Bootloader (EFI) fails to find a valid BFS boot volume on bare metal

[win8linux]

Attempting to boot Haiku on an EFI-based system fails due to the bootloader not being able to find a valid BFS boot volume. This has been done with multiple flash drives, via both imaging the drives and installing Haiku onto flash drives from a VM install. The usual process for setting up UEFI booting was done. Rescanning does not fix the problem.

Unable to replicate this bug in a virtual machine, only when booting directly on bare metal.

[waddlesplash]

Please collect a bootloader log.

[win8linux]

What are the steps to get a bootloader log?

[waddlesplash]

Please read the documentation; it's one of the options in the bootloader menus.

[win8linux]

That option only displays the current bootloader log. Haven't found an option to collect a log yet. What should I be looking for, if what's needed is a photo of part of the log?

[waddlesplash]

Yes, the current bootloader log is what's needed here, especially if it can't find valid partitions.

[win8linux]

BFS mount check

[win8linux]

Haven't found yet an option to get the full log, so here's a photo of the BFS mount check.

[win8linux]

GUID partitioning module scan fails

[waddlesplash]

"could not find partition" is an interesting message.

[madmax]

Both partitions are of type ebd0a0a2-b9e5-4433-87c0-68b6b72699c7, which is a Microsoft data partition, not a Haiku BFS one. Either recreate (not just format) the partition with DriveSetup or use your preferred partitioning tool to set its partition type to 42465331-3BA3-10F1-802A-4861696B7521.

[win8linux]

This did not fix the issue. Same error still happens when attempted with flash drives, even with the correct part type. Also does not explain why directly imaging a drive with official images still causes the issue.

[madmax]

Also does not explain why directly imaging a drive with official images still causes the issue.

It doesn't. Can you get the logs for that case? In the ones provided we have:

• A GPT disk with two MS data partitions and what is reported as a dirty header. This one would not boot as is due to the partition type.
• A GPT disk with 9 partitions, only the last two showing, one of them of the correct type, but for whatever reason it says it couldn't scan it. It looks like the disk is discarded? Can we have full logs?

[win8linux]

Is there a way to save full bootloader logs to a disk? Otherwise, getting full logs would mean uploading multiple photos.

[pulkomandy]

If you have an USB drive with a FAT32 partition connected, the bootloader should offer to save bootloader logs there. However, since the problem you are investigating is about detection of mass storage, it is quite likely that this won't work.

Another option is a serial port, if you have one, but this is very rare on modern machines.

So, only the taking pictures of the screen remains. Yes, it is not very fun to investigate and debug such problems.

[win8linux]

Start of boot log from a portable install

[win8linux]

Start of boot log from drive directly images with a Haiku nightly ISO

[madmax]

Is the HD the system boot device and does the EFI bootloader come from an old version? Remember that it is not updated automatically. I think some of the awkward things I see can only happen if no system is found in the BFS partition, which could happen due to a broken system or a very old loader not supporting current packages.

[win8linux]

In both cases, the external drive is using an updated bootloader. The bootloader in the internal drive is not updated, however it is not being used here.

[madmax]

Could you try ​this other loader and see if the log says something more interesting?

[begasus]

debug output

[begasus]

OK, this looks very similar with the problem I'm having here also getting "Boot volume is not valid"

[begasus]

Syslog data after a boot/reboot from 32bit install with a thumbdrive not wanting to boot, not even after selecting older installed state