Copy stable/9 to releng/9.3 as part of the 9.3-RELEASE cycle.

Approved by: re (implicit)
Sponsored by: The FreeBSD Foundation

MFC r234122:
It seems that libdisk(3) incorrectly sets d_secperunit value.
Automatically fix it like GEOM_PART_BSD does.

PR: bin/165789

MFC r227231:
To be in sync with GEOM_PART_BSD limit the maximum number of supported
partitions to 20.

MFC r227248:
bsdlabel(8) could automatically fill many of disklabel's deprecated
fields, but user could specify some of those fields when edits disklabel
with `bsdlabel -e`. But without -A flag these fields might be
overwritten with default values from the virgin disklabel.
So, don't overwrite such fields if they are not zero. Also add checks
to prevent creating disklabel with less than DEFPARTITIONS and more
than MAXPARTITIONS partitions.

PR: bin/162332
Tested by: Eugene Grosbein

MFC r227262:
Remove unneeded checks.

MFC r227270:
Add recommendation to use gpart(8) when user tries write disklabel
or bootcode to already opened provider.

MFC r227296:
Fix multi-line comment formatting.

Approved by: re (kib)

Copy head to stable/9 as part of 9.0-RELEASE release cycle.

Approved by: re (implicit)

Widen fields that display partition offset/length.

This makes partitions between 50GiB and 2TiB (16TiB for 4k drives) print
correctly aligned.

While here, fix type of secsize. g_sectorsize() returns ssize_t, don't
store this in an unsigned var. Bump WARNS to 6.

MFC after: 4 weeks

Remove dead code in bsdlabel depending on __alpha__

Closing file descriptors when it's done

Fix some more warnings found by clang.

Don't attempt to write label with GEOM_BSD based method if the class is
not available. This improves error reporting when bsdlabel(8) is unable
to open a device for writing. If GEOM_BSD was unavailable, only a rather
obscure error message "Class not found" was printed.

PR: bin/58390
Reviewed by: ae
Discussed with: marcel
MFC after: 1 month

- Check that strtoul(3) succeeds to convert the entire string in a few
places.
- In getasciilabel(), set the disk type only when a valid type is given.

PR: bin/86765
MFC after: 2 weeks

- Don't assign the return value from read(2) to a variable of type
int.
- Use errx(3) instead of err(3) to print the error message on short
reads in readlabel(). errno won't be set on short reads which can
easily occur here due to the fixed size read request.

PR: 144307
Reviewed by: bde

Use setresuid/setresgid to drop privileges

Remove the dependency on the kernel -- in particular the gctl request to
the GEOM_BSD class -- to translate the absolute offsets in the label to
relative ones. This makes bslabel(8) work correctly with GEOM_PART and
also when the BSD label is nested under arbitrary partitioning schemes.

Inspired by: Eygene Ryabinkin <rea-fbsd@codelabs.ru>
Approved by: re (kib)

- A call to close(2) might overwrite errno and thus give a wrong error message
on g_providername failure.

Suggested by: pjd
Approved by: pjd (mentor)

- Improve error message given on g_providername call failure.
- While there, make error messages consistent with the rest.

Approved by: kib (mentor)

- Make bsdlabel use libgeom to determine provider name, device path, the media
size and the sector size.
- Fix a bug where bsdlabel would try to read a regular file using the geom_bsd
class.

Quick review by: phk
Approved by: pjd (mentor)

Remove comment about "-r" flag from readlabel. "-r" is a no-op.
The is comment is left over from the old disklabel command.

Reviewed by: phk

Allow bsdlabel to operate on labels that have at most 26 partitions
by virtue of there not being any (lower-case) letters avaliable for
more partitions.

o '-s' flag was killed in rev. 1.75. Clean getopt(3).

PR: bin/104616
Submitted by: Oliver Fromme
MFC after: 1 week

Attempt to improve the logic for automatically sizing partitions
to take into account the new default of starting the first partition
after the boot blocks instead of at sector 0. If you used automatic
sizing when the first partition did not start at 0, you would get
an error that the automatically sized partition extended beyond the
end of the disk.

Note that there are probably still many more complex cases where
automatic sizing and placement will not work (e.g. non-contiguous
or out of order partitions).

Fix the derivation of the GEOM name from the specified device name by
complementing the existing special case of a not existing /dev prefix
with the recognition of an already existing /dev prefix.

This implicitly solves the following two issues related to working on
GEOM devices /dev/foo/bar (which have the GEOM provider name "foo/bar")
with the expected commands like "bsdlabel /dev/foo/bar":

1. the error "Geom not found" when trying to write or edit the BSD
label (because previously the incorrect GEOM name "bar" instead of
"foo/bar" was derived from "/dev/foo/bar").

2. the multiple times reported "magically introduced" partition offset
of 63 blocks and the resulting errors like "partition extends past
end of unit" and "partition c doesn't start at 0!".

This implicitly resulted because bsdlabel(8) determines the "MBR
offset" via GEOM and (intentionally) silently falls back to an offset
of 0 if it could not be queried (which is the case if the name was
incorrectly derived).

Usually (at least on PCs) the offset for the first slice is 63 blocks
and bsdlabel(8) automatically subtracts them from the absolute
offsets in the read on-disk BSD label, resulting in the display of an
effective offset of 0. If the GEOM query fails, the assumed offset of
0 is subtracted and an incorrect effective offset of 63 is displayed
and tried to be worked upon.

Reviewed by: pjd
MFC after: 1 week

The disk labels generated by bsdlabel can no address more than
0xffffffff sectors. Document this limit and avoid installing bogus
labels on disks with more sectors.

Allowing the installation of labels addressing as much of the disk as
possiable may be a useful addition in some situations, but this was easy
to implement and should reduce confusion.

PR: bin/71408

The multiplier prefix is actually a multiplier suffix.

Use fallthrough to simplify the multiplier logic; optimistically add
support for the T multiplier; improve the error message for unrecognized
multipliers.

Fix some whitespace issues, and move a curly brace out of an #ifdef to
avoid confusing auto-indenting editors.

Implement a '-f' flag to teach bsdlabel to work on files instead of
disk partitions.

When installing boot blocks into an Alpha BSD label, setup the location,
length, and flags fields at the end of the SRM boot sector so that SRM can
find the bootstrap code. This fixes bsdlabel -m alpha to generate bootable
disklabels.

Reviewed by: phk

Only automatically create an 'a' partition when there is nothing
but a 'c' partition.

Remove the hardcoded default block/frag/cpg values from bsdlabel
and the logic for setting them according to the partition size.
Instead, unspecified filesystem values are left at 0 so that newfs
will use its own defaults. It just caused confusion to have the
defaults duplicated in two different places.

Reviewed by: phk

When we initialize a disk with a virgin label, create also an 'a'
partition which starts after the bootstrap area and fills the entire
disk.

Augh! Fix the sparc64 build:

If we don't have a default label location for the compiled architecture,
insist that a -m <architecture> option is specified.

Sanitize setting of labeloffset and labelsector.

Give ia64 the exact same semantics as i386 with respect to non-512
byte sector devices.

Unbreak ia64. 'nuff said.

Fix sectorsize != 512 on i386 and pc98. Add test cases for same.

Simplify the GEOM OAM api: Drop the request type, and let everything
hinge on the "verb" parameter which the class gets to interpret as
it sees fit.

Move the entire request into the kernel and move changed parameters
back when done.

Protext copyright[].

Add pc98 archtecture entry.

Approved by: re/rwatson

Before reading an ascii label, initialize with defaults so that
getasciipartspec() has a sectorisize in case it needs one.

Approved by: re/jhb

Sigh, this shows just how much one can be conditioned my the environment:

Just because we for the last ten years have fought for every byte
in the boot code on i386, doesn't mean that other architectures could
not actually have space to spare there.

Remore debugging message.

Add transparent handling of mbroffset for backwards compatibility.

More axe-work:

Hide all the historical fields of the label, unless people ask for them with -A,
set them to intelligently chosen defaults otherwise.

Distill the manual page to remove inaccuracies, misundertandings and obsolete
information. It can probably still be done better but now at least it is
not misinforming people.

Some minor remodelling with a large axe.

Use new geom.ctl based OAM instead of ioctls.
Various cleanup.

bsd_disklabel_le_dec() takes an extra argument now.

Vastly simplify architecture handling: Use properties of label as
recorded in global variables, rather than checks on the architecture.

Drop horribly code to handle MBR/PC98's embedded in the BSD label area.
If you need to have an MBR or PC98 on your disk, you should not overlap
it with a BSDLABEL, if you don't need it, this code is nothing but trouble.

Default the location to the compiled for architecture if no -m arg specified.

Add three global variables which contain the location, size and a flag
for the alpha checksum, and set them depending on the specified architecture

Don't look for disklabels every 16 bytes, look the only place they should
be for the current architecture.

Always read the label from the raw disk and decode it into struct
disklabel rather than trust a cast from random addresses.

When writing to the raw disk, encode the label properly.

Remove the well-intentioned, but ill thought out check which prevents us
from dd(1)'ing the boot code off one drive and have bsdlabel write it
on another.

First scrub of s/disklabel/bsdlabel/

bsdlabel.5 deliberately exempted, its contents looks less than useful.

Don't die of SIGSEGV on a missing fstype field
in a saved disklabel file.

MFC after: 1 week

Teach disklabel(8) about different hardware architectures.
This is aimed at creating floppies during cross-releases.
For different endianness machines, a tool like bswapfs(8)
is necessary to make the generated floppies readable on
the target machine. While here, fixed unaligned access
on Alphas.

Tested on: i386, alpha

Deal with vestiges of d_boot[01].

If we fail to open O_RDWR, try opening O_RDONLY, if we use ioctls
to fiddle the disk we can get away with it.

Try to use DIOCBSDBB to write boot code.

IA64 still needs to be able to run on MBR+BSD combination, so put disklabel
back.

Pointed out by: peter

Drop the silly notion that i386 has two bootstrap files now that sys/boot
creates a single file named just "boot".

Apart from the fact that the option "-s" is now gone and that "-b" should
be pointed at /boot/boot instead of /boot/boot1, this patch should be
a no-op.

Remove #if checks for NUMBOOT==0, it's not relevant for any architecture
and if we get such an architecture, we can just avoid using the relevant
options.

Offer better advice in #error.

Remove yet a usage message about -N/-W

Remove ia64 and powerpc #ifdefs: this is i386/pc98/alpha only code.

Remove #ifdef __sparc64__ code. GEOM uses native SUN labels on this arch,
so no hacks are needed.

Fix typo in last commit.
Rearrange bits in writelabel() a bit for improved readability.

Remember to remove -N and -W from usage.

Don't fiddle write-protect status of disklabel, it's a discontinued feature.

Retire the "-N" and "-W" options which implemented write-protection
for the disklabel: This facility is OBE.

First of all, we cannot sensibly implement this in a properly stacked
environment.

Second, if we did, it would confuse the heck out of users who
wouldn't be able to "start from scratch" by dd(8)'ing /dev/zero
onto /dev/da0.

Third, the offered protection is not comprehensive: no other software
would respect it.

Fourth and finally, the disklabel is already protected against
tampering if it controls open partitions.

Uselessness of these options discussed with: peter

Clear part_set, part_size_type and part_offset_type before parsing the
ascii label; otherwise checklabel() might report spurious errors because
of leftovers from a previous pass.

Reviewed by: bde
MFC after: 1 week

Rename the dos_partition structure for pc98 to pc98_partition.

Relax partition overlap check to allow Vinum partitions to overlap
other partitiosns. This is necessary when migrating conventional
partitions to Vinum and was broken by recent more stringent overlap
checks. This is arguably the wrong way to do it. A better method
would be to have the loader understand a subset of Vinum partitioning
and allow an install directly to Vinum, but until then, this is the
best we have.

Reviewed by: jhb
Approved by: re (rwatson)

Everything in the disklabel is unsigned so make all the input routines
take unsigned values.
his allows one to label disk with the number of blocks > 31 bits
(though less then 32 bits)

e.g.
# size offset fstype [fsize bsize bps/cpg]
c: 3125755904 0 unused 0 0 # (Cyl. 0 - 194569*)
d: 3125755840 64 unused 0 0 # (Cyl. 0*- 194569*)
which is needd to test UFS2

Fix to support pc98.

Don't be noisy if DIOCWLABEL fails, it's probably because we're writing
the first label to the raw disk.

Sponsored by: DARPA & NAI Labs.

Split MBR and PC98 on-disk sliceformats out from disklabel.h, step 1:

Peter had repocopied sys/disklabel.h to sys/diskpc98.h and sys/diskmbr.h.

These two new copies are still intact copies of disklabel.h and
therefore protected by #ifndef _SYS_DISKLABEL_H_ so #including them
in programs which already include <sys.disklabel.h> is currently a
no-op.

This commit adds a number of such #includes.

Once I have verified that I have fixed all the places which need fixing,
I will commit the updated versions of the three #include files.

Sponsored by: DARPA & NAI Labs.

Construct new disklabels based on the medias stated parameters in
userland, rather than expect all possible GEOMetries to know about
BSD disklabels.

Sponsored by: DARPA & NAI Labs

s/filesystem/file system/g as discussed on -developers

Fix more breakage due to FSTYPENAMES/DKTYPENAMES split.

Also submitted by: David Wolfskill <david@catwhisker.org>

Apply __FBSDID(); fix local variable(names) that stomp on global
definitions; fix some const strings; fix some signedness issues.

Oops, unbreak parsing of the `type' field in getasciilabel(). I had
changed a `goto' to a `continue' in revision 1.52, but it continued
the wrong loop.

Noticed by: bde

correct comment (replace i386 with __i386__)

Fix some serious brain damage in the default block/frag/cpg parameters
that are used if none at all are specified for a partition. Don't
keep replaying the last field if we run out of fields when processing
a line. Use a 8:1 frag:block ratio for both defaults.

More work here is required. I think disklabel should not attempt
to choose default filesystem parameters, and instead let newfs pick
any defaults if required.

PR: i386/38703
Reported by: Martin Kraemer <Martin.Kraemer@Fujitsu-Siemens.com>

Split out the code for parsing the partition specification lines
into a separate function to avoid some over-long lines.

Spread the word of PowerPC.

Retire the bogus uses of the disklabel field d_sbsize and begin to
initialize it to zero so we don't have to have everbody and their
aunt including FFS specific header files.

Sponsored by: DARPA & NAI Labs.

Get us to WARNS=2 with a bit of constification and some printf fiddling.

Pick up the c-partitions magicness from sys/disklabel.h instead
of defining our own magicness for it, remove trivial comment.

Sponsored by: DARPA and NAI Labs.

Nuke a totally pointless optional debug option

Sponsored by: DARPA and NAI Labs.

Remove remaining traces of d_boot[01]

Kill only usage of the undocumnted and unuse d_boot[01] fields of
struct disklabel.

Sponsored by: DARPA & NAI Labs.

Remove stray register

o remove __P
o Use ANSI function definitions
o const poison
o remove register

Add complete Sparc64 support.
Submitted by: tmm

Dike out vax support.

* Support the Sparc64.
* Do not default to any particular platform. Require that we explicitly
support a particular platform.

A bit of premptive GEOM POLA magic: If we don't get a virgin disklabel
from /dev/food0, then try from /dev/f00d0c, in strange cases this work.

Silence non-constant format string warnings by marking functions
as __printflike()/__printf0like(), adding const, or adding missing "%s"
format strings, as appropriate.

MFC after: 2 weeks

Fix the type of the NULL arg to execl()

Idea from: Theo de Raadt <deraadt@openbsd.org>

Nuke unused variables.

Say "add -r" rather than 'use -r' since the former is more correct.
The latter implies to many people that they use only -r, which is
incorrect.

First attempt to not overflow in disk space calculations. Use off_t
for the size variable used to calculate the size of the partition.
Also use ULL suffix for constants to ensure that we use 64 bit math.

Allow for easier configuration when using disklabel. A sample
being:

# size offset fstype [fsize bsize bps/cpg]
a: 400M 0 4.2BSD 4096 16384 75 # (Cyl. 0 - 812*)
b: 1G * swap
c: * * unused
e: 204800 * 4.2BSD
f: 5g * 4.2BSD
g: * * 4.2BSD

These patches are the original work of Randell Jesup, and
I believe Matt Dillon, with additional work by Warner Losh.
Please let me know if I've left someone out.

Incorporated into this is the fix for PR bin/22727.

This patchset still has style issues and a possible problem on
large disks. However, it was a agreed to get these committed before
performing major surgery on them.

PR: bin/22727
Submitted by: Randell Jesup <rjesup@wgate.com>

Add support for virgin disklabels

Submitted by: dillon

Turn dkcksum() into an __inline function.

Change its type to u_int_16_t.

A local WIP snook in rev 1.31.

Noticed by: bde

Found more places where 'r' was being prepended to the device name.

Add include of errno.h where needed, remove extern int errno where not.

These commits were inspired by a similar commit to netbsd.

Don't prepend "r" for the raw device anymore.

10 X's for mkstemp(), and don't redefine _PATH_TMP

Remove duplicate declarations for two internal functions.
Remove delcaration for getenv(), we already get it from stdlib.h

$Id$ -> $FreeBSD$

Fix a gcc stupidity where it thought a variable was being used uninitialized

Add a case for UNSPEC which is in order by the enum definition, but out of
order alphabetically.

Look for boot blocks in new default location.

Oops, don't quite use RAW_PART instead of RAWPARTITION. Use 'a' + RAW_PART
in both places.

Updated the error message for EXDEV to match recent kernel fixes.
Fixed nearby indentation.

Use RAW_PART instead of RAWPARTITION.

Attempt to fix my breakage of the alpha makebootarea() in rev.1.19.
The previous attempt just converted compile time breakage to runtime
breakage.

Try to make this compile on both alpha and i386

Fixed style bugs in previous commit.

Added some comments on #endifs.

Teach disklabel how to install a bootstrap on an alpha with SRM console.

Fixed bugs in `disklabel -R...':
- nonstandard sector sizes didn't work because the sector size in the
ASCII label was not read before a (default, wrong) sector size was used.
- the exit status was 0 after exiting early due to an invalid ASCII label.

Abort `disklabel -B...' if the secondary bootstrap doesn't fit.

Clear d_boot0 and d_boot1 in the virgin label. These are overlaid by
d_packname in in-core labels, so they are garbage if d_packname is
initialized in the dummy label for the whole disk. dsopen() will soon
initialize d_packname to "fictitious" if it is not already initialized.

Fixed nearby error handling. Rev.1.7 apparently confused Perror()
with perror().

Fixed printf format errors.

Use warnx()+fprintf() to handle multi-line messages. Correct K&R support and
KNF continuation indent rule.
Requested by: Bruce.

Add section number to .Xr. Use of .Nm. Typo. Add rcsid. Remove unused
#includes. Use err(3).

Add Id. Use err(3). Costmetic in usage string.

compare return value from getopt against -1 rather than EOF, per the final
posix standard on the topic.

Use mkstemp rather than mktemp to prevent a small race.

Obtained from: OpenBSD

Moved initialization of defaults for the label for the whole disk from
disklabel(8) to the kernel (dsopen()). Drivers should initialize the
hardware values (rpm, interleave, skews). Drivers currently don't do
this, but it usually doesn't matter since rotational position stuff is
normally disabled.

Try to preserve the fdisk table in the primary bootstrap, should it
already exist.

Mention the cases where this doesn't work in the BUGS section of the
man page.

Finally implement the "auto" disktype. It attempts to get all the
required information from the driver, and produce a virgin disklabel
for it. The latter might be further edited with `disklabel -e' to
satisfy the user's need.

The magic sequence is:

disklabel -r -w sdX auto
disklabel -e sdX

Declare functions, resolve printf-format warnings.

Remove trailing whitespace.

Don't clobber d_secperunit in `disklabel -e'. `disklabel -e' replaces all
the values that it doesn't print by defaults. This seems wrong. I want
to be able to see the total number of sectors more than edit it. The
default d_secperunit of (sectors/track * tracks/cylinder * cylinders) is
bogus if sectors/track is only an approximation and more bogus if
sectors/track and tracks/cylinder are dummy values such as 4096 and 1
to defeat ufs's pessimizations.

Allow for the specification of bsize, fsize, and bps (blocks per segment)
for LFS filesystems.

disklabel will dump core if you give it too few numbers to process
Obtained from: NetBSD misc/490 - John Kohl

BSD 4.4 Lite sbin Sources

Note: XNSrouted and routed NOT imported here, they shall be imported with
usr.sbin.