Merge 4142468ff4 into f2233ac33a

sysctl(8) prints a newline after the description, no need for this extra
newline.

MFC after:	1 week

RELNOTES

@@ -10,6 +10,11 @@ newline.  Entries should be separated by a newline.
 
 Changes to this file should not be MFCed.
 
+b2f7c53430c3:
+	Kernel TLS is now enabled by default in kernels including KTLS
+	support.  KTLS is included in GENERIC kernels for aarch64,
+	amd64, powerpc64, and powerpc64le.
+
 f57efe95cc25:
 	New mididump(1) utility which dumps MIDI 1.0 events in real time.
 

etc/mtree/BSD.tests.dist

@@ -505,6 +505,8 @@
         ..
         route
         ..
+        savecore
+        ..
         sysctl
         ..
     ..
@@ -1089,8 +1091,6 @@
         ..
         cut
         ..
-        dc
-        ..
         diff
         ..
         diff3

lib/libsys/fhreadlink.2

@@ -60,7 +60,7 @@ code in the global variable
 .Va errno .
 .Sh ERRORS
 The
-.Fn readlink
+.Fn fhreadlink
 system call
 will fail if:
 .Bl -tag -width Er
@@ -87,7 +87,8 @@ is no longer valid
 .El
 .Sh SEE ALSO
 .Xr fhlink 2 ,
-.Xr fhstat 2
+.Xr fhstat 2 ,
+.Xr readlink 2
 .Sh HISTORY
 The
 .Fn fhreadlink

lib/libsys/getsockopt.2

@@ -25,7 +25,7 @@
 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 .\" SUCH DAMAGE.
 .\"
-.Dd July 8, 2024
+.Dd November 25, 2024
 .Dt GETSOCKOPT 2
 .Os
 .Sh NAME
@@ -568,9 +568,14 @@ struct so_splice {
 .Pp
 Data received on
 .Fa s
-will automatically be transmitted from the socket specified in
+will automatically be transmitted via the socket specified in
 .Fa sp_fd
 without any intervention by userspace.
+That is, the data will be transmitted via
+.Fa sp_fd
+as if userspace had called
+.Xr send 2
+directly.
 Splicing is a one-way operation; a given pair of sockets may be
 spliced in one or both directions.
 Currently only connected

lib/libsys/readlink.2

@@ -138,6 +138,7 @@ is neither
 nor a file descriptor associated with a directory.
 .El
 .Sh SEE ALSO
+.Xr fhreadlink 2 ,
 .Xr lstat 2 ,
 .Xr stat 2 ,
 .Xr symlink 2 ,

libexec/rc/rc.conf

@@ -263,8 +263,8 @@ icmp_log_redirect="NO"		# Set to YES to log ICMP REDIRECT packets
 network_interfaces="auto"	# List of network interfaces (or "auto").
 cloned_interfaces=""		# List of cloned network interfaces to create.
 #cloned_interfaces="gif0 gif1 gif2 gif3" # Pre-cloning GENERIC config.
-#ifconfig_lo0="inet 127.0.0.1"	# default loopback device configuration.
-#ifconfig_lo0_alias0="inet 127.0.0.254 netmask 0xffffffff" # Sample alias entry.
+#ifconfig_lo0="inet 127.0.0.1/8"	# default loopback device configuration.
+#ifconfig_lo0_alias0="inet 127.0.0.254/32" # Sample alias entry.
 #ifconfig_em0_ipv6="inet6 2001:db8:1::1 prefixlen 64" # Sample IPv6 addr entry
 #ifconfig_em0_alias0="inet6 2001:db8:2::1 prefixlen 64" # Sample IPv6 alias
 #ifconfig_em0_name="net0"	# Change interface name from em0 to net0.

libexec/rc/rc.d/var_run

@@ -27,7 +27,7 @@ _var_run_load() {
 
 _var_run_save() {
 	if [ ! -d $(dirname ${var_run_mtree}) ]; then
-		mkdir -p ${var_run_mtree}
+		mkdir -p $(dirname ${var_run_mtree})
 	fi
 	mtree -dcbj -p /var/run > ${var_run_mtree}
 }

sbin/ffsinfo/ffsinfo.8

@@ -36,7 +36,7 @@
 .\"
 .\" $TSHeader: src/sbin/ffsinfo/ffsinfo.8,v 1.3 2000/12/12 19:30:55 tomsoft Exp $
 .\"
-.Dd September 8, 2000
+.Dd November 19, 2024
 .Dt FFSINFO 8
 .Os
 .Sh NAME
@@ -120,6 +120,7 @@ to
 .Pa /var/tmp/ffsinfo
 with all available information.
 .Sh SEE ALSO
+.Xr ffs 4 ,
 .Xr dumpfs 8 ,
 .Xr fsck 8 ,
 .Xr gpart 8 ,

sbin/mount/mount.8

@@ -25,7 +25,7 @@
 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 .\" SUCH DAMAGE.
 .\"
-.Dd January 24, 2024
+.Dd November 19, 2024
 .Dt MOUNT 8
 .Os
 .Sh NAME
@@ -568,6 +568,7 @@ support for a particular file system might be provided either on a static
 .Xr cd9660 4 ,
 .Xr devfs 4 ,
 .Xr ext2fs 4 ,
+.Xr ffs 4 ,
 .Xr mac 4 ,
 .Xr procfs 4 ,
 .Xr tarfs 4 ,

sbin/newfs/newfs.8

@@ -25,7 +25,7 @@
 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 .\" SUCH DAMAGE.
 .\"
-.Dd May 18, 2024
+.Dd November 19, 2024
 .Dt NEWFS 8
 .Os
 .Sh NAME
@@ -350,6 +350,7 @@ than the historical defaults
 This large fragment size may lead to much wasted space
 on file systems that contain many small files.
 .Sh SEE ALSO
+.Xr ffs 4 ,
 .Xr geom 4 ,
 .Xr disktab 5 ,
 .Xr fs 5 ,

sbin/savecore/Makefile

@@ -18,4 +18,7 @@ CFLAGS+=	-DWITH_CASPER
 LIBADD+=	casper cap_fileargs cap_syslog
 .endif
 
+HAS_TESTS=
+SUBDIR.${MK_TESTS}+= tests
+
 .include <bsd.prog.mk>

sbin/savecore/tests/Makefile

@@ -0,0 +1,3 @@
+ATF_TESTS_SH=	livedump_test
+
+.include <bsd.test.mk>

sbin/savecore/tests/livedump_test.sh

@@ -0,0 +1,54 @@
+#
+# SPDX-License-Identifier: BSD-2-Clause
+#
+# Copyright (c) 2024 Mark Johnston <markj@FreeBSD.org>
+#
+
+atf_test_case livedump_kldstat
+livedump_kldstat_head()
+{
+	atf_set "descr" "Test livedump integrity"
+	atf_set "require.progs" kgdb
+	atf_set "require.user" root
+}
+livedump_kldstat_body()
+{
+	atf_check savecore -L .
+
+	kernel=$(sysctl -n kern.bootfile)
+
+	if ! [ -f /usr/lib/debug/${kernel}.debug ]; then
+		atf_skip "No debug symbols for the running kernel"
+	fi
+
+	# Implement kldstat using gdb script.
+	cat >./kldstat.gdb <<'__EOF__'
+printf "Id Refs Address                Size Name\n"
+set $_lf = linker_files.tqh_first
+while ($_lf)
+    printf "%2d %4d %p %8x %s\n", $_lf->id, $_lf->refs, $_lf->address, $_lf->size, $_lf->filename
+    set $_lf = $_lf->link.tqe_next
+end
+__EOF__
+
+	# Ignore stderr since kgdb prints some warnings about inaccessible
+	# source files.
+	#
+	# Use a script to source the main gdb script, otherwise kgdb prints
+	# a bunch of line noise that is annoying to filter out.
+	echo "source ./kldstat.gdb" > ./script.gdb
+	atf_check -o save:out -e ignore \
+	    kgdb -q ${kernel} ./livecore.0 < ./script.gdb
+
+	# Get rid of gunk printed by kgdb.
+	sed -i '' -n -e 's/^(kgdb) //' -e '/^Id Refs /,$p' out
+
+	# The output of kgdb should match the output of kldstat.
+	atf_check -o save:kldstat kldstat
+	atf_check diff kldstat out
+}
+
+atf_init_test_cases()
+{
+	atf_add_test_case livedump_kldstat
+}

share/man/man4/ccr.4

@@ -22,7 +22,7 @@
 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 .\" SUCH DAMAGE.
 .\"
-.Dd June 3, 2019
+.Dd November 25, 2024
 .Dt CCR 4
 .Os
 .Sh NAME
@@ -52,7 +52,10 @@ The driver accelerates AES-CBC, AES-CCM, AES-CTR, AES-GCM, AES-XTS,
 SHA1, SHA2-224, SHA2-256, SHA2-384, SHA2-512,
 SHA1-HMAC, SHA2-224-HMAC, SHA2-256-HMAC, SHA2-384-HMAC, and SHA2-512-HMAC
 operations for
-.Xr crypto 4
+.Xr crypto 9
+consumers such as
+.Xr ktls 4 ,
+.Xr geli 4 ,
 and
 .Xr ipsec 4 .
 The driver also supports chaining one of AES-CBC, AES-CTR, or AES-XTS with
@@ -97,7 +100,11 @@ email all the specific information related to the issue to
 .Sh SEE ALSO
 .Xr crypto 4 ,
 .Xr cxgbe 4 ,
-.Xr ipsec 4
+.Xr geli 4 ,
+.Xr ipsec 4 ,
+.Xr ktls 4 ,
+.Xr crypto 7 ,
+.Xr crypto 9
 .Sh HISTORY
 The
 .Nm

sys/amd64/amd64/pmap.c

@@ -4082,7 +4082,19 @@ pmap_qremove(vm_offset_t sva, int count)
 
 	va = sva;
 	while (count-- > 0) {
+		/*
+		 * pmap_enter() calls within the kernel virtual
+		 * address space happen on virtual addresses from
+		 * subarenas that import superpage-sized and -aligned
+		 * address ranges.  So, the virtual address that we
+		 * allocate to use with pmap_qenter() can't be close
+		 * enough to one of those pmap_enter() calls for it to
+		 * be caught up in a promotion.
+		 */
 		KASSERT(va >= VM_MIN_KERNEL_ADDRESS, ("usermode va %lx", va));
+		KASSERT((*vtopde(va) & X86_PG_PS) == 0,
+		    ("pmap_qremove on promoted va %#lx", va));
+
 		pmap_kremove(va);
 		va += PAGE_SIZE;
 	}
@@ -10506,8 +10518,7 @@ pmap_map_io_transient(vm_page_t page[], vm_offset_t vaddr[], int count,
 {
 	vm_paddr_t paddr;
 	bool needs_mapping;
-	pt_entry_t *pte;
-	int cache_bits, error __unused, i;
+	int error __unused, i;
 
 	/*
 	 * Allocate any KVA space that we need, this is done in a separate
@@ -10552,11 +10563,8 @@ pmap_map_io_transient(vm_page_t page[], vm_offset_t vaddr[], int count,
 				 */
 				pmap_qenter(vaddr[i], &page[i], 1);
 			} else {
-				pte = vtopte(vaddr[i]);
-				cache_bits = pmap_cache_bits(kernel_pmap,
-				    page[i]->md.pat_mode, false);
-				pte_store(pte, paddr | X86_PG_RW | X86_PG_V |
-				    cache_bits);
+				pmap_kenter_attr(vaddr[i], paddr,
+				    page[i]->md.pat_mode);
 				pmap_invlpg(kernel_pmap, vaddr[i]);
 			}
 		}

sys/arm/arm/exec_machdep.c

@@ -101,14 +101,19 @@ get_vfpcontext(struct thread *td, mcontext_vfp_t *vfp)
 	    P_SHOULDSTOP(td->td_proc));
 
 	pcb = td->td_pcb;
-	if ((pcb->pcb_fpflags & PCB_FP_STARTED) != 0 && td == curthread) {
+	if (td == curthread) {
 		critical_enter();
 		vfp_store(&pcb->pcb_vfpstate, false);
 		critical_exit();
 	}
 	KASSERT(pcb->pcb_vfpsaved == &pcb->pcb_vfpstate,
 		("Called get_vfpcontext while the kernel is using the VFP"));
-	memcpy(vfp, &pcb->pcb_vfpstate, sizeof(*vfp));
+
+	memset(vfp, 0, sizeof(*vfp));
+	memcpy(vfp->mcv_reg, pcb->pcb_vfpstate.reg,
+	    sizeof(vfp->mcv_reg));
+	vfp->mcv_fpscr = pcb->pcb_vfpstate.fpscr;
+
 }
 
 /*
@@ -127,7 +132,10 @@ set_vfpcontext(struct thread *td, mcontext_vfp_t *vfp)
 	}
 	KASSERT(pcb->pcb_vfpsaved == &pcb->pcb_vfpstate,
 		("Called set_vfpcontext while the kernel is using the VFP"));
-	memcpy(&pcb->pcb_vfpstate, vfp, sizeof(*vfp));
+	memcpy(pcb->pcb_vfpstate.reg, vfp->mcv_reg,
+	    sizeof(pcb->pcb_vfpstate.reg));
+	pcb->pcb_vfpstate.fpscr = vfp->mcv_fpscr;
+
 }
 #endif
 
@@ -163,8 +171,6 @@ get_mcontext(struct thread *td, mcontext_t *mcp, int clear_ret)
 {
 	struct trapframe *tf = td->td_frame;
 	__greg_t *gr = mcp->__gregs;
-	mcontext_vfp_t	mcontext_vfp;
-	int rv;
 
 	if (clear_ret & GET_MC_CLEAR_RET) {
 		gr[_REG_R0] = 0;
@@ -189,19 +195,9 @@ get_mcontext(struct thread *td, mcontext_t *mcp, int clear_ret)
 	gr[_REG_LR]   = tf->tf_usr_lr;
 	gr[_REG_PC]   = tf->tf_pc;
 
-#ifdef VFP
-	if (mcp->mc_vfp_size != sizeof(mcontext_vfp_t))
-		return (EINVAL);
-	get_vfpcontext(td, &mcontext_vfp);
-#else
-	bzero(&mcontext_vfp, sizeof(mcontext_vfp));
-#endif
-
-	if (mcp->mc_vfp_ptr != NULL) {
-		rv = copyout(&mcontext_vfp, mcp->mc_vfp_ptr,  sizeof(mcontext_vfp));
-		if (rv != 0)
-			return (rv);
-	}
+	mcp->mc_vfp_size = 0;
+	mcp->mc_vfp_ptr = NULL;
+	memset(&mcp->mc_spare, 0, sizeof(mcp->mc_spare));
 
 	return (0);
 }
@@ -315,6 +311,16 @@ sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
 	/* Populate the siginfo frame. */
 	bzero(&frame, sizeof(frame));
 	get_mcontext(td, &frame.sf_uc.uc_mcontext, 0);
+
+#ifdef VFP
+	get_vfpcontext(td, &frame.sf_vfp);
+	frame.sf_uc.uc_mcontext.mc_vfp_size = sizeof(fp->sf_vfp);
+	frame.sf_uc.uc_mcontext.mc_vfp_ptr = &fp->sf_vfp;
+#else
+	frame.sf_uc.uc_mcontext.mc_vfp_size = 0;
+	frame.sf_uc.uc_mcontext.mc_vfp_ptr = NULL;
+#endif
+
 	frame.sf_si = ksi->ksi_info;
 	frame.sf_uc.uc_sigmask = *mask;
 	frame.sf_uc.uc_stack = td->td_sigstk;

sys/compat/linux/linux_xattr.c

@@ -98,7 +98,7 @@ error_to_xattrerror(int attrnamespace, int error)
 }
 
 static int
-xatrr_to_extattr(const char *uattrname, int *attrnamespace, char *attrname)
+xattr_to_extattr(const char *uattrname, int *attrnamespace, char *attrname)
 {
 	char uname[LINUX_XATTR_NAME_MAX + 1], *dot;
 	size_t len, cplen;
@@ -255,7 +255,7 @@ removexattr(struct thread *td, struct removexattr_args *args)
 	char attrname[LINUX_XATTR_NAME_MAX + 1];
 	int attrnamespace, error;
 
-	error = xatrr_to_extattr(args->name, &attrnamespace, attrname);
+	error = xattr_to_extattr(args->name, &attrnamespace, attrname);
 	if (error != 0)
 		return (error);
 	if (args->path != NULL)
@@ -312,7 +312,7 @@ getxattr(struct thread *td, struct getxattr_args *args)
 	char attrname[LINUX_XATTR_NAME_MAX + 1];
 	int attrnamespace, error;
 
-	error = xatrr_to_extattr(args->name, &attrnamespace, attrname);
+	error = xattr_to_extattr(args->name, &attrnamespace, attrname);
 	if (error != 0)
 		return (error);
 	if (args->path != NULL)
@@ -378,7 +378,7 @@ setxattr(struct thread *td, struct setxattr_args *args)
 	if ((args->flags & ~(LINUX_XATTR_FLAGS)) != 0 ||
 	    args->flags == (LINUX_XATTR_FLAGS))
 		return (EINVAL);
-	error = xatrr_to_extattr(args->name, &attrnamespace, attrname);
+	error = xattr_to_extattr(args->name, &attrnamespace, attrname);
 	if (error != 0)
 		return (error);
 

sys/conf/files.riscv

@@ -36,6 +36,7 @@ riscv/riscv/bus_machdep.c	standard
 riscv/riscv/bus_space_asm.S	standard
 riscv/riscv/busdma_bounce.c	standard
 riscv/riscv/busdma_machdep.c	standard
+riscv/riscv/cache.c		standard
 riscv/riscv/clock.c		standard
 riscv/riscv/copyinout.S		standard
 riscv/riscv/cpufunc_asm.S	standard
@@ -83,5 +84,7 @@ riscv/vmm/vmm_riscv.c				optional	vmm
 riscv/vmm/vmm_sbi.c				optional	vmm
 riscv/vmm/vmm_switch.S				optional	vmm
 
+riscv/thead/thead.c		standard
+
 # Zstd
 contrib/zstd/lib/freebsd/zstd_kfreebsd.c		optional zstdio compile-with ${ZSTD_C}

sys/dev/ena/ena_sysctl.c

@@ -158,7 +158,7 @@ SYSCTL_INT(_hw_ena, OID_AUTO, enable_9k_mbufs, CTLFLAG_RDTUN,
 int ena_force_large_llq_header = ENA_LLQ_HEADER_SIZE_POLICY_DEFAULT;
 SYSCTL_INT(_hw_ena, OID_AUTO, force_large_llq_header, CTLFLAG_RDTUN,
     &ena_force_large_llq_header, 0,
-    "Change default LLQ entry size received from the device\n");
+    "Change default LLQ entry size received from the device");
 
 int ena_rss_table_size = ENA_RX_RSS_TABLE_SIZE;
 

sys/dev/hid/hms.c

@@ -122,6 +122,7 @@ struct hms_softc {
 	hid_size_t		isize;
 	uint32_t		drift_cnt;
 	uint32_t		drift_thresh;
+	struct hid_location	wheel_loc;
 #endif
 };
 
@@ -131,6 +132,7 @@ hms_intr(void *context, void *buf, hid_size_t len)
 {
 	struct hidmap *hm = context;
 	struct hms_softc *sc = device_get_softc(hm->dev);
+	int32_t wheel;
 
 	if (len > sc->isize)
 		len = sc->isize;
@@ -140,8 +142,18 @@ hms_intr(void *context, void *buf, hid_size_t len)
 	 * to return last report data in sampling mode even after touch has
 	 * been ended.  That results in cursor drift.  Filter out such a
 	 * reports through comparing with previous one.
+	 *
+	 * Except this results in dropping consecutive mouse wheel events,
+	 * because differently from cursor movement they always move by the
+	 * same amount.  So, don't do it when there's mouse wheel movement.
 	 */
-	if (len == sc->last_irsize && memcmp(buf, sc->last_ir, len) == 0) {
+	if (sc->wheel_loc.size != 0)
+		wheel = hid_get_data(buf, len, &sc->wheel_loc);
+	else
+		wheel = 0;
+
+	if (len == sc->last_irsize && memcmp(buf, sc->last_ir, len) == 0 &&
+	    wheel == 0) {
 		sc->drift_cnt++;
 		if (sc->drift_thresh != 0 && sc->drift_cnt >= sc->drift_thresh)
 			return;
@@ -285,9 +297,25 @@ hms_attach(device_t dev)
 	/* Count number of input usages of variable type mapped to buttons */
 	for (hi = sc->hm.hid_items;
 	     hi < sc->hm.hid_items + sc->hm.nhid_items;
-	     hi++)
+	     hi++) {
 		if (hi->type == HIDMAP_TYPE_VARIABLE && hi->evtype == EV_KEY)
 			nbuttons++;
+#ifdef IICHID_SAMPLING
+		/*
+		 * Make note of which part of the report descriptor is the wheel.
+		 */
+		if (hi->type == HIDMAP_TYPE_VARIABLE &&
+		    hi->evtype == EV_REL && hi->code == REL_WHEEL) {
+			sc->wheel_loc = hi->loc;
+			/*
+			 * Account for the leading Report ID byte
+			 * if it is a multi-report device.
+			 */
+			if (hi->id != 0)
+				sc->wheel_loc.pos += 8;
+		}
+#endif
+	}
 
 	/* announce information about the mouse */
 	device_printf(dev, "%d buttons and [%s%s%s%s%s] coordinates ID=%u\n",

sys/dev/mlx5/mlx5_en/mlx5_en_hw_tls.c

@@ -409,7 +409,7 @@ mlx5e_tls_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params,
 	if (priv->gone != 0 || priv->tls.init == 0)
 		return (EOPNOTSUPP);
 
-	ptag = uma_zalloc(priv->tls.zone, M_WAITOK);
+	ptag = uma_zalloc(priv->tls.zone, M_NOWAIT);
  	if (ptag == NULL)
  		return (ENOMEM);
 

sys/dev/mlx5/mlx5_en/mlx5_en_main.c

@@ -3404,6 +3404,51 @@ mlx5e_set_rx_mode(if_t ifp)
 	queue_work(priv->wq, &priv->set_rx_mode_work);
 }
 
+static bool
+mlx5e_is_ipsec_capable(struct mlx5_core_dev *mdev)
+{
+#ifdef IPSEC_OFFLOAD
+	if ((mlx5_ipsec_device_caps(mdev) & MLX5_IPSEC_CAP_PACKET_OFFLOAD) != 0)
+		return (true);
+#endif
+	return (false);
+}
+
+static bool
+mlx5e_is_ratelimit_capable(struct mlx5_core_dev *mdev)
+{
+#ifdef RATELIMIT
+	if (MLX5_CAP_GEN(mdev, qos) &&
+	    MLX5_CAP_QOS(mdev, packet_pacing))
+		return (true);
+#endif
+	return (false);
+}
+
+static bool
+mlx5e_is_tlstx_capable(struct mlx5_core_dev *mdev)
+{
+#ifdef KERN_TLS
+	if (MLX5_CAP_GEN(mdev, tls_tx) != 0 &&
+	    MLX5_CAP_GEN(mdev, log_max_dek) != 0)
+		return (true);
+#endif
+	return (false);
+}
+
+static bool
+mlx5e_is_tlsrx_capable(struct mlx5_core_dev *mdev)
+{
+#ifdef KERN_TLS
+	if (MLX5_CAP_GEN(mdev, tls_rx) != 0 &&
+	    MLX5_CAP_GEN(mdev, log_max_dek) != 0 &&
+	    MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
+	    ft_field_support.outer_ip_version) != 0)
+		return (true);
+#endif
+	return (false);
+}
+
 static int
 mlx5e_ioctl(if_t ifp, u_long command, caddr_t data)
 {
@@ -3507,6 +3552,24 @@ mlx5e_ioctl(if_t ifp, u_long command, caddr_t data)
 		drv_ioctl_data = (struct siocsifcapnv_driver_data *)data;
 		PRIV_LOCK(priv);
 siocsifcap_driver:
+		if (!mlx5e_is_tlstx_capable(priv->mdev)) {
+			drv_ioctl_data->reqcap &= ~(IFCAP_TXTLS4 |
+			    IFCAP_TXTLS6);
+		}
+		if (!mlx5e_is_tlsrx_capable(priv->mdev)) {
+		        drv_ioctl_data->reqcap &= ~(
+			    IFCAP2_BIT(IFCAP2_RXTLS4) |
+			    IFCAP2_BIT(IFCAP2_RXTLS6));
+		}
+		if (!mlx5e_is_ipsec_capable(priv->mdev)) {
+			drv_ioctl_data->reqcap &=
+			    ~IFCAP2_BIT(IFCAP2_IPSEC_OFFLOAD);
+		}
+		if (!mlx5e_is_ratelimit_capable(priv->mdev)) {
+			drv_ioctl_data->reqcap &= ~(IFCAP_TXTLS_RTLMT |
+			    IFCAP_TXRTLMT);
+		}
+
 		mask = drv_ioctl_data->reqcap ^ if_getcapenable(ifp);
 
 		if (mask & IFCAP_TXCSUM) {
@@ -4535,29 +4598,20 @@ mlx5e_create_ifp(struct mlx5_core_dev *mdev)
 	if_setcapabilitiesbit(ifp, IFCAP_TSO | IFCAP_VLAN_HWTSO, 0);
 	if_setcapabilitiesbit(ifp, IFCAP_HWSTATS | IFCAP_HWRXTSTMP, 0);
 	if_setcapabilitiesbit(ifp, IFCAP_MEXTPG, 0);
-#ifdef KERN_TLS
-	if (MLX5_CAP_GEN(mdev, tls_tx) != 0 &&
-	    MLX5_CAP_GEN(mdev, log_max_dek) != 0)
+	if (mlx5e_is_tlstx_capable(mdev))
 		if_setcapabilitiesbit(ifp, IFCAP_TXTLS4 | IFCAP_TXTLS6, 0);
-	if (MLX5_CAP_GEN(mdev, tls_rx) != 0 &&
-	    MLX5_CAP_GEN(mdev, log_max_dek) != 0 &&
-	    MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
-	    ft_field_support.outer_ip_version) != 0)
+	if (mlx5e_is_tlsrx_capable(mdev))
 		if_setcapabilities2bit(ifp, IFCAP2_BIT(IFCAP2_RXTLS4) |
 		    IFCAP2_BIT(IFCAP2_RXTLS6), 0);
-#endif
-#ifdef RATELIMIT
-	if (MLX5_CAP_GEN(mdev, qos) &&
-	    MLX5_CAP_QOS(mdev, packet_pacing))
-		if_setcapabilitiesbit(ifp, IFCAP_TXRTLMT | IFCAP_TXTLS_RTLMT,
-		    0);
-#endif
+	if (mlx5e_is_ratelimit_capable(mdev)) {
+		if_setcapabilitiesbit(ifp, IFCAP_TXRTLMT, 0);
+		if (mlx5e_is_tlstx_capable(mdev))
+			if_setcapabilitiesbit(ifp, IFCAP_TXTLS_RTLMT, 0);
+	}
 	if_setcapabilitiesbit(ifp, IFCAP_VXLAN_HWCSUM | IFCAP_VXLAN_HWTSO, 0);
-#ifdef IPSEC_OFFLOAD
-	if (mlx5_ipsec_device_caps(mdev) & MLX5_IPSEC_CAP_PACKET_OFFLOAD)
+	if (mlx5e_is_ipsec_capable(mdev))
 		if_setcapabilities2bit(ifp, IFCAP2_BIT(IFCAP2_IPSEC_OFFLOAD),
 		    0);
-#endif
 
 	if_setsndtagallocfn(ifp, mlx5e_snd_tag_alloc);
 #ifdef RATELIMIT

sys/dev/sound/dummy.c

@@ -67,6 +67,24 @@ struct dummy_softc {
 	struct mtx *lock;
 };
 
+static bool
+dummy_active(struct dummy_softc *sc)
+{
+	struct dummy_chan *ch;
+	int i;
+
+	snd_mtxassert(sc->lock);
+
+	for (i = 0; i < sc->chnum; i++) {
+		ch = &sc->chans[i];
+		if (ch->run)
+			return (true);
+	}
+
+	/* No channel is running at the moment. */
+	return (false);
+}
+
 static void
 dummy_chan_io(void *arg)
 {
@@ -74,7 +92,9 @@ dummy_chan_io(void *arg)
 	struct dummy_chan *ch;
 	int i = 0;
 
-	snd_mtxlock(sc->lock);
+	/* Do not reschedule if no channel is running. */
+	if (!dummy_active(sc))
+		return;
 
 	for (i = 0; i < sc->chnum; i++) {
 		ch = &sc->chans[i];
@@ -89,8 +109,6 @@ dummy_chan_io(void *arg)
 		snd_mtxlock(sc->lock);
 	}
 	callout_schedule(&sc->callout, 1);
-
-	snd_mtxunlock(sc->lock);
 }
 
 static int
@@ -179,15 +197,15 @@ dummy_chan_trigger(kobj_t obj, void *data, int go)
 
 	switch (go) {
 	case PCMTRIG_START:
-		if (!callout_active(&sc->callout))
-			callout_reset(&sc->callout, 1, dummy_chan_io, sc);
 		ch->ptr = 0;
 		ch->run = 1;
+		callout_reset(&sc->callout, 1, dummy_chan_io, sc);
 		break;
 	case PCMTRIG_STOP:
 	case PCMTRIG_ABORT:
 		ch->run = 0;
-		if (callout_active(&sc->callout))
+		/* If all channels are stopped, stop the callout as well. */
+		if (!dummy_active(sc))
 			callout_stop(&sc->callout);
 	default:
 		break;
@@ -292,6 +310,7 @@ dummy_attach(device_t dev)
 	sc = device_get_softc(dev);
 	sc->dev = dev;
 	sc->lock = snd_mtxcreate(device_get_nameunit(dev), "snd_dummy softc");
+	callout_init_mtx(&sc->callout, sc->lock, 0);
 
 	sc->cap_fmts[0] = SND_FORMAT(AFMT_S32_LE, 2, 0);
 	sc->cap_fmts[1] = SND_FORMAT(AFMT_S24_LE, 2, 0);
@@ -316,7 +335,6 @@ dummy_attach(device_t dev)
 	if (pcm_register(dev, status))
 		return (ENXIO);
 	mixer_init(dev, &dummy_mixer_class, sc);
-	callout_init(&sc->callout, 1);
 
 	return (0);
 }
@@ -327,8 +345,8 @@ dummy_detach(device_t dev)
 	struct dummy_softc *sc = device_get_softc(dev);
 	int err;
 
-	callout_drain(&sc->callout);
 	err = pcm_unregister(dev);
+	callout_drain(&sc->callout);
 	snd_mtxfree(sc->lock);
 
 	return (err);

sys/dev/sound/pcm/channel.c

@@ -309,14 +309,7 @@ chn_wakeup(struct pcm_channel *c)
 	if (CHN_EMPTY(c, children.busy)) {
 		if (SEL_WAITING(sndbuf_getsel(bs)) && chn_polltrigger(c))
 			selwakeuppri(sndbuf_getsel(bs), PRIBIO);
-		if (c->flags & CHN_F_SLEEPING) {
-			/*
-			 * Ok, I can just panic it right here since it is
-			 * quite obvious that we never allow multiple waiters
-			 * from userland. I'm too generous...
-			 */
-			CHN_BROADCAST(&c->intr_cv);
-		}
+		CHN_BROADCAST(&c->intr_cv);
 	} else {
 		CHN_FOREACH(ch, c, children.busy) {
 			CHN_LOCK(ch);
@@ -332,15 +325,11 @@ chn_sleep(struct pcm_channel *c, int timeout)
 	int ret;
 
 	CHN_LOCKASSERT(c);
-	KASSERT((c->flags & CHN_F_SLEEPING) == 0,
-	    ("%s(): entered with CHN_F_SLEEPING", __func__));
 
 	if (c->flags & CHN_F_DEAD)
 		return (EINVAL);
 
-	c->flags |= CHN_F_SLEEPING;
 	ret = cv_timedwait_sig(&c->intr_cv, c->lock, timeout);
-	c->flags &= ~CHN_F_SLEEPING;
 
 	return ((c->flags & CHN_F_DEAD) ? EINVAL : ret);
 }
@@ -2318,44 +2307,46 @@ chn_trigger(struct pcm_channel *c, int go)
 	if (go == c->trigger)
 		return (0);
 
+	if (snd_verbose > 3) {
+		device_printf(c->dev, "%s() %s: calling go=0x%08x , "
+		    "prev=0x%08x\n", __func__, c->name, go, c->trigger);
+	}
+
+	c->trigger = go;
 	ret = CHANNEL_TRIGGER(c->methods, c->devinfo, go);
 	if (ret != 0)
 		return (ret);
 
+	CHN_UNLOCK(c);
+	PCM_LOCK(d);
+	CHN_LOCK(c);
+
+	/*
+	 * Do nothing if another thread set a different trigger while we had
+	 * dropped the mutex.
+	 */
+	if (go != c->trigger) {
+		PCM_UNLOCK(d);
+		return (0);
+	}
+
+	/*
+	 * Use the SAFE variants to prevent inserting/removing an already
+	 * existing/missing element.
+	 */
 	switch (go) {
 	case PCMTRIG_START:
-		if (snd_verbose > 3)
-			device_printf(c->dev,
-			    "%s() %s: calling go=0x%08x , "
-			    "prev=0x%08x\n", __func__, c->name, go,
-			    c->trigger);
-		if (c->trigger != PCMTRIG_START) {
-			c->trigger = go;
-			CHN_UNLOCK(c);
-			PCM_LOCK(d);
-			CHN_INSERT_HEAD(d, c, channels.pcm.busy);
-			PCM_UNLOCK(d);
-			CHN_LOCK(c);
-			chn_syncstate(c);
-		}
+		CHN_INSERT_HEAD_SAFE(d, c, channels.pcm.busy);
+		PCM_UNLOCK(d);
+		chn_syncstate(c);
 		break;
 	case PCMTRIG_STOP:
 	case PCMTRIG_ABORT:
-		if (snd_verbose > 3)
-			device_printf(c->dev,
-			    "%s() %s: calling go=0x%08x , "
-			    "prev=0x%08x\n", __func__, c->name, go,
-			    c->trigger);
-		if (c->trigger == PCMTRIG_START) {
-			c->trigger = go;
-			CHN_UNLOCK(c);
-			PCM_LOCK(d);
-			CHN_REMOVE(d, c, channels.pcm.busy);
-			PCM_UNLOCK(d);
-			CHN_LOCK(c);
-		}
+		CHN_REMOVE_SAFE(d, c, channels.pcm.busy);
+		PCM_UNLOCK(d);
 		break;
 	default:
+		PCM_UNLOCK(d);
 		break;
 	}
 

sys/dev/sound/pcm/channel.h

@@ -354,7 +354,7 @@ enum {
 #define CHN_F_RUNNING		0x00000004  /* dma is running */
 #define CHN_F_TRIGGERED		0x00000008
 #define CHN_F_NOTRIGGER		0x00000010
-#define CHN_F_SLEEPING		0x00000020
+/* unused			0x00000020 */
 
 #define CHN_F_NBIO              0x00000040  /* do non-blocking i/o */
 #define CHN_F_MMAP		0x00000080  /* has been mmap()ed */
@@ -362,7 +362,7 @@ enum {
 #define CHN_F_BUSY              0x00000100  /* has been opened 	*/
 #define CHN_F_DIRTY		0x00000200  /* need re-config */
 #define CHN_F_DEAD		0x00000400  /* too many errors, dead, mdk */
-#define CHN_F_SILENCE		0x00000800  /* silence, nil, null, yada */
+/* unused			0x00000800 */
 
 #define	CHN_F_HAS_SIZE		0x00001000  /* user set block size */
 #define CHN_F_HAS_VCHAN		0x00002000  /* vchan master */
@@ -381,14 +381,14 @@ enum {
 				"\002ABORTING"				\
 				"\003RUNNING"				\
 				"\004TRIGGERED"				\
+				/* \006 */				\
 				"\005NOTRIGGER"				\
-				"\006SLEEPING"				\
 				"\007NBIO"				\
 				"\010MMAP"				\
 				"\011BUSY"				\
 				"\012DIRTY"				\
 				"\013DEAD"				\
-				"\014SILENCE"				\
+				/* \014 */				\
 				"\015HAS_SIZE"				\
 				"\016HAS_VCHAN"				\
 				"\017VCHAN_PASSTHROUGH"			\

sys/dev/sound/pcm/dsp.c

@@ -137,7 +137,7 @@ dsp_destroy_dev(device_t dev)
 	struct snddev_info *d;
 
 	d = device_get_softc(dev);
-	destroy_dev_sched(d->dsp_dev);
+	destroy_dev(d->dsp_dev);
 }
 
 static void
@@ -177,7 +177,7 @@ dsp_close(void *data)
 
 	d = priv->sc;
 	/* At this point pcm_unregister() will destroy all channels anyway. */
-	if (!DSP_REGISTERED(d) || PCM_DETACHING(d))
+	if (!DSP_REGISTERED(d))
 		goto skip;
 
 	PCM_GIANT_ENTER(d);
@@ -264,7 +264,7 @@ dsp_open(struct cdev *i_dev, int flags, int mode, struct thread *td)
 		return (ENODEV);
 
 	d = i_dev->si_drv1;
-	if (!DSP_REGISTERED(d) || PCM_DETACHING(d))
+	if (!DSP_REGISTERED(d))
 		return (EBADF);
 
 	priv = malloc(sizeof(*priv), M_DEVBUF, M_WAITOK | M_ZERO);
@@ -445,7 +445,7 @@ dsp_io_ops(struct dsp_cdevpriv *priv, struct uio *buf)
 	    ("%s(): io train wreck!", __func__));
 
 	d = priv->sc;
-	if (!DSP_REGISTERED(d) || PCM_DETACHING(d))
+	if (!DSP_REGISTERED(d))
 		return (EBADF);
 
 	PCM_GIANT_ENTER(d);
@@ -664,7 +664,7 @@ dsp_ioctl(struct cdev *i_dev, u_long cmd, caddr_t arg, int mode,
 		return (err);
 
 	d = priv->sc;
-	if (!DSP_REGISTERED(d) || PCM_DETACHING(d))
+	if (!DSP_REGISTERED(d))
 		return (EBADF);
 
 	PCM_GIANT_ENTER(d);
@@ -1783,7 +1783,7 @@ dsp_poll(struct cdev *i_dev, int events, struct thread *td)
 	if ((err = devfs_get_cdevpriv((void **)&priv)) != 0)
 		return (err);
 	d = priv->sc;
-	if (!DSP_REGISTERED(d) || PCM_DETACHING(d)) {
+	if (!DSP_REGISTERED(d)) {
 		/* XXX many clients don't understand POLLNVAL */
 		return (events & (POLLHUP | POLLPRI | POLLIN |
 		    POLLRDNORM | POLLOUT | POLLWRNORM));
@@ -1865,7 +1865,7 @@ dsp_mmap_single(struct cdev *i_dev, vm_ooffset_t *offset,
 	if ((err = devfs_get_cdevpriv((void **)&priv)) != 0)
 		return (err);
 	d = priv->sc;
-	if (!DSP_REGISTERED(d) || PCM_DETACHING(d))
+	if (!DSP_REGISTERED(d))
 		return (EINVAL);
 
 	PCM_GIANT_ENTER(d);

sys/dev/sound/pcm/mixer.c

@@ -146,7 +146,7 @@ mixer_set_softpcmvol(struct snd_mixer *m, struct snddev_info *d,
 	struct pcm_channel *c;
 	int dropmtx, acquiremtx;
 
-	if (!PCM_REGISTERED(d) || PCM_DETACHING(d))
+	if (!PCM_REGISTERED(d))
 		return (EINVAL);
 
 	if (mtx_owned(m->lock))
@@ -199,7 +199,7 @@ mixer_set_eq(struct snd_mixer *m, struct snddev_info *d,
 	else
 		return (EINVAL);
 
-	if (!PCM_REGISTERED(d) || PCM_DETACHING(d))
+	if (!PCM_REGISTERED(d))
 		return (EINVAL);
 
 	if (mtx_owned(m->lock))
@@ -1053,7 +1053,7 @@ mixer_open(struct cdev *i_dev, int flags, int mode, struct thread *td)
 
 	m = i_dev->si_drv1;
 	d = device_get_softc(m->dev);
-	if (!PCM_REGISTERED(d) || PCM_DETACHING(d))
+	if (!PCM_REGISTERED(d))
 		return (EBADF);
 
 	/* XXX Need Giant magic entry ??? */
@@ -1209,7 +1209,7 @@ mixer_ioctl(struct cdev *i_dev, u_long cmd, caddr_t arg, int mode,
 		return (EBADF);
 
 	d = device_get_softc(((struct snd_mixer *)i_dev->si_drv1)->dev);
-	if (!PCM_REGISTERED(d) || PCM_DETACHING(d))
+	if (!PCM_REGISTERED(d))
 		return (EBADF);
 
 	PCM_GIANT_ENTER(d);
@@ -1447,7 +1447,7 @@ mixer_oss_mixerinfo(struct cdev *i_dev, oss_mixerinfo *mi)
 	for (i = 0; pcm_devclass != NULL &&
 	    i < devclass_get_maxunit(pcm_devclass); i++) {
 		d = devclass_get_softc(pcm_devclass, i);
-		if (!PCM_REGISTERED(d) || PCM_DETACHING(d)) {
+		if (!PCM_REGISTERED(d)) {
 			if ((mi->dev == -1 && i == snd_unit) || mi->dev == i) {
 				mixer_oss_mixerinfo_unavail(mi, i);
 				return (0);

sys/dev/sound/pcm/sound.c

@@ -211,40 +211,53 @@ static void
 pcm_killchans(struct snddev_info *d)
 {
 	struct pcm_channel *ch;
-	bool found;
+	bool again;
 
 	PCM_BUSYASSERT(d);
-	do {
-		found = false;
+	KASSERT(!PCM_REGISTERED(d), ("%s(): still registered\n", __func__));
+
+	for (;;) {
+		again = false;
+		/* Make sure all channels are stopped. */
 		CHN_FOREACH(ch, d, channels.pcm) {
 			CHN_LOCK(ch);
-			/*
-			 * Make sure no channel has went to sleep in the
-			 * meantime.
-			 */
-			chn_shutdown(ch);
-			/*
-			 * We have to give a thread sleeping in chn_sleep() a
-			 * chance to observe that the channel is dead.
-			 */
-			if ((ch->flags & CHN_F_SLEEPING) == 0) {
-				found = true;
+			if (ch->intr_cv.cv_waiters == 0 && CHN_STOPPED(ch) &&
+			    ch->inprog == 0) {
 				CHN_UNLOCK(ch);
-				break;
+				continue;
 			}
+			chn_shutdown(ch);
+			if (ch->direction == PCMDIR_PLAY)
+				chn_flush(ch);
+			else
+				chn_abort(ch);
 			CHN_UNLOCK(ch);
+			again = true;
 		}
-
 		/*
-		 * All channels are still sleeping. Sleep for a bit and try
-		 * again to see if any of them is awake now.
+		 * Some channels are still active. Sleep for a bit and try
+		 * again.
 		 */
-		if (!found) {
-			pause_sbt("pcmkillchans", SBT_1MS * 5, 0, 0);
-			continue;
-		}
+		if (again)
+			pause_sbt("pcmkillchans", mstosbt(5), 0, 0);
+		else
+			break;
+	}
+
+	/* All channels are finally dead. */
+	while (!CHN_EMPTY(d, channels.pcm)) {
+		ch = CHN_FIRST(d, channels.pcm);
 		chn_kill(ch);
-	} while (!CHN_EMPTY(d, channels.pcm));
+	}
+
+	if (d->p_unr != NULL)
+		delete_unrhdr(d->p_unr);
+	if (d->vp_unr != NULL)
+		delete_unrhdr(d->vp_unr);
+	if (d->r_unr != NULL)
+		delete_unrhdr(d->r_unr);
+	if (d->vr_unr != NULL)
+		delete_unrhdr(d->vr_unr);
 }
 
 static int
@@ -512,7 +525,6 @@ int
 pcm_unregister(device_t dev)
 {
 	struct snddev_info *d;
-	struct pcm_channel *ch;
 
 	d = device_get_softc(dev);
 
@@ -524,29 +536,14 @@ pcm_unregister(device_t dev)
 	PCM_LOCK(d);
 	PCM_WAIT(d);
 
-	d->flags |= SD_F_DETACHING;
+	d->flags &= ~SD_F_REGISTERED;
 
 	PCM_ACQUIRE(d);
 	PCM_UNLOCK(d);
 
-	CHN_FOREACH(ch, d, channels.pcm) {
-		CHN_LOCK(ch);
-		/*
-		 * Do not wait for the timeout in chn_read()/chn_write(). Wake
-		 * up the sleeping thread and kill the channel.
-		 */
-		chn_shutdown(ch);
-		chn_abort(ch);
-		CHN_UNLOCK(ch);
-	}
+	pcm_killchans(d);
 
-	/* remove /dev/sndstat entry first */
-	sndstat_unregister(dev);
-
-	PCM_LOCK(d);
-	d->flags |= SD_F_DYING;
-	d->flags &= ~SD_F_REGISTERED;
-	PCM_UNLOCK(d);
+	PCM_RELEASE_QUICK(d);
 
 	if (d->play_sysctl_tree != NULL) {
 		sysctl_ctx_free(&d->play_sysctl_ctx);
@@ -557,24 +554,12 @@ pcm_unregister(device_t dev)
 		d->rec_sysctl_tree = NULL;
 	}
 
+	sndstat_unregister(dev);
+	mixer_uninit(dev);
 	dsp_destroy_dev(dev);
-	(void)mixer_uninit(dev);
 
-	pcm_killchans(d);
-
-	PCM_LOCK(d);
-	PCM_RELEASE(d);
 	cv_destroy(&d->cv);
-	PCM_UNLOCK(d);
 	snd_mtxfree(d->lock);
-	if (d->p_unr != NULL)
-		delete_unrhdr(d->p_unr);
-	if (d->vp_unr != NULL)
-		delete_unrhdr(d->vp_unr);
-	if (d->r_unr != NULL)
-		delete_unrhdr(d->r_unr);
-	if (d->vr_unr != NULL)
-		delete_unrhdr(d->vr_unr);
 
 	if (snd_unit == device_get_unit(dev)) {
 		snd_unit = pcm_best_unit(-1);

sys/dev/sound/pcm/sound.h

@@ -104,17 +104,15 @@ struct snd_mixer;
 #define SD_F_SIMPLEX		0x00000001
 #define SD_F_AUTOVCHAN		0x00000002
 #define SD_F_SOFTPCMVOL		0x00000004
-#define SD_F_DYING		0x00000008
-#define SD_F_DETACHING		0x00000010
-#define SD_F_BUSY		0x00000020
-#define SD_F_MPSAFE		0x00000040
-#define SD_F_REGISTERED		0x00000080
-#define SD_F_BITPERFECT		0x00000100
-#define SD_F_VPC		0x00000200	/* volume-per-channel */
-#define SD_F_EQ			0x00000400	/* EQ */
-#define SD_F_EQ_ENABLED		0x00000800	/* EQ enabled */
-#define SD_F_EQ_BYPASSED	0x00001000	/* EQ bypassed */
-#define SD_F_EQ_PC		0x00002000	/* EQ per-channel */
+#define SD_F_BUSY		0x00000008
+#define SD_F_MPSAFE		0x00000010
+#define SD_F_REGISTERED		0x00000020
+#define SD_F_BITPERFECT		0x00000040
+#define SD_F_VPC		0x00000080	/* volume-per-channel */
+#define SD_F_EQ			0x00000100	/* EQ */
+#define SD_F_EQ_ENABLED		0x00000200	/* EQ enabled */
+#define SD_F_EQ_BYPASSED	0x00000400	/* EQ bypassed */
+#define SD_F_EQ_PC		0x00000800	/* EQ per-channel */
 
 #define SD_F_EQ_DEFAULT		(SD_F_EQ | SD_F_EQ_ENABLED)
 #define SD_F_EQ_MASK		(SD_F_EQ | SD_F_EQ_ENABLED |		\
@@ -127,26 +125,20 @@ struct snd_mixer;
 				"\001SIMPLEX"				\
 				"\002AUTOVCHAN"				\
 				"\003SOFTPCMVOL"			\
-				"\004DYING"				\
-				"\005DETACHING"				\
-				"\006BUSY"				\
-				"\007MPSAFE"				\
-				"\010REGISTERED"			\
-				"\011BITPERFECT"			\
-				"\012VPC"				\
-				"\013EQ"				\
-				"\014EQ_ENABLED"			\
-				"\015EQ_BYPASSED"			\
-				"\016EQ_PC"				\
+				"\004BUSY"				\
+				"\005MPSAFE"				\
+				"\006REGISTERED"			\
+				"\007BITPERFECT"			\
+				"\010VPC"				\
+				"\011EQ"				\
+				"\012EQ_ENABLED"			\
+				"\013EQ_BYPASSED"			\
+				"\014EQ_PC"				\
 				"\035PRIO_RD"				\
 				"\036PRIO_WR"
 
-#define PCM_ALIVE(x)		((x) != NULL && (x)->lock != NULL &&	\
-				 !((x)->flags & SD_F_DYING))
-#define PCM_REGISTERED(x)	(PCM_ALIVE(x) &&			\
-				 ((x)->flags & SD_F_REGISTERED))
-
-#define	PCM_DETACHING(x)	((x)->flags & SD_F_DETACHING)
+#define PCM_ALIVE(x)		((x) != NULL && (x)->lock != NULL)
+#define PCM_REGISTERED(x)	(PCM_ALIVE(x) && ((x)->flags & SD_F_REGISTERED))
 
 #define	PCM_CHANCOUNT(d)	\
 	(d->playcount + d->pvchancount + d->reccount + d->rvchancount)

sys/dev/sound/pcm/vchan.c

@@ -146,20 +146,19 @@ vchan_trigger(kobj_t obj, void *data, int go)
 	int ret, otrigger;
 
 	info = data;
+	c = info->channel;
+	p = c->parentchannel;
 
+	CHN_LOCKASSERT(c);
 	if (!PCMTRIG_COMMON(go) || go == info->trigger)
 		return (0);
 
-	c = info->channel;
-	p = c->parentchannel;
-	otrigger = info->trigger;
-	info->trigger = go;
-
-	CHN_LOCKASSERT(c);
-
 	CHN_UNLOCK(c);
 	CHN_LOCK(p);
 
+	otrigger = info->trigger;
+	info->trigger = go;
+
 	switch (go) {
 	case PCMTRIG_START:
 		if (otrigger != PCMTRIG_START)

sys/kern/kern_sig.c

@@ -365,6 +365,16 @@ sigqueue_start(void)
 	SIGFILLSET(fastblock_mask);
 	SIG_CANTMASK(fastblock_mask);
 	ast_register(TDA_SIG, ASTR_UNCOND, 0, ast_sig);
+
+	/*
+	 * TDA_PSELECT is for the case where the signal mask should be restored
+	 * before delivering any signals so that we do not deliver any that are
+	 * blocked by the normal thread mask.  It is mutually exclusive with
+	 * TDA_SIGSUSPEND, which should be used if we *do* want to deliver
+	 * signals that are normally blocked, e.g., if it interrupted our sleep.
+	 */
+	ast_register(TDA_PSELECT, ASTR_ASTF_REQUIRED | ASTR_TDP,
+	    TDP_OLDMASK, ast_sigsuspend);
 	ast_register(TDA_SIGSUSPEND, ASTR_ASTF_REQUIRED | ASTR_TDP,
 	    TDP_OLDMASK, ast_sigsuspend);
 }

sys/kern/kern_vnodedumper.c

@@ -133,8 +133,10 @@ livedump_start_vnode(struct vnode *vp, int flags, uint8_t compression)
 	if (error != 0)
 		goto out;
 
+	curthread->td_pflags2 |= TDP2_SAN_QUIET;
 	dump_savectx();
 	error = minidumpsys(livedi, true);
+	curthread->td_pflags2 &= ~TDP2_SAN_QUIET;
 
 	EVENTHANDLER_INVOKE(livedumper_finish);
 out:

sys/kern/subr_asan.c

@@ -405,6 +405,9 @@ kasan_shadow_check(unsigned long addr, size_t size, bool write,
 
 	if (__predict_false(!kasan_enabled))
 		return;
+	if (__predict_false(curthread != NULL &&
+	    (curthread->td_pflags2 & TDP2_SAN_QUIET) != 0))
+		return;
 	if (__predict_false(size == 0))
 		return;
 	if (__predict_false(kasan_md_unsupported(addr)))

sys/kern/subr_msan.c

@@ -179,6 +179,9 @@ kmsan_report_hook(const void *addr, msan_orig_t *orig, size_t size, size_t off,
 
 	if (__predict_false(KERNEL_PANICKED() || kdb_active || kmsan_reporting))
 		return;
+	if (__predict_false(curthread != NULL &&
+	    (curthread->td_pflags2 & TDP2_SAN_QUIET) != 0))
+		return;
 
 	kmsan_reporting = true;
 	__compiler_membar();
@@ -232,6 +235,9 @@ kmsan_report_inline(msan_orig_t orig, unsigned long pc)
 
 	if (__predict_false(KERNEL_PANICKED() || kdb_active || kmsan_reporting))
 		return;
+	if (__predict_false(curthread != NULL &&
+	    (curthread->td_pflags2 & TDP2_SAN_QUIET) != 0))
+		return;
 
 	kmsan_reporting = true;
 	__compiler_membar();

sys/kern/sys_generic.c

@@ -1049,14 +1049,26 @@ kern_pselect(struct thread *td, int nd, fd_set *in, fd_set *ou, fd_set *ex,
 		if (error != 0)
 			return (error);
 		td->td_pflags |= TDP_OLDMASK;
+	}
+	error = kern_select(td, nd, in, ou, ex, tvp, abi_nfdbits);
+	if (uset != NULL) {
 		/*
 		 * Make sure that ast() is called on return to
 		 * usermode and TDP_OLDMASK is cleared, restoring old
-		 * sigmask.
+		 * sigmask.  If we didn't get interrupted, then the caller is
+		 * likely not expecting a signal to hit that should normally be
+		 * blocked by its signal mask, so we restore the mask before
+		 * any signals could be delivered.
 		 */
-		ast_sched(td, TDA_SIGSUSPEND);
+		if (error == EINTR) {
+			ast_sched(td, TDA_SIGSUSPEND);
+		} else {
+			/* *select(2) should never restart. */
+			MPASS(error != ERESTART);
+			ast_sched(td, TDA_PSELECT);
+		}
 	}
-	error = kern_select(td, nd, in, ou, ex, tvp, abi_nfdbits);
+
 	return (error);
 }
 
@@ -1528,12 +1540,6 @@ kern_poll_kfds(struct thread *td, struct pollfd *kfds, u_int nfds,
 		if (error)
 			return (error);
 		td->td_pflags |= TDP_OLDMASK;
-		/*
-		 * Make sure that ast() is called on return to
-		 * usermode and TDP_OLDMASK is cleared, restoring old
-		 * sigmask.
-		 */
-		ast_sched(td, TDA_SIGSUSPEND);
 	}
 
 	seltdinit(td);
@@ -1556,6 +1562,22 @@ kern_poll_kfds(struct thread *td, struct pollfd *kfds, u_int nfds,
 		error = EINTR;
 	if (error == EWOULDBLOCK)
 		error = 0;
+
+	if (uset != NULL) {
+		/*
+		 * Make sure that ast() is called on return to
+		 * usermode and TDP_OLDMASK is cleared, restoring old
+		 * sigmask.  If we didn't get interrupted, then the caller is
+		 * likely not expecting a signal to hit that should normally be
+		 * blocked by its signal mask, so we restore the mask before
+		 * any signals could be delivered.
+		 */
+		if (error == EINTR)
+			ast_sched(td, TDA_SIGSUSPEND);
+		else
+			ast_sched(td, TDA_PSELECT);
+	}
+
 	return (error);
 }
 

sys/kern/uipc_ktls.c

@@ -136,7 +136,7 @@ SYSCTL_UINT(_kern_ipc_tls, OID_AUTO, ifnet_max_rexmit_pct, CTLFLAG_RWTUN,
     &ktls_ifnet_max_rexmit_pct, 2,
     "Max percent bytes retransmitted before ifnet TLS is disabled");
 
-static bool ktls_offload_enable;
+static bool ktls_offload_enable = true;
 SYSCTL_BOOL(_kern_ipc_tls, OID_AUTO, enable, CTLFLAG_RWTUN,
     &ktls_offload_enable, 0,
     "Enable support for kernel TLS offload");
@@ -273,7 +273,7 @@ SYSCTL_COUNTER_U64(_kern_ipc_tls_ifnet, OID_AUTO, reset_failed, CTLFLAG_RD,
     &ktls_ifnet_reset_failed,
     "TLS sessions that failed to allocate a new ifnet send tag");
 
-static int ktls_ifnet_permitted;
+static int ktls_ifnet_permitted = 1;
 SYSCTL_UINT(_kern_ipc_tls_ifnet, OID_AUTO, permitted, CTLFLAG_RWTUN,
     &ktls_ifnet_permitted, 1,
     "Whether to permit hardware (ifnet) TLS sessions");

sys/kern/vfs_cache.c

@@ -3265,6 +3265,8 @@ kern___realpathat(struct thread *td, int fd, const char *path, char *buf,
 
 	if (nd.ni_vp->v_type == VREG && nd.ni_dvp->v_type != VDIR &&
 	    (nd.ni_vp->v_vflag & VV_ROOT) != 0) {
+		struct vnode *covered_vp;
+
 		/*
 		 * This happens if vp is a file mount. The call to
 		 * vn_fullpath_hardlink can panic if path resolution can't be
@@ -3274,7 +3276,6 @@ kern___realpathat(struct thread *td, int fd, const char *path, char *buf,
 		 * this should have a unique global path since we disallow
 		 * mounting on linked files.
 		 */
-		struct vnode *covered_vp;
 		error = vn_lock(nd.ni_vp, LK_SHARED);
 		if (error != 0)
 			goto out;
@@ -3284,11 +3285,20 @@ kern___realpathat(struct thread *td, int fd, const char *path, char *buf,
 		error = vn_fullpath(covered_vp, &retbuf, &freebuf);
 		vrele(covered_vp);
 	} else {
-		error = vn_fullpath_hardlink(nd.ni_vp, nd.ni_dvp, nd.ni_cnd.cn_nameptr,
-		    nd.ni_cnd.cn_namelen, &retbuf, &freebuf, &size);
+		error = vn_fullpath_hardlink(nd.ni_vp, nd.ni_dvp,
+		    nd.ni_cnd.cn_nameptr, nd.ni_cnd.cn_namelen, &retbuf,
+		    &freebuf, &size);
 	}
 	if (error == 0) {
-		error = copyout(retbuf, buf, size);
+		size_t len;
+
+		len = strlen(retbuf) + 1;
+		if (size < len)
+			error = ENAMETOOLONG;
+		else if (pathseg == UIO_USERSPACE)
+			error = copyout(retbuf, buf, len);
+		else
+			memcpy(buf, retbuf, len);
 		free(freebuf, M_TEMP);
 	}
 out:

sys/net/pfvar.h

@@ -70,6 +70,8 @@
 
 #ifdef _KERNEL
 
+#define        PF_PFIL_NOREFRAGMENT    0x80000000
+
 #if defined(__arm__)
 #define PF_WANT_32_TO_64_COUNTER
 #endif
@@ -2372,7 +2374,8 @@ void	pf_poolmask(struct pf_addr *, struct pf_addr*,
 	    struct pf_addr *, struct pf_addr *, sa_family_t);
 void	pf_addr_inc(struct pf_addr *, sa_family_t);
 int	pf_max_frag_size(struct mbuf *);
-int	pf_refragment6(struct ifnet *, struct mbuf **, struct m_tag *, bool);
+int	pf_refragment6(struct ifnet *, struct mbuf **, struct m_tag *,
+	    struct ifnet *, bool);
 #endif /* INET6 */
 
 int	pf_multihome_scan_init(int, int, struct pf_pdesc *);

sys/netipsec/ipsec_output.c

@@ -1112,7 +1112,7 @@ ipsec_encap(struct mbuf **mp, struct secasindex *saidx)
 #endif
 	struct ip *ip;
 #ifdef INET
-	int setdf;
+	int setdf = V_ip4_ipsec_dfbit == 1 ? 1: 0;
 #endif
 	uint8_t itos, proto;
 
@@ -1122,17 +1122,11 @@ ipsec_encap(struct mbuf **mp, struct secasindex *saidx)
 	case IPVERSION:
 		proto = IPPROTO_IPIP;
 		/*
-		 * Collect IP_DF state from the inner header
-		 * and honor system-wide control of how to handle it.
+		 * Copy IP_DF flag from the inner header if
+		 * system-wide control variable is greater than 1.
 		 */
-		switch (V_ip4_ipsec_dfbit) {
-		case 0:	/* clear in outer header */
-		case 1:	/* set in outer header */
-			setdf = V_ip4_ipsec_dfbit;
-			break;
-		default:/* propagate to outer header */
+		if (V_ip4_ipsec_dfbit > 1)
 			setdf = (ip->ip_off & htons(IP_DF)) != 0;
-		}
 		itos = ip->ip_tos;
 		break;
 #endif

sys/netpfil/pf/pf.c

@@ -7927,6 +7927,7 @@ pf_route6(struct mbuf **m, struct pf_krule *r, struct ifnet *oifp,
     struct pf_kstate *s, struct pf_pdesc *pd, struct inpcb *inp)
 {
 	struct mbuf		*m0, *md;
+	struct m_tag		*mtag;
 	struct sockaddr_in6	dst;
 	struct ip6_hdr		*ip6;
 	struct pfi_kkif		*nkif = NULL;
@@ -8053,8 +8054,8 @@ pf_route6(struct mbuf **m, struct pf_krule *r, struct ifnet *oifp,
 	}
 
 	if (pd->dir == PF_IN) {
-		if (pf_test(AF_INET6, PF_OUT, PFIL_FWD, ifp, &m0, inp,
-		    &pd->act) != PF_PASS) {
+		if (pf_test(AF_INET6, PF_OUT, PFIL_FWD | PF_PFIL_NOREFRAGMENT,
+		    ifp, &m0, inp, &pd->act) != PF_PASS) {
 			SDT_PROBE1(pf, ip6, route_to, drop, __LINE__);
 			goto bad;
 		} else if (m0 == NULL) {
@@ -8087,6 +8088,14 @@ pf_route6(struct mbuf **m, struct pf_krule *r, struct ifnet *oifp,
 	 */
 	if (IN6_IS_SCOPE_EMBED(&dst.sin6_addr))
 		dst.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
+	mtag = m_tag_find(m0, PACKET_TAG_PF_REASSEMBLED, NULL);
+	if (mtag != NULL) {
+		int ret;
+		ret = pf_refragment6(ifp, &m0, mtag, ifp, true);
+		SDT_PROBE2(pf, ip6, route_to, output, ifp, ret);
+		goto done;
+	}
+
 	if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
 		md = m0;
 		pf_dummynet_route(pd, s, r, ifp, sintosa(&dst), &md);
@@ -9474,14 +9483,15 @@ eat_pkt:
 	if (s)
 		PF_STATE_UNLOCK(s);
 
+out:
 #ifdef INET6
 	/* If reassembled packet passed, create new fragments. */
 	if (af == AF_INET6 && action == PF_PASS && *m0 && dir == PF_OUT &&
+	    (! (pflags & PF_PFIL_NOREFRAGMENT)) &&
 	    (mtag = m_tag_find(pd.m, PACKET_TAG_PF_REASSEMBLED, NULL)) != NULL)
-		action = pf_refragment6(ifp, m0, mtag, pflags & PFIL_FWD);
+		action = pf_refragment6(ifp, m0, mtag, NULL, pflags & PFIL_FWD);
 #endif
 
-out:
 	pf_sctp_multihome_delayed(&pd, kif, s, action);
 
 	return (action);

sys/netpfil/pf/pf_norm.c

@@ -42,6 +42,7 @@
 #include <sys/socket.h>
 
 #include <net/if.h>
+#include <net/if_var.h>
 #include <net/vnet.h>
 #include <net/pfvar.h>
 #include <net/if_pflog.h>
@@ -49,6 +50,8 @@
 #include <netinet/in.h>
 #include <netinet/ip.h>
 #include <netinet/ip_var.h>
+#include <netinet6/in6_var.h>
+#include <netinet6/nd6.h>
 #include <netinet6/ip6_var.h>
 #include <netinet6/scope6_var.h>
 #include <netinet/tcp.h>
@@ -958,7 +961,7 @@ pf_max_frag_size(struct mbuf *m)
 
 int
 pf_refragment6(struct ifnet *ifp, struct mbuf **m0, struct m_tag *mtag,
-    bool forward)
+    struct ifnet *rt, bool forward)
 {
 	struct mbuf		*m = *m0, *t;
 	struct ip6_hdr		*hdr;
@@ -1029,16 +1032,27 @@ pf_refragment6(struct ifnet *ifp, struct mbuf **m0, struct m_tag *mtag,
 		m->m_flags |= M_SKIP_FIREWALL;
 		memset(&pd, 0, sizeof(pd));
 		pd.pf_mtag = pf_find_mtag(m);
-		if (error == 0)
-			if (forward) {
-				MPASS(m->m_pkthdr.rcvif != NULL);
-				ip6_forward(m, 0);
-			} else {
-				(void)ip6_output(m, NULL, NULL, 0, NULL, NULL,
-				    NULL);
-			}
-		else
+		if (error != 0) {
 			m_freem(m);
+			continue;
+		}
+		if (rt != NULL) {
+			struct sockaddr_in6	dst;
+			hdr = mtod(m, struct ip6_hdr *);
+
+			bzero(&dst, sizeof(dst));
+			dst.sin6_family = AF_INET6;
+			dst.sin6_len = sizeof(dst);
+			dst.sin6_addr = hdr->ip6_dst;
+
+			nd6_output_ifp(rt, rt, m, &dst, NULL);
+		} else if (forward) {
+			MPASS(m->m_pkthdr.rcvif != NULL);
+			ip6_forward(m, 0);
+		} else {
+			(void)ip6_output(m, NULL, NULL, 0, NULL, NULL,
+			    NULL);
+		}
 	}
 
 	return (action);

sys/riscv/include/asm.h

@@ -69,4 +69,9 @@
 	li	tmp, SSTATUS_SUM;					\
 	csrc	sstatus, tmp
 
+#define	SBI_CALL(ext, func)						\
+	li	a7, ext;						\
+	li	a6, func;						\
+	ecall
+
 #endif /* _MACHINE_ASM_H_ */

sys/riscv/include/cpu.h

@@ -35,9 +35,11 @@
 #ifndef _MACHINE_CPU_H_
 #define	_MACHINE_CPU_H_
 
+#ifndef LOCORE
 #include <machine/atomic.h>
 #include <machine/cpufunc.h>
 #include <machine/frame.h>
+#endif
 
 #define	TRAPF_PC(tfp)		((tfp)->tf_sepc)
 #define	TRAPF_USERMODE(tfp)	(((tfp)->tf_sstatus & SSTATUS_SPP) == 0)
@@ -88,6 +90,7 @@
 #define	MMU_SV57	0x4	/* 5-level paging */
 
 #ifdef _KERNEL
+#ifndef LOCORE
 
 extern char btext[];
 extern char etext[];
@@ -105,6 +108,7 @@ get_cyclecount(void)
 	return (rdcycle());
 }
 
-#endif
+#endif /* !LOCORE */
+#endif /* _KERNEL */
 
 #endif /* !_MACHINE_CPU_H_ */

sys/riscv/include/cpufunc.h

@@ -44,6 +44,8 @@ breakpoint(void)
 
 #ifdef _KERNEL
 
+#include <sys/_null.h>
+
 #include <machine/riscvreg.h>
 
 static __inline register_t
@@ -107,16 +109,42 @@ sfence_vma_page(uintptr_t addr)
 #define	rdinstret()			csr_read64(instret)
 #define	rdhpmcounter(n)			csr_read64(hpmcounter##n)
 
+/* Cache hooks. */
+
 extern int64_t dcache_line_size;
-extern int64_t icache_line_size;
 
-#define	cpu_dcache_wbinv_range(a, s)
-#define	cpu_dcache_inv_range(a, s)
-#define	cpu_dcache_wb_range(a, s)
+typedef void (*cache_op_t)(vm_offset_t start, vm_size_t size);
 
-#define	cpu_idcache_wbinv_range(a, s)
-#define	cpu_icache_sync_range(a, s)
-#define	cpu_icache_sync_range_checked(a, s)
+struct riscv_cache_ops {
+	cache_op_t dcache_wbinv_range;
+	cache_op_t dcache_inv_range;
+	cache_op_t dcache_wb_range;
+};
+
+extern struct riscv_cache_ops cache_ops;
+
+static __inline void
+cpu_dcache_wbinv_range(vm_offset_t addr, vm_size_t size)
+{
+	if (cache_ops.dcache_wbinv_range != NULL)
+		cache_ops.dcache_wbinv_range(addr, size);
+}
+
+static __inline void
+cpu_dcache_inv_range(vm_offset_t addr, vm_size_t size)
+{
+	if (cache_ops.dcache_inv_range != NULL)
+		cache_ops.dcache_inv_range(addr, size);
+}
+
+static __inline void
+cpu_dcache_wb_range(vm_offset_t addr, vm_size_t size)
+{
+	if (cache_ops.dcache_wb_range != NULL)
+		cache_ops.dcache_wb_range(addr, size);
+}
+
+void riscv_cache_install_hooks(struct riscv_cache_ops *, u_int);
 
 #define	cpufunc_nullop()		riscv_nullop()
 

sys/riscv/include/pte.h

@@ -102,6 +102,32 @@ typedef	uint64_t	pn_t;			/* page number */
 #define	PTE_MA_NC		(1ul << PTE_MA_SHIFT)
 #define	PTE_MA_IO		(2ul << PTE_MA_SHIFT)
 
+/*
+ * T-HEAD Custom Memory Attribute (MA) bits [63:59].
+ *
+ * bit 59: Trustable (relating to TEE)
+ * bit 60: Shareable (among CPUs, not configurable)
+ * bit 61: Bufferable (writes to device memory)
+ * bit 62: Cacheable
+ * bit 63: Memory Ordering (1 = strongly ordered (device), 0 = default)
+ *
+ * +------+-------+------------------------------------------------------------+
+ * | Mode | Value | Requested Memory Attributes                                |
+ * +------+-------+------------------------------------------------------------+
+ * | NC   | 00110 | Weakly-ordered, non-cacheable, bufferable, shareable,      |
+ * |      |       | non-trustable                                              |
+ * | PMA  | 01110 | Weakly-ordered, cacheable, bufferable, shareable,          |
+ * |      |       | non-trustable                                              |
+ * | IO   | 10010 | Strongly-ordered, non-cacheable, non-bufferable,           |
+ * |      |       | shareable, non-trustable                                   |
+ * +------+-------+------------------------------------------------------------+
+ */
+#define	PTE_THEAD_MA_SHIFT	59
+#define	PTE_THEAD_MA_MASK	(0x1ful << PTE_THEAD_MA_SHIFT)
+#define	PTE_THEAD_MA_NC		(0x6ul  << PTE_THEAD_MA_SHIFT)
+#define	PTE_THEAD_MA_NONE	(0xeul  << PTE_THEAD_MA_SHIFT)
+#define	PTE_THEAD_MA_IO		(0x12ul << PTE_THEAD_MA_SHIFT)
+
 /* Bits 63 - 54 are reserved for future use. */
 #define PTE_HI_MASK	0xFFC0000000000000ULL
 

sys/riscv/include/sbi.h

@@ -123,6 +123,8 @@
 #define	SBI_REMOTE_SFENCE_VMA_ASID	7
 #define	SBI_SHUTDOWN			8
 
+#ifndef LOCORE
+
 #define	SBI_CALL0(e, f)				SBI_CALL5(e, f, 0, 0, 0, 0, 0)
 #define	SBI_CALL1(e, f, p1)			SBI_CALL5(e, f, p1, 0, 0, 0, 0)
 #define	SBI_CALL2(e, f, p1, p2)			SBI_CALL5(e, f, p1, p2, 0, 0, 0)
@@ -242,4 +244,5 @@ sbi_console_getchar(void)
 void sbi_print_version(void);
 void sbi_init(void);
 
+#endif /* !LOCORE */
 #endif /* !_MACHINE_SBI_H_ */

sys/riscv/include/thead.h

@@ -0,0 +1,37 @@
+/*-
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2024 The FreeBSD Foundation
+ *
+ * This software was developed by Mitchell Horne <mhorne@FreeBSD.org> under
+ * sponsorship from the FreeBSD Foundation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+#ifndef _RISCV_THEAD_H_
+#define	_RISCV_THEAD_H_
+
+extern bool has_errata_thead_pbmt;
+
+void thead_setup_cache(void);
+
+#endif /* _RISCV_THEAD_H_ */

sys/riscv/riscv/cache.c

@@ -0,0 +1,53 @@
+/*-
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2024 The FreeBSD Foundation
+ *
+ * This software was developed by Mitchell Horne <mhorne@FreeBSD.org> under
+ * sponsorship from the FreeBSD Foundation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/types.h>
+
+#include <machine/cpufunc.h>
+
+struct riscv_cache_ops __read_frequently cache_ops;
+
+int64_t __read_frequently dcache_line_size;	/* The minimum D cache line size */
+
+static bool cache_initialized;
+
+void
+riscv_cache_install_hooks(struct riscv_cache_ops *newops, u_int line_size)
+{
+	if (cache_initialized)
+		panic("cache hooks already installed!");
+
+	bcopy(newops, &cache_ops, sizeof(cache_ops));
+	dcache_line_size = line_size;
+
+	cache_initialized = true;
+}

sys/riscv/riscv/identcpu.c

@@ -52,6 +52,7 @@
 #include <machine/cpufunc.h>
 #include <machine/elf.h>
 #include <machine/md_var.h>
+#include <machine/thead.h>
 
 #ifdef FDT
 #include <dev/fdt/fdt_common.h>
@@ -463,6 +464,38 @@ identify_cpu_ids(struct cpu_desc *desc)
 	}
 }
 
+static void
+handle_thead_quirks(u_int cpu, struct cpu_desc *desc)
+{
+	if (cpu != 0)
+		return;
+
+	/*
+	 * For now, it is assumed that T-HEAD CPUs have both marchid and mimpid
+	 * values of zero (although we leave this unchecked). It is true in
+	 * practice for the early generations of this hardware (C906, C910,
+	 * C920). In the future, the identity checks may need to become more
+	 * granular, but until then all known T-HEAD quirks are applied
+	 * indiscriminantly.
+	 *
+	 * Note: any changes in this function relating to has_errata_thead_pbmt
+	 * may need to be applied to get_pte_fixup_bits (in locore.S) as well.
+	 */
+
+	has_errata_thead_pbmt = true;
+	thead_setup_cache();
+}
+
+static void
+handle_cpu_quirks(u_int cpu, struct cpu_desc *desc)
+{
+	switch (mvendorid) {
+	case MVENDORID_THEAD:
+		handle_thead_quirks(cpu, desc);
+		break;
+	}
+}
+
 void
 identify_cpu(u_int cpu)
 {
@@ -472,6 +505,7 @@ identify_cpu(u_int cpu)
 	identify_cpu_features(cpu, desc);
 
 	update_global_capabilities(cpu, desc);
+	handle_cpu_quirks(cpu, desc);
 }
 
 void

sys/riscv/riscv/locore.S

@@ -42,9 +42,11 @@
 #include "assym.inc"
 
 #include <machine/asm.h>
+#include <machine/cpu.h>
 #include <machine/param.h>
-#include <machine/riscvreg.h>
 #include <machine/pte.h>
+#include <machine/riscvreg.h>
+#include <machine/sbi.h>
 
 	.globl	kernbase
 	.set	kernbase, KERNBASE
@@ -83,8 +85,12 @@ _alt_start:
 	lla	t0, boot_hart
 	sw	a0, 0(t0)
 
-	/* Load zero as modulep */
-	mv	a0, zero
+	/*
+	 * Stash the DTB pointer in the callee-saved register s4, and zero s3
+	 * to indicate that we have no loader metadata.
+	 */
+	mv	s4, a1
+	mv	s3, zero
 	j	pagetables
 
 /*
@@ -104,10 +110,12 @@ _start:
 .option pop
 
 	/*
-	 * Zero a1 to indicate that we have no DTB pointer. It is already
-	 * included in the loader(8) metadata.
+	 * Stash modulep in the callee-saved register s3, and zero s4 to
+	 * indicate that we have no DTB pointer. It is already included in the
+	 * loader(8) metadata.
 	 */
-	mv	a1, zero
+	mv	s3, a0
+	mv	s4, zero
 
 	/*
 	 * Set up page tables: Our goal is to enable virtual memory, doing the
@@ -128,19 +136,23 @@ _start:
 	 * here and will conditionally enable Sv48 (or higher) later.
 	 *
 	 * We arrive here with:
-	 *  a0 - modulep or zero
-	 *  a1 - zero or dtbp
+	 *  s3 - modulep or zero
+	 *  s4 - zero or dtbp
 	 */
 pagetables:
 	/* Get the kernel's load address (kernstart) in s9 */
 	jal	get_physmem
 
+	/* Get PTE attribute bits in s8 */
+	jal	get_pte_fixup_bits
+
 	/* Construct 1GB Identity Map (1:1 PA->VA) */
 	lla	s1, bootstrap_pt_l1
 
 	srli	s2, s9, L1_SHIFT	/* kernstart >> L1_SHIFT */
 	andi	a5, s2, Ln_ADDR_MASK	/* & Ln_ADDR_MASK */
 	li	t4, (PTE_KERN)
+	or	t4, t4, s8		/* t4 |= pte bits */
 	slli	s2, s2, PTE_PPN2_S	/* (s2 << PTE_PPN2_S) */
 	or	t6, t4, s2
 
@@ -176,6 +188,7 @@ pagetables:
 	li	t2, Ln_ENTRIES		/* Build 512 entries */
 	add	t3, t4, t2
 	li	t0, (PTE_KERN | PTE_X)
+	or	t0, t0, s8		/* t0 |= pte bits */
 1:
 	slli	t2, t4, PTE_PPN1_S	/* << PTE_PPN1_S */
 	or	t5, t0, t2
@@ -244,8 +257,8 @@ va:
 
 	la	t0, initstack
 	sd	t0, RISCV_BOOTPARAMS_KERN_STACK(sp)
-	sd	a1, RISCV_BOOTPARAMS_DTBP_PHYS(sp)
-	sd	a0, RISCV_BOOTPARAMS_MODULEP(sp)
+	sd	s4, RISCV_BOOTPARAMS_DTBP_PHYS(sp)
+	sd	s3, RISCV_BOOTPARAMS_MODULEP(sp)
 
 	mv	a0, sp
 	call	_C_LABEL(initriscv)	/* Off we go */
@@ -267,6 +280,28 @@ get_physmem:
 	sub	s9, t2, t1	/* s9 = physmem base */
 	ret
 
+/*
+ * T-HEAD CPUs implement an alternate scheme for PTE attributes that is
+ * incompatible with the RISC-V PTE specification (see the definitions in
+ * pte.h). Worse, it defines a non-zero value for "main" memory, and this must
+ * be set in order to proceed with our new page tables.
+ *
+ * Therefore, we are forced to check the CPU identity here, which is both
+ * inconvenient and fragile.
+ *
+ * Return the required attribute bits in s8. For sane implementations this is
+ * zero.
+ */
+get_pte_fixup_bits:
+	mv	s8, zero
+	SBI_CALL(SBI_EXT_ID_BASE, SBI_BASE_GET_MVENDORID)
+	li	t0, MVENDORID_THEAD
+	xor	t0, t0, a1
+	bnez	t0, 1f		/* branch if a1 != t0 */
+	li	s8, PTE_THEAD_MA_NONE
+1:
+	ret
+
 	.align  4
 initstack:
 	.space  (PAGE_SIZE * KSTACK_PAGES)

sys/riscv/riscv/machdep.c

@@ -113,10 +113,6 @@ int cold = 1;
 
 struct kva_md_info kmi;
 
-int64_t dcache_line_size;	/* The minimum D cache line size */
-int64_t icache_line_size;	/* The minimum I cache line size */
-int64_t idcache_line_size;	/* The minimum cache line size */
-
 #define BOOT_HART_INVALID	0xffffffff
 uint32_t boot_hart = BOOT_HART_INVALID;	/* The hart we booted on. */
 
@@ -329,17 +325,6 @@ try_load_dtb(caddr_t kmdp)
 }
 #endif
 
-static void
-cache_setup(void)
-{
-
-	/* TODO */
-
-	dcache_line_size = 0;
-	icache_line_size = 0;
-	idcache_line_size = 0;
-}
-
 /*
  * Fake up a boot descriptor table.
  */
@@ -550,8 +535,6 @@ initriscv(struct riscv_bootparams *rvbp)
 	/* Do basic tuning, hz etc */
 	init_param1();
 
-	cache_setup();
-
 #ifdef FDT
 	/*
 	 * XXX: Unconditionally exclude the lowest 2MB of physical memory, as

sys/riscv/riscv/pmap.c

@@ -156,6 +156,7 @@
 #include <machine/md_var.h>
 #include <machine/pcb.h>
 #include <machine/sbi.h>
+#include <machine/thead.h>
 
 /*
  * Boundary values for the page table page index space:
@@ -867,6 +868,11 @@ pmap_bootstrap(vm_paddr_t kernstart, vm_size_t kernlen)
 		memattr_bits[VM_MEMATTR_UNCACHEABLE] = PTE_MA_NC;
 		memattr_bits[VM_MEMATTR_DEVICE] = PTE_MA_IO;
 		memattr_mask = PTE_MA_MASK;
+	} else if (has_errata_thead_pbmt) {
+		memattr_bits[VM_MEMATTR_PMA] = PTE_THEAD_MA_NONE;
+		memattr_bits[VM_MEMATTR_UNCACHEABLE] = PTE_THEAD_MA_NC;
+		memattr_bits[VM_MEMATTR_DEVICE] = PTE_THEAD_MA_IO;
+		memattr_mask = PTE_THEAD_MA_MASK;
 	}
 
 	/* Create a new set of pagetables to run the kernel in. */
@@ -5013,7 +5019,7 @@ pmap_change_attr_locked(vm_offset_t va, vm_size_t size, int mode)
 	if (anychanged) {
 		pmap_invalidate_range(kernel_pmap, base, tmpva);
 		if (mode == VM_MEMATTR_UNCACHEABLE)
-			cpu_dcache_wbinv_range((void *)base, size);
+			cpu_dcache_wbinv_range(base, size);
 	}
 
 	return (error);

sys/riscv/thead/thead.c

@@ -0,0 +1,101 @@
+/*-
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2024 The FreeBSD Foundation
+ *
+ * This software was developed by Mitchell Horne <mhorne@FreeBSD.org> under
+ * sponsorship from the FreeBSD Foundation.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+#include <sys/param.h>
+#include <sys/systm.h>
+
+#include <machine/thead.h>
+
+bool has_errata_thead_pbmt = false;
+
+/* ----------------- dcache ops --------------------- */
+
+
+/* th.dcache.civa: clean & invalidate at VA stored in t0. */
+#define	THEAD_DCACHE_CIVA	".long 0x0272800b\n"
+
+/* th.dcache.iva: invalidate at VA stored in t0. */
+#define	THEAD_DCACHE_IVA	".long 0x0262800b\n"
+
+/* th.dcache.cva: clean at VA stored in t0. */
+#define	THEAD_DCACHE_CVA	".long 0x0252800b\n"
+
+/* th.sync.s: two-way instruction barrier */
+#define	THEAD_SYNC_S		".long 0x0190000b\n"
+
+/* MHTODO: we could parse this information from the device tree. */
+#define	THEAD_DCACHE_SIZE	64
+
+static void
+thead_cpu_dcache_wbinv_range(vm_offset_t va, vm_size_t len)
+{
+	register vm_offset_t t0 __asm("t0") = rounddown(va, dcache_line_size);
+
+	for (; t0 < va + len; t0 += dcache_line_size) {
+		__asm __volatile(THEAD_DCACHE_CIVA
+		                 :: "r" (t0) : "memory");
+	}
+	__asm __volatile(THEAD_SYNC_S ::: "memory");
+}
+
+static void
+thead_cpu_dcache_inv_range(vm_offset_t va, vm_size_t len)
+{
+	register vm_offset_t t0 __asm("t0") = rounddown(va, dcache_line_size);
+
+	for (; t0 < va + len; t0 += dcache_line_size) {
+		__asm __volatile(THEAD_DCACHE_IVA
+				 :: "r" (t0) : "memory");
+	}
+	__asm __volatile(THEAD_SYNC_S ::: "memory");
+}
+
+static void
+thead_cpu_dcache_wb_range(vm_offset_t va, vm_size_t len)
+{
+	register vm_offset_t t0 __asm("t0") = rounddown(va, dcache_line_size);
+
+	for (; t0 < va + len; t0 += dcache_line_size) {
+		__asm __volatile(THEAD_DCACHE_CVA
+				 :: "r" (t0) : "memory");
+	}
+	__asm __volatile(THEAD_SYNC_S ::: "memory");
+}
+
+void
+thead_setup_cache(void)
+{
+	struct riscv_cache_ops thead_ops;
+
+	thead_ops.dcache_wbinv_range = thead_cpu_dcache_wbinv_range;
+	thead_ops.dcache_inv_range = thead_cpu_dcache_inv_range;
+	thead_ops.dcache_wb_range = thead_cpu_dcache_wb_range;
+
+	riscv_cache_install_hooks(&thead_ops, THEAD_DCACHE_SIZE);
+}

sys/sys/param.h

@@ -73,7 +73,7 @@
  * cannot include sys/param.h and should only be updated here.
  */
 #undef __FreeBSD_version
-#define __FreeBSD_version 1500027
+#define __FreeBSD_version 1500028
 
 /*
  * __FreeBSD_kernel__ indicates that this system uses the kernel of FreeBSD,

sys/sys/proc.h

@@ -494,6 +494,7 @@ enum {
 	TDA_RACCT,
 	TDA_MOD1,		/* For third party use, before signals are */
 	TAD_MOD2,		/* processed .. */
+	TDA_PSELECT,		/* For discarding temporary signal mask */
 	TDA_SIG,
 	TDA_KTRACE,
 	TDA_SUSPEND,
@@ -567,6 +568,7 @@ enum {
 #define	TDP2_SBPAGES	0x00000001 /* Owns sbusy on some pages */
 #define	TDP2_COMPAT32RB	0x00000002 /* compat32 ABI for robust lists */
 #define	TDP2_ACCT	0x00000004 /* Doing accounting */
+#define	TDP2_SAN_QUIET	0x00000008 /* Disable warnings from K(A|M)SAN */
 
 /*
  * Reasons that the current thread can not be run yet.

sys/vm/vm_kern.c

@@ -660,7 +660,7 @@ _kmem_unback(vm_object_t object, vm_offset_t addr, vm_size_t size)
 	    m = vm_page_iter_lookup(&pages, atop(offset))) {
 		vm_page_xbusy_claim(m);
 		vm_page_unwire_noq(m);
-		vm_page_iter_free(&pages);
+		vm_page_iter_free(&pages, m);
 	}
 	VM_OBJECT_WUNLOCK(object);
 

sys/vm/vm_object.c

@@ -2060,7 +2060,7 @@ wired:
 		if ((options & OBJPR_NOTMAPPED) == 0 &&
 		    object->ref_count != 0 && !vm_page_try_remove_all(p))
 			goto wired;
-		vm_page_iter_free(&pages);
+		vm_page_iter_free(&pages, p);
 	}
 	vm_object_pip_wakeup(object);
 

sys/vm/vm_page.c

@@ -1713,14 +1713,12 @@ vm_page_free_object_prep(vm_page_t m)
 /*
  *	vm_page_iter_free:
  *
- *	Free the current page, as identified by iterator.
+ *	Free the given page, and use the iterator to remove it from the radix
+ *	tree.
  */
 void
-vm_page_iter_free(struct pctrie_iter *pages)
+vm_page_iter_free(struct pctrie_iter *pages, vm_page_t m)
 {
-	vm_page_t m;
-
-	m = vm_radix_iter_page(pages);
 	vm_radix_iter_remove(pages);
 	vm_page_free_object_prep(m);
 	vm_page_xunbusy(m);

sys/vm/vm_page.h

@@ -602,7 +602,6 @@ bool vm_page_busy_sleep(vm_page_t m, const char *msg, int allocflags);
 void vm_page_busy_sleep_unlocked(vm_object_t obj, vm_page_t m,
     vm_pindex_t pindex, const char *wmesg, int allocflags);
 void vm_page_free(vm_page_t m);
-void vm_page_iter_free(struct pctrie_iter *);
 void vm_page_free_zero(vm_page_t m);
 
 void vm_page_activate (vm_page_t);
@@ -651,11 +650,13 @@ void vm_page_init_marker(vm_page_t marker, int queue, uint16_t aflags);
 void vm_page_init_page(vm_page_t m, vm_paddr_t pa, int segind, int pool);
 int vm_page_insert (vm_page_t, vm_object_t, vm_pindex_t);
 void vm_page_invalid(vm_page_t m);
-void vm_page_launder(vm_page_t m);
-vm_page_t vm_page_lookup(vm_object_t, vm_pindex_t);
+void vm_page_iter_free(struct pctrie_iter *pages, vm_page_t m);
 void vm_page_iter_init(struct pctrie_iter *, vm_object_t);
 void vm_page_iter_limit_init(struct pctrie_iter *, vm_object_t, vm_pindex_t);
 vm_page_t vm_page_iter_lookup(struct pctrie_iter *, vm_pindex_t);
+bool vm_page_iter_remove(struct pctrie_iter *pages);
+void vm_page_launder(vm_page_t m);
+vm_page_t vm_page_lookup(vm_object_t, vm_pindex_t);
 vm_page_t vm_page_lookup_unlocked(vm_object_t, vm_pindex_t);
 vm_page_t vm_page_next(vm_page_t m);
 void vm_page_pqbatch_drain(void);
@@ -680,7 +681,6 @@ void vm_page_release(vm_page_t m, int flags);
 void vm_page_release_locked(vm_page_t m, int flags);
 vm_page_t vm_page_relookup(vm_object_t, vm_pindex_t);
 bool vm_page_remove(vm_page_t);
-bool vm_page_iter_remove(struct pctrie_iter *);
 bool vm_page_remove_xbusy(vm_page_t);
 int vm_page_rename(struct pctrie_iter *, vm_object_t, vm_pindex_t);
 void vm_page_replace(vm_page_t mnew, vm_object_t object,

sys/x86/iommu/amd_drv.c

@@ -277,7 +277,7 @@ amdiommu_create_dev_tbl(struct amdiommu_unit *sc)
 			pmap_qenter(seg_vaddr, &m, 1);
 		}
 		reg = i == 0 ? AMDIOMMU_DEVTAB_BASE : AMDIOMMU_DEVTAB_S1_BASE +
-		    i - 1;
+		    ((i - 1) << 3);
 		amdiommu_write8(sc, reg, rval);		
 	}
 

tests/sys/netpfil/pf/fragmentation_pass.sh

@@ -155,6 +155,75 @@ v6_cleanup()
 	pft_cleanup
 }
 
+atf_test_case "v6_route_to" "cleanup"
+v6_route_to_head()
+{
+	atf_set descr 'Test IPv6 reassembly combined with route-to'
+	atf_set require.user root
+}
+
+v6_route_to_body()
+{
+	pft_init
+}
+
+v6_route_to_cleanup()
+{
+	pft_cleanup
+
+	epair_send=$(vnet_mkepair)
+	epair_link=$(vnet_mkepair)
+
+	vnet_mkjail alcatraz ${epair_send}b ${epair_link}a
+	vnet_mkjail singsing ${epair_link}b
+
+	ifconfig ${epair_send}a inet6 2001:db8:42::1/64 no_dad up
+
+	jexec alcatraz ifconfig ${epair_send}b inet6 2001:db8:42::2/64 no_dad up
+	jexec alcatraz ifconfig ${epair_link}a inet6 2001:db8:43::2/64 no_dad up
+	jexec alcatraz sysctl net.inet6.ip6.forwarding=1
+
+	jexec singsing ifconfig ${epair_link}b inet6 2001:db8:43::3/64 no_dad up
+	jexec singsing route add -6 2001:db8:42::/64 2001:db8:43::2
+	route add -6 2001:db8:43::/64 2001:db8:42::2
+
+	jexec alcatraz ifconfig ${epair_send}b inet6 -ifdisabled
+	jexec alcatraz ifconfig ${epair_link}a inet6 -ifdisabled
+	jexec singsing ifconfig ${epair_link}b inet6 -ifdisabled
+	ifconfig ${epair_send}a inet6 -ifdisabled
+
+	jexec alcatraz pfctl -e
+	pft_set_rules alcatraz \
+		"set reassemble yes" \
+		"pass" \
+		"pass in route-to (${epair_link}a 2001:db8:43::3) inet6 proto icmp6 from any to 2001:db8:43::3 keep state"
+
+	# Forwarding test
+	atf_check -s exit:0 -o ignore \
+		ping -6 -c 1 2001:db8:43::3
+
+	atf_check -s exit:0 -o ignore \
+		ping -6 -c 1 -s 4500 2001:db8:43::3
+
+	atf_check -s exit:0 -o ignore\
+		ping -6 -c 1 -b 70000 -s 65000 2001:db8:43::3
+
+	# Now test this without fragmentation
+	pft_set_rules alcatraz \
+		"set reassemble no" \
+		"pass" \
+		"pass in route-to (${epair_link}a 2001:db8:43::3) inet6 proto icmp6 from any to 2001:db8:43::3 keep state"
+
+	atf_check -s exit:0 -o ignore \
+		ping -6 -c 1 2001:db8:43::3
+
+	atf_check -s exit:0 -o ignore \
+		ping -6 -c 1 -s 4500 2001:db8:43::3
+
+	atf_check -s exit:0 -o ignore\
+		ping -6 -c 1 -b 70000 -s 65000 2001:db8:43::3
+}
+
 atf_test_case "mtu_diff" "cleanup"
 mtu_diff_head()
 {
@@ -544,6 +613,7 @@ atf_init_test_cases()
 {
 	atf_add_test_case "too_many_fragments"
 	atf_add_test_case "v6"
+	atf_add_test_case "v6_route_to"
 	atf_add_test_case "mtu_diff"
 	atf_add_test_case "overreplace"
 	atf_add_test_case "overindex"

usr.bin/netstat/route.c

@@ -578,7 +578,7 @@ netname4(in_addr_t in, in_addr_t mask)
 	struct netent *np = 0;
 	in_addr_t i;
 
-	if (in == INADDR_ANY && mask == 0) {
+	if (!numeric_addr && in == INADDR_ANY && mask == 0) {
 		strlcpy(line, "default", sizeof(line));
 		return (line);
 	}
@@ -673,7 +673,8 @@ netname6(struct sockaddr_in6 *sa6, struct sockaddr_in6 *mask)
 	else
 		masklen = 128;
 
-	if (masklen == 0 && IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr))
+	if (!numeric_addr && masklen == 0 &&
+	    IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr))
 		return("default");
 
 	getnameinfo((struct sockaddr *)sa6, sa6->sin6_len, nline, sizeof(nline),

usr.bin/truss/setup.c

@@ -160,7 +160,7 @@ setup_and_wait(struct trussinfo *info, char *command[])
 
 	/* Only in the parent here */
 	if (waitpid(pid, NULL, 0) < 0)
-		err(1, "unexpect stop in waitpid");
+		err(1, "unexpected stop in waitpid");
 
 	new_proc(info, pid, 0);
 }
@@ -179,10 +179,10 @@ start_tracing(struct trussinfo *info, pid_t pid)
 		usleep(200);
 	} while (ret && retry-- > 0);
 	if (ret)
-		err(1, "can not attach to target process");
+		err(1, "Cannot attach to target process");
 
 	if (waitpid(pid, NULL, 0) < 0)
-		err(1, "Unexpect stop in waitpid");
+		err(1, "Unexpected stop in waitpid");
 
 	new_proc(info, pid, 0);
 }