[librm] Speed up real-to-protected mode transition under KVM

Ensure that all segment registers have zero in the low two bits before
transitioning to protected mode.  This allows the CPU state to
immediately be deemed to be "valid", and eliminates the need for any
further emulated instructions.

Load the protected-mode interrupt descriptor table after switching to
protected mode, since this avoids triggering an EXCEPTION_NMI and
corresponding VM exit.

This reduces the time taken by real_to_prot under KVM by around 50%.

Signed-off-by: Michael Brown <mcb30@ipxe.org>

src/arch/i386/prefix/libprefix.S

@@ -522,8 +522,11 @@ alloc_basemem:
 	subw	$_data16_memsz_pgh, %ax
 	pushw	%ax
 
-	/* Calculate .text16 segment address */
+	/* Calculate .text16 segment address.  Round down to ensure
+	 * low bits are zero, to speed up mode transitions under KVM.
+	 */
 	subw	$_text16_memsz_pgh, %ax
+	andb	$~0x03, %al
 	pushw	%ax
 
 	/* Update FBMS */

src/arch/i386/transitions/librm.S

@@ -200,10 +200,22 @@ real_to_prot:
 	addr32 leal (%eax,%edx), %esi
 	subl	rm_virt_offset, %esi
 
+	/* Load protected-mode global descriptor table */
+	data32 lgdt gdtr
+
+	/* Zero segment registers.  This wastes around 12 cycles on
+	 * real hardware, but saves a substantial number of emulated
+	 * instructions under KVM.
+	 */
+	xorw	%ax, %ax
+	movw	%ax, %ds
+	movw	%ax, %es
+	movw	%ax, %fs
+	movw	%ax, %gs
+	movw	%ax, %ss
+
 	/* Switch to protected mode */
 	cli
-	data32 lgdt gdtr
-	data32 lidt idtr
 	movl	%cr0, %eax
 	orb	$CR0_PE, %al
 	movl	%eax, %cr0
@@ -220,6 +232,9 @@ r2p_pmode:
 	movw	%ax, %ss
 	movl	pm_esp, %esp
 
+	/* Load protected-mode interrupt descriptor table */
+	lidt	idtr
+
 	/* Record real-mode %ss:sp (after removal of data) */
 	movw	%bp, rm_ss
 	addl	%ecx, %edx

src/arch/i386/transitions/librm_mgmt.c

@@ -26,10 +26,9 @@ static struct interrupt_vector intr_vec[ IRQ_MAX + 1 ];
 struct interrupt_descriptor idt[NUM_INT] __attribute__ (( aligned ( 16 ) ));
 
 /** The interrupt descriptor table register */
-struct idtr __data16 ( idtr ) = {
+struct idtr idtr = {
 	.limit = ( sizeof ( idt ) - 1 ),
 };
-#define idtr __use_data16 ( idtr )
 
 /**
  * Allocate space on the real-mode stack and copy data there from a