linux32_machdep.c revision 293482
1250551Sjeff/*- 2250551Sjeff * Copyright (c) 2004 Tim J. Robbins 3250551Sjeff * Copyright (c) 2002 Doug Rabson 4250551Sjeff * Copyright (c) 2000 Marcel Moolenaar 5250551Sjeff * All rights reserved. 6250551Sjeff * 7250551Sjeff * Redistribution and use in source and binary forms, with or without 8250551Sjeff * modification, are permitted provided that the following conditions 9250551Sjeff * are met: 10250551Sjeff * 1. Redistributions of source code must retain the above copyright 11250551Sjeff * notice, this list of conditions and the following disclaimer 12250551Sjeff * in this position and unchanged. 13250551Sjeff * 2. Redistributions in binary form must reproduce the above copyright 14250551Sjeff * notice, this list of conditions and the following disclaimer in the 15250551Sjeff * documentation and/or other materials provided with the distribution. 16250551Sjeff * 3. The name of the author may not be used to endorse or promote products 17250551Sjeff * derived from this software without specific prior written permission. 18250551Sjeff * 19250551Sjeff * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 20250551Sjeff * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 21250551Sjeff * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 22250551Sjeff * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 23250551Sjeff * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 24250551Sjeff * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 25250551Sjeff * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 26250551Sjeff * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 27250551Sjeff * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 28250551Sjeff * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 29250551Sjeff */ 30250551Sjeff 31250551Sjeff#include <sys/cdefs.h> 32250551Sjeff__FBSDID("$FreeBSD: stable/10/sys/amd64/linux32/linux32_machdep.c 293482 2016-01-09 14:40:38Z dchagin $"); 33250551Sjeff 34250551Sjeff#include <sys/param.h> 35250551Sjeff#include <sys/kernel.h> 36250551Sjeff#include <sys/systm.h> 37250551Sjeff#include <sys/capsicum.h> 38250551Sjeff#include <sys/file.h> 39250551Sjeff#include <sys/fcntl.h> 40250551Sjeff#include <sys/clock.h> 41250578Sjeff#include <sys/imgact.h> 42250578Sjeff#include <sys/limits.h> 43250578Sjeff#include <sys/lock.h> 44250578Sjeff#include <sys/malloc.h> 45250578Sjeff#include <sys/mman.h> 46250551Sjeff#include <sys/mutex.h> 47250551Sjeff#include <sys/priv.h> 48250551Sjeff#include <sys/proc.h> 49250551Sjeff#include <sys/resource.h> 50250551Sjeff#include <sys/resourcevar.h> 51250551Sjeff#include <sys/sched.h> 52250551Sjeff#include <sys/syscallsubr.h> 53250551Sjeff#include <sys/sysproto.h> 54250551Sjeff#include <sys/unistd.h> 55250551Sjeff#include <sys/wait.h> 56250551Sjeff 57250551Sjeff#include <machine/frame.h> 58250551Sjeff#include <machine/pcb.h> 59250551Sjeff#include <machine/psl.h> 60250551Sjeff#include <machine/segments.h> 61250551Sjeff#include <machine/specialreg.h> 62250551Sjeff 63250551Sjeff#include <vm/vm.h> 64250551Sjeff#include <vm/pmap.h> 65250551Sjeff#include <vm/vm_map.h> 66250551Sjeff 67250551Sjeff#include <compat/freebsd32/freebsd32_util.h> 68250551Sjeff#include <amd64/linux32/linux.h> 69250551Sjeff#include <amd64/linux32/linux32_proto.h> 70250551Sjeff#include <compat/linux/linux_ipc.h> 71250551Sjeff#include <compat/linux/linux_misc.h> 72250551Sjeff#include <compat/linux/linux_signal.h> 73250551Sjeff#include <compat/linux/linux_util.h> 74250551Sjeff#include <compat/linux/linux_emul.h> 75250551Sjeff 76250551Sjeffstruct l_old_select_argv { 77250551Sjeff l_int nfds; 78250551Sjeff l_uintptr_t readfds; 79250551Sjeff l_uintptr_t writefds; 80250551Sjeff l_uintptr_t exceptfds; 81250551Sjeff l_uintptr_t timeout; 82250551Sjeff} __packed; 83250551Sjeff 84250551Sjeffint 85250551Sjefflinux_to_bsd_sigaltstack(int lsa) 86260266Sdim{ 87250551Sjeff int bsa = 0; 88250551Sjeff 89250551Sjeff if (lsa & LINUX_SS_DISABLE) 90250551Sjeff bsa |= SS_DISABLE; 91250551Sjeff if (lsa & LINUX_SS_ONSTACK) 92250551Sjeff bsa |= SS_ONSTACK; 93260266Sdim return (bsa); 94250551Sjeff} 95250551Sjeff 96250551Sjeffstatic int linux_mmap_common(struct thread *td, l_uintptr_t addr, 97250551Sjeff l_size_t len, l_int prot, l_int flags, l_int fd, 98250551Sjeff l_loff_t pos); 99250551Sjeff 100250551Sjeffint 101250551Sjeffbsd_to_linux_sigaltstack(int bsa) 102250551Sjeff{ 103250551Sjeff int lsa = 0; 104250551Sjeff 105250551Sjeff if (bsa & SS_DISABLE) 106250551Sjeff lsa |= LINUX_SS_DISABLE; 107250551Sjeff if (bsa & SS_ONSTACK) 108250551Sjeff lsa |= LINUX_SS_ONSTACK; 109250551Sjeff return (lsa); 110250551Sjeff} 111250551Sjeff 112250551Sjeffstatic void 113250551Sjeffbsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru) 114250551Sjeff{ 115250551Sjeff 116250551Sjeff lru->ru_utime.tv_sec = ru->ru_utime.tv_sec; 117250551Sjeff lru->ru_utime.tv_usec = ru->ru_utime.tv_usec; 118250551Sjeff lru->ru_stime.tv_sec = ru->ru_stime.tv_sec; 119250551Sjeff lru->ru_stime.tv_usec = ru->ru_stime.tv_usec; 120250551Sjeff lru->ru_maxrss = ru->ru_maxrss; 121250551Sjeff lru->ru_ixrss = ru->ru_ixrss; 122250551Sjeff lru->ru_idrss = ru->ru_idrss; 123250551Sjeff lru->ru_isrss = ru->ru_isrss; 124 lru->ru_minflt = ru->ru_minflt; 125 lru->ru_majflt = ru->ru_majflt; 126 lru->ru_nswap = ru->ru_nswap; 127 lru->ru_inblock = ru->ru_inblock; 128 lru->ru_oublock = ru->ru_oublock; 129 lru->ru_msgsnd = ru->ru_msgsnd; 130 lru->ru_msgrcv = ru->ru_msgrcv; 131 lru->ru_nsignals = ru->ru_nsignals; 132 lru->ru_nvcsw = ru->ru_nvcsw; 133 lru->ru_nivcsw = ru->ru_nivcsw; 134} 135 136int 137linux_execve(struct thread *td, struct linux_execve_args *args) 138{ 139 struct image_args eargs; 140 struct vmspace *oldvmspace; 141 char *path; 142 int error; 143 144 LCONVPATHEXIST(td, args->path, &path); 145 146#ifdef DEBUG 147 if (ldebug(execve)) 148 printf(ARGS(execve, "%s"), path); 149#endif 150 151 error = pre_execve(td, &oldvmspace); 152 if (error != 0) { 153 free(path, M_TEMP); 154 return (error); 155 } 156 error = freebsd32_exec_copyin_args(&eargs, path, UIO_SYSSPACE, 157 args->argp, args->envp); 158 free(path, M_TEMP); 159 if (error == 0) 160 error = kern_execve(td, &eargs, NULL); 161 if (error == 0) { 162 /* Linux process can execute FreeBSD one, do not attempt 163 * to create emuldata for such process using 164 * linux_proc_init, this leads to a panic on KASSERT 165 * because such process has p->p_emuldata == NULL. 166 */ 167 if (SV_PROC_ABI(td->td_proc) == SV_ABI_LINUX) 168 error = linux_proc_init(td, 0, 0); 169 } 170 post_execve(td, error, oldvmspace); 171 return (error); 172} 173 174CTASSERT(sizeof(struct l_iovec32) == 8); 175 176static int 177linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop) 178{ 179 struct l_iovec32 iov32; 180 struct iovec *iov; 181 struct uio *uio; 182 uint32_t iovlen; 183 int error, i; 184 185 *uiop = NULL; 186 if (iovcnt > UIO_MAXIOV) 187 return (EINVAL); 188 iovlen = iovcnt * sizeof(struct iovec); 189 uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK); 190 iov = (struct iovec *)(uio + 1); 191 for (i = 0; i < iovcnt; i++) { 192 error = copyin(&iovp[i], &iov32, sizeof(struct l_iovec32)); 193 if (error) { 194 free(uio, M_IOV); 195 return (error); 196 } 197 iov[i].iov_base = PTRIN(iov32.iov_base); 198 iov[i].iov_len = iov32.iov_len; 199 } 200 uio->uio_iov = iov; 201 uio->uio_iovcnt = iovcnt; 202 uio->uio_segflg = UIO_USERSPACE; 203 uio->uio_offset = -1; 204 uio->uio_resid = 0; 205 for (i = 0; i < iovcnt; i++) { 206 if (iov->iov_len > INT_MAX - uio->uio_resid) { 207 free(uio, M_IOV); 208 return (EINVAL); 209 } 210 uio->uio_resid += iov->iov_len; 211 iov++; 212 } 213 *uiop = uio; 214 return (0); 215} 216 217int 218linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp, 219 int error) 220{ 221 struct l_iovec32 iov32; 222 struct iovec *iov; 223 uint32_t iovlen; 224 int i; 225 226 *iovp = NULL; 227 if (iovcnt > UIO_MAXIOV) 228 return (error); 229 iovlen = iovcnt * sizeof(struct iovec); 230 iov = malloc(iovlen, M_IOV, M_WAITOK); 231 for (i = 0; i < iovcnt; i++) { 232 error = copyin(&iovp32[i], &iov32, sizeof(struct l_iovec32)); 233 if (error) { 234 free(iov, M_IOV); 235 return (error); 236 } 237 iov[i].iov_base = PTRIN(iov32.iov_base); 238 iov[i].iov_len = iov32.iov_len; 239 } 240 *iovp = iov; 241 return(0); 242 243} 244 245int 246linux_readv(struct thread *td, struct linux_readv_args *uap) 247{ 248 struct uio *auio; 249 int error; 250 251 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio); 252 if (error) 253 return (error); 254 error = kern_readv(td, uap->fd, auio); 255 free(auio, M_IOV); 256 return (error); 257} 258 259int 260linux_writev(struct thread *td, struct linux_writev_args *uap) 261{ 262 struct uio *auio; 263 int error; 264 265 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio); 266 if (error) 267 return (error); 268 error = kern_writev(td, uap->fd, auio); 269 free(auio, M_IOV); 270 return (error); 271} 272 273struct l_ipc_kludge { 274 l_uintptr_t msgp; 275 l_long msgtyp; 276} __packed; 277 278int 279linux_ipc(struct thread *td, struct linux_ipc_args *args) 280{ 281 282 switch (args->what & 0xFFFF) { 283 case LINUX_SEMOP: { 284 struct linux_semop_args a; 285 286 a.semid = args->arg1; 287 a.tsops = args->ptr; 288 a.nsops = args->arg2; 289 return (linux_semop(td, &a)); 290 } 291 case LINUX_SEMGET: { 292 struct linux_semget_args a; 293 294 a.key = args->arg1; 295 a.nsems = args->arg2; 296 a.semflg = args->arg3; 297 return (linux_semget(td, &a)); 298 } 299 case LINUX_SEMCTL: { 300 struct linux_semctl_args a; 301 int error; 302 303 a.semid = args->arg1; 304 a.semnum = args->arg2; 305 a.cmd = args->arg3; 306 error = copyin(args->ptr, &a.arg, sizeof(a.arg)); 307 if (error) 308 return (error); 309 return (linux_semctl(td, &a)); 310 } 311 case LINUX_MSGSND: { 312 struct linux_msgsnd_args a; 313 314 a.msqid = args->arg1; 315 a.msgp = args->ptr; 316 a.msgsz = args->arg2; 317 a.msgflg = args->arg3; 318 return (linux_msgsnd(td, &a)); 319 } 320 case LINUX_MSGRCV: { 321 struct linux_msgrcv_args a; 322 323 a.msqid = args->arg1; 324 a.msgsz = args->arg2; 325 a.msgflg = args->arg3; 326 if ((args->what >> 16) == 0) { 327 struct l_ipc_kludge tmp; 328 int error; 329 330 if (args->ptr == 0) 331 return (EINVAL); 332 error = copyin(args->ptr, &tmp, sizeof(tmp)); 333 if (error) 334 return (error); 335 a.msgp = PTRIN(tmp.msgp); 336 a.msgtyp = tmp.msgtyp; 337 } else { 338 a.msgp = args->ptr; 339 a.msgtyp = args->arg5; 340 } 341 return (linux_msgrcv(td, &a)); 342 } 343 case LINUX_MSGGET: { 344 struct linux_msgget_args a; 345 346 a.key = args->arg1; 347 a.msgflg = args->arg2; 348 return (linux_msgget(td, &a)); 349 } 350 case LINUX_MSGCTL: { 351 struct linux_msgctl_args a; 352 353 a.msqid = args->arg1; 354 a.cmd = args->arg2; 355 a.buf = args->ptr; 356 return (linux_msgctl(td, &a)); 357 } 358 case LINUX_SHMAT: { 359 struct linux_shmat_args a; 360 361 a.shmid = args->arg1; 362 a.shmaddr = args->ptr; 363 a.shmflg = args->arg2; 364 a.raddr = PTRIN((l_uint)args->arg3); 365 return (linux_shmat(td, &a)); 366 } 367 case LINUX_SHMDT: { 368 struct linux_shmdt_args a; 369 370 a.shmaddr = args->ptr; 371 return (linux_shmdt(td, &a)); 372 } 373 case LINUX_SHMGET: { 374 struct linux_shmget_args a; 375 376 a.key = args->arg1; 377 a.size = args->arg2; 378 a.shmflg = args->arg3; 379 return (linux_shmget(td, &a)); 380 } 381 case LINUX_SHMCTL: { 382 struct linux_shmctl_args a; 383 384 a.shmid = args->arg1; 385 a.cmd = args->arg2; 386 a.buf = args->ptr; 387 return (linux_shmctl(td, &a)); 388 } 389 default: 390 break; 391 } 392 393 return (EINVAL); 394} 395 396int 397linux_old_select(struct thread *td, struct linux_old_select_args *args) 398{ 399 struct l_old_select_argv linux_args; 400 struct linux_select_args newsel; 401 int error; 402 403#ifdef DEBUG 404 if (ldebug(old_select)) 405 printf(ARGS(old_select, "%p"), args->ptr); 406#endif 407 408 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 409 if (error) 410 return (error); 411 412 newsel.nfds = linux_args.nfds; 413 newsel.readfds = PTRIN(linux_args.readfds); 414 newsel.writefds = PTRIN(linux_args.writefds); 415 newsel.exceptfds = PTRIN(linux_args.exceptfds); 416 newsel.timeout = PTRIN(linux_args.timeout); 417 return (linux_select(td, &newsel)); 418} 419 420int 421linux_set_cloned_tls(struct thread *td, void *desc) 422{ 423 struct user_segment_descriptor sd; 424 struct l_user_desc info; 425 struct pcb *pcb; 426 int error; 427 int a[2]; 428 429 error = copyin(desc, &info, sizeof(struct l_user_desc)); 430 if (error) { 431 printf(LMSG("copyin failed!")); 432 } else { 433 /* We might copy out the entry_number as GUGS32_SEL. */ 434 info.entry_number = GUGS32_SEL; 435 error = copyout(&info, desc, sizeof(struct l_user_desc)); 436 if (error) 437 printf(LMSG("copyout failed!")); 438 439 a[0] = LINUX_LDT_entry_a(&info); 440 a[1] = LINUX_LDT_entry_b(&info); 441 442 memcpy(&sd, &a, sizeof(a)); 443#ifdef DEBUG 444 if (ldebug(clone)) 445 printf("Segment created in clone with " 446 "CLONE_SETTLS: lobase: %x, hibase: %x, " 447 "lolimit: %x, hilimit: %x, type: %i, " 448 "dpl: %i, p: %i, xx: %i, long: %i, " 449 "def32: %i, gran: %i\n", sd.sd_lobase, 450 sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit, 451 sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx, 452 sd.sd_long, sd.sd_def32, sd.sd_gran); 453#endif 454 pcb = td->td_pcb; 455 pcb->pcb_gsbase = (register_t)info.base_addr; 456 td->td_frame->tf_gs = GSEL(GUGS32_SEL, SEL_UPL); 457 set_pcb_flags(pcb, PCB_32BIT); 458 } 459 460 return (error); 461} 462 463int 464linux_set_upcall_kse(struct thread *td, register_t stack) 465{ 466 467 td->td_frame->tf_rsp = stack; 468 469 return (0); 470} 471 472#define STACK_SIZE (2 * 1024 * 1024) 473#define GUARD_SIZE (4 * PAGE_SIZE) 474 475int 476linux_mmap2(struct thread *td, struct linux_mmap2_args *args) 477{ 478 479#ifdef DEBUG 480 if (ldebug(mmap2)) 481 printf(ARGS(mmap2, "0x%08x, %d, %d, 0x%08x, %d, %d"), 482 args->addr, args->len, args->prot, 483 args->flags, args->fd, args->pgoff); 484#endif 485 486 return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot, 487 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff * 488 PAGE_SIZE)); 489} 490 491int 492linux_mmap(struct thread *td, struct linux_mmap_args *args) 493{ 494 int error; 495 struct l_mmap_argv linux_args; 496 497 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 498 if (error) 499 return (error); 500 501#ifdef DEBUG 502 if (ldebug(mmap)) 503 printf(ARGS(mmap, "0x%08x, %d, %d, 0x%08x, %d, %d"), 504 linux_args.addr, linux_args.len, linux_args.prot, 505 linux_args.flags, linux_args.fd, linux_args.pgoff); 506#endif 507 508 return (linux_mmap_common(td, linux_args.addr, linux_args.len, 509 linux_args.prot, linux_args.flags, linux_args.fd, 510 (uint32_t)linux_args.pgoff)); 511} 512 513static int 514linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot, 515 l_int flags, l_int fd, l_loff_t pos) 516{ 517 struct proc *p = td->td_proc; 518 struct mmap_args /* { 519 caddr_t addr; 520 size_t len; 521 int prot; 522 int flags; 523 int fd; 524 long pad; 525 off_t pos; 526 } */ bsd_args; 527 int error; 528 struct file *fp; 529 cap_rights_t rights; 530 531 error = 0; 532 bsd_args.flags = 0; 533 fp = NULL; 534 535 /* 536 * Linux mmap(2): 537 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE 538 */ 539 if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) 540 return (EINVAL); 541 542 if (flags & LINUX_MAP_SHARED) 543 bsd_args.flags |= MAP_SHARED; 544 if (flags & LINUX_MAP_PRIVATE) 545 bsd_args.flags |= MAP_PRIVATE; 546 if (flags & LINUX_MAP_FIXED) 547 bsd_args.flags |= MAP_FIXED; 548 if (flags & LINUX_MAP_ANON) { 549 /* Enforce pos to be on page boundary, then ignore. */ 550 if ((pos & PAGE_MASK) != 0) 551 return (EINVAL); 552 pos = 0; 553 bsd_args.flags |= MAP_ANON; 554 } else 555 bsd_args.flags |= MAP_NOSYNC; 556 if (flags & LINUX_MAP_GROWSDOWN) 557 bsd_args.flags |= MAP_STACK; 558 559 /* 560 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC 561 * on Linux/i386. We do this to ensure maximum compatibility. 562 * Linux/ia64 does the same in i386 emulation mode. 563 */ 564 bsd_args.prot = prot; 565 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 566 bsd_args.prot |= PROT_READ | PROT_EXEC; 567 568 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ 569 bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd; 570 if (bsd_args.fd != -1) { 571 /* 572 * Linux follows Solaris mmap(2) description: 573 * The file descriptor fildes is opened with 574 * read permission, regardless of the 575 * protection options specified. 576 */ 577 578 error = fget(td, bsd_args.fd, 579 cap_rights_init(&rights, CAP_MMAP), &fp); 580 if (error != 0) 581 return (error); 582 if (fp->f_type != DTYPE_VNODE) { 583 fdrop(fp, td); 584 return (EINVAL); 585 } 586 587 /* Linux mmap() just fails for O_WRONLY files */ 588 if (!(fp->f_flag & FREAD)) { 589 fdrop(fp, td); 590 return (EACCES); 591 } 592 593 fdrop(fp, td); 594 } 595 596 if (flags & LINUX_MAP_GROWSDOWN) { 597 /* 598 * The Linux MAP_GROWSDOWN option does not limit auto 599 * growth of the region. Linux mmap with this option 600 * takes as addr the inital BOS, and as len, the initial 601 * region size. It can then grow down from addr without 602 * limit. However, Linux threads has an implicit internal 603 * limit to stack size of STACK_SIZE. Its just not 604 * enforced explicitly in Linux. But, here we impose 605 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 606 * region, since we can do this with our mmap. 607 * 608 * Our mmap with MAP_STACK takes addr as the maximum 609 * downsize limit on BOS, and as len the max size of 610 * the region. It then maps the top SGROWSIZ bytes, 611 * and auto grows the region down, up to the limit 612 * in addr. 613 * 614 * If we don't use the MAP_STACK option, the effect 615 * of this code is to allocate a stack region of a 616 * fixed size of (STACK_SIZE - GUARD_SIZE). 617 */ 618 619 if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) { 620 /* 621 * Some Linux apps will attempt to mmap 622 * thread stacks near the top of their 623 * address space. If their TOS is greater 624 * than vm_maxsaddr, vm_map_growstack() 625 * will confuse the thread stack with the 626 * process stack and deliver a SEGV if they 627 * attempt to grow the thread stack past their 628 * current stacksize rlimit. To avoid this, 629 * adjust vm_maxsaddr upwards to reflect 630 * the current stacksize rlimit rather 631 * than the maximum possible stacksize. 632 * It would be better to adjust the 633 * mmap'ed region, but some apps do not check 634 * mmap's return value. 635 */ 636 PROC_LOCK(p); 637 p->p_vmspace->vm_maxsaddr = (char *)LINUX32_USRSTACK - 638 lim_cur(p, RLIMIT_STACK); 639 PROC_UNLOCK(p); 640 } 641 642 /* 643 * This gives us our maximum stack size and a new BOS. 644 * If we're using VM_STACK, then mmap will just map 645 * the top SGROWSIZ bytes, and let the stack grow down 646 * to the limit at BOS. If we're not using VM_STACK 647 * we map the full stack, since we don't have a way 648 * to autogrow it. 649 */ 650 if (len > STACK_SIZE - GUARD_SIZE) { 651 bsd_args.addr = (caddr_t)PTRIN(addr); 652 bsd_args.len = len; 653 } else { 654 bsd_args.addr = (caddr_t)PTRIN(addr) - 655 (STACK_SIZE - GUARD_SIZE - len); 656 bsd_args.len = STACK_SIZE - GUARD_SIZE; 657 } 658 } else { 659 bsd_args.addr = (caddr_t)PTRIN(addr); 660 bsd_args.len = len; 661 } 662 bsd_args.pos = pos; 663 664#ifdef DEBUG 665 if (ldebug(mmap)) 666 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", 667 __func__, 668 (void *)bsd_args.addr, (int)bsd_args.len, bsd_args.prot, 669 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); 670#endif 671 error = sys_mmap(td, &bsd_args); 672#ifdef DEBUG 673 if (ldebug(mmap)) 674 printf("-> %s() return: 0x%x (0x%08x)\n", 675 __func__, error, (u_int)td->td_retval[0]); 676#endif 677 return (error); 678} 679 680int 681linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) 682{ 683 struct mprotect_args bsd_args; 684 685 bsd_args.addr = uap->addr; 686 bsd_args.len = uap->len; 687 bsd_args.prot = uap->prot; 688 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 689 bsd_args.prot |= PROT_READ | PROT_EXEC; 690 return (sys_mprotect(td, &bsd_args)); 691} 692 693int 694linux_iopl(struct thread *td, struct linux_iopl_args *args) 695{ 696 int error; 697 698 if (args->level < 0 || args->level > 3) 699 return (EINVAL); 700 if ((error = priv_check(td, PRIV_IO)) != 0) 701 return (error); 702 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 703 return (error); 704 td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) | 705 (args->level * (PSL_IOPL / 3)); 706 707 return (0); 708} 709 710int 711linux_sigaction(struct thread *td, struct linux_sigaction_args *args) 712{ 713 l_osigaction_t osa; 714 l_sigaction_t act, oact; 715 int error; 716 717#ifdef DEBUG 718 if (ldebug(sigaction)) 719 printf(ARGS(sigaction, "%d, %p, %p"), 720 args->sig, (void *)args->nsa, (void *)args->osa); 721#endif 722 723 if (args->nsa != NULL) { 724 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); 725 if (error) 726 return (error); 727 act.lsa_handler = osa.lsa_handler; 728 act.lsa_flags = osa.lsa_flags; 729 act.lsa_restorer = osa.lsa_restorer; 730 LINUX_SIGEMPTYSET(act.lsa_mask); 731 act.lsa_mask.__bits[0] = osa.lsa_mask; 732 } 733 734 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, 735 args->osa ? &oact : NULL); 736 737 if (args->osa != NULL && !error) { 738 osa.lsa_handler = oact.lsa_handler; 739 osa.lsa_flags = oact.lsa_flags; 740 osa.lsa_restorer = oact.lsa_restorer; 741 osa.lsa_mask = oact.lsa_mask.__bits[0]; 742 error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); 743 } 744 745 return (error); 746} 747 748/* 749 * Linux has two extra args, restart and oldmask. We don't use these, 750 * but it seems that "restart" is actually a context pointer that 751 * enables the signal to happen with a different register set. 752 */ 753int 754linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) 755{ 756 sigset_t sigmask; 757 l_sigset_t mask; 758 759#ifdef DEBUG 760 if (ldebug(sigsuspend)) 761 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); 762#endif 763 764 LINUX_SIGEMPTYSET(mask); 765 mask.__bits[0] = args->mask; 766 linux_to_bsd_sigset(&mask, &sigmask); 767 return (kern_sigsuspend(td, sigmask)); 768} 769 770int 771linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) 772{ 773 l_sigset_t lmask; 774 sigset_t sigmask; 775 int error; 776 777#ifdef DEBUG 778 if (ldebug(rt_sigsuspend)) 779 printf(ARGS(rt_sigsuspend, "%p, %d"), 780 (void *)uap->newset, uap->sigsetsize); 781#endif 782 783 if (uap->sigsetsize != sizeof(l_sigset_t)) 784 return (EINVAL); 785 786 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t)); 787 if (error) 788 return (error); 789 790 linux_to_bsd_sigset(&lmask, &sigmask); 791 return (kern_sigsuspend(td, sigmask)); 792} 793 794int 795linux_pause(struct thread *td, struct linux_pause_args *args) 796{ 797 struct proc *p = td->td_proc; 798 sigset_t sigmask; 799 800#ifdef DEBUG 801 if (ldebug(pause)) 802 printf(ARGS(pause, "")); 803#endif 804 805 PROC_LOCK(p); 806 sigmask = td->td_sigmask; 807 PROC_UNLOCK(p); 808 return (kern_sigsuspend(td, sigmask)); 809} 810 811int 812linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) 813{ 814 stack_t ss, oss; 815 l_stack_t lss; 816 int error; 817 818#ifdef DEBUG 819 if (ldebug(sigaltstack)) 820 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); 821#endif 822 823 if (uap->uss != NULL) { 824 error = copyin(uap->uss, &lss, sizeof(l_stack_t)); 825 if (error) 826 return (error); 827 828 ss.ss_sp = PTRIN(lss.ss_sp); 829 ss.ss_size = lss.ss_size; 830 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 831 } 832 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL, 833 (uap->uoss != NULL) ? &oss : NULL); 834 if (!error && uap->uoss != NULL) { 835 lss.ss_sp = PTROUT(oss.ss_sp); 836 lss.ss_size = oss.ss_size; 837 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags); 838 error = copyout(&lss, uap->uoss, sizeof(l_stack_t)); 839 } 840 841 return (error); 842} 843 844int 845linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) 846{ 847 struct ftruncate_args sa; 848 849#ifdef DEBUG 850 if (ldebug(ftruncate64)) 851 printf(ARGS(ftruncate64, "%u, %jd"), args->fd, 852 (intmax_t)args->length); 853#endif 854 855 sa.fd = args->fd; 856 sa.length = args->length; 857 return sys_ftruncate(td, &sa); 858} 859 860int 861linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap) 862{ 863 struct timeval atv; 864 l_timeval atv32; 865 struct timezone rtz; 866 int error = 0; 867 868 if (uap->tp) { 869 microtime(&atv); 870 atv32.tv_sec = atv.tv_sec; 871 atv32.tv_usec = atv.tv_usec; 872 error = copyout(&atv32, uap->tp, sizeof(atv32)); 873 } 874 if (error == 0 && uap->tzp != NULL) { 875 rtz.tz_minuteswest = tz_minuteswest; 876 rtz.tz_dsttime = tz_dsttime; 877 error = copyout(&rtz, uap->tzp, sizeof(rtz)); 878 } 879 return (error); 880} 881 882int 883linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap) 884{ 885 l_timeval atv32; 886 struct timeval atv, *tvp; 887 struct timezone atz, *tzp; 888 int error; 889 890 if (uap->tp) { 891 error = copyin(uap->tp, &atv32, sizeof(atv32)); 892 if (error) 893 return (error); 894 atv.tv_sec = atv32.tv_sec; 895 atv.tv_usec = atv32.tv_usec; 896 tvp = &atv; 897 } else 898 tvp = NULL; 899 if (uap->tzp) { 900 error = copyin(uap->tzp, &atz, sizeof(atz)); 901 if (error) 902 return (error); 903 tzp = &atz; 904 } else 905 tzp = NULL; 906 return (kern_settimeofday(td, tvp, tzp)); 907} 908 909int 910linux_getrusage(struct thread *td, struct linux_getrusage_args *uap) 911{ 912 struct l_rusage s32; 913 struct rusage s; 914 int error; 915 916 error = kern_getrusage(td, uap->who, &s); 917 if (error != 0) 918 return (error); 919 if (uap->rusage != NULL) { 920 bsd_to_linux_rusage(&s, &s32); 921 error = copyout(&s32, uap->rusage, sizeof(s32)); 922 } 923 return (error); 924} 925 926int 927linux_set_thread_area(struct thread *td, 928 struct linux_set_thread_area_args *args) 929{ 930 struct l_user_desc info; 931 struct user_segment_descriptor sd; 932 struct pcb *pcb; 933 int a[2]; 934 int error; 935 936 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 937 if (error) 938 return (error); 939 940#ifdef DEBUG 941 if (ldebug(set_thread_area)) 942 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, " 943 "%i, %i, %i"), info.entry_number, info.base_addr, 944 info.limit, info.seg_32bit, info.contents, 945 info.read_exec_only, info.limit_in_pages, 946 info.seg_not_present, info.useable); 947#endif 948 949 /* 950 * Semantics of Linux version: every thread in the system has array 951 * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. 952 * This syscall loads one of the selected TLS decriptors with a value 953 * and also loads GDT descriptors 6, 7 and 8 with the content of 954 * the per-thread descriptors. 955 * 956 * Semantics of FreeBSD version: I think we can ignore that Linux has 957 * three per-thread descriptors and use just the first one. 958 * The tls_array[] is used only in [gs]et_thread_area() syscalls and 959 * for loading the GDT descriptors. We use just one GDT descriptor 960 * for TLS, so we will load just one. 961 * 962 * XXX: This doesn't work when a user space process tries to use more 963 * than one TLS segment. Comment in the Linux source says wine might 964 * do this. 965 */ 966 967 /* 968 * GLIBC reads current %gs and call set_thread_area() with it. 969 * We should let GUDATA_SEL and GUGS32_SEL proceed as well because 970 * we use these segments. 971 */ 972 switch (info.entry_number) { 973 case GUGS32_SEL: 974 case GUDATA_SEL: 975 case 6: 976 case -1: 977 info.entry_number = GUGS32_SEL; 978 break; 979 default: 980 return (EINVAL); 981 } 982 983 /* 984 * We have to copy out the GDT entry we use. 985 * 986 * XXX: What if a user space program does not check the return value 987 * and tries to use 6, 7 or 8? 988 */ 989 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 990 if (error) 991 return (error); 992 993 if (LINUX_LDT_empty(&info)) { 994 a[0] = 0; 995 a[1] = 0; 996 } else { 997 a[0] = LINUX_LDT_entry_a(&info); 998 a[1] = LINUX_LDT_entry_b(&info); 999 } 1000 1001 memcpy(&sd, &a, sizeof(a)); 1002#ifdef DEBUG 1003 if (ldebug(set_thread_area)) 1004 printf("Segment created in set_thread_area: " 1005 "lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, " 1006 "type: %i, dpl: %i, p: %i, xx: %i, long: %i, " 1007 "def32: %i, gran: %i\n", 1008 sd.sd_lobase, 1009 sd.sd_hibase, 1010 sd.sd_lolimit, 1011 sd.sd_hilimit, 1012 sd.sd_type, 1013 sd.sd_dpl, 1014 sd.sd_p, 1015 sd.sd_xx, 1016 sd.sd_long, 1017 sd.sd_def32, 1018 sd.sd_gran); 1019#endif 1020 1021 pcb = td->td_pcb; 1022 pcb->pcb_gsbase = (register_t)info.base_addr; 1023 set_pcb_flags(pcb, PCB_32BIT); 1024 update_gdt_gsbase(td, info.base_addr); 1025 1026 return (0); 1027} 1028 1029int 1030linux_wait4(struct thread *td, struct linux_wait4_args *args) 1031{ 1032 int error, options; 1033 struct rusage ru, *rup; 1034 struct l_rusage lru; 1035 1036#ifdef DEBUG 1037 if (ldebug(wait4)) 1038 printf(ARGS(wait4, "%d, %p, %d, %p"), 1039 args->pid, (void *)args->status, args->options, 1040 (void *)args->rusage); 1041#endif 1042 1043 options = (args->options & (WNOHANG | WUNTRACED)); 1044 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 1045 if (args->options & __WCLONE) 1046 options |= WLINUXCLONE; 1047 1048 if (args->rusage != NULL) 1049 rup = &ru; 1050 else 1051 rup = NULL; 1052 error = linux_common_wait(td, args->pid, args->status, options, rup); 1053 if (error) 1054 return (error); 1055 if (args->rusage != NULL) { 1056 bsd_to_linux_rusage(rup, &lru); 1057 error = copyout(&lru, args->rusage, sizeof(lru)); 1058 } 1059 1060 return (error); 1061} 1062