Create a separate structure for per-CPU state saved across suspend and

resume that is a superset of a pcb.  Move the FPU state out of the pcb and
into this new structure.  As part of this, move the FPU resume code on
amd64 into a C function.  This allows resumectx() to still operate only on
a pcb and more closely mirrors the i386 code.

Reviewed by:	kib (earlier version)

11 files changed, 76 insertions, 68 deletions

diff --git a/sys/amd64/amd64/cpu_switch.S b/sys/amd64/amd64/cpu_switch.S
index 7c37a41054ec..eb0ee8b5fdd5 100644
--- a/sys/amd64/amd64/cpu_switch.S
+++ b/sys/amd64/amd64/cpu_switch.S
@@ -399,10 +399,6 @@ ENTRY(savectx)
 	rdmsr
 	movl	%eax,PCB_SFMASK(%rdi)
 	movl	%edx,PCB_SFMASK+4(%rdi)
-	movl	xsave_mask,%eax
-	movl	%eax,PCB_XSMASK(%rdi)
-	movl	xsave_mask+4,%eax
-	movl	%eax,PCB_XSMASK+4(%rdi)
 
 	sgdt	PCB_GDT(%rdi)
 	sidt	PCB_IDT(%rdi)
@@ -467,12 +463,9 @@ ENTRY(resumectx)
 	movl	PCB_SFMASK(%rdi),%eax
 	wrmsr
 
-	/* Restore CR0 except for FPU mode. */
+	/* Restore CR0, CR2, CR4 and CR3. */
 	movq	PCB_CR0(%rdi),%rax
-	andq	$~(CR0_EM | CR0_TS),%rax
 	movq	%rax,%cr0
-
-	/* Restore CR2, CR4 and CR3. */
 	movq	PCB_CR2(%rdi),%rax
 	movq	%rax,%cr2
 	movq	PCB_CR4(%rdi),%rax
@@ -510,26 +503,6 @@ ENTRY(resumectx)
 	movq	PCB_DR7(%rdi),%rax
 	movq	%rax,%dr7
 
-	/* Restore FPU state. */
-	fninit
-	movq	PCB_FPUSUSPEND(%rdi),%rbx
-	movq	PCB_XSMASK(%rdi),%rax
-	testq	%rax,%rax
-	jz	1f
-	movq	%rax,%rdx
-	shrq	$32,%rdx
-	movl	$XCR0,%ecx
-	xsetbv
-	xrstor	(%rbx)
-	jmp	2f
-1:
-	fxrstor	(%rbx)
-2:
-
-	/* Reload CR0. */
-	movq	PCB_CR0(%rdi),%rax
-	movq	%rax,%cr0
-
 	/* Restore other callee saved registers. */
 	movq	PCB_R15(%rdi),%r15
 	movq	PCB_R14(%rdi),%r14
diff --git a/sys/amd64/amd64/fpu.c b/sys/amd64/amd64/fpu.c
index 0f4b2b52aeef..fef47a09a134 100644
--- a/sys/amd64/amd64/fpu.c
+++ b/sys/amd64/amd64/fpu.c
@@ -173,6 +173,20 @@ fpususpend(void *addr)
 	load_cr0(cr0);
 }
 
+void
+fpuresume(void *addr)
+{
+	u_long cr0;
+
+	cr0 = rcr0();
+	stop_emulating();
+	fninit();
+	if (use_xsave)
+		load_xcr(XCR0, xsave_mask);
+	fpurestore(addr);
+	load_cr0(cr0);
+}
+
 /*
  * Enable XSAVE if supported and allowed by user.
  * Calculate the xsave_mask.
diff --git a/sys/amd64/amd64/genassym.c b/sys/amd64/amd64/genassym.c
index ff1f28287b88..c4b1c2878497 100644
--- a/sys/amd64/amd64/genassym.c
+++ b/sys/amd64/amd64/genassym.c
@@ -163,8 +163,6 @@ ASSYM(PCB_STAR, offsetof(struct pcb, pcb_star));
 ASSYM(PCB_LSTAR, offsetof(struct pcb, pcb_lstar));
 ASSYM(PCB_CSTAR, offsetof(struct pcb, pcb_cstar));
 ASSYM(PCB_SFMASK, offsetof(struct pcb, pcb_sfmask));
-ASSYM(PCB_XSMASK, offsetof(struct pcb, pcb_xsmask));
-ASSYM(PCB_FPUSUSPEND, offsetof(struct pcb, pcb_fpususpend));
 ASSYM(PCB_SIZE, sizeof(struct pcb));
 ASSYM(PCB_FULL_IRET, PCB_FULL_IRET);
 ASSYM(PCB_DBREGS, PCB_DBREGS);
diff --git a/sys/amd64/amd64/mp_machdep.c b/sys/amd64/amd64/mp_machdep.c
index 9b12449ef196..15d3a7b7895c 100644
--- a/sys/amd64/amd64/mp_machdep.c
+++ b/sys/amd64/amd64/mp_machdep.c
@@ -102,7 +102,7 @@ char *nmi_stack;
 void *dpcpu;
 
 struct pcb stoppcbs[MAXCPU];
-struct pcb **susppcbs;
+struct susppcb **susppcbs;
 
 /* Variables needed for SMP tlb shootdown. */
 vm_offset_t smp_tlb_addr2;
@@ -1461,11 +1461,12 @@ cpususpend_handler(void)
 	mtx_assert(&smp_ipi_mtx, MA_NOTOWNED);
 
 	cpu = PCPU_GET(cpuid);
-	if (savectx(susppcbs[cpu])) {
-		fpususpend(susppcbs[cpu]->pcb_fpususpend);
+	if (savectx(&susppcbs[cpu]->sp_pcb)) {
+		fpususpend(susppcbs[cpu]->sp_fpususpend);
 		wbinvd();
 		CPU_SET_ATOMIC(cpu, &suspended_cpus);
 	} else {
+		fpuresume(susppcbs[cpu]->sp_fpususpend);
 		pmap_init_pat();
 		initializecpu();
 		PCPU_SET(switchtime, 0);
diff --git a/sys/amd64/include/fpu.h b/sys/amd64/include/fpu.h
index be1bdc6c5d7f..78d2fee2714d 100644
--- a/sys/amd64/include/fpu.h
+++ b/sys/amd64/include/fpu.h
@@ -58,6 +58,7 @@ int	fpuformat(void);
 int	fpugetregs(struct thread *td);
 void	fpuinit(void);
 void	fpurestore(void *addr);
+void	fpuresume(void *addr);
 void	fpusave(void *addr);
 int	fpusetregs(struct thread *td, struct savefpu *addr,
 	    char *xfpustate, size_t xfpustate_size);
diff --git a/sys/amd64/include/pcb.h b/sys/amd64/include/pcb.h
index 80aff86eb685..153393ffc15b 100644
--- a/sys/amd64/include/pcb.h
+++ b/sys/amd64/include/pcb.h
@@ -97,14 +97,18 @@ struct pcb {
 	register_t	pcb_lstar;
 	register_t	pcb_cstar;
 	register_t	pcb_sfmask;
-	register_t	pcb_xsmask;
-
-	/* fpu context for suspend/resume */
-	void		*pcb_fpususpend;
 
 	struct savefpu	*pcb_save;
 
-	uint64_t	pcb_pad[3];
+	uint64_t	pcb_pad[5];
+};
+
+/* Per-CPU state saved during suspend and resume. */
+struct susppcb {
+	struct pcb	sp_pcb;
+
+	/* fpu context for suspend/resume */
+	void		*sp_fpususpend;
 };
 #endif
 
diff --git a/sys/dev/acpica/acpi.c b/sys/dev/acpica/acpi.c
index b37e252f2138..47ab4c63da18 100644
--- a/sys/dev/acpica/acpi.c
+++ b/sys/dev/acpica/acpi.c
@@ -1211,6 +1211,11 @@ acpi_set_resource(device_t dev, device_t child, int type, int rid,
 		if (!(type == SYS_RES_IOPORT && start == CONF1_ADDR_PORT))
 #endif
 		{
+		    if (bootverbose)
+			device_printf(dev,
+			    "Ignoring %s range %#lx-%#lx for %s\n",
+			    type == SYS_RES_MEMORY ? "memory" : "I/O",
+			    start, start + count - 1, acpi_name(ad->ad_handle));
 		    AcpiOsFree(devinfo);
 		    return (0);
 		}
diff --git a/sys/i386/i386/mp_machdep.c b/sys/i386/i386/mp_machdep.c
index 1ea8f9572d6a..51e8f4948fbb 100644
--- a/sys/i386/i386/mp_machdep.c
+++ b/sys/i386/i386/mp_machdep.c
@@ -147,7 +147,7 @@ void *bootstacks[MAXCPU];
 static void *dpcpu;
 
 struct pcb stoppcbs[MAXCPU];
-struct pcb **susppcbs = NULL;
+struct susppcb **susppcbs;
 
 /* Variables needed for SMP tlb shootdown. */
 vm_offset_t smp_tlb_addr1;
@@ -1521,12 +1521,12 @@ cpususpend_handler(void)
 	mtx_assert(&smp_ipi_mtx, MA_NOTOWNED);
 
 	cpu = PCPU_GET(cpuid);
-	if (savectx(susppcbs[cpu])) {
-		npxsuspend(&susppcbs[cpu]->pcb_fpususpend);
+	if (savectx(&susppcbs[cpu]->sp_pcb)) {
+		npxsuspend(&susppcbs[cpu]->sp_fpususpend);
 		wbinvd();
 		CPU_SET_ATOMIC(cpu, &suspended_cpus);
 	} else {
-		npxresume(&susppcbs[cpu]->pcb_fpususpend);
+		npxresume(&susppcbs[cpu]->sp_fpususpend);
 		pmap_init_pat();
 		PCPU_SET(switchtime, 0);
 		PCPU_SET(switchticks, ticks);
diff --git a/sys/i386/include/pcb.h b/sys/i386/include/pcb.h
index a654ad3e6c39..01ace723e501 100644
--- a/sys/i386/include/pcb.h
+++ b/sys/i386/include/pcb.h
@@ -90,8 +90,11 @@ struct pcb {
 	struct region_descriptor pcb_idt;
 	uint16_t	pcb_ldt;
 	uint16_t	pcb_tr;
+};
 
-	union	savefpu pcb_fpususpend;
+struct susppcb {
+	struct pcb	sp_pcb;
+	union savefpu	sp_fpususpend;
 };
 
 #ifdef _KERNEL
diff --git a/sys/sparc64/pci/psycho.c b/sys/sparc64/pci/psycho.c
index 57e0f4817b5a..cd04f2899735 100644
--- a/sys/sparc64/pci/psycho.c
+++ b/sys/sparc64/pci/psycho.c
@@ -448,24 +448,30 @@ psycho_attach(device_t dev)
 
 	i = OF_getprop_alloc(node, "ranges", sizeof(*range), (void **)&range);
 	/*
-	 * Make sure that the expected ranges are present.  The
-	 * OFW_PCI_CS_MEM64 one is not currently used though.
-	 */
-	if (i != PSYCHO_NRANGE)
-		panic("%s: unsupported number of ranges", __func__);
-	/*
 	 * Find the addresses of the various bus spaces.
 	 * There should not be multiple ones of one kind.
 	 * The physical start addresses of the ranges are the configuration,
 	 * memory and I/O handles.
 	 */
-	for (i = 0; i < PSYCHO_NRANGE; i++) {
+	for (; i >= 0; i--) {
 		j = OFW_PCI_RANGE_CS(&range[i]);
 		if (sc->sc_pci_bh[j] != 0)
 			panic("%s: duplicate range for space %d",
 			    __func__, j);
 		sc->sc_pci_bh[j] = OFW_PCI_RANGE_PHYS(&range[i]);
 	}
+
+	/*
+	 * Make sure that the expected ranges are present.  The
+	 * OFW_PCI_CS_MEM64 one is not currently used.
+	 */
+	if (sc->sc_pci_bh[OFW_PCI_CS_CONFIG] == 0)
+		panic("%s: missing CONFIG range", __func__);
+	if (sc->sc_pci_bh[OFW_PCI_CS_IO] == 0)
+		panic("%s: missing IO range", __func__);
+	if (sc->sc_pci_bh[OFW_PCI_CS_MEM32] == 0)
+		panic("%s: missing MEM32 range", __func__);
+	
 	free(range, M_OFWPROP);
 
 	/* Register the softc, this is needed for paired Psychos. */
diff --git a/sys/x86/acpica/acpi_wakeup.c b/sys/x86/acpica/acpi_wakeup.c
index fb4698396db8..3d5e1fe37a77 100644
--- a/sys/x86/acpica/acpi_wakeup.c
+++ b/sys/x86/acpica/acpi_wakeup.c
@@ -75,10 +75,10 @@ extern int		acpi_resume_beep;
 extern int		acpi_reset_video;
 
 #ifdef SMP
-extern struct pcb	**susppcbs;
+extern struct susppcb	**susppcbs;
 static cpuset_t		suspcpus;
 #else
-static struct pcb	**susppcbs;
+static struct susppcb	**susppcbs;
 #endif
 
 static void		*acpi_alloc_wakeup_handler(void);
@@ -117,14 +117,15 @@ acpi_stop_beep(void *arg)
 static int
 acpi_wakeup_ap(struct acpi_softc *sc, int cpu)
 {
+	struct pcb *pcb;
 	int		vector = (WAKECODE_PADDR(sc) >> 12) & 0xff;
 	int		apic_id = cpu_apic_ids[cpu];
 	int		ms;
 
-	WAKECODE_FIXUP(wakeup_pcb, struct pcb *, susppcbs[cpu]);
-	WAKECODE_FIXUP(wakeup_gdt, uint16_t, susppcbs[cpu]->pcb_gdt.rd_limit);
-	WAKECODE_FIXUP(wakeup_gdt + 2, uint64_t,
-	    susppcbs[cpu]->pcb_gdt.rd_base);
+	pcb = &susppcbs[cpu]->sp_pcb;
+	WAKECODE_FIXUP(wakeup_pcb, struct pcb *, pcb);
+	WAKECODE_FIXUP(wakeup_gdt, uint16_t, pcb->pcb_gdt.rd_limit);
+	WAKECODE_FIXUP(wakeup_gdt + 2, uint64_t, pcb->pcb_gdt.rd_base);
 
 	ipi_startup(apic_id, vector);
 
@@ -188,6 +189,7 @@ int
 acpi_sleep_machdep(struct acpi_softc *sc, int state)
 {
 	ACPI_STATUS	status;
+	struct pcb	*pcb;
 
 	if (sc->acpi_wakeaddr == 0ul)
 		return (-1);	/* couldn't alloc wake memory */
@@ -204,11 +206,12 @@ acpi_sleep_machdep(struct acpi_softc *sc, int state)
 
 	intr_suspend();
 
-	if (savectx(susppcbs[0])) {
+	pcb = &susppcbs[0]->sp_pcb;
+	if (savectx(pcb)) {
 #ifdef __amd64__
-		fpususpend(susppcbs[0]->pcb_fpususpend);
+		fpususpend(susppcbs[0]->sp_fpususpend);
 #elif defined(DEV_NPX)
-		npxsuspend(&susppcbs[0]->pcb_fpususpend);
+		npxsuspend(&susppcbs[0]->sp_fpususpend);
 #endif
 #ifdef SMP
 		if (!CPU_EMPTY(&suspcpus) && suspend_cpus(suspcpus) == 0) {
@@ -221,13 +224,11 @@ acpi_sleep_machdep(struct acpi_softc *sc, int state)
 		WAKECODE_FIXUP(reset_video, uint8_t, (acpi_reset_video != 0));
 
 #ifndef __amd64__
-		WAKECODE_FIXUP(wakeup_cr4, register_t, susppcbs[0]->pcb_cr4);
+		WAKECODE_FIXUP(wakeup_cr4, register_t, pcb->pcb_cr4);
 #endif
-		WAKECODE_FIXUP(wakeup_pcb, struct pcb *, susppcbs[0]);
-		WAKECODE_FIXUP(wakeup_gdt, uint16_t,
-		    susppcbs[0]->pcb_gdt.rd_limit);
-		WAKECODE_FIXUP(wakeup_gdt + 2, uint64_t,
-		    susppcbs[0]->pcb_gdt.rd_base);
+		WAKECODE_FIXUP(wakeup_pcb, struct pcb *, pcb);
+		WAKECODE_FIXUP(wakeup_gdt, uint16_t, pcb->pcb_gdt.rd_limit);
+		WAKECODE_FIXUP(wakeup_gdt + 2, uint64_t, pcb->pcb_gdt.rd_base);
 
 		/* Call ACPICA to enter the desired sleep state */
 		if (state == ACPI_STATE_S4 && sc->acpi_s4bios)
@@ -244,8 +245,10 @@ acpi_sleep_machdep(struct acpi_softc *sc, int state)
 		for (;;)
 			ia32_pause();
 	} else {
-#ifdef DEV_NPX
-		npxresume(&susppcbs[0]->pcb_fpususpend);
+#ifdef __amd64__
+		fpuresume(susppcbs[0]->sp_fpususpend);
+#elif defined(DEV_NPX)
+		npxresume(&susppcbs[0]->sp_fpususpend);
 #endif
 	}
 
@@ -325,7 +328,7 @@ acpi_alloc_wakeup_handler(void)
 	for (i = 0; i < mp_ncpus; i++) {
 		susppcbs[i] = malloc(sizeof(**susppcbs), M_DEVBUF, M_WAITOK);
 #ifdef __amd64__
-		susppcbs[i]->pcb_fpususpend = alloc_fpusave(M_WAITOK);
+		susppcbs[i]->sp_fpususpend = alloc_fpusave(M_WAITOK);
 #endif
 	}