24 files changed, 120 insertions, 106 deletions

diff --git a/sys/amd64/amd64/machdep.c b/sys/amd64/amd64/machdep.c
index 3175616d0ce05..0c16da1327dc7 100644
--- a/sys/amd64/amd64/machdep.c
+++ b/sys/amd64/amd64/machdep.c
@@ -604,7 +604,6 @@ exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 	regs->tf_fs = _ufssel;
 	regs->tf_gs = _ugssel;
 	regs->tf_flags = TF_HASSEGS;
-	td->td_retval[1] = 0;
 
 	/*
 	 * Reset the hardware debug registers if they were in use.
diff --git a/sys/amd64/ia32/ia32_signal.c b/sys/amd64/ia32/ia32_signal.c
index 79131f3f561ce..b1fc8debc1f95 100644
--- a/sys/amd64/ia32/ia32_signal.c
+++ b/sys/amd64/ia32/ia32_signal.c
@@ -967,5 +967,4 @@ ia32_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 
 	/* Return via doreti so that we can change to a different %cs */
 	set_pcb_flags(pcb, PCB_32BIT | PCB_FULL_IRET);
-	td->td_retval[1] = 0;
 }
diff --git a/sys/amd64/linux32/linux32_sysvec.c b/sys/amd64/linux32/linux32_sysvec.c
index 8b0a35ea0bfae..1a0dc1c954b7a 100644
--- a/sys/amd64/linux32/linux32_sysvec.c
+++ b/sys/amd64/linux32/linux32_sysvec.c
@@ -832,7 +832,6 @@ exec_linux_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 
 	/* Do full restore on return so that we can change to a different %cs */
 	set_pcb_flags(pcb, PCB_32BIT | PCB_FULL_IRET);
-	td->td_retval[1] = 0;
 }
 
 /*
diff --git a/sys/amd64/vmm/amd/amdvi_hw.c b/sys/amd64/vmm/amd/amdvi_hw.c
index 22d4a12e32103..3478262a73468 100644
--- a/sys/amd64/vmm/amd/amdvi_hw.c
+++ b/sys/amd64/vmm/amd/amdvi_hw.c
@@ -582,7 +582,7 @@ amdvi_decode_evt_flag(uint16_t flag)
 {
 
 	flag &= AMDVI_EVENT_FLAG_MASK;
-	printf("0x%b]\n", flag,
+	printf(" 0x%b]\n", flag,
 		"\020"
 		"\001GN"
 		"\002NX"
@@ -692,7 +692,7 @@ amdvi_decode_evt(struct amdvi_event *evt)
 	case AMDVI_EVENT_ILLEGAL_CMD:
 		/* FALL THROUGH */
 	case AMDVI_EVENT_CMD_HW_ERROR:
-		printf("\t[%s EVT]", (evt->opcode == AMDVI_EVENT_ILLEGAL_CMD) ?
+		printf("\t[%s EVT]\n", (evt->opcode == AMDVI_EVENT_ILLEGAL_CMD) ?
 		    "ILLEGAL CMD" : "CMD HW ERR");
 		cmd = (struct amdvi_cmd *)PHYS_TO_DMAP(evt->addr);
 		printf("\tCMD opcode= 0x%x 0x%x 0x%x 0x%lx\n",
@@ -700,13 +700,14 @@ amdvi_decode_evt(struct amdvi_event *evt)
 		break;
 
 	case AMDVI_EVENT_IOTLB_TIMEOUT:
-		printf("\t[IOTLB_INV_TIMEOUT devid:0x%x addr:0x%lx",
+		printf("\t[IOTLB_INV_TIMEOUT devid:0x%x addr:0x%lx]\n",
 		    evt->devid, evt->addr);
 		break;
 
 	case AMDVI_EVENT_INVALID_DTE_REQ:
-		printf("\t[INV_DTE devid:0x%x addr:0x%lx",
-		    evt->devid, evt->addr);
+		printf("\t[INV_DTE devid:0x%x addr:0x%lx type:0x%x tr:%d]\n",
+		    evt->devid, evt->addr, evt->flag >> 9,
+		    (evt->flag >> 8) & 1);
 		break;
 
 	case AMDVI_EVENT_INVALID_PPR_REQ:
@@ -715,7 +716,7 @@ amdvi_decode_evt(struct amdvi_event *evt)
 		break;
 
 	default:
-		printf("Unsupported AMD-Vi event:%d", evt->opcode);
+		printf("Unsupported AMD-Vi event:%d\n", evt->opcode);
 	}
 }
 
@@ -988,15 +989,12 @@ amdvi_add_sysctl(struct amdvi_softc *softc)
 	    &softc->event_intr_cnt, "Event interrupt count");
 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "command_count", CTLFLAG_RD,
 	    &softc->total_cmd, "Command submitted count");
-	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "pci_rid", CTLFLAG_RD,
-	    (int *)&softc->pci_rid, 0,
-	    "IOMMU RID");
-	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "start_dev_rid", CTLFLAG_RD,
-	    (int *)&softc->start_dev_rid, 0,
-	    "Start of device under this IOMMU");
-	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "end_dev_rid", CTLFLAG_RD,
-	    (int *)&softc->end_dev_rid, 0,
-	    "End of device under this IOMMU");
+	SYSCTL_ADD_U16(ctx, child, OID_AUTO, "pci_rid", CTLFLAG_RD,
+	    &softc->pci_rid, 0, "IOMMU RID");
+	SYSCTL_ADD_U16(ctx, child, OID_AUTO, "start_dev_rid", CTLFLAG_RD,
+	    &softc->start_dev_rid, 0, "Start of device under this IOMMU");
+	SYSCTL_ADD_U16(ctx, child, OID_AUTO, "end_dev_rid", CTLFLAG_RD,
+	    &softc->end_dev_rid, 0, "End of device under this IOMMU");
 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "command_head",
 	    CTLTYPE_UINT | CTLFLAG_RD, softc, 0,
 	    amdvi_handle_sysctl, "IU", "Command head");
diff --git a/sys/amd64/vmm/amd/amdvi_priv.h b/sys/amd64/vmm/amd/amdvi_priv.h
index c292d3ef9881c..8d4553fe1a3c7 100755
--- a/sys/amd64/vmm/amd/amdvi_priv.h
+++ b/sys/amd64/vmm/amd/amdvi_priv.h
@@ -230,8 +230,8 @@ struct amdvi_ctrl {
 	uint64_t :45;
 	uint32_t evt_tail:19;
 	uint64_t :45;
-	uint64_t :56;
-	uint8_t	 status:8;
+	uint32_t status:19;
+	uint64_t :45;
 	uint64_t pad2;
 	uint8_t  :4;
 	uint16_t ppr_head:15;
diff --git a/sys/amd64/vmm/amd/ivrs_drv.c b/sys/amd64/vmm/amd/ivrs_drv.c
index dc47717006a05..bc30d83c6ca5a 100755
--- a/sys/amd64/vmm/amd/ivrs_drv.c
+++ b/sys/amd64/vmm/amd/ivrs_drv.c
@@ -88,7 +88,7 @@ ivrs_hdr_iterate_tbl(ivhd_iter_t iter, void *arg)
 			if (!iter(ivrs_hdr, arg))
 				return;
 			break;
-		
+
 		case ACPI_IVRS_TYPE_MEMORY1:
 		case ACPI_IVRS_TYPE_MEMORY2:
 		case ACPI_IVRS_TYPE_MEMORY3:
@@ -96,7 +96,7 @@ ivrs_hdr_iterate_tbl(ivhd_iter_t iter, void *arg)
 				return;
 
 			break;
-		
+
 		default:
 			printf("AMD-Vi:Not IVHD/IVMD type(%d)", ivrs_hdr->Type);
 
@@ -186,7 +186,8 @@ ivhd_dev_add_entry(struct amdvi_softc *softc, uint32_t start_id,
 static int
 ivhd_dev_parse(ACPI_IVRS_HARDWARE * ivhd, struct amdvi_softc *softc)
 {
-	ACPI_IVRS_DE_HEADER *de, *end;
+	ACPI_IVRS_DE_HEADER *de;
+	uint8_t *p, *end;
 	int range_start_id = 0, range_end_id = 0;
 	uint32_t *extended;
 	uint8_t all_data = 0, range_data = 0;
@@ -195,12 +196,15 @@ ivhd_dev_parse(ACPI_IVRS_HARDWARE * ivhd, struct amdvi_softc *softc)
 	softc->start_dev_rid = ~0;
 	softc->end_dev_rid = 0;
 
-	de = (ACPI_IVRS_DE_HEADER *) ((uint8_t *)ivhd +
-	    sizeof(ACPI_IVRS_HARDWARE));
-	end = (ACPI_IVRS_DE_HEADER *) ((uint8_t *)ivhd +
-	    ivhd->Header.Length);
+	/*
+	 * XXX The following actually depends on Header.Type and
+	 * is only true for 0x10.
+	 */
+	p = (uint8_t *)ivhd + sizeof(ACPI_IVRS_HARDWARE);
+	end = (uint8_t *)ivhd + ivhd->Header.Length;
 
-	while (de < (ACPI_IVRS_DE_HEADER *) end) {
+	while (p < end) {
+		de = (ACPI_IVRS_DE_HEADER *)p;
 		softc->start_dev_rid = MIN(softc->start_dev_rid, de->Id);
 		softc->end_dev_rid = MAX(softc->end_dev_rid, de->Id);
 		switch (de->Type) {
@@ -263,7 +267,15 @@ ivhd_dev_parse(ACPI_IVRS_HARDWARE * ivhd, struct amdvi_softc *softc)
 			    "WARN Too many device entries.\n");
 			return (EINVAL);
 		}
-		de++;
+		if (de->Type < 0x40)
+			p += sizeof(ACPI_IVRS_DEVICE4);
+		else if (de->Type < 0x80)
+			p += sizeof(ACPI_IVRS_DEVICE8A);
+		else {
+			printf("Variable size IVHD type 0x%x not supported\n",
+			    de->Type);
+			break;
+		}
 	}
 
 	KASSERT((softc->end_dev_rid >= softc->start_dev_rid),
diff --git a/sys/arm/allwinner/clkng/aw_ccung.c b/sys/arm/allwinner/clkng/aw_ccung.c
index 416a3c08bcf95..41ee55b4baf2e 100644
--- a/sys/arm/allwinner/clkng/aw_ccung.c
+++ b/sys/arm/allwinner/clkng/aw_ccung.c
@@ -76,6 +76,10 @@ __FBSDID("$FreeBSD$");
 #include <arm/allwinner/clkng/ccu_sun8i_r.h>
 #endif
 
+#if defined(SOC_ALLWINNER_A83T)
+#include <arm/allwinner/clkng/ccu_a83t.h>
+#endif
+
 #include "clkdev_if.h"
 #include "hwreset_if.h"
 
@@ -102,6 +106,10 @@ static struct resource_spec aw_ccung_spec[] = {
 #define	A13_CCU		6
 #endif
 
+#if defined(SOC_ALLWINNER_A83T)
+#define	A83T_CCU	7
+#endif
+
 static struct ofw_compat_data compat_data[] = {
 #if defined(SOC_ALLWINNER_A31)
 	{ "allwinner,sun5i-a13-ccu", A13_CCU},
@@ -117,6 +125,9 @@ static struct ofw_compat_data compat_data[] = {
 	{ "allwinner,sun50i-a64-ccu", A64_CCU },
 	{ "allwinner,sun50i-a64-r-ccu", A64_R_CCU },
 #endif
+#if defined(SOC_ALLWINNER_A83T)
+	{ "allwinner,sun8i-a83t-ccu", A83T_CCU },
+#endif
 	{NULL, 0 }
 };
 
@@ -359,6 +370,11 @@ aw_ccung_attach(device_t dev)
 		ccu_sun8i_r_register_clocks(sc);
 		break;
 #endif
+#if defined(SOC_ALLWINNER_A83T)
+	case A83T_CCU:
+		ccu_a83t_register_clocks(sc);
+		break;
+#endif
 	}
 
 	if (sc->gates)
diff --git a/sys/arm/arm/machdep.c b/sys/arm/arm/machdep.c
index 689f7686d4c9b..04fd482170e74 100644
--- a/sys/arm/arm/machdep.c
+++ b/sys/arm/arm/machdep.c
@@ -518,6 +518,16 @@ set_mcontext(struct thread *td, mcontext_t *mcp)
 	mcontext_vfp_t mc_vfp, *vfp;
 	struct trapframe *tf = td->td_frame;
 	const __greg_t *gr = mcp->__gregs;
+	int spsr;
+
+	/*
+	 * Make sure the processor mode has not been tampered with and
+	 * interrupts have not been disabled.
+	 */
+	spsr = gr[_REG_CPSR];
+	if ((spsr & PSR_MODE) != PSR_USR32_MODE ||
+	    (spsr & (PSR_I | PSR_F)) != 0)
+		return (EINVAL);
 
 #ifdef WITNESS
 	if (mcp->mc_vfp_size != 0 && mcp->mc_vfp_size != sizeof(mc_vfp)) {
@@ -677,22 +687,16 @@ sys_sigreturn(td, uap)
 	} */ *uap;
 {
 	ucontext_t uc;
-	int spsr;
+	int error;
 
 	if (uap == NULL)
 		return (EFAULT);
 	if (copyin(uap->sigcntxp, &uc, sizeof(uc)))
 		return (EFAULT);
-	/*
-	 * Make sure the processor mode has not been tampered with and
-	 * interrupts have not been disabled.
-	 */
-	spsr = uc.uc_mcontext.__gregs[_REG_CPSR];
-	if ((spsr & PSR_MODE) != PSR_USR32_MODE ||
-	    (spsr & (PSR_I | PSR_F)) != 0)
-		return (EINVAL);
 	/* Restore register context. */
-	set_mcontext(td, &uc.uc_mcontext);
+	error = set_mcontext(td, &uc.uc_mcontext);
+	if (error != 0)
+		return (error);
 
 	/* Restore signal mask. */
 	kern_sigprocmask(td, SIG_SETMASK, &uc.uc_sigmask, NULL, 0);
diff --git a/sys/arm/cloudabi32/cloudabi32_sysvec.c b/sys/arm/cloudabi32/cloudabi32_sysvec.c
index a0bebcc996414..eeeb93c188c79 100644
--- a/sys/arm/cloudabi32/cloudabi32_sysvec.c
+++ b/sys/arm/cloudabi32/cloudabi32_sysvec.c
@@ -61,7 +61,7 @@ cloudabi32_proc_setregs(struct thread *td, struct image_params *imgp,
 	 * tpidrurw to the TCB.
 	 */
 	regs = td->td_frame;
-	regs->tf_r0 = td->td_retval[0] =
+	regs->tf_r0 =
 	    stack + roundup(sizeof(cloudabi32_tcb_t), sizeof(register_t));
 	(void)cpu_set_user_tls(td, (void *)stack);
 }
diff --git a/sys/arm64/arm64/machdep.c b/sys/arm64/arm64/machdep.c
index 886792a9681db..452f224797af4 100644
--- a/sys/arm64/arm64/machdep.c
+++ b/sys/arm64/arm64/machdep.c
@@ -211,7 +211,8 @@ set_regs(struct thread *td, struct reg *regs)
 	frame->tf_sp = regs->sp;
 	frame->tf_lr = regs->lr;
 	frame->tf_elr = regs->elr;
-	frame->tf_spsr = regs->spsr;
+	frame->tf_spsr &= ~PSR_FLAGS;
+	frame->tf_spsr |= regs->spsr & PSR_FLAGS;
 
 	memcpy(frame->tf_x, regs->x, sizeof(frame->tf_x));
 
@@ -310,12 +311,7 @@ exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 
 	memset(tf, 0, sizeof(struct trapframe));
 
-	/*
-	 * We need to set x0 for init as it doesn't call
-	 * cpu_set_syscall_retval to copy the value. We also
-	 * need to set td_retval for the cases where we do.
-	 */
-	tf->tf_x[0] = td->td_retval[0] = stack;
+	tf->tf_x[0] = stack;
 	tf->tf_sp = STACKALIGN(stack);
 	tf->tf_lr = imgp->entry_addr;
 	tf->tf_elr = imgp->entry_addr;
@@ -354,6 +350,12 @@ int
 set_mcontext(struct thread *td, mcontext_t *mcp)
 {
 	struct trapframe *tf = td->td_frame;
+	uint32_t spsr;
+
+	spsr = mcp->mc_gpregs.gp_spsr;
+	if ((spsr & PSR_M_MASK) != PSR_M_EL0t ||
+	    (spsr & (PSR_F | PSR_I | PSR_A | PSR_D)) != 0)
+		return (EINVAL); 
 
 	memcpy(tf->tf_x, mcp->mc_gpregs.gp_x, sizeof(tf->tf_x));
 
@@ -530,19 +532,16 @@ int
 sys_sigreturn(struct thread *td, struct sigreturn_args *uap)
 {
 	ucontext_t uc;
-	uint32_t spsr;
+	int error;
 
 	if (uap == NULL)
 		return (EFAULT);
 	if (copyin(uap->sigcntxp, &uc, sizeof(uc)))
 		return (EFAULT);
 
-	spsr = uc.uc_mcontext.mc_gpregs.gp_spsr;
-	if ((spsr & PSR_M_MASK) != PSR_M_EL0t ||
-	    (spsr & (PSR_F | PSR_I | PSR_A | PSR_D)) != 0)
-		return (EINVAL); 
-
-	set_mcontext(td, &uc.uc_mcontext);
+	error = set_mcontext(td, &uc.uc_mcontext);
+	if (error != 0)
+		return (error);
 	set_fpcontext(td, &uc.uc_mcontext);
 
 	/* Restore signal mask. */
diff --git a/sys/arm64/cloudabi64/cloudabi64_sysvec.c b/sys/arm64/cloudabi64/cloudabi64_sysvec.c
index f6e75f563668b..a98ff0e2c305f 100644
--- a/sys/arm64/cloudabi64/cloudabi64_sysvec.c
+++ b/sys/arm64/cloudabi64/cloudabi64_sysvec.c
@@ -61,7 +61,7 @@ cloudabi64_proc_setregs(struct thread *td, struct image_params *imgp,
 	 * tpidr_el0 to the TCB.
 	 */
 	regs = td->td_frame;
-	regs->tf_x[0] = td->td_retval[0] =
+	regs->tf_x[0] =
 	    stack + roundup(sizeof(cloudabi64_tcb_t), sizeof(register_t));
 	(void)cpu_set_user_tls(td, (void *)stack);
 }
diff --git a/sys/arm64/include/armreg.h b/sys/arm64/include/armreg.h
index a7a15fc87850c..cf24fadf16239 100644
--- a/sys/arm64/include/armreg.h
+++ b/sys/arm64/include/armreg.h
@@ -572,6 +572,7 @@
 #define	PSR_C		0x20000000
 #define	PSR_Z		0x40000000
 #define	PSR_N		0x80000000
+#define	PSR_FLAGS	0xf0000000
 
 /* TCR_EL1 - Translation Control Register */
 #define	TCR_ASID_16	(1 << 36)
diff --git a/sys/cddl/contrib/opensolaris/uts/common/dtrace/fasttrap.c b/sys/cddl/contrib/opensolaris/uts/common/dtrace/fasttrap.c
index 82353cbb235df..9e6bde2eac482 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/dtrace/fasttrap.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/dtrace/fasttrap.c
@@ -601,8 +601,8 @@ fasttrap_fork(proc_t *p, proc_t *cp)
 	pid_t ppid = p->p_pid;
 	int i;
 
-#ifdef illumos
 	ASSERT(curproc == p);
+#ifdef illumos
 	ASSERT(p->p_proc_flag & P_PR_LOCK);
 #else
 	PROC_LOCK_ASSERT(p, MA_OWNED);
@@ -610,26 +610,15 @@ fasttrap_fork(proc_t *p, proc_t *cp)
 #ifdef illumos
 	ASSERT(p->p_dtrace_count > 0);
 #else
-	if (p->p_dtrace_helpers) {
-		/*
-		 * dtrace_helpers_duplicate() allocates memory.
-		 */
-		_PHOLD(cp);
-		PROC_UNLOCK(p);
-		PROC_UNLOCK(cp);
-		dtrace_helpers_duplicate(p, cp);
-		PROC_LOCK(cp);
-		PROC_LOCK(p);
-		_PRELE(cp);
-	}
 	/*
 	 * This check is purposely here instead of in kern_fork.c because,
 	 * for legal resons, we cannot include the dtrace_cddl.h header
 	 * inside kern_fork.c and insert if-clause there.
 	 */
-	if (p->p_dtrace_count == 0)
+	if (p->p_dtrace_count == 0 && p->p_dtrace_helpers == NULL)
 		return;
 #endif
+
 	ASSERT(cp->p_dtrace_count == 0);
 
 	/*
@@ -658,6 +647,8 @@ fasttrap_fork(proc_t *p, proc_t *cp)
 	_PHOLD(cp);
 	PROC_UNLOCK(cp);
 	PROC_UNLOCK(p);
+	if (p->p_dtrace_count == 0)
+		goto dup_helpers;
 #endif
 
 	/*
@@ -711,6 +702,9 @@ fasttrap_fork(proc_t *p, proc_t *cp)
 	mutex_enter(&cp->p_lock);
 	sprunlock(cp);
 #else
+dup_helpers:
+	if (p->p_dtrace_helpers != NULL)
+		dtrace_helpers_duplicate(p, cp);
 	PROC_LOCK(p);
 	PROC_LOCK(cp);
 	_PRELE(cp);
diff --git a/sys/compat/linux/linux_emul.c b/sys/compat/linux/linux_emul.c
index b244eea3b677d..8be4fa2df0c93 100644
--- a/sys/compat/linux/linux_emul.c
+++ b/sys/compat/linux/linux_emul.c
@@ -186,7 +186,7 @@ linux_common_execve(struct thread *td, struct image_args *eargs)
 
 	error = kern_execve(td, eargs, NULL);
 	post_execve(td, error, oldvmspace);
-	if (error != 0)
+	if (error != EJUSTRETURN)
 		return (error);
 
 	/*
@@ -213,7 +213,7 @@ linux_common_execve(struct thread *td, struct image_args *eargs)
 		free(em, M_TEMP);
 		free(pem, M_LINUX);
 	}
-	return (0);
+	return (EJUSTRETURN);
 }
 
 void 
diff --git a/sys/geom/mirror/g_mirror.c b/sys/geom/mirror/g_mirror.c
index 7f8f670d7461a..4a5f542c1de52 100644
--- a/sys/geom/mirror/g_mirror.c
+++ b/sys/geom/mirror/g_mirror.c
@@ -54,8 +54,8 @@ static MALLOC_DEFINE(M_MIRROR, "mirror_data", "GEOM_MIRROR Data");
 SYSCTL_DECL(_kern_geom);
 static SYSCTL_NODE(_kern_geom, OID_AUTO, mirror, CTLFLAG_RW, 0,
     "GEOM_MIRROR stuff");
-u_int g_mirror_debug = 0;
-SYSCTL_UINT(_kern_geom_mirror, OID_AUTO, debug, CTLFLAG_RWTUN, &g_mirror_debug, 0,
+int g_mirror_debug = 0;
+SYSCTL_INT(_kern_geom_mirror, OID_AUTO, debug, CTLFLAG_RWTUN, &g_mirror_debug, 0,
     "Debug level");
 static u_int g_mirror_timeout = 4;
 SYSCTL_UINT(_kern_geom_mirror, OID_AUTO, timeout, CTLFLAG_RWTUN, &g_mirror_timeout,
diff --git a/sys/geom/mirror/g_mirror.h b/sys/geom/mirror/g_mirror.h
index f3a4a00700936..3b9664035c984 100644
--- a/sys/geom/mirror/g_mirror.h
+++ b/sys/geom/mirror/g_mirror.h
@@ -72,7 +72,7 @@
 					 G_MIRROR_DEVICE_FLAG_NOFAILSYNC)
 
 #ifdef _KERNEL
-extern u_int g_mirror_debug;
+extern int g_mirror_debug;
 
 #define	G_MIRROR_DEBUG(lvl, ...)	do {				\
 	if (g_mirror_debug >= (lvl)) {					\
diff --git a/sys/i386/i386/machdep.c b/sys/i386/i386/machdep.c
index 68a4522085826..dac5fd6967ec4 100644
--- a/sys/i386/i386/machdep.c
+++ b/sys/i386/i386/machdep.c
@@ -1126,6 +1126,7 @@ exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 	set_fsbase(td, 0);
 	set_gsbase(td, 0);
 
+	/* Make sure edx is 0x0 on entry. Linux binaries depend on it. */
 	bzero((char *)regs, sizeof(struct trapframe));
 	regs->tf_eip = imgp->entry_addr;
 	regs->tf_esp = stack;
@@ -1168,13 +1169,6 @@ exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 	 * clean FP state if it uses the FPU again.
 	 */
 	fpstate_drop(td);
-
-	/*
-	 * XXX - Linux emulator
-	 * Make sure sure edx is 0x0 on entry. Linux binaries depend
-	 * on it.
-	 */
-	td->td_retval[1] = 0;
 }
 
 void
diff --git a/sys/i386/include/atomic.h b/sys/i386/include/atomic.h
index c167cf5c0fae1..7cd5b0d2a7e46 100644
--- a/sys/i386/include/atomic.h
+++ b/sys/i386/include/atomic.h
@@ -129,6 +129,7 @@ int		atomic_cmpset_64(volatile uint64_t *, uint64_t, uint64_t);
 uint64_t	atomic_load_acq_64(volatile uint64_t *);
 void		atomic_store_rel_64(volatile uint64_t *, uint64_t);
 uint64_t	atomic_swap_64(volatile uint64_t *, uint64_t);
+uint64_t	atomic_fetchadd_64(volatile uint64_t *, uint64_t);
 
 #else /* !KLD_MODULE && __GNUCLIKE_ASM */
 
@@ -565,6 +566,17 @@ atomic_swap_64(volatile uint64_t *p, uint64_t v)
 		return (atomic_swap_64_i586(p, v));
 }
 
+static __inline uint64_t
+atomic_fetchadd_64(volatile uint64_t *p, uint64_t v)
+{
+
+	for (;;) {
+		uint64_t t = *p;
+		if (atomic_cmpset_64(p, t, t + v))
+			return (t);
+	}
+}
+
 #endif /* _KERNEL */
 
 #endif /* KLD_MODULE || !__GNUCLIKE_ASM */
diff --git a/sys/kern/init_main.c b/sys/kern/init_main.c
index 645d120d9807d..dbd7a4ac07c76 100644
--- a/sys/kern/init_main.c
+++ b/sys/kern/init_main.c
@@ -797,7 +797,7 @@ start_init(void *dummy)
 		 * Otherwise, return via fork_trampoline() all the way
 		 * to user mode as init!
 		 */
-		if ((error = sys_execve(td, &args)) == 0) {
+		if ((error = sys_execve(td, &args)) == EJUSTRETURN) {
 			mtx_unlock(&Giant);
 			return;
 		}
diff --git a/sys/kern/kern_exec.c b/sys/kern/kern_exec.c
index f844a61967c64..284e08a801afd 100644
--- a/sys/kern/kern_exec.c
+++ b/sys/kern/kern_exec.c
@@ -318,7 +318,7 @@ post_execve(struct thread *td, int error, struct vmspace *oldvmspace)
 		 * If success, we upgrade to SINGLE_EXIT state to
 		 * force other threads to suicide.
 		 */
-		if (error == 0)
+		if (error == EJUSTRETURN)
 			thread_single(p, SINGLE_EXIT);
 		else
 			thread_single_end(p, SINGLE_BOUNDARY);
@@ -962,7 +962,13 @@ exec_fail:
 		ktrprocctor(p);
 #endif
 
-	return (error);
+	/*
+	 * We don't want cpu_set_syscall_retval() to overwrite any of
+	 * the register values put in place by exec_setregs().
+	 * Implementations of cpu_set_syscall_retval() will leave
+	 * registers unmodified when returning EJUSTRETURN.
+	 */
+	return (error == 0 ? EJUSTRETURN : error);
 }
 
 int
diff --git a/sys/netinet/sctp_ss_functions.c b/sys/netinet/sctp_ss_functions.c
index 22d3deb58fd93..3099c4893c478 100644
--- a/sys/netinet/sctp_ss_functions.c
+++ b/sys/netinet/sctp_ss_functions.c
@@ -1,4 +1,6 @@
 /*-
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
  * Copyright (c) 2010-2012, by Michael Tuexen. All rights reserved.
  * Copyright (c) 2010-2012, by Randall Stewart. All rights reserved.
  * Copyright (c) 2010-2012, by Robin Seggelmann. All rights reserved.
diff --git a/sys/powerpc/powerpc/exec_machdep.c b/sys/powerpc/powerpc/exec_machdep.c
index 596ae8079f19b..724ae849a0699 100644
--- a/sys/powerpc/powerpc/exec_machdep.c
+++ b/sys/powerpc/powerpc/exec_machdep.c
@@ -520,22 +520,11 @@ exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 	 *	- ps_strings is a NetBSD extention, and will be
 	 * 	  ignored by executables which are strictly
 	 *	  compliant with the SVR4 ABI.
-	 *
-	 * XXX We have to set both regs and retval here due to different
-	 * XXX calling convention in trap.c and init_main.c.
 	 */
 
 	/* Collect argc from the user stack */
 	argc = fuword((void *)stack);
 
-        /*
-         * XXX PG: these get overwritten in the syscall return code.
-         * execve() should return EJUSTRETURN, like it does on NetBSD.
-         * Emulate by setting the syscall return value cells. The
-         * registers still have to be set for init's fork trampoline.
-         */
-        td->td_retval[0] = argc;
-        td->td_retval[1] = stack + sizeof(register_t);
 	tf->fixreg[3] = argc;
 	tf->fixreg[4] = stack + sizeof(register_t);
 	tf->fixreg[5] = stack + (2 + argc)*sizeof(register_t);
@@ -572,8 +561,6 @@ ppc32_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 
 	argc = fuword32((void *)stack);
 
-        td->td_retval[0] = argc;
-        td->td_retval[1] = stack + sizeof(uint32_t);
 	tf->fixreg[3] = argc;
 	tf->fixreg[4] = stack + sizeof(uint32_t);
 	tf->fixreg[5] = stack + (2 + argc)*sizeof(uint32_t);
diff --git a/sys/riscv/riscv/machdep.c b/sys/riscv/riscv/machdep.c
index d8a968791a050..de1a71e4d2eb8 100644
--- a/sys/riscv/riscv/machdep.c
+++ b/sys/riscv/riscv/machdep.c
@@ -279,12 +279,7 @@ exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 
 	memset(tf, 0, sizeof(struct trapframe));
 
-	/*
-	 * We need to set a0 for init as it doesn't call
-	 * cpu_set_syscall_retval to copy the value. We also
-	 * need to set td_retval for the cases where we do.
-	 */
-	tf->tf_a[0] = td->td_retval[0] = stack;
+	tf->tf_a[0] = stack;
 	tf->tf_sp = STACKALIGN(stack);
 	tf->tf_ra = imgp->entry_addr;
 	tf->tf_sepc = imgp->entry_addr;
diff --git a/sys/sparc64/sparc64/machdep.c b/sys/sparc64/sparc64/machdep.c
index 9d917afafbbd6..d450d436379e9 100644
--- a/sys/sparc64/sparc64/machdep.c
+++ b/sys/sparc64/sparc64/machdep.c
@@ -1009,9 +1009,6 @@ exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
 	 * header, it turns out that just always using TSO performs best.
 	 */
 	tf->tf_tstate = TSTATE_IE | TSTATE_PEF | TSTATE_MM_TSO;
-
-	td->td_retval[0] = tf->tf_out[0];
-	td->td_retval[1] = tf->tf_out[1];
 }
 
 int