vmm(4), bhyve(8): Expose kernel-emulated special devices to userspace

Expose the special kernel LAPIC, IOAPIC, and HPET devices to userspace
for use in, e.g., fallback instruction emulation (when userspace has a
newer instruction decode/emulation layer than the kernel vmm(4)).

Plumb the ioctl through libvmmapi and register the memory ranges in
bhyve(8).

Reviewed by:	grehan
Differential Revision:	https://reviews.freebsd.org/D24525

2 files changed, 54 insertions, 0 deletions

diff --git a/sys/amd64/include/vmm_dev.h b/sys/amd64/include/vmm_dev.h
index 21775b70718e..250eef9e7afc 100644
--- a/sys/amd64/include/vmm_dev.h
+++ b/sys/amd64/include/vmm_dev.h
@@ -235,6 +235,15 @@ struct vm_cpu_topology {
 	uint16_t	maxcpus;
 };
 
+struct vm_readwrite_kernemu_device {
+	int		vcpuid;
+	unsigned	access_width : 3;
+	unsigned	_unused : 29;
+	uint64_t	gpa;
+	uint64_t	value;
+};
+_Static_assert(sizeof(struct vm_readwrite_kernemu_device) == 24, "ABI");
+
 enum {
 	/* general routines */
 	IOCNUM_ABIVERS = 0,
@@ -262,6 +271,8 @@ enum {
 	IOCNUM_GET_SEGMENT_DESCRIPTOR = 23,
 	IOCNUM_SET_REGISTER_SET = 24,
 	IOCNUM_GET_REGISTER_SET = 25,
+	IOCNUM_GET_KERNEMU_DEV = 26,
+	IOCNUM_SET_KERNEMU_DEV = 27,
 
 	/* interrupt injection */
 	IOCNUM_GET_INTINFO = 28,
@@ -347,6 +358,12 @@ enum {
 	_IOW('v', IOCNUM_SET_REGISTER_SET, struct vm_register_set)
 #define	VM_GET_REGISTER_SET \
 	_IOWR('v', IOCNUM_GET_REGISTER_SET, struct vm_register_set)
+#define	VM_SET_KERNEMU_DEV \
+	_IOW('v', IOCNUM_SET_KERNEMU_DEV, \
+	    struct vm_readwrite_kernemu_device)
+#define	VM_GET_KERNEMU_DEV \
+	_IOWR('v', IOCNUM_GET_KERNEMU_DEV, \
+	    struct vm_readwrite_kernemu_device)
 #define	VM_INJECT_EXCEPTION	\
 	_IOW('v', IOCNUM_INJECT_EXCEPTION, struct vm_exception)
 #define	VM_LAPIC_IRQ 		\
diff --git a/sys/amd64/vmm/vmm_dev.c b/sys/amd64/vmm/vmm_dev.c
index e47b7081b795..27849d883527 100644
--- a/sys/amd64/vmm/vmm_dev.c
+++ b/sys/amd64/vmm/vmm_dev.c
@@ -58,6 +58,7 @@ __FBSDID("$FreeBSD$");
 #include <machine/vmm_dev.h>
 #include <machine/vmm_instruction_emul.h>
 #include <machine/vmm_snapshot.h>
+#include <x86/apicreg.h>
 
 #include "vmm_lapic.h"
 #include "vmm_stat.h"
@@ -382,6 +383,7 @@ vmmdev_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag,
 	struct vm_rtc_data *rtcdata;
 	struct vm_memmap *mm;
 	struct vm_cpu_topology *topology;
+	struct vm_readwrite_kernemu_device *kernemu;
 	uint64_t *regvals;
 	int *regnums;
 #ifdef BHYVE_SNAPSHOT
@@ -561,6 +563,41 @@ vmmdev_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag,
 	case VM_IOAPIC_PINCOUNT:
 		*(int *)data = vioapic_pincount(sc->vm);
 		break;
+	case VM_SET_KERNEMU_DEV:
+	case VM_GET_KERNEMU_DEV: {
+		mem_region_write_t mwrite;
+		mem_region_read_t mread;
+		bool arg;
+
+		kernemu = (void *)data;
+
+		if (kernemu->access_width > 0)
+			size = (1u << kernemu->access_width);
+		else
+			size = 1;
+
+		if (kernemu->gpa >= DEFAULT_APIC_BASE && kernemu->gpa < DEFAULT_APIC_BASE + PAGE_SIZE) {
+			mread = lapic_mmio_read;
+			mwrite = lapic_mmio_write;
+		} else if (kernemu->gpa >= VIOAPIC_BASE && kernemu->gpa < VIOAPIC_BASE + VIOAPIC_SIZE) {
+			mread = vioapic_mmio_read;
+			mwrite = vioapic_mmio_write;
+		} else if (kernemu->gpa >= VHPET_BASE && kernemu->gpa < VHPET_BASE + VHPET_SIZE) {
+			mread = vhpet_mmio_read;
+			mwrite = vhpet_mmio_write;
+		} else {
+			error = EINVAL;
+			break;
+		}
+
+		if (cmd == VM_SET_KERNEMU_DEV)
+			error = mwrite(sc->vm, kernemu->vcpuid, kernemu->gpa,
+			    kernemu->value, size, &arg);
+		else
+			error = mread(sc->vm, kernemu->vcpuid, kernemu->gpa,
+			    &kernemu->value, size, &arg);
+		break;
+		}
 	case VM_ISA_ASSERT_IRQ:
 		isa_irq = (struct vm_isa_irq *)data;
 		error = vatpic_assert_irq(sc->vm, isa_irq->atpic_irq);