Partitions not created #159
Comments
marshall-mcmullen
changed the title
marshall-mcmullen
changed the title
|
I have also tried I feel like I must be missing another important step here after calling |
Thank you indeed. It started off as an offshoot of my work on linuxkit, where we not only need to shell out to docker, but, for things like ext4, we need to spin up a VM. Unfortunately, et4 has been half done for a long time. Let's say that filesystem work is challenging. |
Just as a small aside, most of these types are defined in |
No, and it is strange that is failing. But the CI (for obvious reasons) runs all of its tests on files rather than block devices, so it is possible we simply broke something at some point and missed it because CI did not catch it. Can you do the following? Whip up a quick fully-functional test program, or at least the complete |
|
Any progress here? I have the same issues from the 1.2.0 release to the latest master. I think it never worked. |
|
@deitch Here is an example which doesn't work https://github.com/immune-gmbh/isogen |
|
If I write the ProtectiveMBR it works with GPT generation |
@zaolin is that a private repo? I get a 404 for it. |
Can you modify that example to make writing the protective MBR optional (once you give me access to it)? Did you recreate the problem with block devices only? Or with file as well? |
|
@deitch It's now public. Yes I did it. It fixed the GPT generation issue. I just tried to generate an iso file. My problem here is that I want to generate a bootable ISO for UEFI systems similar to ubuntu/fedora bootable iso. I gave up at the moment because I would need to dig in even more and I don't understand how they made an eltrorito image on the same disk with your library. |
I remember digging into this a few years ago. I do not remember what the solution was. There are websites that describe how to create a UEFI-bootable ISO, but there was more to it than that. If you do dig into it, and manage to figure it out, do open an issue or (better yet) a PR here? Whether it means changes to the library, or just a good example, very happy to take it. In the meantime, let's look at the issue you raised. |
|
@zaolin I took your sample repo (you can remove it now, if you want) and simplified it to the below: package main
import (
"flag"
"fmt"
"os"
diskfs "github.com/diskfs/go-diskfs"
"github.com/diskfs/go-diskfs/partition/gpt"
)
func main() {
out := flag.String("out", "", "output file, required")
doParts := flag.Bool("parts", false, "whether or not to make partitions")
flag.Parse()
if out == nil || *out == "" {
os.Exit(-1)
}
if err := makeGPTDisk(*out, *doParts); err != nil {
fmt.Printf("failed to make efi bootloader iso: %v", err)
os.Exit(-1)
}
}
func makeGPTDisk(out string, makeParts bool) error {
var (
padding int64 = 5 * 1024 * 1024
diskSize int64 = 20 * 1024 * 1024
partitionSectors int64 = 100
partition1Start int64 = 2048
partition1End int64 = partitionSectors + partition1Start - 1
partition2Start int64 = partition1End + 1
partition2End int64 = ((diskSize - padding) / int64(diskfs.SectorSize512)) - partition2Start + 1
)
_ = os.Remove(out)
disk, err := diskfs.Create(out, diskSize, diskfs.Raw, diskfs.SectorSize512)
if err != nil {
return fmt.Errorf("failed to create disk file: %w", err)
}
disk.LogicalBlocksize = 2048
table := &gpt.Table{}
if makeParts {
table.Partitions = []*gpt.Partition{
{
Start: uint64(partition1Start),
End: uint64(partition1End),
Type: gpt.MicrosoftBasicData,
Name: "ISO9660",
},
{
Start: uint64(partition2Start),
End: uint64(partition2End),
Type: gpt.EFISystemPartition,
Name: "EFI System",
},
}
}
err = disk.Partition(table)
if err != nil {
return fmt.Errorf("failed to create partitiont table: %w", err)
}
return nil
}I ran it both with and without partitions, and it worked fine. I checked with $ go run . --out ./foo.img
$ gdisk -l ./foo.img
GPT fdisk (gdisk) version 1.0.9.1
Partition table scan:
MBR: not present
BSD: not present
APM: not present
GPT: present
Found valid GPT with corrupt MBR; using GPT and will write new
protective MBR on save.
Disk ./foo.img: 40960 sectors, 20.0 MiB
Sector size (logical): 512 bytes
Disk identifier (GUID): 9305AEB4-3DAC-43C0-9CA3-03DDE3769B74
Partition table holds up to 128 entries
Main partition table begins at sector 2 and ends at sector 33
First usable sector is 34, last usable sector is 40927
Partitions will be aligned on 2048-sector boundaries
Total free space is 40894 sectors (20.0 MiB)
Number Start (sector) End (sector) Size Code Name
$ go run . --out ./foo.img --parts true
$ gdisk -l ./foo.img
GPT fdisk (gdisk) version 1.0.9.1
Partition table scan:
MBR: not present
BSD: not present
APM: not present
GPT: present
Found valid GPT with corrupt MBR; using GPT and will write new
protective MBR on save.
Disk ./foo.img: 40960 sectors, 20.0 MiB
Sector size (logical): 512 bytes
Disk identifier (GUID): FEE5E1A1-E31D-4398-9E2E-491299C478E1
Partition table holds up to 128 entries
Main partition table begins at sector 2 and ends at sector 33
First usable sector is 34, last usable sector is 40927
Partitions will be aligned on 4-sector boundaries
Total free space is 14368 sectors (7.0 MiB)
Number Start (sector) End (sector) Size Code Name
1 2048 2147 50.0 KiB 0700 ISO9660
2 2148 28573 12.9 MiB EF00 EFI SystemIt might be due to using gdisk vs fdisk, or it might be something in the numbers passed (I changed them a bit in my sample). |
|
edit: the issue is related to
Original problem description with snippet
Here is a snippet of the code I use to create my partition table and partition: package main
import (
"fmt"
"io"
"os"
"github.com/diskfs/go-diskfs"
"github.com/diskfs/go-diskfs/disk"
"github.com/diskfs/go-diskfs/filesystem"
"github.com/diskfs/go-diskfs/partition/gpt"
)
func main() {
err := PartitionAndFormat("/dev/sdf")
if err != nil {
panic(err)
}
}
func PartitionAndFormat(device string) error {
d, err := diskfs.Open(device)
if err != nil {
return fmt.Errorf("opening device: %w", err)
}
defer d.File.Close()
startSector := uint64(PartitionAlignmentBoundary / int64(diskfs.SectorSize512))
endSector := uint64(d.Size/int64(diskfs.SectorSize512)) - 1
pt := &gpt.Table{
Partitions: []*gpt.Partition{{
Type: gpt.EFISystemPartition,
Name: "EFI System",
Start: startSector,
End: endSector,
}},
}
if err = d.Partition(pt); err != nil {
return fmt.Errorf("writing partition table: %w", err)
}
fs, err := d.CreateFilesystem(disk.FilesystemSpec{
Partition: 1,
FSType: filesystem.TypeFat32,
VolumeLabel: "boot",
})
if err != nil {
return fmt.Errorf("creating FAT32 filesystem: %w", err)
}
return nil
}Reported by // disk
&{
File: 0xc0000a4150,
Info: 0xc0002a6750,
Type: 1,
Size: 1073741824,
LogicalBlocksize: 512,
PhysicalBlocksize: 512,
Table: 0xc0002934d0,
Writable: true,
DefaultBlocks: false,
}
// partition(s)
gpt
[&{
Start: 2048,
End: 2097151,
Size: 1072693248,
Type: "C12A7328-F81F-11D2-BA4B-00A0C93EC93B",
Name: "EFI System",
GUID: "F49DF9DD-BCC6-4D5F-8499-C8E59A512B56",
Attributes: 0,
logicalSectorSize: 512,
physicalSectorSize: 512,
},&{
Start: 0,
End: 0,
Size: 0,
Type: "00000000-0000-0000-0000-000000000000",
Name: "",
GUID: "00000000-0000-0000-0000-000000000000",
Attributes: 0,
logicalSectorSize: 512,
physicalSectorSize: 512,
},
// ...
]Reported by |
|
Now I am confused. Does it work for you, or not? And if not, do you know why? |
|
@deitch sorry if my edit wasn't clear enough. I just amended it. Here is an update and recap:
In conclusion the lib works flawlessly, but one needs to pay attention to setting Also relevant:
which makes me wonder: should the lib maybe write the protective MBR by default, but allow users to opt out? The UEFI spec seems to suggest that this should be the case. I learned a lot by reading through this implementation by the way, stellar work folks 👍 |
Greetings, just found this project and am amazed at what you've built so far. LOVE the idea of being able to partition disks natively inside Go without having to have all the tools installed on a machine and having to shell out. Great stuff.
Problem is that the partitions don't appear to be getting created. Maybe I'm doing something wrong? I'd really love if you can double check how I'm using this and see if anything obvious jumps out at you.
Here's a simplified version of what I'm doing:
There are NO errors reported. And the log message at the end looks like everything is proper:
However, when I exit the Go program above, and try to print the partitions via
fdiskthey are not there:This is running inside an Alpine Install ISO with busybox.
Thank you for any help you can provide!!
The text was updated successfully, but these errors were encountered: