Kernel Cannot Find Any Block Devices

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

Hi,

First of all, thanks for all of the support you guys have, it's been really helpful in getting as far as I have. Now for my query: I'm trying to dual boot gentoo with windows 10 (windows 10 installed first), and I'm pretty sure everything is set up correctly (the kernel functions well, the boot drive shows up in rEFInd, etc). My issue is that on selected the boot drive from rEFInd, it runs into an error "cannot find root drive in ."

I assume it's a mounting issue with my /etc/fstab/

Here's a structure of my partitions (all on one drive)

NeddySeagoon · Posted: Sat Jul 27, 2019 8:08 pm Post subject:

ijcon40,

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

Thanks for your advice, I went ahead and added the respective nvme driver support and recompiled the kernel. My process in doing so it make menuconfig, make modules_install, make install in that order when building the kernel.

This brings me to my next issue which I assume is directly related to me not having any idea what im doing mounting-wise. My new error is

Jaglover · Posted: Sun Jul 28, 2019 12:57 am Post subject:

You have:

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

Hi,

Thanks for all your help thus far. For the next chapter of my incompetence, I went ahead and added ext4 file support in the kernel. I would prefer to not use initramfs as I don't really need it with the setup I have (pretty sure about that). I was also getting the error:

NeddySeagoon · Posted: Sun Jul 28, 2019 11:17 am Post subject:

ijcon40,

I asked for your entire kernel .config because what is missing is just as important to what is present.
I can't see what is missing from filtered output.

Don't worry about incompetence. That can be fixed by education and practice. All the helpers here have been through the learning experiences you are going through now.
Most of us will teach you to fish, rather than feed you, so we expect you to contribute to the solutions to your problems too.

Contrast incompetence to stupidity. Stupidity is forever. With the philosophy out of the way, back to your issues. You have several.

Some background. The boot sequence solves the problem of loading an operating system stored on some block device using a few tools that don't know very much about the block device.
There are lots of circular dependencies to resolve on the way and that's where your current issues arise.

Something loads your boot loader. BIOS and UEFI do it differently but bit its all the same from the grub menu being displayed to the login prompt.
Grub loads your kernel and optionally initrd(s). You can have several.
Now its your kernel and optional initrd all alone in RAM. This combination has to be able to mount the root filesystem, since until its mounted nothing can be read from it.
You can infer that with no initrd, the kernel is on its own. In turn, this means that everything needed to mount root must be configured as <*>, since <M> puts the loadable modules on the HDD.
Its a very bad thing to need to read the HDD to load a module that is needed to read the HDD. That's one of the circular dependencies that has to be broken.

Looking at your kernel

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

Alright, so I made those changes as well as removing some very extraneous options (ones that clearly were not applicable to my setup). My next question is around the bootloader. The guide for dual booting windows and gentoo https://wiki.gentoo.org/wiki/UEFI_Dual_boot_with_Windows_7/8 specifically highlights skipping the bootloader configuration step in the general amd64 handbook. As such, I don't have a grub.cfg set up. I do have the override of commandline for the bootloader,

Tony0945 · Posted: Sun Jul 28, 2019 5:13 pm Post subject:

If you are using refind, you can (should? must?) set the boot device and other parameters in linux.conf.
The path to mine is /boot/refind_linux.conf but yours may be different based on refind version.

Here is mine from my server :

NeddySeagoon · Posted: Sun Jul 28, 2019 5:23 pm Post subject:

ijcon40,

I guessed at grub for your boot loader

You have

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

Alright, so I added the refind_linx.conf in /boot as recommended by Tony as well as adding the rootwait to the command line although I'm not entirely sure how to take a screenshot from command line, so I wasn't able to do that. After adding the refind_linux.conf, the error has now changed to unknown-block(259, 7) rather than unknown-block(0,0). In addition, I noticed that it is picking up the nvme drive and partitions along with their partition UUIDs, but it has the tag (driver?), which I figured could mean that I have not installed a driver to aptly read from them. It also stated:

NeddySeagoon · Posted: Sun Jul 28, 2019 9:55 pm Post subject:

ijcon40,

Tony0945

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

First of all, apologies for the typo in my previous message, it actually is refind_linux.conf, that was just me being dumb in the message.
I went ahead and disabled the initrd and initramfs in the kernel config following the resolution of this thread https://forums.gentoo.org/viewtopic-t-1059434.html. That removed some of the error text from the boot up error screen. I also took a picture of what I'm getting error wise with rootwait (not entirely sure that it's working though because it doesn't permanently pause) https://imgur.com/a/2ZpsjXH. I don't know if this is visible, but it also labels (driver?) in the error screen above the nvme partitions. I don't know if that means I don't have driver support enabled, but it seems likely given that it can't read the filesystem. I checked out the root fs and it appears to be an EXT4, and EXT4 support is included in the kernel (non-modularly). Thanks for all the advice guys.

NeddySeagoon · Posted: Mon Jul 29, 2019 11:00 am Post subject:

ijcon40,

The image is very useful. It lists all the partitions the kernel can see.
There are your 16 RAM drives ram0..ram15
Then there are your seven partitions on your NVME drive.

One of those partitions satisfies rootwait, so rootwait allows booting to continue.

Together, that tells that the kernel can see your NVME drive and read the partition table (or it could not list the partitions).
That only leaves two things,

a) The kernel does not contain the filesystem to read the root partition.
b) The partition given as root= on the kernel command line is not the root partition.

It could be both :)

Boot into your install media, mount your partitions then run df -Th. It will tell the filesystem types.

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

Alright, so I've made sure that EXT4 is <*> is in the kernel and that pretty much all the options for NVMe are supported. So my next question is a little stupid: which drive is supposed to be root? My general way of editing the files is to mount what I think is the root drive, and then follow the mounting stages of the amd64 handbook base installation section. This is a really dumb question, but what is the purposes of each of the partitions, I know that swap is supposed to serve as a hard drive based paging solution and that boot is supposed to deal with all the mounting, but I'm not entirely sure what files/ contents root should contain. I know that I very rarely touch the boot partition manually, and I thought I was actively using root, but looking at the actual commands I run, very rarely am I actually editing anything in root, boot or swap. Unless /usr/src/linux or /etc/fstab are somehow related to root, I don't think I'm really editing it at all. It's also a little worrying that my root partition shows up in EFI, although I presume that is a good sign given that I have EFI_STUB=y. I guess my question relates more to what exactly the boot partition should be doing and why what I thought is my root partition is showing up as an executable in rEFInd when I thought that the kernel was located solely in /boot?

Again, thanks for all the help.

NeddySeagoon · Posted: Tue Jul 30, 2019 1:43 pm Post subject:

ijcon40,

Unlike windows, where everything gets a drive letter and things start to break with more than 26 drives, *NIX attaches everything somewhere in the filesystem tree.
The filesystem tree is rooted at root, denoted /.

After that, what is on separate partitions or drives or shared from other systems is hidden by the system.
There are a few ground rules, like everything needed to boot should be on /. That way, you can fix your system without access to any other filesystem.
Its up to the system admin - thats you.
Its a good idea to make /home its own partition. That's all your user data. Then you can reinstall without messing up your user data, should you ever need to reinstall.

My complete filesystem tree is

Tony0945 · Posted: Tue Jul 30, 2019 1:49 pm Post subject:

NeddySeagoon · Posted: Tue Jul 30, 2019 2:50 pm Post subject:

Tony0945,

<nit>

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

So correct me if I'm wrong, but since I'm using rEFInd, would it possible for me to use EFI_STUB and remove the boot drive entirely? If I place the /boot folder on the root filesystem when it mounts on / it should be able to share the same functionality but simplify the system a fair bit.

This is just for clarification: Here's a timeline of how I think this is functioning.

Computer Start->Bios Mananger/Drive Select->Windows EFI-> rEFInd->boot->refind_linux.conf->kernel->load drivers->rootfs->modules->Running Gentoo

Would it be possible to bypass the "boot->refind_linx.conf->kernel->rootfs" pathway by just having the boot drive mounted on root? Would it just make sense to restart this install from the ground up and use a LVM next time? I'm very confident that I'm mounting the root drive and that it is actually the root drive. I also know the ext4 is <*> in my kernel config. I may just restart the install from the ground up with the exception of leaving rEFInd installed (as it is on the windows EFI partition), unless you (the more experienced users) would recommend just continuing with the install/ if you think this is a simple issue and readily fixable. Thanks for the help.

Jaglover · Posted: Tue Jul 30, 2019 5:13 pm Post subject:

I don't know what Windows EFI is and why you think it is chainloading rEFInd.
the shortest boot sequence would be:

Power on;
EFI firmware runs and loads EFI stub kernel;
kernel loads drivers, mounts rootfs and runs init system.
_________________
My Gentoo installation notes.
Please learn how to denote units correctly!

NeddySeagoon · Posted: Tue Jul 30, 2019 5:23 pm Post subject:

ijcon40,

Your timeline is correct but its missing some constraints.
You need a partition formatted as VFAT to hold your EPI programs.
You can indeed make your kernel an EFI compliant program but EFI can read exactly one partition type and that's VFAT.

Neither *NIX nor Windows can be installed on VFAT. Windows insists on NTFS and *NIX gives you choices. VFAT does not support permissions or symbolic links, so its a really bad idea to try to install any operating system on a VFAT volume. That's what you are suggesting by combining the two.

If we dig into recent history before UEFI, BIOS allows you to combine root and /boot is the way you suggest.
At runtime, the grub boot loader can read exactly one filesystem type but that filesystem type is selected at install time, not dictated by the firmware itself.
Here, merging boot and root works, most of the time.

The down side of making your kernel a UEFI program and having the UEFI firmware load it directly is that the kernel command line must be built into the kernel, so its a kernel rebuild to change it.

Do not reinstall. At least, not yet. You will come to this point again and still won't know what is wrong?

Did you turn off

Tony0945 · Posted: Tue Jul 30, 2019 9:46 pm Post subject:

Re no separate boot. yes and no

Yes, you don't have to have a seperate boot partition, no you can't have one big root partition.
As Neddy says, you can't have root on VFAT and for UEFI you must have /boot/efi on VFAT
This leads me to this fstab:

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

Ok, so I've removed a ton of irrelevant modules from my config, ran make modules_install, make install to update the kernel and rebooted (after unmounting). I got the same error unknown-block(259,7). Here is all the information I can think to include.

Here are some pictures of my errors, hopefully they are better than the other. https://imgur.com/gallery/ugmiyAL

Here's what my root partition (/dev/nvme0n1p7) looks like:

ijcon40 · n00b Joined: 27 Jul 2019 Posts: 18

To answer your partition question, here's an fdisk -l of my partitions:

Tony0945 · Posted: Tue Jul 30, 2019 9:58 pm Post subject:

Ah! Forgot you are dual booting. I only do that on one old BIOS boot system using grub that chain-loads windows.