uipc_usrreq: Fix unix-domain socket inode assignment.
1. `if (unp->unp_ino == 0) unp->unp_ino = unp_ino++' would assign
zero the first time around by mistake, and then assign nonzero
after that. `if (unp->unp_ino == 0) unp->unp_ino = ++unp_ino'
avoids this silliness.
2. Make the ++unp_ino part atomic.
PR kern/60729: incosistent st_ino from the first stat(2) on a socket
x86/sysbeep(4): Don't detach at shutdown.
This reverts sys/arch/x86/isa/clock.c rev. 1.32, which was the first
of a spate of changes to sprinkle DVF_DETACH_SHUTDOWN for reasons
that were unclear and apparently interfere with other functionality:
> Module Name: src
> Committed By: dyoung
> Date: Tue Apr 7 17:51:46 UTC 2009
>
> Modified Files:
> src/sys/arch/x86/isa: clock.c
>
> Log Message:
> Detach sysbeep0 at shutdown.
>
>
> To generate a diff of this commit:
> cvs rdiff -u -r1.31 -r1.32 src/sys/arch/x86/isa/clock.c
[18 lines not shown]
adjtime.2 & gettimeofday.2: minor wording and style tweaks
(Various man pages can't decide between "super user", "super-user", and
"superuser"; let's at least be consistent from line to line in one
man page.)
ld.elf_so: g/c empty conditional
It was not empty for about two hours before the commit that introduced
it had been reverted. Its content was revered, but the conditional
itself was forgoten.
While here, add "compat32" in the comment for the MLIBDIR block to
make this place more easily findable.
risc-v: revert the change for generic_bs_r_4
According to the RISC-V ABI Calling Convention an integer scalars
narrower than XLEN bits are widened according to the sign of their
type up to 32 bits, then sign-extended to XLEN bits.
wg(4): Assert task is nonzero before queueing work.
Prompted by (though I can't see any way it could be the source of):
PR kern/60728: workqueue panic in wg
arm64: don't consult mvfr1_el1 for AArch64 fp capabilities
The MVFR1_EL1 register describes features for the AArch32 AdvSIMD and FP
implementation and the contents of this register are UNKNOWN when FEAT_AA32
is not implemented.
So let's ignore this register and just set FPCR defaults for IEEE 754
compatibility.
Before this change, 7 tests in lib/libm failed on a VMware Fusion VM
(Apple M4 Pro) where MVFR1_EL1 reads as zero. With this change, all
tests pass.
Add interrupt support to pcfiic.
Make the repeat-start logic in the non-interrupt case simpler.
Add comments to softc fields.
By using interrupts, we greatly reduce the system time and wall clock time
used during i2c operations.
On machines where interrupts are not supported, the front end can set
the sc_poll flag to true, and the driver will use the polling path.
tzdata2netbsd - hg repo processing updates (mostly)
More updates to handle using mercurial as the repo, rather than cvs.
Mostly related to my gradual learning process of dealing with merges
and conflicts from the merge, correctly, using hg instead of cvs.
This also separated the merge from an attempt to update the sets list
at the same time (when an update is needed, which is rare), and (for
hg only) commit both in a single transaction, which would be ideal,
as if a rollback is needed, both parts need to be rolled back. But
it doesn't seem as if that is possible (or not without treating the
sets list as files to also be merged, which would have its own issues).
The only change here that affects usage when using a cvs repo, is that
the doc/3RDPARTY update now also updates the Date: field of the tz*
block, along with the Version (and if needed, Current-Version) field(s).
pci_resource: use existing BARs even if mem/io space is left disabled
It's possible that firmware assigned resources to a device and left
memory / io space accesses disabled. Let's assume if we see a non-zero
address in a BAR that it was assigned by firmware.
Should help PR# kern/60427 and kern/60463
