1 /*- 2 * Copyright (c) 1994-1996 S�ren Schmidt 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software withough specific prior written permission 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 * 28 * $Id: linux_sysvec.c,v 1.18 1997/08/25 23:41:39 bde Exp $ 29 */ 30 31 /* XXX we use functions that might not exist. */ 32 #define COMPAT_43 1 33 34 #include <sys/param.h> 35 #include <sys/buf.h> 36 #include <sys/proc.h> 37 #include <sys/systm.h> 38 #include <sys/sysent.h> 39 #include <sys/imgact.h> 40 #include <sys/imgact_elf.h> 41 #include <sys/signalvar.h> 42 #include <sys/malloc.h> 43 #include <vm/vm.h> 44 #include <vm/vm_param.h> 45 #include <vm/vm_prot.h> 46 #include <vm/vm_page.h> 47 #include <vm/vm_extern.h> 48 #include <sys/exec.h> 49 #include <sys/kernel.h> 50 #include <machine/cpu.h> 51 52 #include <i386/linux/linux.h> 53 #include <i386/linux/linux_proto.h> 54 55 int linux_fixup __P((int **stack_base, struct image_params *iparams)); 56 int elf_linux_fixup __P((int **stack_base, struct image_params *iparams)); 57 void linux_prepsyscall __P((struct trapframe *tf, int *args, u_int *code, caddr_t *params)); 58 void linux_sendsig __P((sig_t catcher, int sig, int mask, u_long code)); 59 60 /* 61 * Linux syscalls return negative errno's, we do positive and map them 62 */ 63 int bsd_to_linux_errno[ELAST] = { 64 -0, -1, -2, -3, -4, -5, -6, -7, -8, -9, 65 -10, -35, -12, -13, -14, -15, -16, -17, -18, -19, 66 -20, -21, -22, -23, -24, -25, -26, -27, -28, -29, 67 -30, -31, -32, -33, -34, -11,-115,-114, -88, -89, 68 -90, -91, -92, -93, -94, -95, -96, -97, -98, -99, 69 -100,-101,-102,-103,-104,-105,-106,-107,-108,-109, 70 -110,-111, -40, -36,-112,-113, -39, -11, -87,-122, 71 -116, -66, -6, -6, -6, -6, -6, -37, -38, -9, 72 -6, 73 }; 74 75 int bsd_to_linux_signal[NSIG] = { 76 0, LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT, 77 LINUX_SIGILL, LINUX_SIGTRAP, LINUX_SIGABRT, 0, 78 LINUX_SIGFPE, LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV, 79 0, LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM, 80 LINUX_SIGURG, LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT, 81 LINUX_SIGCHLD, LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO, 82 LINUX_SIGXCPU, LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF, 83 LINUX_SIGWINCH, 0, LINUX_SIGUSR1, LINUX_SIGUSR2 84 }; 85 86 int linux_to_bsd_signal[LINUX_NSIG] = { 87 0, SIGHUP, SIGINT, SIGQUIT, SIGILL, SIGTRAP, SIGABRT, SIGEMT, 88 SIGFPE, SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2, SIGPIPE, SIGALRM, SIGTERM, 89 SIGBUS, SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU, SIGIO, 90 SIGXCPU, SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH, SIGURG, SIGURG, 0 91 }; 92 93 int linux_fixup(int **stack_base, struct image_params *imgp) 94 { 95 int *argv, *envp; 96 97 argv = *stack_base; 98 envp = *stack_base + (imgp->argc + 1); 99 (*stack_base)--; 100 **stack_base = (int)envp; 101 (*stack_base)--; 102 **stack_base = (int)argv; 103 (*stack_base)--; 104 **stack_base = (int)imgp->argc; 105 return 0; 106 } 107 108 int elf_linux_fixup(int **stack_base, struct image_params *imgp) 109 { 110 Elf32_Auxargs *args = (Elf32_Auxargs *)imgp->auxargs; 111 int *pos; 112 113 pos = *stack_base + (imgp->argc + imgp->envc + 2); 114 115 if (args->trace) { 116 AUXARGS_ENTRY(pos, AT_DEBUG, 1); 117 } 118 if (args->execfd != -1) { 119 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); 120 } 121 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); 122 AUXARGS_ENTRY(pos, AT_PHENT, args->phent); 123 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); 124 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); 125 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); 126 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); 127 AUXARGS_ENTRY(pos, AT_BASE, args->base); 128 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_cred->p_ruid); 129 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_cred->p_svuid); 130 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_cred->p_rgid); 131 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_cred->p_svgid); 132 AUXARGS_ENTRY(pos, AT_NULL, 0); 133 134 free(imgp->auxargs, M_TEMP); 135 imgp->auxargs = NULL; 136 137 (*stack_base)--; 138 **stack_base = (int)imgp->argc; 139 return 0; 140 } 141 142 extern int _ucodesel, _udatasel; 143 144 /* 145 * Send an interrupt to process. 146 * 147 * Stack is set up to allow sigcode stored 148 * in u. to call routine, followed by kcall 149 * to sigreturn routine below. After sigreturn 150 * resets the signal mask, the stack, and the 151 * frame pointer, it returns to the user 152 * specified pc, psl. 153 */ 154 155 void 156 linux_sendsig(sig_t catcher, int sig, int mask, u_long code) 157 { 158 register struct proc *p = curproc; 159 register struct trapframe *regs; 160 struct linux_sigframe *fp, frame; 161 struct sigacts *psp = p->p_sigacts; 162 int oonstack; 163 164 regs = p->p_md.md_regs; 165 oonstack = psp->ps_sigstk.ss_flags & SS_ONSTACK; 166 167 #ifdef DEBUG 168 printf("Linux-emul(%d): linux_sendsig(%8x, %d, %d, %ld)\n", 169 p->p_pid, catcher, sig, mask, code); 170 #endif 171 /* 172 * Allocate space for the signal handler context. 173 */ 174 if ((psp->ps_flags & SAS_ALTSTACK) && !oonstack && 175 (psp->ps_sigonstack & sigmask(sig))) { 176 fp = (struct linux_sigframe *)(psp->ps_sigstk.ss_sp + 177 psp->ps_sigstk.ss_size - sizeof(struct linux_sigframe)); 178 psp->ps_sigstk.ss_flags |= SS_ONSTACK; 179 } else { 180 fp = (struct linux_sigframe *)regs->tf_esp - 1; 181 } 182 183 /* 184 * grow() will return FALSE if the fp will not fit inside the stack 185 * and the stack can not be grown. useracc will return FALSE 186 * if access is denied. 187 */ 188 if ((grow(p, (int)fp) == FALSE) || 189 (useracc((caddr_t)fp, sizeof (struct linux_sigframe), B_WRITE) == FALSE)) { 190 /* 191 * Process has trashed its stack; give it an illegal 192 * instruction to halt it in its tracks. 193 */ 194 SIGACTION(p, SIGILL) = SIG_DFL; 195 sig = sigmask(SIGILL); 196 p->p_sigignore &= ~sig; 197 p->p_sigcatch &= ~sig; 198 p->p_sigmask &= ~sig; 199 psignal(p, SIGILL); 200 return; 201 } 202 203 /* 204 * Build the argument list for the signal handler. 205 */ 206 if (p->p_sysent->sv_sigtbl) { 207 if (sig < p->p_sysent->sv_sigsize) 208 sig = p->p_sysent->sv_sigtbl[sig]; 209 else 210 sig = p->p_sysent->sv_sigsize + 1; 211 } 212 213 frame.sf_handler = catcher; 214 frame.sf_sig = sig; 215 216 /* 217 * Build the signal context to be used by sigreturn. 218 */ 219 frame.sf_sc.sc_mask = mask; 220 __asm("movl %%gs,%w0" : "=r" (frame.sf_sc.sc_gs)); 221 __asm("movl %%fs,%w0" : "=r" (frame.sf_sc.sc_fs)); 222 frame.sf_sc.sc_es = regs->tf_es; 223 frame.sf_sc.sc_ds = regs->tf_ds; 224 frame.sf_sc.sc_edi = regs->tf_edi; 225 frame.sf_sc.sc_esi = regs->tf_esi; 226 frame.sf_sc.sc_ebp = regs->tf_ebp; 227 frame.sf_sc.sc_ebx = regs->tf_ebx; 228 frame.sf_sc.sc_edx = regs->tf_edx; 229 frame.sf_sc.sc_ecx = regs->tf_ecx; 230 frame.sf_sc.sc_eax = regs->tf_eax; 231 frame.sf_sc.sc_eip = regs->tf_eip; 232 frame.sf_sc.sc_cs = regs->tf_cs; 233 frame.sf_sc.sc_eflags = regs->tf_eflags; 234 frame.sf_sc.sc_esp_at_signal = regs->tf_esp; 235 frame.sf_sc.sc_ss = regs->tf_ss; 236 frame.sf_sc.sc_err = regs->tf_err; 237 frame.sf_sc.sc_trapno = code; /* XXX ???? */ 238 239 if (copyout(&frame, fp, sizeof(frame)) != 0) { 240 /* 241 * Process has trashed its stack; give it an illegal 242 * instruction to halt it in its tracks. 243 */ 244 sigexit(p, SIGILL); 245 /* NOTREACHED */ 246 } 247 248 /* 249 * Build context to run handler in. 250 */ 251 regs->tf_esp = (int)fp; 252 regs->tf_eip = (int)(((char *)PS_STRINGS) - *(p->p_sysent->sv_szsigcode)); 253 regs->tf_eflags &= ~PSL_VM; 254 regs->tf_cs = _ucodesel; 255 regs->tf_ds = _udatasel; 256 regs->tf_es = _udatasel; 257 regs->tf_ss = _udatasel; 258 } 259 260 /* 261 * System call to cleanup state after a signal 262 * has been taken. Reset signal mask and 263 * stack state from context left by sendsig (above). 264 * Return to previous pc and psl as specified by 265 * context left by sendsig. Check carefully to 266 * make sure that the user has not modified the 267 * psl to gain improper privileges or to cause 268 * a machine fault. 269 */ 270 int 271 linux_sigreturn(p, args, retval) 272 struct proc *p; 273 struct linux_sigreturn_args *args; 274 int *retval; 275 { 276 struct linux_sigcontext *scp, context; 277 register struct trapframe *regs; 278 int eflags; 279 280 regs = p->p_md.md_regs; 281 282 #ifdef DEBUG 283 printf("Linux-emul(%d): linux_sigreturn(%8x)\n", p->p_pid, args->scp); 284 #endif 285 /* 286 * The trampoline code hands us the context. 287 * It is unsafe to keep track of it ourselves, in the event that a 288 * program jumps out of a signal handler. 289 */ 290 scp = args->scp; 291 if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0) 292 return (EFAULT); 293 294 /* 295 * Check for security violations. 296 */ 297 #define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0) 298 eflags = context.sc_eflags; 299 /* 300 * XXX do allow users to change the privileged flag PSL_RF. The 301 * cpu sets PSL_RF in tf_eflags for faults. Debuggers should 302 * sometimes set it there too. tf_eflags is kept in the signal 303 * context during signal handling and there is no other place 304 * to remember it, so the PSL_RF bit may be corrupted by the 305 * signal handler without us knowing. Corruption of the PSL_RF 306 * bit at worst causes one more or one less debugger trap, so 307 * allowing it is fairly harmless. 308 */ 309 if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs->tf_eflags & ~PSL_RF)) { 310 return(EINVAL); 311 } 312 313 /* 314 * Don't allow users to load a valid privileged %cs. Let the 315 * hardware check for invalid selectors, excess privilege in 316 * other selectors, invalid %eip's and invalid %esp's. 317 */ 318 #define CS_SECURE(cs) (ISPL(cs) == SEL_UPL) 319 if (!CS_SECURE(context.sc_cs)) { 320 trapsignal(p, SIGBUS, T_PROTFLT); 321 return(EINVAL); 322 } 323 324 p->p_sigacts->ps_sigstk.ss_flags &= ~SS_ONSTACK; 325 p->p_sigmask = context.sc_mask &~ 326 (sigmask(SIGKILL)|sigmask(SIGCONT)|sigmask(SIGSTOP)); 327 /* 328 * Restore signal context. 329 */ 330 /* %fs and %gs were restored by the trampoline. */ 331 regs->tf_es = context.sc_es; 332 regs->tf_ds = context.sc_ds; 333 regs->tf_edi = context.sc_edi; 334 regs->tf_esi = context.sc_esi; 335 regs->tf_ebp = context.sc_ebp; 336 regs->tf_ebx = context.sc_ebx; 337 regs->tf_edx = context.sc_edx; 338 regs->tf_ecx = context.sc_ecx; 339 regs->tf_eax = context.sc_eax; 340 regs->tf_eip = context.sc_eip; 341 regs->tf_cs = context.sc_cs; 342 regs->tf_eflags = eflags; 343 regs->tf_esp = context.sc_esp_at_signal; 344 regs->tf_ss = context.sc_ss; 345 346 return (EJUSTRETURN); 347 } 348 349 void 350 linux_prepsyscall(struct trapframe *tf, int *args, u_int *code, caddr_t *params) 351 { 352 args[0] = tf->tf_ebx; 353 args[1] = tf->tf_ecx; 354 args[2] = tf->tf_edx; 355 args[3] = tf->tf_esi; 356 args[4] = tf->tf_edi; 357 *params = NULL; /* no copyin */ 358 } 359 360 struct sysentvec linux_sysvec = { 361 LINUX_SYS_MAXSYSCALL, 362 linux_sysent, 363 0xff, 364 NSIG, 365 bsd_to_linux_signal, 366 ELAST, 367 bsd_to_linux_errno, 368 linux_fixup, 369 linux_sendsig, 370 linux_sigcode, 371 &linux_szsigcode, 372 linux_prepsyscall, 373 "Linux a.out" 374 }; 375 376 struct sysentvec elf_linux_sysvec = { 377 LINUX_SYS_MAXSYSCALL, 378 linux_sysent, 379 0xff, 380 NSIG, 381 bsd_to_linux_signal, 382 ELAST, 383 bsd_to_linux_errno, 384 elf_linux_fixup, 385 linux_sendsig, 386 linux_sigcode, 387 &linux_szsigcode, 388 linux_prepsyscall, 389 "Linux ELF" 390 }; 391 392 /* 393 * Installed either via SYSINIT() or via LKM stubs. 394 */ 395 Elf32_Brandinfo linux_brand = { 396 "Linux", 397 "/compat/linux", 398 "/lib/ld-linux.so.1", 399 &elf_linux_sysvec 400 }; 401 402 #ifndef LKM 403 /* 404 * XXX: this is WRONG, it needs to be SI_SUB_EXEC, but this is just at the 405 * "proof of concept" stage and will be fixed shortly 406 */ 407 static void linux_elf_init __P((void *dummy)); 408 409 static void 410 linux_elf_init(dummy) 411 void *dummy; 412 { 413 if (elf_insert_brand_entry(&linux_brand) < 0) 414 printf("cannot insert Linux elf brand handler\n"); 415 else if (bootverbose) 416 printf("Linux-ELF exec handler installed\n"); 417 } 418 419 SYSINIT(linuxelf, SI_SUB_VFS, SI_ORDER_ANY, linux_elf_init, NULL); 420 #endif 421