Opened 14 years ago

Last modified 14 years ago

#5968 new bug

DriveSetup install problem

Reported by: oxoocoffee Owned by: stippi
Priority: normal Milestone: R1
Component: Applications/DriveSetup Version: R1/alpha2
Keywords: Unknown Partition Type Cc:
Blocked By: Blocking:
Platform: x86

Description

DriveSetup fails to setup proper partition. System Setup: Shuttle X27D Atom with 300Gb SATA drive.

Here are steps:

  1. Boot to safe mode and enable fail safe video
  2. Continue Booting
  3. In Drive Setup select "/dev/dis/ata/0/master/raw" not the <empty> option
  4. Go to Partition and select Initialize->Be File System. Create option is grayed out.
  5. Mount Partition - Volume Name Haiku, Mounted @ /Haiku1 298.09 GiB
  6. Exit Drive Setup
  7. Choose mounted partition "Onto: " - Haiku - 298.09 GiB
  8. Press Begin.
  9. After installation is done reboot.
  10. On reboot I get prompted with Haiku boot loader not finding the partition.
  11. Try to reinstall again this time DriveSetup shows that 39 GiB partition is created with Unknown Type 0x5f and rest is <empty>

Attached console output is from such machine. The only thin is that when enabling "console output" and fail safe video mode console output is produced but VESA driver is not activated (screen freezes the same way as fail safe would not be activated - Intel gfx problem)

Attachments (2)

BugX27D.log (51.6 KB ) - added by oxoocoffee 14 years ago.
Screen.7z (412.0 KB ) - added by oxoocoffee 14 years ago.
Screen Shots from Haiku

Download all attachments as: .zip

Change History (11)

by oxoocoffee, 14 years ago

Attachment: BugX27D.log added

comment:1 by stippi, 14 years ago

Hm. I am a bit at a loss about what could have happened here. Is it possible for you to get a partition table listing from another OS? Perhaps with the GParted Live CD or something?

comment:2 by oxoocoffee, 14 years ago

Ok I did booted GParted live and after opening terminal I did.

# fdisk -l

Disk /dev/sda: 320.1 GB, 320072933376 255 heads, 63 sectors/track, 38913 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Disk identifier: 0x00000000

Device Boot Start End Blocks Id System

That shows that there are no partitions.

I am attaching screen caps from Haiku.

Screen 1 - Show I am trying to create one partition

Select <empty> go to Partition -> Create

Screen 2 - Show error message I got after trying to write changes after Screen 1 Screen 3 - Show Initialize <empty> to Be File. Not sure if this is needed but reversing the steps (Screen 3, Screen 1 and 2 does not makes any difference). Screen 3 show Selecting <empty> and selecting Partition -> Initialize -> Be File

By The way the GParted was taken right after reboot from attached (Screen.7z) screen shots

by oxoocoffee, 14 years ago

Attachment: Screen.7z added

Screen Shots from Haiku

comment:3 by oxoocoffee, 14 years ago

I noticed something similar in Virtual box 3.2 Beta 3 when I was trying to create new Haiku image. Also some post install errors that were happening in 3.1.6 are gone now (but that is a different issue)

comment:4 by axeld, 14 years ago

I just talked to Robert, and we found out that the reason this happens is because he did not create an Intel partition in the empty spot, but instead initialized it with BFS.

This shouldn't be allowed by DriveSetup, but the better solution would be to automatically create a partition in the spot if the user does that.

in reply to:  4 ; comment:5 by bonefish, 14 years ago

Replying to axeld:

I just talked to Robert, and we found out that the reason this happens is because he did not create an Intel partition in the empty spot, but instead initialized it with BFS.

This shouldn't be allowed by DriveSetup, but the better solution would be to automatically create a partition in the spot if the user does that.

Sure this should be allowed. That's how one would initialize a partition-less USB stick. According to the syslog the BFS volume is detected correctly, BTW:

580	device 0: /dev/disk/ata/0/master/raw
581	  media status:      No error
582	  device flags:      2
583	  offset:            0
584	  size:              320072933376 (305245.335 MB)
585	  content size:      320072933376
586	  block size:        2048
587	  child count:       0
588	  index:             -1
589	  status:            0
590	  flags:             3
591	  volume:            -1
592	  disk system:       file_systems/bfs/v1
593	  name:              <NULL>
594	  content name:      Haiku
595	  type:              <NULL>
596	  content type:      Be File System
597	  params:            <NULL>
598	  content params:    <NULL>

The syslog is missing the boot loader part BTW. It would be interesting to see the syslog of a boot from the hard disk (continued on CD).

Installing an MBR on a disk initialized this way renders it unbootable, BTW. Just in case that happened.

After following your instructions in qemu (with the R1 alpha 2 ISO CD image and a 700 MB disk image) up to point 9, the system boots fine from disk and DriveSetup shows the setup correctly.

in reply to:  5 ; comment:6 by davidsaunders, 14 years ago

Replying to bonefish:

Replying to axeld:

While it is nessecary to allow for initializing a disk with-out any partitions, Drive-Setup should check to see if the device is a SATA, PATA, or SCSI Hard Disk Drive, and if it is dissable this option.

There are some BIOSes that will not load the root sector of an HDD if there is not a valid Partition table (I have run into this problem a couple of times, and have no idea why this is). As such the bug is not in drive setup per-se though rather it is a common BIOS bootloader bug.

in reply to:  6 comment:7 by umccullough, 14 years ago

Replying to davidsaunders:

While it is nessecary to allow for initializing a disk with-out any partitions, Drive-Setup should check to see if the device is a SATA, PATA, or SCSI Hard Disk Drive, and if it is dissable this option.

There are some BIOSes that will not load the root sector of an HDD if there is not a valid Partition table (I have run into this problem a couple of times, and have no idea why this is). As such the bug is not in drive setup per-se though rather it is a common BIOS bootloader bug.

I have run into many of those same BIOS that refuse to boot a USB stick without a partition table as well - so the conclusion that this should be disabled only for certain disk types seems a bit weak. It would probably be better to at least display some kind of warning in all cases where there is no MBR/partition table and let the user decide what to do.

comment:8 by davidsaunders, 14 years ago

I would agree that user input is the only reasonable solution.

in reply to:  5 comment:9 by bebop, 14 years ago

Replying to bonefish:

This shouldn't be allowed by DriveSetup, but the better solution would be to automatically create a partition in the spot if the user does that.

Sure this should be allowed. That's how one would initialize a partition-less USB stick. According to the syslog the BFS volume is detected correctly, BTW:

Agreeing with bonefish here. I am also having a hard time understanding the disk initialization part, were you able to initialize the raw device to bfs? Did you then initialize an MBR over the top of that partition?

Note: See TracTickets for help on using tickets.