Details of Disk Utility’s partitioning decisions
I decided it was time I became familiar with the nuts and bolts of how disks on Macs are partitioned by Disk Utility, and why.
After reading up on MBR, GPT, and how they compare, I assembled these tools:
- Disk Utility: the “new” version, as shipped with each version of OS X/macOS
- Disk Utility: the “original” (or “good”) version, patched so it’ll still run on OS X 10.11
- diskutil: Apple’s official disk management tool
- gpt: the FreeBSD tool for GPT disks, shipped with macOS
- fdisk: a much older tool for MBR disks, also shipped with macOS
- gdisk: a newer fdisk-based tool for GPT disks
- installable via Homebrew with
brew install gptfdisk
- installable via Homebrew with
Partition tables, partition tables everywhere
For reference, my internal HFS+ drive according to diskutil:
$ diskutil list
/dev/disk0 (internal, physical):
#: TYPE NAME SIZE IDENTIFIER
0: GUID_partition_scheme *1.0 TB disk0
1: EFI EFI 209.7 MB disk0s1
2: Apple_CoreStorage Macintosh HD 999.7 GB disk0s2
3: Apple_Boot Recovery HD 650.0 MB disk0s3
/dev/disk1 (internal, virtual):
#: TYPE NAME SIZE IDENTIFIER
0: Apple_HFS Macintosh SSD +999.3 GB disk1
Logical Volume on disk0s2
FB27242A-A745-4666-AB33-9A934785F0CB
Unencrypted
Here we clearly see the internal drive is GPT with a 200MB EFI system partition (ESP), a Core Storage partition containing the boot volume, and a recovery partition. (Run diskutil cs list for a better view of how Core Storage volumes are structured.)
The same drive according to gpt:
$ sudo gpt show /dev/disk0
start size index contents
0 1 PMBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 1952530904 2 GPT part - 53746F72-6167-11AA-AA11-00306543ECAC
1952940544 1269536 3 GPT part - 426F6F74-0000-11AA-AA11-00306543ECAC
1954210080 7
1954210087 32 Sec GPT table
1954210119 1 Sec GPT header
The gpt tool shows everything, and counts according to 512-byte sectors. The protective MBR occupies one 512-byte sector, the GPT header occupies the next sector, and the partition table entries follow. Then at sector 34 there’s a 3KB gap so that the first partition starts on a 4KB boundary (i.e. 20KB from the start of the disk). The ESP is 409600 sectors * 512 bytes per sector / 1024 bytes per KB / 1024 KB per MB = 200MB. Then our remaining partitions, another boundary alignment gap, and the secondary GPT table and header.
This follows the behaviour described in Secrets of the GPT. If you’ve read that and are wondering why there are no 128MB gaps after each non-ESP partition, note the article doesn’t mention that those gaps are only added after HFS+ partitions when defined directly within the partition table, and not when encapsulated within a Core Storage logical volume group.
For example, an 8GB USB drive partitioned with original Disk Utility to use GPT for two HFS+ partitions looks like:
$ sudo gpt show /dev/disk2
start size index contents
0 1 PMBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 5234376 2 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
5644016 262144
5906160 9490664 3 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
15396824 262151
15658975 32 Sec GPT table
15659007 1 Sec GPT header
Note the 128MB gap after the first partition, and the 128MB + 7 sector gap after the second. This gap is slightly larger because making the end of the second HFS+ partition align to the next 4KB boundary (8 sectors later) would have made the gap one sector short of 128MB.
But if we repartition so the first is FAT:
$ diskutil list /dev/disk2
/dev/disk2 (external, physical):
#: TYPE NAME SIZE IDENTIFIER
0: GUID_partition_scheme *8.0 GB disk2
1: EFI EFI 209.7 MB disk2s1
2: Microsoft Basic Data UNTITLED 1 4.0 GB disk2s2
3: Apple_HFS Untitled 2 3.7 GB disk2s3
we get:
$ sudo gpt show /dev/disk2
gpt show: /dev/disk2: Suspicious MBR at sector 0
start size index contents
0 1 MBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 2008
411648 7829504 2 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
8241152 7155672 3 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
15396824 262151
15658975 32 Sec GPT table
15659007 1 Sec GPT header
…a just-under 1MB gap before the FAT volume, but not after. This is OS X following Windows’ practice of aligning partitions it creates on 1MB boundaries for disks over 4GB, which translates to the location of the partition’s first sector (411648) being divisible by 2048. (Note that OS X does this for all disk sizes rather than on 64KB/128 sector boundaries for smaller disks, as Windows does.)
As for that “suspicious MBR”, we can use fdisk to read its master boot record partition table:
$ sudo fdisk /dev/disk2
Disk: /dev/disk2 geometry: 974/255/63 [15659008 sectors]
Signature: 0xAA55
Starting Ending
#: id cyl hd sec - cyl hd sec [ start - size]
------------------------------------------------------------------------
1: EE 1023 254 63 - 1023 254 63 [ 1 - 409639] <Unknown ID>
2: 0B 1023 254 63 - 1023 254 63 [ 411648 - 7829504] Win95 FAT-32
3: AF 1023 254 63 - 1023 254 63 [ 8241152 - 7155672] HFS+
4: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
Aha: the presence of a FAT partition made Disk Utility create a hybrid MBR. This would in theory allow the disk to mount on older systems that don’t support GPT.
- #1 covers everything from the start of the disk to the end of the ESP and is marked as part of the protective MBR (EE)
- #2 is the FAT (0B) partition
- #3 is the HFS+ (AF) partition
As the linked article explains, hybrid MBRs can be a dangerous hack, and Windows XP can’t actually mount this disk as-is unless the hybrid MBR table is reordered with gdisk to list any mountable partitions before the EE partition. (Otherwise you’ll be told “this partition is not enabled” when assigning a drive letter in Disk Management, and reformatting what shows in My Computer will wipe out the primary GPT structures and the ESP in favour of a 200MB FAT volume!)
The protective MBR that Disk Utility usually creates on disks (meant to actively deter non-GPT-aware disk utilities from thinking the disk is unformatted) looks like this:
$ sudo fdisk /dev/disk0
Disk: /dev/disk0 geometry: 121643/255/63 [1954210120 sectors]
Signature: 0xAA55
Starting Ending
#: id cyl hd sec - cyl hd sec [ start - size]
------------------------------------------------------------------------
1: EE 1023 254 63 - 1023 254 63 [ 1 - 1954210119] <Unknown ID>
2: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
3: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
4: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
where the sole EE partition covers the entire disk.
And if we reformat our 8GB USB drive as GPT with HFS+ and then FAT:
$ sudo gpt show /dev/disk2
gpt show: /dev/disk2: Suspicious MBR at sector 0
start size index contents
0 1 MBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 7829504 2 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
8239144 264152
8503296 7153664 3 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
15656960 2015
15658975 32 Sec GPT table
15659007 1 Sec GPT header
The gap between #2 and #3 is 262144 sectors trailing the HFS+ partition (the 128MB gap mentioned above) plus 2008 sectors leading the FAT partition so it starts on a megabyte boundary. Also note the 2015 sector trailing gap, which wasn’t present when FAT came first. This gives the FAT partition a size that also ends on a megabyte boundary.
And if we use HFS+ followed instead by ExFAT, do we get the same gap and hybrid MBR?
$ sudo gpt show /dev/disk2
gpt show: /dev/disk2: Suspicious MBR at sector 0
start size index contents
0 1 MBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 7826976 2 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
8236616 262584
8499200 7157760 3 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
15656960 2015
15658975 32 Sec GPT table
15659007 1 Sec GPT header
Here, the gap between #2 and #3 is 262144 sectors (128MB) trailing the HFS+ partition plus 440 sectors leading the ExFAT partition (a 1MB boundary gap, as before), and 2015 sectors trailing it. And “Suspicious MBR” indicates the presence of a hybrid MBR.
Further down the MBR rabbit hole
Having gone this far, we may as well keep digging. Using the original Disk Utility, I’ll format the same 8GB drive as MBR with a variety of partitions.
$ diskutil list /dev/disk2
/dev/disk2 (external, physical):
#: TYPE NAME SIZE IDENTIFIER
0: FDisk_partition_scheme *8.0 GB disk2
1: Apple_HFS HFS 2.1 GB disk2s1
2: Apple_HFS HFSCS 2.1 GB disk2s2
3: DOS_FAT_32 FAT 2.0 GB disk2s3
4: Windows_NTFS EXFAT 1.7 GB disk2s5
Note that:
- diskutil refers to MBR as
FDisk_partition_scheme - both standard and case-sensitive HFS+ are
Apple_HFS - FAT is marked as
DOS_FAT_32(so presumably it’s not FAT12 or FAT16) - ExFAT is marked as
Windows_NTFS(what???) - all four partitions were supposed to be 2GB, but the last got shorted by a few hundred MB because of the 128MB trailing gap for each HFS+ partition, which diskutil includes in their total sizes
Where is diskutil getting the partition type info? Presumably from the MBR partition type IDs. Can we see those directly?
$ sudo gpt show /dev/disk2
start size index contents
0 1 MBR
1 1
2 4176896 1 MBR part 175
4176898 1
4176899 4176896 2 MBR part 175
8353795 1
8353796 3914752 3 MBR part 11
12268548 3390460 4 MBR part 5
Neat, gpt can read MBR tables. And it does show partition IDs, but in decimal for some reason.
- first sector is the MBR
- second sector is the (unlabelled) EBR which points to the final partition (see below)
- then our HFS+ and case-sensitive HFS+ partitions with just a 1-sector boundary after each, because 128MB gaps are only added on GPT disks (175 == 0xAF)
- then our FAT/ExFAT partitions (11 == 0x0B)
To see both the MBR and EBR contents, use fdisk:
$ sudo fdisk /dev/disk2
Disk: /dev/disk2 geometry: 974/255/63 [15659008 sectors]
Signature: 0xAA55
Starting Ending
#: id cyl hd sec - cyl hd sec [ start - size]
------------------------------------------------------------------------
1: AF 1023 254 63 - 1023 254 63 [ 2 - 4176896] HFS+
2: AF 1023 254 63 - 1023 254 63 [ 4176899 - 4176896] HFS+
3: 0B 1023 254 63 - 1023 254 63 [ 8353796 - 3914752] Win95 FAT-32
4: 05 1023 254 63 - 1023 254 63 [ 12268548 - 3390460] Extended DOS
Signature: 0xAA55
Starting Ending
#: id cyl hd sec - cyl hd sec [ start - size]
------------------------------------------------------------------------
1: 07 1023 254 63 - 1023 254 63 [ 12268549 - 3390459] HPFS/QNX/AUX
2: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
3: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
4: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
This lets us see the actual hex value partition type IDs in the MBR that Disk Utility chose to assign.
- AF is HFS or HFS+
- 0B is FAT32 with CHS addressing (instead of the newer 0C for LBA addressing; I’m guessing this is for max backwards compatibility)
- 05 is the EBR, covering everything past the third partition
And then in the extended partition table:
- 07 is ambiguous — it could be IFS, HPFS, NTFS, ExFAT, or old QNX. And now we know why diskutil just shows the last partition as
Windows_NTFS, because it can’t tell which it actually is without reading the partition’s data itself.
What does gdisk say?
$ sudo gdisk /dev/disk2
GPT fdisk (gdisk) version 1.0.4
Warning: Devices opened with shared lock will not have their
partition table automatically reloaded!
Partition table scan:
MBR: MBR only
BSD: not present
APM: not present
GPT: not present
***************************************************************
Found invalid GPT and valid MBR; converting MBR to GPT format
in memory. THIS OPERATION IS POTENTIALLY DESTRUCTIVE! Exit by
typing 'q' if you don't want to convert your MBR partitions
to GPT format!
***************************************************************
Warning! Main partition table overlaps the first partition by 32 blocks!
You will need to delete this partition or resize it in another utility.
Warning! Secondary partition table overlaps the last partition by
33 blocks!
You will need to delete this partition or resize it in another utility.
Command (? for help): p
Disk /dev/disk2: 15659008 sectors, 7.5 GiB
Sector size (logical): 512 bytes
Disk identifier (GUID): 082F14ED-F53A-4922-8382-3BF9908F4433
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 15658974
Partitions will be aligned on 1-sector boundaries
Total free space is 3 sectors (1.5 KiB)
Number Start (sector) End (sector) Size Code Name
1 2 4176897 2.0 GiB AF00 Apple HFS/HFS+
2 4176899 8353794 2.0 GiB AF00 Apple HFS/HFS+
3 8353796 12268547 1.9 GiB 0700 Microsoft basic data
5 12268549 15659007 1.6 GiB 0700 Microsoft basic data
What a weird message. Is it detecting an old secondary GPT table at the end of the disk, making it think it should still be GPT? This would cause it to think there’s partition overlap even though the disk isn’t GPT at all. And the last two partitions are labelled the same because gdisk is translating the MBR type IDs to their GPT equivalents, which are the same for FAT and ExFAT.
Once more time, with GPT
And now we’ll repartition with the same volume types, but with GPT instead.
$ diskutil list /dev/disk2
/dev/disk2 (external, physical):
#: TYPE NAME SIZE IDENTIFIER
0: GUID_partition_scheme *8.0 GB disk2
1: EFI EFI 209.7 MB disk2s1
2: Apple_HFS HFS 2.0 GB disk2s2
3: Apple_HFS HFSCS 2.0 GB disk2s3
4: Microsoft Basic Data FAT 2.0 GB disk2s4
5: Microsoft Basic Data EXFAT 1.5 GB disk2s5
Note:
- addition of the standard 200MB ESP
- both the standard and case-sensitive HFS+ partitions have the same label
- FAT and ExFAT partitions are both labelled “Microsoft Basic Data”
$ sudo gpt show /dev/disk2
gpt show: /dev/disk2: Suspicious MBR at sector 0
start size index contents
0 1 MBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 3923256 2 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
4332896 262144
4595040 3914752 3 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
8509792 263840
8773632 3913728 4 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
12687360 2969600 5 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
15656960 2015
15658975 32 Sec GPT table
15659007 1 Sec GPT header
Note:
- we still have a hybrid MBR
- only the HFS+ partitions have a trailing 128MB gap
- the second HFS+ partition’s trailing gap is 1696 sectors larger than it needs to be so that the FAT partition that follows starts on a 1MB (2048-sector) boundary
- the full GPT GUIDs are:
C12A7328-F81F-11D2-BA4B-00A0C93EC93B- ESP48465300-0000-11AA-AA11-00306543ECAC- HFS+EBD0A0A2-B9E5-4433-87C0-68B6B72699C7- Microsoft Basic Data, which could equal MBR type IDs 06, 07, 0B, 01, 04, 0C, 0E
$ sudo fdisk /dev/disk2
Disk: /dev/disk2 geometry: 974/255/63 [15659008 sectors]
Signature: 0xAA55
Starting Ending
#: id cyl hd sec - cyl hd sec [ start - size]
------------------------------------------------------------------------
1: EE 1023 254 63 - 1023 254 63 [ 1 - 409639] <Unknown ID>
2: AF 1023 254 63 - 1023 254 63 [ 409640 - 3923256] HFS+
3: AF 1023 254 63 - 1023 254 63 [ 4595040 - 3914752] HFS+
4: 0B 1023 254 63 - 1023 254 63 [ 8773632 - 3913728] Win95 FAT-32
This hybrid MBR only lists the first four partitions, leaving out the ExFAT one entirely.
$ sudo gdisk /dev/disk2
GPT fdisk (gdisk) version 1.0.4
Warning: Devices opened with shared lock will not have their
partition table automatically reloaded!
Partition table scan:
MBR: hybrid
BSD: not present
APM: not present
GPT: present
Found valid GPT with hybrid MBR; using GPT.
Command (? for help): p
Disk /dev/disk2: 15659008 sectors, 7.5 GiB
Sector size (logical): 512 bytes
Disk identifier (GUID): 53C15232-A63B-437C-A5E3-8110AC399A91
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 15658974
Partitions will be aligned on 8-sector boundaries
Total free space is 528005 sectors (257.8 MiB)
Number Start (sector) End (sector) Size Code Name
1 40 409639 200.0 MiB EF00 EFI System Partition
2 409640 4332895 1.9 GiB AF00 HFS
3 4595040 8509791 1.9 GiB AF00 HFSCS
4 8773632 12687359 1.9 GiB 0700 FAT
5 12687360 15656959 1.4 GiB 0700 EXFAT
Nothing special here, except to note that this Disk Utility has set each partition’s GPT label, which is distinct from its name that shows in Finder and is not set by the new Disk Utility (except for “EFI System Partition”).
By the way, I also ran through the above exercises with a 250GB spinning disk without seeing any changes in behaviour.
So, with original Disk Utility:
- 128MB gaps are only created after HFS+ partitions, and not for EF (ESP), AF05 (Apple Core Storage), or AB00 (Recovery HD) partitions
- the presence of a FAT or ExFAT partition on a GPT disk has it create a hybrid MBR for the first four partitions, including the ESP
- FAT and ExFAT partitions get leading gaps so they start on a 1MB boundary, plus a trailing gap if it’s the last partition on the disk
Trying out the new Disk Utility because we have to
The rewritten Disk Utility that shipped with OS X 10.11 is buggy and underpowered, i.e. changing the partition map only lets you create one volume, and because “free space” is not an option, creating custom-sized and -formatted disk partitions is only possible if you format it as HFS+ first, then click [Partition] to add and change partitions. Even then it won’t let you reduce the first to less than 5.4GB, and disks smaller than that can’t be repartitioned at all.
macOS 10.12
Maybe some of its bugs were fixed with the release of macOS 10.12 Sierra, but I still found it not respecting requested partition sizes, not seeing newly inserted disks without a relaunch, not being able to erase a disk on the first attempt, and so on.
Anyway, let’s see how it behaves with formatting a USB drive as GPT with HFS+, FAT, and ExFAT.
$ diskutil list /dev/disk2
/dev/disk2 (external, physical):
#: TYPE NAME SIZE IDENTIFIER
0: GUID_partition_scheme *7.7 GB disk2
1: EFI EFI 209.7 MB disk2s1
2: Apple_HFS HFS 5.4 GB disk2s2
3: Microsoft Basic Data FAT 993.8 MB disk2s3
4: Microsoft Basic Data ExFAT 993.8 MB disk2s4
Note the uneven partition sizes. I had to reformat as GPT+HFS, then split the last third of the disk twice because of the arbitrary limitation on shrinking the first partition.
$ sudo gpt show /dev/disk2
gpt show: /dev/disk2: Suspicious MBR at sector 0
start size index contents
0 1 MBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 10577728 2 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
10987368 262144
11249512 1941040 3 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
13190552 1941048 4 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
15131600 3
15131603 32 Sec GPT table
15131635 1 Sec GPT header
If you check Disk Utility’s output as it does its repartitioning, you’ll see it shrinks the initial HFS+ partition, creates a new 2GB HFS+ partition in the empty space, then shrinks that partition to 1GB and creates a third HFS+ partition; only then does it reformat the latter two volumes as FAT and ExFAT. This behaviour may be a factor in why I was seeing inconsistent results. On this drive only the HFS+ partition has its 128MB trailing gap, but on other drives that same gap would still be there after the FAT or ExFAT partitions. Also note that it didn’t align those two partitions on 1MB boundaries, as the original Disk Utility would.
$ sudo fdisk /dev/disk2
Disk: /dev/disk2 geometry: 941/255/63 [15131636 sectors]
Signature: 0xAA55
Starting Ending
#: id cyl hd sec - cyl hd sec [ start - size]
------------------------------------------------------------------------
1: EE 1023 254 63 - 1023 254 63 [ 1 - 409639] <Unknown ID>
2: AF 1023 254 63 - 1023 254 63 [ 409640 - 10577728] HFS+
3: 0B 1023 254 63 - 1023 254 63 [ 11249512 - 1941040] Win95 FAT-32
4: 07 1023 254 63 - 1023 254 63 [ 13190552 - 1941048] HPFS/QNX/AUX
But, it’s still creating a hybrid MBR.
macOS 10.13
Can macOS High Sierra do any better? I found that Disk Utility’s worst interface quirks had been resolved at this point.
For example, in 10.12 while erasing a drive as GPT+FAT, the dialog failed to convert the default disk name “Untitled” to uppercase, resulting in the volume just being named “U”. The 10.13 version now auto-uppercases text in the Erase dialog, although it still doesn’t in the Partition dialog, which can lead to failure if anything other than uppercase letters are in your FAT volume’s name. (This can happen even if you set a name with lowercase, tab out of the field, and then uppercase the name, because Disk Utility will inexplicably use the first name you typed when it does the partitioning, and then update the volume name at the end of the process.)
It also now allows repartitioning 4GB and smaller drives. Just remember to select View > Show All Devices to see more than just formatted volumes.
$ diskutil list /dev/disk2
/dev/disk2 (external, physical):
#: TYPE NAME SIZE IDENTIFIER
0: GUID_partition_scheme *7.7 GB disk2
1: EFI EFI 209.7 MB disk2s1
2: Apple_HFS HFS 3.7 GB disk2s2
3: Microsoft Basic Data FAT 1.8 GB disk2s3
4: Microsoft Basic Data ExFAT 1.6 GB disk2s4
$ sudo gpt show /dev/disk2
start size index contents
0 1 PMBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 7229904 2 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
7639544 263688
7903232 3612672 3 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
11515904 264192
11780096 3088384 4 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
14868480 263123
15131603 32 Sec GPT table
15131635 1 Sec GPT header
$ sudo fdisk /dev/disk2
Disk: /dev/disk2 geometry: 941/255/63 [15131636 sectors]
Signature: 0xAA55
Starting Ending
#: id cyl hd sec - cyl hd sec [ start - size]
------------------------------------------------------------------------
1: EE 1023 254 63 - 1023 254 63 [ 1 - 15131635] <Unknown ID>
2: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
3: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
4: 00 0 0 0 - 0 0 0 [ 0 - 0] unused
You can see that Disk Utility isn’t bothering to reclaim the 128MB gap left over from when each partition was formatted as HFS+. It is at least adjusting the start and end sectors for FAT and ExFAT partitions to align on 1MB boundaries again. Also note that it’s no longer creating a hybrid MBR.
A similar quirk comes up if you erase a disk as GPT+(Ex)FAT, it’ll add a leading 1MB boundary gap (and won’t format as HFS+ first!):
$ sudo gpt show /dev/disk2
start size index contents
0 1 PMBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 2008
411648 14718976 2 GPT part - EBD0A0A2-B9E5-4433-87C0-68B6B72699C7
15130624 979
15131603 32 Sec GPT table
15131635 1 Sec GPT header
If you then use Partition to change it to HFS+, or select just the volume and use Erase (and so not rewrite the partition table):
$ sudo gpt show /dev/disk2
start size index contents
0 1 PMBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 2008
411648 14456832 2 GPT part - 48465300-0000-11AA-AA11-00306543ECAC
14868480 263123
15131603 32 Sec GPT table
15131635 1 Sec GPT header
it’ll add a trailing 128MB gap, but won’t reclaim that leading gap.
And of course, as of 10.13 we can create APFS volumes, which always require GPT and are actually containers for sub-partitions that all draw from the same storage pool. This is how a 4GB USB drive with a single APFS container shows up. (Another bug: after formatting, Disk Utility displayed the drive as two separate devices, but after an unplug/replug it displayed properly as SanDisk > Container disk3 > apfs.)
$ diskutil list
...
/dev/disk2 (external, physical):
#: TYPE NAME SIZE IDENTIFIER
0: GUID_partition_scheme *4.0 GB disk2
1: EFI EFI 209.7 MB disk2s1
2: Apple_APFS Container disk3 3.8 GB disk2s2
/dev/disk3 (synthesized):
#: TYPE NAME SIZE IDENTIFIER
0: APFS Container Scheme - +3.8 GB disk3
Physical Store disk2s2
1: APFS Volume apfs 905.2 KB disk3s1
Note how an APFS disk takes up two /dev entries.
$ sudo gpt show /dev/disk2
start size index contents
0 1 PMBR
1 1 Pri GPT header
2 32 Pri GPT table
34 6
40 409600 1 GPT part - C12A7328-F81F-11D2-BA4B-00A0C93EC93B
409640 7403440 2 GPT part - 7C3457EF-0000-11AA-AA11-00306543ECAC
7813080 7
7813087 32 Sec GPT table
7813119 1 Sec GPT header
No trailing gap for APFS containers, just enough to align on 4KB blocks. And still no hybrid MBR, obviously.
My testing on macOS 10.14 didn’t find any changes in how its Disk Utility formats disks. I did dig up some more bugs in its interface though, which I’ve filed with Apple. If any bugfixes or other significant changes surface in Disk Utility with macOS 10.15 I’ll update this post.