Pull up following revision(s) (requested by riastradh in ticket #382):
external/bsd/unbound/etc/rc.d/unbound: revision 1.4
/etc/rc.d/unbound: Fix order of migration and trust anchor setup.
Add some comments explaining the order, and make sure we propagate
various failures back to the caller.
Based on a patch by Erik LaBine.
PR bin/60447: unbound rc.d fails on fresh install before
configuration migration can run
Pull up the following, requested by kre in ticket #387:
external/public-domain/tz/dist/Makefile up to 1.6
external/public-domain/tz/dist/NEWS up to 1.8
external/public-domain/tz/dist/TZDATA_VERSION up to 1.46
external/public-domain/tz/dist/africa up to 1.4
external/public-domain/tz/dist/australasia up to 1.13
external/public-domain/tz/dist/europe up to 1.6
external/public-domain/tz/dist/leap-seconds.list up to 1.12
external/public-domain/tz/dist/leapseconds up to 1.12
external/public-domain/tz/dist/northamerica up to 1.7
external/public-domain/tz/dist/theory.html up to 1.6
external/public-domain/tz/dist/version up to 1.19
external/public-domain/tz/dist/ziguard.awk up to 1.1.1.12
external/public-domain/tz/dist/zone.tab up to 1.6
external/public-domain/tz/dist/zone1970.tab up to 1.7
external/public-domain/tz/dist/zonenow.tab up to 1.8
Import tzdata2026cgtz
/bin/sh - fix a comment
The showjobs() function once (long ago) had a parameter "change"
which caused it to only show jobs whose status had altered since the
last call. It has since (in 2002) had that parameter's name changed
to "mode" and become a bit field, one of the bits of which is
SHOW_CHANGED, which when set does what the old change parameter did.
However, the comment that describes the function wasn't updated to match.
Now it has been. NFC.
Pull up following revision(s) (requested by thorpej in ticket #385):
sys/arch/sparc/conf/GENERIC: revision 1.277
sys/dev/sbus/if_gem_sbus.c: revision 1.16
Adapt the SBus config register for 32-bit SPARC systems and enable
GEM Ethernet in the GENERIC SPARC configuration.
From Erik LaBine.
PR kern/60453
Pull up following revision(s) (requested by thorpej in ticket #384):
sys/arch/sparc/sparc/iommu.c: revision 1.103
Add support for bus_dmamap_load_mbuf(), needed to support GEM Ethernet.
From Erik LaBine.
PR kern/60453
/bin/sh - minor -i & -m changes
When enabling job control (set -m) at the top level (not in any kind
of subshell) in a non-interactive shell, sh would sometimes send itself
a SIGTTIN (suspending itself) if it was unable to grab the controlling
terminal. Only interactive shells should be concerned with that,
so an interactive sh placed in the background, trying to manipulate
the controlling tty will simply suspend, until it is resumed in fg.
Scripts shouldn't need to care.
So only do controlling tty manipulation in interactive shells (-i is set).
While here, clean up the manual description of the -i flag, setting it on
the command line does not always make a shell interactive (the shell will
simply clear it if it is inappropriate), but clearing it (+i) always
prevents the shell becoming interactive. Manipulating it later using
'set' (which is a non-standard usage) is not recommended, and can cause
the shell to behave in unexpected ways, so advise against attempts to
do so (in the manual).
[3 lines not shown]
Pull up following revision(s) (requested by riastradh in ticket #1313):
sys/arch/xen/xen/xbd_xenbus.c: revision 1.136
xbd(4): Don't assert about indirect blocks if not supported.
PR port-xen/60448: xl suspend: panic: kernel diagnostic assertion
"incnt == __arraycount(sc->sc_indirect)" failed
Pull up following revision(s) (requested by riastradh in ticket #383):
sys/arch/xen/xen/xbd_xenbus.c: revision 1.136
xbd(4): Don't assert about indirect blocks if not supported.
PR port-xen/60448: xl suspend: panic: kernel diagnostic assertion
"incnt == __arraycount(sc->sc_indirect)" failed
Pull up following revision(s) (requested by kbowling in ticket #380):
sys/netinet/tcp_timer.h: revision 1.31
sys/netinet/tcp_subr.c: revision 1.300
tcp: set initial RTO to 1s per RFC 6298
Lower TCPTV_SRTTDFLT from 3s to 1s and adjust the t_rttvar seed in
tcp_tcpcb_template() so the pre-measurement TCP_REXMTVAL() equals
TCPTV_SRTTDFLT (was 2*TCPTV_SRTTDFLT). Active opens and SYN|ACK
retransmits now start at 1s and back off 1, 2, 4, ... per RFC 6298
sections 2.1 and 5.5. Post-first-sample RTO is unchanged.
Pin tcp_msl_remote_threshold to its previous numeric default (3*PR_SLOWHZ)
so the SRTTDFLT change does not move an unrelated TIME_WAIT-shortening
tunable.
This is conceptually https://reviews.freebsd.org/D18941 which we ran at a
large CDN, modified for NetBSD's stack.
Pull up following revision(s) (requested by riastradh in ticket #1312):
sys/kern/kern_event.c: revision 1.153
kqueue(2): Fix null pointer deref under memory pressure.
PR kern/60436: kqueue(2): random kernel null pointer deref in
EVFILT_PROC NOTE_TRACK
Pull up following revision(s) (requested by riastradh in ticket #379):
sys/kern/kern_event.c: revision 1.153
kqueue(2): Fix null pointer deref under memory pressure.
PR kern/60436: kqueue(2): random kernel null pointer deref in
EVFILT_PROC NOTE_TRACK
