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 Søren 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("$FreeBSD$"); 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; 251 252 arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS; 253 uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szplatform); 254 255 KASSERT(curthread->td_proc == imgp->proc, 256 ("unsafe elf_linux_fixup(), should be curproc")); 257 base = (Elf32_Addr *)*stack_base; 258 args = (Elf32_Auxargs *)imgp->auxargs; 259 pos = base + (imgp->args->argc + imgp->args->envc + 2); 260 261 AUXARGS_ENTRY_32(pos, LINUX_AT_HWCAP, cpu_feature); 262 263 /* 264 * Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0, 265 * as it has appeared in the 2.4.0-rc7 first time. 266 * Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK), 267 * glibc falls back to the hard-coded CLK_TCK value when aux entry 268 * is not present. 269 * Also see linux_times() implementation. 270 */ 271 if (linux_kernver(curthread) >= LINUX_KERNVER_2004000) 272 AUXARGS_ENTRY_32(pos, LINUX_AT_CLKTCK, stclohz); 273 AUXARGS_ENTRY_32(pos, AT_PHDR, args->phdr); 274 AUXARGS_ENTRY_32(pos, AT_PHENT, args->phent); 275 AUXARGS_ENTRY_32(pos, AT_PHNUM, args->phnum); 276 AUXARGS_ENTRY_32(pos, AT_PAGESZ, args->pagesz); 277 AUXARGS_ENTRY_32(pos, AT_FLAGS, args->flags); 278 AUXARGS_ENTRY_32(pos, AT_ENTRY, args->entry); 279 AUXARGS_ENTRY_32(pos, AT_BASE, args->base); 280 AUXARGS_ENTRY_32(pos, LINUX_AT_SECURE, 0); 281 AUXARGS_ENTRY_32(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); 282 AUXARGS_ENTRY_32(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); 283 AUXARGS_ENTRY_32(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); 284 AUXARGS_ENTRY_32(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); 285 AUXARGS_ENTRY_32(pos, LINUX_AT_PLATFORM, PTROUT(uplatform)); 286 if (args->execfd != -1) 287 AUXARGS_ENTRY_32(pos, AT_EXECFD, args->execfd); 288 AUXARGS_ENTRY_32(pos, AT_NULL, 0); 289 290 free(imgp->auxargs, M_TEMP); 291 imgp->auxargs = NULL; 292 293 base--; 294 suword32(base, (uint32_t)imgp->args->argc); 295 *stack_base = (register_t *)base; 296 return 0; 297} 298 299extern unsigned long linux_sznonrtsigcode; 300 301static void 302linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 303{ 304 struct thread *td = curthread; 305 struct proc *p = td->td_proc; 306 struct sigacts *psp; 307 struct trapframe *regs; 308 struct l_rt_sigframe *fp, frame; 309 int oonstack; 310 int sig; 311 int code; 312 313 sig = ksi->ksi_signo; 314 code = ksi->ksi_code; 315 PROC_LOCK_ASSERT(p, MA_OWNED); 316 psp = p->p_sigacts; 317 mtx_assert(&psp->ps_mtx, MA_OWNED); 318 regs = td->td_frame; 319 oonstack = sigonstack(regs->tf_rsp); 320 321#ifdef DEBUG 322 if (ldebug(rt_sendsig)) 323 printf(ARGS(rt_sendsig, "%p, %d, %p, %u"), 324 catcher, sig, (void*)mask, code); 325#endif 326 /* 327 * Allocate space for the signal handler context. 328 */ 329 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && 330 SIGISMEMBER(psp->ps_sigonstack, sig)) { 331 fp = (struct l_rt_sigframe *)(td->td_sigstk.ss_sp + 332 td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe)); 333 } else 334 fp = (struct l_rt_sigframe *)regs->tf_rsp - 1; 335 mtx_unlock(&psp->ps_mtx); 336 337 /* 338 * Build the argument list for the signal handler. 339 */ 340 if (p->p_sysent->sv_sigtbl) 341 if (sig <= p->p_sysent->sv_sigsize) 342 sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)]; 343 344 bzero(&frame, sizeof(frame)); 345 346 frame.sf_handler = PTROUT(catcher); 347 frame.sf_sig = sig; 348 frame.sf_siginfo = PTROUT(&fp->sf_si); 349 frame.sf_ucontext = PTROUT(&fp->sf_sc); 350 351 /* Fill in POSIX parts */ 352 ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig); 353 354 /* 355 * Build the signal context to be used by sigreturn. 356 */ 357 frame.sf_sc.uc_flags = 0; /* XXX ??? */ 358 frame.sf_sc.uc_link = 0; /* XXX ??? */ 359 360 frame.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp); 361 frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size; 362 frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) 363 ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE; 364 PROC_UNLOCK(p); 365 366 bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask); 367 368 frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__bits[0]; 369 frame.sf_sc.uc_mcontext.sc_edi = regs->tf_rdi; 370 frame.sf_sc.uc_mcontext.sc_esi = regs->tf_rsi; 371 frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_rbp; 372 frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_rbx; 373 frame.sf_sc.uc_mcontext.sc_edx = regs->tf_rdx; 374 frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_rcx; 375 frame.sf_sc.uc_mcontext.sc_eax = regs->tf_rax; 376 frame.sf_sc.uc_mcontext.sc_eip = regs->tf_rip; 377 frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs; 378 frame.sf_sc.uc_mcontext.sc_gs = regs->tf_gs; 379 frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs; 380 frame.sf_sc.uc_mcontext.sc_es = regs->tf_es; 381 frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds; 382 frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_rflags; 383 frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_rsp; 384 frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss; 385 frame.sf_sc.uc_mcontext.sc_err = regs->tf_err; 386 frame.sf_sc.uc_mcontext.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr; 387 frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code); 388 389#ifdef DEBUG 390 if (ldebug(rt_sendsig)) 391 printf(LMSG("rt_sendsig flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x"), 392 frame.sf_sc.uc_stack.ss_flags, td->td_sigstk.ss_sp, 393 td->td_sigstk.ss_size, frame.sf_sc.uc_mcontext.sc_mask); 394#endif 395 396 if (copyout(&frame, fp, sizeof(frame)) != 0) { 397 /* 398 * Process has trashed its stack; give it an illegal 399 * instruction to halt it in its tracks. 400 */ 401#ifdef DEBUG 402 if (ldebug(rt_sendsig)) 403 printf(LMSG("rt_sendsig: bad stack %p, oonstack=%x"), 404 fp, oonstack); 405#endif 406 PROC_LOCK(p); 407 sigexit(td, SIGILL); 408 } 409 410 /* 411 * Build context to run handler in. 412 */ 413 regs->tf_rsp = PTROUT(fp); 414 regs->tf_rip = p->p_sysent->sv_sigcode_base + linux_sznonrtsigcode; 415 regs->tf_rflags &= ~(PSL_T | PSL_D); 416 regs->tf_cs = _ucode32sel; 417 regs->tf_ss = _udatasel; 418 regs->tf_ds = _udatasel; 419 regs->tf_es = _udatasel; 420 regs->tf_fs = _ufssel; 421 regs->tf_gs = _ugssel; 422 regs->tf_flags = TF_HASSEGS; 423 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 424 PROC_LOCK(p); 425 mtx_lock(&psp->ps_mtx); 426} 427 428 429/* 430 * Send an interrupt to process. 431 * 432 * Stack is set up to allow sigcode stored 433 * in u. to call routine, followed by kcall 434 * to sigreturn routine below. After sigreturn 435 * resets the signal mask, the stack, and the 436 * frame pointer, it returns to the user 437 * specified pc, psl. 438 */ 439static void 440linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 441{ 442 struct thread *td = curthread; 443 struct proc *p = td->td_proc; 444 struct sigacts *psp; 445 struct trapframe *regs; 446 struct l_sigframe *fp, frame; 447 l_sigset_t lmask; 448 int oonstack, i; 449 int sig, code; 450 451 sig = ksi->ksi_signo; 452 code = ksi->ksi_code; 453 PROC_LOCK_ASSERT(p, MA_OWNED); 454 psp = p->p_sigacts; 455 mtx_assert(&psp->ps_mtx, MA_OWNED); 456 if (SIGISMEMBER(psp->ps_siginfo, sig)) { 457 /* Signal handler installed with SA_SIGINFO. */ 458 linux_rt_sendsig(catcher, ksi, mask); 459 return; 460 } 461 462 regs = td->td_frame; 463 oonstack = sigonstack(regs->tf_rsp); 464 465#ifdef DEBUG 466 if (ldebug(sendsig)) 467 printf(ARGS(sendsig, "%p, %d, %p, %u"), 468 catcher, sig, (void*)mask, code); 469#endif 470 471 /* 472 * Allocate space for the signal handler context. 473 */ 474 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && 475 SIGISMEMBER(psp->ps_sigonstack, sig)) { 476 fp = (struct l_sigframe *)(td->td_sigstk.ss_sp + 477 td->td_sigstk.ss_size - sizeof(struct l_sigframe)); 478 } else 479 fp = (struct l_sigframe *)regs->tf_rsp - 1; 480 mtx_unlock(&psp->ps_mtx); 481 PROC_UNLOCK(p); 482 483 /* 484 * Build the argument list for the signal handler. 485 */ 486 if (p->p_sysent->sv_sigtbl) 487 if (sig <= p->p_sysent->sv_sigsize) 488 sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)]; 489 490 bzero(&frame, sizeof(frame)); 491 492 frame.sf_handler = PTROUT(catcher); 493 frame.sf_sig = sig; 494 495 bsd_to_linux_sigset(mask, &lmask); 496 497 /* 498 * Build the signal context to be used by sigreturn. 499 */ 500 frame.sf_sc.sc_mask = lmask.__bits[0]; 501 frame.sf_sc.sc_gs = regs->tf_gs; 502 frame.sf_sc.sc_fs = regs->tf_fs; 503 frame.sf_sc.sc_es = regs->tf_es; 504 frame.sf_sc.sc_ds = regs->tf_ds; 505 frame.sf_sc.sc_edi = regs->tf_rdi; 506 frame.sf_sc.sc_esi = regs->tf_rsi; 507 frame.sf_sc.sc_ebp = regs->tf_rbp; 508 frame.sf_sc.sc_ebx = regs->tf_rbx; 509 frame.sf_sc.sc_edx = regs->tf_rdx; 510 frame.sf_sc.sc_ecx = regs->tf_rcx; 511 frame.sf_sc.sc_eax = regs->tf_rax; 512 frame.sf_sc.sc_eip = regs->tf_rip; 513 frame.sf_sc.sc_cs = regs->tf_cs; 514 frame.sf_sc.sc_eflags = regs->tf_rflags; 515 frame.sf_sc.sc_esp_at_signal = regs->tf_rsp; 516 frame.sf_sc.sc_ss = regs->tf_ss; 517 frame.sf_sc.sc_err = regs->tf_err; 518 frame.sf_sc.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr; 519 frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(code); 520 521 for (i = 0; i < (LINUX_NSIG_WORDS-1); i++) 522 frame.sf_extramask[i] = lmask.__bits[i+1]; 523 524 if (copyout(&frame, fp, sizeof(frame)) != 0) { 525 /* 526 * Process has trashed its stack; give it an illegal 527 * instruction to halt it in its tracks. 528 */ 529 PROC_LOCK(p); 530 sigexit(td, SIGILL); 531 } 532 533 /* 534 * Build context to run handler in. 535 */ 536 regs->tf_rsp = PTROUT(fp); 537 regs->tf_rip = p->p_sysent->sv_sigcode_base; 538 regs->tf_rflags &= ~(PSL_T | PSL_D); 539 regs->tf_cs = _ucode32sel; 540 regs->tf_ss = _udatasel; 541 regs->tf_ds = _udatasel; 542 regs->tf_es = _udatasel; 543 regs->tf_fs = _ufssel; 544 regs->tf_gs = _ugssel; 545 regs->tf_flags = TF_HASSEGS; 546 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 547 PROC_LOCK(p); 548 mtx_lock(&psp->ps_mtx); 549} 550 551/* 552 * System call to cleanup state after a signal 553 * has been taken. Reset signal mask and 554 * stack state from context left by sendsig (above). 555 * Return to previous pc and psl as specified by 556 * context left by sendsig. Check carefully to 557 * make sure that the user has not modified the 558 * psl to gain improper privileges or to cause 559 * a machine fault. 560 */ 561int 562linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args) 563{ 564 struct l_sigframe frame; 565 struct trapframe *regs; 566 sigset_t bmask; 567 l_sigset_t lmask; 568 int eflags, i; 569 ksiginfo_t ksi; 570 571 regs = td->td_frame; 572 573#ifdef DEBUG 574 if (ldebug(sigreturn)) 575 printf(ARGS(sigreturn, "%p"), (void *)args->sfp); 576#endif 577 /* 578 * The trampoline code hands us the sigframe. 579 * It is unsafe to keep track of it ourselves, in the event that a 580 * program jumps out of a signal handler. 581 */ 582 if (copyin(args->sfp, &frame, sizeof(frame)) != 0) 583 return (EFAULT); 584 585 /* 586 * Check for security violations. 587 */ 588#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) 589 eflags = frame.sf_sc.sc_eflags; 590 if (!EFLAGS_SECURE(eflags, regs->tf_rflags)) 591 return(EINVAL); 592 593 /* 594 * Don't allow users to load a valid privileged %cs. Let the 595 * hardware check for invalid selectors, excess privilege in 596 * other selectors, invalid %eip's and invalid %esp's. 597 */ 598#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) 599 if (!CS_SECURE(frame.sf_sc.sc_cs)) { 600 ksiginfo_init_trap(&ksi); 601 ksi.ksi_signo = SIGBUS; 602 ksi.ksi_code = BUS_OBJERR; 603 ksi.ksi_trapno = T_PROTFLT; 604 ksi.ksi_addr = (void *)regs->tf_rip; 605 trapsignal(td, &ksi); 606 return(EINVAL); 607 } 608 609 lmask.__bits[0] = frame.sf_sc.sc_mask; 610 for (i = 0; i < (LINUX_NSIG_WORDS-1); i++) 611 lmask.__bits[i+1] = frame.sf_extramask[i]; 612 linux_to_bsd_sigset(&lmask, &bmask); 613 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); 614 615 /* 616 * Restore signal context. 617 */ 618 regs->tf_rdi = frame.sf_sc.sc_edi; 619 regs->tf_rsi = frame.sf_sc.sc_esi; 620 regs->tf_rbp = frame.sf_sc.sc_ebp; 621 regs->tf_rbx = frame.sf_sc.sc_ebx; 622 regs->tf_rdx = frame.sf_sc.sc_edx; 623 regs->tf_rcx = frame.sf_sc.sc_ecx; 624 regs->tf_rax = frame.sf_sc.sc_eax; 625 regs->tf_rip = frame.sf_sc.sc_eip; 626 regs->tf_cs = frame.sf_sc.sc_cs; 627 regs->tf_ds = frame.sf_sc.sc_ds; 628 regs->tf_es = frame.sf_sc.sc_es; 629 regs->tf_fs = frame.sf_sc.sc_fs; 630 regs->tf_gs = frame.sf_sc.sc_gs; 631 regs->tf_rflags = eflags; 632 regs->tf_rsp = frame.sf_sc.sc_esp_at_signal; 633 regs->tf_ss = frame.sf_sc.sc_ss; 634 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 635 636 return (EJUSTRETURN); 637} 638 639/* 640 * System call to cleanup state after a signal 641 * has been taken. Reset signal mask and 642 * stack state from context left by rt_sendsig (above). 643 * Return to previous pc and psl as specified by 644 * context left by sendsig. Check carefully to 645 * make sure that the user has not modified the 646 * psl to gain improper privileges or to cause 647 * a machine fault. 648 */ 649int 650linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 651{ 652 struct l_ucontext uc; 653 struct l_sigcontext *context; 654 sigset_t bmask; 655 l_stack_t *lss; 656 stack_t ss; 657 struct trapframe *regs; 658 int eflags; 659 ksiginfo_t ksi; 660 661 regs = td->td_frame; 662 663#ifdef DEBUG 664 if (ldebug(rt_sigreturn)) 665 printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp); 666#endif 667 /* 668 * The trampoline code hands us the ucontext. 669 * It is unsafe to keep track of it ourselves, in the event that a 670 * program jumps out of a signal handler. 671 */ 672 if (copyin(args->ucp, &uc, sizeof(uc)) != 0) 673 return (EFAULT); 674 675 context = &uc.uc_mcontext; 676 677 /* 678 * Check for security violations. 679 */ 680#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) 681 eflags = context->sc_eflags; 682 if (!EFLAGS_SECURE(eflags, regs->tf_rflags)) 683 return(EINVAL); 684 685 /* 686 * Don't allow users to load a valid privileged %cs. Let the 687 * hardware check for invalid selectors, excess privilege in 688 * other selectors, invalid %eip's and invalid %esp's. 689 */ 690#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) 691 if (!CS_SECURE(context->sc_cs)) { 692 ksiginfo_init_trap(&ksi); 693 ksi.ksi_signo = SIGBUS; 694 ksi.ksi_code = BUS_OBJERR; 695 ksi.ksi_trapno = T_PROTFLT; 696 ksi.ksi_addr = (void *)regs->tf_rip; 697 trapsignal(td, &ksi); 698 return(EINVAL); 699 } 700 701 linux_to_bsd_sigset(&uc.uc_sigmask, &bmask); 702 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); 703 704 /* 705 * Restore signal context 706 */ 707 regs->tf_gs = context->sc_gs; 708 regs->tf_fs = context->sc_fs; 709 regs->tf_es = context->sc_es; 710 regs->tf_ds = context->sc_ds; 711 regs->tf_rdi = context->sc_edi; 712 regs->tf_rsi = context->sc_esi; 713 regs->tf_rbp = context->sc_ebp; 714 regs->tf_rbx = context->sc_ebx; 715 regs->tf_rdx = context->sc_edx; 716 regs->tf_rcx = context->sc_ecx; 717 regs->tf_rax = context->sc_eax; 718 regs->tf_rip = context->sc_eip; 719 regs->tf_cs = context->sc_cs; 720 regs->tf_rflags = eflags; 721 regs->tf_rsp = context->sc_esp_at_signal; 722 regs->tf_ss = context->sc_ss; 723 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 724 725 /* 726 * call sigaltstack & ignore results.. 727 */ 728 lss = &uc.uc_stack; 729 ss.ss_sp = PTRIN(lss->ss_sp); 730 ss.ss_size = lss->ss_size; 731 ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags); 732 733#ifdef DEBUG 734 if (ldebug(rt_sigreturn)) 735 printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x"), 736 ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask); 737#endif 738 (void)kern_sigaltstack(td, &ss, NULL); 739 740 return (EJUSTRETURN); 741} 742 743static int 744linux32_fetch_syscall_args(struct thread *td, struct syscall_args *sa) 745{ 746 struct proc *p; 747 struct trapframe *frame; 748 749 p = td->td_proc; 750 frame = td->td_frame; 751 752 sa->args[0] = frame->tf_rbx; 753 sa->args[1] = frame->tf_rcx; 754 sa->args[2] = frame->tf_rdx; 755 sa->args[3] = frame->tf_rsi; 756 sa->args[4] = frame->tf_rdi; 757 sa->args[5] = frame->tf_rbp; /* Unconfirmed */ 758 sa->code = frame->tf_rax; 759 760 if (sa->code >= p->p_sysent->sv_size) 761 sa->callp = &p->p_sysent->sv_table[0]; 762 else 763 sa->callp = &p->p_sysent->sv_table[sa->code]; 764 sa->narg = sa->callp->sy_narg; 765 766 td->td_retval[0] = 0; 767 td->td_retval[1] = frame->tf_rdx; 768 769 return (0); 770} 771 772/* 773 * If a linux binary is exec'ing something, try this image activator 774 * first. We override standard shell script execution in order to 775 * be able to modify the interpreter path. We only do this if a linux 776 * binary is doing the exec, so we do not create an EXEC module for it. 777 */ 778static int exec_linux_imgact_try(struct image_params *iparams); 779 780static int 781exec_linux_imgact_try(struct image_params *imgp) 782{ 783 const char *head = (const char *)imgp->image_header; 784 char *rpath; 785 int error = -1; 786 787 /* 788 * The interpreter for shell scripts run from a linux binary needs 789 * to be located in /compat/linux if possible in order to recursively 790 * maintain linux path emulation. 791 */ 792 if (((const short *)head)[0] == SHELLMAGIC) { 793 /* 794 * Run our normal shell image activator. If it succeeds attempt 795 * to use the alternate path for the interpreter. If an 796 * alternate * path is found, use our stringspace to store it. 797 */ 798 if ((error = exec_shell_imgact(imgp)) == 0) { 799 linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc), 800 imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0, 801 AT_FDCWD); 802 if (rpath != NULL) 803 imgp->args->fname_buf = 804 imgp->interpreter_name = rpath; 805 } 806 } 807 return (error); 808} 809 810/* 811 * Clear registers on exec 812 * XXX copied from ia32_signal.c. 813 */ 814static void 815exec_linux_setregs(struct thread *td, struct image_params *imgp, u_long stack) 816{ 817 struct trapframe *regs = td->td_frame; 818 struct pcb *pcb = td->td_pcb; 819 820 mtx_lock(&dt_lock); 821 if (td->td_proc->p_md.md_ldt != NULL) 822 user_ldt_free(td); 823 else 824 mtx_unlock(&dt_lock); 825 826 critical_enter(); 827 wrmsr(MSR_FSBASE, 0); 828 wrmsr(MSR_KGSBASE, 0); /* User value while we're in the kernel */ 829 pcb->pcb_fsbase = 0; 830 pcb->pcb_gsbase = 0; 831 critical_exit(); 832 pcb->pcb_initial_fpucw = __LINUX_NPXCW__; 833 834 bzero((char *)regs, sizeof(struct trapframe)); 835 regs->tf_rip = imgp->entry_addr; 836 regs->tf_rsp = stack; 837 regs->tf_rflags = PSL_USER | (regs->tf_rflags & PSL_T); 838 regs->tf_gs = _ugssel; 839 regs->tf_fs = _ufssel; 840 regs->tf_es = _udatasel; 841 regs->tf_ds = _udatasel; 842 regs->tf_ss = _udatasel; 843 regs->tf_flags = TF_HASSEGS; 844 regs->tf_cs = _ucode32sel; 845 regs->tf_rbx = imgp->ps_strings; 846 847 fpstate_drop(td); 848 849 /* Do full restore on return so that we can change to a different %cs */ 850 set_pcb_flags(pcb, PCB_32BIT | PCB_FULL_IRET); 851 td->td_retval[1] = 0; 852} 853 854/* 855 * XXX copied from ia32_sysvec.c. 856 */ 857static register_t * 858linux_copyout_strings(struct image_params *imgp) 859{ 860 int argc, envc; 861 u_int32_t *vectp; 862 char *stringp, *destp; 863 u_int32_t *stack_base; 864 struct linux32_ps_strings *arginfo; 865 866 /* 867 * Calculate string base and vector table pointers. 868 * Also deal with signal trampoline code for this exec type. 869 */ 870 arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS; 871 destp = (caddr_t)arginfo - SPARE_USRSPACE - linux_szplatform - 872 roundup((ARG_MAX - imgp->args->stringspace), 873 sizeof(char *)); 874 875 /* 876 * Install LINUX_PLATFORM 877 */ 878 copyout(linux_platform, ((caddr_t)arginfo - linux_szplatform), 879 linux_szplatform); 880 881 /* 882 * If we have a valid auxargs ptr, prepare some room 883 * on the stack. 884 */ 885 if (imgp->auxargs) { 886 /* 887 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for 888 * lower compatibility. 889 */ 890 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size : 891 (LINUX_AT_COUNT * 2); 892 /* 893 * The '+ 2' is for the null pointers at the end of each of 894 * the arg and env vector sets,and imgp->auxarg_size is room 895 * for argument of Runtime loader. 896 */ 897 vectp = (u_int32_t *) (destp - (imgp->args->argc + 898 imgp->args->envc + 2 + imgp->auxarg_size) * 899 sizeof(u_int32_t)); 900 901 } else 902 /* 903 * The '+ 2' is for the null pointers at the end of each of 904 * the arg and env vector sets 905 */ 906 vectp = (u_int32_t *)(destp - (imgp->args->argc + 907 imgp->args->envc + 2) * sizeof(u_int32_t)); 908 909 /* 910 * vectp also becomes our initial stack base 911 */ 912 stack_base = vectp; 913 914 stringp = imgp->args->begin_argv; 915 argc = imgp->args->argc; 916 envc = imgp->args->envc; 917 /* 918 * Copy out strings - arguments and environment. 919 */ 920 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace); 921 922 /* 923 * Fill in "ps_strings" struct for ps, w, etc. 924 */ 925 suword32(&arginfo->ps_argvstr, (uint32_t)(intptr_t)vectp); 926 suword32(&arginfo->ps_nargvstr, argc); 927 928 /* 929 * Fill in argument portion of vector table. 930 */ 931 for (; argc > 0; --argc) { 932 suword32(vectp++, (uint32_t)(intptr_t)destp); 933 while (*stringp++ != 0) 934 destp++; 935 destp++; 936 } 937 938 /* a null vector table pointer separates the argp's from the envp's */ 939 suword32(vectp++, 0); 940 941 suword32(&arginfo->ps_envstr, (uint32_t)(intptr_t)vectp); 942 suword32(&arginfo->ps_nenvstr, envc); 943 944 /* 945 * Fill in environment portion of vector table. 946 */ 947 for (; envc > 0; --envc) { 948 suword32(vectp++, (uint32_t)(intptr_t)destp); 949 while (*stringp++ != 0) 950 destp++; 951 destp++; 952 } 953 954 /* end of vector table is a null pointer */ 955 suword32(vectp, 0); 956 957 return ((register_t *)stack_base); 958} 959 960static SYSCTL_NODE(_compat, OID_AUTO, linux32, CTLFLAG_RW, 0, 961 "32-bit Linux emulation"); 962 963static u_long linux32_maxdsiz = LINUX32_MAXDSIZ; 964SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxdsiz, CTLFLAG_RW, 965 &linux32_maxdsiz, 0, ""); 966static u_long linux32_maxssiz = LINUX32_MAXSSIZ; 967SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxssiz, CTLFLAG_RW, 968 &linux32_maxssiz, 0, ""); 969static u_long linux32_maxvmem = LINUX32_MAXVMEM; 970SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxvmem, CTLFLAG_RW, 971 &linux32_maxvmem, 0, ""); 972 973static void 974linux32_fixlimit(struct rlimit *rl, int which) 975{ 976 977 switch (which) { 978 case RLIMIT_DATA: 979 if (linux32_maxdsiz != 0) { 980 if (rl->rlim_cur > linux32_maxdsiz) 981 rl->rlim_cur = linux32_maxdsiz; 982 if (rl->rlim_max > linux32_maxdsiz) 983 rl->rlim_max = linux32_maxdsiz; 984 } 985 break; 986 case RLIMIT_STACK: 987 if (linux32_maxssiz != 0) { 988 if (rl->rlim_cur > linux32_maxssiz) 989 rl->rlim_cur = linux32_maxssiz; 990 if (rl->rlim_max > linux32_maxssiz) 991 rl->rlim_max = linux32_maxssiz; 992 } 993 break; 994 case RLIMIT_VMEM: 995 if (linux32_maxvmem != 0) { 996 if (rl->rlim_cur > linux32_maxvmem) 997 rl->rlim_cur = linux32_maxvmem; 998 if (rl->rlim_max > linux32_maxvmem) 999 rl->rlim_max = linux32_maxvmem; 1000 } 1001 break; 1002 } 1003} 1004 1005struct sysentvec elf_linux_sysvec = { 1006 .sv_size = LINUX_SYS_MAXSYSCALL, 1007 .sv_table = linux_sysent, 1008 .sv_mask = 0, 1009 .sv_sigsize = LINUX_SIGTBLSZ, 1010 .sv_sigtbl = bsd_to_linux_signal, 1011 .sv_errsize = ELAST + 1, 1012 .sv_errtbl = bsd_to_linux_errno, 1013 .sv_transtrap = translate_traps, 1014 .sv_fixup = elf_linux_fixup, 1015 .sv_sendsig = linux_sendsig, 1016 .sv_sigcode = linux_sigcode, 1017 .sv_szsigcode = &linux_szsigcode, 1018 .sv_prepsyscall = NULL, 1019 .sv_name = "Linux ELF32", 1020 .sv_coredump = elf32_coredump, 1021 .sv_imgact_try = exec_linux_imgact_try, 1022 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 1023 .sv_pagesize = PAGE_SIZE, 1024 .sv_minuser = VM_MIN_ADDRESS, 1025 .sv_maxuser = LINUX32_MAXUSER, 1026 .sv_usrstack = LINUX32_USRSTACK, 1027 .sv_psstrings = LINUX32_PS_STRINGS, 1028 .sv_stackprot = VM_PROT_ALL, 1029 .sv_copyout_strings = linux_copyout_strings, 1030 .sv_setregs = exec_linux_setregs, 1031 .sv_fixlimit = linux32_fixlimit, 1032 .sv_maxssiz = &linux32_maxssiz, 1033 .sv_flags = SV_ABI_LINUX | SV_ILP32 | SV_IA32 | SV_SHP, 1034 .sv_set_syscall_retval = cpu_set_syscall_retval, 1035 .sv_fetch_syscall_args = linux32_fetch_syscall_args, 1036 .sv_syscallnames = NULL, 1037 .sv_shared_page_base = LINUX32_SHAREDPAGE, 1038 .sv_shared_page_len = PAGE_SIZE, 1039 .sv_schedtail = linux_schedtail, 1040}; 1041INIT_SYSENTVEC(elf_sysvec, &elf_linux_sysvec); 1042 1043static char GNU_ABI_VENDOR[] = "GNU"; 1044static int GNULINUX_ABI_DESC = 0; 1045 1046static boolean_t 1047linux32_trans_osrel(const Elf_Note *note, int32_t *osrel) 1048{ 1049 const Elf32_Word *desc; 1050 uintptr_t p; 1051 1052 p = (uintptr_t)(note + 1); 1053 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 1054 1055 desc = (const Elf32_Word *)p; 1056 if (desc[0] != GNULINUX_ABI_DESC) 1057 return (FALSE); 1058 1059 /* 1060 * For linux we encode osrel as follows (see linux_mib.c): 1061 * VVVMMMIII (version, major, minor), see linux_mib.c. 1062 */ 1063 *osrel = desc[1] * 1000000 + desc[2] * 1000 + desc[3]; 1064 1065 return (TRUE); 1066} 1067 1068static Elf_Brandnote linux32_brandnote = { 1069 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 1070 .hdr.n_descsz = 16, /* XXX at least 16 */ 1071 .hdr.n_type = 1, 1072 .vendor = GNU_ABI_VENDOR, 1073 .flags = BN_TRANSLATE_OSREL, 1074 .trans_osrel = linux32_trans_osrel 1075}; 1076 1077static Elf32_Brandinfo linux_brand = { 1078 .brand = ELFOSABI_LINUX, 1079 .machine = EM_386, 1080 .compat_3_brand = "Linux", 1081 .emul_path = "/compat/linux", 1082 .interp_path = "/lib/ld-linux.so.1", 1083 .sysvec = &elf_linux_sysvec, 1084 .interp_newpath = NULL, 1085 .brand_note = &linux32_brandnote, 1086 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 1087}; 1088 1089static Elf32_Brandinfo linux_glibc2brand = { 1090 .brand = ELFOSABI_LINUX, 1091 .machine = EM_386, 1092 .compat_3_brand = "Linux", 1093 .emul_path = "/compat/linux", 1094 .interp_path = "/lib/ld-linux.so.2", 1095 .sysvec = &elf_linux_sysvec, 1096 .interp_newpath = NULL, 1097 .brand_note = &linux32_brandnote, 1098 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 1099}; 1100 1101Elf32_Brandinfo *linux_brandlist[] = { 1102 &linux_brand, 1103 &linux_glibc2brand, 1104 NULL 1105}; 1106 1107static int 1108linux_elf_modevent(module_t mod, int type, void *data) 1109{ 1110 Elf32_Brandinfo **brandinfo; 1111 int error; 1112 struct linux_ioctl_handler **lihp; 1113 struct linux_device_handler **ldhp; 1114 1115 error = 0; 1116 1117 switch(type) { 1118 case MOD_LOAD: 1119 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 1120 ++brandinfo) 1121 if (elf32_insert_brand_entry(*brandinfo) < 0) 1122 error = EINVAL; 1123 if (error == 0) { 1124 SET_FOREACH(lihp, linux_ioctl_handler_set) 1125 linux_ioctl_register_handler(*lihp); 1126 SET_FOREACH(ldhp, linux_device_handler_set) 1127 linux_device_register_handler(*ldhp); 1128 mtx_init(&emul_lock, "emuldata lock", NULL, MTX_DEF); 1129 sx_init(&emul_shared_lock, "emuldata->shared lock"); 1130 LIST_INIT(&futex_list); 1131 mtx_init(&futex_mtx, "ftllk", NULL, MTX_DEF); 1132 linux_exit_tag = EVENTHANDLER_REGISTER(process_exit, 1133 linux_proc_exit, NULL, 1000); 1134 linux_exec_tag = EVENTHANDLER_REGISTER(process_exec, 1135 linux_proc_exec, NULL, 1000); 1136 linux_szplatform = roundup(strlen(linux_platform) + 1, 1137 sizeof(char *)); 1138 linux_osd_jail_register(); 1139 stclohz = (stathz ? stathz : hz); 1140 if (bootverbose) 1141 printf("Linux ELF exec handler installed\n"); 1142 } else 1143 printf("cannot insert Linux ELF brand handler\n"); 1144 break; 1145 case MOD_UNLOAD: 1146 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 1147 ++brandinfo) 1148 if (elf32_brand_inuse(*brandinfo)) 1149 error = EBUSY; 1150 if (error == 0) { 1151 for (brandinfo = &linux_brandlist[0]; 1152 *brandinfo != NULL; ++brandinfo) 1153 if (elf32_remove_brand_entry(*brandinfo) < 0) 1154 error = EINVAL; 1155 } 1156 if (error == 0) { 1157 SET_FOREACH(lihp, linux_ioctl_handler_set) 1158 linux_ioctl_unregister_handler(*lihp); 1159 SET_FOREACH(ldhp, linux_device_handler_set) 1160 linux_device_unregister_handler(*ldhp); 1161 mtx_destroy(&emul_lock); 1162 sx_destroy(&emul_shared_lock); 1163 mtx_destroy(&futex_mtx); 1164 EVENTHANDLER_DEREGISTER(process_exit, linux_exit_tag); 1165 EVENTHANDLER_DEREGISTER(process_exec, linux_exec_tag); 1166 linux_osd_jail_deregister(); 1167 if (bootverbose) 1168 printf("Linux ELF exec handler removed\n"); 1169 } else 1170 printf("Could not deinstall ELF interpreter entry\n"); 1171 break; 1172 default: 1173 return EOPNOTSUPP; 1174 } 1175 return error; 1176} 1177 1178static moduledata_t linux_elf_mod = { 1179 "linuxelf", 1180 linux_elf_modevent, 1181 0 1182}; 1183 1184DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 1185