virt/kvm/arm/psci.c

   1 /*
   2  * Copyright (C) 2012 - ARM Ltd
   3  * Author: Marc Zyngier <marc.zyngier@arm.com>
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License version 2 as
   7  * published by the Free Software Foundation.
   8  *
   9  * This program is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12  * GNU General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU General Public License
  15  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  16  */
  17
  18 #include <linux/preempt.h>
  19 #include <linux/kvm_host.h>
  20 #include <linux/wait.h>
  21
  22 #include <asm/cputype.h>
  23 #include <asm/kvm_emulate.h>
  24 #include <asm/kvm_psci.h>
  25 #include <asm/kvm_host.h>
  26
  27 #include <uapi/linux/psci.h>
  28
  29 /*
  30  * This is an implementation of the Power State Coordination Interface
  31  * as described in ARM document number ARM DEN 0022A.
  32  */
  33
  34 #define AFFINITY_MASK(level)    ~((0x1UL << ((level) * MPIDR_LEVEL_BITS)) - 1)
  35
  36 static unsigned long psci_affinity_mask(unsigned long affinity_level)
  37 {
  38         if (affinity_level <= 3)
  39                 return MPIDR_HWID_BITMASK & AFFINITY_MASK(affinity_level);
  40
  41         return 0;
  42 }
  43
  44 static unsigned long kvm_psci_vcpu_suspend(struct kvm_vcpu *vcpu)
  45 {
  46         /*
  47          * NOTE: For simplicity, we make VCPU suspend emulation to be
  48          * same-as WFI (Wait-for-interrupt) emulation.
  49          *
  50          * This means for KVM the wakeup events are interrupts and
  51          * this is consistent with intended use of StateID as described
  52          * in section 5.4.1 of PSCI v0.2 specification (ARM DEN 0022A).
  53          *
  54          * Further, we also treat power-down request to be same as
  55          * stand-by request as-per section 5.4.2 clause 3 of PSCI v0.2
  56          * specification (ARM DEN 0022A). This means all suspend states
  57          * for KVM will preserve the register state.
  58          */
  59         kvm_vcpu_block(vcpu);
  60         kvm_clear_request(KVM_REQ_UNHALT, vcpu);
  61
  62         return PSCI_RET_SUCCESS;
  63 }
  64
  65 static void kvm_psci_vcpu_off(struct kvm_vcpu *vcpu)
  66 {
  67         vcpu->arch.power_off = true;
  68         kvm_make_request(KVM_REQ_SLEEP, vcpu);
  69         kvm_vcpu_kick(vcpu);
  70 }
  71
  72 static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
  73 {
  74         struct kvm *kvm = source_vcpu->kvm;
  75         struct kvm_vcpu *vcpu = NULL;
  76         struct swait_queue_head *wq;
  77         unsigned long cpu_id;
  78         unsigned long context_id;
  79         phys_addr_t target_pc;
  80
  81         cpu_id = vcpu_get_reg(source_vcpu, 1) & MPIDR_HWID_BITMASK;
  82         if (vcpu_mode_is_32bit(source_vcpu))
  83                 cpu_id &= ~((u32) 0);
  84
  85         vcpu = kvm_mpidr_to_vcpu(kvm, cpu_id);
  86
  87         /*
  88          * Make sure the caller requested a valid CPU and that the CPU is
  89          * turned off.
  90          */
  91         if (!vcpu)
  92                 return PSCI_RET_INVALID_PARAMS;
  93         if (!vcpu->arch.power_off) {
  94                 if (kvm_psci_version(source_vcpu) != KVM_ARM_PSCI_0_1)
  95                         return PSCI_RET_ALREADY_ON;
  96                 else
  97                         return PSCI_RET_INVALID_PARAMS;
  98         }
  99
 100         target_pc = vcpu_get_reg(source_vcpu, 2);
 101         context_id = vcpu_get_reg(source_vcpu, 3);
 102
 103         kvm_reset_vcpu(vcpu);
 104
 105         /* Gracefully handle Thumb2 entry point */
 106         if (vcpu_mode_is_32bit(vcpu) && (target_pc & 1)) {
 107                 target_pc &= ~((phys_addr_t) 1);
 108                 vcpu_set_thumb(vcpu);
 109         }
 110
 111         /* Propagate caller endianness */
 112         if (kvm_vcpu_is_be(source_vcpu))
 113                 kvm_vcpu_set_be(vcpu);
 114
 115         *vcpu_pc(vcpu) = target_pc;
 116         /*
 117          * NOTE: We always update r0 (or x0) because for PSCI v0.1
 118          * the general puspose registers are undefined upon CPU_ON.
 119          */
 120         vcpu_set_reg(vcpu, 0, context_id);
 121         vcpu->arch.power_off = false;
 122         smp_mb();               /* Make sure the above is visible */
 123
 124         wq = kvm_arch_vcpu_wq(vcpu);
 125         swake_up(wq);
 126
 127         return PSCI_RET_SUCCESS;
 128 }
 129
 130 static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
 131 {
 132         int i, matching_cpus = 0;
 133         unsigned long mpidr;
 134         unsigned long target_affinity;
 135         unsigned long target_affinity_mask;
 136         unsigned long lowest_affinity_level;
 137         struct kvm *kvm = vcpu->kvm;
 138         struct kvm_vcpu *tmp;
 139
 140         target_affinity = vcpu_get_reg(vcpu, 1);
 141         lowest_affinity_level = vcpu_get_reg(vcpu, 2);
 142
 143         /* Determine target affinity mask */
 144         target_affinity_mask = psci_affinity_mask(lowest_affinity_level);
 145         if (!target_affinity_mask)
 146                 return PSCI_RET_INVALID_PARAMS;
 147
 148         /* Ignore other bits of target affinity */
 149         target_affinity &= target_affinity_mask;
 150
 151         /*
 152          * If one or more VCPU matching target affinity are running
 153          * then ON else OFF
 154          */
 155         kvm_for_each_vcpu(i, tmp, kvm) {
 156                 mpidr = kvm_vcpu_get_mpidr_aff(tmp);
 157                 if ((mpidr & target_affinity_mask) == target_affinity) {
 158                         matching_cpus++;
 159                         if (!tmp->arch.power_off)
 160                                 return PSCI_0_2_AFFINITY_LEVEL_ON;
 161                 }
 162         }
 163
 164         if (!matching_cpus)
 165                 return PSCI_RET_INVALID_PARAMS;
 166
 167         return PSCI_0_2_AFFINITY_LEVEL_OFF;
 168 }
 169
 170 static void kvm_prepare_system_event(struct kvm_vcpu *vcpu, u32 type)
 171 {
 172         int i;
 173         struct kvm_vcpu *tmp;
 174
 175         /*
 176          * The KVM ABI specifies that a system event exit may call KVM_RUN
 177          * again and may perform shutdown/reboot at a later time that when the
 178          * actual request is made.  Since we are implementing PSCI and a
 179          * caller of PSCI reboot and shutdown expects that the system shuts
 180          * down or reboots immediately, let's make sure that VCPUs are not run
 181          * after this call is handled and before the VCPUs have been
 182          * re-initialized.
 183          */
 184         kvm_for_each_vcpu(i, tmp, vcpu->kvm)
 185                 tmp->arch.power_off = true;
 186         kvm_make_all_cpus_request(vcpu->kvm, KVM_REQ_SLEEP);
 187
 188         memset(&vcpu->run->system_event, 0, sizeof(vcpu->run->system_event));
 189         vcpu->run->system_event.type = type;
 190         vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT;
 191 }
 192
 193 static void kvm_psci_system_off(struct kvm_vcpu *vcpu)
 194 {
 195         kvm_prepare_system_event(vcpu, KVM_SYSTEM_EVENT_SHUTDOWN);
 196 }
 197
 198 static void kvm_psci_system_reset(struct kvm_vcpu *vcpu)
 199 {
 200         kvm_prepare_system_event(vcpu, KVM_SYSTEM_EVENT_RESET);
 201 }
 202
 203 int kvm_psci_version(struct kvm_vcpu *vcpu)
 204 {
 205         if (test_bit(KVM_ARM_VCPU_PSCI_0_2, vcpu->arch.features))
 206                 return KVM_ARM_PSCI_0_2;
 207
 208         return KVM_ARM_PSCI_0_1;
 209 }
 210
 211 static int kvm_psci_0_2_call(struct kvm_vcpu *vcpu)
 212 {
 213         struct kvm *kvm = vcpu->kvm;
 214         unsigned long psci_fn = vcpu_get_reg(vcpu, 0) & ~((u32) 0);
 215         unsigned long val;
 216         int ret = 1;
 217
 218         switch (psci_fn) {
 219         case PSCI_0_2_FN_PSCI_VERSION:
 220                 /*
 221                  * Bits[31:16] = Major Version = 0
 222                  * Bits[15:0] = Minor Version = 2
 223                  */
 224                 val = 2;
 225                 break;
 226         case PSCI_0_2_FN_CPU_SUSPEND:
 227         case PSCI_0_2_FN64_CPU_SUSPEND:
 228                 val = kvm_psci_vcpu_suspend(vcpu);
 229                 break;
 230         case PSCI_0_2_FN_CPU_OFF:
 231                 kvm_psci_vcpu_off(vcpu);
 232                 val = PSCI_RET_SUCCESS;
 233                 break;
 234         case PSCI_0_2_FN_CPU_ON:
 235         case PSCI_0_2_FN64_CPU_ON:
 236                 mutex_lock(&kvm->lock);
 237                 val = kvm_psci_vcpu_on(vcpu);
 238                 mutex_unlock(&kvm->lock);
 239                 break;
 240         case PSCI_0_2_FN_AFFINITY_INFO:
 241         case PSCI_0_2_FN64_AFFINITY_INFO:
 242                 val = kvm_psci_vcpu_affinity_info(vcpu);
 243                 break;
 244         case PSCI_0_2_FN_MIGRATE_INFO_TYPE:
 245                 /*
 246                  * Trusted OS is MP hence does not require migration
 247                  * or
 248                  * Trusted OS is not present
 249                  */
 250                 val = PSCI_0_2_TOS_MP;
 251                 break;
 252         case PSCI_0_2_FN_SYSTEM_OFF:
 253                 kvm_psci_system_off(vcpu);
 254                 /*
 255                  * We should'nt be going back to guest VCPU after
 256                  * receiving SYSTEM_OFF request.
 257                  *
 258                  * If user space accidently/deliberately resumes
 259                  * guest VCPU after SYSTEM_OFF request then guest
 260                  * VCPU should see internal failure from PSCI return
 261                  * value. To achieve this, we preload r0 (or x0) with
 262                  * PSCI return value INTERNAL_FAILURE.
 263                  */
 264                 val = PSCI_RET_INTERNAL_FAILURE;
 265                 ret = 0;
 266                 break;
 267         case PSCI_0_2_FN_SYSTEM_RESET:
 268                 kvm_psci_system_reset(vcpu);
 269                 /*
 270                  * Same reason as SYSTEM_OFF for preloading r0 (or x0)
 271                  * with PSCI return value INTERNAL_FAILURE.
 272                  */
 273                 val = PSCI_RET_INTERNAL_FAILURE;
 274                 ret = 0;
 275                 break;
 276         default:
 277                 val = PSCI_RET_NOT_SUPPORTED;
 278                 break;
 279         }
 280
 281         vcpu_set_reg(vcpu, 0, val);
 282         return ret;
 283 }
 284
 285 static int kvm_psci_0_1_call(struct kvm_vcpu *vcpu)
 286 {
 287         struct kvm *kvm = vcpu->kvm;
 288         unsigned long psci_fn = vcpu_get_reg(vcpu, 0) & ~((u32) 0);
 289         unsigned long val;
 290
 291         switch (psci_fn) {
 292         case KVM_PSCI_FN_CPU_OFF:
 293                 kvm_psci_vcpu_off(vcpu);
 294                 val = PSCI_RET_SUCCESS;
 295                 break;
 296         case KVM_PSCI_FN_CPU_ON:
 297                 mutex_lock(&kvm->lock);
 298                 val = kvm_psci_vcpu_on(vcpu);
 299                 mutex_unlock(&kvm->lock);
 300                 break;
 301         default:
 302                 val = PSCI_RET_NOT_SUPPORTED;
 303                 break;
 304         }
 305
 306         vcpu_set_reg(vcpu, 0, val);
 307         return 1;
 308 }
 309
 310 /**
 311  * kvm_psci_call - handle PSCI call if r0 value is in range
 312  * @vcpu: Pointer to the VCPU struct
 313  *
 314  * Handle PSCI calls from guests through traps from HVC instructions.
 315  * The calling convention is similar to SMC calls to the secure world
 316  * where the function number is placed in r0.
 317  *
 318  * This function returns: > 0 (success), 0 (success but exit to user
 319  * space), and < 0 (errors)
 320  *
 321  * Errors:
 322  * -EINVAL: Unrecognized PSCI function
 323  */
 324 int kvm_psci_call(struct kvm_vcpu *vcpu)
 325 {
 326         switch (kvm_psci_version(vcpu)) {
 327         case KVM_ARM_PSCI_0_2:
 328                 return kvm_psci_0_2_call(vcpu);
 329         case KVM_ARM_PSCI_0_1:
 330                 return kvm_psci_0_1_call(vcpu);
 331         default:
 332                 return -EINVAL;
 333         };
 334 }