41_snapshots-btrfs aborts on LUKS-on-Btrfs with detached header & external /boot (grub-probe: cannot get cryptodisk …)

## Environment

| Item | Value |
|------|-------|
| **Distro** | Arch Linux |
| **grub-btrfs** | 4.13-yabsnap_info_support-2024-03-06 |
| **GRUB** | grub-improved-luks2-git 2.12.r303.g86e8f2c4b-1 |
| **btrfs-progs** | 6.15|
| **Snapper** | 0.12.2|
| **Layout** | single encrypted container on SSD, no partitions, formatted with Btrfs → LUKS (header on USB) → <br>same USB stick holds **/boot** (vfat) |
| **Devices** | `/dev/disk/by-id/<REDACTED>` → `/dev/mapper/enc` (UUID =`REDACTED`)<br>`/dev/sdb1` (USB, UUID =`REDACTED`) → `/boot` & header |
```
[user@hostname ~]$ lsblk
NAME        MAJ:MIN RM   SIZE RO TYPE  MOUNTPOINTS
sda           8:0    0 953.9G  0 disk
└─enc       254:0    0 953.9G  0 crypt /.snapshots
                                       /var
                                       /home
                                       /pkg
                                       /swap
                                       /
sdb           8:16   1  29.1G  0 disk
└─sdb1        8:17   1     1G  0 part  /boot
```
## Symptom
Running 

```bash
sudo grub-mkconfig -o /boot/grub/grub.cfg
```
terminates during execution of `41_snapshots-btrfs` with only
```bash
Detecting snapshots ...
```
When `set -x` tracing is enabled the last command executed is

```bash
root_uuid=$(${grub_probe} --device ${root_device} --target=fs_uuid)
```
which fails with 
```bash
/usr/bin/grub-probe: error: cannot get cryptodisk from source disk `hostdisk//dev/sda'.
Internal error: Unreleased memory pool(s) found.
```
No `grub-btrfs.cfg` is produced, so the snapshot submenu never appears

## Why `grub-probe --target=fs_uuid` can fail

I would first like to explain why `grub-probe --target=fs_uuid` can fail in this scenario 

- `grub-probe --target=fs_uuid` tries to parse the LUKS header in-place on the disk.

- When it finds the disk it expects to find a LUKS header in the
first sector (inline header).
With a detached header that sector is just random data, so parsing fails
and `grub-probe` exits.

## My solution to `grub-probe --target=fs_uuid` failing

```bash
# 1. Try the original, fast path first
root_uuid=$(${grub_probe} --device "${root_device}" --target=fs_uuid 2>/dev/null) || true
#          └──────────────────┬───────────────────────────────────────────────┘
#               may fail → we ignore the exit code with `|| true`

# 2. Fallback 1: blkid
[ -z "$root_uuid" ] && root_uuid=$(blkid -s UUID -o value "${root_device}")

# 3. Fallback 2: lsblk   (rarely needed, but covers busybox systems without blkid)
[ -z "$root_uuid" ] && root_uuid=$(lsblk -no UUID "${root_device}")

# 4. Hard stop if we *still* have nothing
[ -z "$root_uuid" ] && print_error "Cannot determine UUID of ${root_device}"

```
Same block is applied to `boot_uuid`.
This does not change the underlying logic so script will continue on both inline-header and detached-header systems.

## What this achieves

- Normal systems (inline header) still use GRUB’s own probe 

- Detached-header or exotic file-system configurations fall back to kernel knowledge 

- Script exits with a error message only if all methods fail

## Continuing the debugging.

After applying the above patch I re-ran `41_snapshots-btrfs` with the `set -x` flag to continue monitoring execution tracing.
I was then caught up on this command 
```bash
list_insmods+=("cryptomount -u $(echo $GRUB_CMDLINE_LINUX_DEFAULT | grep -o -P '(?<=cryptdevice=UUID=).*(?=:cryptdev)')")
```
It was failing to `grep` the UUID from `GRUB_CMDLINE_LINUX_DEFAULT` string in `/etc/default/grub` because it didnt exist there. `GRUB_CMDLINE_LINUX_DEFAULT` on my system is only `loglevel=3 quiet ` so it has no `cryptdevice=` string. 
My `cryptdevice=` string is part of `GRUB_CMDLINE_LINUX`.

also with a detached header Arch recommends:
```bash
cryptdevice=/dev/disk/by-id/<SSD>:enc:header
```
so `grep '(?<=cryptdevice=UUID=)…'` would return nothing anyways and grep exits so script
dies again.

You may see why this caused incompatibility with the script so the following patch block was made.

## Cryptomount patch block and changes

The following code block snippet
```bash
# Enable LUKS encrypted devices support
case "$(echo "$GRUB_BTRFS_ENABLE_CRYPTODISK" | tr '[:upper:]' '[:lower:]')" in
    true)
            list_insmods=()
            list_insmods+=("insmod gzio")
            list_insmods+=("insmod part_gpt")
            list_insmods+=("insmod cryptodisk")
            list_insmods+=("insmod luks")
            list_insmods+=("insmod gcry_rijndael")
            list_insmods+=("insmod gcry_rijndael")
            list_insmods+=("insmod gcry_sha256")
            list_insmods+=("insmod ${boot_fs}")
            list_insmods+=("cryptomount -u $(echo $GRUB_CMDLINE_LINUX_DEFAULT | grep -o -P '(?<=cryptdevice=UUID=).*(?=:cryptdev)')")
            ;;
    *)
            list_insmods=("insmod ${boot_fs}")
            ;;
esac
```
is replaced with
```bash
## Enable LUKS encrypted devices support
case "$(echo "$GRUB_BTRFS_ENABLE_CRYPTODISK" | tr '[:upper:]' '[:lower:]')" in
    true)
        list_insmods=(
            "insmod gzio"
            "insmod part_gpt"
            "insmod cryptodisk"
            "insmod luks"
            "insmod gcry_rijndael"
            "insmod gcry_rijndael"
            "insmod gcry_sha256"
            "insmod ${boot_fs}"
        )

        #  Extract the <source> field of cryptdevice=<source>:<name>[:header]
        crypt_source="$(printf '%s %s\n' "$GRUB_CMDLINE_LINUX_DEFAULT" "$GRUB_CMDLINE_LINUX" \
                          | grep -o -P 'cryptdevice=\K[^:]+' || true)"

        # Turn the source into a UUID that cryptomount -u understands
        crypt_uuid=""
        if [[ "$crypt_source" =~ ^UUID=.* ]]; then            # already UUID=…
            crypt_uuid="${crypt_source#UUID=}"
        elif [[ "$crypt_source" == /dev/* ]]; then            # path → resolve → blkid
            real_dev=$(readlink -f "$crypt_source" 2>/dev/null || true)
            [ -b "$real_dev" ] && crypt_uuid=$(blkid -s UUID -o value "$real_dev" 2>/dev/null || true)
        fi

        # Exec the proper cryptomount command
        if [[ "$crypt_uuid" =~ ^[0-9a-fA-F-]{36}$ ]]; then
            list_insmods+=("cryptomount -u ${crypt_uuid}")
        else
            # last-resort: scan all crypto containers (works but a bit slower)
            list_insmods+=("cryptomount -a")
        fi
        ;;
    *)
        list_insmods=("insmod ${boot_fs}")
        ;;
esac
```

This section -
```bash
        crypt_source="$(printf '%s %s\n' "$GRUB_CMDLINE_LINUX_DEFAULT" \
                                          "$GRUB_CMDLINE_LINUX" \
                        | grep -o -P 'cryptdevice=\K[^:]+' || true)"
```

- Extracts the `<source>` part of `cryptdevice=`

- Combines both GRUB cmdline variables (`DEFAULT` and `LINUX`).

- `grep -o -P 'cryptdevice=\K[^:]+'` keeps everything between
`cryptdevice=` and the next `:`.
Matches both forms:

   - `cryptdevice=UUID=1111-2222-…:` → returns `UUID=1111-2222-…`

   - `cryptdevice=/dev/disk/by-id/nvme-SN123:` → returns `/dev/disk/by-id/…`

- `|| true` prevents `set -e` from killing the script if no match is found.

For this snippet -
```bash
        crypt_uuid=""
        if [[ "$crypt_source" =~ ^UUID=.* ]]; then
            crypt_uuid="${crypt_source#UUID=}"       # strip leading “UUID=”
        elif [[ "$crypt_source" == /dev/* ]]; then
            real_dev=$(readlink -f "$crypt_source" 2>/dev/null || true)  # resolve symlinks
            [ -b "$real_dev" ] &&                               # ensure block device
              crypt_uuid=$(blkid -s UUID -o value "$real_dev" 2>/dev/null || true)
        fi
```
We try to extract a usable UUID that grub can understand.

For instance,
1. We resolved a `UUID=…` so we just remove the prefix.

2. we've resolved a device path (`/dev/disk/by-id/…`):
check symlinks and ask `blkid` for its file-system UUID..

3. If neither yields a UUID, `crypt_uuid` remains empty.

Next we have -
```bash
        if [[ "$crypt_uuid" =~ ^[0-9a-fA-F-]{36}$ ]]; then
            list_insmods+=("cryptomount -u ${crypt_uuid}")  # fast, direct
        else
            list_insmods+=("cryptomount -a")                # fallback autoscan
        fi
```
Here we append the correct `cryptomount` command depending on if we found a UUID or not.

- `cryptomount -u <uuid> ` if we found a UUID
or
- `cryptomount -a` if we didn't.

`cryptomount -a` scans every disk for LUKS headers. This is kinda like a last resort 

## Outcome
Boom
```bash
[user@hostname ~]$ sudo grub-mkconfig -o /boot/grub/grub.cfg
Generating grub configuration file ...
Found linux image: /boot/vmlinuz-linux-hardened
Found initrd image: /boot/initramfs-linux-hardened.img
Found fallback initrd image(s) in /boot: initramfs-linux-hardened-fallback.img
Warning: os-prober will be executed to detect other bootable partitions.
Its output will be used to detect bootable binaries on them and create new boot entries.
Found Windows Boot Manager on /dev/nvme0n1p1@/efi/Microsoft/Boot/bootmgfw.efi
Adding boot menu entry for UEFI Firmware Settings ...
Detecting snapshots ...
/usr/bin/grub-probe: error: cannot get cryptodisk from source disk `hostdisk//dev/sda'.
You have a memory leak (not released memory pool):
 [0x78b0c2866b0] dtree
Internal error: Unreleased memory pool(s) found.
Found snapshot: 2025-07-08 03:00:41 | @snapshots/17/snapshot                          | single   | timeline                                      |
Found snapshot: 2025-07-08 02:00:01 | @snapshots/16/snapshot                          | single   | timeline                                      |
Found snapshot: 2025-07-08 01:00:46 | @snapshots/15/snapshot                          | single   | timeline                                      |
Found snapshot: 2025-07-08 00:00:57 | @snapshots/14/snapshot                          | single   | timeline                                      |
Found snapshot: 2025-07-07 23:00:27 | @snapshots/13/snapshot                          | single   | timeline                                      |
Found snapshot: 2025-07-07 22:00:27 | @snapshots/12/snapshot                          | single   | timeline                                      |
Found snapshot: 2025-07-07 21:50:35 | @snapshots/11/snapshot                          | single   | timeline                                      |
Found snapshot: 2025-07-07 16:00:01 | @snapshots/10/snapshot                          | single   | timeline                                      |
Found snapshot: 2025-07-07 15:00:26 | @snapshots/9/snapshot                           | single   | timeline                                      |
Found snapshot: 2025-07-07 14:00:04 | @snapshots/8/snapshot                           | single   | timeline                                      |
Found snapshot: 2025-07-07 13:00:15 | @snapshots/7/snapshot                           | single   | timeline                                      |
Found snapshot: 2025-07-07 12:00:37 | @snapshots/6/snapshot                           | single   | timeline                                      |
Found snapshot: 2025-07-07 11:00:07 | @snapshots/5/snapshot                           | single   | timeline                                      |
Found snapshot: 2025-07-07 10:00:08 | @snapshots/4/snapshot                           | single   | timeline                                      |
Found snapshot: 2025-07-07 09:33:14 | timeshift-btrfs/snapshots/2025-07-07_09-33-14/@ | ondemand | {timeshift-autosnap} {created before upgrade} |
Found snapshot: 2025-07-07 09:12:08 | @snapshots/3/snapshot                           | single   | test snapshot 2                               |
Found snapshot: 2025-07-07 09:00:10 | @snapshots/2/snapshot                           | single   | timeline                                      |
Found snapshot: 2025-07-07 08:59:43 | @snapshots/1/snapshot                           | single   | test snapshot                                 |
Found 18 snapshot(s)
Unmount /tmp/grub-btrfs.tAroDMMCMK .. Success
done
```
Menu generation succeeds, 18 snapshot entries appear.
grub-btrfs.cfg is created and passed script checks.
Im sure you've noticed there's that same grub-probe error from before but its no longer an issue because we extract the UUID using blkid  in my scenario, for normal users with inline-header that command wont fail the check and that error wont appear.

I've been struggling with this issue for a couple years now and finally learned enough to remedy the problem. I'll be making a pull request soon but knowing it might not get merged soon or even at all I've decided to make this issue to hopefully help the next person or somebody that's facing a similar issue. 

[41_snapshots-btrfs.txt](https://github.com/user-attachments/files/21123285/41_snapshots-btrfs.txt)

Item	Value
Distro	Arch Linux
grub-btrfs	4.13-yabsnap_info_support-2024-03-06
GRUB	grub-improved-luks2-git 2.12.r303.g86e8f2c4b-1
btrfs-progs	6.15
Snapper	0.12.2
Layout	single encrypted container on SSD, no partitions, formatted with Btrfs → LUKS (header on USB) → same USB stick holds /boot (vfat)
Devices	`/dev/disk/by-id/<REDACTED>` → `/dev/mapper/enc` (UUID =`REDACTED`) `/dev/sdb1` (USB, UUID =`REDACTED`) → `/boot` & header

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

41_snapshots-btrfs aborts on LUKS-on-Btrfs with detached header & external /boot (grub-probe: cannot get cryptodisk …) #377

Environment

Symptom

Why `grub-probe --target=fs_uuid` can fail

My solution to `grub-probe --target=fs_uuid` failing

What this achieves

Continuing the debugging.

Cryptomount patch block and changes

Outcome

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Uh oh!

41_snapshots-btrfs aborts on LUKS-on-Btrfs with detached header & external /boot (grub-probe: cannot get cryptodisk …) #377

Description

Environment

Symptom

Why grub-probe --target=fs_uuid can fail

My solution to grub-probe --target=fs_uuid failing

What this achieves

Continuing the debugging.

Cryptomount patch block and changes

Outcome

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions

Why `grub-probe --target=fs_uuid` can fail

My solution to `grub-probe --target=fs_uuid` failing