GRUB prevents booting a degraded RAID1 array atop LUKS

Open

Details

2 participants

Giovanni Biscuolo
maxim.cournoyer

Owner: unassigned

Submitted by: maxim.cournoyer

Severity: normal

maxim.cournoyer wrote on 1 May 2020 15:56

Recipients:(name . bug-guix)(address . bug-guix@gnu.org)

Message-ID:87o8r7kbbz.fsf@hurd.i-did-not-set--mail-host-address--so-tickle-me

On a system where:

1) Each disks comprising the array is fully LUKS encrypted

2) Each mapped disk is made part of a Btrfs RAID1 array

When attempting to boot the system after pulling out (in BIOS or using

the cable) the drive to simulate a complete disk failure, GRUB hangs,

prompting for the LUKS password of the disappeared drive and

(unsurprisingly) failing to open it.

This prevents booting in a degraded LUKS encrypted, Btrfs RAID1 on Guix

System.

Maxim

Maxim Cournoyer wrote on 7 Aug 2021 07:06

Recipients:(address . 40999@debbugs.gnu.org)

Message-ID:874kc14zd9.fsf@gmail.com

Hello,

maxim.cournoyer@gmail.com writes:

Toggle quote (13 lines)> On a system where:
>
> 1) Each disks comprising the array is fully LUKS encrypted
> 2) Each mapped disk is made part of a Btrfs RAID1 array
>
> When attempting to boot the system after pulling out (in BIOS or using
> the cable) the drive to simulate a complete disk failure, GRUB hangs,
> prompting for the LUKS password of the disappeared drive and
> (unsurprisingly) failing to open it.
>
> This prevents booting in a degraded LUKS encrypted, Btrfs RAID1 on Guix
> System.

I retested this today, and the problem still occurs. Here's a

screenshot from the failed boot (GRUB):

Attachment: IMG_20210807_004828_1.jpg

Ideally, GRUB (or is it our boot script?) should be smart enough to

realize that oh, that's Btrfs RAID1, it ought to work in degraded mode,

so let's keep going.

Thanks,

Maxim

Giovanni Biscuolo wrote on 11 Aug 2021 16:45

Recipients:

Message-ID:87mtpof3bk.fsf@xelera.eu

Hello Maxim,

Maxim Cournoyer <maxim.cournoyer@gmail.com> writes:

[...]

Toggle quote (10 lines)>> On a system where:
>>
>> 1) Each disks comprising the array is fully LUKS encrypted
>> 2) Each mapped disk is made part of a Btrfs RAID1 array
>>
>> When attempting to boot the system after pulling out (in BIOS or using
>> the cable) the drive to simulate a complete disk failure, GRUB hangs,
>> prompting for the LUKS password of the disappeared drive and
>> (unsurprisingly) failing to open it.

[...]

Toggle quote (2 lines)

> Ideally, GRUB (or is it our boot script?)

Since the end result is your system entered "grub rescue" mode AFAIU

it's a GRUB issue

Toggle quote (3 lines)

> should be smart enough to realize that oh, that's Btrfs RAID1, it

> ought to work in degraded mode, so let's keep going.

I (still) don't have a Guix System to test your setup and (try to) patch

thing up, so we need more info to debug the situation.

Can you please provide the output of the "ls" command and the "set"

command from the grub rescue shell?

Also, please what is your /proc/cmdline (when Linux correcly boots)?

Best regards, Gio

Giovanni Biscuolo

Xelera IT Infrastructures

Maxim Cournoyer wrote on 12 Aug 2021 04:25

Recipients:(name . Giovanni Biscuolo)(address . g@xelera.eu)(address . 40999@debbugs.gnu.org)

Message-ID:87fsvfz9eb.fsf@gmail.com

Hello Giovanni,

Giovanni Biscuolo <g@xelera.eu> writes:

Toggle quote (23 lines)> Hello Maxim,
>
> Maxim Cournoyer <maxim.cournoyer@gmail.com> writes:
>
> [...]
>
>>> On a system where:
>>>
>>> 1) Each disks comprising the array is fully LUKS encrypted
>>> 2) Each mapped disk is made part of a Btrfs RAID1 array
>>>
>>> When attempting to boot the system after pulling out (in BIOS or using
>>> the cable) the drive to simulate a complete disk failure, GRUB hangs,
>>> prompting for the LUKS password of the disappeared drive and
>>> (unsurprisingly) failing to open it.
>
> [...]
>
>> Ideally, GRUB (or is it our boot script?)
>
> Since the end result is your system entered "grub rescue" mode AFAIU
> it's a GRUB issue

Yeah, it looks like it. The grub.cfg file only has basic things in it,

nothing that could explain the failure.

Toggle quote (6 lines)

>> should be smart enough to realize that oh, that's Btrfs RAID1, it

>> ought to work in degraded mode, so let's keep going.

> I (still) don't have a Guix System to test your setup and (try to) patch

> thing up, so we need more info to debug the situation.

I believe the basic recipe to reproduce is there:

1. Partition two drives like so (GPT with 2MiB BIOS boot):

$ sudo sfdisk -l /dev/sda

Disk /dev/sda: 931.53 GiB, 1000204886016 bytes, 1953525168 sectors

Disk model: WDC WD1002FAEX-0

Units: sectors of 1 * 512 = 512 bytes

Sector size (logical/physical): 512 bytes / 512 bytes

I/O size (minimum/optimal): 512 bytes / 512 bytes

Disklabel type: gpt

Disk identifier: B5BB7BA4-23A3-4E7C-87BB-8339B02C5905

Device Start End Sectors Size Type

/dev/sda1 2048 6143 4096 2M BIOS boot

/dev/sda2 6144 1953523711 1953517568 931.5G Linux filesystem

$ sudo sfdisk -l /dev/sdb

Disk /dev/sdb: 931.53 GiB, 1000204886016 bytes, 1953525168 sectors

Disk model: WDC WD1002FAEX-0

Units: sectors of 1 * 512 = 512 bytes

Sector size (logical/physical): 512 bytes / 512 bytes

I/O size (minimum/optimal): 512 bytes / 512 bytes

Disklabel type: gpt

Disk identifier: 45C58C18-7B39-A745-B22F-6A2321FB1999

Device Start End Sectors Size Type

/dev/sdb1 2048 6143 4096 2M BIOS boot

/dev/sdb2 6144 1953523711 1953517568 931.5G Linux filesystem

2. LUKS encrypt the whole 2nd (main) partition of each drive.

3. Format the mapped devices as Btrfs RAID1.

4. Reconfigure a Guix system on top of that (see a config snippet below)

5. Disconnect one of the two drives and reboot.

6. Contemplate the failure to get past GRUB.

Toggle quote (3 lines)

> Can you please provide the output of the "ls" command and the "set"

> command from the grub rescue shell?

I'll post after rebooting.

Toggle quote (2 lines)

> Also, please what is your /proc/cmdline (when Linux correcly boots)?

Toggle snippet (7 lines)

BOOT_IMAGE=/@root/gnu/store/1c0dkkkv5vdnyp73gvcl9k1kym5jjm54-linux-libre-5.13.8/bzImage

--root=/dev/mapper/cryptroot

--system=/gnu/store/815481yf1kfacwgkh4aa11rlb3lm6gvi-system

--load=/gnu/store/815481yf1kfacwgkh4aa11rlb3lm6gvi-system/boot quiet

snd_hda_intel.dmic_detect=0 modprobe.blacklist=rtl8187

The system config relevant sections are:

Toggle snippet (57 lines)(operating-system
    (host-name "hurd")
    (timezone "America/Montreal")
    (keyboard-layout (keyboard-layout "dvorak"))
    (bootloader (bootloader-configuration
                 (bootloader grub-bootloader)
                 (target "/dev/sda")
                 (terminal-outputs '(console))
		 (keyboard-layout keyboard-layout)))
    (kernel-arguments '("quiet" "snd_hda_intel.dmic_detect=0"
                        "modprobe.blacklist=rtl8187"))
    (mapped-devices
     (list (mapped-device
            (source "/dev/sda2")
            (target "cryptroot")
            (type luks-device-mapping))
           (mapped-device
            (source "/dev/sdb2")
            (target "cryptroot-mirror")
            (type luks-device-mapping))
           (mapped-device
            (source "/dev/sdc2")
            (target "cryptroot-mirror2")
            (type luks-device-mapping))))

    ;; Note: Using any of the LUKS encrypted drives exposed under
    ;; /dev/mapper is enough to reference the Btrfs RAID-1 array,
    ;; since the 'btrfs device scan' command is executed in the init
    ;; RAM disk and takes care of assembling the array.
    (file-systems (cons* (file-system
                           (mount-point "/")
                           (device "/dev/mapper/cryptroot")
                           (type "btrfs")
			   (options (alist->file-system-options
                                     (cons '("subvol" . "@root")
                                           %common-btrfs-options)))
			   (dependencies mapped-devices))
                         (file-system
                           (device "/dev/mapper/cryptroot")
                           (mount-point "/home")
                           (type "btrfs")
                           (options (alist->file-system-options
                                     (cons '("subvol" . "@home")
                                           %common-btrfs-options)))
                           (dependencies mapped-devices))
                         (file-system
                           (device "/dev/mapper/cryptroot")
                           (mount-point "/data")
                           (type "btrfs")
                           (options (alist->file-system-options
                                     (cons '("subvol" . "@data")
                                           %common-btrfs-options)))
                           (dependencies mapped-devices))
                         %base-file-systems))
   [...]

Thanks,

Maxim

Giovanni Biscuolo wrote on 13 Aug 2021 17:05

Recipients:(name . Maxim Cournoyer)(address . maxim.cournoyer@gmail.com)(address . 40999@debbugs.gnu.org)

Message-ID:877dgpfkrb.fsf@xelera.eu

Hi Maxim,

I'd "debug" the issue trying to compare my Debian system config with

yours since I'm also using a BTRFS RAID1 filesystem on LUKS.

I've still not unplugged one of the two disks on mine to simulate a

drive failure, Soon™ I'd like to test this condition... but it's a

busy machine so I don't know when.

Maxim Cournoyer <maxim.cournoyer@gmail.com> writes:

[...]

Toggle quote (8 lines)

>>> Ideally, GRUB (or is it our boot script?)

>> Since the end result is your system entered "grub rescue" mode AFAIU

>> it's a GRUB issue

> Yeah, it looks like it. The grub.cfg file only has basic things in it,

> nothing that could explain the failure.

Please could you also provide the result of "lsblk -f"?

This is (part of) my disks layout:

Toggle snippet (22 lines)
sdc                                                                                  
??sdc1                                                                               
??sdc2 vfat                      F6D8-67E3                             470.8M     1% /boot/efi
??sdc3 crypto_L                  e554b806-19ac-48b2-b521-b4e89839a756                
? ??crypt_swap01
?      swap                      a43ce70c-dd35-47d8-a2ef-ef9d3c6d0885                [SWAP]
??sdc4 crypto_L                  820bfdf7-46f7-46f5-8536-7e1b0f04e70e                
  ??crypt_btrfs01_03
       btrfs    btrfs_pool01     82afe97a-bb97-4b3d-90cb-93a058185b97                
sdd                                                                                  
??sdd1                                                                               
??sdd2                                                                               
??sdd3 crypto_L                  960aa919-182b-4604-a8be-8477c86386cc                
? ??crypt_swap02
?      swap                      3f8f6974-05a9-4047-993a-c4ccb27eaa1d                [SWAP]
??sdd4 crypto_L                  c590c62e-6ac8-418c-9ea7-7ae9c79058c8                
  ??crypt_btrfs01_04
       btrfs    btrfs_pool01     82afe97a-bb97-4b3d-90cb-93a058185b97  802.3G    57% /mnt/btrfs

btrfs_pool01 is my BTRFS RAID1 filesystem, it includes /boot and /

(root) and is on two ancrypted LUKS partitions, as you can see.

Also, please what's your grub.cfg?

This is the config of a menuentry of mine:

Toggle snippet (27 lines)
menuentry 'Debian GNU/Linux' --class debian --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-simple-82afe97a-bb97-4b3d-90cb-93a058185b97' {
	load_video
	insmod gzio
	if [ x$grub_platform = xxen ]; then insmod xzio; insmod lzopio; fi
	insmod part_gpt
	insmod cryptodisk
	insmod luks
	insmod gcry_rijndael
	insmod gcry_rijndael
	insmod gcry_sha256
	insmod btrfs
	cryptomount -u c590c62e6ac8418c9ea77ae9c79058c8
	set root='cryptouuid/c590c62e6ac8418c9ea77ae9c79058c8'
	if [ x$feature_platform_search_hint = xy ]; then
	  search --no-floppy --fs-uuid --set=root --hint='cryptouuid/c590c62e6ac8418c9ea77ae9c79058c8'  82afe97a-bb97-4b3d-90cb-93a058185b97
	else
	  search --no-floppy --fs-uuid --set=root 82afe97a-bb97-4b3d-90cb-93a058185b97
	fi
	echo	'Loading Linux 5.10.0-0.bpo.3-amd64 ...'
	linux	/debian_root/boot/vmlinuz-5.10.0-0.bpo.3-amd64 root=UUID=82afe97a-bb97-4b3d-90cb-93a058185b97 ro rootflags=subvol=debian_root ip=10.38.2.2::10.38.2.1:255.255.255.0:anemone:eth0:none quiet
	echo	'Loading initial ramdisk ...'
	initrd	/debian_root/boot/initrd.img-5.10.0-0.bpo.3-amd64
}

AFAIU this code (from the snippet above):

Toggle snippet (9 lines)

if [ x$feature_platform_search_hint = xy ]; then

search --no-floppy --fs-uuid --set=root --hint='cryptouuid/c590c62e6ac8418c9ea77ae9c79058c8' 82afe97a-bb97-4b3d-90cb-93a058185b97

else

search --no-floppy --fs-uuid --set=root 82afe97a-bb97-4b3d-90cb-93a058185b97

sets [1] the root GRUB env variable to the first found device containing

the UUID 82afe97a-bb97-4b3d-90cb-93a058185b97, that is the UUID of my

BTRFS filesystem

AFAIU (but still not tested) this means that if the device with UUID

c590c62[...] is missing the search ensures that GRUB will find the next

device containing the BTRFS filesystem identified by UUID 82afe97a[...]

WDYT?

[1] https://www.gnu.org/software/grub/manual/grub/grub.html#search

[...]

Toggle quote (5 lines)

>> Can you please provide the output of the "ls" command and the "set"

>> command from the grub rescue shell?

> I'll post after rebooting.

OK thanks.

Toggle quote (10 lines)>> Also, please what is your /proc/cmdline (when Linux correcly boots)?
>
> --8<---------------cut here---------------start------------->8---
> BOOT_IMAGE=/@root/gnu/store/1c0dkkkv5vdnyp73gvcl9k1kym5jjm54-linux-libre-5.13.8/bzImage
> --root=/dev/mapper/cryptroot
> --system=/gnu/store/815481yf1kfacwgkh4aa11rlb3lm6gvi-system
> --load=/gnu/store/815481yf1kfacwgkh4aa11rlb3lm6gvi-system/boot quiet
> snd_hda_intel.dmic_detect=0 modprobe.blacklist=rtl8187
> --8<---------------cut here---------------end--------------->8---

This is mine (derived from the GRUB menu entry shown above):

Toggle snippet (5 lines)

BOOT_IMAGE=/debian_root/boot/vmlinuz-5.10.0-0.bpo.3-amd64 root=UUID=82afe97a-bb97-4b3d-90cb-93a058185b97 ro rootflags=subvol=debian_root ip=10.38.2.2::10.38.2.1:255.255.255.0:anemone:eth0:none quiet

AFAIU using "root=UUID=..." is more robust than using the (possibly

missing) device mapper path.

[...]

Hope this helps.

Best regards, Gio'

Giovanni Biscuolo

Xelera IT Infrastructures

Maxim Cournoyer wrote on 29 Aug 2021 08:15

Recipients:(name . Giovanni Biscuolo)(address . g@xelera.eu)(address . 40999@debbugs.gnu.org)

Message-ID:87wno4u3km.fsf@gmail.com

Hello Giovanni!

I've finally reboot the machine, so am I sharing the information

requested:

Giovanni Biscuolo <g@xelera.eu> writes:

Toggle quote (5 lines)

> Hi Maxim,

> I'd "debug" the issue trying to compare my Debian system config with

> yours since I'm also using a BTRFS RAID1 filesystem on LUKS.

Sounds useful!

[...]

Toggle quote (2 lines)

> Please could you also provide the result of "lsblk -f"?

NAME          FSTYPE    FSVER LABEL      UUID                                 FSAVAIL FSUSE% MOUNTPOINT
sda
sda1
sda2        crypto_LU                  0792432c-78d8-4dcc-87c5-30200c3d02db
  cryptroot btrfs           my-root    2e97fbbd-fa4e-4858-948b-b3a89278a39b  201.2G    77% /var/lib/dock
sdb
sdb1
sdb2        crypto_LU                  a9aead40-9d01-4f7a-bb83-be70dd192b7b
  cryptroot-mirror
              btrfs           my-root    2e97fbbd-fa4e-4858-948b-b3a89278a39b
sdc
sdc1
sdc2        crypto_LU                  f0afd5c9-da70-46a7-9c6f-5d22913638bf
  cryptroot-mirror2
              btrfs           my-root    2e97fbbd-fa4e-4858-948b-b3a89278a39b
sdd           crypto_LU                  f04928db-90aa-458c-8908-036a620b74f6
luks-f04928db-90aa-458c-8908-036a620b74f6
              btrfs           Seagate2TB 231e9e86-e841-4c97-81f1-013a2b8d99c2    1.6T    12% /media/maxim/
sr0
sr1
zram0         swap                       76423fb7-9d60-47fc-b64c-313f0a7b1f55                [SWAP]

Toggle snippet (32 lines)
The Btrfs file system in my case is labeled 'my-root' and composed of 3
drives in a raid1c3 btrfs array (3 copies).  @root is a subvolume on
which the root file system lives.

> This is (part of) my disks layout:
>
>
> sdc
> ..sdc1
> ..sdc2 vfat                      F6D8-67E3                             470.8M     1% /boot/efi
> ..sdc3 crypto_L                  e554b806-19ac-48b2-b521-b4e89839a756
> . ..crypt_swap01
> .      swap                      a43ce70c-dd35-47d8-a2ef-ef9d3c6d0885                [SWAP]
> ..sdc4 crypto_L                  820bfdf7-46f7-46f5-8536-7e1b0f04e70e
>   ..crypt_btrfs01_03
>        btrfs    btrfs_pool01     82afe97a-bb97-4b3d-90cb-93a058185b97
> sdd
> ..sdd1
> ..sdd2
> ..sdd3 crypto_L                  960aa919-182b-4604-a8be-8477c86386cc
> . ..crypt_swap02
> .      swap                      3f8f6974-05a9-4047-993a-c4ccb27eaa1d                [SWAP]
> ..sdd4 crypto_L                  c590c62e-6ac8-418c-9ea7-7ae9c79058c8
>   ..crypt_btrfs01_04
>        btrfs    btrfs_pool01     82afe97a-bb97-4b3d-90cb-93a058185b97  802.3G    57% /mnt/btrfs
>
>
> btrfs_pool01 is my BTRFS RAID1 filesystem, it includes /boot and /
> (root) and is on two ancrypted LUKS partitions, as you can see.

Toggle quote (2 lines)

> Also, please what's your grub.cfg?

Here it is:

Toggle snippet (50 lines)

# This file was generated from your Guix configuration. Any changes

# will be lost upon reconfiguration.

# Set 'root' to the partition that contains /gnu/store.

search --file --set /@root/gnu/store/wlf9ccsl9pmch1dyv5x8c2gdngwn9m5i-grub-image.png

terminal_output console

insmod png

if background_image /@root/gnu/store/wlf9ccsl9pmch1dyv5x8c2gdngwn9m5i-grub-image.png; then

set color_normal=light-gray/black

set color_highlight=yellow/black

else

set menu_color_normal=cyan/blue

set menu_color_highlight=white/blue

# Localization configuration.

# search --file --set /@root/gnu/store/q1cf63j2az4wlajg0caqy4nbndp0mvpm-grub-locales/en@quot.mo

set locale_dir=/@root/gnu/store/q1cf63j2az4wlajg0caqy4nbndp0mvpm-grub-locales

set lang=en_US

insmod keylayouts

keymap /@root/gnu/store/25s8pbpv2fnidrgir26mn97g0ciq52gz-grub-keymap.dvorak

set default=0

set timeout=5

menuentry "GNU with Linux-Libre 5.13.12" {

search --file --set /@root/gnu/store/hvmyb8maz32dy6ra5g68gr4wd08pzq3r-linux-libre-5.13.12/bzImage

linux /@root/gnu/store/hvmyb8maz32dy6ra5g68gr4wd08pzq3r-linux-libre-5.13.12/bzImage --root=/dev/mapper/cryptroot --system=/gnu/store/6qa5ga0pkjbmz8ix8gfrpy65zkl16xi7-system --load=/gnu/store/6qa5ga0pkjbmz8ix8gfrpy65zkl16xi7-system/boot quiet snd_hda_intel.dmic_detect=0 modprobe.blacklist=rtl8187

initrd /@root/gnu/store/kllyldndnazfxxrhkabgifx5zvgyz82q-raw-initrd/initrd.cpio.gz

}

submenu "GNU system, old configurations..." {

menuentry "GNU with Linux-Libre 5.13.11 (#275, 2021-08-23 23:17)" {

search --file --set /@root/gnu/store/fznnj7bgs46czizzhn186606jgr52qnp-linux-libre-5.13.11/bzImage

linux /@root/gnu/store/fznnj7bgs46czizzhn186606jgr52qnp-linux-libre-5.13.11/bzImage --root=/dev/mapper/cryptroot --system=/var/guix/profiles/system-275-link --load=/var/guix/profiles/system-275-link/boot quiet snd_hda_intel.dmic_detect=0 modprobe.blacklist=rtl8187

initrd /@root/gnu/store/g73vj8qy6kfrgmr8gnmmzh2q59cbnf2w-raw-initrd/initrd.cpio.gz

}

[...]

if [ "${grub_platform}" == efi ]; then

menuentry "Firmware setup" {

fwsetup

}

Toggle quote (48 lines)> This is the config of a menuentry of mine:
>
>
> menuentry 'Debian GNU/Linux' --class debian --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-simple-82afe97a-bb97-4b3d-90cb-93a058185b97' {
> 	load_video
> 	insmod gzio
> 	if [ x$grub_platform = xxen ]; then insmod xzio; insmod lzopio; fi
> 	insmod part_gpt
> 	insmod cryptodisk
> 	insmod luks
> 	insmod gcry_rijndael
> 	insmod gcry_rijndael
> 	insmod gcry_sha256
> 	insmod btrfs
> 	cryptomount -u c590c62e6ac8418c9ea77ae9c79058c8
> 	set root='cryptouuid/c590c62e6ac8418c9ea77ae9c79058c8'
> 	if [ x$feature_platform_search_hint = xy ]; then
> 	  search --no-floppy --fs-uuid --set=root --hint='cryptouuid/c590c62e6ac8418c9ea77ae9c79058c8'  82afe97a-bb97-4b3d-90cb-93a058185b97
> 	else
> 	  search --no-floppy --fs-uuid --set=root 82afe97a-bb97-4b3d-90cb-93a058185b97
> 	fi
> 	echo	'Loading Linux 5.10.0-0.bpo.3-amd64 ...'
> 	linux	/debian_root/boot/vmlinuz-5.10.0-0.bpo.3-amd64 root=UUID=82afe97a-bb97-4b3d-90cb-93a058185b97 ro rootflags=subvol=debian_root ip=10.38.2.2::10.38.2.1:255.255.255.0:anemone:eth0:none quiet
> 	echo	'Loading initial ramdisk ...'
> 	initrd	/debian_root/boot/initrd.img-5.10.0-0.bpo.3-amd64
> }
>
>
> AFAIU this code (from the snippet above):
>
>
> 	if [ x$feature_platform_search_hint = xy ]; then
> 	  search --no-floppy --fs-uuid --set=root --hint='cryptouuid/c590c62e6ac8418c9ea77ae9c79058c8'  82afe97a-bb97-4b3d-90cb-93a058185b97
> 	else
> 	  search --no-floppy --fs-uuid --set=root 82afe97a-bb97-4b3d-90cb-93a058185b97
> 	fi
>
>
> sets [1] the root GRUB env variable to the first found device containing
> the UUID 82afe97a-bb97-4b3d-90cb-93a058185b97, that is the UUID of my
> BTRFS filesystem
>
> AFAIU (but still not tested) this means that if the device with UUID
> c590c62[...] is missing the search ensures that GRUB will find the next
> device containing the BTRFS filesystem identified by UUID 82afe97a[...]
>
> WDYT?

[...]

Toggle quote (3 lines)

>>> Can you please provide the output of the "ls" command and the "set"

>>> command from the grub rescue shell?

See the attached screenshot of the result:

Attachment: IMG_20210829_023531.jpg

I was about to mess around in GRUB to edit the prefix, cmdline and
root values and do `insmod normal`, `normal` to proceed to boot, but
then the init RAM disk failed like so:

Attachment: IMG_20210829_024344.jpg

So there are more than one things to be adjusted :-).

Thank you, I'll look at the data with a fresh head later, but it seems

to me that we'd need to have GRUB fallback logic for the root devices

when RAID setups are detected. I'll read on what GRUB has in store for

this kind of thing when I have a chance.

Your Debian GRUB config also has me wondering about the 'btrfs' modules,

and the others than Guix System is not using.

Thank you!

Maxim

Maxim Cournoyer wrote on 5 Mar 2022 04:33

Recipients:(name . Giovanni Biscuolo)(address . g@xelera.eu)(address . 40999@debbugs.gnu.org)

Message-ID:87ilst13aa.fsf@gmail.com

Hi,

I'm writing here because I just found a much easier way to trigger this

than by opening the case of my desktop and pulling a drive out with this

QEMU script:

Toggle snippet (14 lines)

#!/usr/bin/env bash

devices=(sda sdb sdc)

args=(-enable-kvm -snapshot -m 2G)

i=0

for d in "${devices[@]}"; do

args+=(-drive file=/dev/$d,index=$i,media=disk)

let i++

done

qemu-system-x86_64 "${args[@]}" "$@"

This attempts to boot the drives of the *live* system in QEMU; don't

fret, it's not dangerous as the '-snapshot' option ensure no actual

writes reach the drives. It seems to fail at the mount command in our

initrd, but it at least allow testing GRUB easily.

With the above script and my Btrfs RAIDc3 array on drives /dev/sda,

/dev/sdb and /dev/sdc, after removing 'sdb' from the devices list for

example I get:

Toggle snippet (24 lines)Booting from Hard Disk...
GRUB loading...
Welcome to GRUB!

Attempting to decrypt master key...
Enter passphrase for hd0,gpt2 (0792432c78d84dcc87c530200c3d02db):
Slot 0 opened
error: failure reading sector 0x0 from `fd0'.
error: no such cryptodisk found.
Attempting to decrypt master key...
Enter passphrase for hd1,gpt2 (f0afd5c9da7046a79c6f5d22913638bf):
Slot 0 opened
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.
error: failure reading sector 0x80 from `fd0'.

Dropping just sdc instead, I get:

Toggle snippet (17 lines)Booting from Hard Disk...
GRUB loading...
Welcome to GRUB!

Attempting to decrypt master key...
Enter passphrase for hd0,gpt2 (0792432c78d84dcc87c530200c3d02db): 
Slot 0 opened
Attempting to decrypt master key...
Enter passphrase for hd1,gpt2 (a9aead409d014f7abb83be70dd192b7b): 
Slot 0 opened
error: failure reading sector 0x0 from `fd0'.
error: no such cryptodisk found.
error: failure reading sector 0x80 from `fd0'.
error: unknown filesystem.
Entering rescue mode...

This should make a future fix cheaper to try (but a system test will be

best anyway :-)).

Thanks,

Maxim

Maxim Cournoyer wrote on 27 Mar 2022 06:07

Recipients:(address . 40999@debbugs.gnu.org)

Message-ID:878rswvy0p.fsf@gmail.com

Hi,

maxim.cournoyer@gmail.com writes:

Toggle quote (13 lines)> On a system where:
>
> 1) Each disks comprising the array is fully LUKS encrypted
> 2) Each mapped disk is made part of a Btrfs RAID1 array
>
> When attempting to boot the system after pulling out (in BIOS or using
> the cable) the drive to simulate a complete disk failure, GRUB hangs,
> prompting for the LUKS password of the disappeared drive and
> (unsurprisingly) failing to open it.
>
> This prevents booting in a degraded LUKS encrypted, Btrfs RAID1 on Guix
> System.

It seems this is a problem not unknown to other (non-Btrfs) software
RAID as well, such as mdadm.  There was recently a fix for it in Ubuntu
[0].  It can probably provide cues about how to go to fix it in Guix
System.

[0]  https://bugs.launchpad.net/ubuntu/+source/cryptsetup/+bug/1879980

Your comment

Commenting via the web interface is currently disabled.

To comment on this conversation send an email to 40999@debbugs.gnu.org

To respond to this issue using the mumi CLI, first switch to it

mumi current 40999

Then, you may apply the latest patchset in this issue (with sign off)

mumi am -- -s

Or, compose a reply to this issue

mumi compose

Or, send patches to this issue

mumi send-email *.patch

is:open	open issues
is:done	closed issues
submitter:<who>	search issue submitter
author:<who>	search by message author
date:yesterday..now	search by issue date
mdate:3m..2d	search by message date