linux32_sysvec.c revision 294904
1/*- 2 * Copyright (c) 2004 Tim J. Robbins 3 * Copyright (c) 2003 Peter Wemm 4 * Copyright (c) 2002 Doug Rabson 5 * Copyright (c) 1998-1999 Andrew Gallatin 6 * Copyright (c) 1994-1996 S��ren Schmidt 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer 14 * in this position and unchanged. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. The name of the author may not be used to endorse or promote products 19 * derived from this software without specific prior written permission 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 22 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 23 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 24 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 26 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 27 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 28 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 29 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 30 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33#include <sys/cdefs.h> 34__FBSDID("$FreeBSD: releng/10.2/sys/amd64/linux32/linux32_sysvec.c 294904 2016-01-27 07:41:31Z delphij $"); 35#include "opt_compat.h" 36 37#ifndef COMPAT_FREEBSD32 38#error "Unable to compile Linux-emulator due to missing COMPAT_FREEBSD32 option!" 39#endif 40 41#define __ELF_WORD_SIZE 32 42 43#include <sys/param.h> 44#include <sys/systm.h> 45#include <sys/exec.h> 46#include <sys/fcntl.h> 47#include <sys/imgact.h> 48#include <sys/imgact_elf.h> 49#include <sys/kernel.h> 50#include <sys/lock.h> 51#include <sys/malloc.h> 52#include <sys/module.h> 53#include <sys/mutex.h> 54#include <sys/proc.h> 55#include <sys/resourcevar.h> 56#include <sys/signalvar.h> 57#include <sys/sysctl.h> 58#include <sys/syscallsubr.h> 59#include <sys/sysent.h> 60#include <sys/sysproto.h> 61#include <sys/vnode.h> 62#include <sys/eventhandler.h> 63 64#include <vm/vm.h> 65#include <vm/pmap.h> 66#include <vm/vm_extern.h> 67#include <vm/vm_map.h> 68#include <vm/vm_object.h> 69#include <vm/vm_page.h> 70#include <vm/vm_param.h> 71 72#include <machine/cpu.h> 73#include <machine/md_var.h> 74#include <machine/pcb.h> 75#include <machine/specialreg.h> 76 77#include <amd64/linux32/linux.h> 78#include <amd64/linux32/linux32_proto.h> 79#include <compat/linux/linux_emul.h> 80#include <compat/linux/linux_futex.h> 81#include <compat/linux/linux_ioctl.h> 82#include <compat/linux/linux_mib.h> 83#include <compat/linux/linux_misc.h> 84#include <compat/linux/linux_signal.h> 85#include <compat/linux/linux_util.h> 86 87MODULE_VERSION(linux, 1); 88 89MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures"); 90 91#define AUXARGS_ENTRY_32(pos, id, val) \ 92 do { \ 93 suword32(pos++, id); \ 94 suword32(pos++, val); \ 95 } while (0) 96 97#if BYTE_ORDER == LITTLE_ENDIAN 98#define SHELLMAGIC 0x2123 /* #! */ 99#else 100#define SHELLMAGIC 0x2321 101#endif 102 103/* 104 * Allow the sendsig functions to use the ldebug() facility 105 * even though they are not syscalls themselves. Map them 106 * to syscall 0. This is slightly less bogus than using 107 * ldebug(sigreturn). 108 */ 109#define LINUX_SYS_linux_rt_sendsig 0 110#define LINUX_SYS_linux_sendsig 0 111 112const char *linux_platform = "i686"; 113static int linux_szplatform; 114extern char linux_sigcode[]; 115extern int linux_szsigcode; 116 117extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL]; 118 119SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); 120SET_DECLARE(linux_device_handler_set, struct linux_device_handler); 121 122static int elf_linux_fixup(register_t **stack_base, 123 struct image_params *iparams); 124static register_t *linux_copyout_strings(struct image_params *imgp); 125static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask); 126static void exec_linux_setregs(struct thread *td, 127 struct image_params *imgp, u_long stack); 128static void linux32_fixlimit(struct rlimit *rl, int which); 129static boolean_t linux32_trans_osrel(const Elf_Note *note, int32_t *osrel); 130 131static eventhandler_tag linux_exit_tag; 132static eventhandler_tag linux_exec_tag; 133 134/* 135 * Linux syscalls return negative errno's, we do positive and map them 136 * Reference: 137 * FreeBSD: src/sys/sys/errno.h 138 * Linux: linux-2.6.17.8/include/asm-generic/errno-base.h 139 * linux-2.6.17.8/include/asm-generic/errno.h 140 */ 141static int bsd_to_linux_errno[ELAST + 1] = { 142 -0, -1, -2, -3, -4, -5, -6, -7, -8, -9, 143 -10, -35, -12, -13, -14, -15, -16, -17, -18, -19, 144 -20, -21, -22, -23, -24, -25, -26, -27, -28, -29, 145 -30, -31, -32, -33, -34, -11,-115,-114, -88, -89, 146 -90, -91, -92, -93, -94, -95, -96, -97, -98, -99, 147 -100,-101,-102,-103,-104,-105,-106,-107,-108,-109, 148 -110,-111, -40, -36,-112,-113, -39, -11, -87,-122, 149 -116, -66, -6, -6, -6, -6, -6, -37, -38, -9, 150 -6, -6, -43, -42, -75,-125, -84, -95, -16, -74, 151 -72, -67, -71 152}; 153 154int bsd_to_linux_signal[LINUX_SIGTBLSZ] = { 155 LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT, LINUX_SIGILL, 156 LINUX_SIGTRAP, LINUX_SIGABRT, 0, LINUX_SIGFPE, 157 LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV, LINUX_SIGSYS, 158 LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM, LINUX_SIGURG, 159 LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT, LINUX_SIGCHLD, 160 LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO, LINUX_SIGXCPU, 161 LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF, LINUX_SIGWINCH, 162 0, LINUX_SIGUSR1, LINUX_SIGUSR2 163}; 164 165int linux_to_bsd_signal[LINUX_SIGTBLSZ] = { 166 SIGHUP, SIGINT, SIGQUIT, SIGILL, 167 SIGTRAP, SIGABRT, SIGBUS, SIGFPE, 168 SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2, 169 SIGPIPE, SIGALRM, SIGTERM, SIGBUS, 170 SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP, 171 SIGTTIN, SIGTTOU, SIGURG, SIGXCPU, 172 SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH, 173 SIGIO, SIGURG, SIGSYS 174}; 175 176#define LINUX_T_UNKNOWN 255 177static int _bsd_to_linux_trapcode[] = { 178 LINUX_T_UNKNOWN, /* 0 */ 179 6, /* 1 T_PRIVINFLT */ 180 LINUX_T_UNKNOWN, /* 2 */ 181 3, /* 3 T_BPTFLT */ 182 LINUX_T_UNKNOWN, /* 4 */ 183 LINUX_T_UNKNOWN, /* 5 */ 184 16, /* 6 T_ARITHTRAP */ 185 254, /* 7 T_ASTFLT */ 186 LINUX_T_UNKNOWN, /* 8 */ 187 13, /* 9 T_PROTFLT */ 188 1, /* 10 T_TRCTRAP */ 189 LINUX_T_UNKNOWN, /* 11 */ 190 14, /* 12 T_PAGEFLT */ 191 LINUX_T_UNKNOWN, /* 13 */ 192 17, /* 14 T_ALIGNFLT */ 193 LINUX_T_UNKNOWN, /* 15 */ 194 LINUX_T_UNKNOWN, /* 16 */ 195 LINUX_T_UNKNOWN, /* 17 */ 196 0, /* 18 T_DIVIDE */ 197 2, /* 19 T_NMI */ 198 4, /* 20 T_OFLOW */ 199 5, /* 21 T_BOUND */ 200 7, /* 22 T_DNA */ 201 8, /* 23 T_DOUBLEFLT */ 202 9, /* 24 T_FPOPFLT */ 203 10, /* 25 T_TSSFLT */ 204 11, /* 26 T_SEGNPFLT */ 205 12, /* 27 T_STKFLT */ 206 18, /* 28 T_MCHK */ 207 19, /* 29 T_XMMFLT */ 208 15 /* 30 T_RESERVED */ 209}; 210#define bsd_to_linux_trapcode(code) \ 211 ((code)<sizeof(_bsd_to_linux_trapcode)/sizeof(*_bsd_to_linux_trapcode)? \ 212 _bsd_to_linux_trapcode[(code)]: \ 213 LINUX_T_UNKNOWN) 214 215struct linux32_ps_strings { 216 u_int32_t ps_argvstr; /* first of 0 or more argument strings */ 217 u_int ps_nargvstr; /* the number of argument strings */ 218 u_int32_t ps_envstr; /* first of 0 or more environment strings */ 219 u_int ps_nenvstr; /* the number of environment strings */ 220}; 221 222/* 223 * If FreeBSD & Linux have a difference of opinion about what a trap 224 * means, deal with it here. 225 * 226 * MPSAFE 227 */ 228static int 229translate_traps(int signal, int trap_code) 230{ 231 if (signal != SIGBUS) 232 return signal; 233 switch (trap_code) { 234 case T_PROTFLT: 235 case T_TSSFLT: 236 case T_DOUBLEFLT: 237 case T_PAGEFLT: 238 return SIGSEGV; 239 default: 240 return signal; 241 } 242} 243 244static int 245elf_linux_fixup(register_t **stack_base, struct image_params *imgp) 246{ 247 Elf32_Auxargs *args; 248 Elf32_Addr *base; 249 Elf32_Addr *pos, *uplatform; 250 struct linux32_ps_strings *arginfo; 251 int issetugid; 252 253 arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS; 254 uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szplatform); 255 256 KASSERT(curthread->td_proc == imgp->proc, 257 ("unsafe elf_linux_fixup(), should be curproc")); 258 base = (Elf32_Addr *)*stack_base; 259 args = (Elf32_Auxargs *)imgp->auxargs; 260 pos = base + (imgp->args->argc + imgp->args->envc + 2); 261 262 issetugid = imgp->proc->p_flag & P_SUGID ? 1 : 0; 263 AUXARGS_ENTRY_32(pos, LINUX_AT_HWCAP, cpu_feature); 264 265 /* 266 * Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0, 267 * as it has appeared in the 2.4.0-rc7 first time. 268 * Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK), 269 * glibc falls back to the hard-coded CLK_TCK value when aux entry 270 * is not present. 271 * Also see linux_times() implementation. 272 */ 273 if (linux_kernver(curthread) >= LINUX_KERNVER_2004000) 274 AUXARGS_ENTRY_32(pos, LINUX_AT_CLKTCK, stclohz); 275 AUXARGS_ENTRY_32(pos, AT_PHDR, args->phdr); 276 AUXARGS_ENTRY_32(pos, AT_PHENT, args->phent); 277 AUXARGS_ENTRY_32(pos, AT_PHNUM, args->phnum); 278 AUXARGS_ENTRY_32(pos, AT_PAGESZ, args->pagesz); 279 AUXARGS_ENTRY_32(pos, AT_FLAGS, args->flags); 280 AUXARGS_ENTRY_32(pos, AT_ENTRY, args->entry); 281 AUXARGS_ENTRY_32(pos, AT_BASE, args->base); 282 AUXARGS_ENTRY_32(pos, LINUX_AT_SECURE, issetugid); 283 AUXARGS_ENTRY_32(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); 284 AUXARGS_ENTRY_32(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); 285 AUXARGS_ENTRY_32(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); 286 AUXARGS_ENTRY_32(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); 287 AUXARGS_ENTRY_32(pos, LINUX_AT_PLATFORM, PTROUT(uplatform)); 288 if (args->execfd != -1) 289 AUXARGS_ENTRY_32(pos, AT_EXECFD, args->execfd); 290 AUXARGS_ENTRY_32(pos, AT_NULL, 0); 291 292 free(imgp->auxargs, M_TEMP); 293 imgp->auxargs = NULL; 294 295 base--; 296 suword32(base, (uint32_t)imgp->args->argc); 297 *stack_base = (register_t *)base; 298 return 0; 299} 300 301extern unsigned long linux_sznonrtsigcode; 302 303static void 304linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 305{ 306 struct thread *td = curthread; 307 struct proc *p = td->td_proc; 308 struct sigacts *psp; 309 struct trapframe *regs; 310 struct l_rt_sigframe *fp, frame; 311 int oonstack; 312 int sig; 313 int code; 314 315 sig = ksi->ksi_signo; 316 code = ksi->ksi_code; 317 PROC_LOCK_ASSERT(p, MA_OWNED); 318 psp = p->p_sigacts; 319 mtx_assert(&psp->ps_mtx, MA_OWNED); 320 regs = td->td_frame; 321 oonstack = sigonstack(regs->tf_rsp); 322 323#ifdef DEBUG 324 if (ldebug(rt_sendsig)) 325 printf(ARGS(rt_sendsig, "%p, %d, %p, %u"), 326 catcher, sig, (void*)mask, code); 327#endif 328 /* 329 * Allocate space for the signal handler context. 330 */ 331 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && 332 SIGISMEMBER(psp->ps_sigonstack, sig)) { 333 fp = (struct l_rt_sigframe *)(td->td_sigstk.ss_sp + 334 td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe)); 335 } else 336 fp = (struct l_rt_sigframe *)regs->tf_rsp - 1; 337 mtx_unlock(&psp->ps_mtx); 338 339 /* 340 * Build the argument list for the signal handler. 341 */ 342 if (p->p_sysent->sv_sigtbl) 343 if (sig <= p->p_sysent->sv_sigsize) 344 sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)]; 345 346 bzero(&frame, sizeof(frame)); 347 348 frame.sf_handler = PTROUT(catcher); 349 frame.sf_sig = sig; 350 frame.sf_siginfo = PTROUT(&fp->sf_si); 351 frame.sf_ucontext = PTROUT(&fp->sf_sc); 352 353 /* Fill in POSIX parts */ 354 ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig); 355 356 /* 357 * Build the signal context to be used by sigreturn. 358 */ 359 frame.sf_sc.uc_flags = 0; /* XXX ??? */ 360 frame.sf_sc.uc_link = 0; /* XXX ??? */ 361 362 frame.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp); 363 frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size; 364 frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) 365 ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE; 366 PROC_UNLOCK(p); 367 368 bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask); 369 370 frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__bits[0]; 371 frame.sf_sc.uc_mcontext.sc_edi = regs->tf_rdi; 372 frame.sf_sc.uc_mcontext.sc_esi = regs->tf_rsi; 373 frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_rbp; 374 frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_rbx; 375 frame.sf_sc.uc_mcontext.sc_edx = regs->tf_rdx; 376 frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_rcx; 377 frame.sf_sc.uc_mcontext.sc_eax = regs->tf_rax; 378 frame.sf_sc.uc_mcontext.sc_eip = regs->tf_rip; 379 frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs; 380 frame.sf_sc.uc_mcontext.sc_gs = regs->tf_gs; 381 frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs; 382 frame.sf_sc.uc_mcontext.sc_es = regs->tf_es; 383 frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds; 384 frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_rflags; 385 frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_rsp; 386 frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss; 387 frame.sf_sc.uc_mcontext.sc_err = regs->tf_err; 388 frame.sf_sc.uc_mcontext.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr; 389 frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code); 390 391#ifdef DEBUG 392 if (ldebug(rt_sendsig)) 393 printf(LMSG("rt_sendsig flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x"), 394 frame.sf_sc.uc_stack.ss_flags, td->td_sigstk.ss_sp, 395 td->td_sigstk.ss_size, frame.sf_sc.uc_mcontext.sc_mask); 396#endif 397 398 if (copyout(&frame, fp, sizeof(frame)) != 0) { 399 /* 400 * Process has trashed its stack; give it an illegal 401 * instruction to halt it in its tracks. 402 */ 403#ifdef DEBUG 404 if (ldebug(rt_sendsig)) 405 printf(LMSG("rt_sendsig: bad stack %p, oonstack=%x"), 406 fp, oonstack); 407#endif 408 PROC_LOCK(p); 409 sigexit(td, SIGILL); 410 } 411 412 /* 413 * Build context to run handler in. 414 */ 415 regs->tf_rsp = PTROUT(fp); 416 regs->tf_rip = p->p_sysent->sv_sigcode_base + linux_sznonrtsigcode; 417 regs->tf_rflags &= ~(PSL_T | PSL_D); 418 regs->tf_cs = _ucode32sel; 419 regs->tf_ss = _udatasel; 420 regs->tf_ds = _udatasel; 421 regs->tf_es = _udatasel; 422 regs->tf_fs = _ufssel; 423 regs->tf_gs = _ugssel; 424 regs->tf_flags = TF_HASSEGS; 425 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 426 PROC_LOCK(p); 427 mtx_lock(&psp->ps_mtx); 428} 429 430 431/* 432 * Send an interrupt to process. 433 * 434 * Stack is set up to allow sigcode stored 435 * in u. to call routine, followed by kcall 436 * to sigreturn routine below. After sigreturn 437 * resets the signal mask, the stack, and the 438 * frame pointer, it returns to the user 439 * specified pc, psl. 440 */ 441static void 442linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 443{ 444 struct thread *td = curthread; 445 struct proc *p = td->td_proc; 446 struct sigacts *psp; 447 struct trapframe *regs; 448 struct l_sigframe *fp, frame; 449 l_sigset_t lmask; 450 int oonstack, i; 451 int sig, code; 452 453 sig = ksi->ksi_signo; 454 code = ksi->ksi_code; 455 PROC_LOCK_ASSERT(p, MA_OWNED); 456 psp = p->p_sigacts; 457 mtx_assert(&psp->ps_mtx, MA_OWNED); 458 if (SIGISMEMBER(psp->ps_siginfo, sig)) { 459 /* Signal handler installed with SA_SIGINFO. */ 460 linux_rt_sendsig(catcher, ksi, mask); 461 return; 462 } 463 464 regs = td->td_frame; 465 oonstack = sigonstack(regs->tf_rsp); 466 467#ifdef DEBUG 468 if (ldebug(sendsig)) 469 printf(ARGS(sendsig, "%p, %d, %p, %u"), 470 catcher, sig, (void*)mask, code); 471#endif 472 473 /* 474 * Allocate space for the signal handler context. 475 */ 476 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && 477 SIGISMEMBER(psp->ps_sigonstack, sig)) { 478 fp = (struct l_sigframe *)(td->td_sigstk.ss_sp + 479 td->td_sigstk.ss_size - sizeof(struct l_sigframe)); 480 } else 481 fp = (struct l_sigframe *)regs->tf_rsp - 1; 482 mtx_unlock(&psp->ps_mtx); 483 PROC_UNLOCK(p); 484 485 /* 486 * Build the argument list for the signal handler. 487 */ 488 if (p->p_sysent->sv_sigtbl) 489 if (sig <= p->p_sysent->sv_sigsize) 490 sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)]; 491 492 bzero(&frame, sizeof(frame)); 493 494 frame.sf_handler = PTROUT(catcher); 495 frame.sf_sig = sig; 496 497 bsd_to_linux_sigset(mask, &lmask); 498 499 /* 500 * Build the signal context to be used by sigreturn. 501 */ 502 frame.sf_sc.sc_mask = lmask.__bits[0]; 503 frame.sf_sc.sc_gs = regs->tf_gs; 504 frame.sf_sc.sc_fs = regs->tf_fs; 505 frame.sf_sc.sc_es = regs->tf_es; 506 frame.sf_sc.sc_ds = regs->tf_ds; 507 frame.sf_sc.sc_edi = regs->tf_rdi; 508 frame.sf_sc.sc_esi = regs->tf_rsi; 509 frame.sf_sc.sc_ebp = regs->tf_rbp; 510 frame.sf_sc.sc_ebx = regs->tf_rbx; 511 frame.sf_sc.sc_edx = regs->tf_rdx; 512 frame.sf_sc.sc_ecx = regs->tf_rcx; 513 frame.sf_sc.sc_eax = regs->tf_rax; 514 frame.sf_sc.sc_eip = regs->tf_rip; 515 frame.sf_sc.sc_cs = regs->tf_cs; 516 frame.sf_sc.sc_eflags = regs->tf_rflags; 517 frame.sf_sc.sc_esp_at_signal = regs->tf_rsp; 518 frame.sf_sc.sc_ss = regs->tf_ss; 519 frame.sf_sc.sc_err = regs->tf_err; 520 frame.sf_sc.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr; 521 frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(code); 522 523 for (i = 0; i < (LINUX_NSIG_WORDS-1); i++) 524 frame.sf_extramask[i] = lmask.__bits[i+1]; 525 526 if (copyout(&frame, fp, sizeof(frame)) != 0) { 527 /* 528 * Process has trashed its stack; give it an illegal 529 * instruction to halt it in its tracks. 530 */ 531 PROC_LOCK(p); 532 sigexit(td, SIGILL); 533 } 534 535 /* 536 * Build context to run handler in. 537 */ 538 regs->tf_rsp = PTROUT(fp); 539 regs->tf_rip = p->p_sysent->sv_sigcode_base; 540 regs->tf_rflags &= ~(PSL_T | PSL_D); 541 regs->tf_cs = _ucode32sel; 542 regs->tf_ss = _udatasel; 543 regs->tf_ds = _udatasel; 544 regs->tf_es = _udatasel; 545 regs->tf_fs = _ufssel; 546 regs->tf_gs = _ugssel; 547 regs->tf_flags = TF_HASSEGS; 548 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 549 PROC_LOCK(p); 550 mtx_lock(&psp->ps_mtx); 551} 552 553/* 554 * System call to cleanup state after a signal 555 * has been taken. Reset signal mask and 556 * stack state from context left by sendsig (above). 557 * Return to previous pc and psl as specified by 558 * context left by sendsig. Check carefully to 559 * make sure that the user has not modified the 560 * psl to gain improper privileges or to cause 561 * a machine fault. 562 */ 563int 564linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args) 565{ 566 struct l_sigframe frame; 567 struct trapframe *regs; 568 sigset_t bmask; 569 l_sigset_t lmask; 570 int eflags, i; 571 ksiginfo_t ksi; 572 573 regs = td->td_frame; 574 575#ifdef DEBUG 576 if (ldebug(sigreturn)) 577 printf(ARGS(sigreturn, "%p"), (void *)args->sfp); 578#endif 579 /* 580 * The trampoline code hands us the sigframe. 581 * It is unsafe to keep track of it ourselves, in the event that a 582 * program jumps out of a signal handler. 583 */ 584 if (copyin(args->sfp, &frame, sizeof(frame)) != 0) 585 return (EFAULT); 586 587 /* 588 * Check for security violations. 589 */ 590#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) 591 eflags = frame.sf_sc.sc_eflags; 592 if (!EFLAGS_SECURE(eflags, regs->tf_rflags)) 593 return(EINVAL); 594 595 /* 596 * Don't allow users to load a valid privileged %cs. Let the 597 * hardware check for invalid selectors, excess privilege in 598 * other selectors, invalid %eip's and invalid %esp's. 599 */ 600#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) 601 if (!CS_SECURE(frame.sf_sc.sc_cs)) { 602 ksiginfo_init_trap(&ksi); 603 ksi.ksi_signo = SIGBUS; 604 ksi.ksi_code = BUS_OBJERR; 605 ksi.ksi_trapno = T_PROTFLT; 606 ksi.ksi_addr = (void *)regs->tf_rip; 607 trapsignal(td, &ksi); 608 return(EINVAL); 609 } 610 611 lmask.__bits[0] = frame.sf_sc.sc_mask; 612 for (i = 0; i < (LINUX_NSIG_WORDS-1); i++) 613 lmask.__bits[i+1] = frame.sf_extramask[i]; 614 linux_to_bsd_sigset(&lmask, &bmask); 615 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); 616 617 /* 618 * Restore signal context. 619 */ 620 regs->tf_rdi = frame.sf_sc.sc_edi; 621 regs->tf_rsi = frame.sf_sc.sc_esi; 622 regs->tf_rbp = frame.sf_sc.sc_ebp; 623 regs->tf_rbx = frame.sf_sc.sc_ebx; 624 regs->tf_rdx = frame.sf_sc.sc_edx; 625 regs->tf_rcx = frame.sf_sc.sc_ecx; 626 regs->tf_rax = frame.sf_sc.sc_eax; 627 regs->tf_rip = frame.sf_sc.sc_eip; 628 regs->tf_cs = frame.sf_sc.sc_cs; 629 regs->tf_ds = frame.sf_sc.sc_ds; 630 regs->tf_es = frame.sf_sc.sc_es; 631 regs->tf_fs = frame.sf_sc.sc_fs; 632 regs->tf_gs = frame.sf_sc.sc_gs; 633 regs->tf_rflags = eflags; 634 regs->tf_rsp = frame.sf_sc.sc_esp_at_signal; 635 regs->tf_ss = frame.sf_sc.sc_ss; 636 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 637 638 return (EJUSTRETURN); 639} 640 641/* 642 * System call to cleanup state after a signal 643 * has been taken. Reset signal mask and 644 * stack state from context left by rt_sendsig (above). 645 * Return to previous pc and psl as specified by 646 * context left by sendsig. Check carefully to 647 * make sure that the user has not modified the 648 * psl to gain improper privileges or to cause 649 * a machine fault. 650 */ 651int 652linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 653{ 654 struct l_ucontext uc; 655 struct l_sigcontext *context; 656 sigset_t bmask; 657 l_stack_t *lss; 658 stack_t ss; 659 struct trapframe *regs; 660 int eflags; 661 ksiginfo_t ksi; 662 663 regs = td->td_frame; 664 665#ifdef DEBUG 666 if (ldebug(rt_sigreturn)) 667 printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp); 668#endif 669 /* 670 * The trampoline code hands us the ucontext. 671 * It is unsafe to keep track of it ourselves, in the event that a 672 * program jumps out of a signal handler. 673 */ 674 if (copyin(args->ucp, &uc, sizeof(uc)) != 0) 675 return (EFAULT); 676 677 context = &uc.uc_mcontext; 678 679 /* 680 * Check for security violations. 681 */ 682#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) 683 eflags = context->sc_eflags; 684 if (!EFLAGS_SECURE(eflags, regs->tf_rflags)) 685 return(EINVAL); 686 687 /* 688 * Don't allow users to load a valid privileged %cs. Let the 689 * hardware check for invalid selectors, excess privilege in 690 * other selectors, invalid %eip's and invalid %esp's. 691 */ 692#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) 693 if (!CS_SECURE(context->sc_cs)) { 694 ksiginfo_init_trap(&ksi); 695 ksi.ksi_signo = SIGBUS; 696 ksi.ksi_code = BUS_OBJERR; 697 ksi.ksi_trapno = T_PROTFLT; 698 ksi.ksi_addr = (void *)regs->tf_rip; 699 trapsignal(td, &ksi); 700 return(EINVAL); 701 } 702 703 linux_to_bsd_sigset(&uc.uc_sigmask, &bmask); 704 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); 705 706 /* 707 * Restore signal context 708 */ 709 regs->tf_gs = context->sc_gs; 710 regs->tf_fs = context->sc_fs; 711 regs->tf_es = context->sc_es; 712 regs->tf_ds = context->sc_ds; 713 regs->tf_rdi = context->sc_edi; 714 regs->tf_rsi = context->sc_esi; 715 regs->tf_rbp = context->sc_ebp; 716 regs->tf_rbx = context->sc_ebx; 717 regs->tf_rdx = context->sc_edx; 718 regs->tf_rcx = context->sc_ecx; 719 regs->tf_rax = context->sc_eax; 720 regs->tf_rip = context->sc_eip; 721 regs->tf_cs = context->sc_cs; 722 regs->tf_rflags = eflags; 723 regs->tf_rsp = context->sc_esp_at_signal; 724 regs->tf_ss = context->sc_ss; 725 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 726 727 /* 728 * call sigaltstack & ignore results.. 729 */ 730 lss = &uc.uc_stack; 731 ss.ss_sp = PTRIN(lss->ss_sp); 732 ss.ss_size = lss->ss_size; 733 ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags); 734 735#ifdef DEBUG 736 if (ldebug(rt_sigreturn)) 737 printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x"), 738 ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask); 739#endif 740 (void)kern_sigaltstack(td, &ss, NULL); 741 742 return (EJUSTRETURN); 743} 744 745static int 746linux32_fetch_syscall_args(struct thread *td, struct syscall_args *sa) 747{ 748 struct proc *p; 749 struct trapframe *frame; 750 751 p = td->td_proc; 752 frame = td->td_frame; 753 754 sa->args[0] = frame->tf_rbx; 755 sa->args[1] = frame->tf_rcx; 756 sa->args[2] = frame->tf_rdx; 757 sa->args[3] = frame->tf_rsi; 758 sa->args[4] = frame->tf_rdi; 759 sa->args[5] = frame->tf_rbp; /* Unconfirmed */ 760 sa->code = frame->tf_rax; 761 762 if (sa->code >= p->p_sysent->sv_size) 763 sa->callp = &p->p_sysent->sv_table[0]; 764 else 765 sa->callp = &p->p_sysent->sv_table[sa->code]; 766 sa->narg = sa->callp->sy_narg; 767 768 td->td_retval[0] = 0; 769 td->td_retval[1] = frame->tf_rdx; 770 771 return (0); 772} 773 774/* 775 * If a linux binary is exec'ing something, try this image activator 776 * first. We override standard shell script execution in order to 777 * be able to modify the interpreter path. We only do this if a linux 778 * binary is doing the exec, so we do not create an EXEC module for it. 779 */ 780static int exec_linux_imgact_try(struct image_params *iparams); 781 782static int 783exec_linux_imgact_try(struct image_params *imgp) 784{ 785 const char *head = (const char *)imgp->image_header; 786 char *rpath; 787 int error = -1; 788 789 /* 790 * The interpreter for shell scripts run from a linux binary needs 791 * to be located in /compat/linux if possible in order to recursively 792 * maintain linux path emulation. 793 */ 794 if (((const short *)head)[0] == SHELLMAGIC) { 795 /* 796 * Run our normal shell image activator. If it succeeds attempt 797 * to use the alternate path for the interpreter. If an 798 * alternate * path is found, use our stringspace to store it. 799 */ 800 if ((error = exec_shell_imgact(imgp)) == 0) { 801 linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc), 802 imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0, 803 AT_FDCWD); 804 if (rpath != NULL) 805 imgp->args->fname_buf = 806 imgp->interpreter_name = rpath; 807 } 808 } 809 return (error); 810} 811 812/* 813 * Clear registers on exec 814 * XXX copied from ia32_signal.c. 815 */ 816static void 817exec_linux_setregs(struct thread *td, struct image_params *imgp, u_long stack) 818{ 819 struct trapframe *regs = td->td_frame; 820 struct pcb *pcb = td->td_pcb; 821 822 mtx_lock(&dt_lock); 823 if (td->td_proc->p_md.md_ldt != NULL) 824 user_ldt_free(td); 825 else 826 mtx_unlock(&dt_lock); 827 828 critical_enter(); 829 wrmsr(MSR_FSBASE, 0); 830 wrmsr(MSR_KGSBASE, 0); /* User value while we're in the kernel */ 831 pcb->pcb_fsbase = 0; 832 pcb->pcb_gsbase = 0; 833 critical_exit(); 834 pcb->pcb_initial_fpucw = __LINUX_NPXCW__; 835 836 bzero((char *)regs, sizeof(struct trapframe)); 837 regs->tf_rip = imgp->entry_addr; 838 regs->tf_rsp = stack; 839 regs->tf_rflags = PSL_USER | (regs->tf_rflags & PSL_T); 840 regs->tf_gs = _ugssel; 841 regs->tf_fs = _ufssel; 842 regs->tf_es = _udatasel; 843 regs->tf_ds = _udatasel; 844 regs->tf_ss = _udatasel; 845 regs->tf_flags = TF_HASSEGS; 846 regs->tf_cs = _ucode32sel; 847 regs->tf_rbx = imgp->ps_strings; 848 849 fpstate_drop(td); 850 851 /* Do full restore on return so that we can change to a different %cs */ 852 set_pcb_flags(pcb, PCB_32BIT | PCB_FULL_IRET); 853 td->td_retval[1] = 0; 854} 855 856/* 857 * XXX copied from ia32_sysvec.c. 858 */ 859static register_t * 860linux_copyout_strings(struct image_params *imgp) 861{ 862 int argc, envc; 863 u_int32_t *vectp; 864 char *stringp, *destp; 865 u_int32_t *stack_base; 866 struct linux32_ps_strings *arginfo; 867 868 /* 869 * Calculate string base and vector table pointers. 870 * Also deal with signal trampoline code for this exec type. 871 */ 872 arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS; 873 destp = (caddr_t)arginfo - SPARE_USRSPACE - linux_szplatform - 874 roundup((ARG_MAX - imgp->args->stringspace), 875 sizeof(char *)); 876 877 /* 878 * Install LINUX_PLATFORM 879 */ 880 copyout(linux_platform, ((caddr_t)arginfo - linux_szplatform), 881 linux_szplatform); 882 883 /* 884 * If we have a valid auxargs ptr, prepare some room 885 * on the stack. 886 */ 887 if (imgp->auxargs) { 888 /* 889 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for 890 * lower compatibility. 891 */ 892 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size : 893 (LINUX_AT_COUNT * 2); 894 /* 895 * The '+ 2' is for the null pointers at the end of each of 896 * the arg and env vector sets,and imgp->auxarg_size is room 897 * for argument of Runtime loader. 898 */ 899 vectp = (u_int32_t *) (destp - (imgp->args->argc + 900 imgp->args->envc + 2 + imgp->auxarg_size) * 901 sizeof(u_int32_t)); 902 903 } else 904 /* 905 * The '+ 2' is for the null pointers at the end of each of 906 * the arg and env vector sets 907 */ 908 vectp = (u_int32_t *)(destp - (imgp->args->argc + 909 imgp->args->envc + 2) * sizeof(u_int32_t)); 910 911 /* 912 * vectp also becomes our initial stack base 913 */ 914 stack_base = vectp; 915 916 stringp = imgp->args->begin_argv; 917 argc = imgp->args->argc; 918 envc = imgp->args->envc; 919 /* 920 * Copy out strings - arguments and environment. 921 */ 922 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace); 923 924 /* 925 * Fill in "ps_strings" struct for ps, w, etc. 926 */ 927 suword32(&arginfo->ps_argvstr, (uint32_t)(intptr_t)vectp); 928 suword32(&arginfo->ps_nargvstr, argc); 929 930 /* 931 * Fill in argument portion of vector table. 932 */ 933 for (; argc > 0; --argc) { 934 suword32(vectp++, (uint32_t)(intptr_t)destp); 935 while (*stringp++ != 0) 936 destp++; 937 destp++; 938 } 939 940 /* a null vector table pointer separates the argp's from the envp's */ 941 suword32(vectp++, 0); 942 943 suword32(&arginfo->ps_envstr, (uint32_t)(intptr_t)vectp); 944 suword32(&arginfo->ps_nenvstr, envc); 945 946 /* 947 * Fill in environment portion of vector table. 948 */ 949 for (; envc > 0; --envc) { 950 suword32(vectp++, (uint32_t)(intptr_t)destp); 951 while (*stringp++ != 0) 952 destp++; 953 destp++; 954 } 955 956 /* end of vector table is a null pointer */ 957 suword32(vectp, 0); 958 959 return ((register_t *)stack_base); 960} 961 962static SYSCTL_NODE(_compat, OID_AUTO, linux32, CTLFLAG_RW, 0, 963 "32-bit Linux emulation"); 964 965static u_long linux32_maxdsiz = LINUX32_MAXDSIZ; 966SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxdsiz, CTLFLAG_RW, 967 &linux32_maxdsiz, 0, ""); 968static u_long linux32_maxssiz = LINUX32_MAXSSIZ; 969SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxssiz, CTLFLAG_RW, 970 &linux32_maxssiz, 0, ""); 971static u_long linux32_maxvmem = LINUX32_MAXVMEM; 972SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxvmem, CTLFLAG_RW, 973 &linux32_maxvmem, 0, ""); 974 975static void 976linux32_fixlimit(struct rlimit *rl, int which) 977{ 978 979 switch (which) { 980 case RLIMIT_DATA: 981 if (linux32_maxdsiz != 0) { 982 if (rl->rlim_cur > linux32_maxdsiz) 983 rl->rlim_cur = linux32_maxdsiz; 984 if (rl->rlim_max > linux32_maxdsiz) 985 rl->rlim_max = linux32_maxdsiz; 986 } 987 break; 988 case RLIMIT_STACK: 989 if (linux32_maxssiz != 0) { 990 if (rl->rlim_cur > linux32_maxssiz) 991 rl->rlim_cur = linux32_maxssiz; 992 if (rl->rlim_max > linux32_maxssiz) 993 rl->rlim_max = linux32_maxssiz; 994 } 995 break; 996 case RLIMIT_VMEM: 997 if (linux32_maxvmem != 0) { 998 if (rl->rlim_cur > linux32_maxvmem) 999 rl->rlim_cur = linux32_maxvmem; 1000 if (rl->rlim_max > linux32_maxvmem) 1001 rl->rlim_max = linux32_maxvmem; 1002 } 1003 break; 1004 } 1005} 1006 1007struct sysentvec elf_linux_sysvec = { 1008 .sv_size = LINUX_SYS_MAXSYSCALL, 1009 .sv_table = linux_sysent, 1010 .sv_mask = 0, 1011 .sv_sigsize = LINUX_SIGTBLSZ, 1012 .sv_sigtbl = bsd_to_linux_signal, 1013 .sv_errsize = ELAST + 1, 1014 .sv_errtbl = bsd_to_linux_errno, 1015 .sv_transtrap = translate_traps, 1016 .sv_fixup = elf_linux_fixup, 1017 .sv_sendsig = linux_sendsig, 1018 .sv_sigcode = linux_sigcode, 1019 .sv_szsigcode = &linux_szsigcode, 1020 .sv_prepsyscall = NULL, 1021 .sv_name = "Linux ELF32", 1022 .sv_coredump = elf32_coredump, 1023 .sv_imgact_try = exec_linux_imgact_try, 1024 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 1025 .sv_pagesize = PAGE_SIZE, 1026 .sv_minuser = VM_MIN_ADDRESS, 1027 .sv_maxuser = LINUX32_MAXUSER, 1028 .sv_usrstack = LINUX32_USRSTACK, 1029 .sv_psstrings = LINUX32_PS_STRINGS, 1030 .sv_stackprot = VM_PROT_ALL, 1031 .sv_copyout_strings = linux_copyout_strings, 1032 .sv_setregs = exec_linux_setregs, 1033 .sv_fixlimit = linux32_fixlimit, 1034 .sv_maxssiz = &linux32_maxssiz, 1035 .sv_flags = SV_ABI_LINUX | SV_ILP32 | SV_IA32 | SV_SHP, 1036 .sv_set_syscall_retval = cpu_set_syscall_retval, 1037 .sv_fetch_syscall_args = linux32_fetch_syscall_args, 1038 .sv_syscallnames = NULL, 1039 .sv_shared_page_base = LINUX32_SHAREDPAGE, 1040 .sv_shared_page_len = PAGE_SIZE, 1041 .sv_schedtail = linux_schedtail, 1042}; 1043INIT_SYSENTVEC(elf_sysvec, &elf_linux_sysvec); 1044 1045static char GNU_ABI_VENDOR[] = "GNU"; 1046static int GNULINUX_ABI_DESC = 0; 1047 1048static boolean_t 1049linux32_trans_osrel(const Elf_Note *note, int32_t *osrel) 1050{ 1051 const Elf32_Word *desc; 1052 uintptr_t p; 1053 1054 p = (uintptr_t)(note + 1); 1055 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 1056 1057 desc = (const Elf32_Word *)p; 1058 if (desc[0] != GNULINUX_ABI_DESC) 1059 return (FALSE); 1060 1061 /* 1062 * For linux we encode osrel as follows (see linux_mib.c): 1063 * VVVMMMIII (version, major, minor), see linux_mib.c. 1064 */ 1065 *osrel = desc[1] * 1000000 + desc[2] * 1000 + desc[3]; 1066 1067 return (TRUE); 1068} 1069 1070static Elf_Brandnote linux32_brandnote = { 1071 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 1072 .hdr.n_descsz = 16, /* XXX at least 16 */ 1073 .hdr.n_type = 1, 1074 .vendor = GNU_ABI_VENDOR, 1075 .flags = BN_TRANSLATE_OSREL, 1076 .trans_osrel = linux32_trans_osrel 1077}; 1078 1079static Elf32_Brandinfo linux_brand = { 1080 .brand = ELFOSABI_LINUX, 1081 .machine = EM_386, 1082 .compat_3_brand = "Linux", 1083 .emul_path = "/compat/linux", 1084 .interp_path = "/lib/ld-linux.so.1", 1085 .sysvec = &elf_linux_sysvec, 1086 .interp_newpath = NULL, 1087 .brand_note = &linux32_brandnote, 1088 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 1089}; 1090 1091static Elf32_Brandinfo linux_glibc2brand = { 1092 .brand = ELFOSABI_LINUX, 1093 .machine = EM_386, 1094 .compat_3_brand = "Linux", 1095 .emul_path = "/compat/linux", 1096 .interp_path = "/lib/ld-linux.so.2", 1097 .sysvec = &elf_linux_sysvec, 1098 .interp_newpath = NULL, 1099 .brand_note = &linux32_brandnote, 1100 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 1101}; 1102 1103Elf32_Brandinfo *linux_brandlist[] = { 1104 &linux_brand, 1105 &linux_glibc2brand, 1106 NULL 1107}; 1108 1109static int 1110linux_elf_modevent(module_t mod, int type, void *data) 1111{ 1112 Elf32_Brandinfo **brandinfo; 1113 int error; 1114 struct linux_ioctl_handler **lihp; 1115 struct linux_device_handler **ldhp; 1116 1117 error = 0; 1118 1119 switch(type) { 1120 case MOD_LOAD: 1121 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 1122 ++brandinfo) 1123 if (elf32_insert_brand_entry(*brandinfo) < 0) 1124 error = EINVAL; 1125 if (error == 0) { 1126 SET_FOREACH(lihp, linux_ioctl_handler_set) 1127 linux_ioctl_register_handler(*lihp); 1128 SET_FOREACH(ldhp, linux_device_handler_set) 1129 linux_device_register_handler(*ldhp); 1130 mtx_init(&emul_lock, "emuldata lock", NULL, MTX_DEF); 1131 sx_init(&emul_shared_lock, "emuldata->shared lock"); 1132 LIST_INIT(&futex_list); 1133 mtx_init(&futex_mtx, "ftllk", NULL, MTX_DEF); 1134 linux_exit_tag = EVENTHANDLER_REGISTER(process_exit, 1135 linux_proc_exit, NULL, 1000); 1136 linux_exec_tag = EVENTHANDLER_REGISTER(process_exec, 1137 linux_proc_exec, NULL, 1000); 1138 linux_szplatform = roundup(strlen(linux_platform) + 1, 1139 sizeof(char *)); 1140 linux_osd_jail_register(); 1141 stclohz = (stathz ? stathz : hz); 1142 if (bootverbose) 1143 printf("Linux ELF exec handler installed\n"); 1144 } else 1145 printf("cannot insert Linux ELF brand handler\n"); 1146 break; 1147 case MOD_UNLOAD: 1148 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 1149 ++brandinfo) 1150 if (elf32_brand_inuse(*brandinfo)) 1151 error = EBUSY; 1152 if (error == 0) { 1153 for (brandinfo = &linux_brandlist[0]; 1154 *brandinfo != NULL; ++brandinfo) 1155 if (elf32_remove_brand_entry(*brandinfo) < 0) 1156 error = EINVAL; 1157 } 1158 if (error == 0) { 1159 SET_FOREACH(lihp, linux_ioctl_handler_set) 1160 linux_ioctl_unregister_handler(*lihp); 1161 SET_FOREACH(ldhp, linux_device_handler_set) 1162 linux_device_unregister_handler(*ldhp); 1163 mtx_destroy(&emul_lock); 1164 sx_destroy(&emul_shared_lock); 1165 mtx_destroy(&futex_mtx); 1166 EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag); 1167 EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag); 1168 linux_osd_jail_deregister(); 1169 if (bootverbose) 1170 printf("Linux ELF exec handler removed\n"); 1171 } else 1172 printf("Could not deinstall ELF interpreter entry\n"); 1173 break; 1174 default: 1175 return EOPNOTSUPP; 1176 } 1177 return error; 1178} 1179 1180static moduledata_t linux_elf_mod = { 1181 "linuxelf", 1182 linux_elf_modevent, 1183 0 1184}; 1185 1186DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 1187