KVM: PPC: Book3S: Controls for in-kernel sPAPR hypercall handling

This provides a way for userspace controls which sPAPR hcalls get handled in the kernel. Each hcall can be individually enabled or disabled for in-kernel handling, except for H_RTAS. The exception for H_RTAS is because userspace can already control whether individual RTAS functions are handled in-kernel or not via the KVM_PPC_RTAS_DEFINE_TOKEN ioctl, and because the numeric value for H_RTAS is out of the normal sequence of hcall numbers. Hcalls are enabled or disabled using the KVM_ENABLE_CAP ioctl for the KVM_CAP_PPC_ENABLE_HCALL capability on the file descriptor for the VM. The args field of the struct kvm_enable_cap specifies the hcall number in args[0] and the enable/disable flag in args[1]; 0 means disable in-kernel handling (so that the hcall will always cause an exit to userspace) and 1 means enable. Enabling or disabling in-kernel handling of an hcall is effective across the whole VM. The ability for KVM_ENABLE_CAP to be used on a VM file descriptor on PowerPC is new, added by this commit. The KVM_CAP_ENABLE_CAP_VM capability advertises that this ability exists. When a VM is created, an initial set of hcalls are enabled for in-kernel handling. The set that is enabled is the set that have an in-kernel implementation at this point. Any new hcall implementations from this point onwards should not be added to the default set without a good reason. No distinction is made between real-mode and virtual-mode hcall implementations; the one setting controls them both.

Paul Mackerras committed 12 years ago

Commit 699a0ea1 parent 1f0eeb710 files · +193 −2Browse files

Unified · Split

10 files

Documentation/virtual/kvm/api.txt +39 −2
arch/powerpc/include/asm/kvm_book3s.h +1 −0
arch/powerpc/include/asm/kvm_host.h +2 −0
arch/powerpc/kernel/asm-offsets.c +1 −0
arch/powerpc/kvm/book3s_hv.c +51 −0
arch/powerpc/kvm/book3s_hv_rmhandlers.S +11 −0
arch/powerpc/kvm/book3s_pr.c +5 −0
arch/powerpc/kvm/book3s_pr_papr.c +37 −0
arch/powerpc/kvm/powerpc.c +45 −0
include/uapi/linux/kvm.h +1 −0

Documentation/virtual/kvm/api.txt+39−2

@@ -2863,8 +2863,8 @@ The fields in each entry are defined as follows: · 2862 unmodified lines
2863	2863	this function/index combination
2864	2864
2865	2865
2866		−6. Capabilities that can be enabled
2867		−-----------------------------------
	2866	+6. Capabilities that can be enabled on vCPUs
	2867	+--------------------------------------------
2868	2868
2869	2869	There are certain capabilities that change the behavior of the virtual CPU when
2870	2870	enabled. To enable them, please see section 4.37. Below you can find a list of
@@ -3002,3 +3002,40 @@ Parameters: args[0] is the XICS device fd · 131 unmodified lines
3002	3002	args[1] is the XICS CPU number (server ID) for this vcpu
3003	3003
3004	3004	This capability connects the vcpu to an in-kernel XICS device.
	3005	+
	3006	+
	3007	+7. Capabilities that can be enabled on VMs
	3008	+------------------------------------------
	3009	+
	3010	+There are certain capabilities that change the behavior of the virtual
	3011	+machine when enabled. To enable them, please see section 4.37. Below
	3012	+you can find a list of capabilities and what their effect on the VM
	3013	+is when enabling them.
	3014	+
	3015	+The following information is provided along with the description:
	3016	+
	3017	+ Architectures: which instruction set architectures provide this ioctl.
	3018	+ x86 includes both i386 and x86_64.
	3019	+
	3020	+ Parameters: what parameters are accepted by the capability.
	3021	+
	3022	+ Returns: the return value. General error numbers (EBADF, ENOMEM, EINVAL)
	3023	+ are not detailed, but errors with specific meanings are.
	3024	+
	3025	+
	3026	+7.1 KVM_CAP_PPC_ENABLE_HCALL
	3027	+
	3028	+Architectures: ppc
	3029	+Parameters: args[0] is the sPAPR hcall number
	3030	+ args[1] is 0 to disable, 1 to enable in-kernel handling
	3031	+
	3032	+This capability controls whether individual sPAPR hypercalls (hcalls)
	3033	+get handled by the kernel or not. Enabling or disabling in-kernel
	3034	+handling of an hcall is effective across the VM. On creation, an
	3035	+initial set of hcalls are enabled for in-kernel handling, which
	3036	+consists of those hcalls for which in-kernel handlers were implemented
	3037	+before this capability was implemented. If disabled, the kernel will
	3038	+not to attempt to handle the hcall, but will always exit to userspace
	3039	+to handle it. Note that it may not make sense to enable some and
	3040	+disable others of a group of related hcalls, but KVM does not prevent
	3041	+userspace from doing that.

arch/powerpc/include/asm/kvm_book3s.h+1−0

@@ -187,6 +187,7 @@ extern void kvmppc_hv_entry_trampoline(void); · 186 unmodified lines
187	187	extern u32 kvmppc_alignment_dsisr(struct kvm_vcpu *vcpu, unsigned int inst);
188	188	extern ulong kvmppc_alignment_dar(struct kvm_vcpu *vcpu, unsigned int inst);
189	189	extern int kvmppc_h_pr(struct kvm_vcpu *vcpu, unsigned long cmd);
	190	+extern void kvmppc_pr_init_default_hcalls(struct kvm *kvm);
190	191	extern void kvmppc_copy_to_svcpu(struct kvmppc_book3s_shadow_vcpu *svcpu,
191	192	struct kvm_vcpu *vcpu);
192	193	extern void kvmppc_copy_from_svcpu(struct kvm_vcpu *vcpu,

arch/powerpc/include/asm/kvm_host.h+2−0

@@ -34,6 +34,7 @@ · 33 unmodified lines
34	34	#include <asm/processor.h>
35	35	#include <asm/page.h>
36	36	#include <asm/cacheflush.h>
	37	+#include <asm/hvcall.h>
37	38
38	39	#define KVM_MAX_VCPUS NR_CPUS
39	40	#define KVM_MAX_VCORES NR_CPUS
@@ -263,6 +264,7 @@ struct kvm_arch { · 223 unmodified lines
263	264	#ifdef CONFIG_PPC_BOOK3S_64
264	265	struct list_head spapr_tce_tables;
265	266	struct list_head rtas_tokens;
	267	+ DECLARE_BITMAP(enabled_hcalls, MAX_HCALL_OPCODE/4 + 1);
266	268	#endif
267	269	#ifdef CONFIG_KVM_MPIC
268	270	struct openpic *mpic;

arch/powerpc/kernel/asm-offsets.c+1−0

@@ -493,6 +493,7 @@ int main(void) · 492 unmodified lines
493	493	DEFINE(KVM_HOST_SDR1, offsetof(struct kvm, arch.host_sdr1));
494	494	DEFINE(KVM_TLBIE_LOCK, offsetof(struct kvm, arch.tlbie_lock));
495	495	DEFINE(KVM_NEED_FLUSH, offsetof(struct kvm, arch.need_tlb_flush.bits));
	496	+ DEFINE(KVM_ENABLED_HCALLS, offsetof(struct kvm, arch.enabled_hcalls));
496	497	DEFINE(KVM_LPCR, offsetof(struct kvm, arch.lpcr));
497	498	DEFINE(KVM_RMOR, offsetof(struct kvm, arch.rmor));
498	499	DEFINE(KVM_VRMA_SLB_V, offsetof(struct kvm, arch.vrma_slb_v));

arch/powerpc/kvm/book3s_hv.c+51−0

@@ -67,6 +67,8 @@ · 66 unmodified lines
67	67	/* Used as a "null" value for timebase values */
68	68	#define TB_NIL (~(u64)0)
69	69
	70	+static DECLARE_BITMAP(default_enabled_hcalls, MAX_HCALL_OPCODE/4 + 1);
	71	+
70	72	static void kvmppc_end_cede(struct kvm_vcpu *vcpu);
71	73	static int kvmppc_hv_setup_htab_rma(struct kvm_vcpu *vcpu);
72	74
@@ -562,6 +564,10 @@ int kvmppc_pseries_do_hcall(struct kvm_vcpu *vcpu) · 489 unmodified lines
562	564	struct kvm_vcpu *tvcpu;
563	565	int idx, rc;
564	566
	567	+ if (req <= MAX_HCALL_OPCODE &&
	568	+ !test_bit(req/4, vcpu->kvm->arch.enabled_hcalls))
	569	+ return RESUME_HOST;
	570	+
565	571	switch (req) {
566	572	case H_ENTER:
567	573	idx = srcu_read_lock(&vcpu->kvm->srcu);
@@ -2269,6 +2275,10 @@ static int kvmppc_core_init_vm_hv(struct kvm *kvm) · 1701 unmodified lines
2269	2275	*/
2270	2276	cpumask_setall(&kvm->arch.need_tlb_flush);
2271	2277
	2278	+ /* Start out with the default set of hcalls enabled */
	2279	+ memcpy(kvm->arch.enabled_hcalls, default_enabled_hcalls,
	2280	+ sizeof(kvm->arch.enabled_hcalls));
	2281	+
2272	2282	kvm->arch.rma = NULL;
2273	2283
2274	2284	kvm->arch.host_sdr1 = mfspr(SPRN_SDR1);
@@ -2407,6 +2417,45 @@ static long kvm_arch_vm_ioctl_hv(struct file *filp, · 132 unmodified lines
2407	2417	return r;
2408	2418	}
2409	2419
	2420	+/*
	2421	+ * List of hcall numbers to enable by default.
	2422	+ * For compatibility with old userspace, we enable by default
	2423	+ * all hcalls that were implemented before the hcall-enabling
	2424	+ * facility was added. Note this list should not include H_RTAS.
	2425	+ */
	2426	+static unsigned int default_hcall_list[] = {
	2427	+ H_REMOVE,
	2428	+ H_ENTER,
	2429	+ H_READ,
	2430	+ H_PROTECT,
	2431	+ H_BULK_REMOVE,
	2432	+ H_GET_TCE,
	2433	+ H_PUT_TCE,
	2434	+ H_SET_DABR,
	2435	+ H_SET_XDABR,
	2436	+ H_CEDE,
	2437	+ H_PROD,
	2438	+ H_CONFER,
	2439	+ H_REGISTER_VPA,
	2440	+#ifdef CONFIG_KVM_XICS
	2441	+ H_EOI,
	2442	+ H_CPPR,
	2443	+ H_IPI,
	2444	+ H_IPOLL,
	2445	+ H_XIRR,
	2446	+ H_XIRR_X,
	2447	+#endif
	2448	+ 0
	2449	+};
	2450	+
	2451	+static void init_default_hcalls(void)
	2452	+{
	2453	+ int i;
	2454	+
	2455	+ for (i = 0; default_hcall_list[i]; ++i)
	2456	+ __set_bit(default_hcall_list[i] / 4, default_enabled_hcalls);
	2457	+}
	2458	+
2410	2459	static struct kvmppc_ops kvm_ops_hv = {
2411	2460	.get_sregs = kvm_arch_vcpu_ioctl_get_sregs_hv,
2412	2461	.set_sregs = kvm_arch_vcpu_ioctl_set_sregs_hv,
@@ -2454,6 +2503,8 @@ static int kvmppc_book3s_init_hv(void) · 41 unmodified lines
2454	2503	kvm_ops_hv.owner = THIS_MODULE;
2455	2504	kvmppc_hv_ops = &kvm_ops_hv;
2456	2505
	2506	+ init_default_hcalls();
	2507	+
2457	2508	r = kvmppc_mmu_hv_init();
2458	2509	return r;
2459	2510	}

arch/powerpc/kvm/book3s_hv_rmhandlers.S+11−0

@@ -1909,6 +1909,17 @@ hcall_try_real_mode: · 1908 unmodified lines
1909	1909	clrrdi r3,r3,2
1910	1910	cmpldi r3,hcall_real_table_end - hcall_real_table
1911	1911	bge guest_exit_cont
	1912	+ /* See if this hcall is enabled for in-kernel handling */
	1913	+ ld r4, VCPU_KVM(r9)
	1914	+ srdi r0, r3, 8 /* r0 = (r3 / 4) >> 6 */
	1915	+ sldi r0, r0, 3 /* index into kvm->arch.enabled_hcalls[] */
	1916	+ add r4, r4, r0
	1917	+ ld r0, KVM_ENABLED_HCALLS(r4)
	1918	+ rlwinm r4, r3, 32-2, 0x3f /* r4 = (r3 / 4) & 0x3f */
	1919	+ srd r0, r0, r4
	1920	+ andi. r0, r0, 1
	1921	+ beq guest_exit_cont
	1922	+ /* Get pointer to handler, if any, and call it */
1912	1923	LOAD_REG_ADDR(r4, hcall_real_table)
1913	1924	lwax r3,r3,r4
1914	1925	cmpwi r3,0

arch/powerpc/kvm/book3s_pr.c+5−0

@@ -1597,6 +1597,11 @@ static int kvmppc_core_init_vm_pr(struct kvm *kvm) · 1596 unmodified lines
1597	1597	{
1598	1598	mutex_init(&kvm->arch.hpt_mutex);
1599	1599
	1600	+#ifdef CONFIG_PPC_BOOK3S_64
	1601	+ /* Start out with the default set of hcalls enabled */
	1602	+ kvmppc_pr_init_default_hcalls(kvm);
	1603	+#endif
	1604	+
1600	1605	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {
1601	1606	spin_lock(&kvm_global_user_count_lock);
1602	1607	if (++kvm_global_user_count == 1)

arch/powerpc/kvm/book3s_pr_papr.c+37−0

@@ -267,6 +267,10 @@ static int kvmppc_h_pr_xics_hcall(struct kvm_vcpu *vcpu, u32 cmd) · 266 unmodified lines
267	267
268	268	int kvmppc_h_pr(struct kvm_vcpu *vcpu, unsigned long cmd)
269	269	{
	270	+ if (cmd <= MAX_HCALL_OPCODE &&
	271	+ !test_bit(cmd/4, vcpu->kvm->arch.enabled_hcalls))
	272	+ return EMULATE_FAIL;
	273	+
270	274	switch (cmd) {
271	275	case H_ENTER:
272	276	return kvmppc_h_pr_enter(vcpu);
@@ -304,3 +308,36 @@ int kvmppc_h_pr(struct kvm_vcpu *vcpu, unsigned long cmd) · 31 unmodified lines
304	308
305	309	return EMULATE_FAIL;
306	310	}
	311	+
	312	+
	313	+/*
	314	+ * List of hcall numbers to enable by default.
	315	+ * For compatibility with old userspace, we enable by default
	316	+ * all hcalls that were implemented before the hcall-enabling
	317	+ * facility was added. Note this list should not include H_RTAS.
	318	+ */
	319	+static unsigned int default_hcall_list[] = {
	320	+ H_ENTER,
	321	+ H_REMOVE,
	322	+ H_PROTECT,
	323	+ H_BULK_REMOVE,
	324	+ H_PUT_TCE,
	325	+ H_CEDE,
	326	+#ifdef CONFIG_KVM_XICS
	327	+ H_XIRR,
	328	+ H_CPPR,
	329	+ H_EOI,
	330	+ H_IPI,
	331	+ H_IPOLL,
	332	+ H_XIRR_X,
	333	+#endif
	334	+ 0
	335	+};
	336	+
	337	+void kvmppc_pr_init_default_hcalls(struct kvm *kvm)
	338	+{
	339	+ int i;
	340	+
	341	+ for (i = 0; default_hcall_list[i]; ++i)
	342	+ __set_bit(default_hcall_list[i] / 4, kvm->arch.enabled_hcalls);
	343	+}

arch/powerpc/kvm/powerpc.c+45−0

@@ -387,6 +387,7 @@ int kvm_dev_ioctl_check_extension(long ext) · 386 unmodified lines
387	387	case KVM_CAP_PPC_UNSET_IRQ:
388	388	case KVM_CAP_PPC_IRQ_LEVEL:
389	389	case KVM_CAP_ENABLE_CAP:
	390	+ case KVM_CAP_ENABLE_CAP_VM:
390	391	case KVM_CAP_ONE_REG:
391	392	case KVM_CAP_IOEVENTFD:
392	393	case KVM_CAP_DEVICE_CTRL:
@@ -417,6 +418,7 @@ int kvm_dev_ioctl_check_extension(long ext) · 24 unmodified lines
417	418	case KVM_CAP_PPC_ALLOC_HTAB:
418	419	case KVM_CAP_PPC_RTAS:
419	420	case KVM_CAP_PPC_FIXUP_HCALL:
	421	+ case KVM_CAP_PPC_ENABLE_HCALL:
420	422	#ifdef CONFIG_KVM_XICS
421	423	case KVM_CAP_IRQ_XICS:
422	424	#endif
@@ -1099,6 +1101,40 @@ int kvm_vm_ioctl_irq_line(struct kvm kvm, struct kvm_irq_level irq_event, · 676 unmodified lines
1099	1101	return 0;
1100	1102	}
1101	1103
	1104	+
	1105	+static int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
	1106	+ struct kvm_enable_cap *cap)
	1107	+{
	1108	+ int r;
	1109	+
	1110	+ if (cap->flags)
	1111	+ return -EINVAL;
	1112	+
	1113	+ switch (cap->cap) {
	1114	+#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
	1115	+ case KVM_CAP_PPC_ENABLE_HCALL: {
	1116	+ unsigned long hcall = cap->args[0];
	1117	+
	1118	+ r = -EINVAL;
	1119	+ if (hcall > MAX_HCALL_OPCODE \|\| (hcall & 3) \|\|
	1120	+ cap->args[1] > 1)
	1121	+ break;
	1122	+ if (cap->args[1])
	1123	+ set_bit(hcall / 4, kvm->arch.enabled_hcalls);
	1124	+ else
	1125	+ clear_bit(hcall / 4, kvm->arch.enabled_hcalls);
	1126	+ r = 0;
	1127	+ break;
	1128	+ }
	1129	+#endif
	1130	+ default:
	1131	+ r = -EINVAL;
	1132	+ break;
	1133	+ }
	1134	+
	1135	+ return r;
	1136	+}
	1137	+
1102	1138	long kvm_arch_vm_ioctl(struct file *filp,
1103	1139	unsigned int ioctl, unsigned long arg)
1104	1140	{
@@ -1118,6 +1154,15 @@ long kvm_arch_vm_ioctl(struct file *filp, · 13 unmodified lines
1118	1154
1119	1155	break;
1120	1156	}
	1157	+ case KVM_ENABLE_CAP:
	1158	+ {
	1159	+ struct kvm_enable_cap cap;
	1160	+ r = -EFAULT;
	1161	+ if (copy_from_user(&cap, argp, sizeof(cap)))
	1162	+ goto out;
	1163	+ r = kvm_vm_ioctl_enable_cap(kvm, &cap);
	1164	+ break;
	1165	+ }
1121	1166	#ifdef CONFIG_PPC_BOOK3S_64
1122	1167	case KVM_CREATE_SPAPR_TCE: {
1123	1168	struct kvm_create_spapr_tce create_tce;

include/uapi/linux/kvm.h+1−0

@@ -758,6 +758,7 @@ struct kvm_ppc_smmu_info { · 757 unmodified lines
758	758	#define KVM_CAP_VM_ATTRIBUTES 101
759	759	#define KVM_CAP_ARM_PSCI_0_2 102
760	760	#define KVM_CAP_PPC_FIXUP_HCALL 103
	761	+#define KVM_CAP_PPC_ENABLE_HCALL 104
761	762
762	763	#ifdef KVM_CAP_IRQ_ROUTING
763	764

KVM: PPC: Book3S: Controls for in-kernel sPAPR hypercall handling

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code