/ folder=phb mitigations=off nvme_core.default_ps_max_latency_us=5500 nvidia=detect changes=/phb
initrd /phb/initrd1.xz
}

lang=en_US just ignored. It is from Ventoy variable $vtCurLang
Other lang= will be loaded if present in /live/lang/

---

Code: Select all

FOLDER=`value folder`
if [ -z "$FOLDER" ]; then FOLDER=live
CMDIR=live
[ `uname -m` = i686 ] && FOLDER=live32
else CMDIR=$FOLDER
fi

This code load /live/firmware

Code: Select all

# Find firmware modules
find $DATADEV/$CMDIR/firmware -name "*.squashfs" 2>/dev/null | egrep -ve "$NOLOAD" | sort >>/tmp/modules

I need to modify this code somehow.

Code: Select all

    if [ $FROM ]; then
	locate -d $FROM/$FOLDER
	if [ $? -eq 0 ]; then
	    DIR=`echo $LPTH | rev | cut -d/ -f3- | rev`; [ $DIR ] && FOLDER=$DIR/$FOLDER

I thick i will add CMDIR=$DIR/$CMDIR in case of locate success.

[Clarity]

Hi @gumanzoy

Like @fredx181 I booted the ISO file on bare-metal and KVMs. I tested via Ventoy booting on a non-nvidia test PC taking ONLY the COPY2RAM option for the initial tests. Synaptic run to check for upgrades only. Things are swift and stable.

Only reporting the following, thus far.

SWAP not picked up.jpg (87.91 KiB) Viewed 159 times

.

[gumanzoy]

@Clarity

Hi

Yes code for enable swap temporary disabled in v0006

I never use swap in my use-cases

I will re-write code for enable swap

[fredx181]

Hi @gumanzoy

But i add / to from=/phb/ and it boot fine.
Except squashfs from /phb/live/firmware not loaded.

Yes, with appending / it works.
And I moved the firmware module to "live", so all fine.

Next I tested pfull= option and initially didn't work but fixed by adding this to linuxrc (on line 272):

Code: Select all

################################################################################
mkdir -p /union/dev /union/mnt /union/live /union/proc /union/sys /union/srv /union/media /union/run /union/tmp
chmod 1777 /union/tmp
################################################################################

Re-created initrd1.xz and then it booted with pfull, but... X wouldn't start.
Found that disabling mesa-24.2.2-drm-2.4.123.squashfs fixed it (somehow conflicting with pfull filesystem )

Well, so it works, but not sure how much interest you or anyone have in a working pfull option for your project.
I'd say that for such a typical modular setup like yours it doesn't make much sense, but up to you of course.

[gumanzoy]

Next I tested pfull= option and initially didn't work but fixed by adding this to linuxrc (on line 272):

Code: Select all

################################################################################
mkdir -p /union/dev /union/mnt /union/live /union/proc /union/sys /union/srv /union/media /union/run /union/tmp
chmod 1777 /union/tmp
################################################################################

Before i delete it, because this directories already present in 01-filesystem.squashfs (created by debootstrap) Commit cefe809

I add this code again, but modify it
GitHub Commit 408550d Upd initramfs/linuxrc (Add mkdir for pfull=)

Code: Select all

# mkdir for pfull=
if [ $PFULL ]; then
mkdir -p /union/{dev,etc,media,mnt,proc,run,srv,sys,tmp,var/log}
chmod 1777 /union/tmp
fi

Well, so it works, but not sure how much interest you or anyone have in a working pfull option for your project.
I'd say that for such a typical modular setup like yours it doesn't make much sense, but up to you of course.

I am not sure for what. But i will keep this code just for keep compatibly with DogLinux.

Re-created initrd1.xz and then it booted with pfull, but... X wouldn't start.
Found that disabling mesa-24.2.2-drm-2.4.123.squashfs fixed it (somehow conflicting with pfull filesystem )

This needs for future investigation.
Maybe copy contents of mesa-24.2.2-drm-2.4.123.squashfs to pfull= ?

[fredx181]

Next I tested pfull= option and initially didn't work but fixed by adding this to linuxrc (on line 272):

Code: Select all

################################################################################
mkdir -p /union/dev /union/mnt /union/live /union/proc /union/sys /union/srv /union/media /union/run /union/tmp
chmod 1777 /union/tmp
################################################################################

Before i delete it, because this directories already present in 01-filesystem.squashfs (created by debootstrap) Commit cefe809

With the pfull setup, the 01-filesystem.squashfs is not loaded (just empty directory /mnt/live/memory/images/01-filesystem.squashfs). So 01-filesystem.squashfs can actually be deleted or disabled in "live".
Can't really explain (forgot) though why creating these directories in /union makes the difference. edit: perhaps to make the 'switch_root' properly

Well, so it works, but not sure how much interest you or anyone have in a working pfull option for your project.
I'd say that for such a typical modular setup like yours it doesn't make much sense, but up to you of course.

I am not sure for what. But i will keep this code just for keep compatibly with DogLinux.

OK!

Re-created initrd1.xz and then it booted with pfull, but... X wouldn't start.
Found that disabling mesa-24.2.2-drm-2.4.123.squashfs fixed it (somehow conflicting with pfull filesystem )

This needs for future investigation.
Maybe copy contents of mesa-24.2.2-drm-2.4.123.squashfs to pfull= ?

Probably yes, in this case. I guess that the reason that the modular setup can cause problems is because pfull/filesystem is on the upper layer, same as changes, not sure though.
(edit: now I did copy contents of mesa-24.2.2-drm-2.4.123.squashfs to pfull/filesystem (overwriting existing), booted, and X started ok)
edit2 : to be safe, it's best IMO that the pfull/filesystem contains everything, so without any .squashfs modules involved.

[gumanzoy]

Re-created initrd1.xz and then it booted with pfull, but... X wouldn't start.
Found that disabling mesa-24.2.2-drm-2.4.123.squashfs fixed it (somehow conflicting with pfull filesystem )

This needs for future investigation.
Maybe copy contents of mesa-24.2.2-drm-2.4.123.squashfs to pfull= ?

Probably yes, in this case. I guess that the reason that the modular setup can cause problems is because pfull/filesystem is on the upper layer, same as changes, not sure though.
(edit: now I did copy contents of mesa-24.2.2-drm-2.4.123.squashfs to pfull/filesystem (overwriting existing), booted, and X started ok)
edit2 : to be safe, it's best IMO that the pfull/filesystem contains everything, so without any .squashfs modules involved.

I think it is because mesa-24.2.2-drm-2.4.123.squashfs contains updated libdrm from backports. And 01-filesystem.squashfs contains stable libdrm version.
If contents of 01-filesystem.squashfs in upper layer - stable libdrm present and updated mesa not work and X not start.