Red Hat Bugzilla – Bug 752702
"you have not created a bootloader stage1 device" error
Last modified: 2013-09-07 09:21:14 EDT
+++ This bug was initially created as a clone of Bug #739389 +++
I am trying to install from the Fedora 16 x86_64 Live ISO image.
I selected manual partitioning and had already partitioned my disk, so I just set the mountpoints and filesystem-to-format-as:
# parted /dev/sda print
Model: ATA M4-CT256M4SSD2 (scsi)
Disk /dev/sda: 256GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos
Number Start End Size Type File system Flags
1 4194kB 1000MB 996MB primary fat32 boot, lba
2 1000MB 5000MB 3999MB primary
3 5000MB 100GB 95.0GB primary
4 100GB 255GB 155GB primary
(Background: partition #1 is where I have livecd-iso-to-disk'd the Live ISO, and anaconda is showing its "Type" as "EFI System Partition")
I get the following partitioning error: "you have not created a bootloader stage1 device".
I noticed in /tmp/storage.log that my disklabel is detected as msdos:
# grep " format = existing" /tmp/storage.log | sort | uniq
format = existing efi filesystem
format = existing ext4 filesystem
format = existing msdos disklabel
format = existing None
...and given that, I'm not sure why I'm being asked about a BIOS boot partition.
Is this a bug or am I misunderstanding something?
I will attach
/tmp/syslog if non-empty, else /var/log/messages
PS - This is the same error message as bug #739389, but because I actually do not have a GPT disk I opened a new bug.
Created attachment 532736 [details]
/tmp/anaconda.log as requested
Created attachment 532737 [details]
/tmp/storage.log as requested
Created attachment 532739 [details]
/tmp/program.log as requested
Because of the widespread use of USB install media we do not allow bootloader installation to the device that contains the install media (see bug 750469 ). As a result of that, you will have to use removable media.
dlehman: so, NOTABUG?
Fedora Bugzappers volunteer triage team
(In reply to comment #5)
> dlehman: so, NOTABUG?
Well, I'm not sure why we wouldn't want to do what I'm doing (running the installer from a partition on the same device as the partition to which one is installing Fedora), so it would be nice to be able to do that. I'm going to manually back out anaconda commit http://git.fedorahosted.org/git/?p=anaconda.git;a=blobdiff;f=pyanaconda/bootloader.py;h=547f7b81a4a35205b522ce9542ad5e5fc5e5b59d;hp=6250f74e91a64d542c4fc9d02dacf15b6dd450bf;hb=74c634f1f90fbe1ee4c283deabd65fb66e5bbebe;hpb=fc87b77ab69befe900d502bc2a17a5d675548eb1 from my Live image's anaconda to try to finish the install, but there is probably a better way.
A straw man: a scary-looking option to "allow bootloader installation to same device as holds installation media" would make it possible.
I'm open to suggestions as to what patch might be acceptable and I can try to help.
(In reply to comment #6)
> I'm going to
> manually back out anaconda commit
> from my Live image's anaconda to try to finish the install
That didn't work, and even creating a bios boot partition still results in the message "you have not created a bootloader stage1 device".
I will attach logs and a screenshot.
Created attachment 533008 [details]
screenshot of error message despite creating bios boot partition
Created attachment 533009 [details]
storage log after creating bios boot partition
Created attachment 533010 [details]
anaconda.log after creating bios boot partition
Created attachment 533011 [details]
program.log after creating bios boot partition
This breaks a very nice to maintain linux desktops.
When a system has been fubar'ed it is nice to have a small partition on the disk set up to boot anaconda with the necessary options to start a kickstart-install that wipes the system partitions and reinstalls a clean system.
If the system is still up and on the net it is very easy to remotely log in and set up grub to initiate a reinstall on reboot. Otherwise, it is usually possible to guide a user to fire the reinstall manually from the grub menu.
We used this at the university where I used to work.
At the very least this needs to be something we can override in the kickstart file. We need to be able to install bootloader on the disk the installer is booted from.
Martin, you'll also want to revert these commits to install the bootloader to sda:
This is a regression (right?) and undesired behavior. Please can you re-open and give some guidance as to what type of patch you might accept. Installing from a small partition on the installation-target media is quite a valid use case.
I will attach a straw-man patch.
Created attachment 533693 [details]
straw man patch
This is problematic in several ways.
1) The message is unhelpful.
* It is not clear if it is a partition or a complete device, that is not created
* It is not clear why this is a problem
* It is not clear what can satisfy the Anaconda checks.
On Fedora 19 (LXDE spin) this message is shown regardless if a boot loader is chosen to be installed or not. On the screen "Installation destination", there is a clickable text link to a dialog, where a disk can be set as "boot device". The choice here does not affect Anaconda, as it throws the error "you have not created a bootloader stage1 target device" either way.
If this error message is triggered by running the live image off one partition and setting the install target to other partitions not in use, but on the same disk, then the error message is unusable to regular users.