sys_machdep.c revision 296954
1/*- 2 * Copyright (c) 2003 Peter Wemm. 3 * Copyright (c) 1990 The Regents of the University of California. 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 4. Neither the name of the University nor the names of its contributors 15 * may be used to endorse or promote products derived from this software 16 * without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 * 30 * from: @(#)sys_machdep.c 5.5 (Berkeley) 1/19/91 31 */ 32 33#include <sys/cdefs.h> 34__FBSDID("$FreeBSD: releng/10.1/sys/amd64/amd64/sys_machdep.c 296954 2016-03-16 22:30:56Z glebius $"); 35 36#include "opt_capsicum.h" 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/capability.h> 41#include <sys/kernel.h> 42#include <sys/lock.h> 43#include <sys/malloc.h> 44#include <sys/mutex.h> 45#include <sys/priv.h> 46#include <sys/proc.h> 47#include <sys/sysproto.h> 48#include <sys/uio.h> 49 50#include <vm/vm.h> 51#include <vm/pmap.h> 52#include <vm/vm_kern.h> /* for kernel_map */ 53#include <vm/vm_extern.h> 54 55#include <machine/frame.h> 56#include <machine/md_var.h> 57#include <machine/pcb.h> 58#include <machine/specialreg.h> 59#include <machine/sysarch.h> 60#include <machine/tss.h> 61#include <machine/vmparam.h> 62 63#include <security/audit/audit.h> 64 65#define MAX_LD 8192 66 67int max_ldt_segment = 1024; 68SYSCTL_INT(_machdep, OID_AUTO, max_ldt_segment, CTLFLAG_RDTUN, 69 &max_ldt_segment, 0, 70 "Maximum number of allowed LDT segments in the single address space"); 71 72static void 73max_ldt_segment_init(void *arg __unused) 74{ 75 76 TUNABLE_INT_FETCH("machdep.max_ldt_segment", &max_ldt_segment); 77 if (max_ldt_segment <= 0) 78 max_ldt_segment = 1; 79 if (max_ldt_segment > MAX_LD) 80 max_ldt_segment = MAX_LD; 81} 82SYSINIT(maxldt, SI_SUB_VM_CONF, SI_ORDER_ANY, max_ldt_segment_init, NULL); 83 84#ifdef notyet 85#ifdef SMP 86static void set_user_ldt_rv(struct vmspace *vmsp); 87#endif 88#endif 89static void user_ldt_derefl(struct proc_ldt *pldt); 90 91#ifndef _SYS_SYSPROTO_H_ 92struct sysarch_args { 93 int op; 94 char *parms; 95}; 96#endif 97 98int 99sysarch_ldt(struct thread *td, struct sysarch_args *uap, int uap_space) 100{ 101 struct i386_ldt_args *largs, la; 102 struct user_segment_descriptor *lp; 103 int error = 0; 104 105 /* 106 * XXXKIB check that the BSM generation code knows to encode 107 * the op argument. 108 */ 109 AUDIT_ARG_CMD(uap->op); 110 if (uap_space == UIO_USERSPACE) { 111 error = copyin(uap->parms, &la, sizeof(struct i386_ldt_args)); 112 if (error != 0) 113 return (error); 114 largs = &la; 115 } else 116 largs = (struct i386_ldt_args *)uap->parms; 117 118 switch (uap->op) { 119 case I386_GET_LDT: 120 error = amd64_get_ldt(td, largs); 121 break; 122 case I386_SET_LDT: 123 if (largs->descs != NULL && largs->num > max_ldt_segment) 124 return (EINVAL); 125 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 126 if (largs->descs != NULL) { 127 lp = malloc(largs->num * sizeof(struct 128 user_segment_descriptor), M_TEMP, M_WAITOK); 129 error = copyin(largs->descs, lp, largs->num * 130 sizeof(struct user_segment_descriptor)); 131 if (error == 0) 132 error = amd64_set_ldt(td, largs, lp); 133 free(lp, M_TEMP); 134 } else { 135 error = amd64_set_ldt(td, largs, NULL); 136 } 137 break; 138 } 139 return (error); 140} 141 142void 143update_gdt_gsbase(struct thread *td, uint32_t base) 144{ 145 struct user_segment_descriptor *sd; 146 147 if (td != curthread) 148 return; 149 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 150 critical_enter(); 151 sd = PCPU_GET(gs32p); 152 sd->sd_lobase = base & 0xffffff; 153 sd->sd_hibase = (base >> 24) & 0xff; 154 critical_exit(); 155} 156 157void 158update_gdt_fsbase(struct thread *td, uint32_t base) 159{ 160 struct user_segment_descriptor *sd; 161 162 if (td != curthread) 163 return; 164 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 165 critical_enter(); 166 sd = PCPU_GET(fs32p); 167 sd->sd_lobase = base & 0xffffff; 168 sd->sd_hibase = (base >> 24) & 0xff; 169 critical_exit(); 170} 171 172int 173sysarch(td, uap) 174 struct thread *td; 175 register struct sysarch_args *uap; 176{ 177 int error = 0; 178 struct pcb *pcb = curthread->td_pcb; 179 uint32_t i386base; 180 uint64_t a64base; 181 struct i386_ioperm_args iargs; 182 struct i386_get_xfpustate i386xfpu; 183 struct amd64_get_xfpustate a64xfpu; 184 185#ifdef CAPABILITY_MODE 186 /* 187 * When adding new operations, add a new case statement here to 188 * explicitly indicate whether or not the operation is safe to 189 * perform in capability mode. 190 */ 191 if (IN_CAPABILITY_MODE(td)) { 192 switch (uap->op) { 193 case I386_GET_LDT: 194 case I386_SET_LDT: 195 case I386_GET_IOPERM: 196 case I386_GET_FSBASE: 197 case I386_SET_FSBASE: 198 case I386_GET_GSBASE: 199 case I386_SET_GSBASE: 200 case I386_GET_XFPUSTATE: 201 case AMD64_GET_FSBASE: 202 case AMD64_SET_FSBASE: 203 case AMD64_GET_GSBASE: 204 case AMD64_SET_GSBASE: 205 case AMD64_GET_XFPUSTATE: 206 break; 207 208 case I386_SET_IOPERM: 209 default: 210#ifdef KTRACE 211 if (KTRPOINT(td, KTR_CAPFAIL)) 212 ktrcapfail(CAPFAIL_SYSCALL, NULL, NULL); 213#endif 214 return (ECAPMODE); 215 } 216 } 217#endif 218 219 if (uap->op == I386_GET_LDT || uap->op == I386_SET_LDT) 220 return (sysarch_ldt(td, uap, UIO_USERSPACE)); 221 /* 222 * XXXKIB check that the BSM generation code knows to encode 223 * the op argument. 224 */ 225 AUDIT_ARG_CMD(uap->op); 226 switch (uap->op) { 227 case I386_GET_IOPERM: 228 case I386_SET_IOPERM: 229 if ((error = copyin(uap->parms, &iargs, 230 sizeof(struct i386_ioperm_args))) != 0) 231 return (error); 232 break; 233 case I386_GET_XFPUSTATE: 234 if ((error = copyin(uap->parms, &i386xfpu, 235 sizeof(struct i386_get_xfpustate))) != 0) 236 return (error); 237 a64xfpu.addr = (void *)(uintptr_t)i386xfpu.addr; 238 a64xfpu.len = i386xfpu.len; 239 break; 240 case AMD64_GET_XFPUSTATE: 241 if ((error = copyin(uap->parms, &a64xfpu, 242 sizeof(struct amd64_get_xfpustate))) != 0) 243 return (error); 244 break; 245 default: 246 break; 247 } 248 249 switch (uap->op) { 250 case I386_GET_IOPERM: 251 error = amd64_get_ioperm(td, &iargs); 252 if (error == 0) 253 error = copyout(&iargs, uap->parms, 254 sizeof(struct i386_ioperm_args)); 255 break; 256 case I386_SET_IOPERM: 257 error = amd64_set_ioperm(td, &iargs); 258 break; 259 case I386_GET_FSBASE: 260 i386base = pcb->pcb_fsbase; 261 error = copyout(&i386base, uap->parms, sizeof(i386base)); 262 break; 263 case I386_SET_FSBASE: 264 error = copyin(uap->parms, &i386base, sizeof(i386base)); 265 if (!error) { 266 pcb->pcb_fsbase = i386base; 267 td->td_frame->tf_fs = _ufssel; 268 update_gdt_fsbase(td, i386base); 269 } 270 break; 271 case I386_GET_GSBASE: 272 i386base = pcb->pcb_gsbase; 273 error = copyout(&i386base, uap->parms, sizeof(i386base)); 274 break; 275 case I386_SET_GSBASE: 276 error = copyin(uap->parms, &i386base, sizeof(i386base)); 277 if (!error) { 278 pcb->pcb_gsbase = i386base; 279 td->td_frame->tf_gs = _ugssel; 280 update_gdt_gsbase(td, i386base); 281 } 282 break; 283 case AMD64_GET_FSBASE: 284 error = copyout(&pcb->pcb_fsbase, uap->parms, sizeof(pcb->pcb_fsbase)); 285 break; 286 287 case AMD64_SET_FSBASE: 288 error = copyin(uap->parms, &a64base, sizeof(a64base)); 289 if (!error) { 290 if (a64base < VM_MAXUSER_ADDRESS) { 291 pcb->pcb_fsbase = a64base; 292 set_pcb_flags(pcb, PCB_FULL_IRET); 293 td->td_frame->tf_fs = _ufssel; 294 } else 295 error = EINVAL; 296 } 297 break; 298 299 case AMD64_GET_GSBASE: 300 error = copyout(&pcb->pcb_gsbase, uap->parms, sizeof(pcb->pcb_gsbase)); 301 break; 302 303 case AMD64_SET_GSBASE: 304 error = copyin(uap->parms, &a64base, sizeof(a64base)); 305 if (!error) { 306 if (a64base < VM_MAXUSER_ADDRESS) { 307 pcb->pcb_gsbase = a64base; 308 set_pcb_flags(pcb, PCB_FULL_IRET); 309 td->td_frame->tf_gs = _ugssel; 310 } else 311 error = EINVAL; 312 } 313 break; 314 315 case I386_GET_XFPUSTATE: 316 case AMD64_GET_XFPUSTATE: 317 if (a64xfpu.len > cpu_max_ext_state_size - 318 sizeof(struct savefpu)) 319 return (EINVAL); 320 fpugetregs(td); 321 error = copyout((char *)(get_pcb_user_save_td(td) + 1), 322 a64xfpu.addr, a64xfpu.len); 323 return (error); 324 325 default: 326 error = EINVAL; 327 break; 328 } 329 return (error); 330} 331 332int 333amd64_set_ioperm(td, uap) 334 struct thread *td; 335 struct i386_ioperm_args *uap; 336{ 337 int i, error; 338 char *iomap; 339 struct amd64tss *tssp; 340 struct system_segment_descriptor *tss_sd; 341 u_long *addr; 342 struct pcb *pcb; 343 344 if ((error = priv_check(td, PRIV_IO)) != 0) 345 return (error); 346 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 347 return (error); 348 if (uap->start + uap->length > IOPAGES * PAGE_SIZE * NBBY) 349 return (EINVAL); 350 351 /* 352 * XXX 353 * While this is restricted to root, we should probably figure out 354 * whether any other driver is using this i/o address, as so not to 355 * cause confusion. This probably requires a global 'usage registry'. 356 */ 357 pcb = td->td_pcb; 358 if (pcb->pcb_tssp == NULL) { 359 tssp = (struct amd64tss *)kmem_malloc(kernel_arena, 360 ctob(IOPAGES+1), M_WAITOK); 361 if (tssp == NULL) 362 return (ENOMEM); 363 iomap = (char *)&tssp[1]; 364 addr = (u_long *)iomap; 365 for (i = 0; i < (ctob(IOPAGES) + 1) / sizeof(u_long); i++) 366 *addr++ = ~0; 367 critical_enter(); 368 /* Takes care of tss_rsp0. */ 369 memcpy(tssp, &common_tss[PCPU_GET(cpuid)], 370 sizeof(struct amd64tss)); 371 tssp->tss_iobase = sizeof(*tssp); 372 pcb->pcb_tssp = tssp; 373 tss_sd = PCPU_GET(tss); 374 tss_sd->sd_lobase = (u_long)tssp & 0xffffff; 375 tss_sd->sd_hibase = ((u_long)tssp >> 24) & 0xfffffffffful; 376 tss_sd->sd_type = SDT_SYSTSS; 377 ltr(GSEL(GPROC0_SEL, SEL_KPL)); 378 PCPU_SET(tssp, tssp); 379 critical_exit(); 380 } else 381 iomap = (char *)&pcb->pcb_tssp[1]; 382 for (i = uap->start; i < uap->start + uap->length; i++) { 383 if (uap->enable) 384 iomap[i >> 3] &= ~(1 << (i & 7)); 385 else 386 iomap[i >> 3] |= (1 << (i & 7)); 387 } 388 return (error); 389} 390 391int 392amd64_get_ioperm(td, uap) 393 struct thread *td; 394 struct i386_ioperm_args *uap; 395{ 396 int i, state; 397 char *iomap; 398 399 if (uap->start >= IOPAGES * PAGE_SIZE * NBBY) 400 return (EINVAL); 401 if (td->td_pcb->pcb_tssp == NULL) { 402 uap->length = 0; 403 goto done; 404 } 405 406 iomap = (char *)&td->td_pcb->pcb_tssp[1]; 407 408 i = uap->start; 409 state = (iomap[i >> 3] >> (i & 7)) & 1; 410 uap->enable = !state; 411 uap->length = 1; 412 413 for (i = uap->start + 1; i < IOPAGES * PAGE_SIZE * NBBY; i++) { 414 if (state != ((iomap[i >> 3] >> (i & 7)) & 1)) 415 break; 416 uap->length++; 417 } 418 419done: 420 return (0); 421} 422 423/* 424 * Update the GDT entry pointing to the LDT to point to the LDT of the 425 * current process. 426 */ 427void 428set_user_ldt(struct mdproc *mdp) 429{ 430 431 critical_enter(); 432 *PCPU_GET(ldt) = mdp->md_ldt_sd; 433 lldt(GSEL(GUSERLDT_SEL, SEL_KPL)); 434 critical_exit(); 435} 436 437#ifdef notyet 438#ifdef SMP 439static void 440set_user_ldt_rv(struct vmspace *vmsp) 441{ 442 struct thread *td; 443 444 td = curthread; 445 if (vmsp != td->td_proc->p_vmspace) 446 return; 447 448 set_user_ldt(&td->td_proc->p_md); 449} 450#endif 451#endif 452 453struct proc_ldt * 454user_ldt_alloc(struct proc *p, int force) 455{ 456 struct proc_ldt *pldt, *new_ldt; 457 struct mdproc *mdp; 458 struct soft_segment_descriptor sldt; 459 460 mtx_assert(&dt_lock, MA_OWNED); 461 mdp = &p->p_md; 462 if (!force && mdp->md_ldt != NULL) 463 return (mdp->md_ldt); 464 mtx_unlock(&dt_lock); 465 new_ldt = malloc(sizeof(struct proc_ldt), M_SUBPROC, M_WAITOK); 466 new_ldt->ldt_base = (caddr_t)kmem_malloc(kernel_arena, 467 max_ldt_segment * sizeof(struct user_segment_descriptor), 468 M_WAITOK | M_ZERO); 469 if (new_ldt->ldt_base == NULL) { 470 FREE(new_ldt, M_SUBPROC); 471 mtx_lock(&dt_lock); 472 return (NULL); 473 } 474 new_ldt->ldt_refcnt = 1; 475 sldt.ssd_base = (uint64_t)new_ldt->ldt_base; 476 sldt.ssd_limit = max_ldt_segment * 477 sizeof(struct user_segment_descriptor) - 1; 478 sldt.ssd_type = SDT_SYSLDT; 479 sldt.ssd_dpl = SEL_KPL; 480 sldt.ssd_p = 1; 481 sldt.ssd_long = 0; 482 sldt.ssd_def32 = 0; 483 sldt.ssd_gran = 0; 484 mtx_lock(&dt_lock); 485 pldt = mdp->md_ldt; 486 if (pldt != NULL && !force) { 487 kmem_free(kernel_arena, (vm_offset_t)new_ldt->ldt_base, 488 max_ldt_segment * sizeof(struct user_segment_descriptor)); 489 free(new_ldt, M_SUBPROC); 490 return (pldt); 491 } 492 493 if (pldt != NULL) { 494 bcopy(pldt->ldt_base, new_ldt->ldt_base, max_ldt_segment * 495 sizeof(struct user_segment_descriptor)); 496 user_ldt_derefl(pldt); 497 } 498 ssdtosyssd(&sldt, &p->p_md.md_ldt_sd); 499 atomic_store_rel_ptr((volatile uintptr_t *)&mdp->md_ldt, 500 (uintptr_t)new_ldt); 501 if (p == curproc) 502 set_user_ldt(mdp); 503 504 return (mdp->md_ldt); 505} 506 507void 508user_ldt_free(struct thread *td) 509{ 510 struct proc *p = td->td_proc; 511 struct mdproc *mdp = &p->p_md; 512 struct proc_ldt *pldt; 513 514 mtx_assert(&dt_lock, MA_OWNED); 515 if ((pldt = mdp->md_ldt) == NULL) { 516 mtx_unlock(&dt_lock); 517 return; 518 } 519 520 mdp->md_ldt = NULL; 521 bzero(&mdp->md_ldt_sd, sizeof(mdp->md_ldt_sd)); 522 if (td == curthread) 523 lldt(GSEL(GNULL_SEL, SEL_KPL)); 524 user_ldt_deref(pldt); 525} 526 527static void 528user_ldt_derefl(struct proc_ldt *pldt) 529{ 530 531 if (--pldt->ldt_refcnt == 0) { 532 kmem_free(kernel_arena, (vm_offset_t)pldt->ldt_base, 533 max_ldt_segment * sizeof(struct user_segment_descriptor)); 534 free(pldt, M_SUBPROC); 535 } 536} 537 538void 539user_ldt_deref(struct proc_ldt *pldt) 540{ 541 542 mtx_assert(&dt_lock, MA_OWNED); 543 user_ldt_derefl(pldt); 544 mtx_unlock(&dt_lock); 545} 546 547/* 548 * Note for the authors of compat layers (linux, etc): copyout() in 549 * the function below is not a problem since it presents data in 550 * arch-specific format (i.e. i386-specific in this case), not in 551 * the OS-specific one. 552 */ 553int 554amd64_get_ldt(td, uap) 555 struct thread *td; 556 struct i386_ldt_args *uap; 557{ 558 int error = 0; 559 struct proc_ldt *pldt; 560 int num; 561 struct user_segment_descriptor *lp; 562 563#ifdef DEBUG 564 printf("amd64_get_ldt: start=%d num=%d descs=%p\n", 565 uap->start, uap->num, (void *)uap->descs); 566#endif 567 568 if ((pldt = td->td_proc->p_md.md_ldt) != NULL) { 569 lp = &((struct user_segment_descriptor *)(pldt->ldt_base)) 570 [uap->start]; 571 num = min(uap->num, max_ldt_segment); 572 } else 573 return (EINVAL); 574 575 if ((uap->start > (unsigned int)max_ldt_segment) || 576 ((unsigned int)num > (unsigned int)max_ldt_segment) || 577 ((unsigned int)(uap->start + num) > (unsigned int)max_ldt_segment)) 578 return(EINVAL); 579 580 error = copyout(lp, uap->descs, num * 581 sizeof(struct user_segment_descriptor)); 582 if (!error) 583 td->td_retval[0] = num; 584 585 return(error); 586} 587 588int 589amd64_set_ldt(td, uap, descs) 590 struct thread *td; 591 struct i386_ldt_args *uap; 592 struct user_segment_descriptor *descs; 593{ 594 int error = 0; 595 unsigned int largest_ld, i; 596 struct mdproc *mdp = &td->td_proc->p_md; 597 struct proc_ldt *pldt; 598 struct user_segment_descriptor *dp; 599 struct proc *p; 600 601#ifdef DEBUG 602 printf("amd64_set_ldt: start=%d num=%d descs=%p\n", 603 uap->start, uap->num, (void *)uap->descs); 604#endif 605 606 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 607 p = td->td_proc; 608 if (descs == NULL) { 609 /* Free descriptors */ 610 if (uap->start == 0 && uap->num == 0) 611 uap->num = max_ldt_segment; 612 if (uap->num == 0) 613 return (EINVAL); 614 if ((pldt = mdp->md_ldt) == NULL || 615 uap->start >= max_ldt_segment) 616 return (0); 617 largest_ld = uap->start + uap->num; 618 if (largest_ld > max_ldt_segment) 619 largest_ld = max_ldt_segment; 620 i = largest_ld - uap->start; 621 mtx_lock(&dt_lock); 622 bzero(&((struct user_segment_descriptor *)(pldt->ldt_base)) 623 [uap->start], sizeof(struct user_segment_descriptor) * i); 624 mtx_unlock(&dt_lock); 625 return (0); 626 } 627 628 if (!(uap->start == LDT_AUTO_ALLOC && uap->num == 1)) { 629 /* verify range of descriptors to modify */ 630 largest_ld = uap->start + uap->num; 631 if (uap->start >= max_ldt_segment || 632 largest_ld > max_ldt_segment) 633 return (EINVAL); 634 } 635 636 /* Check descriptors for access violations */ 637 for (i = 0; i < uap->num; i++) { 638 dp = &descs[i]; 639 640 switch (dp->sd_type) { 641 case SDT_SYSNULL: /* system null */ 642 dp->sd_p = 0; 643 break; 644 case SDT_SYS286TSS: 645 case SDT_SYSLDT: 646 case SDT_SYS286BSY: 647 case SDT_SYS286CGT: 648 case SDT_SYSTASKGT: 649 case SDT_SYS286IGT: 650 case SDT_SYS286TGT: 651 case SDT_SYSNULL2: 652 case SDT_SYSTSS: 653 case SDT_SYSNULL3: 654 case SDT_SYSBSY: 655 case SDT_SYSCGT: 656 case SDT_SYSNULL4: 657 case SDT_SYSIGT: 658 case SDT_SYSTGT: 659 /* I can't think of any reason to allow a user proc 660 * to create a segment of these types. They are 661 * for OS use only. 662 */ 663 return (EACCES); 664 /*NOTREACHED*/ 665 666 /* memory segment types */ 667 case SDT_MEMEC: /* memory execute only conforming */ 668 case SDT_MEMEAC: /* memory execute only accessed conforming */ 669 case SDT_MEMERC: /* memory execute read conforming */ 670 case SDT_MEMERAC: /* memory execute read accessed conforming */ 671 /* Must be "present" if executable and conforming. */ 672 if (dp->sd_p == 0) 673 return (EACCES); 674 break; 675 case SDT_MEMRO: /* memory read only */ 676 case SDT_MEMROA: /* memory read only accessed */ 677 case SDT_MEMRW: /* memory read write */ 678 case SDT_MEMRWA: /* memory read write accessed */ 679 case SDT_MEMROD: /* memory read only expand dwn limit */ 680 case SDT_MEMRODA: /* memory read only expand dwn lim accessed */ 681 case SDT_MEMRWD: /* memory read write expand dwn limit */ 682 case SDT_MEMRWDA: /* memory read write expand dwn lim acessed */ 683 case SDT_MEME: /* memory execute only */ 684 case SDT_MEMEA: /* memory execute only accessed */ 685 case SDT_MEMER: /* memory execute read */ 686 case SDT_MEMERA: /* memory execute read accessed */ 687 break; 688 default: 689 return(EINVAL); 690 /*NOTREACHED*/ 691 } 692 693 /* Only user (ring-3) descriptors may be present. */ 694 if ((dp->sd_p != 0) && (dp->sd_dpl != SEL_UPL)) 695 return (EACCES); 696 } 697 698 if (uap->start == LDT_AUTO_ALLOC && uap->num == 1) { 699 /* Allocate a free slot */ 700 mtx_lock(&dt_lock); 701 pldt = user_ldt_alloc(p, 0); 702 if (pldt == NULL) { 703 mtx_unlock(&dt_lock); 704 return (ENOMEM); 705 } 706 707 /* 708 * start scanning a bit up to leave room for NVidia and 709 * Wine, which still user the "Blat" method of allocation. 710 */ 711 i = 16; 712 dp = &((struct user_segment_descriptor *)(pldt->ldt_base))[i]; 713 for (; i < max_ldt_segment; ++i, ++dp) { 714 if (dp->sd_type == SDT_SYSNULL) 715 break; 716 } 717 if (i >= max_ldt_segment) { 718 mtx_unlock(&dt_lock); 719 return (ENOSPC); 720 } 721 uap->start = i; 722 error = amd64_set_ldt_data(td, i, 1, descs); 723 mtx_unlock(&dt_lock); 724 } else { 725 largest_ld = uap->start + uap->num; 726 if (largest_ld > max_ldt_segment) 727 return (EINVAL); 728 mtx_lock(&dt_lock); 729 if (user_ldt_alloc(p, 0) != NULL) { 730 error = amd64_set_ldt_data(td, uap->start, uap->num, 731 descs); 732 } 733 mtx_unlock(&dt_lock); 734 } 735 if (error == 0) 736 td->td_retval[0] = uap->start; 737 return (error); 738} 739 740int 741amd64_set_ldt_data(struct thread *td, int start, int num, 742 struct user_segment_descriptor *descs) 743{ 744 struct mdproc *mdp = &td->td_proc->p_md; 745 struct proc_ldt *pldt = mdp->md_ldt; 746 747 mtx_assert(&dt_lock, MA_OWNED); 748 749 /* Fill in range */ 750 bcopy(descs, 751 &((struct user_segment_descriptor *)(pldt->ldt_base))[start], 752 num * sizeof(struct user_segment_descriptor)); 753 return (0); 754} 755