You are not logged in.
Pretext: I have done several Arch installations now with a systemd-boot installation, XBOOTLDR partition and discoverable partitions Discoverable Partitions .
There have been no issues until my new PC, a ASUS B660Plus Wifi D4. I can setup the machine as usual (it's a double-boot with Windows 11), and after installation, I get the systemd-boot menu. Except that the Arch entry is missing. I can boot to Windows and to the BIOS setup (automatically generated entries). I have tried several things recommended in the Wiki, like disabling Fast Boot, also disabling Intel RST/VMD to make sure that the disks (3x NVMe) are passed through without getting captured by a RAID controller.
Whatever I do, I can't get user entries for systemd-boot to show up, only the autogenerated ones. From a rescue system with `gdisk`:
Number Start (sector) End (sector) Size Code Name
1 256 25855 100.0 MiB EF00 EFI system partition
2 25856 29951 16.0 MiB 0C01 Microsoft reserved ...
3 29952 25402111 96.8 GiB 0700 Basic data partition
4 25402112 25599999 773.0 MiB 2700
5 25600000 25676799 300.0 MiB EA00 XBOOTLDR partition
6 25676800 27571199 7.2 GiB 8200 Linux swap
7 27571200 34124799 25.0 GiB 8304 Linux x86-64 root (/)
8 34124800 47231999 50.0 GiB 8302 Linux /home
9 47232000 53785599 25.0 GiB 8300 Linux filesystem
10 53785600 60339199 25.0 GiB 8300 Linux filesystem
11 60339200 488378623 1.6 TiB 8306 Linux /srv
codes for the partitions
nvme0n1
|-nvme0n1p1 vfat FAT12 4EB0-AF32 68.7M 31% /mnt/install/efi
|-nvme0n1p2
|-nvme0n1p3 ntfs Win11OS 641843BC18438BCC
|-nvme0n1p4 ntfs DC76F47776F453AC
|-nvme0n1p5 vfat FAT32 C734-993C 227.8M 24% /mnt/install/boot
|-nvme0n1p6 swap 1 SWAP 21ad2ad6-5181-4b78-b813-19a4e953269b [SWAP]
|-nvme0n1p7 ext4 1.0 19f696d5-ba9a-4acc-8f48-f8cff2073d83 18.7G 18% /mnt/install
|-nvme0n1p8 ext4 1.0 51ba5e58-2371-4ec8-847c-593ad9434783 46.4G 0% /mnt/install/home
|-nvme0n1p9
|-nvme0n1p10
`-nvme0n1p11 ext4 1.0 ef54b389-59ad-45c2-89e7-33b3791898d4 1.5T 0% /mnt/install/srv
filesystems on the partitions
boot
|-- initramfs-linux.img
|-- intel-ucode.img
|-- loader
| |-- entries
| | |-- archlinux-core-main.conf
| | `-- memtest.conf
| |-- entries.srel
| |-- loader.conf
| `-- random-seed
|-- memtest86+
| `-- memtest.efi
`-- vmlinuz-linux
structure of `/boot`
[root@archboot /mnt/install]# tree -L 3 efi
efi
|-- EFI
| |-- Boot
| | `-- bootx64.efi
| |-- Microsoft
| | |-- Boot
| | `-- Recovery
| `-- systemd
| `-- systemd-bootx64.efi
`-- shellx64.efi
7 directories, 3 files
structure of `/efi`
[root@archboot /mnt/install]# cat boot/loader/entries/archlinux-core-main.conf
title Arch Linux
linux /vmlinuz-linux
initrd /intel-ucode.img
initrd /initramfs-linux.img
options quiet loglevel=3 consoleblank=300
contents of `archlinux-core-main.conf` (to boot Arch)
[root@archboot /mnt/install]# cat boot/loader/loader.conf
default archlinux-core-main
timeout 5
console-mode keep
#default @saved
contents of `loader.conf` (systemd-boot loader configuration)
Please let me know what information is missing.
I also tried once to give up on the Discoverable Partitions and boot Arch with systemd-boot and the UUID in the options line (`options root=UUID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx rw`), but this didn't work either.
Any ideas?
Offline
although I'm no expert in systemd-boot as I use grub when tested in a vm I used ext4 for the xbootldr partition and had to add the driver from the efifs package
but as uefi is meant to support fat and there's no fat fs driver in efifs I guess an issue could be the xbootldr partition is skipped
can you find the xbootldr partition by calling map from with the efi shell? if it doesn't show up it's not loaded properly
also: how EXACTLY do you install bootctl?
Offline
After I mounted all partitions to their respective places, I arch-chroot into the system and then use
bootctl --esp-path=/efi --boot-path=/boot install
.
XBOOTLDR is supposed by the spec to have vfat, for the same reasons as the EFI partition (needs to be accessible). This is why it's vfat and therefore needs no additional drivers. EFI partition is also accessible w/o additional drivers.
I try the map command later and edit this then.
Offline
Agreed with the above. Why are you using FAT12 ? Chances are mainboard vendors (i.e. ASUS) don't explicitly support this and without a FAT driver in efifs, if the UEFI says it doesn't know this filesystem because they limit themselves to FAT32 you're not going to find the partition.
Offline
Agreed with the above. Why are you using FAT12 ? Chances are mainboard vendors (i.e. ASUS) don't explicitly support this and without a FAT driver in efifs, if the UEFI says it doesn't know this filesystem because they limit themselves to FAT32 you're not going to find the partition.
I think this came from the Windows setup. I definitely didn't set this on purpose. Anyway, the system boots to MSWindows. The UEFI should know about the format then, otherwise Windows wouldn't boot from this partition if the UEFI couldn't find it at all.
But I can back up the contents and reformat / restore; maybe worth a try.
Offline
@v1del
the current fat12 with the size of just 100mb is what the windows installer creates unless you do it manual with diskpart and bcdboot
however: a point often came up lately it seems many oem design and test thier firmware windows only and due to untested bugs breaks tge moment something different shows up - so I see potential issues:
- the uefi just skips all other partitions because it already found its ESP
- similar to the above: the uefi can't handle more than one fat partition correctly and expects the esp to be the first
in both cases using a different filesystem for xbootldr partition and a driver from efifs package might could do the trick
I know this is not how it's supposed to work - but with windows still holding more than 80% marketshare there's no incentive for any of the big oem to verify thier firmwares with linux
if you want that level of support you have to move from consumer hardware to those targeted to professionals and business like asrock rack or supermicro or similar
Offline
Since it was requested, here the mappings from the UEFI shell:
Mapping table
FS0: Alias(s):HD0x0b0b:;BLK1:
PciRoot(0x0)/Pci(0x14,0x0)/USB(0x17,0x0)/USB(0x1,0x0)/HD(1,GPT,52FE17B9-E327-1199-F10F-3459C020493B,0x800,0x5B3028F)
FS1: Alias(s):HD0x0b0c:;BLK2:
PciRoot(0x0)/Pci(0x14,0x0)/USB(0x17,0x0)/USB(0x1,0x0)/HD(2,GPT,C4678A72-7419-B328-F346-EB2B55769E66,0x5B30A8F,0x10000)
FS4: Alias(s):HD3b:;BLK10:
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(1,GPT,48DC55E3-95F8-4463-B690-55030B5E2027,0x100,0x6400)
FS5: Alias(s):HD3d:;BLK14:
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(3,GPT,2D2431A2-1029-4F83-9C1E-6D1CBE016B63,0x7500,0x1832600)
FS6: Alias(s):HD3e:;BLK15:
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(4,GPT,7B3B536E-3E3E-4188-A007-A4470CD340D6,0x1839B00,0x30500)
FS7: Alias(s):HD3f:;BLK16:
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(5,GPT,861F195D-EA20-4B6C-A73B-A84F53DB07C8,0x186A000,0x12C00)
FS2: Alias(s):HD1c:;BLK6:
PciRoot(0x0)/Pci(0x1A,0x0)/Pci(0x0,0x0)/NVMe(0x1,B8-77-91-31-53-38-25-00)/HD(2,GPT,8C191AC8-FC76-4394-BB18-F47B78547AFC,0x8000,0xE8E00000)
FS3: Alias(s):HD2b:;BLK8:
PciRoot(0x0)/Pci(0x1D,0x0)/Pci(0x0,0x0)/NVMe(0x1,45-D8-DB-2A-01-75-A0-00)/HD(1,GPT,D38CCCAE-E7D5-46A0-8584-CD7DE407722D,0x800,0xE8E08000)
BLK0: Alias(s):
PciRoot(0x0)/Pci(0x14,0x0)/USB(0x17,0x0)/USB(0x1,0x0)
BLK3: Alias(s):
PciRoot(0x0)/Pci(0x14,0x0)/USB(0x2,0x0)
BLK9: Alias(s):
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)
BLK13: Alias(s):
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(2,GPT,41D09BE1-397A-4493-A357-B73CF986DC94,0x6500,0x1000)
BLK17: Alias(s):
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(6,GPT,036FC676-8762-4FC3-AFF5-40EF60BBCBB3,0x187CC00,0x1CE800)
BLK18: Alias(s):
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(7,GPT,D4AB315B-2AC8-44BA-B5A0-515F0CBF8DA6,0x1A4B400,0x640000)
BLK19: Alias(s):
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(8,GPT,4C98BD69-BBC1-43DC-AE65-3D2FA1972CB0,0x208B400,0xC80000)
BLK20: Alias(s):
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(9,GPT,7CE19987-4698-4C6C-A9B9-AD7BB9948CEA,0x2D0B400,0x640000)
BLK11: Alias(s):
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(10,GPT,398B4473-B44F-421A-8FB6-2866A8B1E1DB,0x334B400,0x640000)
BLK12: Alias(s):
PciRoot(0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/NVMe(0x1,F9-91-95-45-8B-44-1B-00)/HD(11,GPT,0CCB6850-7DC4-47CB-90EC-DFC7E53E1B95,0x398B400,0x19835D00)
BLK4: Alias(s):
PciRoot(0x0)/Pci(0x1A,0x0)/Pci(0x0,0x0)/NVMe(0x1,B8-77-91-31-53-38-25-00)
BLK5: Alias(s):
PciRoot(0x0)/Pci(0x1A,0x0)/Pci(0x0,0x0)/NVMe(0x1,B8-77-91-31-53-38-25-00)/HD(1,GPT,0D5E68B3-7832-443B-8F72-266704FF0820,0x22,0x7FDE)
BLK7: Alias(s):
PciRoot(0x0)/Pci(0x1D,0x0)/Pci(0x0,0x0)/NVMe(0x1,45-D8-DB-2A-01-75-A0-00)
@cryptearth: Fully agree with what you say. Using a different fs and a driver from efifs is something I didn't think of, thanks for the suggestion. I keep this in mind when I experiment further. Also agree re: consumer vs pro hardware; my Supermicro boards didn't give me that much headache in years than what this one gave me just in a few weeks.
But this board is what I have in hand right now. I certainly try your suggestion later on since I am curious, but for the moment, I have settled on tried-and-true GRUB which works flawless for this situation.
I would extend your market share / testing argument to the systemd-boot mechanism. It's no exaggeration to say that the GRUB user base is at least hundred times bigger than the systemd-boot user base. Look at the UEFI map above, this system has quite a lot of devices at different places which could trigger bugs. Also, systemd-boot and especially the Discoverable Partitions mechanism was meant for automated deployments; I am not sure if it's sufficiently tested on systems with more than one drive, all different controllers (SATA/NVMe) etc. Thus my switch to GRUB.
Thanks for all the help I got. The suggestions helped to shape my assessment of the situation and avoid wasting even more time.
TL;DR for future readers in a similar situation: Cut your losses and switch to GRUB, this is better tested and has higher chances to work
Last edited by emk2203 (2024-10-18 08:58:31)
Offline
I gave several different types of booting a try in a vm as I wanted to figure if one is "better" (or at least more convenient) than others - but couldn't find any as "best" without also comming with its own fair share of issues
I got best results from grub - but I think this is mostly due to I have the most experience with it while still struggle with others like systemd
on the other side I also had one test setup with systemd with auto discover and no fstab at all - and at least within the vm it worked without issues (although not tested on bare metal)
as for your issue: I'm currently on phone so it's a bit hard to understand the mapping - but as the shell has a FS mapping for the additional xbootldr partition it seems that at least the shell sees it and can use it - which should be inhired from the uefi and hence should be the case for systemd, too
so, aside from the uefi firmware being the issue here it also could be the combination of both systemd and the firmware causing bugs - although I do believe systemd being very robust
anyway - its yet another of those "according to specs it should work - but due to some weird quirks and bugs it just doesn't"
recently there's a topic about an acer laptop which for sone reason locked the OP out of accessing the uefi because of an additional boot-entry - and only after it was cleared OP could access the settings again
the only reply from acer support was: "we don't support linux" - they completely missed the point it's thier crap firmware
Offline
btw - I looked up uefi spec: section 13 doesn't help here - seems an implementation bug
Offline