Samsung ssd 950 pro m.2 hard drive

[arnesp]

At least one person was able to apply the installer patches to a USB drive and boot from the USB to install Mageia 5

Sorry for the imprecise description in the bug report. Since I only needed to do one install, I booted an unmodified live USB and did the patch in the virtual file system (in /usr/lib/libDrakX) before double clicking the "Install to Hard Disk" icon.
To allow diskdrake to detect the nvme drive, /usr/lib/libDrakX/detect_devices.pm needs to be patched again after install and update.

[RagingRaven]

Thanks for your reply.

I've created a bootable live USB disk, but seeing as I've never changed anything during installation, how do I change the virtual file system (in /usr/lib/libDrakX) during setup?
I've tried ALT-F1-F5, but I don't know how to get a console to edit the virtual file system.

Could you provide me with a step-by-step noob guide on how to do this? Or is it perhaps possible to edit the files before booting from the usb drive?

[arnesp]

With the live USB, you do not need to change anything during installation. You can just boot into the live system and do the changes from the desktop environment and then start the install from there.
I modified the files (2 perl module files) manually using kwrite, but if you are not comfortable with doing that, I could try to re-create and upload the updated files. You would then just need to download and replace them while running the live system (assuming you trust me not to sneak malware into your installation ).

[RagingRaven]

Ah ok, I get it now, I did make a live usb drive, but chose 'install mageia' during boot instead of booting into the live environment
Stupid me, now I get it.

[RagingRaven]

Ok, this is annoying, now i've run into the following problem while booting the live stick:

During boot it hangs on: Using '/etc/ld.so.conf.d/GL/standard.conf' to provide 'gl.conf.'
Which is probably a problem with Proprietary graphics card drivers, seeing as I have a GeForce card, so I should add 'xdriver=fglrx', but ofcourse I can't change these Kernel Options, because it's a UEFI motherboard and the live stick uses grub2.
The solution would be: https://wiki.mageia.org/en/How_to_set_u ... oot_Loader but this seems to only apply to installation, not booting into live environment, because pressing e after choosing 'boot mageia 5 live' doesn't seem to work.

Is there anything else I could try? Or is there any idea on when there's going to be new ISO's where the bugfixes are included?

[arnesp]

Not that it is of much use to you, but I have no problem getting to the line editor by pressing e (also on an UEFI board).
I am using the KDE live DVD image.

[RagingRaven]

Again, it appears I'm stupid, I misunderstood 'Select "Start Mageia 5 Install" ' as meaning select and continue and then press e.
But you don't actually have to select either option, you just need to press e during the selection screen.

Well, lets see if I can boot into live now...

[RagingRaven]

Ok, I managed to boot into the live environment (with xdriver=nvidia kernel option, misread ati part earlier) and have now changed both /usr/lib/libDrakX/detect_devices.pm and /usr/lib/libDrakX/devices.pm to reflect changes in these commits:

http://gitweb.mageia.org/software/drakx ... c4b5e5b97c
http://gitweb.mageia.org/software/drakx ... 8d1d23da80
http://gitweb.mageia.org/software/drakx ... aebdd538fd
http://gitweb.mageia.org/software/drakx ... ad1b9193dd
http://gitweb.mageia.org/software/drakx ... 9fcd17b446

From what I understood of those, it seems some are iterations of a previous commit so the final changes I made to both files are as follows:

devices.pm:
changed

Code: Select all

$part->{device} =~ /((?:[hsv]|xv)d[a-z]|mmcblk|nvme\d+p)(\d+)$/;

to

Code: Select all

$part->{device} =~ /((?:[hsv]|xv)d[a-z]|mmcblk|(nvme\d+n)\d+p)(\d+)$/;

detect_devices.pm:
changed

Code: Select all

getIDE(), getSCSI(), getMmcBlk(), getNVMe() getXenBlk(), getVirtIO(), getDAC960(), getCompaqSmartArray();

to

Code: Select all

getIDE(), getSCSI(), getMmcBlk(), getNVMe(), getXenBlk(), getVirtIO(), getDAC960(), getCompaqSmartArray();

added

Code: Select all

sub getNVMe() {
    -d '/sys/class/nvme' or return;
    map {
            { device => basename($_), info => "NVMe block device", media_type => 'hd', bus => 'pci_express' };
    }
    glob("/sys/block/nvme*");
}


after function getMmcBlk()

Before I continue with the installation, is this correct? Do I need to change anything else?

I also noticed this in the bug report, is this still needed with the above changes?:

2: /sys/bus/pci_express/devices was empty in my system. Replacing pci_express with pci allowed the installer to recognize the NVMe device.

Note: this was with the kernel from the live ISO. After upgrade to latest kernel, there was no /sys/bus/pci/devices/*/block directory and /sys/bus/pci_express/devices was still empty

And lastly:

To allow diskdrake to detect the nvme drive, /usr/lib/libDrakX/detect_devices.pm needs to be patched again after install and update.

Does this also go for the changes to devices.pm?

[arnesp]

Looks good. Just one problem:
The virtual directory /sys/class/nvme does not exist when running the 3.19 kernel that comes with the live system.

I think the check for /sys/class/nvme is just to detect if an NVMe device is present, so it should be safe to replace it with some other existing NVMe related directory like /sys/module/nvme.

You probably do not need to change anything after updating the system, but you may want to undo the above hack

[RagingRaven]

So change

Code: Select all

sub getNVMe() {
    -d '/sys/class/nvme' or return;
    map {
            { device => basename($_), info => "NVMe block device", media_type => 'hd', bus => 'pci_express' };
    }
    glob("/sys/block/nvme*");
}


to

Code: Select all

sub getNVMe() {
    -d '/sys/module/nvme' or return;
    map {
            { device => basename($_), info => "NVMe block device", media_type => 'hd', bus => 'pci_express' };
    }
    glob("/sys/module/nvme*");
}


And then after installation change it back / redo the changes to detect_devices.pm?

[RagingRaven]

Just did the above change, but the 'install on hard disk' wizard just crashes when clicking next.
If I leave it at /sys/class/nvme it does continue, but it doesn't find my samsung 950.

[arnesp]

You should only replace /sys/class/nvme, not /sys/block/nvme in the last statement

[RagingRaven]

Ok thanks, it sees my drive now.

However there appears to be a problem with the installer, according to https://wiki.mageia.org/en/Installing_o ... I_firmware I should be able to choose 'Use free space':

If "Use free space" is selected, you have nothing else to do, the system is going to automatically create three or four partitions according to the available space and using it entirely:
a ESP partition for UEFI boot loader of 300 Mb, mounted on /boot/EFI
a partition for / of at most 50GB
a partition for /home if the total available space is more than 30 GB, under 30 GB, / and /home are on the same partition.
a partition for swap of at most 4 Gb

However I get a message stating:

you must have a esp fat32 partition mounted in /boot/efi

So I had to create the partitions manually, but then the installation finally started.

[RagingRaven]

I'm going from one problem to the next.

So as said the installation finally started, I then got a screen which asked me for bootloader main options (never saw that before, but i'm guessing that's related to grub2/uefi).
It showed me a couple options on my nvme drive, I chose the one that didn't have any mount points mentioned as I didn't see /boot/EFI as one of the options.

It then showed a screen in which I could choose for Mageia and Mageia with linux desktop, so I chose the latter.
After completing it told me to halt the system and remove the live media, which I did.

It now boots into the bootloader (grub2 I think) where I can choose Mageia and Advanced options for Mageia.
When I choose Mageia I get a message stating something like "ignoring bgrt: invalid status 0 (expected 1)"

This also happens when I choose Advanced options and choose 'Mageia with linux desktop'.
I also changed the kernel option of that line so that it included xdriver=nvidia, but that didn't help.

All I can find about it relating to Mageia is this:
viewtopic.php?t=8912&p=54167
But there's no solution/reply given there.

P.S. it's getting quite late over here (Netherlands) so I'll continue experimenting tomorrow

[arnesp]

@RagingRaven
Did you manage to get your installation booted?
If you need more advice, you probably need to provide more details of what happens, when you try to boot

[RagingRaven]

I have been a bit busy with other things so I only got to it again today.

To answer your question, no I've not been able to get my system to boot yet.
I don't think it's related to the NVME drive at the moment, looking at the little info I can find about "ignoring bgrt" it seems to be related to the graphics, but I'm not having much luck in finding a solution.

The exact message I'm getting after choosing (or letting it automatically continue with) "Mageia" in grub2 is:

Code: Select all

[     0.034751] Ignoring BGRT: invalid status 0 (expected 1)

As stated I've tried to change the kernel options to include xdriver=
I've tried:

Code: Select all

xdriver=nouveau
xdriver=nvidia
xdriver=intel
xdriver=efifb
xdriver=vesa
xdriver=fglrx
xdriver=radeon


Just to be clear (as perhaps I'm placing it at the wrong location), I've added it at the end of the line:

Code: Select all

Linux          /boot/vmlinuz-desktop root=UUID=<long code here, don't feel like typing>\<some more code> ro    splash quiet noiswmd

So the line would be:

Code: Select all

Linux          /boot/vmlinuz-desktop root=UUID=<long code here, don't feel like typing>\<some more code> ro    splash quiet noiswmd xdriver=<specific type>

And pressed F10 after to boot.

Seeing as I have an Intel Core i5-6600 and a Nvidia GTX950, I'd think it should be 'intel', nouveau or nvidia.

Following some more research online I've also tried:

Code: Select all

rd.blacklist=nouveau rd.blacklist=nvidia rd.blacklist=intel xdriver=efifb
rd.blacklist=nouveau rd.blacklist=nvidia rd.blacklist=efifb xdriver=intel
rd.blacklist=intel rd.blacklist=nvidia rd.blacklist=efifb xdriver=nouveau
rd.blacklist=nouveau rd.blacklist=intel rd.blacklist=efifb xdriver=nvidia


And seeing as https://wiki.mageia.org/en/Mageia_5_Err ... d_graphics mentions both rd.blacklist= and rdblacklist= I've also tried it without the dot:

Code: Select all

rdblacklist=nouveau rdblacklist=nvidia rdblacklist=intel xdriver=efifb
rdblacklist=nouveau rdblacklist=nvidia rdblacklist=efifb xdriver=intel
rdblacklist=intel rdblacklist=nvidia rdblacklist=efifb xdriver=nouveau
rdblacklist=nouveau rdblacklist=intel rdblacklist=efifb xdriver=nvidia


As soon as I chose

Code: Select all

rdblacklist=nouveau rdblacklist=nvidia rdblacklist=intel xdriver=efifb
or
rdblacklist=nouveau rdblacklist=nvidia rdblacklist=efifb xdriver=intel


The boot process went a bit further, I then got:

Code: Select all

[     0.034751] Ignoring BGRT: invalid status 0 (expected 1)
Starting netprofile:   [     4.281556] kvm: disabled by bios
The default is the current profile.
[  OK  ]
[     4.297342] kvm disabled by bios
Checking for new hardware_


Edit:The second time I tried this it also mentioned something about an unexpected inconsistency: run fsck manually, but this was gone after running the below kernel options.

With

Code: Select all

rdblacklist=nouveau rdblacklist=intel rdblacklist=efifb xdriver=nvidia


It goes even further, I see multiple [ OK ]'s, but it ends in:

Code: Select all

[  OK  ] Mounted /boot/EFI.
[  OK  ] Started filesystem check on /dev/nvme0n1p3.
            Mounting /home...
Starting netprofile:   [     4.287659] kvm: disabled by bios
The default is the current profile.
[  OK  ]
Checking for new hardware[     4.304337] kvm disabled by bios


And then it hangs

Edit: Seeing as

Code: Select all

rdblacklist=nouveau rdblacklist=intel rdblacklist=efifb xdriver=nvidia


went the furthest, I lastly tried:

Code: Select all

rdblacklist=nouveau rdblacklist=intel rdblacklist=efifb xdriver=nvidia nouveau.runpm=0


as this was mentioned at https://wiki.mageia.org/en/Mageia_5_Err ... d_graphics as well, but still no go.

Sorry about the long read, but I think it's better to be complete when menioning what I did.

[arnesp]

Sorry about the long read, but I think it's better to be complete when menioning what I did.

definitely

It might be an idea to see if you can boot into "CLI mode" (replace xdriver=nvidia with 3 on the boot line).
If this is successful, you may configure and start network and perform an update from the command line. This will give a 4.4 kernel and newer graphical drivers, which might better support your HW.
You might then be able to successfully configure and start a graphical interface.

If you cannot boot to command line mode, maybe it is worth booting the live USB, mounting the partition where root was installed and look through log files.

[RagingRaven]

When I try

Code: Select all

rdblacklist=nouveau rdblacklist=intel rdblacklist=efifb 3
or
rdblacklist=nouveau rdblacklist=intel rdblacklist=efifb 3 nouveau.runpm=0
or
rdblacklist=nouveau rdblacklist=intel rdblacklist=efifb xdriver=nvidia 3
or just
3


at the end of the boot line, it goes back to

Code: Select all

[     0.034763] Ignoring BGRT: invalid status 0 (expected 1)
[  OK  ]
Checking for new hardware_


I'll see if I can find anything when booting from the live USB, but it is looking more and more like something that's growing beyond my ability to fix, I might have to wait for someone to make new installation iso's with the nvme fix included.

Edit: I found another topic which stated you could remove quiet from the boot line to see more info, so I did and now it pretty consistently ends in the following message:

Code: Select all

MXM: GUID detected in BIOS
fb: switching to nouveaufb from EFI VGA


No matter which combination of rbblacklist and xdriver I enter.

Edit 2: Read some more topics on nouveaufb switching and read somewhere I could enter nomodeset to kernel options, so I did, it now appears to continue, but then goes into emergency mode, but I have no idea what to do there.
I can press Enter for maintenace or CTRL-D to continue, pressing CTRL-D gives me a message:

Code: Select all

EXT4-fs (nvme0n1p3): bad geometry: block count 54573398 exceeds size of device (54573393 blocks)

and then goes back to emergency mode, pressing enter gives me a console.

Edit 3: I did some more surfing and came to the following pages:
https://nouveau.freedesktop.org/wiki/VideoAcceleration/
http://my30daysoflinux.blogspot.nl/2015 ... rs-in.html
So it seems the GTX950 isn't really supported with the nouveau drivers and I should use the nvidia drivers, the second link shows how to install those driver in ubuntu, perhaps I need to do something similar in mageia? Or does mageia include the nvidia drivers?

Edit 4: According to https://wiki.mageia.org/en/Setup_the_graphical_server the nvidia drivers are not bundled with Mageia, but there they only provide an installation method if you are allready able to boot.

Edit 5: I tried doing a CLI upgrade following https://wiki.mageia.org/en/Mageia_5_Rel ... _.28CLI.29 but I don't appear to have internet access while in the maintenance mode.
I'm stumped, maybe I'm better off waiting for new ISO releases, I would be really happy if somebody could make them. I might have to switch to another distri otherwise and I really don't want to

[arnesp]

I don't really know about detailed plans, so somebody please correct me, if the below assumptions are wrong.

Bug 17743 has been closed without new ISOs (or relevant RPMs) being released, so I am guessing that it has been decided not to solve the issue for mga5.
In the bug report Adrien Guichard reports success with a network install, but I suspect this is with cauldron.
A first stable cauldron ISO containing the necessary NVMe support is currently being tested, but is, I believe, not yet available to the general public.
I have cauldron installed (on a SATA disk) and it seems to be absolutely usable, so depending on your needs, you may try to install that.

If you decide to spend more time on mga5, I think the "bad geometry" issue should be the first to be dealt with. For example by starting installation from scratch and perform fsck on and test mount of the NVMe partitions in between installation and first boot.

Since you are able to get to a desktop with the live USB boot, I think it should also be possible (unless you are hit by a strange timing issue) when you are booting more or less the same code from a different media. One of the first differences is the initrd image, so it might be interesting booting the installed system using the live system initrd.img. This could be done from the live USB boot-loader by replacing the root specification on the boot line with root=/dev/nvme0n1pX (X number of partition where root was installed) and of course adding what you previously found to be needed for the live system boot.

[arnesp]

I just noticed https://bugs.mageia.org/show_bug.cgi?id=18666.
If your corrupted partition nvme0n1p3 (/home ?) is caused by this bug, apparently reduction of the partition size by 1 block could be a workaround.

[RagingRaven]

Ill give the smaller partition size a try when I have a bit of time again.
Thanks for your help so far.