1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2004 Tim J. Robbins 5 * Copyright (c) 2003 Peter Wemm 6 * Copyright (c) 2002 Doug Rabson 7 * Copyright (c) 1998-1999 Andrew Gallatin 8 * Copyright (c) 1994-1996 Søren Schmidt 9 * All rights reserved. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer 16 * in this position and unchanged. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. The name of the author may not be used to endorse or promote products 21 * derived from this software without specific prior written permission 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 24 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 25 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 26 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 27 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 28 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 32 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 33 */ 34 35 #include "opt_compat.h" 36 37 #include <sys/cdefs.h> 38 __FBSDID("$FreeBSD$"); 39 40 #ifndef COMPAT_FREEBSD32 41 #error "Unable to compile Linux-emulator due to missing COMPAT_FREEBSD32 option!" 42 #endif 43 44 #define __ELF_WORD_SIZE 32 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/exec.h> 49 #include <sys/fcntl.h> 50 #include <sys/imgact.h> 51 #include <sys/imgact_elf.h> 52 #include <sys/kernel.h> 53 #include <sys/lock.h> 54 #include <sys/malloc.h> 55 #include <sys/module.h> 56 #include <sys/mutex.h> 57 #include <sys/proc.h> 58 #include <sys/resourcevar.h> 59 #include <sys/stddef.h> 60 #include <sys/signalvar.h> 61 #include <sys/syscallsubr.h> 62 #include <sys/sysctl.h> 63 #include <sys/sysent.h> 64 #include <sys/sysproto.h> 65 #include <sys/vnode.h> 66 #include <sys/eventhandler.h> 67 68 #include <vm/vm.h> 69 #include <vm/pmap.h> 70 #include <vm/vm_extern.h> 71 #include <vm/vm_map.h> 72 #include <vm/vm_object.h> 73 #include <vm/vm_page.h> 74 #include <vm/vm_param.h> 75 76 #include <machine/cpu.h> 77 #include <machine/md_var.h> 78 #include <machine/pcb.h> 79 #include <machine/specialreg.h> 80 #include <machine/trap.h> 81 82 #include <x86/linux/linux_x86.h> 83 #include <amd64/linux32/linux.h> 84 #include <amd64/linux32/linux32_proto.h> 85 #include <compat/linux/linux_emul.h> 86 #include <compat/linux/linux_fork.h> 87 #include <compat/linux/linux_ioctl.h> 88 #include <compat/linux/linux_mib.h> 89 #include <compat/linux/linux_misc.h> 90 #include <compat/linux/linux_signal.h> 91 #include <compat/linux/linux_util.h> 92 #include <compat/linux/linux_vdso.h> 93 94 MODULE_VERSION(linux, 1); 95 96 #define LINUX32_MAXUSER ((1ul << 32) - PAGE_SIZE) 97 #define LINUX32_VDSOPAGE_SIZE PAGE_SIZE * 2 98 #define LINUX32_VDSOPAGE (LINUX32_MAXUSER - LINUX32_VDSOPAGE_SIZE) 99 #define LINUX32_SHAREDPAGE (LINUX32_VDSOPAGE - PAGE_SIZE) 100 /* 101 * PAGE_SIZE - the size 102 * of the native SHAREDPAGE 103 */ 104 #define LINUX32_USRSTACK LINUX32_SHAREDPAGE 105 106 static int linux_szsigcode; 107 static vm_object_t linux_vdso_obj; 108 static char *linux_vdso_mapping; 109 extern char _binary_linux32_vdso_so_o_start; 110 extern char _binary_linux32_vdso_so_o_end; 111 static vm_offset_t linux_vdso_base; 112 113 extern struct sysent linux32_sysent[LINUX32_SYS_MAXSYSCALL]; 114 115 SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); 116 117 static int linux_fixup_elf(uintptr_t *stack_base, 118 struct image_params *iparams); 119 static int linux_copyout_strings(struct image_params *imgp, 120 uintptr_t *stack_base); 121 static void linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask); 122 static void linux_exec_setregs(struct thread *td, 123 struct image_params *imgp, uintptr_t stack); 124 static void linux_exec_sysvec_init(void *param); 125 static int linux_on_exec_vmspace(struct proc *p, 126 struct image_params *imgp); 127 static void linux32_fixlimit(struct rlimit *rl, int which); 128 static bool linux32_trans_osrel(const Elf_Note *note, int32_t *osrel); 129 static void linux_vdso_install(const void *param); 130 static void linux_vdso_deinstall(const void *param); 131 static void linux_vdso_reloc(char *mapping, Elf_Addr offset); 132 static void linux32_set_fork_retval(struct thread *td); 133 static void linux32_set_syscall_retval(struct thread *td, int error); 134 135 #define LINUX_T_UNKNOWN 255 136 static int _bsd_to_linux_trapcode[] = { 137 LINUX_T_UNKNOWN, /* 0 */ 138 6, /* 1 T_PRIVINFLT */ 139 LINUX_T_UNKNOWN, /* 2 */ 140 3, /* 3 T_BPTFLT */ 141 LINUX_T_UNKNOWN, /* 4 */ 142 LINUX_T_UNKNOWN, /* 5 */ 143 16, /* 6 T_ARITHTRAP */ 144 254, /* 7 T_ASTFLT */ 145 LINUX_T_UNKNOWN, /* 8 */ 146 13, /* 9 T_PROTFLT */ 147 1, /* 10 T_TRCTRAP */ 148 LINUX_T_UNKNOWN, /* 11 */ 149 14, /* 12 T_PAGEFLT */ 150 LINUX_T_UNKNOWN, /* 13 */ 151 17, /* 14 T_ALIGNFLT */ 152 LINUX_T_UNKNOWN, /* 15 */ 153 LINUX_T_UNKNOWN, /* 16 */ 154 LINUX_T_UNKNOWN, /* 17 */ 155 0, /* 18 T_DIVIDE */ 156 2, /* 19 T_NMI */ 157 4, /* 20 T_OFLOW */ 158 5, /* 21 T_BOUND */ 159 7, /* 22 T_DNA */ 160 8, /* 23 T_DOUBLEFLT */ 161 9, /* 24 T_FPOPFLT */ 162 10, /* 25 T_TSSFLT */ 163 11, /* 26 T_SEGNPFLT */ 164 12, /* 27 T_STKFLT */ 165 18, /* 28 T_MCHK */ 166 19, /* 29 T_XMMFLT */ 167 15 /* 30 T_RESERVED */ 168 }; 169 #define bsd_to_linux_trapcode(code) \ 170 ((code)<nitems(_bsd_to_linux_trapcode)? \ 171 _bsd_to_linux_trapcode[(code)]: \ 172 LINUX_T_UNKNOWN) 173 174 struct linux32_ps_strings { 175 u_int32_t ps_argvstr; /* first of 0 or more argument strings */ 176 u_int ps_nargvstr; /* the number of argument strings */ 177 u_int32_t ps_envstr; /* first of 0 or more environment strings */ 178 u_int ps_nenvstr; /* the number of environment strings */ 179 }; 180 #define LINUX32_PS_STRINGS (LINUX32_USRSTACK - \ 181 sizeof(struct linux32_ps_strings)) 182 183 LINUX_VDSO_SYM_INTPTR(__kernel_vsyscall); 184 LINUX_VDSO_SYM_INTPTR(__kernel_sigreturn); 185 LINUX_VDSO_SYM_INTPTR(__kernel_rt_sigreturn); 186 LINUX_VDSO_SYM_INTPTR(kern_timekeep_base); 187 LINUX_VDSO_SYM_INTPTR(kern_tsc_selector); 188 LINUX_VDSO_SYM_CHAR(linux_platform); 189 190 /* 191 * If FreeBSD & Linux have a difference of opinion about what a trap 192 * means, deal with it here. 193 * 194 * MPSAFE 195 */ 196 static int 197 linux_translate_traps(int signal, int trap_code) 198 { 199 if (signal != SIGBUS) 200 return (signal); 201 switch (trap_code) { 202 case T_PROTFLT: 203 case T_TSSFLT: 204 case T_DOUBLEFLT: 205 case T_PAGEFLT: 206 return (SIGSEGV); 207 default: 208 return (signal); 209 } 210 } 211 212 static int 213 linux_copyout_auxargs(struct image_params *imgp, uintptr_t base) 214 { 215 Elf32_Auxargs *args; 216 Elf32_Auxinfo *argarray, *pos; 217 int error, issetugid; 218 219 args = (Elf32_Auxargs *)imgp->auxargs; 220 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, 221 M_WAITOK | M_ZERO); 222 223 issetugid = imgp->proc->p_flag & P_SUGID ? 1 : 0; 224 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO, __kernel_vsyscall); 225 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, linux_vdso_base); 226 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature); 227 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); 228 229 /* 230 * Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0, 231 * as it has appeared in the 2.4.0-rc7 first time. 232 * Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK), 233 * glibc falls back to the hard-coded CLK_TCK value when aux entry 234 * is not present. 235 * Also see linux_times() implementation. 236 */ 237 if (linux_kernver(curthread) >= LINUX_KERNVER_2004000) 238 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); 239 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); 240 AUXARGS_ENTRY(pos, AT_PHENT, args->phent); 241 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); 242 AUXARGS_ENTRY(pos, AT_BASE, args->base); 243 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); 244 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); 245 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); 246 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); 247 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); 248 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); 249 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); 250 AUXARGS_ENTRY(pos, LINUX_AT_RANDOM, PTROUT(imgp->canary)); 251 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, 0); 252 if (imgp->execpathp != 0) 253 AUXARGS_ENTRY(pos, LINUX_AT_EXECFN, PTROUT(imgp->execpathp)); 254 if (args->execfd != -1) 255 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); 256 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); 257 AUXARGS_ENTRY(pos, AT_NULL, 0); 258 259 free(imgp->auxargs, M_TEMP); 260 imgp->auxargs = NULL; 261 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); 262 263 error = copyout(argarray, (void *)base, 264 sizeof(*argarray) * LINUX_AT_COUNT); 265 free(argarray, M_TEMP); 266 return (error); 267 } 268 269 static int 270 linux_fixup_elf(uintptr_t *stack_base, struct image_params *imgp) 271 { 272 Elf32_Addr *base; 273 274 base = (Elf32_Addr *)*stack_base; 275 base--; 276 if (suword32(base, (uint32_t)imgp->args->argc) == -1) 277 return (EFAULT); 278 *stack_base = (uintptr_t)base; 279 return (0); 280 } 281 282 static void 283 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 284 { 285 struct thread *td = curthread; 286 struct proc *p = td->td_proc; 287 struct sigacts *psp; 288 struct trapframe *regs; 289 struct l_rt_sigframe *fp, frame; 290 int oonstack; 291 int sig; 292 int code; 293 294 sig = ksi->ksi_signo; 295 code = ksi->ksi_code; 296 PROC_LOCK_ASSERT(p, MA_OWNED); 297 psp = p->p_sigacts; 298 mtx_assert(&psp->ps_mtx, MA_OWNED); 299 regs = td->td_frame; 300 oonstack = sigonstack(regs->tf_rsp); 301 302 /* Allocate space for the signal handler context. */ 303 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && 304 SIGISMEMBER(psp->ps_sigonstack, sig)) { 305 fp = (struct l_rt_sigframe *)((uintptr_t)td->td_sigstk.ss_sp + 306 td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe)); 307 } else 308 fp = (struct l_rt_sigframe *)regs->tf_rsp - 1; 309 mtx_unlock(&psp->ps_mtx); 310 311 /* Build the argument list for the signal handler. */ 312 sig = bsd_to_linux_signal(sig); 313 314 bzero(&frame, sizeof(frame)); 315 316 frame.sf_handler = PTROUT(catcher); 317 frame.sf_sig = sig; 318 frame.sf_siginfo = PTROUT(&fp->sf_si); 319 frame.sf_ucontext = PTROUT(&fp->sf_sc); 320 321 /* Fill in POSIX parts. */ 322 siginfo_to_lsiginfo(&ksi->ksi_info, &frame.sf_si, sig); 323 324 /* 325 * Build the signal context to be used by sigreturn and libgcc unwind. 326 */ 327 frame.sf_sc.uc_flags = 0; /* XXX ??? */ 328 frame.sf_sc.uc_link = 0; /* XXX ??? */ 329 330 frame.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp); 331 frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size; 332 frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) 333 ? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE; 334 PROC_UNLOCK(p); 335 336 bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask); 337 338 frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__mask; 339 frame.sf_sc.uc_mcontext.sc_edi = regs->tf_rdi; 340 frame.sf_sc.uc_mcontext.sc_esi = regs->tf_rsi; 341 frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_rbp; 342 frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_rbx; 343 frame.sf_sc.uc_mcontext.sc_esp = regs->tf_rsp; 344 frame.sf_sc.uc_mcontext.sc_edx = regs->tf_rdx; 345 frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_rcx; 346 frame.sf_sc.uc_mcontext.sc_eax = regs->tf_rax; 347 frame.sf_sc.uc_mcontext.sc_eip = regs->tf_rip; 348 frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs; 349 frame.sf_sc.uc_mcontext.sc_gs = regs->tf_gs; 350 frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs; 351 frame.sf_sc.uc_mcontext.sc_es = regs->tf_es; 352 frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds; 353 frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_rflags; 354 frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_rsp; 355 frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss; 356 frame.sf_sc.uc_mcontext.sc_err = regs->tf_err; 357 frame.sf_sc.uc_mcontext.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr; 358 frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code); 359 360 if (copyout(&frame, fp, sizeof(frame)) != 0) { 361 /* 362 * Process has trashed its stack; give it an illegal 363 * instruction to halt it in its tracks. 364 */ 365 PROC_LOCK(p); 366 sigexit(td, SIGILL); 367 } 368 369 /* Build context to run handler in. */ 370 regs->tf_rsp = PTROUT(fp); 371 regs->tf_rip = __kernel_rt_sigreturn; 372 regs->tf_rflags &= ~(PSL_T | PSL_D); 373 regs->tf_cs = _ucode32sel; 374 regs->tf_ss = _udatasel; 375 regs->tf_ds = _udatasel; 376 regs->tf_es = _udatasel; 377 regs->tf_fs = _ufssel; 378 regs->tf_gs = _ugssel; 379 regs->tf_flags = TF_HASSEGS; 380 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 381 PROC_LOCK(p); 382 mtx_lock(&psp->ps_mtx); 383 } 384 385 /* 386 * Send an interrupt to process. 387 * 388 * Stack is set up to allow sigcode stored 389 * in u. to call routine, followed by kcall 390 * to sigreturn routine below. After sigreturn 391 * resets the signal mask, the stack, and the 392 * frame pointer, it returns to the user 393 * specified pc, psl. 394 */ 395 static void 396 linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 397 { 398 struct thread *td = curthread; 399 struct proc *p = td->td_proc; 400 struct sigacts *psp; 401 struct trapframe *regs; 402 struct l_sigframe *fp, frame; 403 l_sigset_t lmask; 404 int oonstack; 405 int sig, code; 406 407 sig = ksi->ksi_signo; 408 code = ksi->ksi_code; 409 PROC_LOCK_ASSERT(p, MA_OWNED); 410 psp = p->p_sigacts; 411 mtx_assert(&psp->ps_mtx, MA_OWNED); 412 if (SIGISMEMBER(psp->ps_siginfo, sig)) { 413 /* Signal handler installed with SA_SIGINFO. */ 414 linux_rt_sendsig(catcher, ksi, mask); 415 return; 416 } 417 418 regs = td->td_frame; 419 oonstack = sigonstack(regs->tf_rsp); 420 421 /* Allocate space for the signal handler context. */ 422 if ((td->td_pflags & TDP_ALTSTACK) && !oonstack && 423 SIGISMEMBER(psp->ps_sigonstack, sig)) { 424 fp = (struct l_sigframe *)((uintptr_t)td->td_sigstk.ss_sp + 425 td->td_sigstk.ss_size - sizeof(struct l_sigframe)); 426 } else 427 fp = (struct l_sigframe *)regs->tf_rsp - 1; 428 mtx_unlock(&psp->ps_mtx); 429 PROC_UNLOCK(p); 430 431 /* Build the argument list for the signal handler. */ 432 sig = bsd_to_linux_signal(sig); 433 434 bzero(&frame, sizeof(frame)); 435 436 frame.sf_handler = PTROUT(catcher); 437 frame.sf_sig = sig; 438 439 bsd_to_linux_sigset(mask, &lmask); 440 441 /* Build the signal context to be used by sigreturn. */ 442 frame.sf_sc.sc_mask = lmask.__mask; 443 frame.sf_sc.sc_gs = regs->tf_gs; 444 frame.sf_sc.sc_fs = regs->tf_fs; 445 frame.sf_sc.sc_es = regs->tf_es; 446 frame.sf_sc.sc_ds = regs->tf_ds; 447 frame.sf_sc.sc_edi = regs->tf_rdi; 448 frame.sf_sc.sc_esi = regs->tf_rsi; 449 frame.sf_sc.sc_ebp = regs->tf_rbp; 450 frame.sf_sc.sc_ebx = regs->tf_rbx; 451 frame.sf_sc.sc_esp = regs->tf_rsp; 452 frame.sf_sc.sc_edx = regs->tf_rdx; 453 frame.sf_sc.sc_ecx = regs->tf_rcx; 454 frame.sf_sc.sc_eax = regs->tf_rax; 455 frame.sf_sc.sc_eip = regs->tf_rip; 456 frame.sf_sc.sc_cs = regs->tf_cs; 457 frame.sf_sc.sc_eflags = regs->tf_rflags; 458 frame.sf_sc.sc_esp_at_signal = regs->tf_rsp; 459 frame.sf_sc.sc_ss = regs->tf_ss; 460 frame.sf_sc.sc_err = regs->tf_err; 461 frame.sf_sc.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr; 462 frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(code); 463 464 frame.sf_extramask[0] = lmask.__mask; 465 466 if (copyout(&frame, fp, sizeof(frame)) != 0) { 467 /* 468 * Process has trashed its stack; give it an illegal 469 * instruction to halt it in its tracks. 470 */ 471 PROC_LOCK(p); 472 sigexit(td, SIGILL); 473 } 474 475 /* Build context to run handler in. */ 476 regs->tf_rsp = PTROUT(fp); 477 regs->tf_rip = __kernel_sigreturn; 478 regs->tf_rflags &= ~(PSL_T | PSL_D); 479 regs->tf_cs = _ucode32sel; 480 regs->tf_ss = _udatasel; 481 regs->tf_ds = _udatasel; 482 regs->tf_es = _udatasel; 483 regs->tf_fs = _ufssel; 484 regs->tf_gs = _ugssel; 485 regs->tf_flags = TF_HASSEGS; 486 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 487 PROC_LOCK(p); 488 mtx_lock(&psp->ps_mtx); 489 } 490 491 /* 492 * System call to cleanup state after a signal 493 * has been taken. Reset signal mask and 494 * stack state from context left by sendsig (above). 495 * Return to previous pc and psl as specified by 496 * context left by sendsig. Check carefully to 497 * make sure that the user has not modified the 498 * psl to gain improper privileges or to cause 499 * a machine fault. 500 */ 501 int 502 linux_sigreturn(struct thread *td, struct linux_sigreturn_args *args) 503 { 504 struct l_sigframe frame; 505 struct trapframe *regs; 506 sigset_t bmask; 507 l_sigset_t lmask; 508 int eflags; 509 ksiginfo_t ksi; 510 511 regs = td->td_frame; 512 513 /* 514 * The trampoline code hands us the sigframe. 515 * It is unsafe to keep track of it ourselves, in the event that a 516 * program jumps out of a signal handler. 517 */ 518 if (copyin(args->sfp, &frame, sizeof(frame)) != 0) 519 return (EFAULT); 520 521 /* Check for security violations. */ 522 #define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) 523 eflags = frame.sf_sc.sc_eflags; 524 if (!EFLAGS_SECURE(eflags, regs->tf_rflags)) 525 return(EINVAL); 526 527 /* 528 * Don't allow users to load a valid privileged %cs. Let the 529 * hardware check for invalid selectors, excess privilege in 530 * other selectors, invalid %eip's and invalid %esp's. 531 */ 532 #define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) 533 if (!CS_SECURE(frame.sf_sc.sc_cs)) { 534 ksiginfo_init_trap(&ksi); 535 ksi.ksi_signo = SIGBUS; 536 ksi.ksi_code = BUS_OBJERR; 537 ksi.ksi_trapno = T_PROTFLT; 538 ksi.ksi_addr = (void *)regs->tf_rip; 539 trapsignal(td, &ksi); 540 return(EINVAL); 541 } 542 543 lmask.__mask = frame.sf_sc.sc_mask; 544 lmask.__mask = frame.sf_extramask[0]; 545 linux_to_bsd_sigset(&lmask, &bmask); 546 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); 547 548 /* Restore signal context. */ 549 regs->tf_rdi = frame.sf_sc.sc_edi; 550 regs->tf_rsi = frame.sf_sc.sc_esi; 551 regs->tf_rbp = frame.sf_sc.sc_ebp; 552 regs->tf_rbx = frame.sf_sc.sc_ebx; 553 regs->tf_rdx = frame.sf_sc.sc_edx; 554 regs->tf_rcx = frame.sf_sc.sc_ecx; 555 regs->tf_rax = frame.sf_sc.sc_eax; 556 regs->tf_rip = frame.sf_sc.sc_eip; 557 regs->tf_cs = frame.sf_sc.sc_cs; 558 regs->tf_ds = frame.sf_sc.sc_ds; 559 regs->tf_es = frame.sf_sc.sc_es; 560 regs->tf_fs = frame.sf_sc.sc_fs; 561 regs->tf_gs = frame.sf_sc.sc_gs; 562 regs->tf_rflags = eflags; 563 regs->tf_rsp = frame.sf_sc.sc_esp_at_signal; 564 regs->tf_ss = frame.sf_sc.sc_ss; 565 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 566 567 return (EJUSTRETURN); 568 } 569 570 /* 571 * System call to cleanup state after a signal 572 * has been taken. Reset signal mask and 573 * stack state from context left by rt_sendsig (above). 574 * Return to previous pc and psl as specified by 575 * context left by sendsig. Check carefully to 576 * make sure that the user has not modified the 577 * psl to gain improper privileges or to cause 578 * a machine fault. 579 */ 580 int 581 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 582 { 583 struct l_ucontext uc; 584 struct l_sigcontext *context; 585 sigset_t bmask; 586 l_stack_t *lss; 587 stack_t ss; 588 struct trapframe *regs; 589 int eflags; 590 ksiginfo_t ksi; 591 592 regs = td->td_frame; 593 594 /* 595 * The trampoline code hands us the ucontext. 596 * It is unsafe to keep track of it ourselves, in the event that a 597 * program jumps out of a signal handler. 598 */ 599 if (copyin(args->ucp, &uc, sizeof(uc)) != 0) 600 return (EFAULT); 601 602 context = &uc.uc_mcontext; 603 604 /* Check for security violations. */ 605 #define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) 606 eflags = context->sc_eflags; 607 if (!EFLAGS_SECURE(eflags, regs->tf_rflags)) 608 return(EINVAL); 609 610 /* 611 * Don't allow users to load a valid privileged %cs. Let the 612 * hardware check for invalid selectors, excess privilege in 613 * other selectors, invalid %eip's and invalid %esp's. 614 */ 615 #define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) 616 if (!CS_SECURE(context->sc_cs)) { 617 ksiginfo_init_trap(&ksi); 618 ksi.ksi_signo = SIGBUS; 619 ksi.ksi_code = BUS_OBJERR; 620 ksi.ksi_trapno = T_PROTFLT; 621 ksi.ksi_addr = (void *)regs->tf_rip; 622 trapsignal(td, &ksi); 623 return(EINVAL); 624 } 625 626 linux_to_bsd_sigset(&uc.uc_sigmask, &bmask); 627 kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0); 628 629 /* 630 * Restore signal context 631 */ 632 regs->tf_gs = context->sc_gs; 633 regs->tf_fs = context->sc_fs; 634 regs->tf_es = context->sc_es; 635 regs->tf_ds = context->sc_ds; 636 regs->tf_rdi = context->sc_edi; 637 regs->tf_rsi = context->sc_esi; 638 regs->tf_rbp = context->sc_ebp; 639 regs->tf_rbx = context->sc_ebx; 640 regs->tf_rdx = context->sc_edx; 641 regs->tf_rcx = context->sc_ecx; 642 regs->tf_rax = context->sc_eax; 643 regs->tf_rip = context->sc_eip; 644 regs->tf_cs = context->sc_cs; 645 regs->tf_rflags = eflags; 646 regs->tf_rsp = context->sc_esp_at_signal; 647 regs->tf_ss = context->sc_ss; 648 set_pcb_flags(td->td_pcb, PCB_FULL_IRET); 649 650 /* 651 * call sigaltstack & ignore results.. 652 */ 653 lss = &uc.uc_stack; 654 ss.ss_sp = PTRIN(lss->ss_sp); 655 ss.ss_size = lss->ss_size; 656 ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags); 657 658 (void)kern_sigaltstack(td, &ss, NULL); 659 660 return (EJUSTRETURN); 661 } 662 663 static int 664 linux32_fetch_syscall_args(struct thread *td) 665 { 666 struct proc *p; 667 struct trapframe *frame; 668 struct syscall_args *sa; 669 670 p = td->td_proc; 671 frame = td->td_frame; 672 sa = &td->td_sa; 673 674 sa->args[0] = frame->tf_rbx; 675 sa->args[1] = frame->tf_rcx; 676 sa->args[2] = frame->tf_rdx; 677 sa->args[3] = frame->tf_rsi; 678 sa->args[4] = frame->tf_rdi; 679 sa->args[5] = frame->tf_rbp; /* Unconfirmed */ 680 sa->code = frame->tf_rax; 681 sa->original_code = sa->code; 682 683 if (sa->code >= p->p_sysent->sv_size) 684 /* nosys */ 685 sa->callp = &p->p_sysent->sv_table[p->p_sysent->sv_size - 1]; 686 else 687 sa->callp = &p->p_sysent->sv_table[sa->code]; 688 689 td->td_retval[0] = 0; 690 td->td_retval[1] = frame->tf_rdx; 691 692 return (0); 693 } 694 695 static void 696 linux32_set_syscall_retval(struct thread *td, int error) 697 { 698 struct trapframe *frame = td->td_frame; 699 700 cpu_set_syscall_retval(td, error); 701 702 if (__predict_false(error != 0)) { 703 if (error != ERESTART && error != EJUSTRETURN) 704 frame->tf_rax = bsd_to_linux_errno(error); 705 } 706 } 707 708 static void 709 linux32_set_fork_retval(struct thread *td) 710 { 711 struct trapframe *frame = td->td_frame; 712 713 frame->tf_rax = 0; 714 } 715 716 /* 717 * Clear registers on exec 718 * XXX copied from ia32_signal.c. 719 */ 720 static void 721 linux_exec_setregs(struct thread *td, struct image_params *imgp, 722 uintptr_t stack) 723 { 724 struct trapframe *regs = td->td_frame; 725 struct pcb *pcb = td->td_pcb; 726 register_t saved_rflags; 727 728 regs = td->td_frame; 729 pcb = td->td_pcb; 730 731 if (td->td_proc->p_md.md_ldt != NULL) 732 user_ldt_free(td); 733 734 critical_enter(); 735 wrmsr(MSR_FSBASE, 0); 736 wrmsr(MSR_KGSBASE, 0); /* User value while we're in the kernel */ 737 pcb->pcb_fsbase = 0; 738 pcb->pcb_gsbase = 0; 739 critical_exit(); 740 pcb->pcb_initial_fpucw = __LINUX_NPXCW__; 741 742 saved_rflags = regs->tf_rflags & PSL_T; 743 bzero((char *)regs, sizeof(struct trapframe)); 744 regs->tf_rip = imgp->entry_addr; 745 regs->tf_rsp = stack; 746 regs->tf_rflags = PSL_USER | saved_rflags; 747 regs->tf_gs = _ugssel; 748 regs->tf_fs = _ufssel; 749 regs->tf_es = _udatasel; 750 regs->tf_ds = _udatasel; 751 regs->tf_ss = _udatasel; 752 regs->tf_flags = TF_HASSEGS; 753 regs->tf_cs = _ucode32sel; 754 regs->tf_rbx = (register_t)imgp->ps_strings; 755 756 x86_clear_dbregs(pcb); 757 758 fpstate_drop(td); 759 760 /* Do full restore on return so that we can change to a different %cs */ 761 set_pcb_flags(pcb, PCB_32BIT | PCB_FULL_IRET); 762 } 763 764 /* 765 * XXX copied from ia32_sysvec.c. 766 */ 767 static int 768 linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) 769 { 770 int argc, envc, error; 771 u_int32_t *vectp; 772 char *stringp; 773 uintptr_t destp, ustringp; 774 struct linux32_ps_strings *arginfo; 775 char canary[LINUX_AT_RANDOM_LEN]; 776 size_t execpath_len; 777 778 /* Calculate string base and vector table pointers. */ 779 if (imgp->execpath != NULL && imgp->auxargs != NULL) 780 execpath_len = strlen(imgp->execpath) + 1; 781 else 782 execpath_len = 0; 783 784 arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS; 785 destp = (uintptr_t)arginfo; 786 787 if (execpath_len != 0) { 788 destp -= execpath_len; 789 destp = rounddown2(destp, sizeof(uint32_t)); 790 imgp->execpathp = (void *)destp; 791 error = copyout(imgp->execpath, imgp->execpathp, execpath_len); 792 if (error != 0) 793 return (error); 794 } 795 796 /* Prepare the canary for SSP. */ 797 arc4rand(canary, sizeof(canary), 0); 798 destp -= roundup(sizeof(canary), sizeof(uint32_t)); 799 imgp->canary = (void *)destp; 800 error = copyout(canary, imgp->canary, sizeof(canary)); 801 if (error != 0) 802 return (error); 803 804 /* Allocate room for the argument and environment strings. */ 805 destp -= ARG_MAX - imgp->args->stringspace; 806 destp = rounddown2(destp, sizeof(uint32_t)); 807 ustringp = destp; 808 809 if (imgp->auxargs) { 810 /* 811 * Allocate room on the stack for the ELF auxargs 812 * array. It has LINUX_AT_COUNT entries. 813 */ 814 destp -= LINUX_AT_COUNT * sizeof(Elf32_Auxinfo); 815 destp = rounddown2(destp, sizeof(uint32_t)); 816 } 817 818 vectp = (uint32_t *)destp; 819 820 /* 821 * Allocate room for the argv[] and env vectors including the 822 * terminating NULL pointers. 823 */ 824 vectp -= imgp->args->argc + 1 + imgp->args->envc + 1; 825 826 /* vectp also becomes our initial stack base. */ 827 *stack_base = (uintptr_t)vectp; 828 829 stringp = imgp->args->begin_argv; 830 argc = imgp->args->argc; 831 envc = imgp->args->envc; 832 833 /* Copy out strings - arguments and environment. */ 834 error = copyout(stringp, (void *)ustringp, 835 ARG_MAX - imgp->args->stringspace); 836 if (error != 0) 837 return (error); 838 839 /* Fill in "ps_strings" struct for ps, w, etc. */ 840 if (suword32(&arginfo->ps_argvstr, (uint32_t)(intptr_t)vectp) != 0 || 841 suword32(&arginfo->ps_nargvstr, argc) != 0) 842 return (EFAULT); 843 844 /* Fill in argument portion of vector table. */ 845 for (; argc > 0; --argc) { 846 if (suword32(vectp++, ustringp) != 0) 847 return (EFAULT); 848 while (*stringp++ != 0) 849 ustringp++; 850 ustringp++; 851 } 852 853 /* A null vector table pointer separates the argp's from the envp's. */ 854 if (suword32(vectp++, 0) != 0) 855 return (EFAULT); 856 857 if (suword32(&arginfo->ps_envstr, (uint32_t)(intptr_t)vectp) != 0 || 858 suword32(&arginfo->ps_nenvstr, envc) != 0) 859 return (EFAULT); 860 861 /* Fill in environment portion of vector table. */ 862 for (; envc > 0; --envc) { 863 if (suword32(vectp++, ustringp) != 0) 864 return (EFAULT); 865 while (*stringp++ != 0) 866 ustringp++; 867 ustringp++; 868 } 869 870 /* The end of the vector table is a null pointer. */ 871 if (suword32(vectp, 0) != 0) 872 return (EFAULT); 873 874 if (imgp->auxargs) { 875 vectp++; 876 error = imgp->sysent->sv_copyout_auxargs(imgp, 877 (uintptr_t)vectp); 878 if (error != 0) 879 return (error); 880 } 881 882 return (0); 883 } 884 885 static SYSCTL_NODE(_compat, OID_AUTO, linux32, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 886 "32-bit Linux emulation"); 887 888 static u_long linux32_maxdsiz = LINUX32_MAXDSIZ; 889 SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxdsiz, CTLFLAG_RW, 890 &linux32_maxdsiz, 0, ""); 891 static u_long linux32_maxssiz = LINUX32_MAXSSIZ; 892 SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxssiz, CTLFLAG_RW, 893 &linux32_maxssiz, 0, ""); 894 static u_long linux32_maxvmem = LINUX32_MAXVMEM; 895 SYSCTL_ULONG(_compat_linux32, OID_AUTO, maxvmem, CTLFLAG_RW, 896 &linux32_maxvmem, 0, ""); 897 898 static void 899 linux32_fixlimit(struct rlimit *rl, int which) 900 { 901 902 switch (which) { 903 case RLIMIT_DATA: 904 if (linux32_maxdsiz != 0) { 905 if (rl->rlim_cur > linux32_maxdsiz) 906 rl->rlim_cur = linux32_maxdsiz; 907 if (rl->rlim_max > linux32_maxdsiz) 908 rl->rlim_max = linux32_maxdsiz; 909 } 910 break; 911 case RLIMIT_STACK: 912 if (linux32_maxssiz != 0) { 913 if (rl->rlim_cur > linux32_maxssiz) 914 rl->rlim_cur = linux32_maxssiz; 915 if (rl->rlim_max > linux32_maxssiz) 916 rl->rlim_max = linux32_maxssiz; 917 } 918 break; 919 case RLIMIT_VMEM: 920 if (linux32_maxvmem != 0) { 921 if (rl->rlim_cur > linux32_maxvmem) 922 rl->rlim_cur = linux32_maxvmem; 923 if (rl->rlim_max > linux32_maxvmem) 924 rl->rlim_max = linux32_maxvmem; 925 } 926 break; 927 } 928 } 929 930 struct sysentvec elf_linux_sysvec = { 931 .sv_size = LINUX32_SYS_MAXSYSCALL, 932 .sv_table = linux32_sysent, 933 .sv_transtrap = linux_translate_traps, 934 .sv_fixup = linux_fixup_elf, 935 .sv_sendsig = linux_sendsig, 936 .sv_sigcode = &_binary_linux32_vdso_so_o_start, 937 .sv_szsigcode = &linux_szsigcode, 938 .sv_name = "Linux ELF32", 939 .sv_coredump = elf32_coredump, 940 .sv_elf_core_osabi = ELFOSABI_NONE, 941 .sv_elf_core_abi_vendor = LINUX_ABI_VENDOR, 942 .sv_elf_core_prepare_notes = linux32_prepare_notes, 943 .sv_imgact_try = linux_exec_imgact_try, 944 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 945 .sv_minuser = VM_MIN_ADDRESS, 946 .sv_maxuser = LINUX32_MAXUSER, 947 .sv_usrstack = LINUX32_USRSTACK, 948 .sv_psstrings = LINUX32_PS_STRINGS, 949 .sv_stackprot = VM_PROT_ALL, 950 .sv_copyout_auxargs = linux_copyout_auxargs, 951 .sv_copyout_strings = linux_copyout_strings, 952 .sv_setregs = linux_exec_setregs, 953 .sv_fixlimit = linux32_fixlimit, 954 .sv_maxssiz = &linux32_maxssiz, 955 .sv_flags = SV_ABI_LINUX | SV_ILP32 | SV_IA32 | SV_SHP | 956 SV_SIG_DISCIGN | SV_SIG_WAITNDQ | SV_TIMEKEEP, 957 .sv_set_syscall_retval = linux32_set_syscall_retval, 958 .sv_fetch_syscall_args = linux32_fetch_syscall_args, 959 .sv_syscallnames = NULL, 960 .sv_shared_page_base = LINUX32_SHAREDPAGE, 961 .sv_shared_page_len = PAGE_SIZE, 962 .sv_schedtail = linux_schedtail, 963 .sv_thread_detach = linux_thread_detach, 964 .sv_trap = NULL, 965 .sv_onexec = linux_on_exec_vmspace, 966 .sv_onexit = linux_on_exit, 967 .sv_ontdexit = linux_thread_dtor, 968 .sv_setid_allowed = &linux_setid_allowed_query, 969 .sv_set_fork_retval = linux32_set_fork_retval, 970 }; 971 972 static int 973 linux_on_exec_vmspace(struct proc *p, struct image_params *imgp) 974 { 975 int error; 976 977 error = linux_map_vdso(p, linux_vdso_obj, linux_vdso_base, 978 LINUX32_VDSOPAGE_SIZE, imgp); 979 if (error == 0) 980 linux_on_exec(p, imgp); 981 return (error); 982 } 983 984 /* 985 * linux_vdso_install() and linux_exec_sysvec_init() must be called 986 * after exec_sysvec_init() which is SI_SUB_EXEC (SI_ORDER_ANY). 987 */ 988 static void 989 linux_exec_sysvec_init(void *param) 990 { 991 l_uintptr_t *ktimekeep_base, *ktsc_selector; 992 struct sysentvec *sv; 993 ptrdiff_t tkoff; 994 995 sv = param; 996 /* Fill timekeep_base */ 997 exec_sysvec_init(sv); 998 999 tkoff = kern_timekeep_base - linux_vdso_base; 1000 ktimekeep_base = (l_uintptr_t *)(linux_vdso_mapping + tkoff); 1001 *ktimekeep_base = sv->sv_timekeep_base; 1002 1003 tkoff = kern_tsc_selector - linux_vdso_base; 1004 ktsc_selector = (l_uintptr_t *)(linux_vdso_mapping + tkoff); 1005 *ktsc_selector = linux_vdso_tsc_selector_idx(); 1006 if (bootverbose) 1007 printf("Linux i386 vDSO tsc_selector: %u\n", *ktsc_selector); 1008 } 1009 SYSINIT(elf_linux_exec_sysvec_init, SI_SUB_EXEC + 1, SI_ORDER_ANY, 1010 linux_exec_sysvec_init, &elf_linux_sysvec); 1011 1012 static void 1013 linux_vdso_install(const void *param) 1014 { 1015 char *vdso_start = &_binary_linux32_vdso_so_o_start; 1016 char *vdso_end = &_binary_linux32_vdso_so_o_end; 1017 1018 linux_szsigcode = vdso_end - vdso_start; 1019 MPASS(linux_szsigcode <= LINUX32_VDSOPAGE_SIZE); 1020 1021 linux_vdso_base = LINUX32_VDSOPAGE; 1022 1023 __elfN(linux_vdso_fixup)(vdso_start, linux_vdso_base); 1024 1025 linux_vdso_obj = __elfN(linux_shared_page_init) 1026 (&linux_vdso_mapping, LINUX32_VDSOPAGE_SIZE); 1027 bcopy(vdso_start, linux_vdso_mapping, linux_szsigcode); 1028 1029 linux_vdso_reloc(linux_vdso_mapping, linux_vdso_base); 1030 } 1031 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC + 1, SI_ORDER_FIRST, 1032 linux_vdso_install, NULL); 1033 1034 static void 1035 linux_vdso_deinstall(const void *param) 1036 { 1037 1038 __elfN(linux_shared_page_fini)(linux_vdso_obj, 1039 linux_vdso_mapping, LINUX32_VDSOPAGE_SIZE); 1040 } 1041 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, 1042 linux_vdso_deinstall, NULL); 1043 1044 static void 1045 linux_vdso_reloc(char *mapping, Elf_Addr offset) 1046 { 1047 const Elf_Shdr *shdr; 1048 const Elf_Rel *rel; 1049 const Elf_Ehdr *ehdr; 1050 Elf32_Addr *where; 1051 Elf_Size rtype, symidx; 1052 Elf32_Addr addr, addend; 1053 int i, relcnt; 1054 1055 MPASS(offset != 0); 1056 1057 relcnt = 0; 1058 ehdr = (const Elf_Ehdr *)mapping; 1059 shdr = (const Elf_Shdr *)(mapping + ehdr->e_shoff); 1060 for (i = 0; i < ehdr->e_shnum; i++) 1061 { 1062 switch (shdr[i].sh_type) { 1063 case SHT_REL: 1064 rel = (const Elf_Rel *)(mapping + shdr[i].sh_offset); 1065 relcnt = shdr[i].sh_size / sizeof(*rel); 1066 break; 1067 case SHT_RELA: 1068 printf("Linux i386 vDSO: unexpected Rela section\n"); 1069 break; 1070 } 1071 } 1072 1073 for (i = 0; i < relcnt; i++, rel++) { 1074 where = (Elf32_Addr *)(mapping + rel->r_offset); 1075 addend = *where; 1076 rtype = ELF_R_TYPE(rel->r_info); 1077 symidx = ELF_R_SYM(rel->r_info); 1078 1079 switch (rtype) { 1080 case R_386_NONE: /* none */ 1081 break; 1082 1083 case R_386_RELATIVE: /* B + A */ 1084 addr = (Elf32_Addr)PTROUT(offset + addend); 1085 if (*where != addr) 1086 *where = addr; 1087 break; 1088 1089 case R_386_IRELATIVE: 1090 printf("Linux i386 vDSO: unexpected ifunc relocation, " 1091 "symbol index %ld\n", (intmax_t)symidx); 1092 break; 1093 default: 1094 printf("Linux i386 vDSO: unexpected relocation type %ld, " 1095 "symbol index %ld\n", (intmax_t)rtype, (intmax_t)symidx); 1096 } 1097 } 1098 } 1099 1100 static char GNU_ABI_VENDOR[] = "GNU"; 1101 static int GNULINUX_ABI_DESC = 0; 1102 1103 static bool 1104 linux32_trans_osrel(const Elf_Note *note, int32_t *osrel) 1105 { 1106 const Elf32_Word *desc; 1107 uintptr_t p; 1108 1109 p = (uintptr_t)(note + 1); 1110 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 1111 1112 desc = (const Elf32_Word *)p; 1113 if (desc[0] != GNULINUX_ABI_DESC) 1114 return (false); 1115 1116 /* 1117 * For Linux we encode osrel using the Linux convention of 1118 * (version << 16) | (major << 8) | (minor) 1119 * See macro in linux_mib.h 1120 */ 1121 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); 1122 1123 return (true); 1124 } 1125 1126 static Elf_Brandnote linux32_brandnote = { 1127 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 1128 .hdr.n_descsz = 16, /* XXX at least 16 */ 1129 .hdr.n_type = 1, 1130 .vendor = GNU_ABI_VENDOR, 1131 .flags = BN_TRANSLATE_OSREL, 1132 .trans_osrel = linux32_trans_osrel 1133 }; 1134 1135 static Elf32_Brandinfo linux_brand = { 1136 .brand = ELFOSABI_LINUX, 1137 .machine = EM_386, 1138 .compat_3_brand = "Linux", 1139 .emul_path = linux_emul_path, 1140 .interp_path = "/lib/ld-linux.so.1", 1141 .sysvec = &elf_linux_sysvec, 1142 .interp_newpath = NULL, 1143 .brand_note = &linux32_brandnote, 1144 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 1145 }; 1146 1147 static Elf32_Brandinfo linux_glibc2brand = { 1148 .brand = ELFOSABI_LINUX, 1149 .machine = EM_386, 1150 .compat_3_brand = "Linux", 1151 .emul_path = linux_emul_path, 1152 .interp_path = "/lib/ld-linux.so.2", 1153 .sysvec = &elf_linux_sysvec, 1154 .interp_newpath = NULL, 1155 .brand_note = &linux32_brandnote, 1156 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 1157 }; 1158 1159 static Elf32_Brandinfo linux_muslbrand = { 1160 .brand = ELFOSABI_LINUX, 1161 .machine = EM_386, 1162 .compat_3_brand = "Linux", 1163 .emul_path = linux_emul_path, 1164 .interp_path = "/lib/ld-musl-i386.so.1", 1165 .sysvec = &elf_linux_sysvec, 1166 .interp_newpath = NULL, 1167 .brand_note = &linux32_brandnote, 1168 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE | 1169 LINUX_BI_FUTEX_REQUEUE 1170 }; 1171 1172 Elf32_Brandinfo *linux_brandlist[] = { 1173 &linux_brand, 1174 &linux_glibc2brand, 1175 &linux_muslbrand, 1176 NULL 1177 }; 1178 1179 static int 1180 linux_elf_modevent(module_t mod, int type, void *data) 1181 { 1182 Elf32_Brandinfo **brandinfo; 1183 int error; 1184 struct linux_ioctl_handler **lihp; 1185 1186 error = 0; 1187 1188 switch(type) { 1189 case MOD_LOAD: 1190 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 1191 ++brandinfo) 1192 if (elf32_insert_brand_entry(*brandinfo) < 0) 1193 error = EINVAL; 1194 if (error == 0) { 1195 SET_FOREACH(lihp, linux_ioctl_handler_set) 1196 linux32_ioctl_register_handler(*lihp); 1197 stclohz = (stathz ? stathz : hz); 1198 if (bootverbose) 1199 printf("Linux i386 ELF exec handler installed\n"); 1200 } else 1201 printf("cannot insert Linux i386 ELF brand handler\n"); 1202 break; 1203 case MOD_UNLOAD: 1204 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 1205 ++brandinfo) 1206 if (elf32_brand_inuse(*brandinfo)) 1207 error = EBUSY; 1208 if (error == 0) { 1209 for (brandinfo = &linux_brandlist[0]; 1210 *brandinfo != NULL; ++brandinfo) 1211 if (elf32_remove_brand_entry(*brandinfo) < 0) 1212 error = EINVAL; 1213 } 1214 if (error == 0) { 1215 SET_FOREACH(lihp, linux_ioctl_handler_set) 1216 linux32_ioctl_unregister_handler(*lihp); 1217 if (bootverbose) 1218 printf("Linux i386 ELF exec handler removed\n"); 1219 } else 1220 printf("Could not deinstall Linux i386 ELF interpreter entry\n"); 1221 break; 1222 default: 1223 return (EOPNOTSUPP); 1224 } 1225 return (error); 1226 } 1227 1228 static moduledata_t linux_elf_mod = { 1229 "linuxelf", 1230 linux_elf_modevent, 1231 0 1232 }; 1233 1234 DECLARE_MODULE_TIED(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 1235 MODULE_DEPEND(linuxelf, linux_common, 1, 1, 1); 1236 FEATURE(linux, "Linux 32bit support"); 1237