arch/arm/kvm/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
  61         return PSCI_RET_SUCCESS;
  62 }
  63
  64 static void kvm_psci_vcpu_off(struct kvm_vcpu *vcpu)
  65 {
  66         vcpu->arch.pause = true;
  67 }
  68
  69 static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
  70 {
  71         struct kvm *kvm = source_vcpu->kvm;
  72         struct kvm_vcpu *vcpu = NULL;
  73         wait_queue_head_t *wq;
  74         unsigned long cpu_id;
  75         unsigned long context_id;
  76         phys_addr_t target_pc;
  77
  78         cpu_id = *vcpu_reg(source_vcpu, 1) & MPIDR_HWID_BITMASK;
  79         if (vcpu_mode_is_32bit(source_vcpu))
  80                 cpu_id &= ~((u32) 0);
  81
  82         vcpu = kvm_mpidr_to_vcpu(kvm, cpu_id);
  83
  84         /*
  85          * Make sure the caller requested a valid CPU and that the CPU is
  86          * turned off.
  87          */
  88         if (!vcpu)
  89                 return PSCI_RET_INVALID_PARAMS;
  90         if (!vcpu->arch.pause) {
  91                 if (kvm_psci_version(source_vcpu) != KVM_ARM_PSCI_0_1)
  92                         return PSCI_RET_ALREADY_ON;
  93                 else
  94                         return PSCI_RET_INVALID_PARAMS;
  95         }
  96
  97         target_pc = *vcpu_reg(source_vcpu, 2);
  98         context_id = *vcpu_reg(source_vcpu, 3);
  99
 100         kvm_reset_vcpu(vcpu);
 101
 102         /* Gracefully handle Thumb2 entry point */
 103         if (vcpu_mode_is_32bit(vcpu) && (target_pc & 1)) {
 104                 target_pc &= ~((phys_addr_t) 1);
 105                 vcpu_set_thumb(vcpu);
 106         }
 107
 108         /* Propagate caller endianness */
 109         if (kvm_vcpu_is_be(source_vcpu))
 110                 kvm_vcpu_set_be(vcpu);
 111
 112         *vcpu_pc(vcpu) = target_pc;
 113         /*
 114          * NOTE: We always update r0 (or x0) because for PSCI v0.1
 115          * the general puspose registers are undefined upon CPU_ON.
 116          */
 117         *vcpu_reg(vcpu, 0) = context_id;
 118         vcpu->arch.pause = false;
 119         smp_mb();               /* Make sure the above is visible */
 120
 121         wq = kvm_arch_vcpu_wq(vcpu);
 122         wake_up_interruptible(wq);
 123
 124         return PSCI_RET_SUCCESS;
 125 }
 126
 127 static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
 128 {
 129         int i;
 130         unsigned long mpidr;
 131         unsigned long target_affinity;
 132         unsigned long target_affinity_mask;
 133         unsigned long lowest_affinity_level;
 134         struct kvm *kvm = vcpu->kvm;
 135         struct kvm_vcpu *tmp;
 136
 137         target_affinity = *vcpu_reg(vcpu, 1);
 138         lowest_affinity_level = *vcpu_reg(vcpu, 2);
 139
 140         /* Determine target affinity mask */
 141         target_affinity_mask = psci_affinity_mask(lowest_affinity_level);
 142         if (!target_affinity_mask)
 143                 return PSCI_RET_INVALID_PARAMS;
 144
 145         /* Ignore other bits of target affinity */
 146         target_affinity &= target_affinity_mask;
 147
 148         /*
 149          * If one or more VCPU matching target affinity are running
 150          * then ON else OFF
 151          */
 152         kvm_for_each_vcpu(i, tmp, kvm) {
 153                 mpidr = kvm_vcpu_get_mpidr_aff(tmp);
 154                 if (((mpidr & target_affinity_mask) == target_affinity) &&
 155                     !tmp->arch.pause) {
 156                         return PSCI_0_2_AFFINITY_LEVEL_ON;
 157                 }
 158         }
 159
 160         return PSCI_0_2_AFFINITY_LEVEL_OFF;
 161 }
 162
 163 static void kvm_prepare_system_event(struct kvm_vcpu *vcpu, u32 type)
 164 {
 165         int i;
 166         struct kvm_vcpu *tmp;
 167
 168         /*
 169          * The KVM ABI specifies that a system event exit may call KVM_RUN
 170          * again and may perform shutdown/reboot at a later time that when the
 171          * actual request is made.  Since we are implementing PSCI and a
 172          * caller of PSCI reboot and shutdown expects that the system shuts
 173          * down or reboots immediately, let's make sure that VCPUs are not run
 174          * after this call is handled and before the VCPUs have been
 175          * re-initialized.
 176          */
 177         kvm_for_each_vcpu(i, tmp, vcpu->kvm) {
 178                 tmp->arch.pause = true;
 179                 kvm_vcpu_kick(tmp);
 180         }
 181
 182         memset(&vcpu->run->system_event, 0, sizeof(vcpu->run->system_event));
 183         vcpu->run->system_event.type = type;
 184         vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT;
 185 }
 186
 187 static void kvm_psci_system_off(struct kvm_vcpu *vcpu)
 188 {
 189         kvm_prepare_system_event(vcpu, KVM_SYSTEM_EVENT_SHUTDOWN);
 190 }
 191
 192 static void kvm_psci_system_reset(struct kvm_vcpu *vcpu)
 193 {
 194         kvm_prepare_system_event(vcpu, KVM_SYSTEM_EVENT_RESET);
 195 }
 196
 197 int kvm_psci_version(struct kvm_vcpu *vcpu)
 198 {
 199         if (test_bit(KVM_ARM_VCPU_PSCI_0_2, vcpu->arch.features))
 200                 return KVM_ARM_PSCI_0_2;
 201
 202         return KVM_ARM_PSCI_0_1;
 203 }
 204
 205 static int kvm_psci_0_2_call(struct kvm_vcpu *vcpu)
 206 {
 207         int ret = 1;
 208         unsigned long psci_fn = *vcpu_reg(vcpu, 0) & ~((u32) 0);
 209         unsigned long val;
 210
 211         switch (psci_fn) {
 212         case PSCI_0_2_FN_PSCI_VERSION:
 213                 /*
 214                  * Bits[31:16] = Major Version = 0
 215                  * Bits[15:0] = Minor Version = 2
 216                  */
 217                 val = 2;
 218                 break;
 219         case PSCI_0_2_FN_CPU_SUSPEND:
 220         case PSCI_0_2_FN64_CPU_SUSPEND:
 221                 val = kvm_psci_vcpu_suspend(vcpu);
 222                 break;
 223         case PSCI_0_2_FN_CPU_OFF:
 224                 kvm_psci_vcpu_off(vcpu);
 225                 val = PSCI_RET_SUCCESS;
 226                 break;
 227         case PSCI_0_2_FN_CPU_ON:
 228         case PSCI_0_2_FN64_CPU_ON:
 229                 val = kvm_psci_vcpu_on(vcpu);
 230                 break;
 231         case PSCI_0_2_FN_AFFINITY_INFO:
 232         case PSCI_0_2_FN64_AFFINITY_INFO:
 233                 val = kvm_psci_vcpu_affinity_info(vcpu);
 234                 break;
 235         case PSCI_0_2_FN_MIGRATE_INFO_TYPE:
 236                 /*
 237                  * Trusted OS is MP hence does not require migration
 238                  * or
 239                  * Trusted OS is not present
 240                  */
 241                 val = PSCI_0_2_TOS_MP;
 242                 break;
 243         case PSCI_0_2_FN_SYSTEM_OFF:
 244                 kvm_psci_system_off(vcpu);
 245                 /*
 246                  * We should'nt be going back to guest VCPU after
 247                  * receiving SYSTEM_OFF request.
 248                  *
 249                  * If user space accidently/deliberately resumes
 250                  * guest VCPU after SYSTEM_OFF request then guest
 251                  * VCPU should see internal failure from PSCI return
 252                  * value. To achieve this, we preload r0 (or x0) with
 253                  * PSCI return value INTERNAL_FAILURE.
 254                  */
 255                 val = PSCI_RET_INTERNAL_FAILURE;
 256                 ret = 0;
 257                 break;
 258         case PSCI_0_2_FN_SYSTEM_RESET:
 259                 kvm_psci_system_reset(vcpu);
 260                 /*
 261                  * Same reason as SYSTEM_OFF for preloading r0 (or x0)
 262                  * with PSCI return value INTERNAL_FAILURE.
 263                  */
 264                 val = PSCI_RET_INTERNAL_FAILURE;
 265                 ret = 0;
 266                 break;
 267         default:
 268                 val = PSCI_RET_NOT_SUPPORTED;
 269                 break;
 270         }
 271
 272         *vcpu_reg(vcpu, 0) = val;
 273         return ret;
 274 }
 275
 276 static int kvm_psci_0_1_call(struct kvm_vcpu *vcpu)
 277 {
 278         unsigned long psci_fn = *vcpu_reg(vcpu, 0) & ~((u32) 0);
 279         unsigned long val;
 280
 281         switch (psci_fn) {
 282         case KVM_PSCI_FN_CPU_OFF:
 283                 kvm_psci_vcpu_off(vcpu);
 284                 val = PSCI_RET_SUCCESS;
 285                 break;
 286         case KVM_PSCI_FN_CPU_ON:
 287                 val = kvm_psci_vcpu_on(vcpu);
 288                 break;
 289         default:
 290                 val = PSCI_RET_NOT_SUPPORTED;
 291                 break;
 292         }
 293
 294         *vcpu_reg(vcpu, 0) = val;
 295         return 1;
 296 }
 297
 298 /**
 299  * kvm_psci_call - handle PSCI call if r0 value is in range
 300  * @vcpu: Pointer to the VCPU struct
 301  *
 302  * Handle PSCI calls from guests through traps from HVC instructions.
 303  * The calling convention is similar to SMC calls to the secure world
 304  * where the function number is placed in r0.
 305  *
 306  * This function returns: > 0 (success), 0 (success but exit to user
 307  * space), and < 0 (errors)
 308  *
 309  * Errors:
 310  * -EINVAL: Unrecognized PSCI function
 311  */
 312 int kvm_psci_call(struct kvm_vcpu *vcpu)
 313 {
 314         switch (kvm_psci_version(vcpu)) {
 315         case KVM_ARM_PSCI_0_2:
 316                 return kvm_psci_0_2_call(vcpu);
 317         case KVM_ARM_PSCI_0_1:
 318                 return kvm_psci_0_1_call(vcpu);
 319         default:
 320                 return -EINVAL;
 321         };
 322 }