NetBSD/src n0bOrsh — usr.sbin/rdate rdate.8

   rdate.8: note RFC 868 is obsolescent

   From OpenBSD, specifically a mailing list post, which corrects the
   maximum effective date.
   https://marc.info/?l=openbsd-bugs&m=177447243611195&w=2

   US NIST still supports RFC 868 queries via time.nist.gov. (For
   comparison, the Canadian equivalent via NRC dropped support, citing
   security concerns.)
VersionDeltaFile
1.15+4-2usr.sbin/rdate/rdate.8
+4-21 files

NetBSD/src m3L0Nvz — sys/arch/hb68k/pg68k pmap_pgmmu.c, sys/arch/m68k/m68k pmap_68k.c pmap_motorola.c

   Change pmap_update() from a "delete call sites with a do nothing macro"
   to a do-nothing function.
VersionDeltaFile
1.115+15-2sys/arch/m68k/m68k/pmap_motorola.c
1.136+12-2sys/arch/sun3/sun3x/pmap.c
1.9+8-3sys/arch/sun68k/sun68k/pmap_sunmmu.c
1.68+8-3sys/arch/sun2/sun2/pmap.c
1.98+7-3sys/arch/m68k/m68k/pmap_68k.c
1.6+7-3sys/arch/hb68k/pg68k/pmap_pgmmu.c
+57-165 files not shown
+62-5111 files

NetBSD/src faXIQCZ — sys/kern kern_drvctl.c subr_autoconf.c, sys/sys device.h

   autoconf(9): Take device references in iteration for detach.

   This avoids a race with concurrent detach that can lead to tripping
   KASSERT(!dv->dv_detach_done) in config_detach -> device_acquire, even
   if the device_t itself remains stable while the device iteration is
   ongoing (because alldev_nwrite or alldev_nread is raised).

   While here, add:

   - deviter_first_acquire
   - deviter_next_acquire
   - device_find_by_xname_acquire
   - device_find_by_driver_unit_acquire

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.322+114-26sys/kern/subr_autoconf.c
1.54+6-5sys/kern/kern_drvctl.c
1.195+5-1sys/sys/device.h
+125-323 files

NetBSD/src wPrKY2g — sys/dev fss.c

   fss(4): Revoke vnodes when unconfiguring.

   This way, if you unconfigure an fss instance and reconfigure it,
   anything that had it open (e.g., a file system mounted from it) won't
   suddenly start mixing old cached data with new data -- instead, new
   I/O operations will fail.

   Not perfect -- see comments.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.133+61-11sys/dev/fss.c
+61-111 files

NetBSD/src kpwtvQJ — sys/dev fss.c

   fss(4): Use LK_SHARED, not LK_EXCLUSIVE, to read out mount point.

   No need to exclude other readers.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.132+6-4sys/dev/fss.c
+6-41 files

NetBSD/src bOXTPlm — sys/dev fss.c

   fss(4): Nix needless sc == NULL conditionals.

   fss_strategy and fss_ioctl can be reached only after fss_open has
   returned and (now that fss_cancel exists) before fss_close has been
   called (on both block _and_ all character devices that have been
   opened); between (first) open and (last) close, the autoconf unit
   must exist.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.131+3-12sys/dev/fss.c
+3-121 files

NetBSD/src E6LYKlM — sys/dev fss.c

   fss(4): Tighten fss_close a bit.

   Clear last FSS_CDEV_OPEN/FSS_BDEV_OPEN in fss_close so we can assert
   it in fss_detach, and pass DETACH_FORCE to fss_detach so it knows
   when to expect the flags to be cleared.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.130+15-19sys/dev/fss.c
+15-191 files

NetBSD/src ljijtuE — sys/dev fss.c

   fss(4): Tighten fss_close -> fss_detach path.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.129+15-6sys/dev/fss.c
+15-61 files

NetBSD/src 6MGkTef — sys/dev fss.c

   fss(4): Install a cancel handler.

   This way, fss_close doesn't run until all concurrent I/O operations
   (read, write, ioctl, strategy) have returned.  Note that this applies
   independently to the bdev and the cdev, so (e.g.) fss_close may run
   on the cdev while there's still an fss_strategy running in parallel.
   But in that case, sc->sc_flags will have FSS_BDEV_OPEN set, so
   fss_close won't try to detach it too soon.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.128+21-2sys/dev/fss.c
+21-21 files

NetBSD/src eGJ2Dn7 — sys/dev fss.c

   fss(4): Ensure newly attached autoconf unit matches dev_t minor.

   Setting cf->cf_fstate = FSTATE_STAR means autoconf(9) may select any
   unit from cf->cf_unit onward, rather than only cf->cf_unit.  Since
   the attach is always serialized by fss_device_lock, this proably
   doesn't make a difference in practice -- and we can't get rid of
   fss_device_lock without a config_intern_pseudo_acquire or something
   that atomically acquires the existing unit or if there is none
   attaches a new one.  But let's make the intent clearer in case any of
   that changes.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.127+4-3sys/dev/fss.c
+4-31 files

NetBSD/src XNnRo9y — sys/dev fss.c

   fss(4): Gather module init/fini logic into one place.

   Fixes some little-used error branches and avoids duplication that can
   get out of sync.

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.126+27-13sys/dev/fss.c
+27-131 files

NetBSD/src TYfSPXO — sys/dev fssvar.h fss.c

   fss(4): Omit needless kmem_alloc for struct disk sc_dkdev.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.125+13-14sys/dev/fss.c
1.33+2-2sys/dev/fssvar.h
+15-162 files

NetBSD/src r3hFG1k — sys/dev fss.c

   fss(4): Serialize fss_num_attached / fss_vfs_hooks access.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.124+11-2sys/dev/fss.c
+11-21 files

NetBSD/src yUt0BQr — sys/dev fss.c

   fss(4): Avoid races with drvctl detach.

   Attach and detach by opening and closing /dev/fssN is serialized by
   fss_device_lock, but concurrent `drvctl -d fssN' can cause an fss
   instance to be detached and freed while fss_open or fss_close is
   still trying to use it.

   Prevent concurrent detach during other operations like fss_ioctl or
   fss_strategy by making fss_detach fail if sc_flags shows it is held
   open.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.123+21-10sys/dev/fss.c
+21-101 files

NetBSD/src 70sXtzG — sys/dev fss.c

   fss(4): Rework creation path to avoid various races.

   1. Look up each vnode exactly once; the identity of the vnode may
      change between lookups otherwise.

   2. Don't read v_vflag or v_mount without holding the vnode lock;
      otherwise they may be changed by concurrent revoke.

   3. Hold vfs_busy(mp) while we don't have one of its vnode's locked
      and we haven't committed the fss instance to FSS_ACTIVE (at which
      point fss_unmount_hook will take care of it).

   4. Set sc->sc_uflags at the same time the transition to FSS_ACTIVE,
      and loosen assertion after fss_create_snapshot, to avoid a race
      between FSSIOCSET and FSSIOCCLR.  See comments in FSSIOCSET for
      details.

   XXX Should consider using namei and VOP_REMOVE instead of
   do_sys_unlink to remove only the vnode that we actually looked up,

    [3 lines not shown]
VersionDeltaFile
1.122+244-84sys/dev/fss.c
+244-841 files

NetBSD/src J54UD2v — sys/dev fss.c

   fss(4): Handle fss_softc_alloc failure.

   Shouldn't successfully transition to FSS_ACTIVE in that case; should
   instead back out and fail.

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.121+4-3sys/dev/fss.c
+4-31 files

NetBSD/src nRtfEeD — sys/dev fss.c

   fss(4): Free cfdata_t in error branch.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.120+3-2sys/dev/fss.c
+3-21 files

NetBSD/src YEFl86A — sys/dev fss.c

   fss(4): Initialize sc_state in fss_attach, not fss_open.

   Access without sc_slock is forbidden.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.119+3-3sys/dev/fss.c
+3-31 files

NetBSD/src h2upnTl — sys/dev fss.c

   fss(4): Sprinkle assertions in state transitions.

   Prompted by:

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.118+14-5sys/dev/fss.c
+14-51 files

NetBSD/src yRhbJ2b — sys/dev fss.c

   fss(4): Close vn_open with vn_close, not with vrele.

   Fixes

   # fssconfig fss0 / /dev/ld4
   ...
   panic: vrelel: bad ref count

   (The mistake is using the block device /dev/ld4 instead of the
   character device /dev/rld4, but this ought to be handled gracefully,
   not by panicking.)

   PR kern/60986: fssconfig panics on bad backing store type
VersionDeltaFile
1.117+5-5sys/dev/fss.c
+5-51 files

NetBSD/src aDjfrF1 — sys/arch/hp300/hp300 bus_space.c

   Oops, typo.
VersionDeltaFile
1.27+3-3sys/arch/hp300/hp300/bus_space.c
+3-31 files

NetBSD/src QQbiwEj — sys/arch/amiga/amiga amiga_init.c, sys/arch/hp300/hp300 bus_space.c autoconf.c

   Sprinkle a few missing pmap_update() call sites.
VersionDeltaFile
1.97+10-2sys/arch/m68k/m68k/pmap_68k.c
1.26+4-2sys/arch/hp300/hp300/bus_space.c
1.120+4-2sys/arch/hp300/hp300/autoconf.c
1.35+3-2sys/arch/mac68k/mac68k/bus_space.c
1.143+3-2sys/arch/amiga/amiga/amiga_init.c
+24-105 files

NetBSD/src q6IGoac — lib/libc/regex engine.c

   PR/61017: Juan Manuel Mendez Rey: regexec reads past the string after a
   multibyte character. Provide a prevchar() function that fetches the previous
   multibyte character.
VersionDeltaFile
1.31+41-10lib/libc/regex/engine.c
+41-101 files

NetBSD/src IW6ukhU — distrib/sets/lists/debug mi, distrib/sets/lists/tests mi

   New regex rest for PR/61017
VersionDeltaFile
1.1436+2-1distrib/sets/lists/tests/mi
1.526+2-1distrib/sets/lists/debug/mi
+4-22 files

NetBSD/src w6aFJEA — tests/lib/libc/regex Makefile t_regex.sh

   PR/61017: Juan Manuel Mendez Rey: regexec reads past the string after a
   multibyte character
VersionDeltaFile
1.1+57-0tests/lib/libc/regex/h_match.c
1.2+33-1tests/lib/libc/regex/t_regex.sh
1.14+2-2tests/lib/libc/regex/Makefile
+92-33 files

NetBSD/src ZkI5BUM — usr.sbin/npf/npfctl npf_data.c, usr.sbin/npf/npftest npftest.conf

   implement /32 IPv4 parsing in NPF.

   Timo Burhmester (timo@) requested in a recent tech-net thread on NPF.

   during the extraction of address/mask data, if mask is equal to max mask,
   both v4 and v6, it is assumed to be no net mask set.

   reviewed by martin@ and tests also included.
VersionDeltaFile
1.35+7-2usr.sbin/npf/npfctl/npf_data.c
1.15+9-0usr.sbin/npf/npftest/libnpftest/npf_nat_test.c
1.20+6-1usr.sbin/npf/npftest/npftest.conf
1.27+2-0usr.sbin/npf/npftest/libnpftest/npf_test.h
+24-34 files

NetBSD/src aDlIWsF — sys/arch/m68k/m68k genassym.cf

   Only define CI_LEV1MAP when TBL40_L1_SIZE is available.
VersionDeltaFile
1.12+3-1sys/arch/m68k/m68k/genassym.cf
+3-11 files

NetBSD/src nvJQzhq — sbin/newfs_lfs Makefile extern.h

   newfs_lfs(8): teach it to use findspecial so it can take NAME= syntax

   PR bin/61018 newfs_lfs(8) cannot take a wedge NAME=abc
VersionDeltaFile
1.31+6-4sbin/newfs_lfs/newfs.c
1.15+3-3sbin/newfs_lfs/extern.h
1.17+2-2sbin/newfs_lfs/Makefile
+11-93 files

NetBSD/src NYDw6Xx — sys/compat/linux/common linux_signal.c

   PR/61005: Bobby Bingham: compat_linux fails to set si_pid when delivering
   SI_QUEUE signal to Linux process because of cut-n-paste mistake
VersionDeltaFile
1.93+3-3sys/compat/linux/common/linux_signal.c
+3-31 files

NetBSD/src poVzNqB — usr.bin/kdump kdump.c

   PR/61007: tuxifan at posteo.de: Don't attempt to translate syscall arguments to
   names for non-native emulations
VersionDeltaFile
1.153+13-2usr.bin/kdump/kdump.c
+13-21 files