Boot method: USB stick
Image version:
https://chuangtzu.ftp.acc.umu.se/cdimage/weekly-builds/amd64/iso-cd/debian-testing-amd64-netinst.iso
Date: File downloaded on the 18/Jan/2023

Machine: HP Victus by HP Gaming Laptop 15-fb0xxx/8A3D, BIOS F.16
12/13/2022
Processor: AMD Ryzen 5 5600H with Radeon Graphics
Memory: 64G (2x32G)
Partitions:
Filesystem           1K-blocks      Used Available Use% Mounted on
tmpfs                  6516464       128   6516336   0% /run
devtmpfs              32565932         0  32565932   0% /dev
/dev/sdb1               488448    488448         0 100% /cdrom
dev/sda1                102174       288    101886   0% /mnt/tmp

It should be noted I'm installing to a NVME which has been set to 4K
format. Not sure if that is relevant but the dmesg does mention errors
related to 4K

It should also be noted that while the 64G is seen and I can run tests
on it and windows can also see it and is happy with it; the bios is
reporting it as 0GB.

Base System Installation Checklist:
[O] = OK, [E] = Error (please elaborate below), [ ] = didn't try it

Initial boot:           [O]
Detect network card:    [E]
Configure network:      [ ]
Detect media:           [ ]
Load installer modules: [ ]
Detect hard drives:     [ ]
Partition hard drives:  [ ]
Install base system:    [ ]
Clock/timezone setup:   [ ]
User/password setup:    [ ]
Install tasks:          [ ]
Install boot loader:    [ ]
Overall install:        [ ]

Comments/Problems:

Its impossible to install Bookworm on to the fairly new (last year?) HP
laptop. I can install a minimal Bullseye; boot that; and then edit the
apt sources file to upgrade to Bookworm.

There are, however, a few bugs even with that method.

To get a working system I installed just the base Bullseye, then
changed the apt, then dist-upgrade and reboot. Then installed amd,
nvidia, headers, xfce, then dist-upgrade again.

The amd graphics drivers and, possibly, the nvidia graphics drivers
must be installed otherwise there is an "issue" with the boot.
Installing xfce without amd/nvidia causes a hard power off crash.

Sorry can't be more specific as I currently have it working and am
loathed to re-install again unless I can re-install directly to a clean
Bookworm. 

The nouveau seems to crash somewhere between starting the kernel (I
don't use "quiet" so I see a load of messages; then the screen blanks
that usually denotes switching away from VGA; then more output then a
power off hard crash) and bringing up the signon screen for xfce.

There is also an issue with the brightness of the screen (but I can
workaround that using amdgpu.backlight=1 to force full brightness) but
I figured I'll wait till the new backlight code comes live in a newer
kernel before I log that bug if it still exists.

I am also getting an issue with a random hard power off which seems to
happen if I don't sign on to the xfce but instead switch to vt1 and do
stuff there. There are no logs because when I say hard power off I
mean... I'm doing stuff like upgrades or rsync's and then - boom its
off. It seems to do it after about 6 mins or so.

I have attached two dmesg's. "dmesg" is the dmesg of the failing
Bookworm and "dmesg_bullseye" is from bullseye (not a complete log but
to the point where the Bookworm fails).

The install has crashed at various points. At one point I managed to
get as far as scanning for disks before the crash but now it seems to
be crashing as soon as I scan for network (I have used the very latest
version of the installer which overwrote the earlier version I had). I
even tried not checking the network but that still crashed on me.

The installer on vt1 just hangs with a blank screen but the log on vt4
carries on and the interactive session on vt2 works (hence the various
log files).

I am more than willing to help debug this further, but if I need to do
something (say apply a patch, or add something to the installer iso)
then I am going to need to be walked through it in detail. eg: do this,
then this, then this, then that, now send me...

I have used attachments but if this isn't correct please let me know.

And finally. I think I've sent this to the correct email address but I
am not subscribed to any debian lists so please let me know (cc) which
list this goes to so I can subscribe or view it online.

Jon Wilson.

It seems this is a problem with other laptops as well as mine as others
have posted to this bug tracker with similar problems (which I hadn't
seen before I posted). But I shall post the workaround here so if
others search for it it may help.

I hadn't realised that there is more information than the normal
"dmesg" within "syslog" (something to remember).

For the record pressing "e" on the installer boot option that you want
to run and then inserting "modprobe.blacklist=mt7921e" (without quotes)
on the line before the ---- allows the installer to work beyond the
point of the error message (as far as the scanning disks which found
the nvme and the partitions on it successfully; I didn't go any further
as I don't want to reinstall again).

The issue seems to be that the driver isn't loaded, and then is being
removed which is causing the crash dump. At this point this is way
beyond my abilities... but as the workaround fixes the issue (however
[in]correctly) I'm happy.

I vaguely remember that when I was running Bullseye for a short time it
didn't have a driver for the wireless/Bluetooth and required me to
download a driver from another site and DKMS it so I'm guessing that is
why the install works on Bullseye (it doesn't even bother to try and
load it as its much newer than that kernel?) and not Bookworm. Although
it seems bad that failing to remove a module that is not loaded causes
the whole thing to fall down in a big messy heap.

At least now I know about the bug tracker and how to actually search it
(the layout really isn't for the uninitiated!) and that it exists (does
google even index the tracker?) the next time I have an issue this will
be the first port of call.

Jon

Not exactly. The module is loaded but fails to unload gracefully when
the firmware is missing. This is actually a bug in the Linux kernel, not
in the installer.

Yes, the offending module does not exist in Bullseye's kernel (5.10).