KVM: PPC: Add OSI hypercall interface

MOL uses its own hypercall interface to call back into userspace when the guest wants to do something. So let's implement that as an exit reason, specify it with a CAP and only really use it when userspace wants us to. The only user of it so far is MOL. Signed-off-by: Alexander Graf <agraf@suse.de> Signed-off-by: Avi Kivity <avi@redhat.com>

KVM: PPC: Add OSI hypercall interface
MOL uses its own hypercall interface to call back into userspace when the guest wants to do something. So let's implement that as an exit reason, specify it with a CAP and only really use it when userspace wants us to. The only user of it so far is MOL. Signed-off-by: Alexander Graf <agraf@suse.de> Signed-off-by: Avi Kivity <avi@redhat.com>
Alexander Graf · Avi Kivity
1 parent 71fbfd5f38
Showing 6 changed files with 59 additions and 9 deletions Side-by-side Diff
Documentation/kvm/api.txt
arch/powerpc/include/asm/kvm_book3s.h
arch/powerpc/include/asm/kvm_host.h
arch/powerpc/kvm/book3s.c
arch/powerpc/kvm/powerpc.c
include/linux/kvm.h
@@ -958,9 +958,9 @@
 by kvm.  The 'data' member contains the written data if 'is_write' is
 true, and should be filled by application code otherwise.
  
-NOTE: For KVM_EXIT_IO and KVM_EXIT_MMIO, the corresponding operations
-are complete (and guest state is consistent) only after userspace has
-re-entered the kernel with KVM_RUN.  The kernel side will first finish
+NOTE: For KVM_EXIT_IO, KVM_EXIT_MMIO and KVM_EXIT_OSI, the corresponding
+operations are complete (and guest state is consistent) only after userspace
+has re-entered the kernel with KVM_RUN.  The kernel side will first finish
 incomplete operations and then check for pending signals.  Userspace
 can re-enter the guest with an unmasked signal pending to complete
 pending operations.
@@ -1014,6 +1014,19 @@
 		} dcr;
  
 powerpc specific.
+
+		/* KVM_EXIT_OSI */
+		struct {
+			__u64 gprs[32];
+		} osi;
+
+MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch
+hypercalls and exit with this exit struct that contains all the guest gprs.
+
+If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall.
+Userspace can now handle the hypercall and when it's done modify the gprs as
+necessary. Upon guest entry all guest GPRs will then be replaced by the values
+in this struct.
  
 		/* Fix the size of the union. */
 		char padding[256];
@@ -148,6 +148,11 @@
  
 extern void kvm_return_point(void);
  
+/* Magic register values loaded into r3 and r4 before the 'sc' assembly
+ * instruction for the OSI hypercalls */
+#define OSI_SC_MAGIC_R3			0x113724FA
+#define OSI_SC_MAGIC_R4			0x77810F9B
+
 #define INS_DCBZ			0x7c0007ec
  
 #endif /* __ASM_KVM_BOOK3S_H__ */
@@ -273,6 +273,8 @@
 	u8 mmio_sign_extend;
 	u8 dcr_needed;
 	u8 dcr_is_write;
+	u8 osi_needed;
+	u8 osi_enabled;
  
 	u32 cpr0_cfgaddr; /* holds the last set cpr0_cfgaddr */
  
@@ -872,12 +872,24 @@
 		break;
 	}
 	case BOOK3S_INTERRUPT_SYSCALL:
-#ifdef EXIT_DEBUG
-		printk(KERN_INFO "Syscall Nr %d\n", (int)kvmppc_get_gpr(vcpu, 0));
-#endif
-		vcpu->stat.syscall_exits++;
-		kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
-		r = RESUME_GUEST;
+		// XXX make user settable
+		if (vcpu->arch.osi_enabled &&
+		    (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
+		    (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
+			u64 *gprs = run->osi.gprs;
+			int i;
+
+			run->exit_reason = KVM_EXIT_OSI;
+			for (i = 0; i < 32; i++)
+				gprs[i] = kvmppc_get_gpr(vcpu, i);
+			vcpu->arch.osi_needed = 1;
+			r = RESUME_HOST_NV;
+
+		} else {
+			vcpu->stat.syscall_exits++;
+			kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
+			r = RESUME_GUEST;
+		}
 		break;
 	case BOOK3S_INTERRUPT_FP_UNAVAIL:
 	case BOOK3S_INTERRUPT_ALTIVEC:
@@ -151,6 +151,7 @@
 	case KVM_CAP_PPC_PAIRED_SINGLES:
 	case KVM_CAP_PPC_UNSET_IRQ:
 	case KVM_CAP_ENABLE_CAP:
+	case KVM_CAP_PPC_OSI:
 		r = 1;
 		break;
 	case KVM_CAP_COALESCED_MMIO:
@@ -433,6 +434,13 @@
 		if (!vcpu->arch.dcr_is_write)
 			kvmppc_complete_dcr_load(vcpu, run);
 		vcpu->arch.dcr_needed = 0;
+	} else if (vcpu->arch.osi_needed) {
+		u64 *gprs = run->osi.gprs;
+		int i;
+
+		for (i = 0; i < 32; i++)
+			kvmppc_set_gpr(vcpu, i, gprs[i]);
+		vcpu->arch.osi_needed = 0;
 	}
  
 	kvmppc_core_deliver_interrupts(vcpu);
@@ -475,6 +483,10 @@
 		return -EINVAL;
  
 	switch (cap->cap) {
+	case KVM_CAP_PPC_OSI:
+		r = 0;
+		vcpu->arch.osi_enabled = true;
+		break;
 	default:
 		r = -EINVAL;
 		break;
@@ -160,6 +160,7 @@
 #define KVM_EXIT_DCR              15
 #define KVM_EXIT_NMI              16
 #define KVM_EXIT_INTERNAL_ERROR   17
+#define KVM_EXIT_OSI              18
  
 /* For KVM_EXIT_INTERNAL_ERROR */
 #define KVM_INTERNAL_ERROR_EMULATION 1
@@ -259,6 +260,10 @@
 			__u32 ndata;
 			__u64 data[16];
 		} internal;
+		/* KVM_EXIT_OSI */
+		struct {
+			__u64 gprs[32];
+		} osi;
 		/* Fix the size of the union. */
 		char padding[256];
 	};
@@ -516,6 +521,7 @@
 #define KVM_CAP_DEBUGREGS 50
 #endif
 #define KVM_CAP_X86_ROBUST_SINGLESTEP 51
+#define KVM_CAP_PPC_OSI 52
 #define KVM_CAP_PPC_UNSET_IRQ 53
 #define KVM_CAP_ENABLE_CAP 54
...	...	@@ -958,9 +958,9 @@
958	958	by kvm. The 'data' member contains the written data if 'is_write' is
959	959	true, and should be filled by application code otherwise.
960	960
961		-NOTE: For KVM_EXIT_IO and KVM_EXIT_MMIO, the corresponding operations
962		-are complete (and guest state is consistent) only after userspace has
963		-re-entered the kernel with KVM_RUN. The kernel side will first finish
	961	+NOTE: For KVM_EXIT_IO, KVM_EXIT_MMIO and KVM_EXIT_OSI, the corresponding
	962	+operations are complete (and guest state is consistent) only after userspace
	963	+has re-entered the kernel with KVM_RUN. The kernel side will first finish
964	964	incomplete operations and then check for pending signals. Userspace
965	965	can re-enter the guest with an unmasked signal pending to complete
966	966	pending operations.
...	...	@@ -1014,6 +1014,19 @@
1014	1014	} dcr;
1015	1015
1016	1016	powerpc specific.
	1017	+
	1018	+ /* KVM_EXIT_OSI */
	1019	+ struct {
	1020	+ __u64 gprs[32];
	1021	+ } osi;
	1022	+
	1023	+MOL uses a special hypercall interface it calls 'OSI'. To enable it, we catch
	1024	+hypercalls and exit with this exit struct that contains all the guest gprs.
	1025	+
	1026	+If exit_reason is KVM_EXIT_OSI, then the vcpu has triggered such a hypercall.
	1027	+Userspace can now handle the hypercall and when it's done modify the gprs as
	1028	+necessary. Upon guest entry all guest GPRs will then be replaced by the values
	1029	+in this struct.
1017	1030
1018	1031	/* Fix the size of the union. */
1019	1032	char padding[256];
...	...	@@ -148,6 +148,11 @@
148	148
149	149	extern void kvm_return_point(void);
150	150
	151	+/* Magic register values loaded into r3 and r4 before the 'sc' assembly
	152	+ * instruction for the OSI hypercalls */
	153	+#define OSI_SC_MAGIC_R3 0x113724FA
	154	+#define OSI_SC_MAGIC_R4 0x77810F9B
	155	+
151	156	#define INS_DCBZ 0x7c0007ec
152	157
153	158	#endif /* __ASM_KVM_BOOK3S_H__ */
...	...	@@ -273,6 +273,8 @@
273	273	u8 mmio_sign_extend;
274	274	u8 dcr_needed;
275	275	u8 dcr_is_write;
	276	+ u8 osi_needed;
	277	+ u8 osi_enabled;
276	278
277	279	u32 cpr0_cfgaddr; /* holds the last set cpr0_cfgaddr */
278	280
...	...	@@ -872,12 +872,24 @@
872	872	break;
873	873	}
874	874	case BOOK3S_INTERRUPT_SYSCALL:
875		-#ifdef EXIT_DEBUG
876		- printk(KERN_INFO "Syscall Nr %d\n", (int)kvmppc_get_gpr(vcpu, 0));
877		-#endif
878		- vcpu->stat.syscall_exits++;
879		- kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
880		- r = RESUME_GUEST;
	875	+ // XXX make user settable
	876	+ if (vcpu->arch.osi_enabled &&
	877	+ (((u32)kvmppc_get_gpr(vcpu, 3)) == OSI_SC_MAGIC_R3) &&
	878	+ (((u32)kvmppc_get_gpr(vcpu, 4)) == OSI_SC_MAGIC_R4)) {
	879	+ u64 *gprs = run->osi.gprs;
	880	+ int i;
	881	+
	882	+ run->exit_reason = KVM_EXIT_OSI;
	883	+ for (i = 0; i < 32; i++)
	884	+ gprs[i] = kvmppc_get_gpr(vcpu, i);
	885	+ vcpu->arch.osi_needed = 1;
	886	+ r = RESUME_HOST_NV;
	887	+
	888	+ } else {
	889	+ vcpu->stat.syscall_exits++;
	890	+ kvmppc_book3s_queue_irqprio(vcpu, exit_nr);
	891	+ r = RESUME_GUEST;
	892	+ }
881	893	break;
882	894	case BOOK3S_INTERRUPT_FP_UNAVAIL:
883	895	case BOOK3S_INTERRUPT_ALTIVEC:
...	...	@@ -151,6 +151,7 @@
151	151	case KVM_CAP_PPC_PAIRED_SINGLES:
152	152	case KVM_CAP_PPC_UNSET_IRQ:
153	153	case KVM_CAP_ENABLE_CAP:
	154	+ case KVM_CAP_PPC_OSI:
154	155	r = 1;
155	156	break;
156	157	case KVM_CAP_COALESCED_MMIO:
...	...	@@ -433,6 +434,13 @@
433	434	if (!vcpu->arch.dcr_is_write)
434	435	kvmppc_complete_dcr_load(vcpu, run);
435	436	vcpu->arch.dcr_needed = 0;
	437	+ } else if (vcpu->arch.osi_needed) {
	438	+ u64 *gprs = run->osi.gprs;
	439	+ int i;
	440	+
	441	+ for (i = 0; i < 32; i++)
	442	+ kvmppc_set_gpr(vcpu, i, gprs[i]);
	443	+ vcpu->arch.osi_needed = 0;
436	444	}
437	445
438	446	kvmppc_core_deliver_interrupts(vcpu);
...	...	@@ -475,6 +483,10 @@
475	483	return -EINVAL;
476	484
477	485	switch (cap->cap) {
	486	+ case KVM_CAP_PPC_OSI:
	487	+ r = 0;
	488	+ vcpu->arch.osi_enabled = true;
	489	+ break;
478	490	default:
479	491	r = -EINVAL;
480	492	break;
...	...	@@ -160,6 +160,7 @@
160	160	#define KVM_EXIT_DCR 15
161	161	#define KVM_EXIT_NMI 16
162	162	#define KVM_EXIT_INTERNAL_ERROR 17
	163	+#define KVM_EXIT_OSI 18
163	164
164	165	/* For KVM_EXIT_INTERNAL_ERROR */
165	166	#define KVM_INTERNAL_ERROR_EMULATION 1
...	...	@@ -259,6 +260,10 @@
259	260	__u32 ndata;
260	261	__u64 data[16];
261	262	} internal;
	263	+ /* KVM_EXIT_OSI */
	264	+ struct {
	265	+ __u64 gprs[32];
	266	+ } osi;
262	267	/* Fix the size of the union. */
263	268	char padding[256];
264	269	};
...	...	@@ -516,6 +521,7 @@
516	521	#define KVM_CAP_DEBUGREGS 50
517	522	#endif
518	523	#define KVM_CAP_X86_ROBUST_SINGLESTEP 51
	524	+#define KVM_CAP_PPC_OSI 52
519	525	#define KVM_CAP_PPC_UNSET_IRQ 53
520	526	#define KVM_CAP_ENABLE_CAP 54
521	527