xref: /freebsd/sys/i386/linux/linux_sysvec.c (revision a14a0223ae1b172e96dd2a1d849e22026a98b692)
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  * $FreeBSD$
29  */
30 
31 /* XXX we use functions that might not exist. */
32 #include "opt_compat.h"
33 
34 #ifndef COMPAT_43
35 #error "Unable to compile Linux-emulator due to missing COMPAT_43 option!"
36 #endif
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/buf.h>
41 #include <sys/proc.h>
42 #include <sys/sysent.h>
43 #include <sys/imgact.h>
44 #include <sys/imgact_aout.h>
45 #include <sys/imgact_elf.h>
46 #include <sys/signalvar.h>
47 #include <sys/malloc.h>
48 #include <vm/vm.h>
49 #include <vm/vm_param.h>
50 #include <vm/vm_page.h>
51 #include <vm/vm_extern.h>
52 #include <sys/exec.h>
53 #include <sys/kernel.h>
54 #include <sys/module.h>
55 #include <machine/cpu.h>
56 
57 #include <i386/linux/linux.h>
58 #include <i386/linux/linux_proto.h>
59 
60 static int	linux_fixup __P((long **stack_base,
61 				 struct image_params *iparams));
62 static int	elf_linux_fixup __P((long **stack_base,
63 				     struct image_params *iparams));
64 static void	linux_prepsyscall __P((struct trapframe *tf, int *args,
65 				       u_int *code, caddr_t *params));
66 static void     linux_sendsig __P((sig_t catcher, int sig, sigset_t *mask,
67 				   u_long code));
68 
69 /*
70  * Linux syscalls return negative errno's, we do positive and map them
71  */
72 static int bsd_to_linux_errno[ELAST + 1] = {
73   	-0,  -1,  -2,  -3,  -4,  -5,  -6,  -7,  -8,  -9,
74  	-10, -35, -12, -13, -14, -15, -16, -17, -18, -19,
75  	-20, -21, -22, -23, -24, -25, -26, -27, -28, -29,
76  	-30, -31, -32, -33, -34, -11,-115,-114, -88, -89,
77  	-90, -91, -92, -93, -94, -95, -96, -97, -98, -99,
78 	-100,-101,-102,-103,-104,-105,-106,-107,-108,-109,
79 	-110,-111, -40, -36,-112,-113, -39, -11, -87,-122,
80 	-116, -66,  -6,  -6,  -6,  -6,  -6, -37, -38,  -9,
81   	-6, -6, -43, -42, -75, -6, -84
82 };
83 
84 int bsd_to_linux_signal[LINUX_SIGTBLSZ] = {
85 	LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT, LINUX_SIGILL,
86 	LINUX_SIGTRAP, LINUX_SIGABRT, 0, LINUX_SIGFPE,
87 	LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV, 0,
88 	LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM, LINUX_SIGURG,
89 	LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT, LINUX_SIGCHLD,
90 	LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO, LINUX_SIGXCPU,
91 	LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF, LINUX_SIGWINCH,
92 	0, LINUX_SIGUSR1, LINUX_SIGUSR2
93 };
94 
95 int linux_to_bsd_signal[LINUX_SIGTBLSZ] = {
96 	SIGHUP, SIGINT, SIGQUIT, SIGILL,
97 	SIGTRAP, SIGABRT, SIGBUS, SIGFPE,
98 	SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2,
99 	SIGPIPE, SIGALRM, SIGTERM, SIGBUS,
100 	SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP,
101 	SIGTTIN, SIGTTOU, SIGURG, SIGXCPU,
102 	SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH,
103 	SIGIO, SIGURG, 0
104 };
105 
106 /*
107  * If FreeBSD & Linux have a difference of opinion about what a trap
108  * means, deal with it here.
109  */
110 static int
111 translate_traps(int signal, int trap_code)
112 {
113 	if (signal != SIGBUS)
114 		return signal;
115 	switch (trap_code) {
116 	case T_PROTFLT:
117 	case T_TSSFLT:
118 	case T_DOUBLEFLT:
119 	case T_PAGEFLT:
120 		return SIGSEGV;
121 	default:
122 		return signal;
123 	}
124 }
125 
126 static int
127 linux_fixup(long **stack_base, struct image_params *imgp)
128 {
129 	long *argv, *envp;
130 
131 	argv = *stack_base;
132 	envp = *stack_base + (imgp->argc + 1);
133 	(*stack_base)--;
134 	**stack_base = (intptr_t)(void *)envp;
135 	(*stack_base)--;
136 	**stack_base = (intptr_t)(void *)argv;
137 	(*stack_base)--;
138 	**stack_base = imgp->argc;
139 	return 0;
140 }
141 
142 static int
143 elf_linux_fixup(long **stack_base, struct image_params *imgp)
144 {
145 	Elf32_Auxargs *args = (Elf32_Auxargs *)imgp->auxargs;
146 	long *pos;
147 
148 	pos = *stack_base + (imgp->argc + imgp->envc + 2);
149 
150 	if (args->trace) {
151 		AUXARGS_ENTRY(pos, AT_DEBUG, 1);
152 	}
153 	if (args->execfd != -1) {
154 		AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
155 	}
156 	AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
157 	AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
158 	AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
159 	AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
160 	AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
161 	AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
162 	AUXARGS_ENTRY(pos, AT_BASE, args->base);
163 	AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_cred->p_ruid);
164 	AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_cred->p_svuid);
165 	AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_cred->p_rgid);
166 	AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_cred->p_svgid);
167 	AUXARGS_ENTRY(pos, AT_NULL, 0);
168 
169 	free(imgp->auxargs, M_TEMP);
170 	imgp->auxargs = NULL;
171 
172 	(*stack_base)--;
173 	**stack_base = (long)imgp->argc;
174 	return 0;
175 }
176 
177 extern int _ucodesel, _udatasel;
178 
179 /*
180  * Send an interrupt to process.
181  *
182  * Stack is set up to allow sigcode stored
183  * in u. to call routine, followed by kcall
184  * to sigreturn routine below.  After sigreturn
185  * resets the signal mask, the stack, and the
186  * frame pointer, it returns to the user
187  * specified pc, psl.
188  */
189 
190 static void
191 linux_sendsig(sig_t catcher, int sig, sigset_t *mask, u_long code)
192 {
193 	register struct proc *p = curproc;
194 	register struct trapframe *regs;
195 	struct linux_sigframe *fp, frame;
196 	struct sigacts *psp = p->p_sigacts;
197 	int oonstack;
198 
199 	regs = p->p_md.md_regs;
200 	oonstack = p->p_sigstk.ss_flags & SS_ONSTACK;
201 
202 #ifdef DEBUG
203 	printf("Linux-emul(%ld): linux_sendsig(%p, %d, %p, %lu)\n",
204 	    (long)p->p_pid, catcher, sig, (void*)mask, code);
205 #endif
206 	/*
207 	 * Allocate space for the signal handler context.
208 	 */
209 	if ((p->p_flag & P_ALTSTACK) && !oonstack &&
210 	    SIGISMEMBER(psp->ps_sigonstack, sig)) {
211 		fp = (struct linux_sigframe *)(p->p_sigstk.ss_sp +
212 		    p->p_sigstk.ss_size - sizeof(struct linux_sigframe));
213 		p->p_sigstk.ss_flags |= SS_ONSTACK;
214 	} else {
215 		fp = (struct linux_sigframe *)regs->tf_esp - 1;
216 	}
217 
218 	/*
219 	 * grow() will return FALSE if the fp will not fit inside the stack
220 	 *	and the stack can not be grown. useracc will return FALSE
221 	 *	if access is denied.
222 	 */
223 	if ((grow_stack (p, (int)fp) == FALSE) ||
224 	    !useracc((caddr_t)fp, sizeof (struct linux_sigframe),
225 	    VM_PROT_WRITE)) {
226 		/*
227 		 * Process has trashed its stack; give it an illegal
228 		 * instruction to halt it in its tracks.
229 		 */
230 		SIGACTION(p, SIGILL) = SIG_DFL;
231 		SIGDELSET(p->p_sigignore, SIGILL);
232 		SIGDELSET(p->p_sigcatch, SIGILL);
233 		SIGDELSET(p->p_sigmask, SIGILL);
234 		psignal(p, SIGILL);
235 		return;
236 	}
237 
238 	/*
239 	 * Build the argument list for the signal handler.
240 	 */
241 	if (p->p_sysent->sv_sigtbl)
242 		if (sig <= p->p_sysent->sv_sigsize)
243 			sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];
244 
245 	frame.sf_handler = catcher;
246 	frame.sf_sig = sig;
247 
248 	/*
249 	 * Build the signal context to be used by sigreturn.
250 	 */
251 	frame.sf_sc.sc_mask   = mask->__bits[0];
252 	frame.sf_sc.sc_gs     = rgs();
253 	frame.sf_sc.sc_fs     = regs->tf_fs;
254 	frame.sf_sc.sc_es     = regs->tf_es;
255 	frame.sf_sc.sc_ds     = regs->tf_ds;
256 	frame.sf_sc.sc_edi    = regs->tf_edi;
257 	frame.sf_sc.sc_esi    = regs->tf_esi;
258 	frame.sf_sc.sc_ebp    = regs->tf_ebp;
259 	frame.sf_sc.sc_ebx    = regs->tf_ebx;
260 	frame.sf_sc.sc_edx    = regs->tf_edx;
261 	frame.sf_sc.sc_ecx    = regs->tf_ecx;
262 	frame.sf_sc.sc_eax    = regs->tf_eax;
263 	frame.sf_sc.sc_eip    = regs->tf_eip;
264 	frame.sf_sc.sc_cs     = regs->tf_cs;
265 	frame.sf_sc.sc_eflags = regs->tf_eflags;
266 	frame.sf_sc.sc_esp_at_signal = regs->tf_esp;
267 	frame.sf_sc.sc_ss     = regs->tf_ss;
268 	frame.sf_sc.sc_err    = regs->tf_err;
269 	frame.sf_sc.sc_trapno = code;	/* XXX ???? */
270 
271 	if (copyout(&frame, fp, sizeof(frame)) != 0) {
272 		/*
273 		 * Process has trashed its stack; give it an illegal
274 		 * instruction to halt it in its tracks.
275 		 */
276 		sigexit(p, SIGILL);
277 		/* NOTREACHED */
278 	}
279 
280 	/*
281 	 * Build context to run handler in.
282 	 */
283 	regs->tf_esp = (int)fp;
284 	regs->tf_eip = PS_STRINGS - *(p->p_sysent->sv_szsigcode);
285 	regs->tf_eflags &= ~PSL_VM;
286 	regs->tf_cs = _ucodesel;
287 	regs->tf_ds = _udatasel;
288 	regs->tf_es = _udatasel;
289 	regs->tf_fs = _udatasel;
290 	load_gs(_udatasel);
291 	regs->tf_ss = _udatasel;
292 }
293 
294 /*
295  * System call to cleanup state after a signal
296  * has been taken.  Reset signal mask and
297  * stack state from context left by sendsig (above).
298  * Return to previous pc and psl as specified by
299  * context left by sendsig. Check carefully to
300  * make sure that the user has not modified the
301  * psl to gain improper privileges or to cause
302  * a machine fault.
303  */
304 int
305 linux_sigreturn(p, args)
306 	struct proc *p;
307 	struct linux_sigreturn_args *args;
308 {
309 	struct linux_sigcontext *scp, context;
310 	register struct trapframe *regs;
311 	int eflags;
312 
313 	regs = p->p_md.md_regs;
314 
315 #ifdef DEBUG
316 	printf("Linux-emul(%ld): linux_sigreturn(%p)\n",
317 	    (long)p->p_pid, (void *)args->scp);
318 #endif
319 	/*
320 	 * The trampoline code hands us the context.
321 	 * It is unsafe to keep track of it ourselves, in the event that a
322 	 * program jumps out of a signal handler.
323 	 */
324 	scp = SCARG(args,scp);
325 	if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0)
326 		return (EFAULT);
327 
328 	/*
329 	 * Check for security violations.
330 	 */
331 #define	EFLAGS_SECURE(ef, oef)	((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
332 	eflags = context.sc_eflags;
333 	/*
334 	 * XXX do allow users to change the privileged flag PSL_RF.  The
335 	 * cpu sets PSL_RF in tf_eflags for faults.  Debuggers should
336 	 * sometimes set it there too.  tf_eflags is kept in the signal
337 	 * context during signal handling and there is no other place
338 	 * to remember it, so the PSL_RF bit may be corrupted by the
339 	 * signal handler without us knowing.  Corruption of the PSL_RF
340 	 * bit at worst causes one more or one less debugger trap, so
341 	 * allowing it is fairly harmless.
342 	 */
343 	if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs->tf_eflags & ~PSL_RF)) {
344     		return(EINVAL);
345 	}
346 
347 	/*
348 	 * Don't allow users to load a valid privileged %cs.  Let the
349 	 * hardware check for invalid selectors, excess privilege in
350 	 * other selectors, invalid %eip's and invalid %esp's.
351 	 */
352 #define	CS_SECURE(cs)	(ISPL(cs) == SEL_UPL)
353 	if (!CS_SECURE(context.sc_cs)) {
354 		trapsignal(p, SIGBUS, T_PROTFLT);
355 		return(EINVAL);
356 	}
357 
358 	p->p_sigstk.ss_flags &= ~SS_ONSTACK;
359 	SIGSETOLD(p->p_sigmask, context.sc_mask);
360 	SIG_CANTMASK(p->p_sigmask);
361 
362 	/*
363 	 * Restore signal context.
364 	 */
365 	/* %gs was restored by the trampoline. */
366 	regs->tf_fs     = context.sc_fs;
367 	regs->tf_es     = context.sc_es;
368 	regs->tf_ds     = context.sc_ds;
369 	regs->tf_edi    = context.sc_edi;
370 	regs->tf_esi    = context.sc_esi;
371 	regs->tf_ebp    = context.sc_ebp;
372 	regs->tf_ebx    = context.sc_ebx;
373 	regs->tf_edx    = context.sc_edx;
374 	regs->tf_ecx    = context.sc_ecx;
375 	regs->tf_eax    = context.sc_eax;
376 	regs->tf_eip    = context.sc_eip;
377 	regs->tf_cs     = context.sc_cs;
378 	regs->tf_eflags = eflags;
379 	regs->tf_esp    = context.sc_esp_at_signal;
380 	regs->tf_ss     = context.sc_ss;
381 
382 	return (EJUSTRETURN);
383 }
384 
385 static void
386 linux_prepsyscall(struct trapframe *tf, int *args, u_int *code, caddr_t *params)
387 {
388 	args[0] = tf->tf_ebx;
389 	args[1] = tf->tf_ecx;
390 	args[2] = tf->tf_edx;
391 	args[3] = tf->tf_esi;
392 	args[4] = tf->tf_edi;
393 	*params = NULL;		/* no copyin */
394 }
395 
396 struct sysentvec linux_sysvec = {
397 	LINUX_SYS_MAXSYSCALL,
398 	linux_sysent,
399 	0xff,
400 	LINUX_SIGTBLSZ,
401 	bsd_to_linux_signal,
402 	ELAST + 1,
403 	bsd_to_linux_errno,
404 	translate_traps,
405 	linux_fixup,
406 	linux_sendsig,
407 	linux_sigcode,
408 	&linux_szsigcode,
409 	linux_prepsyscall,
410 	"Linux a.out",
411 	aout_coredump
412 };
413 
414 struct sysentvec elf_linux_sysvec = {
415 	LINUX_SYS_MAXSYSCALL,
416 	linux_sysent,
417 	0xff,
418 	LINUX_SIGTBLSZ,
419 	bsd_to_linux_signal,
420 	ELAST + 1,
421 	bsd_to_linux_errno,
422 	translate_traps,
423 	elf_linux_fixup,
424 	linux_sendsig,
425 	linux_sigcode,
426 	&linux_szsigcode,
427 	linux_prepsyscall,
428 	"Linux ELF",
429 	elf_coredump
430 };
431 
432 static Elf32_Brandinfo linux_brand = {
433 					"Linux",
434 					"/compat/linux",
435 					"/lib/ld-linux.so.1",
436 					&elf_linux_sysvec
437 				 };
438 
439 static Elf32_Brandinfo linux_glibc2brand = {
440 					"Linux",
441 					"/compat/linux",
442 					"/lib/ld-linux.so.2",
443 					&elf_linux_sysvec
444 				 };
445 
446 Elf32_Brandinfo *linux_brandlist[] = {
447 					&linux_brand,
448 					&linux_glibc2brand,
449 					NULL
450 				};
451 
452 static int
453 linux_elf_modevent(module_t mod, int type, void *data)
454 {
455 	Elf32_Brandinfo **brandinfo;
456 	int error;
457 
458 	error = 0;
459 
460 	switch(type) {
461 	case MOD_LOAD:
462 		for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
463 		    ++brandinfo)
464 			if (elf_insert_brand_entry(*brandinfo) < 0)
465 				error = EINVAL;
466 		if (error)
467 			printf("cannot insert Linux elf brand handler\n");
468 		else if (bootverbose)
469 			printf("Linux-ELF exec handler installed\n");
470 		break;
471 	case MOD_UNLOAD:
472 		for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
473 		    ++brandinfo)
474 			if (elf_brand_inuse(*brandinfo)) {
475 				error = EBUSY;
476 			}
477 
478 		if (error == 0) {
479 			for (brandinfo = &linux_brandlist[0];
480 			    *brandinfo != NULL; ++brandinfo)
481 				if (elf_remove_brand_entry(*brandinfo) < 0)
482 					error = EINVAL;
483 		}
484 		if (error)
485 			printf("Could not deinstall ELF interpreter entry\n");
486 		else if (bootverbose)
487 			printf("Linux-elf exec handler removed\n");
488 		break;
489 	default:
490 		break;
491 	}
492 	return error;
493 }
494 static moduledata_t linux_elf_mod = {
495 	"linuxelf",
496 	linux_elf_modevent,
497 	0
498 };
499 DECLARE_MODULE(linuxelf, linux_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
500