Pull up following revision(s) (requested by wiz in ticket #1423):
external/bsd/tmux/usr.bin/tmux/Makefile: revision 1.47 (patch)
tmux: fix TMUX_CONF
Addresses PR 61034.
Pull up following revision(s) (requested by wiz in ticket #584):
external/bsd/tmux/usr.bin/tmux/Makefile: revision 1.47
tmux: fix TMUX_CONF
Addresses PR 61034.
The evaluation of resource shortage in wapbl_register_deallocation()
is incomplete as only "wl_dealloccnt" is checked. Replace this check with
the doflush check from wapbl_begin() testing for 75% instead 50% to
make sure the next wapbl_begin() will always flush.
It is still a mitigation, for a fix the file system must provide an
upper bound for every transaction to wapbl_begin().
PR kern/47146 user-triggerable crash in wapbl_register_deallocation (confi)
PR kern/60621 (panic: ufs_inactive: dirty filesystem?)
Implement a locking protocol for the 68k pmap structures. This essentially
replaces the old PMAP_CRIT_{ENTER,EXIT}() markers with PMAP_{LOCK,UNLOCK}()
calls.
On systems that aren't built for MP or for pmap lock debugging, this is
all a no-op as before.
Add UFS_WAPBL_UPDATE() after block allocations to group the
allocation and the inode update into the same transaction.
PR kern/60621 (panic: ufs_inactive: dirty filesystem?)
The 68010, 68020, and 68030 are have strictly ordered memory operations.
The 68040 and 68060 do not, however -- stores can be observed after
subsequent loads.
This is to say: all loads are load-acquire, all stores are store-release,
but store-before-load must be explicitly serialized. The 68040 User's
Manual states that a NOP insn forces this serialization.
And so: membar_aquire -> do nothing, membar_release -> do nothing,
membar_sync -> do nothing *with feeling*
Chapter and verse of the manual quoted in comments.
The MMU-specific functions that load the user root pointer already take
care of any necessary ATC / cache invalidations, so we don't need to do
it again here.
mmu_load_urp20hp(): No need to explicitly invalidate the external VAC
or the MMU's TLB -- user VAC entries and user TLB entries are automatically
invalidated when the USTP register is written, and there's no need to
invalidate supervisor entries in either of those when switching the user
address space.
pckbc(4) at acpi: Fix build with -Os.
Fixes:
/home/riastradh/netbsd/current/src/sys/dev/acpi/pckbc_acpi.c: In function 'pckbc_acpi_intr_establish':
/home/riastradh/netbsd/current/src/sys/dev/acpi/pckbc_acpi.c:242:63: error: 'handle' may be used uninitialized [-Werror=maybe-uninitialized]
242 | rv = acpi_intr_establish(sc->sc_dv, (uint64_t)(uintptr_t)handle,
| ^~~~~~~~~~~~~~~~~
/home/riastradh/netbsd/current/src/sys/dev/acpi/pckbc_acpi.c:224:21: note: 'handle' was declared here
224 | ACPI_HANDLE handle;
| ^~~~~~
No functional change intended -- the code wasn't _wrong_ before; this
just makes it clearer, with less analysis effort, that handle is
initialized in the path in question.
While here, rename local variable `rv' to `ih' to make it clearer
what it is (not just some arbitrary return value of a subroutine but
specifically an interrupt handler cookie).
[3 lines not shown]