drivers/hv/hv.c

   1 /*
   2  * Copyright (c) 2009, Microsoft Corporation.
   3  *
   4  * This program is free software; you can redistribute it and/or modify it
   5  * under the terms and conditions of the GNU General Public License,
   6  * version 2, as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope it will be useful, but WITHOUT
   9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  11  * more details.
  12  *
  13  * You should have received a copy of the GNU General Public License along with
  14  * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
  15  * Place - Suite 330, Boston, MA 02111-1307 USA.
  16  *
  17  * Authors:
  18  *   Haiyang Zhang <haiyangz@microsoft.com>
  19  *   Hank Janssen  <hjanssen@microsoft.com>
  20  *
  21  */
  22 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  23
  24 #include <linux/kernel.h>
  25 #include <linux/mm.h>
  26 #include <linux/slab.h>
  27 #include <linux/vmalloc.h>
  28 #include <linux/hyperv.h>
  29 #include <linux/version.h>
  30 #include <linux/interrupt.h>
  31 #include <linux/clockchips.h>
  32 #include <asm/hyperv.h>
  33 #include <asm/mshyperv.h>
  34 #include "hyperv_vmbus.h"
  35
  36 /* The one and only */
  37 struct hv_context hv_context = {
  38         .synic_initialized      = false,
  39         .hypercall_page         = NULL,
  40 };
  41
  42 #define HV_TIMER_FREQUENCY (10 * 1000 * 1000) /* 100ns period */
  43 #define HV_MAX_MAX_DELTA_TICKS 0xffffffff
  44 #define HV_MIN_DELTA_TICKS 1
  45
  46 /*
  47  * query_hypervisor_info - Get version info of the windows hypervisor
  48  */
  49 unsigned int host_info_eax;
  50 unsigned int host_info_ebx;
  51 unsigned int host_info_ecx;
  52 unsigned int host_info_edx;
  53
  54 static int query_hypervisor_info(void)
  55 {
  56         unsigned int eax;
  57         unsigned int ebx;
  58         unsigned int ecx;
  59         unsigned int edx;
  60         unsigned int max_leaf;
  61         unsigned int op;
  62
  63         /*
  64         * Its assumed that this is called after confirming that Viridian
  65         * is present. Query id and revision.
  66         */
  67         eax = 0;
  68         ebx = 0;
  69         ecx = 0;
  70         edx = 0;
  71         op = HVCPUID_VENDOR_MAXFUNCTION;
  72         cpuid(op, &eax, &ebx, &ecx, &edx);
  73
  74         max_leaf = eax;
  75
  76         if (max_leaf >= HVCPUID_VERSION) {
  77                 eax = 0;
  78                 ebx = 0;
  79                 ecx = 0;
  80                 edx = 0;
  81                 op = HVCPUID_VERSION;
  82                 cpuid(op, &eax, &ebx, &ecx, &edx);
  83                 host_info_eax = eax;
  84                 host_info_ebx = ebx;
  85                 host_info_ecx = ecx;
  86                 host_info_edx = edx;
  87         }
  88         return max_leaf;
  89 }
  90
  91 /*
  92  * do_hypercall- Invoke the specified hypercall
  93  */
  94 static u64 do_hypercall(u64 control, void *input, void *output)
  95 {
  96 #ifdef CONFIG_X86_64
  97         u64 hv_status = 0;
  98         u64 input_address = (input) ? virt_to_phys(input) : 0;
  99         u64 output_address = (output) ? virt_to_phys(output) : 0;
 100         void *hypercall_page = hv_context.hypercall_page;
 101
 102         __asm__ __volatile__("mov %0, %%r8" : : "r" (output_address) : "r8");
 103         __asm__ __volatile__("call *%3" : "=a" (hv_status) :
 104                              "c" (control), "d" (input_address),
 105                              "m" (hypercall_page));
 106
 107         return hv_status;
 108
 109 #else
 110
 111         u32 control_hi = control >> 32;
 112         u32 control_lo = control & 0xFFFFFFFF;
 113         u32 hv_status_hi = 1;
 114         u32 hv_status_lo = 1;
 115         u64 input_address = (input) ? virt_to_phys(input) : 0;
 116         u32 input_address_hi = input_address >> 32;
 117         u32 input_address_lo = input_address & 0xFFFFFFFF;
 118         u64 output_address = (output) ? virt_to_phys(output) : 0;
 119         u32 output_address_hi = output_address >> 32;
 120         u32 output_address_lo = output_address & 0xFFFFFFFF;
 121         void *hypercall_page = hv_context.hypercall_page;
 122
 123         __asm__ __volatile__ ("call *%8" : "=d"(hv_status_hi),
 124                               "=a"(hv_status_lo) : "d" (control_hi),
 125                               "a" (control_lo), "b" (input_address_hi),
 126                               "c" (input_address_lo), "D"(output_address_hi),
 127                               "S"(output_address_lo), "m" (hypercall_page));
 128
 129         return hv_status_lo | ((u64)hv_status_hi << 32);
 130 #endif /* !x86_64 */
 131 }
 132
 133 /*
 134  * hv_init - Main initialization routine.
 135  *
 136  * This routine must be called before any other routines in here are called
 137  */
 138 int hv_init(void)
 139 {
 140         int max_leaf;
 141         union hv_x64_msr_hypercall_contents hypercall_msr;
 142         void *virtaddr = NULL;
 143
 144         memset(hv_context.synic_event_page, 0, sizeof(void *) * NR_CPUS);
 145         memset(hv_context.synic_message_page, 0,
 146                sizeof(void *) * NR_CPUS);
 147         memset(hv_context.post_msg_page, 0,
 148                sizeof(void *) * NR_CPUS);
 149         memset(hv_context.vp_index, 0,
 150                sizeof(int) * NR_CPUS);
 151         memset(hv_context.event_dpc, 0,
 152                sizeof(void *) * NR_CPUS);
 153         memset(hv_context.clk_evt, 0,
 154                sizeof(void *) * NR_CPUS);
 155
 156         max_leaf = query_hypervisor_info();
 157
 158         /*
 159          * Write our OS ID.
 160          */
 161         hv_context.guestid = generate_guest_id(0, LINUX_VERSION_CODE, 0);
 162         wrmsrl(HV_X64_MSR_GUEST_OS_ID, hv_context.guestid);
 163
 164         /* See if the hypercall page is already set */
 165         rdmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 166
 167         virtaddr = __vmalloc(PAGE_SIZE, GFP_KERNEL, PAGE_KERNEL_EXEC);
 168
 169         if (!virtaddr)
 170                 goto cleanup;
 171
 172         hypercall_msr.enable = 1;
 173
 174         hypercall_msr.guest_physical_address = vmalloc_to_pfn(virtaddr);
 175         wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 176
 177         /* Confirm that hypercall page did get setup. */
 178         hypercall_msr.as_uint64 = 0;
 179         rdmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 180
 181         if (!hypercall_msr.enable)
 182                 goto cleanup;
 183
 184         hv_context.hypercall_page = virtaddr;
 185
 186         return 0;
 187
 188 cleanup:
 189         if (virtaddr) {
 190                 if (hypercall_msr.enable) {
 191                         hypercall_msr.as_uint64 = 0;
 192                         wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 193                 }
 194
 195                 vfree(virtaddr);
 196         }
 197
 198         return -ENOTSUPP;
 199 }
 200
 201 /*
 202  * hv_cleanup - Cleanup routine.
 203  *
 204  * This routine is called normally during driver unloading or exiting.
 205  */
 206 void hv_cleanup(void)
 207 {
 208         union hv_x64_msr_hypercall_contents hypercall_msr;
 209
 210         /* Reset our OS id */
 211         wrmsrl(HV_X64_MSR_GUEST_OS_ID, 0);
 212
 213         if (hv_context.hypercall_page) {
 214                 hypercall_msr.as_uint64 = 0;
 215                 wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 216                 vfree(hv_context.hypercall_page);
 217                 hv_context.hypercall_page = NULL;
 218         }
 219 }
 220
 221 /*
 222  * hv_post_message - Post a message using the hypervisor message IPC.
 223  *
 224  * This involves a hypercall.
 225  */
 226 int hv_post_message(union hv_connection_id connection_id,
 227                   enum hv_message_type message_type,
 228                   void *payload, size_t payload_size)
 229 {
 230
 231         struct hv_input_post_message *aligned_msg;
 232         u16 status;
 233
 234         if (payload_size > HV_MESSAGE_PAYLOAD_BYTE_COUNT)
 235                 return -EMSGSIZE;
 236
 237         aligned_msg = (struct hv_input_post_message *)
 238                         hv_context.post_msg_page[get_cpu()];
 239
 240         aligned_msg->connectionid = connection_id;
 241         aligned_msg->reserved = 0;
 242         aligned_msg->message_type = message_type;
 243         aligned_msg->payload_size = payload_size;
 244         memcpy((void *)aligned_msg->payload, payload, payload_size);
 245
 246         status = do_hypercall(HVCALL_POST_MESSAGE, aligned_msg, NULL)
 247                 & 0xFFFF;
 248
 249         put_cpu();
 250         return status;
 251 }
 252
 253
 254 /*
 255  * hv_signal_event -
 256  * Signal an event on the specified connection using the hypervisor event IPC.
 257  *
 258  * This involves a hypercall.
 259  */
 260 u16 hv_signal_event(void *con_id)
 261 {
 262         u16 status;
 263
 264         status = (do_hypercall(HVCALL_SIGNAL_EVENT, con_id, NULL) & 0xFFFF);
 265
 266         return status;
 267 }
 268
 269 static int hv_ce_set_next_event(unsigned long delta,
 270                                 struct clock_event_device *evt)
 271 {
 272         cycle_t current_tick;
 273
 274         WARN_ON(evt->mode != CLOCK_EVT_MODE_ONESHOT);
 275
 276         rdmsrl(HV_X64_MSR_TIME_REF_COUNT, current_tick);
 277         current_tick += delta;
 278         wrmsrl(HV_X64_MSR_STIMER0_COUNT, current_tick);
 279         return 0;
 280 }
 281
 282 static void hv_ce_setmode(enum clock_event_mode mode,
 283                           struct clock_event_device *evt)
 284 {
 285         union hv_timer_config timer_cfg;
 286
 287         switch (mode) {
 288         case CLOCK_EVT_MODE_PERIODIC:
 289                 /* unsupported */
 290                 break;
 291
 292         case CLOCK_EVT_MODE_ONESHOT:
 293                 timer_cfg.enable = 1;
 294                 timer_cfg.auto_enable = 1;
 295                 timer_cfg.sintx = VMBUS_MESSAGE_SINT;
 296                 wrmsrl(HV_X64_MSR_STIMER0_CONFIG, timer_cfg.as_uint64);
 297                 break;
 298
 299         case CLOCK_EVT_MODE_UNUSED:
 300         case CLOCK_EVT_MODE_SHUTDOWN:
 301                 wrmsrl(HV_X64_MSR_STIMER0_COUNT, 0);
 302                 wrmsrl(HV_X64_MSR_STIMER0_CONFIG, 0);
 303                 break;
 304         case CLOCK_EVT_MODE_RESUME:
 305                 break;
 306         }
 307 }
 308
 309 static void hv_init_clockevent_device(struct clock_event_device *dev, int cpu)
 310 {
 311         dev->name = "Hyper-V clockevent";
 312         dev->features = CLOCK_EVT_FEAT_ONESHOT;
 313         dev->cpumask = cpumask_of(cpu);
 314         dev->rating = 1000;
 315         /*
 316          * Avoid settint dev->owner = THIS_MODULE deliberately as doing so will
 317          * result in clockevents_config_and_register() taking additional
 318          * references to the hv_vmbus module making it impossible to unload.
 319          */
 320
 321         dev->set_mode = hv_ce_setmode;
 322         dev->set_next_event = hv_ce_set_next_event;
 323 }
 324
 325
 326 int hv_synic_alloc(void)
 327 {
 328         size_t size = sizeof(struct tasklet_struct);
 329         size_t ced_size = sizeof(struct clock_event_device);
 330         int cpu;
 331
 332         for_each_online_cpu(cpu) {
 333                 hv_context.event_dpc[cpu] = kmalloc(size, GFP_ATOMIC);
 334                 if (hv_context.event_dpc[cpu] == NULL) {
 335                         pr_err("Unable to allocate event dpc\n");
 336                         goto err;
 337                 }
 338                 tasklet_init(hv_context.event_dpc[cpu], vmbus_on_event, cpu);
 339
 340                 hv_context.clk_evt[cpu] = kzalloc(ced_size, GFP_ATOMIC);
 341                 if (hv_context.clk_evt[cpu] == NULL) {
 342                         pr_err("Unable to allocate clock event device\n");
 343                         goto err;
 344                 }
 345                 hv_init_clockevent_device(hv_context.clk_evt[cpu], cpu);
 346
 347                 hv_context.synic_message_page[cpu] =
 348                         (void *)get_zeroed_page(GFP_ATOMIC);
 349
 350                 if (hv_context.synic_message_page[cpu] == NULL) {
 351                         pr_err("Unable to allocate SYNIC message page\n");
 352                         goto err;
 353                 }
 354
 355                 hv_context.synic_event_page[cpu] =
 356                         (void *)get_zeroed_page(GFP_ATOMIC);
 357
 358                 if (hv_context.synic_event_page[cpu] == NULL) {
 359                         pr_err("Unable to allocate SYNIC event page\n");
 360                         goto err;
 361                 }
 362
 363                 hv_context.post_msg_page[cpu] =
 364                         (void *)get_zeroed_page(GFP_ATOMIC);
 365
 366                 if (hv_context.post_msg_page[cpu] == NULL) {
 367                         pr_err("Unable to allocate post msg page\n");
 368                         goto err;
 369                 }
 370         }
 371
 372         return 0;
 373 err:
 374         return -ENOMEM;
 375 }
 376
 377 static void hv_synic_free_cpu(int cpu)
 378 {
 379         kfree(hv_context.event_dpc[cpu]);
 380         kfree(hv_context.clk_evt[cpu]);
 381         if (hv_context.synic_event_page[cpu])
 382                 free_page((unsigned long)hv_context.synic_event_page[cpu]);
 383         if (hv_context.synic_message_page[cpu])
 384                 free_page((unsigned long)hv_context.synic_message_page[cpu]);
 385         if (hv_context.post_msg_page[cpu])
 386                 free_page((unsigned long)hv_context.post_msg_page[cpu]);
 387 }
 388
 389 void hv_synic_free(void)
 390 {
 391         int cpu;
 392
 393         for_each_online_cpu(cpu)
 394                 hv_synic_free_cpu(cpu);
 395 }
 396
 397 /*
 398  * hv_synic_init - Initialize the Synthethic Interrupt Controller.
 399  *
 400  * If it is already initialized by another entity (ie x2v shim), we need to
 401  * retrieve the initialized message and event pages.  Otherwise, we create and
 402  * initialize the message and event pages.
 403  */
 404 void hv_synic_init(void *arg)
 405 {
 406         u64 version;
 407         union hv_synic_simp simp;
 408         union hv_synic_siefp siefp;
 409         union hv_synic_sint shared_sint;
 410         union hv_synic_scontrol sctrl;
 411         u64 vp_index;
 412
 413         int cpu = smp_processor_id();
 414
 415         if (!hv_context.hypercall_page)
 416                 return;
 417
 418         /* Check the version */
 419         rdmsrl(HV_X64_MSR_SVERSION, version);
 420
 421         /* Setup the Synic's message page */
 422         rdmsrl(HV_X64_MSR_SIMP, simp.as_uint64);
 423         simp.simp_enabled = 1;
 424         simp.base_simp_gpa = virt_to_phys(hv_context.synic_message_page[cpu])
 425                 >> PAGE_SHIFT;
 426
 427         wrmsrl(HV_X64_MSR_SIMP, simp.as_uint64);
 428
 429         /* Setup the Synic's event page */
 430         rdmsrl(HV_X64_MSR_SIEFP, siefp.as_uint64);
 431         siefp.siefp_enabled = 1;
 432         siefp.base_siefp_gpa = virt_to_phys(hv_context.synic_event_page[cpu])
 433                 >> PAGE_SHIFT;
 434
 435         wrmsrl(HV_X64_MSR_SIEFP, siefp.as_uint64);
 436
 437         /* Setup the shared SINT. */
 438         rdmsrl(HV_X64_MSR_SINT0 + VMBUS_MESSAGE_SINT, shared_sint.as_uint64);
 439
 440         shared_sint.as_uint64 = 0;
 441         shared_sint.vector = HYPERVISOR_CALLBACK_VECTOR;
 442         shared_sint.masked = false;
 443         shared_sint.auto_eoi = true;
 444
 445         wrmsrl(HV_X64_MSR_SINT0 + VMBUS_MESSAGE_SINT, shared_sint.as_uint64);
 446
 447         /* Enable the global synic bit */
 448         rdmsrl(HV_X64_MSR_SCONTROL, sctrl.as_uint64);
 449         sctrl.enable = 1;
 450
 451         wrmsrl(HV_X64_MSR_SCONTROL, sctrl.as_uint64);
 452
 453         hv_context.synic_initialized = true;
 454
 455         /*
 456          * Setup the mapping between Hyper-V's notion
 457          * of cpuid and Linux' notion of cpuid.
 458          * This array will be indexed using Linux cpuid.
 459          */
 460         rdmsrl(HV_X64_MSR_VP_INDEX, vp_index);
 461         hv_context.vp_index[cpu] = (u32)vp_index;
 462
 463         INIT_LIST_HEAD(&hv_context.percpu_list[cpu]);
 464
 465         /*
 466          * Register the per-cpu clockevent source.
 467          */
 468         if (ms_hyperv.features & HV_X64_MSR_SYNTIMER_AVAILABLE)
 469                 clockevents_config_and_register(hv_context.clk_evt[cpu],
 470                                                 HV_TIMER_FREQUENCY,
 471                                                 HV_MIN_DELTA_TICKS,
 472                                                 HV_MAX_MAX_DELTA_TICKS);
 473         return;
 474 }
 475
 476 /*
 477  * hv_synic_clockevents_cleanup - Cleanup clockevent devices
 478  */
 479 void hv_synic_clockevents_cleanup(void)
 480 {
 481         int cpu;
 482
 483         if (!(ms_hyperv.features & HV_X64_MSR_SYNTIMER_AVAILABLE))
 484                 return;
 485
 486         for_each_online_cpu(cpu)
 487                 clockevents_unbind_device(hv_context.clk_evt[cpu], cpu);
 488 }
 489
 490 /*
 491  * hv_synic_cleanup - Cleanup routine for hv_synic_init().
 492  */
 493 void hv_synic_cleanup(void *arg)
 494 {
 495         union hv_synic_sint shared_sint;
 496         union hv_synic_simp simp;
 497         union hv_synic_siefp siefp;
 498         union hv_synic_scontrol sctrl;
 499         int cpu = smp_processor_id();
 500
 501         if (!hv_context.synic_initialized)
 502                 return;
 503
 504         /* Turn off clockevent device */
 505         if (ms_hyperv.features & HV_X64_MSR_SYNTIMER_AVAILABLE)
 506                 hv_ce_setmode(CLOCK_EVT_MODE_SHUTDOWN,
 507                               hv_context.clk_evt[cpu]);
 508
 509         rdmsrl(HV_X64_MSR_SINT0 + VMBUS_MESSAGE_SINT, shared_sint.as_uint64);
 510
 511         shared_sint.masked = 1;
 512
 513         /* Need to correctly cleanup in the case of SMP!!! */
 514         /* Disable the interrupt */
 515         wrmsrl(HV_X64_MSR_SINT0 + VMBUS_MESSAGE_SINT, shared_sint.as_uint64);
 516
 517         rdmsrl(HV_X64_MSR_SIMP, simp.as_uint64);
 518         simp.simp_enabled = 0;
 519         simp.base_simp_gpa = 0;
 520
 521         wrmsrl(HV_X64_MSR_SIMP, simp.as_uint64);
 522
 523         rdmsrl(HV_X64_MSR_SIEFP, siefp.as_uint64);
 524         siefp.siefp_enabled = 0;
 525         siefp.base_siefp_gpa = 0;
 526
 527         wrmsrl(HV_X64_MSR_SIEFP, siefp.as_uint64);
 528
 529         /* Disable the global synic bit */
 530         rdmsrl(HV_X64_MSR_SCONTROL, sctrl.as_uint64);
 531         sctrl.enable = 0;
 532         wrmsrl(HV_X64_MSR_SCONTROL, sctrl.as_uint64);
 533
 534         hv_synic_free_cpu(cpu);
 535 }