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.power_off = 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         struct swait_queue_head *wq;
  74         unsigned long cpu_id;
  75         unsigned long context_id;
  76         phys_addr_t target_pc;
  77
  78         cpu_id = vcpu_get_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.power_off) {
  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_get_reg(source_vcpu, 2);
  98         context_id = vcpu_get_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_set_reg(vcpu, 0, context_id);
 118         vcpu->arch.power_off = false;
 119         smp_mb();               /* Make sure the above is visible */
 120
 121         wq = kvm_arch_vcpu_wq(vcpu);
 122         swake_up(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, matching_cpus = 0;
 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_get_reg(vcpu, 1);
 138         lowest_affinity_level = vcpu_get_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                         matching_cpus++;
 156                         if (!tmp->arch.power_off)
 157                                 return PSCI_0_2_AFFINITY_LEVEL_ON;
 158                 }
 159         }
 160
 161         if (!matching_cpus)
 162                 return PSCI_RET_INVALID_PARAMS;
 163
 164         return PSCI_0_2_AFFINITY_LEVEL_OFF;
 165 }
 166
 167 static void kvm_prepare_system_event(struct kvm_vcpu *vcpu, u32 type)
 168 {
 169         int i;
 170         struct kvm_vcpu *tmp;
 171
 172         /*
 173          * The KVM ABI specifies that a system event exit may call KVM_RUN
 174          * again and may perform shutdown/reboot at a later time that when the
 175          * actual request is made.  Since we are implementing PSCI and a
 176          * caller of PSCI reboot and shutdown expects that the system shuts
 177          * down or reboots immediately, let's make sure that VCPUs are not run
 178          * after this call is handled and before the VCPUs have been
 179          * re-initialized.
 180          */
 181         kvm_for_each_vcpu(i, tmp, vcpu->kvm) {
 182                 tmp->arch.power_off = true;
 183                 kvm_vcpu_kick(tmp);
 184         }
 185
 186         memset(&vcpu->run->system_event, 0, sizeof(vcpu->run->system_event));
 187         vcpu->run->system_event.type = type;
 188         vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT;
 189 }
 190
 191 static void kvm_psci_system_off(struct kvm_vcpu *vcpu)
 192 {
 193         kvm_prepare_system_event(vcpu, KVM_SYSTEM_EVENT_SHUTDOWN);
 194 }
 195
 196 static void kvm_psci_system_reset(struct kvm_vcpu *vcpu)
 197 {
 198         kvm_prepare_system_event(vcpu, KVM_SYSTEM_EVENT_RESET);
 199 }
 200
 201 int kvm_psci_version(struct kvm_vcpu *vcpu)
 202 {
 203         if (test_bit(KVM_ARM_VCPU_PSCI_0_2, vcpu->arch.features))
 204                 return KVM_ARM_PSCI_0_2;
 205
 206         return KVM_ARM_PSCI_0_1;
 207 }
 208
 209 static int kvm_psci_0_2_call(struct kvm_vcpu *vcpu)
 210 {
 211         int ret = 1;
 212         unsigned long psci_fn = vcpu_get_reg(vcpu, 0) & ~((u32) 0);
 213         unsigned long val;
 214
 215         switch (psci_fn) {
 216         case PSCI_0_2_FN_PSCI_VERSION:
 217                 /*
 218                  * Bits[31:16] = Major Version = 0
 219                  * Bits[15:0] = Minor Version = 2
 220                  */
 221                 val = 2;
 222                 break;
 223         case PSCI_0_2_FN_CPU_SUSPEND:
 224         case PSCI_0_2_FN64_CPU_SUSPEND:
 225                 val = kvm_psci_vcpu_suspend(vcpu);
 226                 break;
 227         case PSCI_0_2_FN_CPU_OFF:
 228                 kvm_psci_vcpu_off(vcpu);
 229                 val = PSCI_RET_SUCCESS;
 230                 break;
 231         case PSCI_0_2_FN_CPU_ON:
 232         case PSCI_0_2_FN64_CPU_ON:
 233                 val = kvm_psci_vcpu_on(vcpu);
 234                 break;
 235         case PSCI_0_2_FN_AFFINITY_INFO:
 236         case PSCI_0_2_FN64_AFFINITY_INFO:
 237                 val = kvm_psci_vcpu_affinity_info(vcpu);
 238                 break;
 239         case PSCI_0_2_FN_MIGRATE_INFO_TYPE:
 240                 /*
 241                  * Trusted OS is MP hence does not require migration
 242                  * or
 243                  * Trusted OS is not present
 244                  */
 245                 val = PSCI_0_2_TOS_MP;
 246                 break;
 247         case PSCI_0_2_FN_SYSTEM_OFF:
 248                 kvm_psci_system_off(vcpu);
 249                 /*
 250                  * We should'nt be going back to guest VCPU after
 251                  * receiving SYSTEM_OFF request.
 252                  *
 253                  * If user space accidently/deliberately resumes
 254                  * guest VCPU after SYSTEM_OFF request then guest
 255                  * VCPU should see internal failure from PSCI return
 256                  * value. To achieve this, we preload r0 (or x0) with
 257                  * PSCI return value INTERNAL_FAILURE.
 258                  */
 259                 val = PSCI_RET_INTERNAL_FAILURE;
 260                 ret = 0;
 261                 break;
 262         case PSCI_0_2_FN_SYSTEM_RESET:
 263                 kvm_psci_system_reset(vcpu);
 264                 /*
 265                  * Same reason as SYSTEM_OFF for preloading r0 (or x0)
 266                  * with PSCI return value INTERNAL_FAILURE.
 267                  */
 268                 val = PSCI_RET_INTERNAL_FAILURE;
 269                 ret = 0;
 270                 break;
 271         default:
 272                 val = PSCI_RET_NOT_SUPPORTED;
 273                 break;
 274         }
 275
 276         vcpu_set_reg(vcpu, 0, val);
 277         return ret;
 278 }
 279
 280 static int kvm_psci_0_1_call(struct kvm_vcpu *vcpu)
 281 {
 282         unsigned long psci_fn = vcpu_get_reg(vcpu, 0) & ~((u32) 0);
 283         unsigned long val;
 284
 285         switch (psci_fn) {
 286         case KVM_PSCI_FN_CPU_OFF:
 287                 kvm_psci_vcpu_off(vcpu);
 288                 val = PSCI_RET_SUCCESS;
 289                 break;
 290         case KVM_PSCI_FN_CPU_ON:
 291                 val = kvm_psci_vcpu_on(vcpu);
 292                 break;
 293         default:
 294                 val = PSCI_RET_NOT_SUPPORTED;
 295                 break;
 296         }
 297
 298         vcpu_set_reg(vcpu, 0, val);
 299         return 1;
 300 }
 301
 302 /**
 303  * kvm_psci_call - handle PSCI call if r0 value is in range
 304  * @vcpu: Pointer to the VCPU struct
 305  *
 306  * Handle PSCI calls from guests through traps from HVC instructions.
 307  * The calling convention is similar to SMC calls to the secure world
 308  * where the function number is placed in r0.
 309  *
 310  * This function returns: > 0 (success), 0 (success but exit to user
 311  * space), and < 0 (errors)
 312  *
 313  * Errors:
 314  * -EINVAL: Unrecognized PSCI function
 315  */
 316 int kvm_psci_call(struct kvm_vcpu *vcpu)
 317 {
 318         switch (kvm_psci_version(vcpu)) {
 319         case KVM_ARM_PSCI_0_2:
 320                 return kvm_psci_0_2_call(vcpu);
 321         case KVM_ARM_PSCI_0_1:
 322                 return kvm_psci_0_1_call(vcpu);
 323         default:
 324                 return -EINVAL;
 325         };
 326 }