1/* 2 * Copyright (c) 2014 Advanced Micro Devices, Inc. 3 * Copyright (c) 2007 The Hewlett-Packard Development Company 4 * All rights reserved. 5 * 6 * The license below extends only to copyright in the software and shall 7 * not be construed as granting a license to any other intellectual 8 * property including but not limited to intellectual property relating 9 * to a hardware implementation of the functionality of the software 10 * licensed hereunder. You may use the software subject to the license 11 * terms below provided that you ensure that this notice is replicated 12 * unmodified and in its entirety in all distributions of the software, 13 * modified or unmodified, in source code or in binary form. 14 * 15 * Copyright (c) 2003-2006 The Regents of The University of Michigan 16 * All rights reserved. 17 * 18 * Redistribution and use in source and binary forms, with or without 19 * modification, are permitted provided that the following conditions are 20 * met: redistributions of source code must retain the above copyright 21 * notice, this list of conditions and the following disclaimer; 22 * redistributions in binary form must reproduce the above copyright 23 * notice, this list of conditions and the following disclaimer in the 24 * documentation and/or other materials provided with the distribution; 25 * neither the name of the copyright holders nor the names of its 26 * contributors may be used to endorse or promote products derived from 27 * this software without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 30 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 31 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 32 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 33 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 34 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 35 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 36 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 37 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 38 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 39 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 40 * 41 * Authors: Gabe Black 42 * Ali Saidi 43 */ 44 45#include "arch/x86/process.hh" 46 47#include <string> 48#include <vector> 49 50#include "arch/x86/isa_traits.hh" 51#include "arch/x86/regs/misc.hh" 52#include "arch/x86/regs/segment.hh" 53#include "arch/x86/system.hh" 54#include "arch/x86/types.hh" 55#include "base/loader/elf_object.hh" 56#include "base/loader/object_file.hh" 57#include "base/logging.hh" 58#include "base/trace.hh" 59#include "cpu/thread_context.hh" 60#include "debug/Stack.hh" 61#include "mem/multi_level_page_table.hh" 62#include "mem/page_table.hh" 63#include "params/Process.hh" 64#include "sim/aux_vector.hh" 65#include "sim/process_impl.hh" 66#include "sim/syscall_desc.hh" 67#include "sim/syscall_return.hh" 68#include "sim/system.hh" 69 70using namespace std; 71using namespace X86ISA; 72 73static const int ArgumentReg[] = { 74 INTREG_RDI, 75 INTREG_RSI, 76 INTREG_RDX, 77 // This argument register is r10 for syscalls and rcx for C. 78 INTREG_R10W, 79 // INTREG_RCX, 80 INTREG_R8W, 81 INTREG_R9W 82}; 83 84static const int NumArgumentRegs M5_VAR_USED = 85 sizeof(ArgumentReg) / sizeof(const int); 86 87static const int ArgumentReg32[] = { 88 INTREG_EBX, 89 INTREG_ECX, 90 INTREG_EDX, 91 INTREG_ESI, 92 INTREG_EDI, 93 INTREG_EBP 94}; 95 96static const int NumArgumentRegs32 M5_VAR_USED = 97 sizeof(ArgumentReg) / sizeof(const int); 98 99template class MultiLevelPageTable<LongModePTE<47, 39>, 100 LongModePTE<38, 30>, 101 LongModePTE<29, 21>, 102 LongModePTE<20, 12> >; 103typedef MultiLevelPageTable<LongModePTE<47, 39>, 104 LongModePTE<38, 30>, 105 LongModePTE<29, 21>, 106 LongModePTE<20, 12> > ArchPageTable; 107 108X86Process::X86Process(ProcessParams *params, ObjectFile *objFile, 109 SyscallDesc *_syscallDescs, int _numSyscallDescs) 110 : Process(params, params->useArchPT ? 111 static_cast<EmulationPageTable *>( 112 new ArchPageTable(params->name, params->pid, 113 params->system, PageBytes)) : 114 new EmulationPageTable(params->name, params->pid, 115 PageBytes), 116 objFile), 117 syscallDescs(_syscallDescs), numSyscallDescs(_numSyscallDescs) 118{ 119} 120 121void X86Process::clone(ThreadContext *old_tc, ThreadContext *new_tc, 122 Process *p, RegVal flags) 123{ 124 Process::clone(old_tc, new_tc, p, flags); 125 X86Process *process = (X86Process*)p; 126 *process = *this; 127} 128 129X86_64Process::X86_64Process(ProcessParams *params, ObjectFile *objFile, 130 SyscallDesc *_syscallDescs, int _numSyscallDescs) 131 : X86Process(params, objFile, _syscallDescs, _numSyscallDescs) 132{ 133 134 vsyscallPage.base = 0xffffffffff600000ULL; 135 vsyscallPage.size = PageBytes; 136 vsyscallPage.vtimeOffset = 0x400; 137 vsyscallPage.vgettimeofdayOffset = 0x0; 138 139 Addr brk_point = roundUp(objFile->dataBase() + objFile->dataSize() + 140 objFile->bssSize(), PageBytes); 141 Addr stack_base = 0x7FFFFFFFF000ULL; 142 Addr max_stack_size = 8 * 1024 * 1024; 143 Addr next_thread_stack_base = stack_base - max_stack_size; 144 Addr mmap_end = 0x7FFFF7FFF000ULL; 145 146 memState = make_shared<MemState>(brk_point, stack_base, max_stack_size, 147 next_thread_stack_base, mmap_end); 148} 149 150void 151I386Process::syscall(int64_t callnum, ThreadContext *tc, Fault *fault) 152{ 153 TheISA::PCState pc = tc->pcState(); 154 Addr eip = pc.pc(); 155 if (eip >= vsyscallPage.base && 156 eip < vsyscallPage.base + vsyscallPage.size) { 157 pc.npc(vsyscallPage.base + vsyscallPage.vsysexitOffset); 158 tc->pcState(pc); 159 } 160 X86Process::syscall(callnum, tc, fault); 161} 162 163 164I386Process::I386Process(ProcessParams *params, ObjectFile *objFile, 165 SyscallDesc *_syscallDescs, int _numSyscallDescs) 166 : X86Process(params, objFile, _syscallDescs, _numSyscallDescs) 167{ 168 if (kvmInSE) 169 panic("KVM CPU model does not support 32 bit processes"); 170 171 _gdtStart = ULL(0xffffd000); 172 _gdtSize = PageBytes; 173 174 vsyscallPage.base = 0xffffe000ULL; 175 vsyscallPage.size = PageBytes; 176 vsyscallPage.vsyscallOffset = 0x400; 177 vsyscallPage.vsysexitOffset = 0x410; 178 179 Addr brk_point = roundUp(objFile->dataBase() + objFile->dataSize() + 180 objFile->bssSize(), PageBytes); 181 Addr stack_base = _gdtStart; 182 Addr max_stack_size = 8 * 1024 * 1024; 183 Addr next_thread_stack_base = stack_base - max_stack_size; 184 Addr mmap_end = 0xB7FFF000ULL; 185 186 memState = make_shared<MemState>(brk_point, stack_base, max_stack_size, 187 next_thread_stack_base, mmap_end); 188} 189 190SyscallDesc* 191X86Process::getDesc(int callnum) 192{ 193 if (callnum < 0 || callnum >= numSyscallDescs) 194 return NULL; 195 return &syscallDescs[callnum]; 196} 197 198void 199X86_64Process::initState() 200{ 201 X86Process::initState(); 202 203 if (useForClone) 204 return; 205 206 argsInit(PageBytes); 207 208 // Set up the vsyscall page for this process. 209 allocateMem(vsyscallPage.base, vsyscallPage.size); 210 uint8_t vtimeBlob[] = { 211 0x48,0xc7,0xc0,0xc9,0x00,0x00,0x00, // mov $0xc9,%rax 212 0x0f,0x05, // syscall 213 0xc3 // retq 214 }; 215 initVirtMem.writeBlob(vsyscallPage.base + vsyscallPage.vtimeOffset, 216 vtimeBlob, sizeof(vtimeBlob)); 217 218 uint8_t vgettimeofdayBlob[] = { 219 0x48,0xc7,0xc0,0x60,0x00,0x00,0x00, // mov $0x60,%rax 220 0x0f,0x05, // syscall 221 0xc3 // retq 222 }; 223 initVirtMem.writeBlob(vsyscallPage.base + vsyscallPage.vgettimeofdayOffset, 224 vgettimeofdayBlob, sizeof(vgettimeofdayBlob)); 225 226 if (kvmInSE) { 227 PortProxy physProxy = system->physProxy; 228 229 Addr syscallCodePhysAddr = system->allocPhysPages(1); 230 Addr gdtPhysAddr = system->allocPhysPages(1); 231 Addr idtPhysAddr = system->allocPhysPages(1); 232 Addr istPhysAddr = system->allocPhysPages(1); 233 Addr tssPhysAddr = system->allocPhysPages(1); 234 Addr pfHandlerPhysAddr = system->allocPhysPages(1); 235 236 /* 237 * Set up the gdt. 238 */ 239 uint8_t numGDTEntries = 0; 240 uint64_t nullDescriptor = 0; 241 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 242 (uint8_t *)(&nullDescriptor), 8); 243 numGDTEntries++; 244 245 SegDescriptor initDesc = 0; 246 initDesc.type.codeOrData = 0; // code or data type 247 initDesc.type.c = 0; // conforming 248 initDesc.type.r = 1; // readable 249 initDesc.dpl = 0; // privilege 250 initDesc.p = 1; // present 251 initDesc.l = 1; // longmode - 64 bit 252 initDesc.d = 0; // operand size 253 initDesc.s = 1; // system segment 254 initDesc.limit = 0xFFFFFFFF; 255 initDesc.base = 0; 256 257 //64 bit code segment 258 SegDescriptor csLowPLDesc = initDesc; 259 csLowPLDesc.type.codeOrData = 1; 260 csLowPLDesc.dpl = 0; 261 uint64_t csLowPLDescVal = csLowPLDesc; 262 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 263 (uint8_t *)(&csLowPLDescVal), 8); 264 265 numGDTEntries++; 266 267 SegSelector csLowPL = 0; 268 csLowPL.si = numGDTEntries - 1; 269 csLowPL.rpl = 0; 270 271 //64 bit data segment 272 SegDescriptor dsLowPLDesc = initDesc; 273 dsLowPLDesc.type.codeOrData = 0; 274 dsLowPLDesc.dpl = 0; 275 uint64_t dsLowPLDescVal = dsLowPLDesc; 276 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 277 (uint8_t *)(&dsLowPLDescVal), 8); 278 279 numGDTEntries++; 280 281 SegSelector dsLowPL = 0; 282 dsLowPL.si = numGDTEntries - 1; 283 dsLowPL.rpl = 0; 284 285 //64 bit data segment 286 SegDescriptor dsDesc = initDesc; 287 dsDesc.type.codeOrData = 0; 288 dsDesc.dpl = 3; 289 uint64_t dsDescVal = dsDesc; 290 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 291 (uint8_t *)(&dsDescVal), 8); 292 293 numGDTEntries++; 294 295 SegSelector ds = 0; 296 ds.si = numGDTEntries - 1; 297 ds.rpl = 3; 298 299 //64 bit code segment 300 SegDescriptor csDesc = initDesc; 301 csDesc.type.codeOrData = 1; 302 csDesc.dpl = 3; 303 uint64_t csDescVal = csDesc; 304 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 305 (uint8_t *)(&csDescVal), 8); 306 307 numGDTEntries++; 308 309 SegSelector cs = 0; 310 cs.si = numGDTEntries - 1; 311 cs.rpl = 3; 312 313 SegSelector scall = 0; 314 scall.si = csLowPL.si; 315 scall.rpl = 0; 316 317 SegSelector sret = 0; 318 sret.si = dsLowPL.si; 319 sret.rpl = 3; 320 321 /* In long mode the TSS has been extended to 16 Bytes */ 322 TSSlow TSSDescLow = 0; 323 TSSDescLow.type = 0xB; 324 TSSDescLow.dpl = 0; // Privelege level 0 325 TSSDescLow.p = 1; // Present 326 TSSDescLow.limit = 0xFFFFFFFF; 327 TSSDescLow.base = bits(TSSVirtAddr, 31, 0); 328 329 TSShigh TSSDescHigh = 0; 330 TSSDescHigh.base = bits(TSSVirtAddr, 63, 32); 331 332 struct TSSDesc { 333 uint64_t low; 334 uint64_t high; 335 } tssDescVal = {TSSDescLow, TSSDescHigh}; 336 337 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 338 (uint8_t *)(&tssDescVal), sizeof(tssDescVal)); 339 340 numGDTEntries++; 341 342 SegSelector tssSel = 0; 343 tssSel.si = numGDTEntries - 1; 344 345 uint64_t tss_base_addr = (TSSDescHigh.base << 32) | TSSDescLow.base; 346 uint64_t tss_limit = TSSDescLow.limit; 347 348 SegAttr tss_attr = 0; 349 350 tss_attr.type = TSSDescLow.type; 351 tss_attr.dpl = TSSDescLow.dpl; 352 tss_attr.present = TSSDescLow.p; 353 tss_attr.granularity = TSSDescLow.g; 354 tss_attr.unusable = 0; 355 356 for (int i = 0; i < contextIds.size(); i++) { 357 ThreadContext * tc = system->getThreadContext(contextIds[i]); 358 359 tc->setMiscReg(MISCREG_CS, cs); 360 tc->setMiscReg(MISCREG_DS, ds); 361 tc->setMiscReg(MISCREG_ES, ds); 362 tc->setMiscReg(MISCREG_FS, ds); 363 tc->setMiscReg(MISCREG_GS, ds); 364 tc->setMiscReg(MISCREG_SS, ds); 365 366 // LDT 367 tc->setMiscReg(MISCREG_TSL, 0); 368 SegAttr tslAttr = 0; 369 tslAttr.present = 1; 370 tslAttr.type = 2; 371 tc->setMiscReg(MISCREG_TSL_ATTR, tslAttr); 372 373 tc->setMiscReg(MISCREG_TSG_BASE, GDTVirtAddr); 374 tc->setMiscReg(MISCREG_TSG_LIMIT, 8 * numGDTEntries - 1); 375 376 tc->setMiscReg(MISCREG_TR, tssSel); 377 tc->setMiscReg(MISCREG_TR_BASE, tss_base_addr); 378 tc->setMiscReg(MISCREG_TR_EFF_BASE, 0); 379 tc->setMiscReg(MISCREG_TR_LIMIT, tss_limit); 380 tc->setMiscReg(MISCREG_TR_ATTR, tss_attr); 381 382 //Start using longmode segments. 383 installSegDesc(tc, SEGMENT_REG_CS, csDesc, true); 384 installSegDesc(tc, SEGMENT_REG_DS, dsDesc, true); 385 installSegDesc(tc, SEGMENT_REG_ES, dsDesc, true); 386 installSegDesc(tc, SEGMENT_REG_FS, dsDesc, true); 387 installSegDesc(tc, SEGMENT_REG_GS, dsDesc, true); 388 installSegDesc(tc, SEGMENT_REG_SS, dsDesc, true); 389 390 Efer efer = 0; 391 efer.sce = 1; // Enable system call extensions. 392 efer.lme = 1; // Enable long mode. 393 efer.lma = 1; // Activate long mode. 394 efer.nxe = 0; // Enable nx support. 395 efer.svme = 1; // Enable svm support for now. 396 efer.ffxsr = 0; // Turn on fast fxsave and fxrstor. 397 tc->setMiscReg(MISCREG_EFER, efer); 398 399 //Set up the registers that describe the operating mode. 400 CR0 cr0 = 0; 401 cr0.pg = 1; // Turn on paging. 402 cr0.cd = 0; // Don't disable caching. 403 cr0.nw = 0; // This is bit is defined to be ignored. 404 cr0.am = 1; // No alignment checking 405 cr0.wp = 1; // Supervisor mode can write read only pages 406 cr0.ne = 1; 407 cr0.et = 1; // This should always be 1 408 cr0.ts = 0; // We don't do task switching, so causing fp exceptions 409 // would be pointless. 410 cr0.em = 0; // Allow x87 instructions to execute natively. 411 cr0.mp = 1; // This doesn't really matter, but the manual suggests 412 // setting it to one. 413 cr0.pe = 1; // We're definitely in protected mode. 414 tc->setMiscReg(MISCREG_CR0, cr0); 415 416 CR0 cr2 = 0; 417 tc->setMiscReg(MISCREG_CR2, cr2); 418 419 CR3 cr3 = dynamic_cast<ArchPageTable *>(pTable)->basePtr(); 420 tc->setMiscReg(MISCREG_CR3, cr3); 421 422 CR4 cr4 = 0; 423 //Turn on pae. 424 cr4.osxsave = 1; // Enable XSAVE and Proc Extended States 425 cr4.osxmmexcpt = 1; // Operating System Unmasked Exception 426 cr4.osfxsr = 1; // Operating System FXSave/FSRSTOR Support 427 cr4.pce = 0; // Performance-Monitoring Counter Enable 428 cr4.pge = 0; // Page-Global Enable 429 cr4.mce = 0; // Machine Check Enable 430 cr4.pae = 1; // Physical-Address Extension 431 cr4.pse = 0; // Page Size Extensions 432 cr4.de = 0; // Debugging Extensions 433 cr4.tsd = 0; // Time Stamp Disable 434 cr4.pvi = 0; // Protected-Mode Virtual Interrupts 435 cr4.vme = 0; // Virtual-8086 Mode Extensions 436 437 tc->setMiscReg(MISCREG_CR4, cr4); 438 439 CR4 cr8 = 0; 440 tc->setMiscReg(MISCREG_CR8, cr8); 441 442 tc->setMiscReg(MISCREG_MXCSR, 0x1f80); 443 444 tc->setMiscReg(MISCREG_APIC_BASE, 0xfee00900); 445 446 tc->setMiscReg(MISCREG_TSG_BASE, GDTVirtAddr); 447 tc->setMiscReg(MISCREG_TSG_LIMIT, 0xffff); 448 449 tc->setMiscReg(MISCREG_IDTR_BASE, IDTVirtAddr); 450 tc->setMiscReg(MISCREG_IDTR_LIMIT, 0xffff); 451 452 /* enabling syscall and sysret */ 453 RegVal star = ((RegVal)sret << 48) | ((RegVal)scall << 32); 454 tc->setMiscReg(MISCREG_STAR, star); 455 RegVal lstar = (RegVal)syscallCodeVirtAddr; 456 tc->setMiscReg(MISCREG_LSTAR, lstar); 457 RegVal sfmask = (1 << 8) | (1 << 10); // TF | DF 458 tc->setMiscReg(MISCREG_SF_MASK, sfmask); 459 } 460 461 /* Set up the content of the TSS and write it to physical memory. */ 462 463 struct { 464 uint32_t reserved0; // +00h 465 uint32_t RSP0_low; // +04h 466 uint32_t RSP0_high; // +08h 467 uint32_t RSP1_low; // +0Ch 468 uint32_t RSP1_high; // +10h 469 uint32_t RSP2_low; // +14h 470 uint32_t RSP2_high; // +18h 471 uint32_t reserved1; // +1Ch 472 uint32_t reserved2; // +20h 473 uint32_t IST1_low; // +24h 474 uint32_t IST1_high; // +28h 475 uint32_t IST2_low; // +2Ch 476 uint32_t IST2_high; // +30h 477 uint32_t IST3_low; // +34h 478 uint32_t IST3_high; // +38h 479 uint32_t IST4_low; // +3Ch 480 uint32_t IST4_high; // +40h 481 uint32_t IST5_low; // +44h 482 uint32_t IST5_high; // +48h 483 uint32_t IST6_low; // +4Ch 484 uint32_t IST6_high; // +50h 485 uint32_t IST7_low; // +54h 486 uint32_t IST7_high; // +58h 487 uint32_t reserved3; // +5Ch 488 uint32_t reserved4; // +60h 489 uint16_t reserved5; // +64h 490 uint16_t IO_MapBase; // +66h 491 } tss; 492 493 /** setting Interrupt Stack Table */ 494 uint64_t IST_start = ISTVirtAddr + PageBytes; 495 tss.IST1_low = IST_start; 496 tss.IST1_high = IST_start >> 32; 497 tss.RSP0_low = tss.IST1_low; 498 tss.RSP0_high = tss.IST1_high; 499 tss.RSP1_low = tss.IST1_low; 500 tss.RSP1_high = tss.IST1_high; 501 tss.RSP2_low = tss.IST1_low; 502 tss.RSP2_high = tss.IST1_high; 503 physProxy.writeBlob(tssPhysAddr, (uint8_t *)(&tss), sizeof(tss)); 504 505 /* Setting IDT gates */ 506 GateDescriptorLow PFGateLow = 0; 507 PFGateLow.offsetHigh = bits(PFHandlerVirtAddr, 31, 16); 508 PFGateLow.offsetLow = bits(PFHandlerVirtAddr, 15, 0); 509 PFGateLow.selector = csLowPL; 510 PFGateLow.p = 1; 511 PFGateLow.dpl = 0; 512 PFGateLow.type = 0xe; // gate interrupt type 513 PFGateLow.IST = 0; // setting IST to 0 and using RSP0 514 515 GateDescriptorHigh PFGateHigh = 0; 516 PFGateHigh.offset = bits(PFHandlerVirtAddr, 63, 32); 517 518 struct { 519 uint64_t low; 520 uint64_t high; 521 } PFGate = {PFGateLow, PFGateHigh}; 522 523 physProxy.writeBlob(idtPhysAddr + 0xE0, 524 (uint8_t *)(&PFGate), sizeof(PFGate)); 525 526 /* System call handler */ 527 uint8_t syscallBlob[] = { 528 // mov %rax, (0xffffc90000005600) 529 0x48, 0xa3, 0x00, 0x60, 0x00, 530 0x00, 0x00, 0xc9, 0xff, 0xff, 531 // sysret 532 0x48, 0x0f, 0x07 533 }; 534 535 physProxy.writeBlob(syscallCodePhysAddr, 536 syscallBlob, sizeof(syscallBlob)); 537 538 /** Page fault handler */ 539 uint8_t faultBlob[] = { 540 // mov %rax, (0xffffc90000005700) 541 0x48, 0xa3, 0x00, 0x61, 0x00, 542 0x00, 0x00, 0xc9, 0xff, 0xff, 543 // add $0x8, %rsp # skip error 544 0x48, 0x83, 0xc4, 0x08, 545 // iretq 546 0x48, 0xcf 547 }; 548 549 physProxy.writeBlob(pfHandlerPhysAddr, faultBlob, sizeof(faultBlob)); 550 551 /* Syscall handler */ 552 pTable->map(syscallCodeVirtAddr, syscallCodePhysAddr, 553 PageBytes, false); 554 /* GDT */ 555 pTable->map(GDTVirtAddr, gdtPhysAddr, PageBytes, false); 556 /* IDT */ 557 pTable->map(IDTVirtAddr, idtPhysAddr, PageBytes, false); 558 /* TSS */ 559 pTable->map(TSSVirtAddr, tssPhysAddr, PageBytes, false); 560 /* IST */ 561 pTable->map(ISTVirtAddr, istPhysAddr, PageBytes, false); 562 /* PF handler */ 563 pTable->map(PFHandlerVirtAddr, pfHandlerPhysAddr, PageBytes, false); 564 /* MMIO region for m5ops */ 565 pTable->map(MMIORegionVirtAddr, MMIORegionPhysAddr, 566 16 * PageBytes, false); 567 } else { 568 for (int i = 0; i < contextIds.size(); i++) { 569 ThreadContext * tc = system->getThreadContext(contextIds[i]); 570 571 SegAttr dataAttr = 0; 572 dataAttr.dpl = 3; 573 dataAttr.unusable = 0; 574 dataAttr.defaultSize = 1; 575 dataAttr.longMode = 1; 576 dataAttr.avl = 0; 577 dataAttr.granularity = 1; 578 dataAttr.present = 1; 579 dataAttr.type = 3; 580 dataAttr.writable = 1; 581 dataAttr.readable = 1; 582 dataAttr.expandDown = 0; 583 dataAttr.system = 1; 584 585 // Initialize the segment registers. 586 for (int seg = 0; seg < NUM_SEGMENTREGS; seg++) { 587 tc->setMiscRegNoEffect(MISCREG_SEG_BASE(seg), 0); 588 tc->setMiscRegNoEffect(MISCREG_SEG_EFF_BASE(seg), 0); 589 tc->setMiscRegNoEffect(MISCREG_SEG_ATTR(seg), dataAttr); 590 } 591 592 SegAttr csAttr = 0; 593 csAttr.dpl = 3; 594 csAttr.unusable = 0; 595 csAttr.defaultSize = 0; 596 csAttr.longMode = 1; 597 csAttr.avl = 0; 598 csAttr.granularity = 1; 599 csAttr.present = 1; 600 csAttr.type = 10; 601 csAttr.writable = 0; 602 csAttr.readable = 1; 603 csAttr.expandDown = 0; 604 csAttr.system = 1; 605 606 tc->setMiscRegNoEffect(MISCREG_CS_ATTR, csAttr); 607 608 Efer efer = 0; 609 efer.sce = 1; // Enable system call extensions. 610 efer.lme = 1; // Enable long mode. 611 efer.lma = 1; // Activate long mode. 612 efer.nxe = 1; // Enable nx support. 613 efer.svme = 0; // Disable svm support for now. It isn't implemented. 614 efer.ffxsr = 1; // Turn on fast fxsave and fxrstor. 615 tc->setMiscReg(MISCREG_EFER, efer); 616 617 // Set up the registers that describe the operating mode. 618 CR0 cr0 = 0; 619 cr0.pg = 1; // Turn on paging. 620 cr0.cd = 0; // Don't disable caching. 621 cr0.nw = 0; // This is bit is defined to be ignored. 622 cr0.am = 0; // No alignment checking 623 cr0.wp = 0; // Supervisor mode can write read only pages 624 cr0.ne = 1; 625 cr0.et = 1; // This should always be 1 626 cr0.ts = 0; // We don't do task switching, so causing fp exceptions 627 // would be pointless. 628 cr0.em = 0; // Allow x87 instructions to execute natively. 629 cr0.mp = 1; // This doesn't really matter, but the manual suggests 630 // setting it to one. 631 cr0.pe = 1; // We're definitely in protected mode. 632 tc->setMiscReg(MISCREG_CR0, cr0); 633 634 tc->setMiscReg(MISCREG_MXCSR, 0x1f80); 635 } 636 } 637} 638 639void 640I386Process::initState() 641{ 642 X86Process::initState(); 643 644 argsInit(PageBytes); 645 646 /* 647 * Set up a GDT for this process. The whole GDT wouldn't really be for 648 * this process, but the only parts we care about are. 649 */ 650 allocateMem(_gdtStart, _gdtSize); 651 uint64_t zero = 0; 652 assert(_gdtSize % sizeof(zero) == 0); 653 for (Addr gdtCurrent = _gdtStart; 654 gdtCurrent < _gdtStart + _gdtSize; gdtCurrent += sizeof(zero)) { 655 initVirtMem.write(gdtCurrent, zero); 656 } 657 658 // Set up the vsyscall page for this process. 659 allocateMem(vsyscallPage.base, vsyscallPage.size); 660 uint8_t vsyscallBlob[] = { 661 0x51, // push %ecx 662 0x52, // push %edp 663 0x55, // push %ebp 664 0x89, 0xe5, // mov %esp, %ebp 665 0x0f, 0x34 // sysenter 666 }; 667 initVirtMem.writeBlob(vsyscallPage.base + vsyscallPage.vsyscallOffset, 668 vsyscallBlob, sizeof(vsyscallBlob)); 669 670 uint8_t vsysexitBlob[] = { 671 0x5d, // pop %ebp 672 0x5a, // pop %edx 673 0x59, // pop %ecx 674 0xc3 // ret 675 }; 676 initVirtMem.writeBlob(vsyscallPage.base + vsyscallPage.vsysexitOffset, 677 vsysexitBlob, sizeof(vsysexitBlob)); 678 679 for (int i = 0; i < contextIds.size(); i++) { 680 ThreadContext * tc = system->getThreadContext(contextIds[i]); 681 682 SegAttr dataAttr = 0; 683 dataAttr.dpl = 3; 684 dataAttr.unusable = 0; 685 dataAttr.defaultSize = 1; 686 dataAttr.longMode = 0; 687 dataAttr.avl = 0; 688 dataAttr.granularity = 1; 689 dataAttr.present = 1; 690 dataAttr.type = 3; 691 dataAttr.writable = 1; 692 dataAttr.readable = 1; 693 dataAttr.expandDown = 0; 694 dataAttr.system = 1; 695 696 // Initialize the segment registers. 697 for (int seg = 0; seg < NUM_SEGMENTREGS; seg++) { 698 tc->setMiscRegNoEffect(MISCREG_SEG_BASE(seg), 0); 699 tc->setMiscRegNoEffect(MISCREG_SEG_EFF_BASE(seg), 0); 700 tc->setMiscRegNoEffect(MISCREG_SEG_ATTR(seg), dataAttr); 701 tc->setMiscRegNoEffect(MISCREG_SEG_SEL(seg), 0xB); 702 tc->setMiscRegNoEffect(MISCREG_SEG_LIMIT(seg), (uint32_t)(-1)); 703 } 704 705 SegAttr csAttr = 0; 706 csAttr.dpl = 3; 707 csAttr.unusable = 0; 708 csAttr.defaultSize = 1; 709 csAttr.longMode = 0; 710 csAttr.avl = 0; 711 csAttr.granularity = 1; 712 csAttr.present = 1; 713 csAttr.type = 0xa; 714 csAttr.writable = 0; 715 csAttr.readable = 1; 716 csAttr.expandDown = 0; 717 csAttr.system = 1; 718 719 tc->setMiscRegNoEffect(MISCREG_CS_ATTR, csAttr); 720 721 tc->setMiscRegNoEffect(MISCREG_TSG_BASE, _gdtStart); 722 tc->setMiscRegNoEffect(MISCREG_TSG_EFF_BASE, _gdtStart); 723 tc->setMiscRegNoEffect(MISCREG_TSG_LIMIT, _gdtStart + _gdtSize - 1); 724 725 // Set the LDT selector to 0 to deactivate it. 726 tc->setMiscRegNoEffect(MISCREG_TSL, 0); 727 728 Efer efer = 0; 729 efer.sce = 1; // Enable system call extensions. 730 efer.lme = 1; // Enable long mode. 731 efer.lma = 0; // Deactivate long mode. 732 efer.nxe = 1; // Enable nx support. 733 efer.svme = 0; // Disable svm support for now. It isn't implemented. 734 efer.ffxsr = 1; // Turn on fast fxsave and fxrstor. 735 tc->setMiscReg(MISCREG_EFER, efer); 736 737 // Set up the registers that describe the operating mode. 738 CR0 cr0 = 0; 739 cr0.pg = 1; // Turn on paging. 740 cr0.cd = 0; // Don't disable caching. 741 cr0.nw = 0; // This is bit is defined to be ignored. 742 cr0.am = 0; // No alignment checking 743 cr0.wp = 0; // Supervisor mode can write read only pages 744 cr0.ne = 1; 745 cr0.et = 1; // This should always be 1 746 cr0.ts = 0; // We don't do task switching, so causing fp exceptions 747 // would be pointless. 748 cr0.em = 0; // Allow x87 instructions to execute natively. 749 cr0.mp = 1; // This doesn't really matter, but the manual suggests 750 // setting it to one. 751 cr0.pe = 1; // We're definitely in protected mode. 752 tc->setMiscReg(MISCREG_CR0, cr0); 753 754 tc->setMiscReg(MISCREG_MXCSR, 0x1f80); 755 } 756} 757 758template<class IntType> 759void 760X86Process::argsInit(int pageSize, 761 std::vector<AuxVector<IntType> > extraAuxvs) 762{ 763 int intSize = sizeof(IntType); 764
| 1/* 2 * Copyright (c) 2014 Advanced Micro Devices, Inc. 3 * Copyright (c) 2007 The Hewlett-Packard Development Company 4 * All rights reserved. 5 * 6 * The license below extends only to copyright in the software and shall 7 * not be construed as granting a license to any other intellectual 8 * property including but not limited to intellectual property relating 9 * to a hardware implementation of the functionality of the software 10 * licensed hereunder. You may use the software subject to the license 11 * terms below provided that you ensure that this notice is replicated 12 * unmodified and in its entirety in all distributions of the software, 13 * modified or unmodified, in source code or in binary form. 14 * 15 * Copyright (c) 2003-2006 The Regents of The University of Michigan 16 * All rights reserved. 17 * 18 * Redistribution and use in source and binary forms, with or without 19 * modification, are permitted provided that the following conditions are 20 * met: redistributions of source code must retain the above copyright 21 * notice, this list of conditions and the following disclaimer; 22 * redistributions in binary form must reproduce the above copyright 23 * notice, this list of conditions and the following disclaimer in the 24 * documentation and/or other materials provided with the distribution; 25 * neither the name of the copyright holders nor the names of its 26 * contributors may be used to endorse or promote products derived from 27 * this software without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 30 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 31 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 32 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 33 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 34 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 35 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 36 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 37 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 38 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 39 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 40 * 41 * Authors: Gabe Black 42 * Ali Saidi 43 */ 44 45#include "arch/x86/process.hh" 46 47#include <string> 48#include <vector> 49 50#include "arch/x86/isa_traits.hh" 51#include "arch/x86/regs/misc.hh" 52#include "arch/x86/regs/segment.hh" 53#include "arch/x86/system.hh" 54#include "arch/x86/types.hh" 55#include "base/loader/elf_object.hh" 56#include "base/loader/object_file.hh" 57#include "base/logging.hh" 58#include "base/trace.hh" 59#include "cpu/thread_context.hh" 60#include "debug/Stack.hh" 61#include "mem/multi_level_page_table.hh" 62#include "mem/page_table.hh" 63#include "params/Process.hh" 64#include "sim/aux_vector.hh" 65#include "sim/process_impl.hh" 66#include "sim/syscall_desc.hh" 67#include "sim/syscall_return.hh" 68#include "sim/system.hh" 69 70using namespace std; 71using namespace X86ISA; 72 73static const int ArgumentReg[] = { 74 INTREG_RDI, 75 INTREG_RSI, 76 INTREG_RDX, 77 // This argument register is r10 for syscalls and rcx for C. 78 INTREG_R10W, 79 // INTREG_RCX, 80 INTREG_R8W, 81 INTREG_R9W 82}; 83 84static const int NumArgumentRegs M5_VAR_USED = 85 sizeof(ArgumentReg) / sizeof(const int); 86 87static const int ArgumentReg32[] = { 88 INTREG_EBX, 89 INTREG_ECX, 90 INTREG_EDX, 91 INTREG_ESI, 92 INTREG_EDI, 93 INTREG_EBP 94}; 95 96static const int NumArgumentRegs32 M5_VAR_USED = 97 sizeof(ArgumentReg) / sizeof(const int); 98 99template class MultiLevelPageTable<LongModePTE<47, 39>, 100 LongModePTE<38, 30>, 101 LongModePTE<29, 21>, 102 LongModePTE<20, 12> >; 103typedef MultiLevelPageTable<LongModePTE<47, 39>, 104 LongModePTE<38, 30>, 105 LongModePTE<29, 21>, 106 LongModePTE<20, 12> > ArchPageTable; 107 108X86Process::X86Process(ProcessParams *params, ObjectFile *objFile, 109 SyscallDesc *_syscallDescs, int _numSyscallDescs) 110 : Process(params, params->useArchPT ? 111 static_cast<EmulationPageTable *>( 112 new ArchPageTable(params->name, params->pid, 113 params->system, PageBytes)) : 114 new EmulationPageTable(params->name, params->pid, 115 PageBytes), 116 objFile), 117 syscallDescs(_syscallDescs), numSyscallDescs(_numSyscallDescs) 118{ 119} 120 121void X86Process::clone(ThreadContext *old_tc, ThreadContext *new_tc, 122 Process *p, RegVal flags) 123{ 124 Process::clone(old_tc, new_tc, p, flags); 125 X86Process *process = (X86Process*)p; 126 *process = *this; 127} 128 129X86_64Process::X86_64Process(ProcessParams *params, ObjectFile *objFile, 130 SyscallDesc *_syscallDescs, int _numSyscallDescs) 131 : X86Process(params, objFile, _syscallDescs, _numSyscallDescs) 132{ 133 134 vsyscallPage.base = 0xffffffffff600000ULL; 135 vsyscallPage.size = PageBytes; 136 vsyscallPage.vtimeOffset = 0x400; 137 vsyscallPage.vgettimeofdayOffset = 0x0; 138 139 Addr brk_point = roundUp(objFile->dataBase() + objFile->dataSize() + 140 objFile->bssSize(), PageBytes); 141 Addr stack_base = 0x7FFFFFFFF000ULL; 142 Addr max_stack_size = 8 * 1024 * 1024; 143 Addr next_thread_stack_base = stack_base - max_stack_size; 144 Addr mmap_end = 0x7FFFF7FFF000ULL; 145 146 memState = make_shared<MemState>(brk_point, stack_base, max_stack_size, 147 next_thread_stack_base, mmap_end); 148} 149 150void 151I386Process::syscall(int64_t callnum, ThreadContext *tc, Fault *fault) 152{ 153 TheISA::PCState pc = tc->pcState(); 154 Addr eip = pc.pc(); 155 if (eip >= vsyscallPage.base && 156 eip < vsyscallPage.base + vsyscallPage.size) { 157 pc.npc(vsyscallPage.base + vsyscallPage.vsysexitOffset); 158 tc->pcState(pc); 159 } 160 X86Process::syscall(callnum, tc, fault); 161} 162 163 164I386Process::I386Process(ProcessParams *params, ObjectFile *objFile, 165 SyscallDesc *_syscallDescs, int _numSyscallDescs) 166 : X86Process(params, objFile, _syscallDescs, _numSyscallDescs) 167{ 168 if (kvmInSE) 169 panic("KVM CPU model does not support 32 bit processes"); 170 171 _gdtStart = ULL(0xffffd000); 172 _gdtSize = PageBytes; 173 174 vsyscallPage.base = 0xffffe000ULL; 175 vsyscallPage.size = PageBytes; 176 vsyscallPage.vsyscallOffset = 0x400; 177 vsyscallPage.vsysexitOffset = 0x410; 178 179 Addr brk_point = roundUp(objFile->dataBase() + objFile->dataSize() + 180 objFile->bssSize(), PageBytes); 181 Addr stack_base = _gdtStart; 182 Addr max_stack_size = 8 * 1024 * 1024; 183 Addr next_thread_stack_base = stack_base - max_stack_size; 184 Addr mmap_end = 0xB7FFF000ULL; 185 186 memState = make_shared<MemState>(brk_point, stack_base, max_stack_size, 187 next_thread_stack_base, mmap_end); 188} 189 190SyscallDesc* 191X86Process::getDesc(int callnum) 192{ 193 if (callnum < 0 || callnum >= numSyscallDescs) 194 return NULL; 195 return &syscallDescs[callnum]; 196} 197 198void 199X86_64Process::initState() 200{ 201 X86Process::initState(); 202 203 if (useForClone) 204 return; 205 206 argsInit(PageBytes); 207 208 // Set up the vsyscall page for this process. 209 allocateMem(vsyscallPage.base, vsyscallPage.size); 210 uint8_t vtimeBlob[] = { 211 0x48,0xc7,0xc0,0xc9,0x00,0x00,0x00, // mov $0xc9,%rax 212 0x0f,0x05, // syscall 213 0xc3 // retq 214 }; 215 initVirtMem.writeBlob(vsyscallPage.base + vsyscallPage.vtimeOffset, 216 vtimeBlob, sizeof(vtimeBlob)); 217 218 uint8_t vgettimeofdayBlob[] = { 219 0x48,0xc7,0xc0,0x60,0x00,0x00,0x00, // mov $0x60,%rax 220 0x0f,0x05, // syscall 221 0xc3 // retq 222 }; 223 initVirtMem.writeBlob(vsyscallPage.base + vsyscallPage.vgettimeofdayOffset, 224 vgettimeofdayBlob, sizeof(vgettimeofdayBlob)); 225 226 if (kvmInSE) { 227 PortProxy physProxy = system->physProxy; 228 229 Addr syscallCodePhysAddr = system->allocPhysPages(1); 230 Addr gdtPhysAddr = system->allocPhysPages(1); 231 Addr idtPhysAddr = system->allocPhysPages(1); 232 Addr istPhysAddr = system->allocPhysPages(1); 233 Addr tssPhysAddr = system->allocPhysPages(1); 234 Addr pfHandlerPhysAddr = system->allocPhysPages(1); 235 236 /* 237 * Set up the gdt. 238 */ 239 uint8_t numGDTEntries = 0; 240 uint64_t nullDescriptor = 0; 241 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 242 (uint8_t *)(&nullDescriptor), 8); 243 numGDTEntries++; 244 245 SegDescriptor initDesc = 0; 246 initDesc.type.codeOrData = 0; // code or data type 247 initDesc.type.c = 0; // conforming 248 initDesc.type.r = 1; // readable 249 initDesc.dpl = 0; // privilege 250 initDesc.p = 1; // present 251 initDesc.l = 1; // longmode - 64 bit 252 initDesc.d = 0; // operand size 253 initDesc.s = 1; // system segment 254 initDesc.limit = 0xFFFFFFFF; 255 initDesc.base = 0; 256 257 //64 bit code segment 258 SegDescriptor csLowPLDesc = initDesc; 259 csLowPLDesc.type.codeOrData = 1; 260 csLowPLDesc.dpl = 0; 261 uint64_t csLowPLDescVal = csLowPLDesc; 262 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 263 (uint8_t *)(&csLowPLDescVal), 8); 264 265 numGDTEntries++; 266 267 SegSelector csLowPL = 0; 268 csLowPL.si = numGDTEntries - 1; 269 csLowPL.rpl = 0; 270 271 //64 bit data segment 272 SegDescriptor dsLowPLDesc = initDesc; 273 dsLowPLDesc.type.codeOrData = 0; 274 dsLowPLDesc.dpl = 0; 275 uint64_t dsLowPLDescVal = dsLowPLDesc; 276 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 277 (uint8_t *)(&dsLowPLDescVal), 8); 278 279 numGDTEntries++; 280 281 SegSelector dsLowPL = 0; 282 dsLowPL.si = numGDTEntries - 1; 283 dsLowPL.rpl = 0; 284 285 //64 bit data segment 286 SegDescriptor dsDesc = initDesc; 287 dsDesc.type.codeOrData = 0; 288 dsDesc.dpl = 3; 289 uint64_t dsDescVal = dsDesc; 290 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 291 (uint8_t *)(&dsDescVal), 8); 292 293 numGDTEntries++; 294 295 SegSelector ds = 0; 296 ds.si = numGDTEntries - 1; 297 ds.rpl = 3; 298 299 //64 bit code segment 300 SegDescriptor csDesc = initDesc; 301 csDesc.type.codeOrData = 1; 302 csDesc.dpl = 3; 303 uint64_t csDescVal = csDesc; 304 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 305 (uint8_t *)(&csDescVal), 8); 306 307 numGDTEntries++; 308 309 SegSelector cs = 0; 310 cs.si = numGDTEntries - 1; 311 cs.rpl = 3; 312 313 SegSelector scall = 0; 314 scall.si = csLowPL.si; 315 scall.rpl = 0; 316 317 SegSelector sret = 0; 318 sret.si = dsLowPL.si; 319 sret.rpl = 3; 320 321 /* In long mode the TSS has been extended to 16 Bytes */ 322 TSSlow TSSDescLow = 0; 323 TSSDescLow.type = 0xB; 324 TSSDescLow.dpl = 0; // Privelege level 0 325 TSSDescLow.p = 1; // Present 326 TSSDescLow.limit = 0xFFFFFFFF; 327 TSSDescLow.base = bits(TSSVirtAddr, 31, 0); 328 329 TSShigh TSSDescHigh = 0; 330 TSSDescHigh.base = bits(TSSVirtAddr, 63, 32); 331 332 struct TSSDesc { 333 uint64_t low; 334 uint64_t high; 335 } tssDescVal = {TSSDescLow, TSSDescHigh}; 336 337 physProxy.writeBlob(gdtPhysAddr + numGDTEntries * 8, 338 (uint8_t *)(&tssDescVal), sizeof(tssDescVal)); 339 340 numGDTEntries++; 341 342 SegSelector tssSel = 0; 343 tssSel.si = numGDTEntries - 1; 344 345 uint64_t tss_base_addr = (TSSDescHigh.base << 32) | TSSDescLow.base; 346 uint64_t tss_limit = TSSDescLow.limit; 347 348 SegAttr tss_attr = 0; 349 350 tss_attr.type = TSSDescLow.type; 351 tss_attr.dpl = TSSDescLow.dpl; 352 tss_attr.present = TSSDescLow.p; 353 tss_attr.granularity = TSSDescLow.g; 354 tss_attr.unusable = 0; 355 356 for (int i = 0; i < contextIds.size(); i++) { 357 ThreadContext * tc = system->getThreadContext(contextIds[i]); 358 359 tc->setMiscReg(MISCREG_CS, cs); 360 tc->setMiscReg(MISCREG_DS, ds); 361 tc->setMiscReg(MISCREG_ES, ds); 362 tc->setMiscReg(MISCREG_FS, ds); 363 tc->setMiscReg(MISCREG_GS, ds); 364 tc->setMiscReg(MISCREG_SS, ds); 365 366 // LDT 367 tc->setMiscReg(MISCREG_TSL, 0); 368 SegAttr tslAttr = 0; 369 tslAttr.present = 1; 370 tslAttr.type = 2; 371 tc->setMiscReg(MISCREG_TSL_ATTR, tslAttr); 372 373 tc->setMiscReg(MISCREG_TSG_BASE, GDTVirtAddr); 374 tc->setMiscReg(MISCREG_TSG_LIMIT, 8 * numGDTEntries - 1); 375 376 tc->setMiscReg(MISCREG_TR, tssSel); 377 tc->setMiscReg(MISCREG_TR_BASE, tss_base_addr); 378 tc->setMiscReg(MISCREG_TR_EFF_BASE, 0); 379 tc->setMiscReg(MISCREG_TR_LIMIT, tss_limit); 380 tc->setMiscReg(MISCREG_TR_ATTR, tss_attr); 381 382 //Start using longmode segments. 383 installSegDesc(tc, SEGMENT_REG_CS, csDesc, true); 384 installSegDesc(tc, SEGMENT_REG_DS, dsDesc, true); 385 installSegDesc(tc, SEGMENT_REG_ES, dsDesc, true); 386 installSegDesc(tc, SEGMENT_REG_FS, dsDesc, true); 387 installSegDesc(tc, SEGMENT_REG_GS, dsDesc, true); 388 installSegDesc(tc, SEGMENT_REG_SS, dsDesc, true); 389 390 Efer efer = 0; 391 efer.sce = 1; // Enable system call extensions. 392 efer.lme = 1; // Enable long mode. 393 efer.lma = 1; // Activate long mode. 394 efer.nxe = 0; // Enable nx support. 395 efer.svme = 1; // Enable svm support for now. 396 efer.ffxsr = 0; // Turn on fast fxsave and fxrstor. 397 tc->setMiscReg(MISCREG_EFER, efer); 398 399 //Set up the registers that describe the operating mode. 400 CR0 cr0 = 0; 401 cr0.pg = 1; // Turn on paging. 402 cr0.cd = 0; // Don't disable caching. 403 cr0.nw = 0; // This is bit is defined to be ignored. 404 cr0.am = 1; // No alignment checking 405 cr0.wp = 1; // Supervisor mode can write read only pages 406 cr0.ne = 1; 407 cr0.et = 1; // This should always be 1 408 cr0.ts = 0; // We don't do task switching, so causing fp exceptions 409 // would be pointless. 410 cr0.em = 0; // Allow x87 instructions to execute natively. 411 cr0.mp = 1; // This doesn't really matter, but the manual suggests 412 // setting it to one. 413 cr0.pe = 1; // We're definitely in protected mode. 414 tc->setMiscReg(MISCREG_CR0, cr0); 415 416 CR0 cr2 = 0; 417 tc->setMiscReg(MISCREG_CR2, cr2); 418 419 CR3 cr3 = dynamic_cast<ArchPageTable *>(pTable)->basePtr(); 420 tc->setMiscReg(MISCREG_CR3, cr3); 421 422 CR4 cr4 = 0; 423 //Turn on pae. 424 cr4.osxsave = 1; // Enable XSAVE and Proc Extended States 425 cr4.osxmmexcpt = 1; // Operating System Unmasked Exception 426 cr4.osfxsr = 1; // Operating System FXSave/FSRSTOR Support 427 cr4.pce = 0; // Performance-Monitoring Counter Enable 428 cr4.pge = 0; // Page-Global Enable 429 cr4.mce = 0; // Machine Check Enable 430 cr4.pae = 1; // Physical-Address Extension 431 cr4.pse = 0; // Page Size Extensions 432 cr4.de = 0; // Debugging Extensions 433 cr4.tsd = 0; // Time Stamp Disable 434 cr4.pvi = 0; // Protected-Mode Virtual Interrupts 435 cr4.vme = 0; // Virtual-8086 Mode Extensions 436 437 tc->setMiscReg(MISCREG_CR4, cr4); 438 439 CR4 cr8 = 0; 440 tc->setMiscReg(MISCREG_CR8, cr8); 441 442 tc->setMiscReg(MISCREG_MXCSR, 0x1f80); 443 444 tc->setMiscReg(MISCREG_APIC_BASE, 0xfee00900); 445 446 tc->setMiscReg(MISCREG_TSG_BASE, GDTVirtAddr); 447 tc->setMiscReg(MISCREG_TSG_LIMIT, 0xffff); 448 449 tc->setMiscReg(MISCREG_IDTR_BASE, IDTVirtAddr); 450 tc->setMiscReg(MISCREG_IDTR_LIMIT, 0xffff); 451 452 /* enabling syscall and sysret */ 453 RegVal star = ((RegVal)sret << 48) | ((RegVal)scall << 32); 454 tc->setMiscReg(MISCREG_STAR, star); 455 RegVal lstar = (RegVal)syscallCodeVirtAddr; 456 tc->setMiscReg(MISCREG_LSTAR, lstar); 457 RegVal sfmask = (1 << 8) | (1 << 10); // TF | DF 458 tc->setMiscReg(MISCREG_SF_MASK, sfmask); 459 } 460 461 /* Set up the content of the TSS and write it to physical memory. */ 462 463 struct { 464 uint32_t reserved0; // +00h 465 uint32_t RSP0_low; // +04h 466 uint32_t RSP0_high; // +08h 467 uint32_t RSP1_low; // +0Ch 468 uint32_t RSP1_high; // +10h 469 uint32_t RSP2_low; // +14h 470 uint32_t RSP2_high; // +18h 471 uint32_t reserved1; // +1Ch 472 uint32_t reserved2; // +20h 473 uint32_t IST1_low; // +24h 474 uint32_t IST1_high; // +28h 475 uint32_t IST2_low; // +2Ch 476 uint32_t IST2_high; // +30h 477 uint32_t IST3_low; // +34h 478 uint32_t IST3_high; // +38h 479 uint32_t IST4_low; // +3Ch 480 uint32_t IST4_high; // +40h 481 uint32_t IST5_low; // +44h 482 uint32_t IST5_high; // +48h 483 uint32_t IST6_low; // +4Ch 484 uint32_t IST6_high; // +50h 485 uint32_t IST7_low; // +54h 486 uint32_t IST7_high; // +58h 487 uint32_t reserved3; // +5Ch 488 uint32_t reserved4; // +60h 489 uint16_t reserved5; // +64h 490 uint16_t IO_MapBase; // +66h 491 } tss; 492 493 /** setting Interrupt Stack Table */ 494 uint64_t IST_start = ISTVirtAddr + PageBytes; 495 tss.IST1_low = IST_start; 496 tss.IST1_high = IST_start >> 32; 497 tss.RSP0_low = tss.IST1_low; 498 tss.RSP0_high = tss.IST1_high; 499 tss.RSP1_low = tss.IST1_low; 500 tss.RSP1_high = tss.IST1_high; 501 tss.RSP2_low = tss.IST1_low; 502 tss.RSP2_high = tss.IST1_high; 503 physProxy.writeBlob(tssPhysAddr, (uint8_t *)(&tss), sizeof(tss)); 504 505 /* Setting IDT gates */ 506 GateDescriptorLow PFGateLow = 0; 507 PFGateLow.offsetHigh = bits(PFHandlerVirtAddr, 31, 16); 508 PFGateLow.offsetLow = bits(PFHandlerVirtAddr, 15, 0); 509 PFGateLow.selector = csLowPL; 510 PFGateLow.p = 1; 511 PFGateLow.dpl = 0; 512 PFGateLow.type = 0xe; // gate interrupt type 513 PFGateLow.IST = 0; // setting IST to 0 and using RSP0 514 515 GateDescriptorHigh PFGateHigh = 0; 516 PFGateHigh.offset = bits(PFHandlerVirtAddr, 63, 32); 517 518 struct { 519 uint64_t low; 520 uint64_t high; 521 } PFGate = {PFGateLow, PFGateHigh}; 522 523 physProxy.writeBlob(idtPhysAddr + 0xE0, 524 (uint8_t *)(&PFGate), sizeof(PFGate)); 525 526 /* System call handler */ 527 uint8_t syscallBlob[] = { 528 // mov %rax, (0xffffc90000005600) 529 0x48, 0xa3, 0x00, 0x60, 0x00, 530 0x00, 0x00, 0xc9, 0xff, 0xff, 531 // sysret 532 0x48, 0x0f, 0x07 533 }; 534 535 physProxy.writeBlob(syscallCodePhysAddr, 536 syscallBlob, sizeof(syscallBlob)); 537 538 /** Page fault handler */ 539 uint8_t faultBlob[] = { 540 // mov %rax, (0xffffc90000005700) 541 0x48, 0xa3, 0x00, 0x61, 0x00, 542 0x00, 0x00, 0xc9, 0xff, 0xff, 543 // add $0x8, %rsp # skip error 544 0x48, 0x83, 0xc4, 0x08, 545 // iretq 546 0x48, 0xcf 547 }; 548 549 physProxy.writeBlob(pfHandlerPhysAddr, faultBlob, sizeof(faultBlob)); 550 551 /* Syscall handler */ 552 pTable->map(syscallCodeVirtAddr, syscallCodePhysAddr, 553 PageBytes, false); 554 /* GDT */ 555 pTable->map(GDTVirtAddr, gdtPhysAddr, PageBytes, false); 556 /* IDT */ 557 pTable->map(IDTVirtAddr, idtPhysAddr, PageBytes, false); 558 /* TSS */ 559 pTable->map(TSSVirtAddr, tssPhysAddr, PageBytes, false); 560 /* IST */ 561 pTable->map(ISTVirtAddr, istPhysAddr, PageBytes, false); 562 /* PF handler */ 563 pTable->map(PFHandlerVirtAddr, pfHandlerPhysAddr, PageBytes, false); 564 /* MMIO region for m5ops */ 565 pTable->map(MMIORegionVirtAddr, MMIORegionPhysAddr, 566 16 * PageBytes, false); 567 } else { 568 for (int i = 0; i < contextIds.size(); i++) { 569 ThreadContext * tc = system->getThreadContext(contextIds[i]); 570 571 SegAttr dataAttr = 0; 572 dataAttr.dpl = 3; 573 dataAttr.unusable = 0; 574 dataAttr.defaultSize = 1; 575 dataAttr.longMode = 1; 576 dataAttr.avl = 0; 577 dataAttr.granularity = 1; 578 dataAttr.present = 1; 579 dataAttr.type = 3; 580 dataAttr.writable = 1; 581 dataAttr.readable = 1; 582 dataAttr.expandDown = 0; 583 dataAttr.system = 1; 584 585 // Initialize the segment registers. 586 for (int seg = 0; seg < NUM_SEGMENTREGS; seg++) { 587 tc->setMiscRegNoEffect(MISCREG_SEG_BASE(seg), 0); 588 tc->setMiscRegNoEffect(MISCREG_SEG_EFF_BASE(seg), 0); 589 tc->setMiscRegNoEffect(MISCREG_SEG_ATTR(seg), dataAttr); 590 } 591 592 SegAttr csAttr = 0; 593 csAttr.dpl = 3; 594 csAttr.unusable = 0; 595 csAttr.defaultSize = 0; 596 csAttr.longMode = 1; 597 csAttr.avl = 0; 598 csAttr.granularity = 1; 599 csAttr.present = 1; 600 csAttr.type = 10; 601 csAttr.writable = 0; 602 csAttr.readable = 1; 603 csAttr.expandDown = 0; 604 csAttr.system = 1; 605 606 tc->setMiscRegNoEffect(MISCREG_CS_ATTR, csAttr); 607 608 Efer efer = 0; 609 efer.sce = 1; // Enable system call extensions. 610 efer.lme = 1; // Enable long mode. 611 efer.lma = 1; // Activate long mode. 612 efer.nxe = 1; // Enable nx support. 613 efer.svme = 0; // Disable svm support for now. It isn't implemented. 614 efer.ffxsr = 1; // Turn on fast fxsave and fxrstor. 615 tc->setMiscReg(MISCREG_EFER, efer); 616 617 // Set up the registers that describe the operating mode. 618 CR0 cr0 = 0; 619 cr0.pg = 1; // Turn on paging. 620 cr0.cd = 0; // Don't disable caching. 621 cr0.nw = 0; // This is bit is defined to be ignored. 622 cr0.am = 0; // No alignment checking 623 cr0.wp = 0; // Supervisor mode can write read only pages 624 cr0.ne = 1; 625 cr0.et = 1; // This should always be 1 626 cr0.ts = 0; // We don't do task switching, so causing fp exceptions 627 // would be pointless. 628 cr0.em = 0; // Allow x87 instructions to execute natively. 629 cr0.mp = 1; // This doesn't really matter, but the manual suggests 630 // setting it to one. 631 cr0.pe = 1; // We're definitely in protected mode. 632 tc->setMiscReg(MISCREG_CR0, cr0); 633 634 tc->setMiscReg(MISCREG_MXCSR, 0x1f80); 635 } 636 } 637} 638 639void 640I386Process::initState() 641{ 642 X86Process::initState(); 643 644 argsInit(PageBytes); 645 646 /* 647 * Set up a GDT for this process. The whole GDT wouldn't really be for 648 * this process, but the only parts we care about are. 649 */ 650 allocateMem(_gdtStart, _gdtSize); 651 uint64_t zero = 0; 652 assert(_gdtSize % sizeof(zero) == 0); 653 for (Addr gdtCurrent = _gdtStart; 654 gdtCurrent < _gdtStart + _gdtSize; gdtCurrent += sizeof(zero)) { 655 initVirtMem.write(gdtCurrent, zero); 656 } 657 658 // Set up the vsyscall page for this process. 659 allocateMem(vsyscallPage.base, vsyscallPage.size); 660 uint8_t vsyscallBlob[] = { 661 0x51, // push %ecx 662 0x52, // push %edp 663 0x55, // push %ebp 664 0x89, 0xe5, // mov %esp, %ebp 665 0x0f, 0x34 // sysenter 666 }; 667 initVirtMem.writeBlob(vsyscallPage.base + vsyscallPage.vsyscallOffset, 668 vsyscallBlob, sizeof(vsyscallBlob)); 669 670 uint8_t vsysexitBlob[] = { 671 0x5d, // pop %ebp 672 0x5a, // pop %edx 673 0x59, // pop %ecx 674 0xc3 // ret 675 }; 676 initVirtMem.writeBlob(vsyscallPage.base + vsyscallPage.vsysexitOffset, 677 vsysexitBlob, sizeof(vsysexitBlob)); 678 679 for (int i = 0; i < contextIds.size(); i++) { 680 ThreadContext * tc = system->getThreadContext(contextIds[i]); 681 682 SegAttr dataAttr = 0; 683 dataAttr.dpl = 3; 684 dataAttr.unusable = 0; 685 dataAttr.defaultSize = 1; 686 dataAttr.longMode = 0; 687 dataAttr.avl = 0; 688 dataAttr.granularity = 1; 689 dataAttr.present = 1; 690 dataAttr.type = 3; 691 dataAttr.writable = 1; 692 dataAttr.readable = 1; 693 dataAttr.expandDown = 0; 694 dataAttr.system = 1; 695 696 // Initialize the segment registers. 697 for (int seg = 0; seg < NUM_SEGMENTREGS; seg++) { 698 tc->setMiscRegNoEffect(MISCREG_SEG_BASE(seg), 0); 699 tc->setMiscRegNoEffect(MISCREG_SEG_EFF_BASE(seg), 0); 700 tc->setMiscRegNoEffect(MISCREG_SEG_ATTR(seg), dataAttr); 701 tc->setMiscRegNoEffect(MISCREG_SEG_SEL(seg), 0xB); 702 tc->setMiscRegNoEffect(MISCREG_SEG_LIMIT(seg), (uint32_t)(-1)); 703 } 704 705 SegAttr csAttr = 0; 706 csAttr.dpl = 3; 707 csAttr.unusable = 0; 708 csAttr.defaultSize = 1; 709 csAttr.longMode = 0; 710 csAttr.avl = 0; 711 csAttr.granularity = 1; 712 csAttr.present = 1; 713 csAttr.type = 0xa; 714 csAttr.writable = 0; 715 csAttr.readable = 1; 716 csAttr.expandDown = 0; 717 csAttr.system = 1; 718 719 tc->setMiscRegNoEffect(MISCREG_CS_ATTR, csAttr); 720 721 tc->setMiscRegNoEffect(MISCREG_TSG_BASE, _gdtStart); 722 tc->setMiscRegNoEffect(MISCREG_TSG_EFF_BASE, _gdtStart); 723 tc->setMiscRegNoEffect(MISCREG_TSG_LIMIT, _gdtStart + _gdtSize - 1); 724 725 // Set the LDT selector to 0 to deactivate it. 726 tc->setMiscRegNoEffect(MISCREG_TSL, 0); 727 728 Efer efer = 0; 729 efer.sce = 1; // Enable system call extensions. 730 efer.lme = 1; // Enable long mode. 731 efer.lma = 0; // Deactivate long mode. 732 efer.nxe = 1; // Enable nx support. 733 efer.svme = 0; // Disable svm support for now. It isn't implemented. 734 efer.ffxsr = 1; // Turn on fast fxsave and fxrstor. 735 tc->setMiscReg(MISCREG_EFER, efer); 736 737 // Set up the registers that describe the operating mode. 738 CR0 cr0 = 0; 739 cr0.pg = 1; // Turn on paging. 740 cr0.cd = 0; // Don't disable caching. 741 cr0.nw = 0; // This is bit is defined to be ignored. 742 cr0.am = 0; // No alignment checking 743 cr0.wp = 0; // Supervisor mode can write read only pages 744 cr0.ne = 1; 745 cr0.et = 1; // This should always be 1 746 cr0.ts = 0; // We don't do task switching, so causing fp exceptions 747 // would be pointless. 748 cr0.em = 0; // Allow x87 instructions to execute natively. 749 cr0.mp = 1; // This doesn't really matter, but the manual suggests 750 // setting it to one. 751 cr0.pe = 1; // We're definitely in protected mode. 752 tc->setMiscReg(MISCREG_CR0, cr0); 753 754 tc->setMiscReg(MISCREG_MXCSR, 0x1f80); 755 } 756} 757 758template<class IntType> 759void 760X86Process::argsInit(int pageSize, 761 std::vector<AuxVector<IntType> > extraAuxvs) 762{ 763 int intSize = sizeof(IntType); 764
|