Add an "uuid" command to generate a new UUID for a portion. The
primary intention is for use for after cloning disks to prevent
collisions.

branches: 1.18.16;
WARNS=6

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

- automatically sync the wedge information unless -n is specified.
- document the general options in the traditional way.
- split the main program into a separate file.

Remove getdisksize support; we either have the ioctls (current/-7) or we
don't (non-netbsd-current/7+tools).

Add a "header" subcommand that displays information about the size of
the disk along with information from the GPT header if it exists.

Factor out the getdisksize() drvctl method, and provide an alternative that
directly uses the disk ioctl's instead of relying on the drvctl device driver
which is currently not mandatory.

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

more toolification

PR/44218 - David Young -- add "type" subcommand to change a partition type

- make gpt_gpt() available for use directly by subcommands
- create new resizedisk disk subcommand for relocating backup GPT

branches: 1.8.2;
Add restore subcommand.

XXX It does not actually work yet. It is being committed now to make
later pullups easier.

Add the backup subcommand. It dumps the contents of the partition
tables as a plist, which is readable by the restore subcommand.

XXX restore subcommand forthcoming

Add two new subcommands, "set" and "unset". These were inspired by
FreeBSD's gpart(8), but the code is all mine. The purpose of these
is to set and unset partition attributes.

Add a resize command. This command was inspired by FreeBSD's gpart(8),
but the code was written by myself.

branches: 1.4.6; 1.4.12;
Support booting from GPT-partioned disks on PC-BIOS-compatible systems.

Much of the work in this commit was done by Mike Volokhov during GSoC 2009.

Enable WARNS=4 by default except for:
dump dump_lfs fsck_ffs fsck_lfs fsdb mount_smbfs
newfs_ext2fs newfs_lfs resize_lfs setkey

branches: 1.2.28;
Make this work under NetBSD.

branches: 1.1.1;
Initial revision

gpt(8): Fix "useless use of <sys/param.h>" antipattern.

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

Pass some lint.

PR bin/58806 Never create a partition with a null guid

If a new partition entry being added doesn't already have a guid,
make one for it.

branches: 1.44.2; 1.44.10; 1.44.12;
Not all things that write to the disk do something that dk(4) will
be interested in. Let individual commands decide if ioctl(DIOCMWEDGES)
should be done. I was conservative and set the flag on any command
that might create/modify/delete partitions in any way.

branches: 1.43.14; 1.43.16;
- add more informational messages
- handle case where last argument is an option not a device (set -l)

CID 1341561: Argument memory leak

CID 1341560: Free arg.

CID 1341548: Check error returns

WARNS=6

eliminate static globals so that commands can be re-used.

hrmm...

i386devel: {245} ./gpt add -b 1000 -s 1000 disk
Segmentation fault (core dumped)

Where's my flyswatter? .......zzzzzzzap

refactor the utf code so that it does not leak memory.

- write a function to set the entity, so that we don't reset to guid.
- entry 0 means create entry.

merge command line parsers and check all memory allocations.

refactor more duplicated code.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

- use gpt_msg to print informational messages (perhaps these should be printed
only with -v)
- don't print any messages with gpt_msg if quiet
- print a message if we didn't reconfigure the wedges

- automatically sync the wedge information unless -n is specified.
- document the general options in the traditional way.
- split the main program into a separate file.

merge the create errors in one place and fix them.

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

minor toolification: need libprop...

branches: 1.24.4;
Explicitly state which partition was added/modified.

For the add and resize subcommands, change the -s option. If there is
no suffix, or the suffix is 's' or 'S', size is in sectors (as before)
otherwise size is in bytes.

Don't attempt to create/resize an unaligned partition if the attempt
to create/resize an aligned one failed. This simplifies the code
and prevents surprises. If the user wants an unaligned partition
in the case where an aligned one fails, they can simply retry the
command without the "-a" option. This change was requested by
wiz@, and after some thought I agree with it.

when calling dehumanize_number() make sure the resulting number is >= 1

Remove FreeBSD specific code as per message to tech-userlevel.

change the "-b blocknr" parameter to accept a "human number"

In the manpage:
- s/-b number/-b blocknr/
- s/-s count/-s sectors/
- s/-p count/-p partitions/

In the program:
- s/-b lba/-b blocknr/
- s/-s lba/-s sectors/

This makes the documentation in the manpage and the program consistent
and makes it more clear what the parameters are. Also, "-s lba" was
just plain wrong since LBA stands for Logical Block Address[ing], and
the -s option didn't represent any kind of address, but rather a size.

delete an errant where it is obvious what the code is doing

When setting a label on the new partition, be sure to set the label
in the secondary partition table as well.

Add two new options to the "add" subcommand:

-a alignment -- attempt to align the start and size of the partition
-l label -- supply a label for the partition

These options were inspired by FreeBSD's gpart(8) command, but the
code was written by me.

Instead of 'no space', say 'not enough space' in error message.

Make pre-initialized partition type uuid_t symbols 'static const'.

This conserves a few hundred bytes of total program size by
allowing the linker to merge duplicates.

Clean up gpt(8) a bit more

uuid_create_nil(3) and uuid_is_nil(3) take a uuid_t*, not a pointer to
an array of bytes merely casted to compatible type.

Additonally, there no need for extra casts and address-of operations when
using le_uuid_*() functions.

Furthermore, le_uuid_*() are identical to uuid_*_le functions in
libc/libuuid, so use those instead on NetBSD.

branches: 1.11.2; 1.11.8;
Use __dead

Clarify the 'dkctl addwedge' usage message after creating a GPT
partition.

UFS is called FFS in the NetBSD case. Display as such and accept as input.

Let 'gpt add' suggest a collect device name so that opendisk(3) called by
dkctl(8) succeeds.

Make manual and usage consistent. No functional changes.

As discussed on tech-userlevel, add NetBSD GPT partition types for
swap, ffs, lfs, raidframe, ccd and cgd. gpt(8) now defaults to
NetBSD ffs if no type is given, but previous FreeBSD types are
still recognized in the DKWEDGE_AUTODISCOVER code.

Fix the usage strings for each command, and print them all for the
"global" usage. It's still a little ugly, but hopefully not as
fantastically confusing as before...

branches: 1.4.4; 1.4.6;
Use strtoll() when converting a 64-bit quantity. This makes "gpt add"
work with an offset of more than 2147483647 sectors. While I'm here,
use strtoul() when converting an unsigned quantity.

branches: 1.3.2;
Add initialization of a variable where gcc for some archs thinks
that it may be used un-initialized. From cursory inspection it is
not obvious that gcc is wrong.

Make this work under NetBSD.

branches: 1.1.1;
Initial revision

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

Pass some lint.

branches: 1.20.8;
When we add "entries", "index", and "sector_size" values to the dictionary,
add them as signed valaues, rather than unsigned (which is how we keep them
in memory). This causes them be serialized in base-10 (rather than base-16,
which is the default for unsigned). This behavior is documented in
prop_number(3). Fixes t_gpt::backup_2part unit test.

Update for proplib(3) API changes.

use __arraycount

- make sure that the utf16 string is padded with 0's where needed.
- since the utf16 string is not 0 terminated, pass the size of the string.

branches: 1.16.8;
CID 1341558: Fix proplib memory leaks

compare to stdout, thanks kre

eliminate static globals so that commands can be re-used.

Allow backup and restore to operate on files.

split out into smaller functions.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

more toolification changes

more toolification

minor toolification: need libprop...

- dump all partitions, not just ones that are in use
- while here, squash a memory leak

It shouldn't be necessary to backup unused partitions, however
the partition GUID is created at the time the GPT is created and
is never changed. It shouldn't matter if the GUID of an unused
partition changes, but there may be some special case where it
does. Since it isn't a big deal to record the unused partitions,
might as well do it.

- record the sector size of the disk
- correct confusion with end_cylinder and end_sector in MBRs

branches: 1.1.4; 1.1.6; 1.1.8;
Add the backup subcommand. It dumps the contents of the partition
tables as a plist, which is readable by the restore subcommand.

XXX restore subcommand forthcoming

gpt(8): Fix "useless use of <sys/param.h>" antipattern.

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

Pass some lint.

Don't use kernel type daddr_t for disk offsets.

branches: 1.32.2; 1.32.10; 1.32.12;
When biosboot is used on an existing wedge and we are switching to
the gpt of the parent device, do not pass the size of the wedge to
gpt_open, it certainly is smaller than the whole device.
Fixes bin/54312.

Allow specifying the partition via -b startsec for the biosboot cmd.

branches: 1.30.4;
use __arraycount

- make sure that the utf16 string is padded with 0's where needed.
- since the utf16 string is not 0 terminated, pass the size of the string.

only do wedges if not the tools version. fixes build on netbsd-4,
and any other place that might have the same ioctl name while having
something not quite right (in this case, cpuid_t in headers not
working properly..)

XXX: pullup-8

branches: 1.27.4;
Add -T timestamp for reproducible builds.

branches: 1.26.2; 1.26.4;
Rename 'a' to set the active PMBR flag to 'A'

PR/51230: Add the ability to set the active flag in the PMBR.

- add more informational messages
- handle case where last argument is an option not a device (set -l)

more leak

WARNS=6

eliminate static globals so that commands can be re-used.

don't advertise stuff we don't support.

refactor the utf code so that it does not leak memory.

merge command line parsers and check all memory allocations.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

provide a copy of the kernel uuid generator (with portable API calls) for
tools.

Fix tools build on !NetBSD.

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

more toolification

minor toolification: need libprop...

PR/47990 - Dr. Wolfgang Stukenbrock -- add "-L <label>" as selector option

no C99isms in tool code

branches: 1.7.4;
Remove FreeBSD specific code as per message to tech-userlevel.

drop trailing whitespace

branches: 1.5.2; 1.5.8;
Use __dead

Fix sign compare bugs

Ajdust bootcode file name from mbr_gpt to gptmbr.bin.

Switch 'gpt biosboot' to support T13 EDD-4 annex A GPT hybrid MBR instead
of a home-grown MBR/VBR hand off protocol.

'biosboot' now will set the legacy BIOS boot flag instead of patching a
GUID into the MBR.

Support booting from GPT-partioned disks on PC-BIOS-compatible systems.

Much of the work in this commit was done by Mike Volokhov during GSoC 2009.

gpt(8): Fix "useless use of <sys/param.h>" antipattern.

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

Pass some lint.

branches: 1.23.26;
PR/51497: Clare: Allow gpt -p <n> create to work.

branches: 1.22.2;
Rename 'a' to set the active PMBR flag to 'A'

PR/51230: Add the ability to set the active flag in the PMBR.

- add more informational messages
- handle case where last argument is an option not a device (set -l)

CID 1341556: Don't leak map info

WARNS=6

eliminate static globals so that commands can be re-used.

refactor more duplicated code.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

provide a copy of the kernel uuid generator (with portable API calls) for
tools.

"0" is not guaranteed to be unique

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

minor toolification: need libprop...

branches: 1.7.4;
Use MBR_PTYPE_ defines from <sys/bootblock.h>.

Clean up gpt(8) a bit more

uuid_create_nil(3) and uuid_is_nil(3) take a uuid_t*, not a pointer to
an array of bytes merely casted to compatible type.

Additonally, there no need for extra casts and address-of operations when
using le_uuid_*() functions.

Furthermore, le_uuid_*() are identical to uuid_*_le functions in
libc/libuuid, so use those instead on NetBSD.

branches: 1.5.2; 1.5.8;
Use __dead

Use less bogus CHS addresses in PMBR.
With the ending head set at 0xff one machine I have will never leave
the initial startup screen if such a disk is present. Additionally,
Wikipedia suggests without citiation that 254 is the maximium allowable
value for the head, and this seems to be the case.

Fix the usage strings for each command, and print them all for the
"global" usage. It's still a little ugly, but hopefully not as
fantastically confusing as before...

branches: 1.2.2; 1.2.4; 1.2.8; 1.2.10;
Make this work under NetBSD.

branches: 1.1.1;
Initial revision

Pass some lint.

branches: 1.14.4;
Fix destroying and moving GPT header also for truncated/extended
images.

Do the 'dkctl makewedges' dance to drop any wedges assoicated the
destroyed GPT.

PR bin/54088: make "destroy" remove the protective MBR too.
We could make this optional if someone comes up with a good reason
to leave it intact.

Ignore -f with the "destroy" command completely (it was effectively
ignored already, so no functional change). Leave it as valid option
for compatibility.

branches: 1.10.16;
eliminate static globals so that commands can be re-used.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

more toolification

minor toolification: need libprop...

branches: 1.4.20;
Use __dead

Fix the usage strings for each command, and print them all for the
"global" usage. It's still a little ugly, but hopefully not as
fantastically confusing as before...

branches: 1.2.2; 1.2.4; 1.2.8; 1.2.10;
Make this work under NetBSD.

branches: 1.1.1;
Initial revision

Remove getdisksize support; we either have the ioctls (current/-7) or we
don't (non-netbsd-current/7+tools).

branches: 1.1.2;
Factor out the getdisksize() drvctl method, and provide an alternative that
directly uses the disk ioctl's instead of relying on the drvctl device driver
which is currently not mandatory.

Minor cleanups.

Add missing "to" (Anon Ymous)

Additions to guid printing and an option to print start/size in hex
(from Anon Ymous)

Add -U option to specify new UUID
Add OpenBSD
from Anon Ymous

branches: 1.82.2;

Revert unintended commit.

Note need to deal with old libarchive.so files in cdrom images when doing
an update build.

gpt.8: fix grammar in a sentence

Apply patch from Thierry Laronde: add missing suffixes for size values
and clarify that they are case independant.

gpt.8: fix some grammar

(The BUGS section seems perhaps a little outdated in its warnings of
potential incompatible changes.)

gpt.8: drop cross-reference to cvs(1)

It's no more useful to reference cvs(1) than, say, openssl(1), and
should be self-evident, anyway.

branches: 1.76.2;

Note that biosboot without -A clears the PMBR "active" flag.
While here, fix some grammar and make the selection options
for biosboot a little clearer.

Ride nia@'s Dd bump.

gpt.8: Clarify that "windows" partition types are also used for FAT32
("msdos" in NetBSD terms), and UDF partitions. Use the same name as
Wikipedia, "Microsoft basic data".

catch up with source rename of fbsd-zfs -> zfs

Fix typo in macro. Sort option descriptions.

Add -H flag to ignore existing MBR (Hybrid MBR/GPT mode).

Exit gracefully when auto-expanding a partition and it is already the
correct size. Add a -q flag to "resize" and "resizedisk" commands to skip
printing warnings in the already resize paths.

Remove outdated note about sysinst restrictions

Add an "uuid" command to generate a new UUID for a portion. The
primary intention is for use for after cloning disks to prevent
collisions.

Allow specifying the partition via -b startsec for the biosboot cmd.

Allow -b to specify a partition for "gpt unset" as well.

Allow -b startsec to specify the partition to change for the set subcommand
as well.
Add another option, -N, as a quick way to remove all attributes.

Make the "show" subcommand accept -b startsec to identify a partition
(very usefull for scripts and other robotic callers).

Add support for specifying the partition to resize via -b startsec
(similar to label and other subcommands). Do not fully add gpt_find
functionality here, as resizing multiple partitions in one go does not
make sense.

branches: 1.63.2;

Spello. (it is "existence").

ispell also says that we should s/parseable/parsable/ but I'm
not sure about that one, so I left it.

I also left a correct spelling that no-one has bothered to mangle!

Bump date for new EXIT STATUS section.

Sort sections.

Make the "gpt header" command return EXIT_FAILURE when no GPT is present.
This helps sysinst to tell a GPT labeled disk from others.
Very lazy version of a change proposed by kre.

add information about how to boot from gpt. mostly taken from the wiki.

Check device parameter to avoid segfaults. Agument synopsis for -l option.

branches: 1.57.2;
Explain what suffixes are accepted when specifying a size.

clarify that alignment is the number of bytes to align to.

Use Fx/Nx.

Fix a typo, and make a couple of minor wording improvements.
I resisted the (very weak) impulse to Americanise some spellings ...

Add note about bootme flag:

The bootme flag is used to indicate which partiotion should be booted
by UEFI boot code.

branches: 1.52.4;
Add -T timestamp for reproducible builds.

Fix spelling of partition.

branches: 1.50.2;
remove dup line

Document the version gpt first appeared.
Replace contractions.
Bump date.

branches: 1.48.2;
Rename 'a' to set the active PMBR flag to 'A'

Fix rendering of biosboot command synopsis.

PR/51230: Add the ability to set the active flag in the PMBR.

Simplify usage for gpt show.

Add a -a flag to the show command, so we can display the full GPT with
all information in one go.

Bump date for previous.
Remove trailing whitespace.

Add listing commands for type, set and unset.
Add help and formatting for set and unset.
Change show to print all the attribute info in one line.

Allow backup and restore to operate on files.

Reflect reality better; we don't handle multiple devices anymore, and
the partitions option is only for the commands that support it.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

- automatically sync the wedge information unless -n is specified.
- document the general options in the traditional way.
- split the main program into a separate file.

Add a "header" subcommand that displays information about the size of
the disk along with information from the GPT header if it exists.

Bump date for previous.

Add some more common partition types.

Remove trailing whitespace.

Allow specifying sector and media size on the command line.

PR/47990 - Dr. Wolfgang Stukenbrock -- add "-L <label>" as selector option

PR/44218 - David Young -- add "type" subcommand to change a partition type

- make gpt_gpt() available for use directly by subcommands
- create new resizedisk disk subcommand for relocating backup GPT

document the new restore subcommand

branches: 1.28.4;
Bump date for previous.

Add the backup subcommand. It dumps the contents of the partition
tables as a plist, which is readable by the restore subcommand.

XXX restore subcommand forthcoming

Fix typo.

Add two new subcommands, "set" and "unset". These were inspired by
FreeBSD's gpart(8), but the code is all mine. The purpose of these
is to set and unset partition attributes.

For the "show -i <entry>" subcommand, print Start and Size both in
terms of number of sectors and bytes.

For the add and resize subcommands, change the -s option. If there is
no suffix, or the suffix is 's' or 'S', size is in sectors (as before)
otherwise size is in bytes.

Don't attempt to create/resize an unaligned partition if the attempt
to create/resize an aligned one failed. This simplifies the code
and prevents surprises. If the user wants an unaligned partition
in the case where an aligned one fails, they can simply retry the
command without the "-a" option. This change was requested by
wiz@, and after some thought I agree with it.

Add two new options:

-g which shows the GUID for a partition
-i which shows all the gory details for a particular option

XXX the output format of -i is subject to change

bump date

In the manpage:
- s/-b number/-b blocknr/
- s/-s count/-s sectors/
- s/-p count/-p partitions/

In the program:
- s/-b lba/-b blocknr/
- s/-s lba/-s sectors/

This makes the documentation in the manpage and the program consistent
and makes it more clear what the parameters are. Also, "-s lba" was
just plain wrong since LBA stands for Logical Block Address[ing], and
the -s option didn't represent any kind of address, but rather a size.

shrunk.

Add a resize command. This command was inspired by FreeBSD's gpart(8),
but the code was written by myself.

update synopsis for add subcommand

Add two new options to the "add" subcommand:

-a alignment -- attempt to align the start and size of the partition
-l label -- supply a label for the partition

These options were inspired by FreeBSD's gpart(8) command, but the
code was written by me.

- markup
- typo; the label command labels partitions, not remove them
- migrate -s isn't applicable to NetBSD
- add information about the space required for migration
- add a "gpt show -l" example

revert typo fix; teach me to blindly take changes from upstream

type fix: accommodate. -> accomodate.

branches: 1.11.2;
Bump date to date of previous change.

fix a path

branches: 1.9.2;
Add an EXAMPLE section showing how to add a swap file system using gpt. Bump date.

branches: 1.8.2;
Minimally document gpt recover. Bump date.

Document biosboot. Xref boot(8) and installboot(8).

List all the aliases supported for GPT partition types.

branches: 1.5.2;
Don't workaround ancient macro argument limit with .Xo/.Xc.

Fix preamble to match order set out by mdoc(7). Discussed with wiz.

branches: 1.3.28;
fix spelling of accommodate; from Zapher.

Serial comma. Add RCS Id. Remove superfluous quotes.

branches: 1.1.1;
Initial revision

gpt(8): Fix "useless use of <sys/param.h>" antipattern.

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

Pass some lint.

When generating timestamp based GUIDs for reproducable builds
- increment timestamp by 100ns for each partition.
- use the standard time-based UUID format (type 1) and don't
pretend it's a random number (type 4).
- make the -T option actually work for the uuid command.

Random GUIDs:
start size index contents
34 1000 1 GPT part - d93ba067-a788-4ce0-99b8-0ead51f00215
1034 2000 2 GPT part - bebba77a-7fdc-4ca0-a1bf-7450aa871d41

d93ba067-a788-4ce0-99b8-0ead51f00215:
Version 4 Random
Variant 2 RFC 4122
Data D9 3B A0 67 A7 88 4C E0 99 B8 0E AD 51 F0 02 15

bebba77a-7fdc-4ca0-a1bf-7450aa871d41:
Version 4 Random
Variant 2 RFC 4122
Data BE BB A7 7A 7F DC 4C A0 A1 BF 74 50 AA 87 1D 41

Timestamp based GUIDs:
start size index contents
34 1000 1 GPT part - 0a524600-8eba-11ef-8000-000000000000
1034 2000 2 GPT part - 0a524601-8eba-11ef-8000-000000000000

0a524600-8eba-11ef-8000-000000000000:
Version 1 Time and MAC based
Variant 2 RFC 4122
Node 00:00:00:00:00:00
Clock 0
Time 2024-10-20T08:05:16.000000.0Z

0a524601-8eba-11ef-8000-000000000000:
Version 1 Time and MAC based
Variant 2 RFC 4122
Node 00:00:00:00:00:00
Clock 0
Time 2024-10-20T08:05:16.000000.1Z

Node (host MAC address) and clock (sequence number incremented whenever
the time went backwards) are left undefined (all zero) for our purpose.

branches: 1.89.2;

Revert unintended commit.

Note need to deal with old libarchive.so files in cdrom images when doing
an update build.

avoid use after free. skip the QUIET check already in gpt_warn().

Be verbose about errors.

When extracting the attributes (to printable form), avoid simply ignoring
any attribute bits which are unknown, print them as a hex value. This
avoids "Attributes: " for most windows filesystem types, which all seem
to have but 63 set (which is supposed to mean "don't assign a drive letter"
which is akin to "noauto" in fstab - except it is set even on partitions
which do get mounted, so must mean something subtly different).

These upper 16 attribute bits are supposed to be file system type speficic
(in practice, they seem to be common to all filesystem types from one vendor)
but we don't have the info (yet anyway) to treat them like that.

ChromeOS seems to treat some of the bits as bit fields containing numeric
values - add #if 0'd (but compile tested) code to deal with those (maybe,
compile tested - but not execution tested) should someone ever get an
environment where these things occur, and could add the missing definitions
to actually test this.

Big-Endian fixes.

Fix destroying and moving GPT header also for truncated/extended
images.

Add -H flag to ignore existing MBR (Hybrid MBR/GPT mode).

gpt_uuid_parse does not set errno ...

Add gpt_change_hdr() similar to gpt_change_ent() for changing
arbitrary header fields.

Add a third argument to the "cfn" function that is an argument to
gpt_change_ent(). The purpose of the third argument is to specify
whether the entry to be changed is a primary GPT entry or a secondary
GPT entry. It is assumed that a secondary GPT entry will always
follow a corresponding primary entry.

This is in preparation for an upcoming change that will require it.

Do not try to close a NULL gpt. Fixes the core dump part of PR bin/54312

Fix three MBR extended partition related bugs:
1) when walking an extended MBR chain, do not assign duplicate indices
2) the pointer to the next MBR may be any of the MBR_PTYPE_EXT* variants
3) the ext MBR chain links are relative to the primary extended partition,
unlike the contained partitions which are relative to the extended MBR
block address.

When changing entries skip those that do not match the -L option.
Should fix PR 53668.

Not all things that write to the disk do something that dk(4) will
be interested in. Let individual commands decide if ioctl(DIOCMWEDGES)
should be done. I was conservative and set the flag on any command
that might create/modify/delete partitions in any way.

branches: 1.74.2; 1.74.4;
Spelling

use __arraycount

- make sure that the utf16 string is padded with 0's where needed.
- since the utf16 string is not 0 terminated, pass the size of the string.

PR/52522: Piotr Meyer: Don't NUL terminate the gpt label name.
XXX: pullup-8

branches: 1.70.4;
Add -T timestamp for reproducible builds.

branches: 1.69.2;
say that you need to destroy it first.

branches: 1.68.2;
PR/51230: Add the ability to set the active flag in the PMBR.

gpt_vwarnx is printf-like.

- add more informational messages
- handle case where last argument is an option not a device (set -l)

pretty up attribute display a bit

Add listing commands for type, set and unset.
Add help and formatting for set and unset.
Change show to print all the attribute info in one line.

Fix resizedisk.

don't print dkctl message for files
fix typo

CID 1341556: Don't leak map info

Add a space.

WARNS=6

annotate map additions with non-allocated memory.

- check errors from map allocation
- make map_resize return -1 instead of 0, and handle errors locally
explaining what's going wrong

Always set NOSYNC if regular file, even if we got the size from the command
line.

- dereferencing NULL is generally bad, found by visual inspection
- automatically set nosync when opening a file to avoid spurious errors

print the old and the new style usage.

refactor the utf code so that it does not leak memory.

merge command line parsers and check all memory allocations.

refactor more duplicated code.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

- use gpt_msg to print informational messages (perhaps these should be printed
only with -v)
- don't print any messages with gpt_msg if quiet
- print a message if we didn't reconfigure the wedges

only update the wedge information if it is supported.

- automatically sync the wedge information unless -n is specified.
- document the general options in the traditional way.
- split the main program into a separate file.

Do division once

Only use the ioctl's if we have them.

fix build breakage for systems that do not have DIOCG{SECTOR,MEDIA}SIZE

merge the create errors in one place and fix them.

Add a "header" subcommand that displays information about the size of
the disk along with information from the GPT header if it exists.

Factor out the getdisksize() drvctl method, and provide an alternative that
directly uses the disk ioctl's instead of relying on the drvctl device driver
which is currently not mandatory.

handle constness better

handle constness

Report the argument instead of (null) when opendisk fails.

Allow specifying sector and media size on the command line.

Fix tools build on !NetBSD.

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

more toolification

minor toolification: need libprop...

PR/44218 - David Young -- add "type" subcommand to change a partition type

- make gpt_gpt() available for use directly by subcommands
- create new resizedisk disk subcommand for relocating backup GPT

branches: 1.28.2;
Add restore subcommand.

XXX It does not actually work yet. It is being committed now to make
later pullups easier.

Add the backup subcommand. It dumps the contents of the partition
tables as a plist, which is readable by the restore subcommand.

XXX restore subcommand forthcoming

Add two new subcommands, "set" and "unset". These were inspired by
FreeBSD's gpart(8), but the code is all mine. The purpose of these
is to set and unset partition attributes.

Use MBR_PTYPE_ defines from <sys/bootblock.h>.

Remove FreeBSD specific code as per message to tech-userlevel.

Only skip past _PATH_DEV if it is actually present.

Add a resize command. This command was inspired by FreeBSD's gpart(8),
but the code was written by myself.

Add two new options to the "add" subcommand:

-a alignment -- attempt to align the start and size of the partition
-l label -- supply a label for the partition

These options were inspired by FreeBSD's gpart(8) command, but the
code was written by me.

Make pre-initialized partition type uuid_t symbols 'static const'.

This conserves a few hundred bytes of total program size by
allowing the linker to merge duplicates.

Clean up gpt(8) a bit more

uuid_create_nil(3) and uuid_is_nil(3) take a uuid_t*, not a pointer to
an array of bytes merely casted to compatible type.

Additonally, there no need for extra casts and address-of operations when
using le_uuid_*() functions.

Furthermore, le_uuid_*() are identical to uuid_*_le functions in
libc/libuuid, so use those instead on NetBSD.

Now that GPT_ENT_TYPE_LINUX_DATA is diffrent from the Windows one, use it.

branches: 1.17.2;
Fix a few straggling BE/LE problems.

Make sure utf-16 strings are written as little endian words.

branches: 1.15.2; 1.15.4;
Use __dead

UFS is called FFS in the NetBSD case. Display as such and accept as input.

Support booting from GPT-partioned disks on PC-BIOS-compatible systems.

Much of the work in this commit was done by Mike Volokhov during GSoC 2009.

fix printf format.

PR/43105: Peter Kerwien: Destroying a GPT partition with dd will cause gpt
destroy / create to fail. When reading the GPT label from the end of the disk
ignore errors if the GPT label at the beginning of the disk was not found.

Initialize device_name before opendisk to avoid
"unable to open device '(null)'" error messages
when the opendisk fails.

Let 'gpt add' suggest a collect device name so that opendisk(3) called by
dkctl(8) succeeds.

branches: 1.8.6;
PR/38094: Robert Millan: support "BIOS Boot" partition type in gpt(8)

branches: 1.7.2; 1.7.4;
As discussed on tech-userlevel, add NetBSD GPT partition types for
swap, ffs, lfs, raidframe, ccd and cgd. gpt(8) now defaults to
NetBSD ffs if no type is given, but previous FreeBSD types are
still recognized in the DKWEDGE_AUTODISCOVER code.

Your friendly neighborhood 64-bit architecture here to remind you that
strlen returns size_t, and printf's precision specifier uses int.

Fix the usage strings for each command, and print them all for the
"global" usage. It's still a little ugly, but hopefully not as
fantastically confusing as before...

branches: 1.4.4; 1.4.6;
Produce more useful error messages.

Use strtoll() when converting a 64-bit quantity. This makes "gpt add"
work with an offset of more than 2147483647 sectors. While I'm here,
use strtoul() when converting an unsigned quantity.

branches: 1.2.2;
Make this work under NetBSD.

branches: 1.1.1;
Initial revision

Pass some lint.

branches: 1.44.4;
Fix destroying and moving GPT header also for truncated/extended
images.

Whitespace

Add -H flag to ignore existing MBR (Hybrid MBR/GPT mode).

Add rcsid

Add gpt_change_hdr() similar to gpt_change_ent() for changing
arbitrary header fields.

Add a third argument to the "cfn" function that is an argument to
gpt_change_ent(). The purpose of the third argument is to specify
whether the entry to be changed is a primary GPT entry or a secondary
GPT entry. It is assumed that a secondary GPT entry will always
follow a corresponding primary entry.

This is in preparation for an upcoming change that will require it.

Not all things that write to the disk do something that dk(4) will
be interested in. Let individual commands decide if ioctl(DIOCMWEDGES)
should be done. I was conservative and set the flag on any command
that might create/modify/delete partitions in any way.

branches: 1.37.2;
Check device parameter to avoid segfaults. Agument synopsis for -l option.

branches: 1.36.2;
- make sure that the utf16 string is padded with 0's where needed.
- since the utf16 string is not 0 terminated, pass the size of the string.

branches: 1.35.4;
allow 0 timestamp

Add -T timestamp for reproducible builds.

branches: 1.33.2; 1.33.4;
PR/51230: Add the ability to set the active flag in the PMBR.

- add more informational messages
- handle case where last argument is an option not a device (set -l)

Add listing commands for type, set and unset.
Add help and formatting for set and unset.
Change show to print all the attribute info in one line.

Fix resizedisk.

don't print dkctl message for files
fix typo

WARNS=6

refactor the utf code so that it does not leak memory.

merge command line parsers and check all memory allocations.

refactor more duplicated code.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

- automatically sync the wedge information unless -n is specified.
- document the general options in the traditional way.
- split the main program into a separate file.

Remove getdisksize support; we either have the ioctls (current/-7) or we
don't (non-netbsd-current/7+tools).

merge the create errors in one place and fix them.

Add a "header" subcommand that displays information about the size of
the disk along with information from the GPT header if it exists.

Factor out the getdisksize() drvctl method, and provide an alternative that
directly uses the disk ioctl's instead of relying on the drvctl device driver
which is currently not mandatory.

handle constness better

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

more toolification changes

PR/44218 - David Young -- add "type" subcommand to change a partition type

- make gpt_gpt() available for use directly by subcommands
- create new resizedisk disk subcommand for relocating backup GPT

branches: 1.11.2;
Add restore subcommand.

XXX It does not actually work yet. It is being committed now to make
later pullups easier.

Add the backup subcommand. It dumps the contents of the partition
tables as a plist, which is readable by the restore subcommand.

XXX restore subcommand forthcoming

Add two new subcommands, "set" and "unset". These were inspired by
FreeBSD's gpart(8), but the code is all mine. The purpose of these
is to set and unset partition attributes.

Remove FreeBSD specific code as per message to tech-userlevel.

Add a resize command. This command was inspired by FreeBSD's gpart(8),
but the code was written by myself.

Clean up gpt(8) a bit more

uuid_create_nil(3) and uuid_is_nil(3) take a uuid_t*, not a pointer to
an array of bytes merely casted to compatible type.

Additonally, there no need for extra casts and address-of operations when
using le_uuid_*() functions.

Furthermore, le_uuid_*() are identical to uuid_*_le functions in
libc/libuuid, so use those instead on NetBSD.

branches: 1.5.6; 1.5.12;
Support booting from GPT-partioned disks on PC-BIOS-compatible systems.

Much of the work in this commit was done by Mike Volokhov during GSoC 2009.

Let 'gpt add' suggest a collect device name so that opendisk(3) called by
dkctl(8) succeeds.

Don't use sizeof(struct gpt_hdr). It will pad the structure on x86-64,
which causes some systems to reject the GPT.

kern/38997

branches: 1.2.18; 1.2.26;
Make this work under NetBSD.

branches: 1.1.1;
Initial revision

When generating timestamp based GUIDs for reproducable builds
- increment timestamp by 100ns for each partition.
- use the standard time-based UUID format (type 1) and don't
pretend it's a random number (type 4).
- make the -T option actually work for the uuid command.

Random GUIDs:
start size index contents
34 1000 1 GPT part - d93ba067-a788-4ce0-99b8-0ead51f00215
1034 2000 2 GPT part - bebba77a-7fdc-4ca0-a1bf-7450aa871d41

d93ba067-a788-4ce0-99b8-0ead51f00215:
Version 4 Random
Variant 2 RFC 4122
Data D9 3B A0 67 A7 88 4C E0 99 B8 0E AD 51 F0 02 15

bebba77a-7fdc-4ca0-a1bf-7450aa871d41:
Version 4 Random
Variant 2 RFC 4122
Data BE BB A7 7A 7F DC 4C A0 A1 BF 74 50 AA 87 1D 41

Timestamp based GUIDs:
start size index contents
34 1000 1 GPT part - 0a524600-8eba-11ef-8000-000000000000
1034 2000 2 GPT part - 0a524601-8eba-11ef-8000-000000000000

0a524600-8eba-11ef-8000-000000000000:
Version 1 Time and MAC based
Variant 2 RFC 4122
Node 00:00:00:00:00:00
Clock 0
Time 2024-10-20T08:05:16.000000.0Z

0a524601-8eba-11ef-8000-000000000000:
Version 1 Time and MAC based
Variant 2 RFC 4122
Node 00:00:00:00:00:00
Clock 0
Time 2024-10-20T08:05:16.000000.1Z

Node (host MAC address) and clock (sequence number incremented whenever
the time went backwards) are left undefined (all zero) for our purpose.

branches: 1.3.12;
Add rcsid

branches: 1.2.12; 1.2.14;
Add -T timestamp for reproducible builds.

branches: 1.1.2; 1.1.4;
Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

Add a couple of GPT partition type GUIDs one might encounter when
fiddling around with RISC-V systems (Linux XBOOTLDR, SiFive BBL).

Pass some lint.

s/reproducable/reproducible/ in comment.

When generating timestamp based GUIDs for reproducable builds
- increment timestamp by 100ns for each partition.
- use the standard time-based UUID format (type 1) and don't
pretend it's a random number (type 4).
- make the -T option actually work for the uuid command.

Random GUIDs:
start size index contents
34 1000 1 GPT part - d93ba067-a788-4ce0-99b8-0ead51f00215
1034 2000 2 GPT part - bebba77a-7fdc-4ca0-a1bf-7450aa871d41

d93ba067-a788-4ce0-99b8-0ead51f00215:
Version 4 Random
Variant 2 RFC 4122
Data D9 3B A0 67 A7 88 4C E0 99 B8 0E AD 51 F0 02 15

bebba77a-7fdc-4ca0-a1bf-7450aa871d41:
Version 4 Random
Variant 2 RFC 4122
Data BE BB A7 7A 7F DC 4C A0 A1 BF 74 50 AA 87 1D 41

Timestamp based GUIDs:
start size index contents
34 1000 1 GPT part - 0a524600-8eba-11ef-8000-000000000000
1034 2000 2 GPT part - 0a524601-8eba-11ef-8000-000000000000

0a524600-8eba-11ef-8000-000000000000:
Version 1 Time and MAC based
Variant 2 RFC 4122
Node 00:00:00:00:00:00
Clock 0
Time 2024-10-20T08:05:16.000000.0Z

0a524601-8eba-11ef-8000-000000000000:
Version 1 Time and MAC based
Variant 2 RFC 4122
Node 00:00:00:00:00:00
Clock 0
Time 2024-10-20T08:05:16.000000.1Z

Node (host MAC address) and clock (sequence number incremented whenever
the time went backwards) are left undefined (all zero) for our purpose.

Add -U option to specify new UUID
Add OpenBSD
from Anon Ymous

branches: 1.21.2;
gpt(8): Make gpt type array and enum match again.

Add cross-referencing comment to reduce the probability of these
getting out of sync again.

Should fix a slew of failing tests since kre's recent change to add
windows-recovery to the array but not to the enum:

sbin/gpt/t_gpt:create_2part
sbin/gpt/t_gpt:migrate_disklabel
sbin/gpt/t_gpt:recover_backup
sbin/gpt/t_gpt:recover_primary
sbin/gpt/t_gpt:remove_2part
sbin/gpt/t_gpt:resize_2part
sbin/gpt/t_gpt:restore_2part

Fail: stdout does not match golden output
--- /usr/tests/sbin/gpt/gpt.2part.show.normal 2023-09-26 15:48:30.000000000 +0000
+++ /tmp/check.sc6ylB/stdout 2023-09-26 23:30:42.388157924 +0000
@@ -3,6 +3,6 @@
1 1 Pri GPT header
2 32 Pri GPT table
34 1024 1 GPT part - EFI System
- 1058 9150 2 GPT part - NetBSD FFSv1/FFSv2
+ 1058 9150 2 GPT part - NetBSD Cryptographic Disk
10208 32 Sec GPT table
10240 1 Sec GPT header

https://releng.netbsd.org/b5reports/i386/commits-2023.09.html#build-2023.09.26.15.47.11

Recognise Windows Recovery partitions.

Now that we use the same UUID for ZFS as FreeBSD, drop the FreeBSD marker
from the description.

adjust gpt_uuid_generate() to accept gpt==NULL

Add a query function (for external code) to enumerate the know guids.
Add recently added VMware GUIDs to the internal enum type.
Fix some short names (bogus + duplicate).

add support to detect the 3 vmware GPT partition types:

- VMKcore (dump partition)
- VMFS
- VMware reserved

this enables vmkcore partitions for netbsd swap or dump, as well
as naming all of them.

branches: 1.15.10; 1.15.12;
allow 0 timestamp

Add -T timestamp for reproducible builds.

branches: 1.13.2; 1.13.4;
Add listing commands for type, set and unset.
Add help and formatting for set and unset.
Change show to print all the attribute info in one line.

WARNS=6

merge command line parsers and check all memory allocations.

branches: 1.10.2;
Add some more common partition types.

Paranoia: choke noisily on EOF from /dev/urandom.

Make gpt(8) generate v4 (random) uuids.

Don't needlessly leak your date/time and MAC address when you edit
the disk.

Read from /dev/urandom rather than calling arc4random or anything,
since this is a tool. (Cygwin seems to have /dev/urandom, but yell
if this breaks the build on your exotic platform because it lacks
/dev/urandom.)

ok apb

provide a copy of the kernel uuid generator (with portable API calls) for
tools.

"0" is not guaranteed to be unique

include <sys/endian.h> for both the non-tools build,
and the tools build on a host where the file exists.

Fixes a build problem on FreeBSD, reported by Herbert J. Skuhra, where
tools/compat's configure detected that be32dec() and friends were
available in the host's <sys/endian.h>, so tools/compat did not provide
its own versions.

It's HAVE_NBTOOL_CONFIG_H, not HAVE_NBTOOLS_CONFIG_H

actually return the uuid when parsing one that is in numeric format

correctly show partition type in the new world order

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

Add a couple of GPT partition type GUIDs one might encounter when
fiddling around with RISC-V systems (Linux XBOOTLDR, SiFive BBL).

Add -U option to specify new UUID
Add OpenBSD
from Anon Ymous

branches: 1.10.2;
gpt(8): Make gpt type array and enum match again.

Add cross-referencing comment to reduce the probability of these
getting out of sync again.

Should fix a slew of failing tests since kre's recent change to add
windows-recovery to the array but not to the enum:

sbin/gpt/t_gpt:create_2part
sbin/gpt/t_gpt:migrate_disklabel
sbin/gpt/t_gpt:recover_backup
sbin/gpt/t_gpt:recover_primary
sbin/gpt/t_gpt:remove_2part
sbin/gpt/t_gpt:resize_2part
sbin/gpt/t_gpt:restore_2part

Fail: stdout does not match golden output
--- /usr/tests/sbin/gpt/gpt.2part.show.normal 2023-09-26 15:48:30.000000000 +0000
+++ /tmp/check.sc6ylB/stdout 2023-09-26 23:30:42.388157924 +0000
@@ -3,6 +3,6 @@
1 1 Pri GPT header
2 32 Pri GPT table
34 1024 1 GPT part - EFI System
- 1058 9150 2 GPT part - NetBSD FFSv1/FFSv2
+ 1058 9150 2 GPT part - NetBSD Cryptographic Disk
10208 32 Sec GPT table
10240 1 Sec GPT header

https://releng.netbsd.org/b5reports/i386/commits-2023.09.html#build-2023.09.26.15.47.11

Add rcsid

Add a query function (for external code) to enumerate the know guids.
Add recently added VMware GUIDs to the internal enum type.
Fix some short names (bogus + duplicate).

branches: 1.7.14; 1.7.16;
remove more duplicated code.

Add listing commands for type, set and unset.
Add help and formatting for set and unset.
Change show to print all the attribute info in one line.

merge command line parsers and check all memory allocations.

A change to the GPT_TYPE list also requires a change to the GPT_TYPE
index constants.

branches: 1.3.2;
provide a copy of the kernel uuid generator (with portable API calls) for
tools.

"0" is not guaranteed to be unique

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

Pass some lint.

branches: 1.10.12;
print little-endian header fields correctly.

branches: 1.9.2; 1.9.4;
Make the "gpt header" command return EXIT_FAILURE when no GPT is present.
This helps sysinst to tell a GPT labeled disk from others.
Very lazy version of a change proposed by kre.

branches: 1.8.8; 1.8.14;
Fix output in an error case

WARNS=6

eliminate static globals so that commands can be re-used.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

correctly print the gpt header revision as a "float"

Add a "header" subcommand that displays information about the size of
the disk along with information from the GPT header if it exists.

Pass some lint.

branches: 1.30.12;
Add a third argument to the "cfn" function that is an argument to
gpt_change_ent(). The purpose of the third argument is to specify
whether the entry to be changed is a primary GPT entry or a secondary
GPT entry. It is assumed that a secondary GPT entry will always
follow a corresponding primary entry.

This is in preparation for an upcoming change that will require it.

Not all things that write to the disk do something that dk(4) will
be interested in. Let individual commands decide if ioctl(DIOCMWEDGES)
should be done. I was conservative and set the flag on any command
that might create/modify/delete partitions in any way.

branches: 1.28.14; 1.28.16;
put back usage from kre

CID 1341564: Arg memory leak

CID 1341563: fix leak

eliminate static globals so that commands can be re-used.

refactor the utf code so that it does not leak memory.

merge command line parsers and check all memory allocations.

refactor more duplicated code.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

minor toolification: need libprop...

PR/47990 - Dr. Wolfgang Stukenbrock -- add "-L <label>" as selector option

branches: 1.14.4;
when calling dehumanize_number() make sure the resulting number is >= 1

Remove FreeBSD specific code as per message to tech-userlevel.

change the "-b blocknr" parameter to accept a "human number"

In the manpage:
- s/-b number/-b blocknr/
- s/-s count/-s sectors/
- s/-p count/-p partitions/

In the program:
- s/-b lba/-b blocknr/
- s/-s lba/-s sectors/

This makes the documentation in the manpage and the program consistent
and makes it more clear what the parameters are. Also, "-s lba" was
just plain wrong since LBA stands for Logical Block Address[ing], and
the -s option didn't represent any kind of address, but rather a size.

Clean up gpt(8) a bit more

uuid_create_nil(3) and uuid_is_nil(3) take a uuid_t*, not a pointer to
an array of bytes merely casted to compatible type.

Additonally, there no need for extra casts and address-of operations when
using le_uuid_*() functions.

Furthermore, le_uuid_*() are identical to uuid_*_le functions in
libc/libuuid, so use those instead on NetBSD.

Use correct spelling of "secondary" in comments.

branches: 1.8.2; 1.8.8;
Use __dead

Fix spelling: s/parition/partition/.

branches: 1.6.12;
Your friendly neighborhood 64-bit architecture here to remind you that
strlen returns size_t, and printf's precision specifier uses int.

Fix the usage strings for each command, and print them all for the
"global" usage. It's still a little ugly, but hopefully not as
fantastically confusing as before...

branches: 1.4.4; 1.4.6;
Use strtoll() when converting a 64-bit quantity. This makes "gpt add"
work with an offset of more than 2147483647 sectors. While I'm here,
use strtoul() when converting an unsigned quantity.

branches: 1.3.2;
add casts

Make this work under NetBSD.

branches: 1.1.1;
Initial revision

Pass some lint.

branches: 1.14.8;
Add -H flag to ignore existing MBR (Hybrid MBR/GPT mode).

Add an "uuid" command to generate a new UUID for a portion. The
primary intention is for use for after cloning disks to prevent
collisions.

Add the -T flag to usage()

branches: 1.11.2; 1.11.4;
Check device parameter to avoid segfaults. Agument synopsis for -l option.

branches: 1.10.4; 1.10.10;
allow 0 timestamp

Add -T timestamp for reproducible builds.

Make gpt(8) work when compiled on a glibc-based OS.

This restores the ability to build amd64 install-image's under Linux.

branches: 1.7.2; 1.7.4;
- add more informational messages
- handle case where last argument is an option not a device (set -l)

WARNS=6

eliminate static globals so that commands can be re-used.

automatically enforce readonly for the commands that don't need to write.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

- automatically sync the wedge information unless -n is specified.
- document the general options in the traditional way.
- split the main program into a separate file.

Fix an incorrect comment.

Exit gracefully when auto-expanding a partition and it is already the
correct size. Add a -q flag to "resize" and "resizedisk" commands to skip
printing warnings in the already resize paths.

if a new map entry doesn't fit, be more verbose about the sizes.

branches: 1.13.8; 1.13.14;
CID 1341556: Don't leak map info

- check errors from map allocation
- make map_resize return -1 instead of 0, and handle errors locally
explaining what's going wrong

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

Add some debugging.

print the current map type.

- use snprintf
- fix bogus logic on map_type

minor toolification: need libprop...

branches: 1.6.4;
CID 1132762: Add error check for negative, to make zero-divide in module
impossible.

Add a resize command. This command was inspired by FreeBSD's gpart(8),
but the code was written by myself.

Add two new options to the "add" subcommand:

-a alignment -- attempt to align the start and size of the partition
-l label -- supply a label for the partition

These options were inspired by FreeBSD's gpart(8) command, but the
code was written by me.

cosmetic fix from drochner@

branches: 1.2.42; 1.2.48;
Make this work under NetBSD.

branches: 1.1.1;
Initial revision

Add rcsid

branches: 1.6.16;
CID 1341556: Don't leak map info

- check errors from map allocation
- make map_resize return -1 instead of 0, and handle errors locally
explaining what's going wrong

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

branches: 1.3.4;
Add a resize command. This command was inspired by FreeBSD's gpart(8),
but the code was written by myself.

Add two new options to the "add" subcommand:

-a alignment -- attempt to align the start and size of the partition
-l label -- supply a label for the partition

These options were inspired by FreeBSD's gpart(8) command, but the
code was written by me.

branches: 1.1.1;
Initial revision

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

Pass some lint.

branches: 1.35.12;
add some more FAT file system types

Make "gpt migrate" work for disks with FAT32 and NTFS partitions

Not all things that write to the disk do something that dk(4) will
be interested in. Let individual commands decide if ioctl(DIOCMWEDGES)
should be done. I was conservative and set the flag on any command
that might create/modify/delete partitions in any way.

branches: 1.32.14; 1.32.16;
Rename 'a' to set the active PMBR flag to 'A'

PR/51230: Add the ability to set the active flag in the PMBR.

- add more informational messages
- handle case where last argument is an option not a device (set -l)

remove more duplicated code.

Add debugging and more to migration

WARNS=6

eliminate static globals so that commands can be re-used.

refactor more duplicated code.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

provide a copy of the kernel uuid generator (with portable API calls) for
tools.

"0" is not guaranteed to be unique

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

more toolification changes

more toolification

minor toolification: need libprop...

branches: 1.14.4;
Use MBR_PTYPE_ defines from <sys/bootblock.h>.

Mirror my changes of src/sbin/gpt/create.c 1.4 in migrate.c;

Use less bogus CHS addresses in PMBR.
With the ending head set at 0xff one machine I have will never leave
the initial startup screen if such a disk is present. Additionally,
Wikipedia suggests without citiation that 254 is the maximium allowable
value for the head, and this seems to be the case.

CID 1125874: Fix memory leak

In a NetBSD disklabel, p_offset is from the beginning of the disk,
not the beginning of the NetBSD partition.

Convert FreeBSD numeric FS_<type>s to labels by request.

- convert FreeBSD FS_<type> to numbers where they don't match NetBSD
- add support for migrating NetBSD disklabel'ed disks (only 7 years late)
- use labels for partition types

Add code for detecting FreeBSD ZFS partitions.

XXX Need to add code for migrating NetBSD disklabel'ed disks.

Make pre-initialized partition type uuid_t symbols 'static const'.

This conserves a few hundred bytes of total program size by
allowing the linker to merge duplicates.

Clean up gpt(8) a bit more

uuid_create_nil(3) and uuid_is_nil(3) take a uuid_t*, not a pointer to
an array of bytes merely casted to compatible type.

Additonally, there no need for extra casts and address-of operations when
using le_uuid_*() functions.

Furthermore, le_uuid_*() are identical to uuid_*_le functions in
libc/libuuid, so use those instead on NetBSD.

branches: 1.5.2; 1.5.8;
Use __dead

Fix the usage strings for each command, and print them all for the
"global" usage. It's still a little ugly, but hopefully not as
fantastically confusing as before...

branches: 1.3.2; 1.3.4; 1.3.8; 1.3.10;
Some of our ports want <sys/param.h> in order to have DEV_BSIZE defined,
needed for LABELSECTOR expansion.

Make this work under NetBSD.

branches: 1.1.1;
Initial revision

gpt(8): Fix "useless use of <sys/param.h>" antipattern.

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

remove getopt call (noted by rillig@)

Pass some lint.

branches: 1.19.8;
PR/55875 - Valentin -- "gpt recover -r" does nothing

Delete do nothing undocumented option, that was there from initial
import. No idea what it was intended to do and there is no longer
an "upstream".

Not all things that write to the disk do something that dk(4) will
be interested in. Let individual commands decide if ioctl(DIOCMWEDGES)
should be done. I was conservative and set the flag on any command
that might create/modify/delete partitions in any way.

branches: 1.17.10; 1.17.12;
Add missing endianes swap to make the "recover" command work on big endian
machines.

branches: 1.16.2; 1.16.4;
- fix setting of hdr_lba_alt which was broken in revision 1.10 on 2015/12/2
- switch to using gpt_last
- PR/51230 -- recreate the PMBR if it was lost

Fix resizedisk.

CID 1341556: Don't leak map info

WARNS=6

eliminate static globals so that commands can be re-used.

don't hard-code the program name

merge duplicated code and check allocations.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

Instruct user to use resizedisk if media size has changed, as
resizedisk will adjust the media size in the headers, whereas
recover simply copies the existing header over the missing one.

XXX recover and resizedisk should probably be merged (even if just
partially, so that recover can properly handle media size changes).
Also, reading in the GPT should probably be centralised so that
error handling can be centralised, and users aren't given misleading
messages (i.e. they aren't told to run recover when they should be
running resizedisk). However, something that can be quickly pulled
up to netbsd-7 was needed, and there isn't time for a major overhaul
or rewrite.

minor toolification: need libprop...

branches: 1.4.20;
Use __dead

Fix the usage strings for each command, and print them all for the
"global" usage. It's still a little ugly, but hopefully not as
fantastically confusing as before...

branches: 1.2.2; 1.2.4; 1.2.8; 1.2.10;
Make this work under NetBSD.

branches: 1.1.1;
Initial revision

Pass some lint.

branches: 1.23.12;
Add a third argument to the "cfn" function that is an argument to
gpt_change_ent(). The purpose of the third argument is to specify
whether the entry to be changed is a primary GPT entry or a secondary
GPT entry. It is assumed that a secondary GPT entry will always
follow a corresponding primary entry.

This is in preparation for an upcoming change that will require it.

Not all things that write to the disk do something that dk(4) will
be interested in. Let individual commands decide if ioctl(DIOCMWEDGES)
should be done. I was conservative and set the flag on any command
that might create/modify/delete partitions in any way.

branches: 1.21.14; 1.21.16;
eliminate static globals so that commands can be re-used.

refactor more duplicated code.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

merge the create errors in one place and fix them.

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

minor toolification: need libprop...

PR/47990 - Dr. Wolfgang Stukenbrock -- add "-L <label>" as selector option

branches: 1.12.4;
when calling dehumanize_number() make sure the resulting number is >= 1

Remove FreeBSD specific code as per message to tech-userlevel.

change the "-b blocknr" parameter to accept a "human number"

In the manpage:
- s/-b number/-b blocknr/
- s/-s count/-s sectors/
- s/-p count/-p partitions/

In the program:
- s/-b lba/-b blocknr/
- s/-s lba/-s sectors/

This makes the documentation in the manpage and the program consistent
and makes it more clear what the parameters are. Also, "-s lba" was
just plain wrong since LBA stands for Logical Block Address[ing], and
the -s option didn't represent any kind of address, but rather a size.

Clean up gpt(8) a bit more

uuid_create_nil(3) and uuid_is_nil(3) take a uuid_t*, not a pointer to
an array of bytes merely casted to compatible type.

Additonally, there no need for extra casts and address-of operations when
using le_uuid_*() functions.

Furthermore, le_uuid_*() are identical to uuid_*_le functions in
libc/libuuid, so use those instead on NetBSD.

Use correct spelling of "secondary" in comments.

branches: 1.6.2; 1.6.8;
Use __dead

Make manual and usage consistent. No functional changes.

Fix the usage strings for each command, and print them all for the
"global" usage. It's still a little ugly, but hopefully not as
fantastically confusing as before...

branches: 1.3.4; 1.3.6;
Use strtoll() when converting a 64-bit quantity. This makes "gpt add"
work with an offset of more than 2147483647 sectors. While I'm here,
use strtoul() when converting an unsigned quantity.

branches: 1.2.2;
Make this work under NetBSD.

branches: 1.1.1;
Initial revision

gpt(8): Don't assume transclusion of <sys/endian.h> via <sys/types.h>

Prompted by PR pkg/59839

Pass some lint.

branches: 1.25.8;
Exit gracefully when auto-expanding a partition and it is already the
correct size. Add a -q flag to "resize" and "resizedisk" commands to skip
printing warnings in the already resize paths.

Add support for specifying the partition to resize via -b startsec
(similar to label and other subcommands). Do not fully add gpt_find
functionality here, as resizing multiple partitions in one go does not
make sense.

Not all things that write to the disk do something that dk(4) will
be interested in. Let individual commands decide if ioctl(DIOCMWEDGES)
should be done. I was conservative and set the flag on any command
that might create/modify/delete partitions in any way.

branches: 1.22.14; 1.22.16;
Kill newline

WARNS=6

remove extra assignment

eliminate static globals so that commands can be re-used.

- check errors from map allocation
- make map_resize return -1 instead of 0, and handle errors locally
explaining what's going wrong

refactor more duplicated code.

Refactor the command dispatching and help printing logic.

Firs step of refactoring, remove all globals, factor out some common code,
handle alternate usage but not advertise it.

- use gpt_msg to print informational messages (perhaps these should be printed
only with -v)
- don't print any messages with gpt_msg if quiet
- print a message if we didn't reconfigure the wedges

- automatically sync the wedge information unless -n is specified.
- document the general options in the traditional way.
- split the main program into a separate file.

merge the create errors in one place and fix them.

Add internal uuid support, since the linux+macos versions of the library
are different than than *bsd ones, and others might not have it at all.

remove compat defines.

minor toolification: need libprop...

branches: 1.8.4; 1.8.6; 1.8.8;
Explicitly state which partition was added/modified.