| 1 |
/*
|
| 2 |
* Copyright (c) 2007, 2008 University of Tsukuba
|
| 3 |
* All rights reserved.
|
| 4 |
*
|
| 5 |
* Redistribution and use in source and binary forms, with or without
|
| 6 |
* modification, are permitted provided that the following conditions are met:
|
| 7 |
*
|
| 8 |
* 1. Redistributions of source code must retain the above copyright notice,
|
| 9 |
* this list of conditions and the following disclaimer.
|
| 10 |
* 2. Redistributions in binary form must reproduce the above copyright notice,
|
| 11 |
* this list of conditions and the following disclaimer in the documentation
|
| 12 |
* and/or other materials provided with the distribution.
|
| 13 |
* 3. Neither the name of the University of Tsukuba nor the names of its
|
| 14 |
* contributors may be used to endorse or promote products derived from
|
| 15 |
* this software without specific prior written permission.
|
| 16 |
*
|
| 17 |
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
| 18 |
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
| 19 |
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
| 20 |
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
|
| 21 |
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
| 22 |
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
| 23 |
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
| 24 |
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
| 25 |
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
| 26 |
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
| 27 |
* POSSIBILITY OF SUCH DAMAGE.
|
| 28 |
*/
|
| 29 |
/*
|
| 30 |
* Copyright (c) 2010-2012 Yuichi Watanabe
|
| 31 |
*/
|
| 32 |
|
| 33 |
/* MMU emulation, Shadow Page Tables (SPT) */
|
| 34 |
|
| 35 |
#include <common/list.h>
|
| 36 |
#include <core/initfunc.h>
|
| 37 |
#include <core/printf.h>
|
| 38 |
#include <core/spinlock.h>
|
| 39 |
#include <core/string.h>
|
| 40 |
#include "asm.h"
|
| 41 |
#include "constants.h"
|
| 42 |
#include "cpu_mmu.h"
|
| 43 |
#include "cpu_mmu_spt.h"
|
| 44 |
#include "current.h"
|
| 45 |
#include "mm.h"
|
| 46 |
#include "mmio.h"
|
| 47 |
#include "msr_pass.h"
|
| 48 |
#include "panic.h"
|
| 49 |
#include "pcpu.h"
|
| 50 |
#include "vmmcall_status.h"
|
| 51 |
|
| 52 |
struct map_page_data1 {
|
| 53 |
unsigned int write : 1;
|
| 54 |
unsigned int user : 1;
|
| 55 |
unsigned int wp : 1;
|
| 56 |
u32 cache_flag;
|
| 57 |
};
|
| 58 |
|
| 59 |
struct map_page_data2 {
|
| 60 |
unsigned int rw : 1;
|
| 61 |
unsigned int us : 1;
|
| 62 |
unsigned int nx : 1;
|
| 63 |
};
|
| 64 |
|
| 65 |
#define GFN_GLOBAL 0xFFFFFFFFFFFFFFFEULL
|
| 66 |
#define GFN_UNUSED 0xFFFFFFFFFFFFFFFFULL
|
| 67 |
|
| 68 |
static bool guest64 (void);
|
| 69 |
static void update_cr3 (void);
|
| 70 |
static void invalidate_page (ulong virtual_addr);
|
| 71 |
static void map_page (u64 v, struct map_page_data1 m1,
|
| 72 |
struct map_page_data2 m2[5], u64 gfns[5], int glvl,
|
| 73 |
phys_t hphys);
|
| 74 |
static void init_vcpu(void);
|
| 75 |
|
| 76 |
static void
|
| 77 |
invalidate_page (ulong v)
|
| 78 |
{
|
| 79 |
pmap_t p;
|
| 80 |
u64 tmp;
|
| 81 |
|
| 82 |
pmap_open_vmm (&p, current->spt.cr3tbl_phys, current->spt.levels);
|
| 83 |
pmap_setvirt (&p, v, 2);
|
| 84 |
tmp = pmap_read (&p);
|
| 85 |
if (tmp & PDE_P_BIT) {
|
| 86 |
if (tmp & PDE_AVAILABLE1_BIT) {
|
| 87 |
pmap_write (&p, tmp & ~PDE_AVAILABLE1_BIT, 0xFFF);
|
| 88 |
pmap_setlevel (&p, 1);
|
| 89 |
pmap_clear (&p);
|
| 90 |
if (current->spt.levels >= 3) {
|
| 91 |
pmap_setvirt (&p, v ^ PAGESIZE2M, 2);
|
| 92 |
tmp = pmap_read (&p);
|
| 93 |
if (tmp & PDE_P_BIT) {
|
| 94 |
pmap_write (&p,
|
| 95 |
tmp & ~PDE_AVAILABLE1_BIT,
|
| 96 |
0xFFF);
|
| 97 |
pmap_setlevel (&p, 1);
|
| 98 |
pmap_clear (&p);
|
| 99 |
}
|
| 100 |
}
|
| 101 |
} else {
|
| 102 |
pmap_setlevel (&p, 1);
|
| 103 |
tmp = pmap_read (&p);
|
| 104 |
pmap_write (&p, 0, 0xFFF);
|
| 105 |
}
|
| 106 |
}
|
| 107 |
pmap_close (&p);
|
| 108 |
}
|
| 109 |
|
| 110 |
static void
|
| 111 |
invalidate_all_spt(void)
|
| 112 |
{
|
| 113 |
pmap_t p;
|
| 114 |
struct spt_page *spt_page;
|
| 115 |
|
| 116 |
pmap_open_vmm (&p, current->spt.cr3tbl_phys, current->spt.levels);
|
| 117 |
pmap_clear (&p);
|
| 118 |
pmap_close (&p);
|
| 119 |
while((spt_page = LIST2_POP(current->spt.shadow_l1, list)) != NULL) {
|
| 120 |
spt_page->pte_virt = NULL;
|
| 121 |
LIST2_ADD(current->spt.shadow_free, list, spt_page);
|
| 122 |
}
|
| 123 |
while((spt_page = LIST2_POP(current->spt.shadow_lu, list)) != NULL) {
|
| 124 |
spt_page->pte_virt = NULL;
|
| 125 |
LIST2_ADD(current->spt.shadow_free, list, spt_page);
|
| 126 |
}
|
| 127 |
}
|
| 128 |
|
| 129 |
static void
|
| 130 |
map_page (u64 v, struct map_page_data1 m1, struct map_page_data2 m2[5],
|
| 131 |
u64 gfns[5], int glvl, phys_t hphys)
|
| 132 |
{
|
| 133 |
pmap_t p;
|
| 134 |
u64 tmp;
|
| 135 |
int l;
|
| 136 |
ulong cr0;
|
| 137 |
struct spt_page *spt_page;
|
| 138 |
|
| 139 |
current->vmctl.read_control_reg (CONTROL_REG_CR0, &cr0);
|
| 140 |
if (!(cr0 & CR0_WP_BIT)) {
|
| 141 |
if (!m1.user && m1.write) {
|
| 142 |
m2[0].us = 0;
|
| 143 |
m2[0].rw = 1;
|
| 144 |
}
|
| 145 |
}
|
| 146 |
pmap_open_vmm (&p, current->spt.cr3tbl_phys, current->spt.levels);
|
| 147 |
pmap_setvirt (&p, v, 1);
|
| 148 |
tmp = pmap_read (&p);
|
| 149 |
for (; (l = pmap_getreadlevel (&p)) > 1; tmp = pmap_read (&p)) {
|
| 150 |
pmap_setlevel (&p, l);
|
| 151 |
spt_page = LIST2_POP(current->spt.shadow_free, list);
|
| 152 |
if (spt_page == NULL) {
|
| 153 |
spt_page = LIST2_POP(current->spt.shadow_l1, list);
|
| 154 |
if (spt_page == NULL) {
|
| 155 |
printf("DEBUG: shadow_free and shadow_l1 are empty.");
|
| 156 |
invalidate_all_spt();
|
| 157 |
continue;
|
| 158 |
}
|
| 159 |
if (spt_page->pte_virt) {
|
| 160 |
*spt_page->pte_virt = 0;
|
| 161 |
}
|
| 162 |
}
|
| 163 |
spt_page->pte_virt = pmap_pointer(&p);
|
| 164 |
if (l == 1) {
|
| 165 |
LIST2_ADD(current->spt.shadow_l1, list, spt_page);
|
| 166 |
} else {
|
| 167 |
LIST2_ADD(current->spt.shadow_lu, list, spt_page);
|
| 168 |
}
|
| 169 |
pmap_write (&p, spt_page->phys | PDE_P_BIT, PDE_P_BIT);
|
| 170 |
pmap_setlevel (&p, l - 1);
|
| 171 |
pmap_clear (&p);
|
| 172 |
pmap_setlevel (&p, 1);
|
| 173 |
}
|
| 174 |
if (glvl > 1 && (gfns[1] == GFN_GLOBAL || gfns[1] == GFN_UNUSED)) {
|
| 175 |
pmap_setlevel (&p, 2);
|
| 176 |
tmp = pmap_read (&p);
|
| 177 |
if (!(tmp & PDE_AVAILABLE1_BIT))
|
| 178 |
pmap_write (&p, tmp | PDE_AVAILABLE1_BIT, 0xFFF);
|
| 179 |
pmap_setlevel (&p, 1);
|
| 180 |
}
|
| 181 |
pmap_write (&p, (hphys & (~PAGESIZE_MASK)) | PTE_P_BIT |
|
| 182 |
(m2[0].rw ? PTE_RW_BIT : 0) |
|
| 183 |
(m2[0].us ? PTE_US_BIT : 0) |
|
| 184 |
(m2[0].nx ? PTE_NX_BIT : 0) |
|
| 185 |
mm_cache_flag_to_pte_attr (m1.cache_flag),
|
| 186 |
PTE_P_BIT | PTE_RW_BIT | PTE_US_BIT |
|
| 187 |
PTE_PWT_BIT | PTE_PCD_BIT | PTE_PAT_BIT);
|
| 188 |
pmap_close (&p);
|
| 189 |
}
|
| 190 |
|
| 191 |
static void
|
| 192 |
update_cr3 (void)
|
| 193 |
{
|
| 194 |
current->spt.levels = guest64 () ? 4 : 3;
|
| 195 |
invalidate_all_spt();
|
| 196 |
current->vmctl.spt_setcr3 (current->spt.cr3tbl_phys);
|
| 197 |
}
|
| 198 |
|
| 199 |
static bool
|
| 200 |
spt_tlbflush (void)
|
| 201 |
{
|
| 202 |
return true;
|
| 203 |
}
|
| 204 |
|
| 205 |
static void
|
| 206 |
init_vcpu(void)
|
| 207 |
{
|
| 208 |
int i;
|
| 209 |
vmmerr_t err;
|
| 210 |
|
| 211 |
err = alloc_page(¤t->spt.cr3tbl, ¤t->spt.cr3tbl_phys);
|
| 212 |
if (err) {
|
| 213 |
panic("Failed to allocate memory for a cr3-table.");
|
| 214 |
}
|
| 215 |
memset(current->spt.cr3tbl, 0, PAGESIZE);
|
| 216 |
LIST2_HEAD_INIT(current->spt.shadow_free);
|
| 217 |
LIST2_HEAD_INIT(current->spt.shadow_l1);
|
| 218 |
LIST2_HEAD_INIT(current->spt.shadow_lu);
|
| 219 |
for (i = 0; i < NUM_OF_SPTTBL; i++) {
|
| 220 |
err = alloc_page(¤t->spt.spt_page[i].virt, ¤t->spt.spt_page[i].phys);
|
| 221 |
if (err) {
|
| 222 |
panic("Failed to allocate memory for a shadow page table.");
|
| 223 |
}
|
| 224 |
current->spt.spt_page[i].pte_virt = NULL;
|
| 225 |
LIST2_ADD(current->spt.shadow_free, list, ¤t->spt.spt_page[i]);
|
| 226 |
}
|
| 227 |
}
|
| 228 |
|
| 229 |
static bool
|
| 230 |
guest64 (void)
|
| 231 |
{
|
| 232 |
u64 efer;
|
| 233 |
|
| 234 |
current->vmctl.read_msr (MSR_IA32_EFER, &efer);
|
| 235 |
if (efer & MSR_IA32_EFER_LMA_BIT)
|
| 236 |
return true;
|
| 237 |
return false;
|
| 238 |
}
|
| 239 |
|
| 240 |
/* this function is called when shadow page table entries are cleared
|
| 241 |
from TLB. every VM exit in VT. */
|
| 242 |
bool
|
| 243 |
cpu_mmu_spt_tlbflush (void)
|
| 244 |
{
|
| 245 |
return spt_tlbflush ();
|
| 246 |
}
|
| 247 |
|
| 248 |
/* this function is called when a guest sets CR3 */
|
| 249 |
void
|
| 250 |
cpu_mmu_spt_updatecr3 (void)
|
| 251 |
{
|
| 252 |
update_cr3 ();
|
| 253 |
}
|
| 254 |
|
| 255 |
/* this function is called by INVLPG in a guest */
|
| 256 |
/* deletes TLB entry regardless of G bit */
|
| 257 |
void
|
| 258 |
cpu_mmu_spt_invalidate (ulong virtual_addr)
|
| 259 |
{
|
| 260 |
invalidate_page (virtual_addr);
|
| 261 |
}
|
| 262 |
|
| 263 |
static void
|
| 264 |
set_m1 (u64 entry0, bool write, bool user, bool wp, struct map_page_data1 *m1)
|
| 265 |
{
|
| 266 |
m1->write = write;
|
| 267 |
m1->user = user;
|
| 268 |
m1->wp = wp;
|
| 269 |
m1->cache_flag = msr_pte_to_cache_flag(entry0);
|
| 270 |
}
|
| 271 |
|
| 272 |
static void
|
| 273 |
set_m2 (u64 entries[5], int levels, struct map_page_data2 m2[5])
|
| 274 |
{
|
| 275 |
unsigned int rw = 1, us = 1, nx = 0;
|
| 276 |
int i;
|
| 277 |
u64 entry;
|
| 278 |
|
| 279 |
for (i = levels; i >= 0; i--) {
|
| 280 |
entry = entries[i];
|
| 281 |
if (!(entry & PDE_RW_BIT))
|
| 282 |
rw = 0;
|
| 283 |
if (!(entry & PDE_US_BIT))
|
| 284 |
us = 0;
|
| 285 |
if (entry & PDE_NX_BIT)
|
| 286 |
nx = 1;
|
| 287 |
m2[i].rw = rw;
|
| 288 |
m2[i].us = us;
|
| 289 |
m2[i].nx = nx;
|
| 290 |
}
|
| 291 |
if (!(entries[0] & PTE_D_BIT))
|
| 292 |
m2[0].rw = 0;
|
| 293 |
}
|
| 294 |
|
| 295 |
static void
|
| 296 |
set_gfns (u64 entries[5], int levels, u64 gfns[5])
|
| 297 |
{
|
| 298 |
int i;
|
| 299 |
u64 entry;
|
| 300 |
|
| 301 |
for (i = levels; i >= 0; i--) {
|
| 302 |
entry = entries[i];
|
| 303 |
if (i == 1 && (entry & PDE_PS_BIT)) {
|
| 304 |
if (entry & PDE_G_BIT)
|
| 305 |
gfns[i] = GFN_GLOBAL;
|
| 306 |
else
|
| 307 |
gfns[i] = GFN_UNUSED;
|
| 308 |
continue;
|
| 309 |
}
|
| 310 |
gfns[i] = (entry & PTE_ADDR_MASK64) >> PAGESIZE_SHIFT;
|
| 311 |
}
|
| 312 |
}
|
| 313 |
|
| 314 |
/* handling a page fault of a guest */
|
| 315 |
void
|
| 316 |
cpu_mmu_spt_pagefault (ulong err, ulong cr2)
|
| 317 |
{
|
| 318 |
int levels;
|
| 319 |
vmmerr_t ret;
|
| 320 |
bool wr, us, ex, wp;
|
| 321 |
ulong cr0, cr3, cr4;
|
| 322 |
struct map_page_data1 m1;
|
| 323 |
struct map_page_data2 m2[5];
|
| 324 |
u64 efer, gfns[5], entries[5];
|
| 325 |
phys_t hphys;
|
| 326 |
|
| 327 |
current->vmctl.read_control_reg(CONTROL_REG_CR0, &cr0);
|
| 328 |
current->vmctl.read_control_reg(CONTROL_REG_CR3, &cr3);
|
| 329 |
current->vmctl.read_control_reg(CONTROL_REG_CR4, &cr4);
|
| 330 |
current->vmctl.read_msr(MSR_IA32_EFER, &efer);
|
| 331 |
wr = !!(err & PAGEFAULT_ERR_WR_BIT);
|
| 332 |
us = !!(err & PAGEFAULT_ERR_US_BIT);
|
| 333 |
ex = !!(err & PAGEFAULT_ERR_ID_BIT);
|
| 334 |
wp = !!(cr0 & CR0_WP_BIT);
|
| 335 |
ret = mmu_get_pte(cr2, cr0, cr3, cr4, efer, wr, us, ex, entries,
|
| 336 |
&levels);
|
| 337 |
if (ret != VMMERR_SUCCESS) {
|
| 338 |
mmu_generate_pagefault(ret, err & PAGEFAULT_ERR_WR_BIT,
|
| 339 |
err & PAGEFAULT_ERR_US_BIT, cr2);
|
| 340 |
return;
|
| 341 |
}
|
| 342 |
set_m1(entries[0], wr, us, wp, &m1);
|
| 343 |
set_m2(entries, levels, m2);
|
| 344 |
set_gfns(entries, levels, gfns);
|
| 345 |
|
| 346 |
mmio_lock();
|
| 347 |
ret = mmio_pagefault(gfns[0] << PAGESIZE_SHIFT);
|
| 348 |
mmio_unlock();
|
| 349 |
if (ret != VMMERR_NODEV) {
|
| 350 |
if (ret != VMMERR_SUCCESS &&
|
| 351 |
(ret < VMMERR_PAGE_NOT_PRESENT ||
|
| 352 |
ret > VMMERR_PAGE_BAD_RESERVED_BIT)) {
|
| 353 |
panic("Failed to emulate accessing to MMIO area. ret 0x%x",
|
| 354 |
ret);
|
| 355 |
}
|
| 356 |
return;
|
| 357 |
}
|
| 358 |
|
| 359 |
hphys = current->vm->gmm.gp2hp((gfns[0] << PAGESIZE_SHIFT) +
|
| 360 |
(cr2 & PAGESIZE_MASK));
|
| 361 |
if (hphys == GMM_NO_MAPPING) {
|
| 362 |
ret = cpu_interpreter();
|
| 363 |
if (ret != VMMERR_SUCCESS &&
|
| 364 |
(ret < VMMERR_PAGE_NOT_PRESENT ||
|
| 365 |
ret > VMMERR_PAGE_BAD_RESERVED_BIT)) {
|
| 366 |
panic("Failed to emulate accessing to no mapping area. ret 0x%x",
|
| 367 |
ret);
|
| 368 |
}
|
| 369 |
return;
|
| 370 |
}
|
| 371 |
map_page(cr2, m1, m2, gfns, levels, hphys);
|
| 372 |
}
|
| 373 |
|
| 374 |
void
|
| 375 |
cpu_mmu_spt_init(void)
|
| 376 |
{
|
| 377 |
init_vcpu();
|
| 378 |
}
|