[PATCH RFC 2/2] Hardware task switching support

[PATCH RFC 2/2] Hardware task switching support

[Izik Eidus]

>From 6a7207a0f3ee8af6ebafcec9d40a75b87f00a129 Mon Sep 17 00:00:00 2001
From: Izik Eidus <izike@qumranet.com>
Date: Thu, 13 Mar 2008 02:34:21 +0200
Subject: [PATCH] KVM: hardware task switching support

Signed-off-by: Izik Eidus <izike@qumranet.com>
---
 arch/x86/kvm/svm.c         |   11 +-
 arch/x86/kvm/tss_segment.h |   59 +++++++
 arch/x86/kvm/vmx.c         |   15 ++
 arch/x86/kvm/x86.c         |  385 ++++++++++++++++++++++++++++++++++++++++++++
 include/asm-x86/kvm_host.h |    2 +
 5 files changed, 469 insertions(+), 3 deletions(-)
 create mode 100644 arch/x86/kvm/tss_segment.h

diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 4e1dd61..be78278 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -1121,9 +1121,14 @@ static int invalid_op_interception(struct vcpu_svm *svm,
 static int task_switch_interception(struct vcpu_svm *svm,
 				    struct kvm_run *kvm_run)
 {
-	pr_unimpl(&svm->vcpu, "%s: task switch is unsupported\n", __func__);
-	kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
-	return 0;
+	u16 tss_selector;
+
+	tss_selector = (u16)svm->vmcb->control.exit_info_1;
+	if(svm->vmcb->control.exit_info_2 & ((unsigned long)1 << 36))
+		return kvm_task_switch(&svm->vcpu, tss_selector, 1);
+	if(svm->vmcb->control.exit_info_2 & ((unsigned long)1 << 38))
+		return kvm_task_switch(&svm->vcpu, tss_selector, 2);
+	return kvm_task_switch(&svm->vcpu, tss_selector, 0);
 }
 
 static int cpuid_interception(struct vcpu_svm *svm, struct kvm_run *kvm_run)
diff --git a/arch/x86/kvm/tss_segment.h b/arch/x86/kvm/tss_segment.h
new file mode 100644
index 0000000..622aa10
--- /dev/null
+++ b/arch/x86/kvm/tss_segment.h
@@ -0,0 +1,59 @@
+#ifndef __TSS_SEGMENT_H
+#define __TSS_SEGMENT_H
+
+struct tss_segment_32 {
+	u32 prev_task_link;
+	u32 esp0;
+	u32 ss0;
+	u32 esp1;
+	u32 ss1;
+	u32 esp2;
+	u32 ss2;
+	u32 cr3;
+	u32 eip;
+	u32 eflags;
+	u32 eax;
+	u32 ecx;
+	u32 edx;
+	u32 ebx;
+	u32 esp;
+	u32 ebp;
+	u32 esi;
+	u32 edi;
+	u32 es;
+	u32 cs;
+	u32 ss;
+	u32 ds;
+	u32 fs;
+	u32 gs;
+	u32 ldt_selector;
+	u16 t;
+	u16 io_map;
+};
+
+struct tss_segment_16 {
+	u16 prev_task_link;
+	u16 sp0;
+	u16 ss0;
+	u16 sp1;
+	u16 ss1;
+	u16 sp2;
+	u16 ss2;
+	u16 ip;
+	u16 flag;
+	u16 ax;
+	u16 cx;
+	u16 dx;
+	u16 bx;
+	u16 sp;
+	u16 bp;
+	u16 si;
+	u16 di;
+	u16 es;
+	u16 cs;
+	u16 ss;
+	u16 ds;
+	u16 ldt;
+};
+
+#endif
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 459b0bd..e5fe09b 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2246,6 +2246,20 @@ static int handle_apic_access(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 	return 1;
 }
 
+static int handle_task_switch(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	unsigned long exit_qualification;
+	u16 tss_selector;
+	int reason;
+	
+	exit_qualification = vmcs_readl(EXIT_QUALIFICATION);
+
+	reason = (u32)exit_qualification >> 30;
+	tss_selector = (u16)exit_qualification;
+
+	return kvm_task_switch(vcpu, tss_selector, reason);
+}
+
 /*
  * The exit handlers return 1 if the exit was handled fully and guest execution
  * may resume.  Otherwise they set the kvm_run parameter to indicate what needs
@@ -2268,6 +2282,7 @@ static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu,
 	[EXIT_REASON_TPR_BELOW_THRESHOLD]     = handle_tpr_below_threshold,
 	[EXIT_REASON_APIC_ACCESS]             = handle_apic_access,
 	[EXIT_REASON_WBINVD]                  = handle_wbinvd,
+	[EXIT_REASON_TASK_SWITCH]             = handle_task_switch,
 };
 
 static const int kvm_vmx_max_exit_handlers =
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 5339ab1..7feb41e 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -18,6 +18,7 @@
 #include "irq.h"
 #include "mmu.h"
 #include "i8254.h"
+#include "tss_segment.h"
 
 #include <linux/clocksource.h>
 #include <linux/kvm.h>
@@ -3050,6 +3051,390 @@ static void set_segment(struct kvm_vcpu *vcpu,
 	kvm_x86_ops->set_segment(vcpu, var, seg);
 }
 
+static void seg_desct_to_kvm_desct(struct desc_struct *seg_desc, u16 selector,
+				   struct kvm_segment *kvm_desct)
+{
+	kvm_desct->base = seg_desc->base0;
+	kvm_desct->base |= seg_desc->base1 << 16;
+	kvm_desct->base |= seg_desc->base2 << 24;
+	kvm_desct->limit = seg_desc->limit0;
+	kvm_desct->limit |= seg_desc->limit << 16;
+	kvm_desct->selector = selector;
+	kvm_desct->type = seg_desc->type;
+	kvm_desct->present = seg_desc->p;
+	kvm_desct->dpl = seg_desc->dpl;
+	kvm_desct->db = seg_desc->d;
+	kvm_desct->s = seg_desc->s;
+	kvm_desct->l = seg_desc->l;
+	kvm_desct->g = seg_desc->g;
+	kvm_desct->avl = seg_desc->avl;
+	kvm_desct->unusable = 0;
+	kvm_desct->padding = 0;
+}
+
+/* allowed just for 8 bytes segments */
+static int load_guest_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector,
+			      struct desc_struct *seg_desc)
+{
+	struct descriptor_table gdt_ldt;
+	u16 index = selector >> 3;
+
+	if (selector & 1 << 2)
+		kvm_x86_ops->get_ldt(vcpu, &gdt_ldt);
+	else
+		kvm_x86_ops->get_gdt(vcpu, &gdt_ldt);
+	if (gdt_ldt.limit < index * 8 + 7)
+		return 1;
+	return kvm_read_guest(vcpu->kvm, gdt_ldt.base + index * 8, seg_desc, 8);
+}
+
+/* allowed just for 8 bytes segments */
+static int save_guest_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector,
+			      struct desc_struct *seg_desc)
+{
+	struct descriptor_table gdt_ldt;
+	u16 index = selector >> 3;
+
+	if (selector & 1 << 2)
+		kvm_x86_ops->get_ldt(vcpu, &gdt_ldt);
+	else
+		kvm_x86_ops->get_gdt(vcpu, &gdt_ldt);
+	if (gdt_ldt.limit < index * 8 + 7)
+		return 1;
+	return kvm_write_guest(vcpu->kvm, gdt_ldt.base + index * 8, seg_desc, 8);
+}
+
+static int load_tss_segment32(struct kvm_vcpu *vcpu,
+			      struct desc_struct *seg_desc,
+			      struct tss_segment_32 *tss)
+{
+	u32 base_addr;
+
+	base_addr = seg_desc->base0;
+	base_addr |= (seg_desc->base1 << 16);
+	base_addr |= (seg_desc->base2 << 24);
+
+	return kvm_read_guest(vcpu->kvm, base_addr, tss,
+			      sizeof(struct tss_segment_32));
+}
+
+static int save_tss_segment32(struct kvm_vcpu *vcpu,
+			      struct desc_struct *seg_desc,
+			      struct tss_segment_32 *tss)
+{
+	u32 base_addr;
+
+	base_addr = seg_desc->base0;
+	base_addr |= (seg_desc->base1 << 16);
+	base_addr |= (seg_desc->base2 << 24);
+
+	return kvm_write_guest(vcpu->kvm, base_addr, tss,
+			       sizeof(struct tss_segment_32));
+}
+
+static int load_tss_segment16(struct kvm_vcpu *vcpu,
+			      struct desc_struct *seg_desc,
+			      struct tss_segment_16 *tss)
+{
+	u32 base_addr;
+
+	base_addr = seg_desc->base0;
+	base_addr |= (seg_desc->base1 << 16);
+	base_addr |= (seg_desc->base2 << 24);
+
+	return kvm_read_guest(vcpu->kvm, base_addr, tss,
+			      sizeof(struct tss_segment_16));
+}
+
+static int save_tss_segment16(struct kvm_vcpu *vcpu,
+			      struct desc_struct *seg_desc,
+			      struct tss_segment_16 *tss)
+{
+	u32 base_addr;
+
+	base_addr = seg_desc->base0;
+	base_addr |= (seg_desc->base1 << 16);
+	base_addr |= (seg_desc->base2 << 24);
+
+	return kvm_write_guest(vcpu->kvm, base_addr, tss,
+			       sizeof(struct tss_segment_16));
+}
+
+static void save_state_to_tss32(struct kvm_vcpu *vcpu,
+				struct tss_segment_32 *tss)
+{
+	struct kvm_segment kvm_seg;
+
+	tss->cr3 = vcpu->arch.cr3;
+	tss->eip = vcpu->arch.rip;
+	tss->eflags = kvm_x86_ops->get_rflags(vcpu);
+	tss->eax = vcpu->arch.regs[VCPU_REGS_RAX];
+	tss->ecx = vcpu->arch.regs[VCPU_REGS_RCX];
+	tss->edx = vcpu->arch.regs[VCPU_REGS_RDX];
+	tss->ebx = vcpu->arch.regs[VCPU_REGS_RBX];
+	tss->esp = vcpu->arch.regs[VCPU_REGS_RSP];
+	tss->ebp = vcpu->arch.regs[VCPU_REGS_RBP];
+	tss->esi = vcpu->arch.regs[VCPU_REGS_RSI];
+	tss->edi = vcpu->arch.regs[VCPU_REGS_RDI];
+	
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_ES);
+	tss->es = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_CS);
+	tss->cs = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_SS);
+	tss->ss = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_DS);
+	tss->ds = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_FS);
+	tss->fs = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_GS);
+	tss->gs = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_LDTR);
+	tss->ldt_selector = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_TR);
+	tss->prev_task_link = kvm_seg.selector;
+}
+
+static void load_state_from_tss32(struct kvm_vcpu *vcpu,
+				  struct tss_segment_32 *tss)
+{
+	struct desc_struct seg_desc;
+	struct kvm_segment kvm_seg;
+
+	if (vcpu->arch.cr0 & 1 << 31)
+		kvm_set_cr3(vcpu, tss->cr3);
+
+	vcpu->arch.rip = tss->eip;
+	kvm_x86_ops->set_rflags(vcpu, (tss->eflags | 2) & ~0x8028);
+
+	vcpu->arch.regs[VCPU_REGS_RAX] = tss->eax;
+	vcpu->arch.regs[VCPU_REGS_RCX] = tss->ecx;
+	vcpu->arch.regs[VCPU_REGS_RDX] = tss->edx;
+	vcpu->arch.regs[VCPU_REGS_RBX] = tss->ebx;
+	vcpu->arch.regs[VCPU_REGS_RSP] = tss->esp;
+	vcpu->arch.regs[VCPU_REGS_RBP] = tss->ebp;
+	vcpu->arch.regs[VCPU_REGS_RSI] = tss->esi;
+	vcpu->arch.regs[VCPU_REGS_RDI] = tss->edi;
+
+	load_guest_segment_descriptor(vcpu, tss->ldt_selector,
+				      &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->ldt_selector,
+			       &kvm_seg);
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_LDTR);
+
+	load_guest_segment_descriptor(vcpu, tss->es, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->es, &kvm_seg);
+	kvm_seg.type |= 1;
+	if (!kvm_seg.s)
+		kvm_seg.unusable = 1;
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_ES);
+
+	load_guest_segment_descriptor(vcpu, tss->cs, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->cs, &kvm_seg);
+	kvm_seg.type |= 9;
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_CS);
+
+	load_guest_segment_descriptor(vcpu, tss->ss, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->ss, &kvm_seg);
+	kvm_seg.type |= 1;
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_SS);
+
+	load_guest_segment_descriptor(vcpu, tss->ds, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->ds, &kvm_seg);
+	kvm_seg.type |= 1;
+	if (!kvm_seg.s)
+		kvm_seg.unusable = 1;
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_DS);
+
+	load_guest_segment_descriptor(vcpu, tss->fs, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->fs, &kvm_seg);
+	kvm_seg.type |= 1;
+	if (!kvm_seg.s)
+		kvm_seg.unusable = 1;
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_FS);
+
+	load_guest_segment_descriptor(vcpu, tss->gs, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->gs, &kvm_seg);
+	kvm_seg.type |= 1;
+	if (!kvm_seg.s)
+		kvm_seg.unusable = 1;
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_GS);
+}
+
+static void save_state_to_tss16(struct kvm_vcpu *vcpu,
+				struct tss_segment_16 *tss)
+{
+	struct kvm_segment kvm_seg;
+
+	tss->ip = vcpu->arch.rip;
+	tss->flag = kvm_x86_ops->get_rflags(vcpu);
+	tss->ax = vcpu->arch.regs[VCPU_REGS_RAX];
+	tss->cx = vcpu->arch.regs[VCPU_REGS_RCX];
+	tss->dx = vcpu->arch.regs[VCPU_REGS_RDX];
+	tss->bx = vcpu->arch.regs[VCPU_REGS_RBX];
+	tss->sp = vcpu->arch.regs[VCPU_REGS_RSP];
+	tss->bp = vcpu->arch.regs[VCPU_REGS_RBP];
+	tss->si = vcpu->arch.regs[VCPU_REGS_RSI];
+	tss->di = vcpu->arch.regs[VCPU_REGS_RDI];
+	
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_ES);
+	tss->es = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_CS);
+	tss->cs = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_SS);
+	tss->ss = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_DS);
+	tss->ds = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_LDTR);
+	tss->ldt = kvm_seg.selector;
+	get_segment(vcpu, &kvm_seg, VCPU_SREG_TR);
+	tss->prev_task_link = kvm_seg.selector;
+
+}
+
+static void load_state_from_tss16(struct kvm_vcpu *vcpu,
+				  struct tss_segment_16 *tss)
+{
+	struct desc_struct seg_desc;
+	struct kvm_segment kvm_seg;
+
+	vcpu->arch.rip = tss->ip;
+	kvm_x86_ops->set_rflags(vcpu, tss->flag);
+	vcpu->arch.regs[VCPU_REGS_RAX] = tss->ax;
+	vcpu->arch.regs[VCPU_REGS_RCX] = tss->cx;
+	vcpu->arch.regs[VCPU_REGS_RDX] = tss->dx;
+	vcpu->arch.regs[VCPU_REGS_RBX] = tss->bx;
+	vcpu->arch.regs[VCPU_REGS_RSP] = tss->sp;
+	vcpu->arch.regs[VCPU_REGS_RBP] = tss->bp;
+	vcpu->arch.regs[VCPU_REGS_RSI] = tss->si;
+	vcpu->arch.regs[VCPU_REGS_RDI] = tss->di;
+	
+	load_guest_segment_descriptor(vcpu, tss->es, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->es, &kvm_seg);
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_ES);
+
+	load_guest_segment_descriptor(vcpu, tss->cs, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->cs, &kvm_seg);
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_CS);
+
+	load_guest_segment_descriptor(vcpu, tss->ss, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->ss, &kvm_seg);
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_SS);
+
+	load_guest_segment_descriptor(vcpu, tss->ds, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->ds, &kvm_seg);
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_DS);
+
+	load_guest_segment_descriptor(vcpu, tss->ldt, &seg_desc);
+	seg_desct_to_kvm_desct(&seg_desc, tss->ldt, &kvm_seg);
+	set_segment(vcpu, &kvm_seg, VCPU_SREG_LDTR);
+}
+
+int kvm_task_switch_16(struct kvm_vcpu *vcpu, u16 tss_selector, int reason,
+		       struct desc_struct *cseg_desc,
+		       struct desc_struct *nseg_desc)
+{
+	struct tss_segment_16 tss_segment_16;
+	int ret = 0;
+
+	if (load_tss_segment16(vcpu, cseg_desc, &tss_segment_16))
+		goto out;
+
+	save_state_to_tss16(vcpu, &tss_segment_16);
+	save_tss_segment16(vcpu, cseg_desc, &tss_segment_16);
+
+	load_tss_segment16(vcpu, nseg_desc, &tss_segment_16);
+	load_state_from_tss16(vcpu, &tss_segment_16);
+
+	ret = 1;
+out:
+	return ret;
+}
+
+int kvm_task_switch_32(struct kvm_vcpu *vcpu, u16 tss_selector, int reason,
+		       struct desc_struct *cseg_desc,
+		       struct desc_struct *nseg_desc)
+{
+	struct tss_segment_32 tss_segment_32;
+	int ret = 0;
+
+	if (load_tss_segment32(vcpu, cseg_desc, &tss_segment_32))
+		goto out;
+
+	save_state_to_tss32(vcpu, &tss_segment_32);
+	save_tss_segment32(vcpu, cseg_desc, &tss_segment_32);
+
+	if (load_tss_segment32(vcpu, nseg_desc, &tss_segment_32))
+		goto out;
+	load_state_from_tss32(vcpu, &tss_segment_32);
+
+	ret = 1;
+out:
+	return ret;
+}
+
+int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason)
+{
+	struct kvm_segment tr_seg;
+	struct desc_struct cseg_desc;
+	struct desc_struct nseg_desc;
+	int ret = 0;
+
+	get_segment(vcpu, &tr_seg, VCPU_SREG_TR);
+
+	if (load_guest_segment_descriptor(vcpu, tss_selector, &nseg_desc))
+		goto out;
+
+	if (load_guest_segment_descriptor(vcpu, tr_seg.selector, &cseg_desc))
+		goto out;
+
+	if (!nseg_desc.p || (nseg_desc.limit0 | nseg_desc.limit << 16) < 0x67)
+		goto out;
+
+	if (reason == 1 || reason == 2) {
+		cseg_desc.type &= ~(1 << 8); //clear the B flag
+		save_guest_segment_descriptor(vcpu, tr_seg.selector,
+					      &cseg_desc);
+	}
+
+	if (reason == 1) {
+		u32 eflags = kvm_x86_ops->get_rflags(vcpu);
+		kvm_x86_ops->set_rflags(vcpu, eflags &~(1 << 14));
+	}
+
+	kvm_x86_ops->skip_emulated_instruction(vcpu);
+	kvm_x86_ops->cache_regs(vcpu);
+
+	if (nseg_desc.type & 8)
+		ret = kvm_task_switch_32(vcpu, tss_selector, reason,
+					 &cseg_desc, &nseg_desc);
+	else
+		ret = kvm_task_switch_16(vcpu, tss_selector, reason,
+					 &cseg_desc, &nseg_desc);
+
+	if (reason == 0 || reason == 3) {
+		u32 eflags = kvm_x86_ops->get_rflags(vcpu);
+		kvm_x86_ops->set_rflags(vcpu, eflags | (1 << 14));
+	}
+
+	if (reason != 1) {
+		nseg_desc.type &= ~(1 << 8);
+		save_guest_segment_descriptor(vcpu, tss_selector,
+					      &nseg_desc);
+	}
+
+	kvm_x86_ops->set_cr0(vcpu, vcpu->arch.cr0 | 1 << 3);
+	seg_desct_to_kvm_desct(&nseg_desc, tss_selector, &tr_seg);
+	tr_seg.type = 11;
+	set_segment(vcpu, &tr_seg, VCPU_SREG_TR);
+out:
+	kvm_x86_ops->decache_regs(vcpu);
+	if (!ret)
+		kvm_queue_exception(vcpu, GP_VECTOR);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(kvm_task_switch);
+
 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 				  struct kvm_sregs *sregs)
 {
diff --git a/include/asm-x86/kvm_host.h b/include/asm-x86/kvm_host.h
index d9806e2..7ca0462 100644
--- a/include/asm-x86/kvm_host.h
+++ b/include/asm-x86/kvm_host.h
@@ -472,6 +472,8 @@ int emulator_get_dr(struct x86_emulate_ctxt *ctxt, int dr,
 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
 		    unsigned long value);
 
+int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason);
+
 void kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
 void kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
 void kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
-- 
1.5.3.6

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

Re: [PATCH RFC 2/2] Hardware task switching support

[Avi Kivity]

Izik Eidus wrote:
> >From 6a7207a0f3ee8af6ebafcec9d40a75b87f00a129 Mon Sep 17 00:00:00 2001
> From: Izik Eidus <izike@qumranet.com>
> Date: Thu, 13 Mar 2008 02:34:21 +0200
> Subject: [PATCH] KVM: hardware task switching support
>
> Signed-off-by: Izik Eidus <izike@qumranet.com>
> ---
>  arch/x86/kvm/svm.c         |   11 +-
>  arch/x86/kvm/tss_segment.h |   59 +++++++
>  arch/x86/kvm/vmx.c         |   15 ++
>  arch/x86/kvm/x86.c         |  385 ++++++++++++++++++++++++++++++++++++++++++++
>  include/asm-x86/kvm_host.h |    2 +
>  5 files changed, 469 insertions(+), 3 deletions(-)
>  create mode 100644 arch/x86/kvm/tss_segment.h
>
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 4e1dd61..be78278 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -1121,9 +1121,14 @@ static int invalid_op_interception(struct vcpu_svm *svm,
>  static int task_switch_interception(struct vcpu_svm *svm,
>  				    struct kvm_run *kvm_run)
>  {
> -	pr_unimpl(&svm->vcpu, "%s: task switch is unsupported\n", __func__);
> -	kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
> -	return 0;
> +	u16 tss_selector;
> +
> +	tss_selector = (u16)svm->vmcb->control.exit_info_1;
> +	if(svm->vmcb->control.exit_info_2 & ((unsigned long)1 << 36))
> +		return kvm_task_switch(&svm->vcpu, tss_selector, 1);
> +	if(svm->vmcb->control.exit_info_2 & ((unsigned long)1 << 38))
> +		return kvm_task_switch(&svm->vcpu, tss_selector, 2);
> +	return kvm_task_switch(&svm->vcpu, tss_selector, 0);
>   

space after if.  Change the magic numbers (36, 38, 0, 1, 2) inuo 
constants in svm.h and kvm_host.h. (unsigned long)1 << 36 will break on 
i386 (needs 1ULL << 36).

>  }
>  
>  static int cpuid_interception(struct vcpu_svm *svm, struct kvm_run *kvm_run)
> diff --git a/arch/x86/kvm/tss_segment.h b/arch/x86/kvm/tss_segment.h
> new file mode 100644
> index 0000000..622aa10
> --- /dev/null
> +++ b/arch/x86/kvm/tss_segment.h
>   

"tss" already has "segment" in it.  call it just tss.h.

> +
> +/* allowed just for 8 bytes segments */
> +static int load_guest_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector,
> +			      struct desc_struct *seg_desc)
> +{
> +	struct descriptor_table gdt_ldt;
> +	u16 index = selector >> 3;
> +
> +	if (selector & 1 << 2)
> +		kvm_x86_ops->get_ldt(vcpu, &gdt_ldt);
> +	else
> +		kvm_x86_ops->get_gdt(vcpu, &gdt_ldt);
> +	if (gdt_ldt.limit < index * 8 + 7)
> +		return 1;
>   

This part can be in a helper function shared with load_ and 
save_guest_segment_descriptor.  Maybe rename to read_ and write_ to 
avoid confusion.

> +
> +static int load_tss_segment32(struct kvm_vcpu *vcpu,
> +			      struct desc_struct *seg_desc,
> +			      struct tss_segment_32 *tss)
> +{
> +	u32 base_addr;
> +
> +	base_addr = seg_desc->base0;
> +	base_addr |= (seg_desc->base1 << 16);
> +	base_addr |= (seg_desc->base2 << 24);
>   

A helper, again.

> +static void save_state_to_tss32(struct kvm_vcpu *vcpu,
> +				struct tss_segment_32 *tss)
> +{
> +	struct kvm_segment kvm_seg;
> +
> +	tss->cr3 = vcpu->arch.cr3;
> +	tss->eip = vcpu->arch.rip;
> +	tss->eflags = kvm_x86_ops->get_rflags(vcpu);
> +	tss->eax = vcpu->arch.regs[VCPU_REGS_RAX];
> +	tss->ecx = vcpu->arch.regs[VCPU_REGS_RCX];
> +	tss->edx = vcpu->arch.regs[VCPU_REGS_RDX];
> +	tss->ebx = vcpu->arch.regs[VCPU_REGS_RBX];
> +	tss->esp = vcpu->arch.regs[VCPU_REGS_RSP];
> +	tss->ebp = vcpu->arch.regs[VCPU_REGS_RBP];
> +	tss->esi = vcpu->arch.regs[VCPU_REGS_RSI];
> +	tss->edi = vcpu->arch.regs[VCPU_REGS_RDI];
> +	
> +	get_segment(vcpu, &kvm_seg, VCPU_SREG_ES);
> +	tss->es = kvm_seg.selector;
>   

tss->es = get_segment_selector(vcpu, VCPU_SREG_ES);

> +	load_guest_segment_descriptor(vcpu, tss->ldt_selector,
> +				      &seg_desc);
> +	seg_desct_to_kvm_desct(&seg_desc, tss->ldt_selector,
> +			       &kvm_seg);
> +	set_segment(vcpu, &kvm_seg, VCPU_SREG_LDTR);
> +
> +	load_guest_segment_descriptor(vcpu, tss->es, &seg_desc);
> +	seg_desct_to_kvm_desct(&seg_desc, tss->es, &kvm_seg);
> +	kvm_seg.type |= 1;
> +	if (!kvm_seg.s)
> +		kvm_seg.unusable = 1;
> +	set_segment(vcpu, &kvm_seg, VCPU_SREG_ES);
>   

Wrap these into a helper.


-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/