KVM: PPC: Book3S: Add infrastructure to implement kernel-side RTAS calls

For pseries machine emulation, in order to move the interrupt controller code to the kernel, we need to intercept some RTAS calls in the kernel itself. This adds an infrastructure to allow in-kernel handlers to be registered for RTAS services by name. A new ioctl, KVM_PPC_RTAS_DEFINE_TOKEN, then allows userspace to associate token values with those service names. Then, when the guest requests an RTAS service with one of those token values, it will be handled by the relevant in-kernel handler rather than being passed up to userspace as at present.

Michael Ellerman committed 13 years ago

Commit 8e591cb1 parent 911949112 files · +252 −1Browse files

Unified · Split

12 files

Documentation/virtual/kvm/api.txt +19 −0
arch/powerpc/include/asm/hvcall.h +3 −0
arch/powerpc/include/asm/kvm_host.h +1 −0
arch/powerpc/include/asm/kvm_ppc.h +4 −0
arch/powerpc/include/uapi/asm/kvm.h +6 −0
arch/powerpc/kvm/Makefile +1 −0
arch/powerpc/kvm/book3s_hv.c +17 −1
arch/powerpc/kvm/book3s_pr.c +1 −0
arch/powerpc/kvm/book3s_pr_papr.c +7 −0
arch/powerpc/kvm/book3s_rtas.c +182 −0
arch/powerpc/kvm/powerpc.c +8 −0
include/uapi/linux/kvm.h +3 −0

Documentation/virtual/kvm/api.txt+19−0

@@ -2341,6 +2341,25 @@ and distributor interface, the ioctl must be called after calling · 2340 unmodified lines
2341	2341	KVM_CREATE_IRQCHIP, but before calling KVM_RUN on any of the VCPUs. Calling
2342	2342	this ioctl twice for any of the base addresses will return -EEXIST.
2343	2343
	2344	+4.82 KVM_PPC_RTAS_DEFINE_TOKEN
	2345	+
	2346	+Capability: KVM_CAP_PPC_RTAS
	2347	+Architectures: ppc
	2348	+Type: vm ioctl
	2349	+Parameters: struct kvm_rtas_token_args
	2350	+Returns: 0 on success, -1 on error
	2351	+
	2352	+Defines a token value for a RTAS (Run Time Abstraction Services)
	2353	+service in order to allow it to be handled in the kernel. The
	2354	+argument struct gives the name of the service, which must be the name
	2355	+of a service that has a kernel-side implementation. If the token
	2356	+value is non-zero, it will be associated with that service, and
	2357	+subsequent RTAS calls by the guest specifying that token will be
	2358	+handled by the kernel. If the token value is 0, then any token
	2359	+associated with the service will be forgotten, and subsequent RTAS
	2360	+calls by the guest for that service will be passed to userspace to be
	2361	+handled.
	2362	+
2344	2363
2345	2364	5. The kvm_run structure
2346	2365	------------------------

arch/powerpc/include/asm/hvcall.h+3−0

@@ -270,6 +270,9 @@ · 269 unmodified lines
270	270	#define H_SET_MODE 0x31C
271	271	#define MAX_HCALL_OPCODE H_SET_MODE
272	272
	273	+/* Platform specific hcalls, used by KVM */
	274	+#define H_RTAS 0xf000
	275	+
273	276	#ifndef __ASSEMBLY__
274	277
275	278	/**

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

@@ -259,6 +259,7 @@ struct kvm_arch { · 258 unmodified lines
259	259	#endif /* CONFIG_KVM_BOOK3S_64_HV */
260	260	#ifdef CONFIG_PPC_BOOK3S_64
261	261	struct list_head spapr_tce_tables;
	262	+ struct list_head rtas_tokens;
262	263	#endif
263	264	#ifdef CONFIG_KVM_MPIC
264	265	struct openpic *mpic;

arch/powerpc/include/asm/kvm_ppc.h+4−0

@@ -166,6 +166,10 @@ extern int kvm_vm_ioctl_get_htab_fd(struct kvm kvm, struct kvm_get_htab_fd ); · 165 unmodified lines
166	166
167	167	int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu vcpu, struct kvm_interrupt irq);
168	168
	169	+extern int kvm_vm_ioctl_rtas_define_token(struct kvm kvm, void __user argp);
	170	+extern int kvmppc_rtas_hcall(struct kvm_vcpu *vcpu);
	171	+extern void kvmppc_rtas_tokens_free(struct kvm *kvm);
	172	+
169	173	/*
170	174	* Cuts out inst bits with ordering according to spec.
171	175	* That means the leftmost bit is zero. All given bits are included.

arch/powerpc/include/uapi/asm/kvm.h+6−0

@@ -324,6 +324,12 @@ struct kvm_allocate_rma { · 323 unmodified lines
324	324	__u64 rma_size;
325	325	};
326	326
	327	+/* for KVM_CAP_PPC_RTAS */
	328	+struct kvm_rtas_token_args {
	329	+ char name[120];
	330	+ __u64 token; /* Use a token of 0 to undefine a mapping */
	331	+};
	332	+
327	333	struct kvm_book3e_206_tlb_entry {
328	334	__u32 mas8;
329	335	__u32 mas1;

arch/powerpc/kvm/Makefile+1−0

@@ -86,6 +86,7 @@ kvm-book3s_64-module-objs := \ · 85 unmodified lines
86	86	emulate.o \
87	87	book3s.o \
88	88	book3s_64_vio.o \
	89	+ book3s_rtas.o \
89	90	$(kvm-book3s_64-objs-y)
90	91
91	92	kvm-objs-$(CONFIG_KVM_BOOK3S_64) := $(kvm-book3s_64-module-objs)

arch/powerpc/kvm/book3s_hv.c+17−1

@@ -483,7 +483,7 @@ int kvmppc_pseries_do_hcall(struct kvm_vcpu *vcpu) · 482 unmodified lines
483	483	unsigned long req = kvmppc_get_gpr(vcpu, 3);
484	484	unsigned long target, ret = H_SUCCESS;
485	485	struct kvm_vcpu *tvcpu;
486		− int idx;
	486	+ int idx, rc;
487	487
488	488	switch (req) {
489	489	case H_ENTER:
@@ -519,6 +519,19 @@ int kvmppc_pseries_do_hcall(struct kvm_vcpu *vcpu) · 29 unmodified lines
519	519	kvmppc_get_gpr(vcpu, 5),
520	520	kvmppc_get_gpr(vcpu, 6));
521	521	break;
	522	+ case H_RTAS:
	523	+ if (list_empty(&vcpu->kvm->arch.rtas_tokens))
	524	+ return RESUME_HOST;
	525	+
	526	+ rc = kvmppc_rtas_hcall(vcpu);
	527	+
	528	+ if (rc == -ENOENT)
	529	+ return RESUME_HOST;
	530	+ else if (rc == 0)
	531	+ break;
	532	+
	533	+ /* Send the error out to userspace via KVM_RUN */
	534	+ return rc;
522	535	default:
523	536	return RESUME_HOST;
524	537	}
@@ -1829,6 +1842,7 @@ int kvmppc_core_init_vm(struct kvm *kvm) · 1304 unmodified lines
1829	1842	cpumask_setall(&kvm->arch.need_tlb_flush);
1830	1843
1831	1844	INIT_LIST_HEAD(&kvm->arch.spapr_tce_tables);
	1845	+ INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
1832	1846
1833	1847	kvm->arch.rma = NULL;
1834	1848
@@ -1874,6 +1888,8 @@ void kvmppc_core_destroy_vm(struct kvm *kvm) · 39 unmodified lines
1874	1888	kvm->arch.rma = NULL;
1875	1889	}
1876	1890
	1891	+ kvmppc_rtas_tokens_free(kvm);
	1892	+
1877	1893	kvmppc_free_hpt(kvm);
1878	1894	WARN_ON(!list_empty(&kvm->arch.spapr_tce_tables));
1879	1895	}

arch/powerpc/kvm/book3s_pr.c+1−0

@@ -1296,6 +1296,7 @@ int kvmppc_core_init_vm(struct kvm *kvm) · 1295 unmodified lines
1296	1296	{
1297	1297	#ifdef CONFIG_PPC64
1298	1298	INIT_LIST_HEAD(&kvm->arch.spapr_tce_tables);
	1299	+ INIT_LIST_HEAD(&kvm->arch.rtas_tokens);
1299	1300	#endif
1300	1301
1301	1302	if (firmware_has_feature(FW_FEATURE_SET_MODE)) {

arch/powerpc/kvm/book3s_pr_papr.c+7−0

@@ -246,6 +246,13 @@ int kvmppc_h_pr(struct kvm_vcpu *vcpu, unsigned long cmd) · 245 unmodified lines
246	246	clear_bit(KVM_REQ_UNHALT, &vcpu->requests);
247	247	vcpu->stat.halt_wakeup++;
248	248	return EMULATE_DONE;
	249	+ case H_RTAS:
	250	+ if (list_empty(&vcpu->kvm->arch.rtas_tokens))
	251	+ return RESUME_HOST;
	252	+ if (kvmppc_rtas_hcall(vcpu))
	253	+ break;
	254	+ kvmppc_set_gpr(vcpu, 3, 0);
	255	+ return EMULATE_DONE;
249	256	}
250	257
251	258	return EMULATE_FAIL;

arch/powerpc/kvm/book3s_rtas.c+182−0

@@ -0,0 +1,182 @@
	1	+/*
	2	+ * Copyright 2012 Michael Ellerman, IBM Corporation.
	3	+ *
	4	+ * This program is free software; you can redistribute it and/or modify
	5	+ * it under the terms of the GNU General Public License, version 2, as
	6	+ * published by the Free Software Foundation.
	7	+ */
	8	+
	9	+#include <linux/kernel.h>
	10	+#include <linux/kvm_host.h>
	11	+#include <linux/kvm.h>
	12	+#include <linux/err.h>
	13	+
	14	+#include <asm/uaccess.h>
	15	+#include <asm/kvm_book3s.h>
	16	+#include <asm/kvm_ppc.h>
	17	+#include <asm/hvcall.h>
	18	+#include <asm/rtas.h>
	19	+
	20	+
	21	+struct rtas_handler {
	22	+ void (handler)(struct kvm_vcpu vcpu, struct rtas_args *args);
	23	+ char *name;
	24	+};
	25	+
	26	+static struct rtas_handler rtas_handlers[] = { };
	27	+
	28	+struct rtas_token_definition {
	29	+ struct list_head list;
	30	+ struct rtas_handler *handler;
	31	+ u64 token;
	32	+};
	33	+
	34	+static int rtas_name_matches(char s1, char s2)
	35	+{
	36	+ struct kvm_rtas_token_args args;
	37	+ return !strncmp(s1, s2, sizeof(args.name));
	38	+}
	39	+
	40	+static int rtas_token_undefine(struct kvm kvm, char name)
	41	+{
	42	+ struct rtas_token_definition d, tmp;
	43	+
	44	+ lockdep_assert_held(&kvm->lock);
	45	+
	46	+ list_for_each_entry_safe(d, tmp, &kvm->arch.rtas_tokens, list) {
	47	+ if (rtas_name_matches(d->handler->name, name)) {
	48	+ list_del(&d->list);
	49	+ kfree(d);
	50	+ return 0;
	51	+ }
	52	+ }
	53	+
	54	+ /* It's not an error to undefine an undefined token */
	55	+ return 0;
	56	+}
	57	+
	58	+static int rtas_token_define(struct kvm kvm, char name, u64 token)
	59	+{
	60	+ struct rtas_token_definition *d;
	61	+ struct rtas_handler *h = NULL;
	62	+ bool found;
	63	+ int i;
	64	+
	65	+ lockdep_assert_held(&kvm->lock);
	66	+
	67	+ list_for_each_entry(d, &kvm->arch.rtas_tokens, list) {
	68	+ if (d->token == token)
	69	+ return -EEXIST;
	70	+ }
	71	+
	72	+ found = false;
	73	+ for (i = 0; i < ARRAY_SIZE(rtas_handlers); i++) {
	74	+ h = &rtas_handlers[i];
	75	+ if (rtas_name_matches(h->name, name)) {
	76	+ found = true;
	77	+ break;
	78	+ }
	79	+ }
	80	+
	81	+ if (!found)
	82	+ return -ENOENT;
	83	+
	84	+ d = kzalloc(sizeof(*d), GFP_KERNEL);
	85	+ if (!d)
	86	+ return -ENOMEM;
	87	+
	88	+ d->handler = h;
	89	+ d->token = token;
	90	+
	91	+ list_add_tail(&d->list, &kvm->arch.rtas_tokens);
	92	+
	93	+ return 0;
	94	+}
	95	+
	96	+int kvm_vm_ioctl_rtas_define_token(struct kvm kvm, void __user argp)
	97	+{
	98	+ struct kvm_rtas_token_args args;
	99	+ int rc;
	100	+
	101	+ if (copy_from_user(&args, argp, sizeof(args)))
	102	+ return -EFAULT;
	103	+
	104	+ mutex_lock(&kvm->lock);
	105	+
	106	+ if (args.token)
	107	+ rc = rtas_token_define(kvm, args.name, args.token);
	108	+ else
	109	+ rc = rtas_token_undefine(kvm, args.name);
	110	+
	111	+ mutex_unlock(&kvm->lock);
	112	+
	113	+ return rc;
	114	+}
	115	+
	116	+int kvmppc_rtas_hcall(struct kvm_vcpu *vcpu)
	117	+{
	118	+ struct rtas_token_definition *d;
	119	+ struct rtas_args args;
	120	+ rtas_arg_t *orig_rets;
	121	+ gpa_t args_phys;
	122	+ int rc;
	123	+
	124	+ /* r4 contains the guest physical address of the RTAS args */
	125	+ args_phys = kvmppc_get_gpr(vcpu, 4);
	126	+
	127	+ rc = kvm_read_guest(vcpu->kvm, args_phys, &args, sizeof(args));
	128	+ if (rc)
	129	+ goto fail;
	130	+
	131	+ /*
	132	+ * args->rets is a pointer into args->args. Now that we've
	133	+ * copied args we need to fix it up to point into our copy,
	134	+ * not the guest args. We also need to save the original
	135	+ * value so we can restore it on the way out.
	136	+ */
	137	+ orig_rets = args.rets;
	138	+ args.rets = &args.args[args.nargs];
	139	+
	140	+ mutex_lock(&vcpu->kvm->lock);
	141	+
	142	+ rc = -ENOENT;
	143	+ list_for_each_entry(d, &vcpu->kvm->arch.rtas_tokens, list) {
	144	+ if (d->token == args.token) {
	145	+ d->handler->handler(vcpu, &args);
	146	+ rc = 0;
	147	+ break;
	148	+ }
	149	+ }
	150	+
	151	+ mutex_unlock(&vcpu->kvm->lock);
	152	+
	153	+ if (rc == 0) {
	154	+ args.rets = orig_rets;
	155	+ rc = kvm_write_guest(vcpu->kvm, args_phys, &args, sizeof(args));
	156	+ if (rc)
	157	+ goto fail;
	158	+ }
	159	+
	160	+ return rc;
	161	+
	162	+fail:
	163	+ /*
	164	+ * We only get here if the guest has called RTAS with a bogus
	165	+ * args pointer. That means we can't get to the args, and so we
	166	+ * can't fail the RTAS call. So fail right out to userspace,
	167	+ * which should kill the guest.
	168	+ */
	169	+ return rc;
	170	+}
	171	+
	172	+void kvmppc_rtas_tokens_free(struct kvm *kvm)
	173	+{
	174	+ struct rtas_token_definition d, tmp;
	175	+
	176	+ lockdep_assert_held(&kvm->lock);
	177	+
	178	+ list_for_each_entry_safe(d, tmp, &kvm->arch.rtas_tokens, list) {
	179	+ list_del(&d->list);
	180	+ kfree(d);
	181	+ }
	182	+}

arch/powerpc/kvm/powerpc.c+8−0

@@ -341,6 +341,7 @@ int kvm_dev_ioctl_check_extension(long ext) · 340 unmodified lines
341	341	#ifdef CONFIG_PPC_BOOK3S_64
342	342	case KVM_CAP_SPAPR_TCE:
343	343	case KVM_CAP_PPC_ALLOC_HTAB:
	344	+ case KVM_CAP_PPC_RTAS:
344	345	r = 1;
345	346	break;
346	347	#endif /* CONFIG_PPC_BOOK3S_64 */
@@ -986,6 +987,7 @@ long kvm_arch_vm_ioctl(struct file *filp, · 639 unmodified lines
986	987	#ifdef CONFIG_KVM_BOOK3S_64_HV
987	988	case KVM_ALLOCATE_RMA: {
988	989	struct kvm_allocate_rma rma;
	990	+ struct kvm *kvm = filp->private_data;
989	991
990	992	r = kvm_vm_ioctl_allocate_rma(kvm, &rma);
991	993	if (r >= 0 && copy_to_user(argp, &rma, sizeof(rma)))
@@ -1030,6 +1032,12 @@ long kvm_arch_vm_ioctl(struct file *filp, · 38 unmodified lines
1030	1032	r = -EFAULT;
1031	1033	break;
1032	1034	}
	1035	+ case KVM_PPC_RTAS_DEFINE_TOKEN: {
	1036	+ struct kvm *kvm = filp->private_data;
	1037	+
	1038	+ r = kvm_vm_ioctl_rtas_define_token(kvm, argp);
	1039	+ break;
	1040	+ }
1033	1041	#endif /* CONFIG_PPC_BOOK3S_64 */
1034	1042	default:
1035	1043	r = -ENOTTY;

include/uapi/linux/kvm.h+3−0

@@ -668,6 +668,7 @@ struct kvm_ppc_smmu_info { · 667 unmodified lines
668	668	#define KVM_CAP_ARM_SET_DEVICE_ADDR 88
669	669	#define KVM_CAP_DEVICE_CTRL 89
670	670	#define KVM_CAP_IRQ_MPIC 90
	671	+#define KVM_CAP_PPC_RTAS 91
671	672
672	673	#ifdef KVM_CAP_IRQ_ROUTING
673	674
@@ -928,6 +929,8 @@ struct kvm_s390_ucas_mapping { · 254 unmodified lines
928	929	#define KVM_PPC_GET_HTAB_FD _IOW(KVMIO, 0xaa, struct kvm_get_htab_fd)
929	930	/* Available with KVM_CAP_ARM_SET_DEVICE_ADDR */
930	931	#define KVM_ARM_SET_DEVICE_ADDR _IOW(KVMIO, 0xab, struct kvm_arm_device_addr)
	932	+/* Available with KVM_CAP_PPC_RTAS */
	933	+#define KVM_PPC_RTAS_DEFINE_TOKEN _IOW(KVMIO, 0xac, struct kvm_rtas_token_args)
931	934
932	935	/* ioctl for vm fd */
933	936	#define KVM_CREATE_DEVICE _IOWR(KVMIO, 0xe0, struct kvm_create_device)

KVM: PPC: Book3S: Add infrastructure to implement kernel-side RTAS calls

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code