linux_sysvec.c revision 27557
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 1.14 1997/05/07 20:05:45 peter 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#include <machine/psl.h>
52
53#include <i386/linux/linux.h>
54#include <i386/linux/linux_proto.h>
55
56int	linux_fixup __P((int **stack_base, struct image_params *iparams));
57int	elf_linux_fixup __P((int **stack_base, struct image_params *iparams));
58void	linux_prepsyscall __P((struct trapframe *tf, int *args, u_int *code, caddr_t *params));
59void    linux_sendsig __P((sig_t catcher, int sig, int mask, u_long code));
60static void linux_elf_init __P((void *dummy));
61
62/*
63 * Linux syscalls return negative errno's, we do positive and map them
64 */
65int bsd_to_linux_errno[ELAST] = {
66  	-0,  -1,  -2,  -3,  -4,  -5,  -6,  -7,  -8,  -9,
67 	-10, -35, -12, -13, -14, -15, -16, -17, -18, -19,
68 	-20, -21, -22, -23, -24, -25, -26, -27, -28, -29,
69 	-30, -31, -32, -33, -34, -11,-115,-114, -88, -89,
70 	-90, -91, -92, -93, -94, -95, -96, -97, -98, -99,
71	-100,-101,-102,-103,-104,-105,-106,-107,-108,-109,
72	-110,-111, -40, -36,-112,-113, -39, -11, -87,-122,
73	-116, -66,  -6,  -6,  -6,  -6,  -6, -37, -38,  -9,
74  	-6,
75};
76
77int bsd_to_linux_signal[NSIG] = {
78	0, LINUX_SIGHUP, LINUX_SIGINT, LINUX_SIGQUIT,
79	LINUX_SIGILL, LINUX_SIGTRAP, LINUX_SIGABRT, 0,
80	LINUX_SIGFPE, LINUX_SIGKILL, LINUX_SIGBUS, LINUX_SIGSEGV,
81	0, LINUX_SIGPIPE, LINUX_SIGALRM, LINUX_SIGTERM,
82	LINUX_SIGURG, LINUX_SIGSTOP, LINUX_SIGTSTP, LINUX_SIGCONT,
83	LINUX_SIGCHLD, LINUX_SIGTTIN, LINUX_SIGTTOU, LINUX_SIGIO,
84	LINUX_SIGXCPU, LINUX_SIGXFSZ, LINUX_SIGVTALRM, LINUX_SIGPROF,
85	LINUX_SIGWINCH, 0, LINUX_SIGUSR1, LINUX_SIGUSR2
86};
87
88int linux_to_bsd_signal[LINUX_NSIG] = {
89	0, SIGHUP, SIGINT, SIGQUIT, SIGILL, SIGTRAP, SIGABRT, SIGEMT,
90	SIGFPE, SIGKILL, SIGUSR1, SIGSEGV, SIGUSR2, SIGPIPE, SIGALRM, SIGTERM,
91	SIGBUS, SIGCHLD, SIGCONT, SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU, SIGIO,
92	SIGXCPU, SIGXFSZ, SIGVTALRM, SIGPROF, SIGWINCH, SIGURG, SIGURG, 0
93};
94
95int linux_fixup(int **stack_base, struct image_params *imgp)
96{
97	int *argv, *envp;
98
99	argv = *stack_base;
100	envp = *stack_base + (imgp->argc + 1);
101	(*stack_base)--;
102	**stack_base = (int)envp;
103	(*stack_base)--;
104	**stack_base = (int)argv;
105	(*stack_base)--;
106	**stack_base = (int)imgp->argc;
107	return 0;
108}
109
110int elf_linux_fixup(int **stack_base, struct image_params *imgp)
111{
112	Elf32_Auxargs *args = (Elf32_Auxargs *)imgp->auxargs;
113	int *pos;
114
115	pos = *stack_base + (imgp->argc + imgp->envc + 2);
116
117	if (args->trace) {
118		AUXARGS_ENTRY(pos, AT_DEBUG, 1);
119	}
120	if (args->execfd != -1) {
121		AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
122	}
123	AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
124	AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
125	AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
126	AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
127	AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
128	AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
129	AUXARGS_ENTRY(pos, AT_BASE, args->base);
130	AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_cred->p_ruid);
131	AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_cred->p_svuid);
132	AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_cred->p_rgid);
133	AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_cred->p_svgid);
134	AUXARGS_ENTRY(pos, AT_NULL, 0);
135
136	free(imgp->auxargs, M_TEMP);
137	imgp->auxargs = NULL;
138
139	(*stack_base)--;
140	**stack_base = (int)imgp->argc;
141	return 0;
142}
143
144extern int _ucodesel, _udatasel;
145
146/*
147 * Send an interrupt to process.
148 *
149 * Stack is set up to allow sigcode stored
150 * in u. to call routine, followed by kcall
151 * to sigreturn routine below.  After sigreturn
152 * resets the signal mask, the stack, and the
153 * frame pointer, it returns to the user
154 * specified pc, psl.
155 */
156
157void
158linux_sendsig(sig_t catcher, int sig, int mask, u_long code)
159{
160	register struct proc *p = curproc;
161	register struct trapframe *regs;
162	struct linux_sigframe *fp, frame;
163	struct sigacts *psp = p->p_sigacts;
164	int oonstack;
165
166	regs = p->p_md.md_regs;
167	oonstack = psp->ps_sigstk.ss_flags & SS_ONSTACK;
168
169#ifdef DEBUG
170	printf("Linux-emul(%d): linux_sendsig(%8x, %d, %d, %ld)\n",
171		p->p_pid, catcher, sig, mask, code);
172#endif
173	/*
174	 * Allocate space for the signal handler context.
175	 */
176	if ((psp->ps_flags & SAS_ALTSTACK) && !oonstack &&
177	    (psp->ps_sigonstack & sigmask(sig))) {
178		fp = (struct linux_sigframe *)(psp->ps_sigstk.ss_sp +
179		    psp->ps_sigstk.ss_size - sizeof(struct linux_sigframe));
180		psp->ps_sigstk.ss_flags |= SS_ONSTACK;
181	} else {
182		fp = (struct linux_sigframe *)regs->tf_esp - 1;
183	}
184
185	/*
186	 * grow() will return FALSE if the fp will not fit inside the stack
187	 *	and the stack can not be grown. useracc will return FALSE
188	 *	if access is denied.
189	 */
190	if ((grow(p, (int)fp) == FALSE) ||
191	    (useracc((caddr_t)fp, sizeof (struct linux_sigframe), B_WRITE) == FALSE)) {
192		/*
193		 * Process has trashed its stack; give it an illegal
194		 * instruction to halt it in its tracks.
195		 */
196		SIGACTION(p, SIGILL) = SIG_DFL;
197		sig = sigmask(SIGILL);
198		p->p_sigignore &= ~sig;
199		p->p_sigcatch &= ~sig;
200		p->p_sigmask &= ~sig;
201		psignal(p, SIGILL);
202		return;
203	}
204
205	/*
206	 * Build the argument list for the signal handler.
207	 */
208	if (p->p_sysent->sv_sigtbl) {
209		if (sig < p->p_sysent->sv_sigsize)
210			sig = p->p_sysent->sv_sigtbl[sig];
211		else
212			sig = p->p_sysent->sv_sigsize + 1;
213	}
214
215	frame.sf_handler = catcher;
216	frame.sf_sig = sig;
217
218	/*
219	 * Build the signal context to be used by sigreturn.
220	 */
221	frame.sf_sc.sc_mask   = mask;
222	__asm("movl %%gs,%w0" : "=r" (frame.sf_sc.sc_gs));
223	__asm("movl %%fs,%w0" : "=r" (frame.sf_sc.sc_fs));
224	frame.sf_sc.sc_es     = regs->tf_es;
225	frame.sf_sc.sc_ds     = regs->tf_ds;
226	frame.sf_sc.sc_edi    = regs->tf_edi;
227	frame.sf_sc.sc_esi    = regs->tf_esi;
228	frame.sf_sc.sc_ebp    = regs->tf_ebp;
229	frame.sf_sc.sc_ebx    = regs->tf_ebx;
230	frame.sf_sc.sc_edx    = regs->tf_edx;
231	frame.sf_sc.sc_ecx    = regs->tf_ecx;
232	frame.sf_sc.sc_eax    = regs->tf_eax;
233	frame.sf_sc.sc_eip    = regs->tf_eip;
234	frame.sf_sc.sc_cs     = regs->tf_cs;
235	frame.sf_sc.sc_eflags = regs->tf_eflags;
236	frame.sf_sc.sc_esp_at_signal = regs->tf_esp;
237	frame.sf_sc.sc_ss     = regs->tf_ss;
238	frame.sf_sc.sc_err    = regs->tf_err;
239	frame.sf_sc.sc_trapno = code;	/* XXX ???? */
240
241	if (copyout(&frame, fp, sizeof(frame)) != 0) {
242		/*
243		 * Process has trashed its stack; give it an illegal
244		 * instruction to halt it in its tracks.
245		 */
246		sigexit(p, SIGILL);
247		/* NOTREACHED */
248	}
249
250	/*
251	 * Build context to run handler in.
252	 */
253	regs->tf_esp = (int)fp;
254	regs->tf_eip = (int)(((char *)PS_STRINGS) - *(p->p_sysent->sv_szsigcode));
255	regs->tf_eflags &= ~PSL_VM;
256	regs->tf_cs = _ucodesel;
257	regs->tf_ds = _udatasel;
258	regs->tf_es = _udatasel;
259	regs->tf_ss = _udatasel;
260}
261
262/*
263 * System call to cleanup state after a signal
264 * has been taken.  Reset signal mask and
265 * stack state from context left by sendsig (above).
266 * Return to previous pc and psl as specified by
267 * context left by sendsig. Check carefully to
268 * make sure that the user has not modified the
269 * psl to gain improper privileges or to cause
270 * a machine fault.
271 */
272int
273linux_sigreturn(p, args, retval)
274	struct proc *p;
275	struct linux_sigreturn_args *args;
276	int *retval;
277{
278	struct linux_sigcontext *scp, context;
279	register struct trapframe *regs;
280	int eflags;
281
282	regs = p->p_md.md_regs;
283
284#ifdef DEBUG
285	printf("Linux-emul(%d): linux_sigreturn(%8x)\n", p->p_pid, args->scp);
286#endif
287	/*
288	 * The trampoline code hands us the context.
289	 * It is unsafe to keep track of it ourselves, in the event that a
290	 * program jumps out of a signal handler.
291	 */
292	scp = args->scp;
293	if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0)
294		return (EFAULT);
295
296	/*
297	 * Check for security violations.
298	 */
299#define	EFLAGS_SECURE(ef, oef)	((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
300	eflags = context.sc_eflags;
301	/*
302	 * XXX do allow users to change the privileged flag PSL_RF.  The
303	 * cpu sets PSL_RF in tf_eflags for faults.  Debuggers should
304	 * sometimes set it there too.  tf_eflags is kept in the signal
305	 * context during signal handling and there is no other place
306	 * to remember it, so the PSL_RF bit may be corrupted by the
307	 * signal handler without us knowing.  Corruption of the PSL_RF
308	 * bit at worst causes one more or one less debugger trap, so
309	 * allowing it is fairly harmless.
310	 */
311	if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs->tf_eflags & ~PSL_RF)) {
312    		return(EINVAL);
313	}
314
315	/*
316	 * Don't allow users to load a valid privileged %cs.  Let the
317	 * hardware check for invalid selectors, excess privilege in
318	 * other selectors, invalid %eip's and invalid %esp's.
319	 */
320#define	CS_SECURE(cs)	(ISPL(cs) == SEL_UPL)
321	if (!CS_SECURE(context.sc_cs)) {
322		trapsignal(p, SIGBUS, T_PROTFLT);
323		return(EINVAL);
324	}
325
326	p->p_sigacts->ps_sigstk.ss_flags &= ~SS_ONSTACK;
327	p->p_sigmask = context.sc_mask &~
328		(sigmask(SIGKILL)|sigmask(SIGCONT)|sigmask(SIGSTOP));
329	/*
330	 * Restore signal context.
331	 */
332	/* %fs and %gs were restored by the trampoline. */
333	regs->tf_es     = context.sc_es;
334	regs->tf_ds     = context.sc_ds;
335	regs->tf_edi    = context.sc_edi;
336	regs->tf_esi    = context.sc_esi;
337	regs->tf_ebp    = context.sc_ebp;
338	regs->tf_ebx    = context.sc_ebx;
339	regs->tf_edx    = context.sc_edx;
340	regs->tf_ecx    = context.sc_ecx;
341	regs->tf_eax    = context.sc_eax;
342	regs->tf_eip    = context.sc_eip;
343	regs->tf_cs     = context.sc_cs;
344	regs->tf_eflags = eflags;
345	regs->tf_esp    = context.sc_esp_at_signal;
346	regs->tf_ss     = context.sc_ss;
347
348	return (EJUSTRETURN);
349}
350
351void
352linux_prepsyscall(struct trapframe *tf, int *args, u_int *code, caddr_t *params)
353{
354	args[0] = tf->tf_ebx;
355	args[1] = tf->tf_ecx;
356	args[2] = tf->tf_edx;
357	args[3] = tf->tf_esi;
358	args[4] = tf->tf_edi;
359	*params = NULL;		/* no copyin */
360}
361
362struct sysentvec linux_sysvec = {
363	LINUX_SYS_MAXSYSCALL,
364	linux_sysent,
365	0xff,
366	NSIG,
367	bsd_to_linux_signal,
368	ELAST,
369	bsd_to_linux_errno,
370	linux_fixup,
371	linux_sendsig,
372	linux_sigcode,
373	&linux_szsigcode,
374	linux_prepsyscall,
375	"Linux a.out"
376};
377
378struct sysentvec elf_linux_sysvec = {
379        LINUX_SYS_MAXSYSCALL,
380        linux_sysent,
381        0xff,
382        NSIG,
383        bsd_to_linux_signal,
384        ELAST,
385        bsd_to_linux_errno,
386        elf_linux_fixup,
387        linux_sendsig,
388        linux_sigcode,
389        &linux_szsigcode,
390        linux_prepsyscall,
391	"Linux ELF"
392};
393
394/*
395 * Installed either via SYSINIT() or via LKM stubs.
396 */
397Elf32_Brandinfo linux_brand = {
398					"Linux",
399					"/compat/linux",
400					"/lib/ld-linux.so.1",
401					&elf_linux_sysvec
402				 };
403
404#ifndef LKM
405/*
406 * XXX: this is WRONG, it needs to be SI_SUB_EXEC, but this is just at the
407 * "proof of concept" stage and will be fixed shortly
408 */
409static void
410linux_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
419SYSINIT(linuxelf, SI_SUB_VFS, SI_ORDER_ANY, linux_elf_init, NULL);
420#endif
421