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 <asm/hyperv.h>
  32 #include "hyperv_vmbus.h"
  33
  34 /* The one and only */
  35 struct hv_context hv_context = {
  36         .synic_initialized      = false,
  37         .hypercall_page         = NULL,
  38 };
  39
  40 /*
  41  * query_hypervisor_info - Get version info of the windows hypervisor
  42  */
  43 unsigned int host_info_eax;
  44 unsigned int host_info_ebx;
  45 unsigned int host_info_ecx;
  46 unsigned int host_info_edx;
  47
  48 static int query_hypervisor_info(void)
  49 {
  50         unsigned int eax;
  51         unsigned int ebx;
  52         unsigned int ecx;
  53         unsigned int edx;
  54         unsigned int max_leaf;
  55         unsigned int op;
  56
  57         /*
  58         * Its assumed that this is called after confirming that Viridian
  59         * is present. Query id and revision.
  60         */
  61         eax = 0;
  62         ebx = 0;
  63         ecx = 0;
  64         edx = 0;
  65         op = HVCPUID_VENDOR_MAXFUNCTION;
  66         cpuid(op, &eax, &ebx, &ecx, &edx);
  67
  68         max_leaf = eax;
  69
  70         if (max_leaf >= HVCPUID_VERSION) {
  71                 eax = 0;
  72                 ebx = 0;
  73                 ecx = 0;
  74                 edx = 0;
  75                 op = HVCPUID_VERSION;
  76                 cpuid(op, &eax, &ebx, &ecx, &edx);
  77                 host_info_eax = eax;
  78                 host_info_ebx = ebx;
  79                 host_info_ecx = ecx;
  80                 host_info_edx = edx;
  81         }
  82         return max_leaf;
  83 }
  84
  85 /*
  86  * do_hypercall- Invoke the specified hypercall
  87  */
  88 static u64 do_hypercall(u64 control, void *input, void *output)
  89 {
  90 #ifdef CONFIG_X86_64
  91         u64 hv_status = 0;
  92         u64 input_address = (input) ? virt_to_phys(input) : 0;
  93         u64 output_address = (output) ? virt_to_phys(output) : 0;
  94         void *hypercall_page = hv_context.hypercall_page;
  95
  96         __asm__ __volatile__("mov %0, %%r8" : : "r" (output_address) : "r8");
  97         __asm__ __volatile__("call *%3" : "=a" (hv_status) :
  98                              "c" (control), "d" (input_address),
  99                              "m" (hypercall_page));
 100
 101         return hv_status;
 102
 103 #else
 104
 105         u32 control_hi = control >> 32;
 106         u32 control_lo = control & 0xFFFFFFFF;
 107         u32 hv_status_hi = 1;
 108         u32 hv_status_lo = 1;
 109         u64 input_address = (input) ? virt_to_phys(input) : 0;
 110         u32 input_address_hi = input_address >> 32;
 111         u32 input_address_lo = input_address & 0xFFFFFFFF;
 112         u64 output_address = (output) ? virt_to_phys(output) : 0;
 113         u32 output_address_hi = output_address >> 32;
 114         u32 output_address_lo = output_address & 0xFFFFFFFF;
 115         void *hypercall_page = hv_context.hypercall_page;
 116
 117         __asm__ __volatile__ ("call *%8" : "=d"(hv_status_hi),
 118                               "=a"(hv_status_lo) : "d" (control_hi),
 119                               "a" (control_lo), "b" (input_address_hi),
 120                               "c" (input_address_lo), "D"(output_address_hi),
 121                               "S"(output_address_lo), "m" (hypercall_page));
 122
 123         return hv_status_lo | ((u64)hv_status_hi << 32);
 124 #endif /* !x86_64 */
 125 }
 126
 127 /*
 128  * hv_init - Main initialization routine.
 129  *
 130  * This routine must be called before any other routines in here are called
 131  */
 132 int hv_init(void)
 133 {
 134         int max_leaf;
 135         union hv_x64_msr_hypercall_contents hypercall_msr;
 136         void *virtaddr = NULL;
 137
 138         memset(hv_context.synic_event_page, 0, sizeof(void *) * NR_CPUS);
 139         memset(hv_context.synic_message_page, 0,
 140                sizeof(void *) * NR_CPUS);
 141         memset(hv_context.vp_index, 0,
 142                sizeof(int) * NR_CPUS);
 143         memset(hv_context.event_dpc, 0,
 144                sizeof(void *) * NR_CPUS);
 145
 146         max_leaf = query_hypervisor_info();
 147
 148         /*
 149          * Write our OS ID.
 150          */
 151         hv_context.guestid = generate_guest_id(0, LINUX_VERSION_CODE, 0);
 152         wrmsrl(HV_X64_MSR_GUEST_OS_ID, hv_context.guestid);
 153
 154         /* See if the hypercall page is already set */
 155         rdmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 156
 157         virtaddr = __vmalloc(PAGE_SIZE, GFP_KERNEL, PAGE_KERNEL_EXEC);
 158
 159         if (!virtaddr)
 160                 goto cleanup;
 161
 162         hypercall_msr.enable = 1;
 163
 164         hypercall_msr.guest_physical_address = vmalloc_to_pfn(virtaddr);
 165         wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 166
 167         /* Confirm that hypercall page did get setup. */
 168         hypercall_msr.as_uint64 = 0;
 169         rdmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 170
 171         if (!hypercall_msr.enable)
 172                 goto cleanup;
 173
 174         hv_context.hypercall_page = virtaddr;
 175
 176         return 0;
 177
 178 cleanup:
 179         if (virtaddr) {
 180                 if (hypercall_msr.enable) {
 181                         hypercall_msr.as_uint64 = 0;
 182                         wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 183                 }
 184
 185                 vfree(virtaddr);
 186         }
 187
 188         return -ENOTSUPP;
 189 }
 190
 191 /*
 192  * hv_cleanup - Cleanup routine.
 193  *
 194  * This routine is called normally during driver unloading or exiting.
 195  */
 196 void hv_cleanup(void)
 197 {
 198         union hv_x64_msr_hypercall_contents hypercall_msr;
 199
 200         /* Reset our OS id */
 201         wrmsrl(HV_X64_MSR_GUEST_OS_ID, 0);
 202
 203         if (hv_context.hypercall_page) {
 204                 hypercall_msr.as_uint64 = 0;
 205                 wrmsrl(HV_X64_MSR_HYPERCALL, hypercall_msr.as_uint64);
 206                 vfree(hv_context.hypercall_page);
 207                 hv_context.hypercall_page = NULL;
 208         }
 209 }
 210
 211 /*
 212  * hv_post_message - Post a message using the hypervisor message IPC.
 213  *
 214  * This involves a hypercall.
 215  */
 216 int hv_post_message(union hv_connection_id connection_id,
 217                   enum hv_message_type message_type,
 218                   void *payload, size_t payload_size)
 219 {
 220         struct aligned_input {
 221                 u64 alignment8;
 222                 struct hv_input_post_message msg;
 223         };
 224
 225         struct hv_input_post_message *aligned_msg;
 226         u16 status;
 227         unsigned long addr;
 228
 229         if (payload_size > HV_MESSAGE_PAYLOAD_BYTE_COUNT)
 230                 return -EMSGSIZE;
 231
 232         addr = (unsigned long)kmalloc(sizeof(struct aligned_input), GFP_ATOMIC);
 233         if (!addr)
 234                 return -ENOMEM;
 235
 236         aligned_msg = (struct hv_input_post_message *)
 237                         (ALIGN(addr, HV_HYPERCALL_PARAM_ALIGN));
 238
 239         aligned_msg->connectionid = connection_id;
 240         aligned_msg->message_type = message_type;
 241         aligned_msg->payload_size = payload_size;
 242         memcpy((void *)aligned_msg->payload, payload, payload_size);
 243
 244         status = do_hypercall(HVCALL_POST_MESSAGE, aligned_msg, NULL)
 245                 & 0xFFFF;
 246
 247         kfree((void *)addr);
 248
 249         return status;
 250 }
 251
 252
 253 /*
 254  * hv_signal_event -
 255  * Signal an event on the specified connection using the hypervisor event IPC.
 256  *
 257  * This involves a hypercall.
 258  */
 259 u16 hv_signal_event(void *con_id)
 260 {
 261         u16 status;
 262
 263         status = (do_hypercall(HVCALL_SIGNAL_EVENT, con_id, NULL) & 0xFFFF);
 264
 265         return status;
 266 }
 267
 268
 269 int hv_synic_alloc(void)
 270 {
 271         size_t size = sizeof(struct tasklet_struct);
 272         int cpu;
 273
 274         for_each_online_cpu(cpu) {
 275                 hv_context.event_dpc[cpu] = kmalloc(size, GFP_ATOMIC);
 276                 if (hv_context.event_dpc[cpu] == NULL) {
 277                         pr_err("Unable to allocate event dpc\n");
 278                         goto err;
 279                 }
 280                 tasklet_init(hv_context.event_dpc[cpu], vmbus_on_event, cpu);
 281
 282                 hv_context.synic_message_page[cpu] =
 283                         (void *)get_zeroed_page(GFP_ATOMIC);
 284
 285                 if (hv_context.synic_message_page[cpu] == NULL) {
 286                         pr_err("Unable to allocate SYNIC message page\n");
 287                         goto err;
 288                 }
 289
 290                 hv_context.synic_event_page[cpu] =
 291                         (void *)get_zeroed_page(GFP_ATOMIC);
 292
 293                 if (hv_context.synic_event_page[cpu] == NULL) {
 294                         pr_err("Unable to allocate SYNIC event page\n");
 295                         goto err;
 296                 }
 297         }
 298
 299         return 0;
 300 err:
 301         return -ENOMEM;
 302 }
 303
 304 void hv_synic_free_cpu(int cpu)
 305 {
 306         kfree(hv_context.event_dpc[cpu]);
 307         if (hv_context.synic_event_page[cpu])
 308                 free_page((unsigned long)hv_context.synic_event_page[cpu]);
 309         if (hv_context.synic_message_page[cpu])
 310                 free_page((unsigned long)hv_context.synic_message_page[cpu]);
 311 }
 312
 313 void hv_synic_free(void)
 314 {
 315         int cpu;
 316
 317         for_each_online_cpu(cpu)
 318                 hv_synic_free_cpu(cpu);
 319 }
 320
 321 /*
 322  * hv_synic_init - Initialize the Synthethic Interrupt Controller.
 323  *
 324  * If it is already initialized by another entity (ie x2v shim), we need to
 325  * retrieve the initialized message and event pages.  Otherwise, we create and
 326  * initialize the message and event pages.
 327  */
 328 void hv_synic_init(void *arg)
 329 {
 330         u64 version;
 331         union hv_synic_simp simp;
 332         union hv_synic_siefp siefp;
 333         union hv_synic_sint shared_sint;
 334         union hv_synic_scontrol sctrl;
 335         u64 vp_index;
 336
 337         int cpu = smp_processor_id();
 338
 339         if (!hv_context.hypercall_page)
 340                 return;
 341
 342         /* Check the version */
 343         rdmsrl(HV_X64_MSR_SVERSION, version);
 344
 345         /* Setup the Synic's message page */
 346         rdmsrl(HV_X64_MSR_SIMP, simp.as_uint64);
 347         simp.simp_enabled = 1;
 348         simp.base_simp_gpa = virt_to_phys(hv_context.synic_message_page[cpu])
 349                 >> PAGE_SHIFT;
 350
 351         wrmsrl(HV_X64_MSR_SIMP, simp.as_uint64);
 352
 353         /* Setup the Synic's event page */
 354         rdmsrl(HV_X64_MSR_SIEFP, siefp.as_uint64);
 355         siefp.siefp_enabled = 1;
 356         siefp.base_siefp_gpa = virt_to_phys(hv_context.synic_event_page[cpu])
 357                 >> PAGE_SHIFT;
 358
 359         wrmsrl(HV_X64_MSR_SIEFP, siefp.as_uint64);
 360
 361         /* Setup the shared SINT. */
 362         rdmsrl(HV_X64_MSR_SINT0 + VMBUS_MESSAGE_SINT, shared_sint.as_uint64);
 363
 364         shared_sint.as_uint64 = 0;
 365         shared_sint.vector = HYPERVISOR_CALLBACK_VECTOR;
 366         shared_sint.masked = false;
 367         shared_sint.auto_eoi = true;
 368
 369         wrmsrl(HV_X64_MSR_SINT0 + VMBUS_MESSAGE_SINT, shared_sint.as_uint64);
 370
 371         /* Enable the global synic bit */
 372         rdmsrl(HV_X64_MSR_SCONTROL, sctrl.as_uint64);
 373         sctrl.enable = 1;
 374
 375         wrmsrl(HV_X64_MSR_SCONTROL, sctrl.as_uint64);
 376
 377         hv_context.synic_initialized = true;
 378
 379         /*
 380          * Setup the mapping between Hyper-V's notion
 381          * of cpuid and Linux' notion of cpuid.
 382          * This array will be indexed using Linux cpuid.
 383          */
 384         rdmsrl(HV_X64_MSR_VP_INDEX, vp_index);
 385         hv_context.vp_index[cpu] = (u32)vp_index;
 386         return;
 387 }
 388
 389 /*
 390  * hv_synic_cleanup - Cleanup routine for hv_synic_init().
 391  */
 392 void hv_synic_cleanup(void *arg)
 393 {
 394         union hv_synic_sint shared_sint;
 395         union hv_synic_simp simp;
 396         union hv_synic_siefp siefp;
 397         int cpu = smp_processor_id();
 398
 399         if (!hv_context.synic_initialized)
 400                 return;
 401
 402         rdmsrl(HV_X64_MSR_SINT0 + VMBUS_MESSAGE_SINT, shared_sint.as_uint64);
 403
 404         shared_sint.masked = 1;
 405
 406         /* Need to correctly cleanup in the case of SMP!!! */
 407         /* Disable the interrupt */
 408         wrmsrl(HV_X64_MSR_SINT0 + VMBUS_MESSAGE_SINT, shared_sint.as_uint64);
 409
 410         rdmsrl(HV_X64_MSR_SIMP, simp.as_uint64);
 411         simp.simp_enabled = 0;
 412         simp.base_simp_gpa = 0;
 413
 414         wrmsrl(HV_X64_MSR_SIMP, simp.as_uint64);
 415
 416         rdmsrl(HV_X64_MSR_SIEFP, siefp.as_uint64);
 417         siefp.siefp_enabled = 0;
 418         siefp.base_siefp_gpa = 0;
 419
 420         wrmsrl(HV_X64_MSR_SIEFP, siefp.as_uint64);
 421
 422         free_page((unsigned long)hv_context.synic_message_page[cpu]);
 423         free_page((unsigned long)hv_context.synic_event_page[cpu]);
 424 }