uhid: Fix uhidev attach bug and remove hidev_tag_t typedef

(jmcneill)

uhid: Move generic HID code to sys/dev/hid

This change introduces a bus agnostic HID interface (sys/dev/hid/hidev.h)
and splits the uhid driver into a shareable (sys/dev/hid/uhid_common.c) and
USB specific (sys/dev/usb/uhid.c) code.

(jmcneill)

Ticket #115

(martin)

Pull up following revision(s) (requested by jmcneill in ticket #115):

sys/dev/sdmmc/sdmmc_io.c: revision 1.22
sys/dev/sdmmc/sdhc.c: revision 1.123

sdhc: Do not set auto cmd12 enable for SDIO RW extended ops
sdmmc: Use block mode for multi block SDIO transfers.

Change from OpenBSD:
https://github.com/openbsd/src/commit/5e0782b17abc2070c235843e37de911ce3f98ef8

So far the SDIO stack issued one transfer for every 64 byte to be
copied.  This severely limits the speed over the bus.  By using
block mode we can issue block-sized transfers, which bumps the
transfer size to at least 512 bytes.  By using multi-block mode
we can copy up to 511 blocks per transfer.

(martin)

Annotate #112 for additional pullups

(martin)

Additionaly apply the following to fix the build for ticket #112:

usr.sbin/npf/npfctl/npf_build.c 1.63

enclose NPF_VAR_ID case in braces to fix linter build

(martin)

remove some commented out leftovers
NFC

(macallan)

fix XMAP9 access / timings, which is surprisingly difficult since REX and XMAP9
run at different clock speeds ( one on bus clock, the other on half the pixel
clock with no handshake, and both with their own FIFOs

From Adrian Chadd

(macallan)

reverse the GCC 14 test polarity, now just alpha, mips, sh3, and m68k.

mips switching soon, the others have some issues remaining.

(mrg)

note ppc switched.  thanks to martin for making it work!

(mrg)

Fix case of OpenBSD for last rge(4) entry

(leot)

sdmmc: Use block mode for multi block SDIO transfers.

Change from OpenBSD:
https://github.com/openbsd/src/commit/5e0782b17abc2070c235843e37de911ce3f98ef8

So far the SDIO stack issued one transfer for every 64 byte to be
copied.  This severely limits the speed over the bus.  By using
block mode we can issue block-sized transfers, which bumps the
transfer size to at least 512 bytes.  By using multi-block mode
we can copy up to 511 blocks per transfer.

(jmcneill)

sdhc: Do not set auto cmd12 enable for SDIO RW extended ops

(jmcneill)

Note my update of rge(4)

(pgoyette)

doc: Fix spelling of "accommodate".

(nia)

doc: Last month of changes

(nia)

Trailing whitespace

(skrll)

m68k/elf_machdep.h: organize symbols into psABI & local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

elftoolchain/common: Update to upstream [r4303].

(jkoshy)

Add a testcase for PR/59652.

(skrll)

ftp: fix ascii transfers with progress bar

Handle stdio interruption by signals and improve error handling
in getc() and putc() on the control and data channels.
Provide ftp_getc() and ftp_putc() wrappers that:
- Retry the operation on EINTR or EAGAIN instead of failing.
- Store other error codes in a return variable separate to errno,
  and use that variable in ferror() handling, for more correct
  error messages.

Fixes the progress bar display in ascii mode transfers.

(Note that I haven't fixed interrupted reads from stdin or
writes to ttyout; that's a much larger refactor that's out of scope).

Fix PR bin/59587 from Johnny Billquist, with the fix
inspired by the patch in that PR.

(lukem)

Undo previous commit; it needs further refinement.

(perseant)

Remove redundant cache.h header; all of these values are already defined
in generic m68k code.

(thorpej)

or1k/elf_machdep.h: Group symbols into psABI and local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

Appease clang & KNF.

Should be NFC for gcc compiles.

(kre)

Track individual inodes between the time that their accounting is
subtracted from one segment before it is added to the new segment.
Counting duplicates is not sufficient.  Addresses a "negative bytes"
panic.

(perseant)

Use a red/black tree, rather than a linked list, to store deferred
sergment byte count changes in truncation.

(perseant)

mips: Fix R5000SC cache page macro typo that broke boot.

A change in rev 1.5 to mips_r5k_round_page() and mips_r5k_trunc_page()
in cache_r5k.h (for PR/55139) had a fatal typo in the mask, so the
R5000SC cache flush code ended up operating on unintended addresses
and at least R5000SC Indy would no longer boot.
(not sure how my Qube 2700 worked at that time..)

Fix the macros to use the intended mask so that the secondary cache
flushes are done on the correct range again.

Sorry for the long breakage.

Analyzed and reported by Adrian Chadd on port-mips@.
https://mail-index.netbsd.org/port-mips/2025/12/thread1.html#001536

Should be pulled up to netbsd-9, netbsd-10, and netbsd-11.

(tsutsui)

sys/arch/amd64: support ramdisks when booting in GENPVH mode

This adds support for booting NetBSD/amd64 with QEMU窶冱 -kernel/-initrd
combination, in generic PVH mode (GENPVH). This includes the MICROVM
kernel, which now defaults to supporting modules and root on memory
disks.

This patch probes the first few bytes loaded in memory for
differentiating between modules, splash screens (PNG or JPEG), or
assumes filesystem images otherwise. It works with plain filesystem
images, but not with modules yet. Supporting any combination of
filesystem, modules or splash screen images should be possible in the
future.

Tested on NetBSD/amd64 by iMil and myself; measured to gain another 1~3
ms of boot time with the MICROVM kernel on the reference hardware.

As posted on tech-kern@ on 26/11/2025, with no objections.

XXX pull-up to -11 and adapt for -10

(khorben)

ia64/elf_machdep.h: organize symbols into psABI & local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

sparc/elf_machdep.h: organize symbols into psABI & local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

sparc64/elf_machdep.h: organize symbols into psABI & local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

elftoolchain/common: Update to upstream [r4299].

(jkoshy)

Make sure to actually *push* the MMU51_TCR_BITS onto the stack.

(thorpej)

Use the sharable MMU enablement code fragment.

(thorpej)

Add a sharable code fragment for enabling the MMU.

(thorpej)

Some tidy-up of the MMU-enable code:
- Use defined constants rather than magic numbers for the 68040 %cacr
  and 68060 %cacr and %pcr bits.
- Clean up some comments.
- Use a more robust comparison against mmutype for the 68040-or-68060
  case (and improve the comment, as well).
- Make sure all final jumps to Lmmuenabled are non-PC-relative, and add
  such a jump to the final (68851/68030) case rather than letting it simply
  fall through.  This is in service of making this code sharable as an
  includable code fragment on platforms with different requirements (i.e.
  systems that don't have a VA==PA environment when this code runs).

(thorpej)

Add a note about how asm code relies on the ordering of the MMU_*
constants.

(thorpej)

Define the fields in the 68060 Processor Configuration Register.

(thorpej)

Tickets #111 - #114

(martin)

Pull up following revision(s) (requested by riz in ticket #114):

distrib/sets/lists/debug/mi: revision 1.488 (parts of, via patch)

fix debuglib setlists

(martin)

Pull up following revision(s) (requested by nia in ticket #113):

sys/arch/mips/include/bswap.h: revision 1.6
sys/arch/powerpc/include/bswap.h: revision 1.7

mips: Use the inline macro version of bswap.

mips suffers from the same problem as sparc where bswap is slow
and using the GCC builtin does not make it better.

This change improves the speed of disk encryption by around 3 MiB/s
on an EdgeRouter 4.
PR port-sparc64/59789

powerpc: Use GCC builtins for bswap.

This doubled disk encryption performance on an (emulated) G4. Further
improvements are possible if the GCC 2 version of the endian functions
is used, so we might want to consider doing that on more NetBSD ports.

PR port-sparc64/59789 bswap is slow

(martin)

Pull up following revision(s) (requested by joe in ticket #112):

usr.sbin/npf/npftest/npftest.conf: revision 1.18
usr.sbin/npf/npftest/libnpftest/npf_rule_test.c: revision 1.26
usr.sbin/npf/npfctl/npf_var.c: revision 1.16
usr.sbin/npf/npfctl/npf_var.h: revision 1.14
usr.sbin/npf/npfctl/npf_build.c: revision 1.62

PR bin/59511

when extracting variables for filtering in NPF, allow the handler to
recursively extract all variables that might be present in the parent variable
to fully get all the filter elements present in them. this issue poses a  security risk
as intruders can find their way into your machine if you intend to block them
but have their IPs in a nested variable with other IPs as well.

(martin)

Pull up following revision(s) (requested by skrll in ticket #111):

sys/dev/mii/mcommphy.c: revision 1.4
sys/dev/mii/mcommphy.c: revision 1.5
sys/dev/mii/mcommphy.c: revision 1.6

Apply some standard defaults for YT85[23]1

Slight re-arrangement of #defines. NFCI.

Spaces to tab

(martin)

diff(1): use Fl Fl for long options

(wiz)

Avoid inserting spurious whitespace after "etc.)", which doesn't actually
end the sentence here.

(jschauma)

mips/elf_machdep.h: organize symbols into psABI & local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

riscv/elf_machdep.h: organize symbols into psABI & local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

elftoolchain/common: Sync with upstream [r4296].

(jkoshy)

Import BSD-licensed diff from FreeBSD to replace GNU diffutils

Ported to NetBSD, compiler warnings fixed, and some unnecessary
features (for the purpose of replacing our GPLv2 diff) have been
stripped out.

Non-default at the moment, requires building with MKBSDDIFF=yes.

PR bin/59798 We are still using GNU diffutils

(nia)

remove accidental new line.

(andvar)

viaide(4): identify and attach VT6420 IDE (PATA) controller.

The VT6420 is an independent version of the integrated VT8237R SATA RAID and
IDE controller, sharing the same PCI ID as the southbridge SATA controller.
The PATA controller, on the other hand, has a unique PCI ID, since the
integrated one reuses the same PCI ID as many other VIA southbridges and ISA
bridge identification would not work with this controller. The controller is
quite rare (likely due to the SATA autonegotiation bug), so it seems nobody
added this PCI ID to any major OS.

Tested on Biostar M7NCD Ultra.

(andvar)

Enable PaX features in WII kernel configuration.

Seems to work fine, with mmap addresses being randomized per-process
and (PROT_WRITE | PROT_EXEC) resulting in a segfault.

(nia)

regen.

(andvar)

Add VIA VT6420 IDE controller.

(andvar)

Disable caches earlier.  Use relative labels in MMU enablement code.

(thorpej)

Initialize %sfc and %dfc in pmap_bootstrap2(), rather than duplicating
4 lines (including comment) across 11 locore.s files.

(thorpej)

Add setdfc() and setsfc() inlines.

(thorpej)

Fix an error in version 1.295: the size of an IINFO is IINFOSIZE(fs), not
sizeof(IINFOSIZE(fs)).  Addresses a segment summary overwrite in LFS64.

(perseant)

vax/elf_machdep.h: permit usage with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

hppa/elf_machdep.h: organize symbols into psABI & local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

elftoolchain/common: Update to upstream [r4294].

(jkoshy)

The LLVM build is a bit larger than the GCC build.  Bump the size up a bit.

(brad)

_SC_OPEN_MAX can exceed FD_SETSIZE. Use the compat value svc_maxfd+1
from the rpc library similar to ypbind(8).

The custom _svc_run() function is required as the standard svc_run()
may hang for retries.

Fixes PR 59750.

(mlelstv)

Fix error message.

(mlelstv)

I am giving up attempting to bend the kernel build machinery to my will
so that assym.h can be safely included by libkern assembly sources.  So,
instead, redundantly define __HAVE_M68K_BROKEN_RMC in multiple places so
that all the things that need to key off of it can.

(thorpej)

sh3/elf_machdep.h: organize symbols into toolchain & local groups.

Allow this file to be used with <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

Update to Elftoolchain [r4290].

(jkoshy)

Trim sl and swwdog from the RAMDISK kernel in hopes of getting it to fit.

(brad)

Revert previous.  I guess I'll have to solve the m68k build issue
a different way.

(thorpej)

We can use __builtin_bswapX on sparc when using clang, so do so.

PR port-sparc64/59789 bswap is slow

(nia)

Make libkern depend on assym.h.

(thorpej)

Revert the size change of the install ramdisk and instead just remove
a driver from the install kernel.  Same ramdisk size as before.  Not
sure this platform can have a larger ramdisk size.

(brad)

Image size changed.  Bump.

(brad)

MPROTECT=0 misleadingly actually enables MPROTECT.

(nia)

Config glue for __HAVE_NEW_PMAP_68K

(thorpej)

Add a PMBM_F_RO flag to allow machine_bootmap[] to create read-only
mappings.  mac68k uses this to map the Mac ROMs.

(thorpej)

Add __HAVE_NEW_PMAP_68K glue:
- mac68k_init(): Use pmap_protect() to write-protect the kernel text
  (with a comment explaining why we do this here).
- Add a machine_bootmap[] with entries for IOBase, ROMBase, and VIDBase.
  In bootstrap_mac68k(), patch up machine_bootmap[] with run-time computed
  values.

(thorpej)

pmap_machine_check_bootstrap_allocations() and bootstrap_mac68k() are
independent of the pmap implementation; move them out of pmap_bootstrap.c
and into machdep.c

(thorpej)

bus_mem_add_mapping(): Rather than directly fiddling with PTEs, just
pass the appropriate flags to pmap_kenter_pa().

(thorpej)

Revert previous (wrong commit message, obviously).

(thorpej)

cpsw.4: some minor updates

Intervention for PHY configuration hasn't been necessary for some time.

(gutteridge)

bus_mem_add_mapping(): Rather than directly fiddling with PTEs, just
pass the appropriate flags to pmap_kenter_pa().

(thorpej)

Now that pmap_bootstrap1() correctly initializes Sysseg_pa, use that
value rather than computing the PA of from Sysseg.

(thorpej)

Sysseg_pa is supposed to be a physical address, so don't PA2VA()
the value before assigning it.

(thorpej)

* Use the LFS_IFDIRTY flag to keep track of whether the Ifile has
  ungathered changes during checkpoint.  Modify LFS_WRITESEGENTRY and
  make a new LFS_WRITEIENTRY macro to help track this.  Prevents
  an infinite loop in lfs_writeinode, previously caught with a panic.

* Fix a false-positive "negative bytes" panic when writing inodes into
  the same segment they were in before, and consolidate the accounting
  logic into new function lfs_subtract_inode.

* Ensure that we do not clean any inodes that were unavailable during
  the first pass of cleaning, so that we never need to mix cleaning and
  non-cleaning blocks (except for the Ifile) when writing.

* Straighten out the logic in lfs_writefile to make it more readable.

* Include IMNT_WANTRDONLY in tests for closing out writes, not just
  IMNT_UNMOUNT.

* Keep track of whether the kernel cleaner is active on a file system
  with a status flag; use a condvar to synchronously turn this on
  or off.

(perseant)

Remved final return from void-retuning function

(manu)

Bump up the size of the ramdisk.  Maybe fix the build.

(brad)

fix few typos in comments.

(andvar)

Make the sshramdisk a bit bigger.  Hopefully fixes the build.

(brad)

Enable ASLR by default on OCTEON.

Tested on an EdgeRouter 4.

(nia)

Header file glue for __HAVE_NEW_PMAP_68K.

(thorpej)

i386/elf_machdep.h: Organise symbols into psABI and local groups.

Permit this file to be used alongside <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

Add a hook that allows a machine to override the default kernel text
protection bits.  This is needed by mac68k which must map a portion
of the kernel text read/write (roughly the first 8KB of the kernel
address space) as that's where it needs to place the vector table plus
the "low ram" variable area needed by the Mac ROMs.

The hook is such that the entire kernel text is mapped with the specified
PTE bits, and then machdep code can come back later to perform any fixups
as needed (e.g. re-mapping &start to &etext read-only) once the kernel is
running on its MMU tables.

While here, add one more PMAP_MACHINE_CHECK_BOOTSTRAP_ALLOCATIONS check
if allocation of additional segment table pages is needed (unlikely, but
this check is needed in case it does happen).

(thorpej)

G/C P1PAGES

(thorpej)

powerpc/elf_machdep.h: Use psABI spellings and group symbols into psABI/local groups.

The prior spellings have been kept around for backward compatibility.

Permit this file to be used alongside <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

Pull in some code from OpenBSD and adapt to our net80211 stack.

The result seems to be stable (no more firmware panics and
unknown firmware events). The driver now also sets (and
upgrades) transmit rates.

Tested with
- Intel AC 3168
- Intel AC 7265
- Intel AC 8000
- Intel AC 8265

AC 3168 now also uses a more recent firmware (in a separate commit).

(mlelstv)

When using a SC16IS7xx device in interrupt polling mode it was pretty
abusive to the I2C or SPI bus, and beat up the USB bus pretty badly as
the kthread ran all of the time checking to see if an interrupt
occurred.

Take advantage of the enable and disable calls that the general com.c
code supports for the purpose of doing power management and use those
to start, pause and resume the kthread only when the device is
actually open().  There does not effect the use of the GPIO on the
SC16IS7XX as that doesn't use the kthread for anything.  There does
not seem to be any particular down side to this change.  That is, it
there doesn't seem to be a use case for the device not to be open()
and still want interrupts handled.

Also...  apparently the device property is-console and/or
force-console needs to be set to something.  Just letting those float
appears to have the chance of introducing random values which are
sometimes "true", which is unexpected.  For the SC16IS7xx device, just
set these to false.  It is mostly a SMOP to getting this device
working as a console, with limitations, as the I2C or SPI bus must
exist first and be in a working state.  There is some interesting
ability of UEFI to talk to a I2C bus, so it may be possible to even
have early console if you can get that path working on systems that
have UEFI.  While the ability to have a dedicated console, and perhaps
KGDB, port outside of what a system might usually have is interesting,
all of this is beyond my particular use case.

(brad)

Add recent firmware files for some iwm(4) variants.
This needs corresponding changes in the driver.

(mlelstv)

XXX XXX Big, fat, ugly workaround for an unfortunate senario triggered
XXX XXX by the addition of the umcpmio driver on the RISC-V port.

(thorpej)

Only include "assym.h" if _KERNEL_OPT is defined.

(Rump strikes again!)

(thorpej)

if: move if_linkstate_hooks initialization out of if_getindex() (NFC)

if_getindex() is not expected to initialize it.

(ozaki-r)

sparc64: Enable AES_BEAR64 in GENERIC

PR kern/59774 bearssl 32-bit AES is too slow, want 64-bit optimized
version in kernel

(nia)

s/relesection/reselection/ in comments.

(andvar)

s/releated/related/ in the acpi dump output and the kernel config comment.

(andvar)

spaces -> tabs

(skrll)

G/C unused P1PAGES.

(thorpej)

npf.conf.5: fix a minor typo

(gutteridge)

pmap_bootstrap1() has some macros to aid translation between virtual and
physical addresses.  Move these into <m68k/pmap_68k.h> and put them into
two groups.

The first group (PMAP_BOOTSTRAP_RELOC_GLOB() and PMAP_BOOTSTRAP_RELOC_PA())
are for referencing global symbols and physical addresses from within the
context of pmap_bootstrap1() (or similar function) itself.

The second group (PMAP_BOOTSTRAP_VA_TO_PA() and PMAP_BOOTSTRAP_PA_TO_VA())
are for translating addresses that will be used in the context of the running
kernel once the MMU is enabled.

The former group is necessary because some platforms may call pmap_bootstrap1()
with the MMU *enabled* but running on foreign mappings whereby global
references work without relocation but access to physical addresses require
relocation (this is the case on mac68k systems that place the frame buffer
at physical address $0000.0000, in which case MacOS uses the MMU to simulate
the traditional MacOS memory map and we're running on those mappings when
the kernel starts).

Systems with special requirements (see above) can override each of these
macros individually by simply defining them in <machine/pmap.h> before
including <m68k/pmap_68k.k>.

(thorpej)

arm/elf_machdep.h: Use psABI spellings for a few constants.

The prior spellings have been kept around for backward compatibility.

(jkoshy)

Move the mac68k "check bootstrap allocations fit" code into a new function
called pmap_machine_check_bootstrap_allocations(), and call it from the
existing Hibler pmap pmap_bootstrap1() in the appropriate place.

Add a new hook for the new pmap_bootstrap1() enabled by defining
PMAP_MACHINE_CHECK_BOOTSTRAP_ALLOCATIONS in <machine/pmap.h>.  When
enabled, that function will be called once the bulk of the bootstrap
allocations have been completed.

(thorpej)

pmap_bootstrap() -> pmap_bootstrap1().

(thorpej)

Move the initialization of avail_start, avail_end, and last_page
from the end of pmap_bootstrap() to just after the call to pmap_bootstrap().

NFC.

(thorpej)

Don't bother recomputing physmem after all of the bootstrap allocations
are complete.  It's now computed earlier and is intended to reflect the
actual amount of physical memory in the machine in any case.

(thorpej)

physmem is a psize_t (see uvm_page.c).  Make the type used in the RELOC()
macro match, even though there's no practical difference on an IPL32 system.

(thorpej)

physmem is a psize_t (see uvm_page.c).  Make the type used in the RELOC()
macro match, even though there's no practical difference on an IPL32 system.

(thorpej)

Another mac68k adaptation note / TODO: mac68k also needs to map the
last page of RAM VA==PA for doboot(), but this is not at a fixed
location like it is on hp300.  Existing kernel PT allocation should
be sufficient to ensure there's a PTE availabe for this, but it would
be nice to provide a way for cpu_startup() to assert it.

(thorpej)

bump date to reflect changes

(joe)

Rename the "maxaddr" global to "last_page" to reflect what it actually is.
(There is also a "maxaddr" local used in cpu_startup(), so renaming the
variable that points to the last memory page makes things less confusing.)

(thorpej)

pmap_bootstrap1(): Skip machine_bootmap[] entries with size == 0.

(thorpej)

powerpc: Use GCC builtins for bswap.

This doubled disk encryption performance on an (emulated) G4. Further
improvements are possible if the GCC 2 version of the endian functions
is used, so we might want to consider doing that on more NetBSD ports.

PR port-sparc64/59789 bswap is slow

(nia)

mips: Use the inline macro version of bswap.

mips suffers from the same problem as sparc where bswap is slow
and using the GCC builtin does not make it better.

This change improves the speed of disk encryption by around 3 MiB/s
on an EdgeRouter 4.

PR port-sparc64/59789

(nia)

pmap_bootstrap1(): Add a TODO comment describing a hook currently needed
by mac68k.

(thorpej)

Sum up the memory and initalize physmem just before calling pmap_bootstrap().
NFC.

(thorpej)

Move the initialization of vidlen out of pmap_bootstrap() and into
bootstrap_mac68k(), just before the pmap_bootstrap() call.

NFC.

(thorpej)

Spaces to tab

(skrll)

Slight re-arrangement of #defines. NFCI.

(skrll)

extent(9) -> vmem(9)

(thorpej)

extent(9) -> vmem(9)

(thorpej)

install.1: Fix explanation of -p option.

Noted by GAW like a year ago

(dholland)

Consistently specify `-fno-unwind-tables` to shrink kernel binary size.

before:
NetBSD 11.99.4 (GENERIC) #22: Sun Nov 30 08:16:31 JST 2025
  text    data    bss    dec    hex filename
2996019  101388  101984 3199391  30d19f netbsd

after:
NetBSD 11.99.4 (GENERIC) #23: Sun Nov 30 08:18:18 JST 2025
  text    data    bss    dec    hex filename
2689719  101388  101984 2893091  2c2523 netbsd

Worth to pull up to netbsd-11.

XXX: should this be default for all kernels?

(tsutsui)

Remind people to update the SMALL* configs when updating this one.

(thorpej)

Add a trimmed down configuration file for the small 68020-based
machines.

(thorpej)

Initialize interrupt handlers earlier than registering NMI handler.

Otherwise `KASSERT(ih_allocfuncs == NULL);` in m68k_intr_init() is fired.
Also fix build error without "options news1700".

Should be pulled up to netbsd-11.

XXX: maybe this KASSERT should be more consistent with m68k_intr_establish()
XXX: that allows called without m68k_intr_init()?

(tsutsui)

- pte_set() and pte_mask() are only used for HP MMU VAC shenanigans; only
  make them available if MMU_CONFIG_HP_CLASS is true.  Also, don't bother
  with the inline asm; it's not really needed at all.
- Don't use atomic_add_ulong() for pmap_stat_update(), as it is implemented
  in terms of CAS, which must be avoided on __HAVE_M68K_BROKEN_RMC systems.
  It's only used inside pmap critical sections, so just garden-variety C
  arithmetic is fine.
- pmap_reference() / pmap_destroy() get similar treatment -- avoid
  atomic_inc and atomic_dec.  In this case, wrap the arithmentic inside
  PMAP_CRIT_ENTER() / PMAP_CRIT_EXIT().

(thorpej)

remove closing parentheses that I added in the last commit, it is closed in the
next line.

(andvar)

regen.

(andvar)

s/Fujutsu/Fujitsu/ in comment.

(andvar)

Fix various typos, mainly in comments.

(andvar)

Avoid using anything from <sys/atomic.h> if __HAVE_M68K_BROKEN_RMC is
defined.  On m68k, the atomics in question are implemented in terms of
CAS, and that instruction must be avoided on __HAVE_M68K_BROKEN_RMC
systems.

(Really, the use of atomics they're they're used isn't necessary *at
all* here, but I'm looking for the minimal change necessary for possible
pullup-11.)

(thorpej)

alpha/elf_machdep.h: Organize symbols into local and ABI groups.

Permit this file to be used alongside <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

Introduce __HAVE_M68K_BROKEN_RMC, which can be defined by m68k platforms
that cannot perform indivisible READ-MODIFY-WRITE bus cycles.  The
define is named for the /RMC output signal on the CPU that is asserted
when such a cycle is performed.  These cycles are only used by the CAS,
CAS2, and TAS instructions.  The cycles are, as far as I can tell,
otherwise indistinguishable from garden-variety read followed by write
bus cycles, except for the assertion of /RMC.  Note that other single-
instruction "read-modify-write" instructions (e.g. OR, AND, ADD, etc.
with a memory destination operand) do not assert /RMC, and are thus not
considered truly atomic from the perspective of the 68k architecture.

When __HAVE_M68K_BROKEN_RMC is defined, then we must make unavailable
the CAS, CAS2, and TAS instructions, because using them will typically
cause a bus error to occur.  In this case, we handle _atomic_cas_32()
(and its aliases) in the kernel environment the same way as the 68010
does: a restartable atomic sequence.  8- and 16-bit CAS operations are
not available in the kernel environment on such platforms.

We define __HAVE_M68K_BROKEN_RMC for the hp300 platform because 68020-based
hp300 systems will generate a fault for /RMC-type cycles.

Based on work by Izumi Tsutsui.

XXX pullup-11?

(thorpej)

Interpret p_cdsession field in partition info only when the type
is set to FS_ISO9660. Other filesystem types use the field for p_fsize,
and default values there confuse the cd9660 filesystem code.

Fixes PR 59783.

(mlelstv)

amd64/elf_machdep.h: Organize symbols into local and psABI groups.

Permit this file to be used alongside <sys/elfdefinitions.h>.

PR lib/59564

(jkoshy)

Don't need to define __HAVE_RAS here; it's already defined by <m68k/types.h>.

(thorpej)

Make sure to pass an allocation strategy (VM_BESTFIT) to vmem_alloc().

(thorpej)

extent(9) -> vmem(9)

(thorpej)

Add umcpmio to the list of things that spi can attach to.

(brad)

Mention umcpmio(4) version 2.

(brad)

PR bin/59801

(joe)