xref: /freebsd/sys/i386/linux/linux_sysvec.c (revision 7573cb47cf4b1f33b6794f82a8a5e0bcbdcc1ada)
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 1.2 1996/03/10 08:42:48 sos 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/systm.h>
36 #include <sys/sysproto.h>
37 #include <sys/sysent.h>
38 #include <sys/imgact.h>
39 #include <sys/imgact_elf.h>
40 #include <sys/signalvar.h>
41 #include <sys/malloc.h>
42 #include <vm/vm.h>
43 #include <vm/vm_param.h>
44 #include <vm/vm_prot.h>
45 #include <vm/lock.h>
46 #include <vm/vm_kern.h>
47 #include <vm/vm_object.h>
48 #include <vm/vm_page.h>
49 #include <vm/vm_map.h>
50 #include <vm/vm_pager.h>
51 #include <vm/vm_extern.h>
52 #include <sys/user.h>
53 #include <sys/exec.h>
54 #include <sys/kernel.h>
55 #include <machine/cpu.h>
56 #include <machine/frame.h>
57 #include <machine/reg.h>
58 #include <machine/specialreg.h>
59 #include <machine/psl.h>
60 #include <machine/sysarch.h>
61 #include <machine/md_var.h>
62 
63 #include <i386/linux/linux.h>
64 #include <i386/linux/linux_proto.h>
65 
66 int	linux_fixup __P((int **stack_base, struct image_params *iparams));
67 int	elf_linux_fixup __P((int **stack_base, struct image_params *iparams));
68 void	linux_prepsyscall __P((struct trapframe *tf, int *args, u_int *code, caddr_t *params));
69 void    linux_sendsig __P((sig_t catcher, int sig, int mask, unsigned code));
70 
71 /*
72  * Linux syscalls return negative errno's, we do positive and map them
73  */
74 int bsd_to_linux_errno[ELAST] = {
75   	-0,  -1,  -2,  -3,  -4,  -5,  -6,  -7,  -8,  -9,
76  	-10, -35, -12, -13, -14, -15, -16, -17, -18, -19,
77  	-20, -21, -22, -23, -24, -25, -26, -27, -28, -29,
78  	-30, -31, -32, -33, -34, -11,-115,-114, -88, -89,
79  	-90, -91, -92, -93, -94, -95, -96, -97, -98, -99,
80 	-100,-101,-102,-103,-104,-105,-106,-107,-108,-109,
81 	-110,-111, -40, -36,-112,-113, -39, -11, -87,-122,
82 	-116, -66,  -6,  -6,  -6,  -6,  -6, -37, -38,  -9,
83   	-6,
84 };
85 
86 int bsd_to_linux_signal[NSIG] = {
87 	0, LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT,
88 	LINUX_SIGILL, LINUX_SIGTRAP, LINUX_SIGABRT, 0,
89 	LINUX_SIGFPE, LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV,
90 	0, LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM,
91 	LINUX_SIGURG, LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT,
92 	LINUX_SIGCHLD, LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO,
93 	LINUX_SIGXCPU, LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF,
94 	LINUX_SIGWINCH, 0, LINUX_SIGUSR1, LINUX_SIGUSR2
95 };
96 
97 int linux_to_bsd_signal[LINUX_NSIG] = {
98 	0, SIGHUP, SIGINT, SIGQUIT, SIGILL, SIGTRAP, SIGABRT, SIGEMT,
99 	SIGFPE, SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2, SIGPIPE, SIGALRM, SIGTERM,
100 	SIGBUS, SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU, SIGIO,
101 	SIGXCPU, SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH, SIGURG, SIGURG, 0
102 };
103 
104 int linux_fixup(int **stack_base, struct image_params *imgp)
105 {
106 	int *argv, *envp;
107 
108 	argv = *stack_base;
109 	envp = *stack_base + (imgp->argc + 1);
110 	(*stack_base)--;
111 	**stack_base = (int)envp;
112 	(*stack_base)--;
113 	**stack_base = (int)argv;
114 	(*stack_base)--;
115 	**stack_base = (int)imgp->argc;
116 	return 0;
117 }
118 
119 int elf_linux_fixup(int **stack_base, struct image_params *imgp)
120 {
121 	Elf32_Auxargs *args = (Elf32_Auxargs *)imgp->auxargs;
122 	int *pos;
123 
124 	pos = *stack_base + (imgp->argc + imgp->envc + 2);
125 
126 	if (args->trace) {
127 		AUXARGS_ENTRY(pos, AT_DEBUG, 1);
128 	}
129 	if (args->execfd != -1) {
130 		AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
131 	}
132 	AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
133 	AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
134 	AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
135 	AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
136 	AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
137 	AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
138 	AUXARGS_ENTRY(pos, AT_BASE, args->base);
139 	AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_cred->p_ruid);
140 	AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_cred->p_svuid);
141 	AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_cred->p_rgid);
142 	AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_cred->p_svgid);
143 	AUXARGS_ENTRY(pos, AT_NULL, 0);
144 
145 	free(imgp->auxargs, M_TEMP);
146 	imgp->auxargs = NULL;
147 
148 	(*stack_base)--;
149 	**stack_base = (int)imgp->argc;
150 	return 0;
151 }
152 
153 extern int _ucodesel, _udatasel;
154 
155 /*
156  * Send an interrupt to process.
157  *
158  * Stack is set up to allow sigcode stored
159  * in u. to call routine, followed by kcall
160  * to sigreturn routine below.  After sigreturn
161  * resets the signal mask, the stack, and the
162  * frame pointer, it returns to the user
163  * specified pc, psl.
164  */
165 
166 void
167 linux_sendsig(sig_t catcher, int sig, int mask, unsigned code)
168 {
169 	register struct proc *p = curproc;
170 	register int *regs;
171 	struct linux_sigframe *fp, frame;
172 	struct sigacts *psp = p->p_sigacts;
173 	int oonstack;
174 
175 	regs = p->p_md.md_regs;
176 	oonstack = psp->ps_sigstk.ss_flags & SS_ONSTACK;
177 
178 #ifdef DEBUG
179 	printf("Linux-emul(%d): linux_sendsig(%8x, %d, %d, %d)\n",
180 		p->p_pid, catcher, sig, mask, code);
181 #endif
182 	/*
183 	 * Allocate space for the signal handler context.
184 	 */
185 	if ((psp->ps_flags & SAS_ALTSTACK) && !oonstack &&
186 	    (psp->ps_sigonstack & sigmask(sig))) {
187 		fp = (struct linux_sigframe *)(psp->ps_sigstk.ss_sp +
188 		    psp->ps_sigstk.ss_size - sizeof(struct linux_sigframe));
189 		psp->ps_sigstk.ss_flags |= SS_ONSTACK;
190 	} else {
191 		fp = (struct linux_sigframe *)regs[tESP] - 1;
192 	}
193 
194 	/*
195 	 * grow() will return FALSE if the fp will not fit inside the stack
196 	 *	and the stack can not be grown. useracc will return FALSE
197 	 *	if access is denied.
198 	 */
199 	if ((grow(p, (int)fp) == FALSE) ||
200 	    (useracc((caddr_t)fp, sizeof (struct linux_sigframe), B_WRITE) == FALSE)) {
201 		/*
202 		 * Process has trashed its stack; give it an illegal
203 		 * instruction to halt it in its tracks.
204 		 */
205 		SIGACTION(p, SIGILL) = SIG_DFL;
206 		sig = sigmask(SIGILL);
207 		p->p_sigignore &= ~sig;
208 		p->p_sigcatch &= ~sig;
209 		p->p_sigmask &= ~sig;
210 		psignal(p, SIGILL);
211 		return;
212 	}
213 
214 	/*
215 	 * Build the argument list for the signal handler.
216 	 */
217 	if (p->p_sysent->sv_sigtbl) {
218 		if (sig < p->p_sysent->sv_sigsize)
219 			sig = p->p_sysent->sv_sigtbl[sig];
220 		else
221 			sig = p->p_sysent->sv_sigsize + 1;
222 	}
223 
224 	frame.sf_handler = catcher;
225 	frame.sf_sig = sig;
226 
227 	/*
228 	 * Build the signal context to be used by sigreturn.
229 	 */
230 	frame.sf_sc.sc_mask   = mask;
231 	__asm("movl %%gs,%w0" : "=r" (frame.sf_sc.sc_gs));
232 	__asm("movl %%fs,%w0" : "=r" (frame.sf_sc.sc_fs));
233 	frame.sf_sc.sc_es     = regs[tES];
234 	frame.sf_sc.sc_ds     = regs[tDS];
235 	frame.sf_sc.sc_edi    = regs[tEDI];
236 	frame.sf_sc.sc_esi    = regs[tESI];
237 	frame.sf_sc.sc_ebp    = regs[tEBP];
238 	frame.sf_sc.sc_ebx    = regs[tEBX];
239 	frame.sf_sc.sc_edx    = regs[tEDX];
240 	frame.sf_sc.sc_ecx    = regs[tECX];
241 	frame.sf_sc.sc_eax    = regs[tEAX];
242 	frame.sf_sc.sc_eip    = regs[tEIP];
243 	frame.sf_sc.sc_cs     = regs[tCS];
244 	frame.sf_sc.sc_eflags = regs[tEFLAGS];
245 	frame.sf_sc.sc_esp_at_signal = regs[tESP];
246 	frame.sf_sc.sc_ss     = regs[tSS];
247 	frame.sf_sc.sc_err    = regs[tERR];
248 	frame.sf_sc.sc_trapno = code;	/* XXX ???? */
249 
250 	if (copyout(&frame, fp, sizeof(frame)) != 0) {
251 		/*
252 		 * Process has trashed its stack; give it an illegal
253 		 * instruction to halt it in its tracks.
254 		 */
255 		sigexit(p, SIGILL);
256 		/* NOTREACHED */
257 	}
258 
259 	/*
260 	 * Build context to run handler in.
261 	 */
262 	regs[tESP] = (int)fp;
263 	regs[tEIP] = (int)(((char *)PS_STRINGS) - *(p->p_sysent->sv_szsigcode));
264 	regs[tEFLAGS] &= ~PSL_VM;
265 	regs[tCS] = _ucodesel;
266 	regs[tDS] = _udatasel;
267 	regs[tES] = _udatasel;
268 	regs[tSS] = _udatasel;
269 }
270 
271 /*
272  * System call to cleanup state after a signal
273  * has been taken.  Reset signal mask and
274  * stack state from context left by sendsig (above).
275  * Return to previous pc and psl as specified by
276  * context left by sendsig. Check carefully to
277  * make sure that the user has not modified the
278  * psl to gain improper privileges or to cause
279  * a machine fault.
280  */
281 int
282 linux_sigreturn(p, args, retval)
283 	struct proc *p;
284 	struct linux_sigreturn_args *args;
285 	int *retval;
286 {
287 	struct linux_sigcontext *scp, context;
288 	register int *regs;
289 	int eflags;
290 
291 	regs = p->p_md.md_regs;
292 
293 #ifdef DEBUG
294 	printf("Linux-emul(%d): linux_sigreturn(%8x)\n", p->p_pid, args->scp);
295 #endif
296 	/*
297 	 * The trampoline code hands us the context.
298 	 * It is unsafe to keep track of it ourselves, in the event that a
299 	 * program jumps out of a signal handler.
300 	 */
301 	scp = args->scp;
302 	if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0)
303 		return (EFAULT);
304 
305 	/*
306 	 * Check for security violations.
307 	 */
308 #define	EFLAGS_SECURE(ef, oef)	((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
309 	eflags = context.sc_eflags;
310 	/*
311 	 * XXX do allow users to change the privileged flag PSL_RF.  The
312 	 * cpu sets PSL_RF in tf_eflags for faults.  Debuggers should
313 	 * sometimes set it there too.  tf_eflags is kept in the signal
314 	 * context during signal handling and there is no other place
315 	 * to remember it, so the PSL_RF bit may be corrupted by the
316 	 * signal handler without us knowing.  Corruption of the PSL_RF
317 	 * bit at worst causes one more or one less debugger trap, so
318 	 * allowing it is fairly harmless.
319 	 */
320 	if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs[tEFLAGS] & ~PSL_RF)) {
321     		return(EINVAL);
322 	}
323 
324 	/*
325 	 * Don't allow users to load a valid privileged %cs.  Let the
326 	 * hardware check for invalid selectors, excess privilege in
327 	 * other selectors, invalid %eip's and invalid %esp's.
328 	 */
329 #define	CS_SECURE(cs)	(ISPL(cs) == SEL_UPL)
330 	if (!CS_SECURE(context.sc_cs)) {
331 		trapsignal(p, SIGBUS, T_PROTFLT);
332 		return(EINVAL);
333 	}
334 
335 	p->p_sigacts->ps_sigstk.ss_flags &= ~SS_ONSTACK;
336 	p->p_sigmask = context.sc_mask &~
337 		(sigmask(SIGKILL)|sigmask(SIGCONT)|sigmask(SIGSTOP));
338 	/*
339 	 * Restore signal context.
340 	 */
341 	/* %fs and %gs were restored by the trampoline. */
342 	regs[tES]     = context.sc_es;
343 	regs[tDS]     = context.sc_ds;
344 	regs[tEDI]    = context.sc_edi;
345 	regs[tESI]    = context.sc_esi;
346 	regs[tEBP]    = context.sc_ebp;
347 	regs[tEBX]    = context.sc_ebx;
348 	regs[tEDX]    = context.sc_edx;
349 	regs[tECX]    = context.sc_ecx;
350 	regs[tEAX]    = context.sc_eax;
351 	regs[tEIP]    = context.sc_eip;
352 	regs[tCS]     = context.sc_cs;
353 	regs[tEFLAGS] = eflags;
354 	regs[tESP]    = context.sc_esp_at_signal;
355 	regs[tSS]     = context.sc_ss;
356 
357 	return (EJUSTRETURN);
358 }
359 
360 void
361 linux_prepsyscall(struct trapframe *tf, int *args, u_int *code, caddr_t *params)
362 {
363 	int i;
364 	args[0] = tf->tf_ebx;
365 	args[1] = tf->tf_ecx;
366 	args[2] = tf->tf_edx;
367 	args[3] = tf->tf_esi;
368 	args[4] = tf->tf_edi;
369 	*params = NULL;		/* no copyin */
370 }
371 
372 struct sysentvec linux_sysvec = {
373 	LINUX_SYS_MAXSYSCALL,
374 	linux_sysent,
375 	0xff,
376 	NSIG,
377 	bsd_to_linux_signal,
378 	ELAST,
379 	bsd_to_linux_errno,
380 	linux_fixup,
381 	linux_sendsig,
382 	linux_sigcode,
383 	&linux_szsigcode,
384 	linux_prepsyscall,
385 };
386 
387 struct sysentvec elf_linux_sysvec = {
388         LINUX_SYS_MAXSYSCALL,
389         linux_sysent,
390         0xff,
391         NSIG,
392         bsd_to_linux_signal,
393         ELAST,
394         bsd_to_linux_errno,
395         elf_linux_fixup,
396         linux_sendsig,
397         linux_sigcode,
398         &linux_szsigcode,
399         linux_prepsyscall,
400 };
401 
402 /*
403  * Installed either via SYSINIT() or via LKM stubs.
404  */
405 Elf32_Interp_info linux_interp = {
406 					&elf_linux_sysvec,
407 					"/lib/ld-linux.so.1",
408 					"/compat/linux"
409 				 };
410 
411 #ifndef LKM
412 /*
413  * XXX: this is WRONG, it needs to be SI_SUB_EXEC, but this is just at the
414  * "proof of concept" stage and will be fixed shortly
415  */
416 SYSINIT(linuxelf, SI_SUB_VFS, SI_ORDER_ANY, elf_insert_interp, &linux_interp);
417 #endif
418