nvmm_x86_vmx.c: fixes tprof stalls
the current implementation of nvmm nmi handling is
simple; just drops them. it interferes other nmi-using
components in the system, including tprof.
this commit makes nvmm try to dispatch nmis by itself.
while this is imperfect, it's good enough to give
tprof a chance to unmask LAPIC_LVT_PCINT.
i guess svm VMCB_EXITCODE_NMI needs something similar.
but i have no hardware. if someone wants to test a patch,
please let me know.
reproduce recipe:
1. run "tprof top" in a terminal. leave it running.
2. run an nvmm-based vm in another terminal.
3. observe some cpus in the "tprof top" stopped counting.
Re-factor m68k FP exception handling:
- Move the FP exception stubs into the shared trap_subr.s. Remove as
many conditionals from the stubs as feasible, having separate stubs
for each case that needs to be handled. This doesn't really result
in any more (or duplicated) code, and makes it all a lot easier to
read.
- All platforms now call the new fpu_init() function, which is renamed
from fpu_probe(), and centralizes all of the FP-related initialization.
Platforms that don't need to probe for the FPU aren't forced to do so;
the probe only occurs if the fputype is FPU_UNKNOWN at the time fpu_init()
is called.
- Even on platforms that don't actually need to probe the FPU, make sure
the "nofault" code is present in trap() to do so; this is meant to reduce
diffs that need to be resolved later then the trap() function is re-factored.
- On Amiga and Atari, remove the bespoke vectab code and just use the shared
vec_init(). Additional cleanup is possible in this area.
(missed this file in larger commit)
Re-factor m68k FP exception handling:
- Move the FP exception stubs into the shared trap_subr.s. Remove as
many conditionals from the stubs as feasible, having separate stubs
for each case that needs to be handled. This doesn't really result
in any more (or duplicated) code, and makes it all a lot easier to
read.
- All platforms now call the new fpu_init() function, which is renamed
from fpu_probe(), and centralizes all of the FP-related initialization.
Platforms that don't need to probe for the FPU aren't forced to do so;
the probe only occurs if the fputype is FPU_UNKNOWN at the time fpu_init()
is called.
- Even on platforms that don't actually need to probe the FPU, make sure
the "nofault" code is present in trap() to do so; this is meant to reduce
diffs that need to be resolved later then the trap() function is re-factored.
- On Amiga and Atari, remove the bespoke vectab code and just use the shared
vec_init(). Additional cleanup is possible in this area.
ddb(4): improve markup readability for the list of commands
Add empty (single dot) lines around commands to make them more visible
in the man page source. Split long chains of Ij Kl Mn Op into
multiple source lines with Xo/Xc. While here mark up literal comma in
commands with Cm to make it more visible in the PostScript output.
sys_mincore: stop locking potentially huge amount of user memory
the current implementation performs uvm_vslock() on the
user-specified amount of memory. it isn't safe in general.
some might even consider it as a security issue.
this commit fixes it by splitting the user-specified range into
small ones which a temporary kernel buffer can cover. it's ok to
report potentially stale values as the mincore() api is inherently
racy in the first place.
while we still ought to invent a proper "lock user memory for
direct i/o" infrastructue, in this particular use case, it isn't
necessary or appropriate.
zfs: fix case insensitive / utf-8 normalized file names (cont.)
this was intented to be a part of an earlier commit.
("zfs: fix case insensitive / utf-8 normalized file names")
for some reasons, it seems i unintentionally dropped this hunk
when porting the commit from git to cvs.
PR bin/60099 Add more test cases
More test cases to test $* expansion in a variety of
situations (all using $* unquoted - and in contexts
where field splitting could happen).
Several of these new text cases assisted with debugging
the $* expansion code in sh, avoiding issues that otherwise
might not be seen as potential problems.
PR/60069 NetBSD non booting on H3 Allwinner
Disable EFI runtime support on arm as it doesn't play nicely with many
FWs where things go south after/during the call to SetVirtualAddressMap
Also seen by cp- on Odriod-XU4
The Coprocessor Protocol Violation trap is not present on the 68010
(the vector is "unassigned, reserved"), so don't install a handler
for it on that CPU.
Fix a brain-f*rt in the special mac68k "nofault" bus error handling.
The information we need to pass along is packaged up for us neatly
in the stack frame and arguments being passed to trap(), so use those
to extract the %a2 value and faulting address.
Issue raised and fix tested by nat@
zpool_find_import_impl: fix block/character device confusion
this commit fixes zpool import failure in some cases.
the current logic sometimes (eg. "zpool import -d" with a directory
which is not "/dev") ends up with picking character devices
and pass them to ZFS_IOC_POOL_TRYIMPORT/ZFS_IOC_POOL_IMPORT ioctl.
such attempts would fail, marking the corrosponding vdevs UNAVAIL.
this commit fixes it by skipping character devices.
also, this commit makes the label checking logic prefer to use
character devices when available because it seems like the intention
of the upstream logic.
also, this commit fixes import of file-backed vdevs.
zfs: fix file vdev
make solaris compat vn_openat to honor the root vnode specified
by the caller. it's currently only used for vdev_file.c.
this commit fixes "no such pool or dataset" error on zpool create
with files:
```
uma% dd if=/dev/zero of=/tmp/hoge count=100
100+0 records in
100+0 records out
51200 bytes transferred in 0.001 secs (51200000 bytes/sec)
uma% sudo zpool create f /tmp/hoge
cannot create 'f': no such pool or dataset
uma%
```
cf. "file" in "Virtual Devices (vdevs)" in zpool(8).
tests/kernel/t_time_arith.c: update for the recent kernel change
the test updates in the recent change was for some reasons incomplete.
("itimer_transition: do not keep it_value unchanged after firing the event")
https://mail-index.netbsd.org/source-changes/2026/03/17/msg161046.html
this commit updates the rest of tests.
tested with a full build and then the following command:
```
# chroot destdir
# cd /usr/tests/kernel
# atf-run t_time_arith > /dev/null
# echo $?
0
#
```
test failures are pointed out by Taylor R Campbell
[5 lines not shown]
