WARNS=5 is the default in sbin, don't need to set it explicitly.

branches: 1.16.2;
Set WARNS=5. No warnings/errors were found with a test compile.

branches: 1.15.6;
Add rump client support for module utilities. It should be noted
that while modstat and modunload and builtin modules work exactly
the same as in the host case, modload loads file system kernel
modules from the rump kernel namespace. By default, archs which
have rump support for the kernel kernel ABI have the host module
directory mapped into the rump kernel namespace at the same location
(/stand/...). Therefore, if the *host* module directory is populated,
"rump.modload foo" will work as expected. Otherwise, RUMP_MODULEBASE
can be used to point to the module directory.

Remove LKMs and switch to the module framework, pass 1.

Proposed on tech-kern@.

Add the ability to specify load-time properties for modules. This adds
three new flags to the new modload utility: -b, -i and -s, which are used
to pass boolean, integer and string parameters respectively.

branches: 1.12.2; 1.12.4;
Always test MKMODULAR with != "no". Requested by mrg@.

Reverse MKMODULAR checks.

Recognize MKMODULAR and build the appropriate tool.

Pull in my modules code for review/test/hacking.

branches: 1.8.10; 1.8.12;
Move WARNS=3 to the Makefile.inc, and add a little const to the remaining
programs that did not compile before.

add ELF support, mostly from joda@pdc.kth.se (Johan Danielsson) in PR#6149,
updated for the symtab support for a.out (non for ELF yet) and other cleanup
by myself. should work on the alpha as well as other ELF ports...
remove the ld(1) output by default, rather than leaving little binary turds
across the filesystem...

convert to new RCS Id conventions; reduce my headache

specify man pages the new way.

do not -DKERNEL please

Don't reset CFLAGS.

add LKM man pages

add support for terry lambert's loadable kernel modules.
needs a bit of cleanup, but overall: SLICK!

Remove LKMs and switch to the module framework, pass 1.

Proposed on tech-kern@.

branches: 1.6.26;
remove obsolete "register" declarations.

it's not correct to use MAXPATHNAME-sized buffer for link command, as
pointed out by Klaus Klein (original idea was that it should hold at least
the pathname ...); instead, let the ELF and a.out backends allocate memory
for the link command, and get rid of the fixed size buffer altogether

Add support for using an optional linker script (/usr/lkm/ldscript if
found, and overridable by the -T option).

Use "type *" instead of "type*".

ANSIfy, KNF, declare externs in headers only.

add ELF support, mostly from joda@pdc.kth.se (Johan Danielsson) in PR#6149,
updated for the symtab support for a.out (non for ELF yet) and other cleanup
by myself. should work on the alpha as well as other ELF ports...
remove the ld(1) output by default, rather than leaving little binary turds
across the filesystem...

Remove LKMs and switch to the module framework, pass 1.

Proposed on tech-kern@.

branches: 1.19.20;
Replace complicated test for ELFSIZE 64 with #ifdef _LP64 which is
the netbsd convention. Rig the old test temporarily to a #error
sanity check.

Check for allocation failures in malloc, calloc, realloc, asprintf, and
vasprintf and try to handle them.

Move WARNS=3 to the Makefile.inc, and add a little const to the remaining
programs that did not compile before.

it's not correct to use MAXPATHNAME-sized buffer for link command, as
pointed out by Klaus Klein (original idea was that it should hold at least
the pathname ...); instead, let the ELF and a.out backends allocate memory
for the link command, and get rid of the fixed size buffer altogether

Don't compare an integer to NULL.

Skip empty sections. From Allen Briggs.

Add support for using an optional linker script (/usr/lkm/ldscript if
found, and overridable by the -T option).

Use "type *" instead of "type*".

ANSIfy, KNF, declare externs in headers only.

various spelling

Only provide sym_size and sym_symsize if symbols are actually going
to be loaded.
This also avoids the annoying 'Unable to find symbol table slot
for xxx' kernel message when module is unloaded.

make sure string_table is initialized to NULL in read_shstring_table(),
so that NULL is returned if .strtab is not found

PR/14498: Wesley Chen: Support symbol loading on elf lkms.

Use ELFSIZE 64 for x86-64.

Do the last fix correctly: round the beginning of the .data segment,
otherwise an unaligned address gets passed to the linker. (which is
rounded there, so this is harmless)
XXX how about passing "-N" and killing all these hacks?

make sure that elf_mod_sizes() uses data_hole size aligned to section alignment
this fixes a problem where elf_mod_sizes() would report size which would
be different (smaller) that the actual size of LKM code to be loaded in some
cases

Reviewed by: Johan Danielsson

branches: 1.3.6;
Update to match new SVR4-style definition names in <sys/exec_elf.h>.

branches: 1.2.2;
Don't try to play games with the BSS; just make it work.

add ELF support, mostly from joda@pdc.kth.se (Johan Danielsson) in PR#6149,
updated for the symtab support for a.out (non for ELF yet) and other cleanup
by myself. should work on the alpha as well as other ELF ports...
remove the ld(1) output by default, rather than leaving little binary turds
across the filesystem...

Update for proplib(3) API changes.

Drop main() prototype.

More friendly error messages for modload(8) and modunload(8)

Tested on NetBSD/amd64.

From Christian Koch (cfkoch@) of EdgeBSD; thanks!

Don't pass NULL to prop_dictionary_set.
Coverity CID 275196.

branches: 1.14.6; 1.14.12;
Add rump client support for module utilities. It should be noted
that while modstat and modunload and builtin modules work exactly
the same as in the host case, modload loads file system kernel
modules from the rump kernel namespace. By default, archs which
have rump support for the kernel kernel ABI have the host module
directory mapped into the rump kernel namespace at the same location
(/stand/...). Therefore, if the *host* module directory is populated,
"rump.modload foo" will work as expected. Otherwise, RUMP_MODULEBASE
can be used to point to the module directory.

Sync usage with man page.

-p -P doesn't make much sense since no module will be loaded.

Add the MODCTL_NO_PROP flag to tell the kernel to ignore <module>.prop.

Add the '-P' option to modload(8) to set this flag.

Add a new [-d var] option which when combined with -p and -m will allow
you to delete vars from the existing <module>.prop file.

Sort options.

Add a "-m plist" option. This option will read in an existing
<module>.prop file and merge any options supplied on the "command
line".

This code will serve as the basis for in-kernel merging.

fputs() instead of puts() so we don't get the silly extra blank line

Sync usage and man page.

Add -p option which outputs a property list suitable for loading with a
module. This is in preparation for having the kernel load an optional
<module>.prop alongside a module, which is useful for passing options
to autoloaded modules and modules loaded at boot time.

Remove LKMs and switch to the module framework, pass 1.

Proposed on tech-kern@.

Remove clause 3 and 4 from TNF licenses

branches: 1.2.2; 1.2.4;
Add the ability to specify load-time properties for modules. This adds
three new flags to the new modload utility: -b, -i and -s, which are used
to pass boolean, integer and string parameters respectively.

branches: 1.1.2; 1.1.4; 1.1.6;
Pull in my modules code for review/test/hacking.

call it "kernel object linker module framework" not just
"module framework". the latter is generic enough to
also mean the old ld(1) linked loadable kernel modules.

Make -f description actually use the word "force" so that it can be
found when searched for.

Fix Dt argument.

PR/52417: Edgar Pettijohn: modules.conf(5) not mentioned in related manuals

Add .An -nosplit.

Document the version modload appeared and authors.
Correct spelling mistakes.
Bump date.

Add cross-refs from modload(8) amd modunload(8) to modctl(2)

Quote path separator character.

Do not not look for modules in the current working directory first. This is
to prevent from accidentally loading ./module.kmod when we actually wanted to
load module from the system module area.

To load a module from a filesystem path, the module name must contain at
least on path separator character (/), to load a module from the system
module areas, the name must not contain a path separator character:

modload ./mymod.kmod # loads mymod.kmod from the curren directory
modload mymod # loads mymod.kmod from the system module area

Fix xref; module(4) -> module(7).

Xref module(4).

note module changes

Remove -P from -p section, not from -f one.

-p -P doesn't make much sense since no module will be loaded.

Add the MODCTL_NO_PROP flag to tell the kernel to ignore <module>.prop.

Add the '-P' option to modload(8) to set this flag.

Add a new [-d var] option which when combined with -p and -m will allow
you to delete vars from the existing <module>.prop file.

Sort options.

Add a "-m plist" option. This option will read in an existing
<module>.prop file and merge any options supplied on the "command
line".

This code will serve as the basis for in-kernel merging.

Sync usage and man page.

Sort option descriptions. Split usage in SYNOPSIS in two, since there
are two different methods of calling it.
Add arguments to option descriptions.

Add -p option which outputs a property list suitable for loading with a
module. This is in preparation for having the kernel load an optional
<module>.prop alongside a module, which is useful for passing options
to autoloaded modules and modules loaded at boot time.

Punctuation formatting nit.

Remove LKMs and switch to the module framework, pass 1.

Proposed on tech-kern@.

branches: 1.26.20;
In the DIAGNOSTICS section, document that mismatched LKM and kernel
versions will be reported to the console and to the system message
buffer.

And update documentation date.

New sentence, new line; improve grammar.

implement support for forced LKM load, which makes the kernel skip
the version checks, using new -f flag and LMFORCE ioctl

Hopefully the description in the manpage is discouraging enough so that
nobody would dare even _think_ of using forced loads. A warning
message is written to stderr when this is used, just in case.

.Nm does not need a dummy argument ("") before punctuation or
for correct formatting of the SYNOPSIS any longer.

'NetBSD.org' and some mdoc fixes.

Add support for using an optional linker script (/usr/lkm/ldscript if
found, and overridable by the -T option).

New sentence, new line. By Robert Elz with minimal fixes.

For loadable drivers, the 3rd argument is character device major and
the 4th argument is block device major.

Remove one item from BUGS section.

Generate <>& symbolically. I'm avoiding .../dist/... directories for now.

Improve author markup, use standard header.

Whitespace nits

Drop arguments of .Os.

sweep of my licenses (userland files w/o only my copyright) for
consistency. (no functional changes)

fix up NetBSD RCS Ids to match the standard, and the leading comment as
to match as well. No functional changes.

branches: 1.12.6;
add ELF support, mostly from joda@pdc.kth.se (Johan Danielsson) in PR#6149,
updated for the symtab support for a.out (non for ELF yet) and other cleanup
by myself. should work on the alpha as well as other ELF ports...
remove the ld(1) output by default, rather than leaving little binary turds
across the filesystem...

Invert the meaning of -s, from "don't load symbols" to "load symbols".
Document that this is expensive (and why) in the BUGS section.

branches: 1.10.2;
Add missing .El line.

Clean up SYNOPSIS formatting.

Allow modload to load symbols for use by DDB.

if xxxinit is not found in the module, look for <modname>_lkmentry as an
alternate entry point (used in provided modules). Mention this on man page
too.

* cleanup for WARNS=1
* getopt returns -1 not EOF
* use .Nm correctly

convert to new RCS Id conventions; reduce my headache

Bug fixes from John Kohl:
modload didn't honor its -p argument.
It also would destroy the input file when you didn't specify an output
symbol file with -o.

branches: 1.3.2;
Fix spelling error in Copyright notice

Fix spelling errors

add LKM man pages

Remove LKMs and switch to the module framework, pass 1.

Proposed on tech-kern@.

branches: 1.51.20; 1.51.28;
Move WARNS=3 to the Makefile.inc, and add a little const to the remaining
programs that did not compile before.

Terminate variable argument list with NULL, not 0.

Fix a typo in a comment.

Replaced strncpy with strlcpy. Thanks to Peter Postma who
pointed them our in PR #25762. Approved by christos@NetBSD.org.

Remove an extraneous comment before main() that seems to have been
accidentally cut-and-pasted in with rev 1.30.

Put -T in usage output. PR#25875 from Kouichirou Hiratsuka.

Remove superfluous TRUE and FALSE definitions.
Noted by Jeff Ito in PR 24845.

Update the change from rev. 1.42 to reflect that MAXPATH includes room
for the terminating NUL character.

it's not correct to use MAXPATHNAME-sized buffer for link command, as
pointed out by Klaus Klein (original idea was that it should hold at least
the pathname ...); instead, let the ELF and a.out backends allocate memory
for the link command, and get rid of the fixed size buffer altogether

use MAXPATHLEN-sized buffer for module paths
reported and fix provided in bin/24379 by Nicolas Joly

implement support for forced LKM load, which makes the kernel skip
the version checks, using new -f flag and LMFORCE ioctl

Hopefully the description in the manpage is discouraging enough so that
nobody would dare even _think_ of using forced loads. A warning
message is written to stderr when this is used, just in case.

remove obsolete comment

asprintf() is easier than malloc + sprintf

Try first to use /dev/ksyms to get kernel symbols, if that fails mutter a
bit and fallback to the old way.

desupport LMRESERV_O, anything relying on it doesn't have chance of working
anyway due to numerous kernel structure changes

If no other name is specified, and machdep.booted_kernel is present
and names a file, use that as a the default kernel, otherwise fall
back to /netbsd.

Makes lkms work *much* better when you're testing kernels that are not
named /netbsd.

Add support for using an optional linker script (/usr/lkm/ldscript if
found, and overridable by the -T option).

Use "type *" instead of "type*".

ANSIfy, KNF, declare externs in headers only.

uppercase the lkm kernel-userspace dev macros and prefix them with LKM_
make modload print the bdev and cdev major when its a dev lkm

Split the overload block/character device major.
The 3rd argument is character device major and
the 4th argument is block device major.

PR/14498: Wesley Chen: Support symbol loading on elf lkms.

couple cosmetic stylistic changes

(usage): add -s and -S

correct getopt() optstring: -A gets a parameter!

-S isn't supposed to take an argument

add ELF support, mostly from joda@pdc.kth.se (Johan Danielsson) in PR#6149,
updated for the symtab support for a.out (non for ELF yet) and other cleanup
by myself. should work on the alpha as well as other ELF ports...
remove the ld(1) output by default, rather than leaving little binary turds
across the filesystem...

Invert the meaning of -s, from "don't load symbols" to "load symbols".
Document that this is expensive (and why) in the BUGS section.

Fix entry point parsing if output file specified to modload (PR7352)

branches: 1.22.2;
Allow modload to load symbols for use by DDB.

__AUDIT__ cleanup.

const poisoning.

Mods for alpha: doesn't work yet, but this is as good a placeholder as
any to note the ld flags to actually use. Still to do: elf loading.

make these compile on the alpha after WARNS=1.

use memset() not bcopy()

if xxxinit is not found in the module, look for <modname>_lkmentry as an
alternate entry point (used in provided modules). Mention this on man page
too.

* cleanup for WARNS=1
* getopt returns -1 not EOF
* use .Nm correctly

Check for .o extension after the *last* . in a file name.
Reported by Micheal Graff in PR #1868.

<sys/errno.h> -> <errno.h>

convert to new RCS Id conventions; reduce my headache

-Wall cleanups from Jim Jegers

Bug fixes from John Kohl:
modload didn't honor its -p argument.
It also would destroy the input file when you didn't specify an output
symbol file with -o.

branches: 1.9.2;
kill (bogus) multiple inclusions

kill stupidity

Stylistic cleanup; use err*() and atexit().

Support BSS in LKMs

while documentation is nice, including SunOS man pages in your source isn't!

i love ansi

update for present reality

open device for read/write, so we can actually load
devices w/new checks in kern_lkm.c

add support for terry lambert's loadable kernel modules.
needs a bit of cleanup, but overall: SLICK!

Remove LKMs and switch to the module framework, pass 1.

Proposed on tech-kern@.

branches: 1.6.32;
it's not correct to use MAXPATHNAME-sized buffer for link command, as
pointed out by Klaus Klein (original idea was that it should hold at least
the pathname ...); instead, let the ELF and a.out backends allocate memory
for the link command, and get rid of the fixed size buffer altogether

Add support for using an optional linker script (/usr/lkm/ldscript if
found, and overridable by the -T option).

Use "type *" instead of "type*".

ANSIfy, KNF, declare externs in headers only.

PR/14498: Wesley Chen: Support symbol loading on elf lkms.

add ELF support, mostly from joda@pdc.kth.se (Johan Danielsson) in PR#6149,
updated for the symtab support for a.out (non for ELF yet) and other cleanup
by myself. should work on the alpha as well as other ELF ports...
remove the ld(1) output by default, rather than leaving little binary turds
across the filesystem...

Add rump client support for module utilities. It should be noted
that while modstat and modunload and builtin modules work exactly
the same as in the host case, modload loads file system kernel
modules from the rump kernel namespace. By default, archs which
have rump support for the kernel kernel ABI have the host module
directory mapped into the rump kernel namespace at the same location
(/stand/...). Therefore, if the *host* module directory is populated,
"rump.modload foo" will work as expected. Otherwise, RUMP_MODULEBASE
can be used to point to the module directory.

Add rump client support for module utilities. It should be noted
that while modstat and modunload and builtin modules work exactly
the same as in the host case, modload loads file system kernel
modules from the rump kernel namespace. By default, archs which
have rump support for the kernel kernel ABI have the host module
directory mapped into the rump kernel namespace at the same location
(/stand/...). Therefore, if the *host* module directory is populated,
"rump.modload foo" will work as expected. Otherwise, RUMP_MODULEBASE
can be used to point to the module directory.

Remove LKMs and switch to the module framework, pass 1.

Proposed on tech-kern@.

branches: 1.3.32;
Add support for using an optional linker script (/usr/lkm/ldscript if
found, and overridable by the -T option).

convert to new RCS Id conventions; reduce my headache

Forgot to check this in.

use crunchops for crunchables

Add rump client support for module utilities. It should be noted
that while modstat and modunload and builtin modules work exactly
the same as in the host case, modload loads file system kernel
modules from the rump kernel namespace. By default, archs which
have rump support for the kernel kernel ABI have the host module
directory mapped into the rump kernel namespace at the same location
(/stand/...). Therefore, if the *host* module directory is populated,
"rump.modload foo" will work as expected. Otherwise, RUMP_MODULEBASE
can be used to point to the module directory.