1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2003 Peter Wemm. 5 * Copyright (c) 1990 The Regents of the University of California. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * from: @(#)sys_machdep.c 5.5 (Berkeley) 1/19/91 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD$"); 37 38 #include "opt_capsicum.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/capsicum.h> 43 #include <sys/kernel.h> 44 #include <sys/lock.h> 45 #include <sys/malloc.h> 46 #include <sys/mutex.h> 47 #include <sys/priv.h> 48 #include <sys/proc.h> 49 #include <sys/smp.h> 50 #include <sys/sysproto.h> 51 #include <sys/uio.h> 52 53 #include <vm/vm.h> 54 #include <vm/pmap.h> 55 #include <vm/vm_kern.h> /* for kernel_map */ 56 #include <vm/vm_extern.h> 57 58 #include <machine/frame.h> 59 #include <machine/md_var.h> 60 #include <machine/pcb.h> 61 #include <machine/specialreg.h> 62 #include <machine/sysarch.h> 63 #include <machine/tss.h> 64 #include <machine/vmparam.h> 65 66 #include <security/audit/audit.h> 67 68 static void user_ldt_deref(struct proc_ldt *pldt); 69 static void user_ldt_derefl(struct proc_ldt *pldt); 70 71 #define MAX_LD 8192 72 73 int max_ldt_segment = 512; 74 SYSCTL_INT(_machdep, OID_AUTO, max_ldt_segment, CTLFLAG_RDTUN, 75 &max_ldt_segment, 0, 76 "Maximum number of allowed LDT segments in the single address space"); 77 78 static void 79 max_ldt_segment_init(void *arg __unused) 80 { 81 82 if (max_ldt_segment <= 0) 83 max_ldt_segment = 1; 84 if (max_ldt_segment > MAX_LD) 85 max_ldt_segment = MAX_LD; 86 } 87 SYSINIT(maxldt, SI_SUB_VM_CONF, SI_ORDER_ANY, max_ldt_segment_init, NULL); 88 89 #ifndef _SYS_SYSPROTO_H_ 90 struct sysarch_args { 91 int op; 92 char *parms; 93 }; 94 #endif 95 96 int 97 sysarch_ldt(struct thread *td, struct sysarch_args *uap, int uap_space) 98 { 99 struct i386_ldt_args *largs, la; 100 struct user_segment_descriptor *lp; 101 int error = 0; 102 103 /* 104 * XXXKIB check that the BSM generation code knows to encode 105 * the op argument. 106 */ 107 AUDIT_ARG_CMD(uap->op); 108 if (uap_space == UIO_USERSPACE) { 109 error = copyin(uap->parms, &la, sizeof(struct i386_ldt_args)); 110 if (error != 0) 111 return (error); 112 largs = &la; 113 } else 114 largs = (struct i386_ldt_args *)uap->parms; 115 116 switch (uap->op) { 117 case I386_GET_LDT: 118 error = amd64_get_ldt(td, largs); 119 break; 120 case I386_SET_LDT: 121 if (largs->descs != NULL && largs->num > max_ldt_segment) 122 return (EINVAL); 123 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 124 if (largs->descs != NULL) { 125 lp = malloc(largs->num * sizeof(struct 126 user_segment_descriptor), M_TEMP, M_WAITOK); 127 error = copyin(largs->descs, lp, largs->num * 128 sizeof(struct user_segment_descriptor)); 129 if (error == 0) 130 error = amd64_set_ldt(td, largs, lp); 131 free(lp, M_TEMP); 132 } else { 133 error = amd64_set_ldt(td, largs, NULL); 134 } 135 break; 136 } 137 return (error); 138 } 139 140 void 141 update_gdt_gsbase(struct thread *td, uint32_t base) 142 { 143 struct user_segment_descriptor *sd; 144 145 if (td != curthread) 146 return; 147 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 148 critical_enter(); 149 sd = PCPU_GET(gs32p); 150 sd->sd_lobase = base & 0xffffff; 151 sd->sd_hibase = (base >> 24) & 0xff; 152 critical_exit(); 153 } 154 155 void 156 update_gdt_fsbase(struct thread *td, uint32_t base) 157 { 158 struct user_segment_descriptor *sd; 159 160 if (td != curthread) 161 return; 162 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 163 critical_enter(); 164 sd = PCPU_GET(fs32p); 165 sd->sd_lobase = base & 0xffffff; 166 sd->sd_hibase = (base >> 24) & 0xff; 167 critical_exit(); 168 } 169 170 int 171 sysarch(struct thread *td, struct sysarch_args *uap) 172 { 173 int error = 0; 174 struct pcb *pcb = curthread->td_pcb; 175 uint32_t i386base; 176 uint64_t a64base; 177 struct i386_ioperm_args iargs; 178 struct i386_get_xfpustate i386xfpu; 179 struct amd64_get_xfpustate a64xfpu; 180 181 #ifdef CAPABILITY_MODE 182 /* 183 * When adding new operations, add a new case statement here to 184 * explicitly indicate whether or not the operation is safe to 185 * perform in capability mode. 186 */ 187 if (IN_CAPABILITY_MODE(td)) { 188 switch (uap->op) { 189 case I386_GET_LDT: 190 case I386_SET_LDT: 191 case I386_GET_IOPERM: 192 case I386_GET_FSBASE: 193 case I386_SET_FSBASE: 194 case I386_GET_GSBASE: 195 case I386_SET_GSBASE: 196 case I386_GET_XFPUSTATE: 197 case AMD64_GET_FSBASE: 198 case AMD64_SET_FSBASE: 199 case AMD64_GET_GSBASE: 200 case AMD64_SET_GSBASE: 201 case AMD64_GET_XFPUSTATE: 202 break; 203 204 case I386_SET_IOPERM: 205 default: 206 #ifdef KTRACE 207 if (KTRPOINT(td, KTR_CAPFAIL)) 208 ktrcapfail(CAPFAIL_SYSCALL, NULL, NULL); 209 #endif 210 return (ECAPMODE); 211 } 212 } 213 #endif 214 215 if (uap->op == I386_GET_LDT || uap->op == I386_SET_LDT) 216 return (sysarch_ldt(td, uap, UIO_USERSPACE)); 217 /* 218 * XXXKIB check that the BSM generation code knows to encode 219 * the op argument. 220 */ 221 AUDIT_ARG_CMD(uap->op); 222 switch (uap->op) { 223 case I386_GET_IOPERM: 224 case I386_SET_IOPERM: 225 if ((error = copyin(uap->parms, &iargs, 226 sizeof(struct i386_ioperm_args))) != 0) 227 return (error); 228 break; 229 case I386_GET_XFPUSTATE: 230 if ((error = copyin(uap->parms, &i386xfpu, 231 sizeof(struct i386_get_xfpustate))) != 0) 232 return (error); 233 a64xfpu.addr = (void *)(uintptr_t)i386xfpu.addr; 234 a64xfpu.len = i386xfpu.len; 235 break; 236 case AMD64_GET_XFPUSTATE: 237 if ((error = copyin(uap->parms, &a64xfpu, 238 sizeof(struct amd64_get_xfpustate))) != 0) 239 return (error); 240 break; 241 default: 242 break; 243 } 244 245 switch (uap->op) { 246 case I386_GET_IOPERM: 247 error = amd64_get_ioperm(td, &iargs); 248 if (error == 0) 249 error = copyout(&iargs, uap->parms, 250 sizeof(struct i386_ioperm_args)); 251 break; 252 case I386_SET_IOPERM: 253 error = amd64_set_ioperm(td, &iargs); 254 break; 255 case I386_GET_FSBASE: 256 update_pcb_bases(pcb); 257 i386base = pcb->pcb_fsbase; 258 error = copyout(&i386base, uap->parms, sizeof(i386base)); 259 break; 260 case I386_SET_FSBASE: 261 error = copyin(uap->parms, &i386base, sizeof(i386base)); 262 if (!error) { 263 set_pcb_flags(pcb, PCB_FULL_IRET); 264 pcb->pcb_fsbase = i386base; 265 td->td_frame->tf_fs = _ufssel; 266 update_gdt_fsbase(td, i386base); 267 } 268 break; 269 case I386_GET_GSBASE: 270 update_pcb_bases(pcb); 271 i386base = pcb->pcb_gsbase; 272 error = copyout(&i386base, uap->parms, sizeof(i386base)); 273 break; 274 case I386_SET_GSBASE: 275 error = copyin(uap->parms, &i386base, sizeof(i386base)); 276 if (!error) { 277 set_pcb_flags(pcb, PCB_FULL_IRET); 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 update_pcb_bases(pcb); 285 error = copyout(&pcb->pcb_fsbase, uap->parms, 286 sizeof(pcb->pcb_fsbase)); 287 break; 288 289 case AMD64_SET_FSBASE: 290 error = copyin(uap->parms, &a64base, sizeof(a64base)); 291 if (!error) { 292 if (a64base < VM_MAXUSER_ADDRESS) { 293 set_pcb_flags(pcb, PCB_FULL_IRET); 294 pcb->pcb_fsbase = a64base; 295 td->td_frame->tf_fs = _ufssel; 296 } else 297 error = EINVAL; 298 } 299 break; 300 301 case AMD64_GET_GSBASE: 302 update_pcb_bases(pcb); 303 error = copyout(&pcb->pcb_gsbase, uap->parms, 304 sizeof(pcb->pcb_gsbase)); 305 break; 306 307 case AMD64_SET_GSBASE: 308 error = copyin(uap->parms, &a64base, sizeof(a64base)); 309 if (!error) { 310 if (a64base < VM_MAXUSER_ADDRESS) { 311 set_pcb_flags(pcb, PCB_FULL_IRET); 312 pcb->pcb_gsbase = a64base; 313 td->td_frame->tf_gs = _ugssel; 314 } else 315 error = EINVAL; 316 } 317 break; 318 319 case I386_GET_XFPUSTATE: 320 case AMD64_GET_XFPUSTATE: 321 if (a64xfpu.len > cpu_max_ext_state_size - 322 sizeof(struct savefpu)) 323 return (EINVAL); 324 fpugetregs(td); 325 error = copyout((char *)(get_pcb_user_save_td(td) + 1), 326 a64xfpu.addr, a64xfpu.len); 327 break; 328 329 default: 330 error = EINVAL; 331 break; 332 } 333 return (error); 334 } 335 336 int 337 amd64_set_ioperm(td, uap) 338 struct thread *td; 339 struct i386_ioperm_args *uap; 340 { 341 char *iomap; 342 struct amd64tss *tssp; 343 struct system_segment_descriptor *tss_sd; 344 struct pcb *pcb; 345 u_int i; 346 int error; 347 348 if ((error = priv_check(td, PRIV_IO)) != 0) 349 return (error); 350 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 351 return (error); 352 if (uap->start > uap->start + uap->length || 353 uap->start + uap->length > IOPAGES * PAGE_SIZE * NBBY) 354 return (EINVAL); 355 356 /* 357 * XXX 358 * While this is restricted to root, we should probably figure out 359 * whether any other driver is using this i/o address, as so not to 360 * cause confusion. This probably requires a global 'usage registry'. 361 */ 362 pcb = td->td_pcb; 363 if (pcb->pcb_tssp == NULL) { 364 tssp = (struct amd64tss *)kmem_malloc(kernel_arena, 365 ctob(IOPAGES + 1), M_WAITOK); 366 pmap_pti_add_kva((vm_offset_t)tssp, (vm_offset_t)tssp + 367 ctob(IOPAGES + 1), false); 368 iomap = (char *)&tssp[1]; 369 memset(iomap, 0xff, IOPERM_BITMAP_SIZE); 370 critical_enter(); 371 /* Takes care of tss_rsp0. */ 372 memcpy(tssp, &common_tss[PCPU_GET(cpuid)], 373 sizeof(struct amd64tss)); 374 tssp->tss_iobase = sizeof(*tssp); 375 pcb->pcb_tssp = tssp; 376 tss_sd = PCPU_GET(tss); 377 tss_sd->sd_lobase = (u_long)tssp & 0xffffff; 378 tss_sd->sd_hibase = ((u_long)tssp >> 24) & 0xfffffffffful; 379 tss_sd->sd_type = SDT_SYSTSS; 380 ltr(GSEL(GPROC0_SEL, SEL_KPL)); 381 PCPU_SET(tssp, tssp); 382 critical_exit(); 383 } else 384 iomap = (char *)&pcb->pcb_tssp[1]; 385 for (i = uap->start; i < uap->start + uap->length; i++) { 386 if (uap->enable) 387 iomap[i >> 3] &= ~(1 << (i & 7)); 388 else 389 iomap[i >> 3] |= (1 << (i & 7)); 390 } 391 return (error); 392 } 393 394 int 395 amd64_get_ioperm(td, uap) 396 struct thread *td; 397 struct i386_ioperm_args *uap; 398 { 399 int i, state; 400 char *iomap; 401 402 if (uap->start >= IOPAGES * PAGE_SIZE * NBBY) 403 return (EINVAL); 404 if (td->td_pcb->pcb_tssp == NULL) { 405 uap->length = 0; 406 goto done; 407 } 408 409 iomap = (char *)&td->td_pcb->pcb_tssp[1]; 410 411 i = uap->start; 412 state = (iomap[i >> 3] >> (i & 7)) & 1; 413 uap->enable = !state; 414 uap->length = 1; 415 416 for (i = uap->start + 1; i < IOPAGES * PAGE_SIZE * NBBY; i++) { 417 if (state != ((iomap[i >> 3] >> (i & 7)) & 1)) 418 break; 419 uap->length++; 420 } 421 422 done: 423 return (0); 424 } 425 426 /* 427 * Update the GDT entry pointing to the LDT to point to the LDT of the 428 * current process. 429 */ 430 static void 431 set_user_ldt(struct mdproc *mdp) 432 { 433 434 *PCPU_GET(ldt) = mdp->md_ldt_sd; 435 lldt(GSEL(GUSERLDT_SEL, SEL_KPL)); 436 } 437 438 static void 439 set_user_ldt_rv(struct vmspace *vmsp) 440 { 441 struct thread *td; 442 443 td = curthread; 444 if (vmsp != td->td_proc->p_vmspace) 445 return; 446 447 set_user_ldt(&td->td_proc->p_md); 448 } 449 450 struct proc_ldt * 451 user_ldt_alloc(struct proc *p, int force) 452 { 453 struct proc_ldt *pldt, *new_ldt; 454 struct mdproc *mdp; 455 struct soft_segment_descriptor sldt; 456 vm_offset_t sva; 457 vm_size_t sz; 458 459 mtx_assert(&dt_lock, MA_OWNED); 460 mdp = &p->p_md; 461 if (!force && mdp->md_ldt != NULL) 462 return (mdp->md_ldt); 463 mtx_unlock(&dt_lock); 464 new_ldt = malloc(sizeof(struct proc_ldt), M_SUBPROC, M_WAITOK); 465 sz = max_ldt_segment * sizeof(struct user_segment_descriptor); 466 sva = kmem_malloc(kernel_arena, sz, M_WAITOK | M_ZERO); 467 new_ldt->ldt_base = (caddr_t)sva; 468 pmap_pti_add_kva(sva, sva + sz, false); 469 new_ldt->ldt_refcnt = 1; 470 sldt.ssd_base = sva; 471 sldt.ssd_limit = sz - 1; 472 sldt.ssd_type = SDT_SYSLDT; 473 sldt.ssd_dpl = SEL_KPL; 474 sldt.ssd_p = 1; 475 sldt.ssd_long = 0; 476 sldt.ssd_def32 = 0; 477 sldt.ssd_gran = 0; 478 mtx_lock(&dt_lock); 479 pldt = mdp->md_ldt; 480 if (pldt != NULL && !force) { 481 pmap_pti_remove_kva(sva, sva + sz); 482 kmem_free(kernel_arena, sva, sz); 483 free(new_ldt, M_SUBPROC); 484 return (pldt); 485 } 486 487 if (pldt != NULL) { 488 bcopy(pldt->ldt_base, new_ldt->ldt_base, max_ldt_segment * 489 sizeof(struct user_segment_descriptor)); 490 user_ldt_derefl(pldt); 491 } 492 critical_enter(); 493 ssdtosyssd(&sldt, &p->p_md.md_ldt_sd); 494 atomic_thread_fence_rel(); 495 mdp->md_ldt = new_ldt; 496 critical_exit(); 497 smp_rendezvous(NULL, (void (*)(void *))set_user_ldt_rv, NULL, 498 p->p_vmspace); 499 500 return (mdp->md_ldt); 501 } 502 503 void 504 user_ldt_free(struct thread *td) 505 { 506 struct proc *p = td->td_proc; 507 struct mdproc *mdp = &p->p_md; 508 struct proc_ldt *pldt; 509 510 mtx_lock(&dt_lock); 511 if ((pldt = mdp->md_ldt) == NULL) { 512 mtx_unlock(&dt_lock); 513 return; 514 } 515 516 critical_enter(); 517 mdp->md_ldt = NULL; 518 atomic_thread_fence_rel(); 519 bzero(&mdp->md_ldt_sd, sizeof(mdp->md_ldt_sd)); 520 if (td == curthread) 521 lldt(GSEL(GNULL_SEL, SEL_KPL)); 522 critical_exit(); 523 user_ldt_deref(pldt); 524 } 525 526 static void 527 user_ldt_derefl(struct proc_ldt *pldt) 528 { 529 vm_offset_t sva; 530 vm_size_t sz; 531 532 if (--pldt->ldt_refcnt == 0) { 533 sva = (vm_offset_t)pldt->ldt_base; 534 sz = max_ldt_segment * sizeof(struct user_segment_descriptor); 535 pmap_pti_remove_kva(sva, sva + sz); 536 kmem_free(kernel_arena, sva, sz); 537 free(pldt, M_SUBPROC); 538 } 539 } 540 541 static void 542 user_ldt_deref(struct proc_ldt *pldt) 543 { 544 545 mtx_assert(&dt_lock, MA_OWNED); 546 user_ldt_derefl(pldt); 547 mtx_unlock(&dt_lock); 548 } 549 550 /* 551 * Note for the authors of compat layers (linux, etc): copyout() in 552 * the function below is not a problem since it presents data in 553 * arch-specific format (i.e. i386-specific in this case), not in 554 * the OS-specific one. 555 */ 556 int 557 amd64_get_ldt(struct thread *td, struct i386_ldt_args *uap) 558 { 559 struct proc_ldt *pldt; 560 struct user_segment_descriptor *lp; 561 uint64_t *data; 562 u_int i, num; 563 int error; 564 565 #ifdef DEBUG 566 printf("amd64_get_ldt: start=%u num=%u descs=%p\n", 567 uap->start, uap->num, (void *)uap->descs); 568 #endif 569 570 pldt = td->td_proc->p_md.md_ldt; 571 if (pldt == NULL || uap->start >= max_ldt_segment || uap->num == 0) { 572 td->td_retval[0] = 0; 573 return (0); 574 } 575 num = min(uap->num, max_ldt_segment - uap->start); 576 lp = &((struct user_segment_descriptor *)(pldt->ldt_base))[uap->start]; 577 data = malloc(num * sizeof(struct user_segment_descriptor), M_TEMP, 578 M_WAITOK); 579 mtx_lock(&dt_lock); 580 for (i = 0; i < num; i++) 581 data[i] = ((volatile uint64_t *)lp)[i]; 582 mtx_unlock(&dt_lock); 583 error = copyout(data, uap->descs, num * 584 sizeof(struct user_segment_descriptor)); 585 free(data, M_TEMP); 586 if (error == 0) 587 td->td_retval[0] = num; 588 return (error); 589 } 590 591 int 592 amd64_set_ldt(struct thread *td, struct i386_ldt_args *uap, 593 struct user_segment_descriptor *descs) 594 { 595 struct mdproc *mdp; 596 struct proc_ldt *pldt; 597 struct user_segment_descriptor *dp; 598 struct proc *p; 599 u_int largest_ld, i; 600 int error; 601 602 #ifdef DEBUG 603 printf("amd64_set_ldt: start=%u num=%u descs=%p\n", 604 uap->start, uap->num, (void *)uap->descs); 605 #endif 606 mdp = &td->td_proc->p_md; 607 error = 0; 608 609 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 610 p = td->td_proc; 611 if (descs == NULL) { 612 /* Free descriptors */ 613 if (uap->start == 0 && uap->num == 0) 614 uap->num = max_ldt_segment; 615 if (uap->num == 0) 616 return (EINVAL); 617 if ((pldt = mdp->md_ldt) == NULL || 618 uap->start >= max_ldt_segment) 619 return (0); 620 largest_ld = uap->start + uap->num; 621 if (largest_ld > max_ldt_segment) 622 largest_ld = max_ldt_segment; 623 if (largest_ld < uap->start) 624 return (EINVAL); 625 mtx_lock(&dt_lock); 626 for (i = uap->start; i < largest_ld; i++) 627 ((volatile uint64_t *)(pldt->ldt_base))[i] = 0; 628 mtx_unlock(&dt_lock); 629 return (0); 630 } 631 632 if (!(uap->start == LDT_AUTO_ALLOC && uap->num == 1)) { 633 /* verify range of descriptors to modify */ 634 largest_ld = uap->start + uap->num; 635 if (uap->start >= max_ldt_segment || 636 largest_ld > max_ldt_segment || 637 largest_ld < uap->start) 638 return (EINVAL); 639 } 640 641 /* Check descriptors for access violations */ 642 for (i = 0; i < uap->num; i++) { 643 dp = &descs[i]; 644 645 switch (dp->sd_type) { 646 case SDT_SYSNULL: /* system null */ 647 dp->sd_p = 0; 648 break; 649 case SDT_SYS286TSS: 650 case SDT_SYSLDT: 651 case SDT_SYS286BSY: 652 case SDT_SYS286CGT: 653 case SDT_SYSTASKGT: 654 case SDT_SYS286IGT: 655 case SDT_SYS286TGT: 656 case SDT_SYSNULL2: 657 case SDT_SYSTSS: 658 case SDT_SYSNULL3: 659 case SDT_SYSBSY: 660 case SDT_SYSCGT: 661 case SDT_SYSNULL4: 662 case SDT_SYSIGT: 663 case SDT_SYSTGT: 664 return (EACCES); 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 } 691 692 /* Only user (ring-3) descriptors may be present. */ 693 if ((dp->sd_p != 0) && (dp->sd_dpl != SEL_UPL)) 694 return (EACCES); 695 } 696 697 if (uap->start == LDT_AUTO_ALLOC && uap->num == 1) { 698 /* Allocate a free slot */ 699 mtx_lock(&dt_lock); 700 pldt = user_ldt_alloc(p, 0); 701 if (pldt == NULL) { 702 mtx_unlock(&dt_lock); 703 return (ENOMEM); 704 } 705 706 /* 707 * start scanning a bit up to leave room for NVidia and 708 * Wine, which still user the "Blat" method of allocation. 709 */ 710 i = 16; 711 dp = &((struct user_segment_descriptor *)(pldt->ldt_base))[i]; 712 for (; i < max_ldt_segment; ++i, ++dp) { 713 if (dp->sd_type == SDT_SYSNULL) 714 break; 715 } 716 if (i >= max_ldt_segment) { 717 mtx_unlock(&dt_lock); 718 return (ENOSPC); 719 } 720 uap->start = i; 721 error = amd64_set_ldt_data(td, i, 1, descs); 722 mtx_unlock(&dt_lock); 723 } else { 724 largest_ld = uap->start + uap->num; 725 if (largest_ld > max_ldt_segment) 726 return (EINVAL); 727 mtx_lock(&dt_lock); 728 if (user_ldt_alloc(p, 0) != NULL) { 729 error = amd64_set_ldt_data(td, uap->start, uap->num, 730 descs); 731 } 732 mtx_unlock(&dt_lock); 733 } 734 if (error == 0) 735 td->td_retval[0] = uap->start; 736 return (error); 737 } 738 739 int 740 amd64_set_ldt_data(struct thread *td, int start, int num, 741 struct user_segment_descriptor *descs) 742 { 743 struct mdproc *mdp; 744 struct proc_ldt *pldt; 745 volatile uint64_t *dst, *src; 746 int i; 747 748 mtx_assert(&dt_lock, MA_OWNED); 749 750 mdp = &td->td_proc->p_md; 751 pldt = mdp->md_ldt; 752 dst = (volatile uint64_t *)(pldt->ldt_base); 753 src = (volatile uint64_t *)descs; 754 for (i = 0; i < num; i++) 755 dst[start + i] = src[i]; 756 return (0); 757 } 758