Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/i386/i386/sys_machdep.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
1 /*- 2 * Copyright (c) 1990 The Regents of the University of California. 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 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * 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 REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * from: @(#)sys_machdep.c 5.5 (Berkeley) 1/19/91 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD: releng/8.1/sys/i386/i386/sys_machdep.c 199583 2009-11-20 15:27:52Z jhb $"); 34 35 #include "opt_kstack_pages.h" 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/lock.h> 40 #include <sys/malloc.h> 41 #include <sys/mutex.h> 42 #include <sys/priv.h> 43 #include <sys/proc.h> 44 #include <sys/smp.h> 45 #include <sys/sysproto.h> 46 47 #include <vm/vm.h> 48 #include <vm/pmap.h> 49 #include <vm/vm_map.h> 50 #include <vm/vm_extern.h> 51 52 #include <machine/cpu.h> 53 #include <machine/pcb.h> 54 #include <machine/pcb_ext.h> 55 #include <machine/proc.h> 56 #include <machine/sysarch.h> 57 58 #include <security/audit/audit.h> 59 60 #ifdef XEN 61 #include <machine/xen/xenfunc.h> 62 63 void i386_reset_ldt(struct proc_ldt *pldt); 64 65 void 66 i386_reset_ldt(struct proc_ldt *pldt) 67 { 68 xen_set_ldt((vm_offset_t)pldt->ldt_base, pldt->ldt_len); 69 } 70 #else 71 #define i386_reset_ldt(x) 72 #endif 73 74 #include <vm/vm_kern.h> /* for kernel_map */ 75 76 #define MAX_LD 8192 77 #define LD_PER_PAGE 512 78 #define NEW_MAX_LD(num) ((num + LD_PER_PAGE) & ~(LD_PER_PAGE-1)) 79 #define SIZE_FROM_LARGEST_LD(num) (NEW_MAX_LD(num) << 3) 80 #define NULL_LDT_BASE ((caddr_t)NULL) 81 82 #ifdef SMP 83 static void set_user_ldt_rv(struct vmspace *vmsp); 84 #endif 85 static int i386_set_ldt_data(struct thread *, int start, int num, 86 union descriptor *descs); 87 static int i386_ldt_grow(struct thread *td, int len); 88 89 #ifndef _SYS_SYSPROTO_H_ 90 struct sysarch_args { 91 int op; 92 char *parms; 93 }; 94 #endif 95 96 int 97 sysarch(td, uap) 98 struct thread *td; 99 register struct sysarch_args *uap; 100 { 101 int error; 102 union descriptor *lp; 103 union { 104 struct i386_ldt_args largs; 105 struct i386_ioperm_args iargs; 106 } kargs; 107 uint32_t base; 108 struct segment_descriptor sd, *sdp; 109 110 AUDIT_ARG_CMD(uap->op); 111 switch (uap->op) { 112 case I386_GET_IOPERM: 113 case I386_SET_IOPERM: 114 if ((error = copyin(uap->parms, &kargs.iargs, 115 sizeof(struct i386_ioperm_args))) != 0) 116 return (error); 117 break; 118 case I386_GET_LDT: 119 case I386_SET_LDT: 120 if ((error = copyin(uap->parms, &kargs.largs, 121 sizeof(struct i386_ldt_args))) != 0) 122 return (error); 123 if (kargs.largs.num > MAX_LD || kargs.largs.num <= 0) 124 return (EINVAL); 125 break; 126 default: 127 break; 128 } 129 130 switch(uap->op) { 131 case I386_GET_LDT: 132 error = i386_get_ldt(td, &kargs.largs); 133 break; 134 case I386_SET_LDT: 135 if (kargs.largs.descs != NULL) { 136 lp = (union descriptor *)kmem_alloc(kernel_map, 137 kargs.largs.num * sizeof(union descriptor)); 138 if (lp == NULL) { 139 error = ENOMEM; 140 break; 141 } 142 error = copyin(kargs.largs.descs, lp, 143 kargs.largs.num * sizeof(union descriptor)); 144 if (error == 0) 145 error = i386_set_ldt(td, &kargs.largs, lp); 146 kmem_free(kernel_map, (vm_offset_t)lp, 147 kargs.largs.num * sizeof(union descriptor)); 148 } else { 149 error = i386_set_ldt(td, &kargs.largs, NULL); 150 } 151 break; 152 case I386_GET_IOPERM: 153 error = i386_get_ioperm(td, &kargs.iargs); 154 if (error == 0) 155 error = copyout(&kargs.iargs, uap->parms, 156 sizeof(struct i386_ioperm_args)); 157 break; 158 case I386_SET_IOPERM: 159 error = i386_set_ioperm(td, &kargs.iargs); 160 break; 161 case I386_VM86: 162 error = vm86_sysarch(td, uap->parms); 163 break; 164 case I386_GET_FSBASE: 165 sdp = &td->td_pcb->pcb_fsd; 166 base = sdp->sd_hibase << 24 | sdp->sd_lobase; 167 error = copyout(&base, uap->parms, sizeof(base)); 168 break; 169 case I386_SET_FSBASE: 170 error = copyin(uap->parms, &base, sizeof(base)); 171 if (!error) { 172 /* 173 * Construct a descriptor and store it in the pcb for 174 * the next context switch. Also store it in the gdt 175 * so that the load of tf_fs into %fs will activate it 176 * at return to userland. 177 */ 178 sd.sd_lobase = base & 0xffffff; 179 sd.sd_hibase = (base >> 24) & 0xff; 180 #ifdef XEN 181 /* need to do nosegneg like Linux */ 182 sd.sd_lolimit = (HYPERVISOR_VIRT_START >> 12) & 0xffff; 183 #else 184 sd.sd_lolimit = 0xffff; /* 4GB limit, wraps around */ 185 #endif 186 sd.sd_hilimit = 0xf; 187 sd.sd_type = SDT_MEMRWA; 188 sd.sd_dpl = SEL_UPL; 189 sd.sd_p = 1; 190 sd.sd_xx = 0; 191 sd.sd_def32 = 1; 192 sd.sd_gran = 1; 193 critical_enter(); 194 td->td_pcb->pcb_fsd = sd; 195 #ifdef XEN 196 HYPERVISOR_update_descriptor(vtomach(&PCPU_GET(fsgs_gdt)[0]), 197 *(uint64_t *)&sd); 198 #else 199 PCPU_GET(fsgs_gdt)[0] = sd; 200 #endif 201 critical_exit(); 202 td->td_frame->tf_fs = GSEL(GUFS_SEL, SEL_UPL); 203 } 204 break; 205 case I386_GET_GSBASE: 206 sdp = &td->td_pcb->pcb_gsd; 207 base = sdp->sd_hibase << 24 | sdp->sd_lobase; 208 error = copyout(&base, uap->parms, sizeof(base)); 209 break; 210 case I386_SET_GSBASE: 211 error = copyin(uap->parms, &base, sizeof(base)); 212 if (!error) { 213 /* 214 * Construct a descriptor and store it in the pcb for 215 * the next context switch. Also store it in the gdt 216 * because we have to do a load_gs() right now. 217 */ 218 sd.sd_lobase = base & 0xffffff; 219 sd.sd_hibase = (base >> 24) & 0xff; 220 221 #ifdef XEN 222 /* need to do nosegneg like Linux */ 223 sd.sd_lolimit = (HYPERVISOR_VIRT_START >> 12) & 0xffff; 224 #else 225 sd.sd_lolimit = 0xffff; /* 4GB limit, wraps around */ 226 #endif 227 sd.sd_hilimit = 0xf; 228 sd.sd_type = SDT_MEMRWA; 229 sd.sd_dpl = SEL_UPL; 230 sd.sd_p = 1; 231 sd.sd_xx = 0; 232 sd.sd_def32 = 1; 233 sd.sd_gran = 1; 234 critical_enter(); 235 td->td_pcb->pcb_gsd = sd; 236 #ifdef XEN 237 HYPERVISOR_update_descriptor(vtomach(&PCPU_GET(fsgs_gdt)[1]), 238 *(uint64_t *)&sd); 239 #else 240 PCPU_GET(fsgs_gdt)[1] = sd; 241 #endif 242 critical_exit(); 243 load_gs(GSEL(GUGS_SEL, SEL_UPL)); 244 } 245 break; 246 default: 247 error = EINVAL; 248 break; 249 } 250 return (error); 251 } 252 253 int 254 i386_extend_pcb(struct thread *td) 255 { 256 int i, offset; 257 u_long *addr; 258 struct pcb_ext *ext; 259 struct soft_segment_descriptor ssd = { 260 0, /* segment base address (overwritten) */ 261 ctob(IOPAGES + 1) - 1, /* length */ 262 SDT_SYS386TSS, /* segment type */ 263 0, /* priority level */ 264 1, /* descriptor present */ 265 0, 0, 266 0, /* default 32 size */ 267 0 /* granularity */ 268 }; 269 270 ext = (struct pcb_ext *)kmem_alloc(kernel_map, ctob(IOPAGES+1)); 271 if (ext == 0) 272 return (ENOMEM); 273 bzero(ext, sizeof(struct pcb_ext)); 274 /* -16 is so we can convert a trapframe into vm86trapframe inplace */ 275 ext->ext_tss.tss_esp0 = td->td_kstack + ctob(KSTACK_PAGES) - 276 sizeof(struct pcb) - 16; 277 ext->ext_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL); 278 /* 279 * The last byte of the i/o map must be followed by an 0xff byte. 280 * We arbitrarily allocate 16 bytes here, to keep the starting 281 * address on a doubleword boundary. 282 */ 283 offset = PAGE_SIZE - 16; 284 ext->ext_tss.tss_ioopt = 285 (offset - ((unsigned)&ext->ext_tss - (unsigned)ext)) << 16; 286 ext->ext_iomap = (caddr_t)ext + offset; 287 ext->ext_vm86.vm86_intmap = (caddr_t)ext + offset - 32; 288 289 addr = (u_long *)ext->ext_vm86.vm86_intmap; 290 for (i = 0; i < (ctob(IOPAGES) + 32 + 16) / sizeof(u_long); i++) 291 *addr++ = ~0; 292 293 ssd.ssd_base = (unsigned)&ext->ext_tss; 294 ssd.ssd_limit -= ((unsigned)&ext->ext_tss - (unsigned)ext); 295 ssdtosd(&ssd, &ext->ext_tssd); 296 297 KASSERT(td == curthread, ("giving TSS to !curthread")); 298 KASSERT(td->td_pcb->pcb_ext == 0, ("already have a TSS!")); 299 300 /* Switch to the new TSS. */ 301 critical_enter(); 302 td->td_pcb->pcb_ext = ext; 303 PCPU_SET(private_tss, 1); 304 *PCPU_GET(tss_gdt) = ext->ext_tssd; 305 ltr(GSEL(GPROC0_SEL, SEL_KPL)); 306 critical_exit(); 307 308 return 0; 309 } 310 311 int 312 i386_set_ioperm(td, uap) 313 struct thread *td; 314 struct i386_ioperm_args *uap; 315 { 316 int i, error; 317 char *iomap; 318 319 if ((error = priv_check(td, PRIV_IO)) != 0) 320 return (error); 321 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 322 return (error); 323 /* 324 * XXX 325 * While this is restricted to root, we should probably figure out 326 * whether any other driver is using this i/o address, as so not to 327 * cause confusion. This probably requires a global 'usage registry'. 328 */ 329 330 if (td->td_pcb->pcb_ext == 0) 331 if ((error = i386_extend_pcb(td)) != 0) 332 return (error); 333 iomap = (char *)td->td_pcb->pcb_ext->ext_iomap; 334 335 if (uap->start + uap->length > IOPAGES * PAGE_SIZE * NBBY) 336 return (EINVAL); 337 338 for (i = uap->start; i < uap->start + uap->length; i++) { 339 if (uap->enable) 340 iomap[i >> 3] &= ~(1 << (i & 7)); 341 else 342 iomap[i >> 3] |= (1 << (i & 7)); 343 } 344 return (error); 345 } 346 347 int 348 i386_get_ioperm(td, uap) 349 struct thread *td; 350 struct i386_ioperm_args *uap; 351 { 352 int i, state; 353 char *iomap; 354 355 if (uap->start >= IOPAGES * PAGE_SIZE * NBBY) 356 return (EINVAL); 357 358 if (td->td_pcb->pcb_ext == 0) { 359 uap->length = 0; 360 goto done; 361 } 362 363 iomap = (char *)td->td_pcb->pcb_ext->ext_iomap; 364 365 i = uap->start; 366 state = (iomap[i >> 3] >> (i & 7)) & 1; 367 uap->enable = !state; 368 uap->length = 1; 369 370 for (i = uap->start + 1; i < IOPAGES * PAGE_SIZE * NBBY; i++) { 371 if (state != ((iomap[i >> 3] >> (i & 7)) & 1)) 372 break; 373 uap->length++; 374 } 375 376 done: 377 return (0); 378 } 379 380 /* 381 * Update the GDT entry pointing to the LDT to point to the LDT of the 382 * current process. Manage dt_lock holding/unholding autonomously. 383 */ 384 void 385 set_user_ldt(struct mdproc *mdp) 386 { 387 struct proc_ldt *pldt; 388 int dtlocked; 389 390 dtlocked = 0; 391 if (!mtx_owned(&dt_lock)) { 392 mtx_lock_spin(&dt_lock); 393 dtlocked = 1; 394 } 395 396 pldt = mdp->md_ldt; 397 #ifdef XEN 398 i386_reset_ldt(pldt); 399 PCPU_SET(currentldt, (int)pldt); 400 #else 401 #ifdef SMP 402 gdt[PCPU_GET(cpuid) * NGDT + GUSERLDT_SEL].sd = pldt->ldt_sd; 403 #else 404 gdt[GUSERLDT_SEL].sd = pldt->ldt_sd; 405 #endif 406 lldt(GSEL(GUSERLDT_SEL, SEL_KPL)); 407 PCPU_SET(currentldt, GSEL(GUSERLDT_SEL, SEL_KPL)); 408 #endif /* XEN */ 409 if (dtlocked) 410 mtx_unlock_spin(&dt_lock); 411 } 412 413 #ifdef SMP 414 static void 415 set_user_ldt_rv(struct vmspace *vmsp) 416 { 417 struct thread *td; 418 419 td = curthread; 420 if (vmsp != td->td_proc->p_vmspace) 421 return; 422 423 set_user_ldt(&td->td_proc->p_md); 424 } 425 #endif 426 427 #ifdef XEN 428 429 /* 430 * dt_lock must be held. Returns with dt_lock held. 431 */ 432 struct proc_ldt * 433 user_ldt_alloc(struct mdproc *mdp, int len) 434 { 435 struct proc_ldt *pldt, *new_ldt; 436 437 mtx_assert(&dt_lock, MA_OWNED); 438 mtx_unlock_spin(&dt_lock); 439 new_ldt = malloc(sizeof(struct proc_ldt), 440 M_SUBPROC, M_WAITOK); 441 442 new_ldt->ldt_len = len = NEW_MAX_LD(len); 443 new_ldt->ldt_base = (caddr_t)kmem_alloc(kernel_map, 444 round_page(len * sizeof(union descriptor))); 445 if (new_ldt->ldt_base == NULL) { 446 free(new_ldt, M_SUBPROC); 447 mtx_lock_spin(&dt_lock); 448 return (NULL); 449 } 450 new_ldt->ldt_refcnt = 1; 451 new_ldt->ldt_active = 0; 452 453 if ((pldt = mdp->md_ldt)) { 454 if (len > pldt->ldt_len) 455 len = pldt->ldt_len; 456 bcopy(pldt->ldt_base, new_ldt->ldt_base, 457 len * sizeof(union descriptor)); 458 } else { 459 bcopy(ldt, new_ldt->ldt_base, PAGE_SIZE); 460 } 461 pmap_map_readonly(kernel_pmap, (vm_offset_t)new_ldt->ldt_base, 462 new_ldt->ldt_len*sizeof(union descriptor)); 463 return (new_ldt); 464 } 465 #else 466 /* 467 * dt_lock must be held. Returns with dt_lock held. 468 */ 469 struct proc_ldt * 470 user_ldt_alloc(struct mdproc *mdp, int len) 471 { 472 struct proc_ldt *pldt, *new_ldt; 473 474 mtx_assert(&dt_lock, MA_OWNED); 475 mtx_unlock_spin(&dt_lock); 476 new_ldt = malloc(sizeof(struct proc_ldt), 477 M_SUBPROC, M_WAITOK); 478 479 new_ldt->ldt_len = len = NEW_MAX_LD(len); 480 new_ldt->ldt_base = (caddr_t)kmem_alloc(kernel_map, 481 len * sizeof(union descriptor)); 482 if (new_ldt->ldt_base == NULL) { 483 free(new_ldt, M_SUBPROC); 484 mtx_lock_spin(&dt_lock); 485 return (NULL); 486 } 487 new_ldt->ldt_refcnt = 1; 488 new_ldt->ldt_active = 0; 489 490 mtx_lock_spin(&dt_lock); 491 gdt_segs[GUSERLDT_SEL].ssd_base = (unsigned)new_ldt->ldt_base; 492 gdt_segs[GUSERLDT_SEL].ssd_limit = len * sizeof(union descriptor) - 1; 493 ssdtosd(&gdt_segs[GUSERLDT_SEL], &new_ldt->ldt_sd); 494 495 if ((pldt = mdp->md_ldt) != NULL) { 496 if (len > pldt->ldt_len) 497 len = pldt->ldt_len; 498 bcopy(pldt->ldt_base, new_ldt->ldt_base, 499 len * sizeof(union descriptor)); 500 } else 501 bcopy(ldt, new_ldt->ldt_base, sizeof(ldt)); 502 503 return (new_ldt); 504 } 505 #endif /* !XEN */ 506 507 /* 508 * Must be called with dt_lock held. Returns with dt_lock unheld. 509 */ 510 void 511 user_ldt_free(struct thread *td) 512 { 513 struct mdproc *mdp = &td->td_proc->p_md; 514 struct proc_ldt *pldt; 515 516 mtx_assert(&dt_lock, MA_OWNED); 517 if ((pldt = mdp->md_ldt) == NULL) { 518 mtx_unlock_spin(&dt_lock); 519 return; 520 } 521 522 if (td == PCPU_GET(curthread)) { 523 lldt(_default_ldt); 524 PCPU_SET(currentldt, _default_ldt); 525 } 526 527 mdp->md_ldt = NULL; 528 user_ldt_deref(pldt); 529 } 530 531 void 532 user_ldt_deref(struct proc_ldt *pldt) 533 { 534 535 mtx_assert(&dt_lock, MA_OWNED); 536 if (--pldt->ldt_refcnt == 0) { 537 mtx_unlock_spin(&dt_lock); 538 kmem_free(kernel_map, (vm_offset_t)pldt->ldt_base, 539 pldt->ldt_len * sizeof(union descriptor)); 540 free(pldt, M_SUBPROC); 541 } else 542 mtx_unlock_spin(&dt_lock); 543 } 544 545 /* 546 * Note for the authors of compat layers (linux, etc): copyout() in 547 * the function below is not a problem since it presents data in 548 * arch-specific format (i.e. i386-specific in this case), not in 549 * the OS-specific one. 550 */ 551 int 552 i386_get_ldt(td, uap) 553 struct thread *td; 554 struct i386_ldt_args *uap; 555 { 556 int error = 0; 557 struct proc_ldt *pldt; 558 int nldt, num; 559 union descriptor *lp; 560 561 #ifdef DEBUG 562 printf("i386_get_ldt: start=%d num=%d descs=%p\n", 563 uap->start, uap->num, (void *)uap->descs); 564 #endif 565 566 mtx_lock_spin(&dt_lock); 567 if ((pldt = td->td_proc->p_md.md_ldt) != NULL) { 568 nldt = pldt->ldt_len; 569 lp = &((union descriptor *)(pldt->ldt_base))[uap->start]; 570 mtx_unlock_spin(&dt_lock); 571 num = min(uap->num, nldt); 572 } else { 573 mtx_unlock_spin(&dt_lock); 574 nldt = sizeof(ldt)/sizeof(ldt[0]); 575 num = min(uap->num, nldt); 576 lp = &ldt[uap->start]; 577 } 578 579 if ((uap->start > (unsigned int)nldt) || 580 ((unsigned int)num > (unsigned int)nldt) || 581 ((unsigned int)(uap->start + num) > (unsigned int)nldt)) 582 return(EINVAL); 583 584 error = copyout(lp, uap->descs, num * sizeof(union descriptor)); 585 if (!error) 586 td->td_retval[0] = num; 587 588 return(error); 589 } 590 591 int 592 i386_set_ldt(td, uap, descs) 593 struct thread *td; 594 struct i386_ldt_args *uap; 595 union descriptor *descs; 596 { 597 int error = 0, i; 598 int largest_ld; 599 struct mdproc *mdp = &td->td_proc->p_md; 600 struct proc_ldt *pldt; 601 union descriptor *dp; 602 603 #ifdef DEBUG 604 printf("i386_set_ldt: start=%d num=%d descs=%p\n", 605 uap->start, uap->num, (void *)uap->descs); 606 #endif 607 608 if (descs == NULL) { 609 /* Free descriptors */ 610 if (uap->start == 0 && uap->num == 0) { 611 /* 612 * Treat this as a special case, so userland needn't 613 * know magic number NLDT. 614 */ 615 uap->start = NLDT; 616 uap->num = MAX_LD - NLDT; 617 } 618 if (uap->num <= 0) 619 return (EINVAL); 620 mtx_lock_spin(&dt_lock); 621 if ((pldt = mdp->md_ldt) == NULL || 622 uap->start >= pldt->ldt_len) { 623 mtx_unlock_spin(&dt_lock); 624 return (0); 625 } 626 largest_ld = uap->start + uap->num; 627 if (largest_ld > pldt->ldt_len) 628 largest_ld = pldt->ldt_len; 629 i = largest_ld - uap->start; 630 bzero(&((union descriptor *)(pldt->ldt_base))[uap->start], 631 sizeof(union descriptor) * i); 632 mtx_unlock_spin(&dt_lock); 633 return (0); 634 } 635 636 if (!(uap->start == LDT_AUTO_ALLOC && uap->num == 1)) { 637 /* verify range of descriptors to modify */ 638 largest_ld = uap->start + uap->num; 639 if (uap->start >= MAX_LD || 640 uap->num < 0 || largest_ld > MAX_LD) { 641 return (EINVAL); 642 } 643 } 644 645 /* Check descriptors for access violations */ 646 for (i = 0; i < uap->num; i++) { 647 dp = &descs[i]; 648 649 switch (dp->sd.sd_type) { 650 case SDT_SYSNULL: /* system null */ 651 dp->sd.sd_p = 0; 652 break; 653 case SDT_SYS286TSS: /* system 286 TSS available */ 654 case SDT_SYSLDT: /* system local descriptor table */ 655 case SDT_SYS286BSY: /* system 286 TSS busy */ 656 case SDT_SYSTASKGT: /* system task gate */ 657 case SDT_SYS286IGT: /* system 286 interrupt gate */ 658 case SDT_SYS286TGT: /* system 286 trap gate */ 659 case SDT_SYSNULL2: /* undefined by Intel */ 660 case SDT_SYS386TSS: /* system 386 TSS available */ 661 case SDT_SYSNULL3: /* undefined by Intel */ 662 case SDT_SYS386BSY: /* system 386 TSS busy */ 663 case SDT_SYSNULL4: /* undefined by Intel */ 664 case SDT_SYS386IGT: /* system 386 interrupt gate */ 665 case SDT_SYS386TGT: /* system 386 trap gate */ 666 case SDT_SYS286CGT: /* system 286 call gate */ 667 case SDT_SYS386CGT: /* system 386 call gate */ 668 /* I can't think of any reason to allow a user proc 669 * to create a segment of these types. They are 670 * for OS use only. 671 */ 672 return (EACCES); 673 /*NOTREACHED*/ 674 675 /* memory segment types */ 676 case SDT_MEMEC: /* memory execute only conforming */ 677 case SDT_MEMEAC: /* memory execute only accessed conforming */ 678 case SDT_MEMERC: /* memory execute read conforming */ 679 case SDT_MEMERAC: /* memory execute read accessed conforming */ 680 /* Must be "present" if executable and conforming. */ 681 if (dp->sd.sd_p == 0) 682 return (EACCES); 683 break; 684 case SDT_MEMRO: /* memory read only */ 685 case SDT_MEMROA: /* memory read only accessed */ 686 case SDT_MEMRW: /* memory read write */ 687 case SDT_MEMRWA: /* memory read write accessed */ 688 case SDT_MEMROD: /* memory read only expand dwn limit */ 689 case SDT_MEMRODA: /* memory read only expand dwn lim accessed */ 690 case SDT_MEMRWD: /* memory read write expand dwn limit */ 691 case SDT_MEMRWDA: /* memory read write expand dwn lim acessed */ 692 case SDT_MEME: /* memory execute only */ 693 case SDT_MEMEA: /* memory execute only accessed */ 694 case SDT_MEMER: /* memory execute read */ 695 case SDT_MEMERA: /* memory execute read accessed */ 696 break; 697 default: 698 return(EINVAL); 699 /*NOTREACHED*/ 700 } 701 702 /* Only user (ring-3) descriptors may be present. */ 703 if ((dp->sd.sd_p != 0) && (dp->sd.sd_dpl != SEL_UPL)) 704 return (EACCES); 705 } 706 707 if (uap->start == LDT_AUTO_ALLOC && uap->num == 1) { 708 /* Allocate a free slot */ 709 mtx_lock_spin(&dt_lock); 710 if ((pldt = mdp->md_ldt) == NULL) { 711 if ((error = i386_ldt_grow(td, NLDT + 1))) { 712 mtx_unlock_spin(&dt_lock); 713 return (error); 714 } 715 pldt = mdp->md_ldt; 716 } 717 again: 718 /* 719 * start scanning a bit up to leave room for NVidia and 720 * Wine, which still user the "Blat" method of allocation. 721 */ 722 dp = &((union descriptor *)(pldt->ldt_base))[NLDT]; 723 for (i = NLDT; i < pldt->ldt_len; ++i) { 724 if (dp->sd.sd_type == SDT_SYSNULL) 725 break; 726 dp++; 727 } 728 if (i >= pldt->ldt_len) { 729 if ((error = i386_ldt_grow(td, pldt->ldt_len+1))) { 730 mtx_unlock_spin(&dt_lock); 731 return (error); 732 } 733 goto again; 734 } 735 uap->start = i; 736 error = i386_set_ldt_data(td, i, 1, descs); 737 mtx_unlock_spin(&dt_lock); 738 } else { 739 largest_ld = uap->start + uap->num; 740 mtx_lock_spin(&dt_lock); 741 if (!(error = i386_ldt_grow(td, largest_ld))) { 742 error = i386_set_ldt_data(td, uap->start, uap->num, 743 descs); 744 } 745 mtx_unlock_spin(&dt_lock); 746 } 747 if (error == 0) 748 td->td_retval[0] = uap->start; 749 return (error); 750 } 751 #ifdef XEN 752 static int 753 i386_set_ldt_data(struct thread *td, int start, int num, 754 union descriptor *descs) 755 { 756 struct mdproc *mdp = &td->td_proc->p_md; 757 struct proc_ldt *pldt = mdp->md_ldt; 758 759 mtx_assert(&dt_lock, MA_OWNED); 760 761 /* Fill in range */ 762 bcopy(descs, 763 &((union descriptor *)(pldt->ldt_base))[start], 764 num * sizeof(union descriptor)); 765 return (0); 766 } 767 #else 768 static int 769 i386_set_ldt_data(struct thread *td, int start, int num, 770 union descriptor *descs) 771 { 772 struct mdproc *mdp = &td->td_proc->p_md; 773 struct proc_ldt *pldt = mdp->md_ldt; 774 775 mtx_assert(&dt_lock, MA_OWNED); 776 777 /* Fill in range */ 778 bcopy(descs, 779 &((union descriptor *)(pldt->ldt_base))[start], 780 num * sizeof(union descriptor)); 781 return (0); 782 } 783 #endif /* !XEN */ 784 785 static int 786 i386_ldt_grow(struct thread *td, int len) 787 { 788 struct mdproc *mdp = &td->td_proc->p_md; 789 struct proc_ldt *new_ldt, *pldt; 790 caddr_t old_ldt_base = NULL_LDT_BASE; 791 int old_ldt_len = 0; 792 793 mtx_assert(&dt_lock, MA_OWNED); 794 795 if (len > MAX_LD) 796 return (ENOMEM); 797 if (len < NLDT + 1) 798 len = NLDT + 1; 799 800 /* Allocate a user ldt. */ 801 if ((pldt = mdp->md_ldt) == NULL || len > pldt->ldt_len) { 802 new_ldt = user_ldt_alloc(mdp, len); 803 if (new_ldt == NULL) 804 return (ENOMEM); 805 pldt = mdp->md_ldt; 806 807 if (pldt != NULL) { 808 if (new_ldt->ldt_len <= pldt->ldt_len) { 809 /* 810 * We just lost the race for allocation, so 811 * free the new object and return. 812 */ 813 mtx_unlock_spin(&dt_lock); 814 kmem_free(kernel_map, 815 (vm_offset_t)new_ldt->ldt_base, 816 new_ldt->ldt_len * sizeof(union descriptor)); 817 free(new_ldt, M_SUBPROC); 818 mtx_lock_spin(&dt_lock); 819 return (0); 820 } 821 822 /* 823 * We have to substitute the current LDT entry for 824 * curproc with the new one since its size grew. 825 */ 826 old_ldt_base = pldt->ldt_base; 827 old_ldt_len = pldt->ldt_len; 828 pldt->ldt_sd = new_ldt->ldt_sd; 829 pldt->ldt_base = new_ldt->ldt_base; 830 pldt->ldt_len = new_ldt->ldt_len; 831 } else 832 mdp->md_ldt = pldt = new_ldt; 833 #ifdef SMP 834 /* 835 * Signal other cpus to reload ldt. We need to unlock dt_lock 836 * here because other CPU will contest on it since their 837 * curthreads won't hold the lock and will block when trying 838 * to acquire it. 839 */ 840 mtx_unlock_spin(&dt_lock); 841 smp_rendezvous(NULL, (void (*)(void *))set_user_ldt_rv, 842 NULL, td->td_proc->p_vmspace); 843 #else 844 set_user_ldt(&td->td_proc->p_md); 845 mtx_unlock_spin(&dt_lock); 846 #endif 847 if (old_ldt_base != NULL_LDT_BASE) { 848 kmem_free(kernel_map, (vm_offset_t)old_ldt_base, 849 old_ldt_len * sizeof(union descriptor)); 850 free(new_ldt, M_SUBPROC); 851 } 852 mtx_lock_spin(&dt_lock); 853 } 854 return (0); 855 }
Cache object: 8eaca37d7f06647cc293d2e0eef909b7
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]