Install FreeBSD bootcode from Linux rescue system

CyberCr33p

CyberCr33p

I have a system like this:
gpart show
=> 40 7501476448 nvd0 GPT (3.5T)
40 1024 1 freebsd-boot (512K)
1064 984 - free - (492K)
2048 33554432 2 freebsd-swap (16G)
33556480 7467919360 3 freebsd-zfs (3.5T)
7501475840 648 - free - (324K)
Click to expand...
If after a FreeBSD upgrade I forget to update the bootloader then is it possible to install it using a Linux rescue system and command dd?

In the Linux rescue system I can mount the zfs pool:


zpool import -N -R /mnt zroot
zfs mount zroot/ROOT/default


And /mnt/boot/pmbr and /mnt/boot/gptzfsboot are accessible.

And in Linux rescue system the disk is found as: /dev/nvme0n1
 
Do you think these commands are correct?


dd if=/mnt/boot/pmbr of=/dev/nvme0n1 bs=512 count=1
dd if=/mnt/boot/gptzfsboot of=/dev/nvme0n1 bs=512 seek=40
 
CyberCr33p said:
Do you think these commands are correct?


dd if=/mnt/boot/pmbr of=/dev/nvme0n1 bs=512 count=1
Click to expand...
Nooo! Not 512. You will destroy the partition table. Ok, it's a fake one but... The number of bytes to write is precisely 446. I don't know with linux, but the FreeBSD dd can't write less than 512 bytes IIRC, so this command is a bad idea. I will let the current bootcode. The update of this piece of code brings nothing interesting.

CyberCr33p said:
Do you think these commands are correct?


dd if=/mnt/boot/gptzfsboot of=/dev/nvme0n1 bs=512 seek=40
Click to expand...
Should work if you remove "seek" and if the linux nvme0n1 is well the FreeBSD nvd0p1.

That said, ataxa1a is right. It would be simpler and less dangerous to use a FreeBSD booting key and use the suited commands.
 
Linux nvme0n1 = FreeBSD nvd0 (not nvd0p1). So I believe seek=40 is needed in this case to skip the first 40 sectors (MBR and unused space) and write into the freebsd-boot partition starting at sector 40. Am I right?
 
CyberCr33p said:
Linux nvme0n1 = FreeBSD nvd0 (not nvd0p1). So I believe seek=40 is needed in this case to skip the first 40 sectors (MBR and unused space) and write into the freebsd-boot partition starting at sector 40. Am I right?
Click to expand...
Ah ok. Can't you use the partition directly: nvme0n1p1? This is less prone to error.
 
Please use tags code when pasting code, not cmd. It's less error prone to read output such as this one:
Code: 
gpart show
=>        40  7501476448  nvd0  GPT  (3.5T)
         40        1024     1  freebsd-boot  (512K)
       1064         984        - free -  (492K)
       2048    33554432     2  freebsd-swap  (16G)
   33556480  7467919360     3  freebsd-zfs  (3.5T)
 7501475840         648        - free -  (324K)

CyberCr33p said:
Do you think these commands are correct?


dd if=/mnt/boot/pmbr of=/dev/nvme0n1 bs=512 count=1
dd if=/mnt/boot/gptzfsboot of=/dev/nvme0n1 bs=512 seek=40
Click to expand...
We are assuming block size is 512.
In theory these commands do look OK. With few buts and gotchas though..

Emrion did raise few important points, I will address them a bit further.

1) pMBR
It's important to distinguish between legacy boot and UEFI boot with GPT layout. In your case it's the former; pMBR is important.

Simply said pMBR in legacy boot searches for freebsd-boot UUID, loads it into memory and continues there. GPT header/structures are used to get the partition information. Still to preserve the fake partition from your current disk you could do this:
Code: 
dd if=/dev/nvd0 of=pmbr bs=512 count=1                                   # save current pmbr
dd if=/boot/pmbr of=./pmbr bs=1 count=446 conv=notrunc      # write 446 (0x1be) to a file, don't truncate the output
dd if=./pmbr of=/dev/nvd0 bs=512 count=1                   # copy it back to nvd0 ; count is not mandatory but it's safer .. you could also check its size before actual dd-ing
You need to do it like this because you can't write less than a block size to a disk.

2) actual FreeBSD bootloader
If you can you should do it as Emrion pointed out - use partition. It will keep you within its limits if something goes wrong. Now we still have a problem - we can't write any size to a disk directly, it has to be in blocks. My current gptzfsboot:
Code: 
# ll /boot/gptzfsboot
-r--r--r--  1 root wheel 176358 Nov 29 11:22 /boot/gptzfsboot
Is not .. Your dd will not write all the data which can be a problem.
Code: 
dd if=/boot/gptzfsboot of=/dev/nvd0p1 bs=512 conv=sync
Would do the trick.

While it's not mandatory I strictly use iseek/oseek when doing skip/seek. They are aliases that clearly communicate on which side (input/output) skipping/seeking is to be done.
 
Finally in Linux rescue mode (something like mfsBSD but for Linux) I run:

Code: 
dd if=/dev/zero of=/dev/nvme0n1p1 bs=512 conv=sync
dd if=/dev/zero of=/dev/nvme1n1p1 bs=512 conv=sync

Then I verified that the system doesn't boot.

Then I boot again in Linux rescue mode and I run:

Code: 
zpool import -N  -R /mnt zroot
zfs mount zroot/ROOT/default

dd if=/mnt/boot/gptzfsboot of=/dev/nvme0n1p1 bs=512 conv=sync
dd if=/mnt/boot/gptzfsboot of=/dev/nvme1n1p1 bs=512 conv=sync

Then I verified that the system boots successfully.

Thank you all for the help.
 
You must log in or register to reply here.
Back
Top