sgdisk - Unexpected Behavior with Decimal Point in Partition Size Specification

[adrya]

When using the sgdisk command to create a new partition with a specified size, the tool behaves unexpectedly when a size is specified with a decimal point. Specifically, using a size like 7.5G results in the partition size being set to 3.5 KiB instead.

Steps to Reproduce:

    Run the following command to create a partition:

    sgdisk --new=1:0:+7.5G --typecode=1:8300 --change-name=1:boot /dev/sda

    Verify the size of the newly created partition using sgdisk --print /dev/sda.

Expected Behavior:
The partition size should be set to 7.5 gigabytes, as specified in the --new command.

Actual Behavior:
The partition size is set to 3.5 KiB.

Additional Information:

    The issue can be worked around by specifying the size in megabytes, such as 7500M.

System Information:

    Operating System: Ubuntu 22.04.3 LTS
    sgdisk Version: GPT fdisk (sgdisk) version 1.0.8

[Rod Smith (rodsmith)]

This is fixed in an upstream commit:

https://sourceforge.net/p/gptfdisk/code/ci/e1cc654ef71996d836c5d051278130f50f768f84/tree/

This commit causes input to be truncated to the nearest whole value; e.g., "+7.5G" becomes "+7G". Decimal processing would be imprecise and/or would add code that might hide more bugs. If that level of precision is needed, then using a lower unit (e.g., M or K rather than G) is more appropriate.

The upstream commit will eventually be released, presumably as version 1.0.10, but it's not out yet. Note that the current released version of GPT fdisk is 1.0.9, released in April of 2022; version 1.0.8 is out of date, although it was current as of Ubuntu 22.04's release.

[Sergio Durigan Junior (sergiodj)]

Thanks for taking the time to report this bug.

I can confirm it on all supported Ubuntu releases (F/J/L/M). Here's the reproducer:

# fallocate -l 10G a.img
# sgdisk --new=1:0:+7.5G --typecode=1:8300 ./a.img
# fdisk -l ./a.img

I haven't tried the patch mentioned by Rod, but it does seem to address the issue.

I'll add this to our backlog.

[Brian Murray (brian-murray)]

Ubuntu 23.04 (Lunar Lobster) has reached end of life, so this bug will not be fixed for that specific release.

[Mitchell Dzurick (mitchdz)]

Rod, do you know when the presumed 1.0.10 release will come out?

[Mitchell Dzurick (mitchdz)]

Let's wait for this upstream fix to go into a release and then Debian can pick it up - from there, our archive will sync with Debian. I'll ask upstream to make a release.

[Mitchell Dzurick (mitchdz)]

Made bug in sourceforge here asking for new release - https://sourceforge.net/p/gptfdisk/discussion/939590/thread/84321440ed/

[Rod Smith (rodsmith)]

I've released version 1.0.10 yesterday.

[Mitchell Dzurick (mitchdz)]

Thanks a ton Rod! Addding server-triage-discuss to get this properly tracked.

[Bryce Harrington (bryce)]

Sounds like this should be included with the next merge of this package (presumably next cycle).
Process for tracking these kinds of bugs is to tag them 'packaging'. The tooling for automatically attaching these to the generated 'needs-merge' card is not in place, so I'm also tagging *this* bug 'needs-merge' but with milestone 'later'.