f5b00833 mav Nov. 30, 2021, 1:20 a.m.
Length of some string buffers was insufficient for cases of more that
99 targets per HBA or slots per enclosure.  Some others are tuned just
for better alignment.  While there also fix output formatting issues.

MFC after:	1 week
Sponsored by:	iXsystems, Inc.
cgit
bd6dce97 debdrup Nov. 30, 2021, 1:06 a.m.
Add a few very useful variables that might easily be overlooked, since
they're only documented in rc.subr(8) which might not be the first place
that people look.

At least _oomprotect has existed since 11.0-RELEASE, and doesn't appear
to be very well-known. While the others aren't as new, in my estimation,
a lot more people would use them if they knew about them.

While here, also add a reference to rc.subr(8) and login.conf(5), and
sort the variables alphabetically.

Reported by: 	Daniel Dettlaff <dmilith at gmail.com>
Reviewed by:	ceri, gbe, 0mp, ygy, a.wolk, pauamma
cgit
0e765d9b brooks Nov. 29, 2021, 10:04 p.m.
6d37a167 brooks Nov. 29, 2021, 10:04 p.m.
The mprotect syscall decleration is not const.  I added this one
incorrectly in a944d28d0edf7ceb1bef4d789dfa4e8e18331658.

Reported by:	kib
Reviewed by:	kib, imp
cgit
401eec36 brooks Nov. 29, 2021, 10:04 p.m.
a8efd4d1 brooks Nov. 29, 2021, 10:04 p.m.
Rather than combining the declearation of nosys with the registration
of SYS_syscall, declare syscall(2) and __syscall(2) with the new
SYSMUX type in syscalls.master and declare nosys directly.  This
eliminates the last use of syscall aliases in the tree.

Reviewed by:	kib, imp
cgit
d7f306c5 brooks Nov. 29, 2021, 10:04 p.m.
This type is for system call multiplexers (syscall(2), __syscall(2))
that don't have a normal handler and instead are handled in the
machine-dependent syscall code.

Reviewed by:	kib, imp
cgit
5c1835b1 brooks Nov. 29, 2021, 10:04 p.m.
cffb55f0 brooks Nov. 29, 2021, 10:04 p.m.
Declare the exit system call normally.  This results in the
implementation being named sys_exit rather than sys_sys_exit and
being decalred as returning an int.  Infact it does not return
at all because exit1 does not, so add an __unreachable() to let the
compiler know that.

Reviewed by:	kib, imp
cgit
7fb006e7 brooks Nov. 29, 2021, 10:04 p.m.
638c5fa8 brooks Nov. 29, 2021, 10:04 p.m.
Declare normal <foo>_args structs rather than going out of the way
to declare __<foo>_args.

Reviewed by:	kib, imp
cgit
c2996f8a brooks Nov. 29, 2021, 10:04 p.m.
ba4e5253 brooks Nov. 29, 2021, 10:04 p.m.
Declare o<foo>_args rather than reusing the equivalent <foo>_args
structs.  Avoiding the addition of a new type isn't worth the
gratutious differences.

Reviewed by:	kib, imp
cgit
28f04718 brooks Nov. 29, 2021, 10:04 p.m.
Stop using <foo>_args structs as part of internal kernel APIs.  Add
a kern_recvfrom and adjust getsockname and getpeername's equivalent
functions to take individual arguments rather than a uap pointer.

Adopt a convention from CheriBSD that a function interacting with
userspace pointers and sitting between the sys_<foo> syscall and
kern_<foo> implementation is named user_<foo>.

Reviewed by:	kib, imp
cgit
3660e76a brooks Nov. 29, 2021, 10:04 p.m.
Reviewed by:	kib, imp
cgit