netbsd32_execve.c revision 1.16
1/*	$NetBSD: netbsd32_execve.c,v 1.16 2003/06/29 13:35:38 martin Exp $	*/
2
3/*
4 * Copyright (c) 1998, 2001 Matthew R. Green
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. The name of the author may not be used to endorse or promote products
16 *    derived from this software without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
23 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
24 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
25 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
26 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * SUCH DAMAGE.
29 */
30
31#include <sys/cdefs.h>
32__KERNEL_RCSID(0, "$NetBSD: netbsd32_execve.c,v 1.16 2003/06/29 13:35:38 martin Exp $");
33
34#if defined(_KERNEL_OPT)
35#include "opt_ktrace.h"
36#endif
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/malloc.h>
41#include <sys/mount.h>
42#include <sys/stat.h>
43#include <sys/wait.h>
44#include <sys/ktrace.h>
45#include <sys/vnode.h>
46#include <sys/file.h>
47#include <sys/filedesc.h>
48#include <sys/namei.h>
49
50#include <uvm/uvm_extern.h>
51
52#include <sys/sa.h>
53#include <sys/syscallargs.h>
54#include <sys/proc.h>
55#include <sys/acct.h>
56#include <sys/exec.h>
57
58#include <compat/netbsd32/netbsd32.h>
59#include <compat/netbsd32/netbsd32_syscall.h>
60#include <compat/netbsd32/netbsd32_syscallargs.h>
61
62/* this is provided by kern/kern_exec.c */
63extern u_int exec_maxhdrsz;
64#if defined(LKM) || defined(_LKM)
65extern struct lock exec_lock;
66#endif
67
68/*
69 * Need to completly reimplement this syscall due to argument copying.
70 */
71/* ARGSUSED */
72int
73netbsd32_execve(l, v, retval)
74	struct lwp *l;
75	void *v;
76	register_t *retval;
77{
78	struct netbsd32_execve_args /* {
79		syscallarg(const netbsd32_charp) path;
80		syscallarg(netbsd32_charpp) argp;
81		syscallarg(netbsd32_charpp) envp;
82	} */ *uap = v;
83	struct sys_execve_args ua;
84	caddr_t sg;
85	struct proc *p = l->l_proc;
86
87	NETBSD32TOP_UAP(path, const char);
88	NETBSD32TOP_UAP(argp, char *);
89	NETBSD32TOP_UAP(envp, char *);
90	sg = stackgap_init(p, 0);
91	CHECK_ALT_EXIST(l, &sg, SCARG(&ua, path));
92
93	return netbsd32_execve2(l, &ua, retval);
94}
95
96int
97netbsd32_execve2(l, uap, retval)
98	struct lwp *l;
99	struct sys_execve_args *uap;
100	register_t *retval;
101{
102	/* Function args */
103	struct proc *p = l->l_proc;
104	int error, i;
105	struct exec_package pack;
106	struct nameidata nid;
107	struct vattr attr;
108	struct ucred *cred = p->p_ucred;
109	char *argp;
110	netbsd32_charp const *cpp;
111	char *dp;
112	netbsd32_charp sp;
113	long argc, envc;
114	size_t len;
115	char *stack;
116	struct ps_strings arginfo;
117	struct vmspace *vm;
118	char **tmpfap;
119	int szsigcode;
120	struct exec_vmcmd *base_vcp = NULL;
121
122	/*
123	 * Init the namei data to point the file user's program name.
124	 * This is done here rather than in check_exec(), so that it's
125	 * possible to override this settings if any of makecmd/probe
126	 * functions call check_exec() recursively - for example,
127	 * see exec_script_makecmds().
128	 */
129	NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l);
130
131	/*
132	 * initialize the fields of the exec package.
133	 */
134	pack.ep_name = SCARG(uap, path);
135	pack.ep_hdr = malloc(exec_maxhdrsz, M_EXEC, M_WAITOK);
136	pack.ep_hdrlen = exec_maxhdrsz;
137	pack.ep_hdrvalid = 0;
138	pack.ep_ndp = &nid;
139	pack.ep_emul_arg = NULL;
140	pack.ep_vmcmds.evs_cnt = 0;
141	pack.ep_vmcmds.evs_used = 0;
142	pack.ep_vap = &attr;
143	pack.ep_flags = 0;
144
145#if defined(LKM) || defined(_LKM)
146	lockmgr(&exec_lock, LK_SHARED, NULL);
147#endif
148
149	/* see if we can run it. */
150	if ((error = check_exec(l, &pack)) != 0)
151		goto freehdr;
152
153	/* XXX -- THE FOLLOWING SECTION NEEDS MAJOR CLEANUP */
154
155	/* allocate an argument buffer */
156	argp = (char *) uvm_km_valloc_wait(exec_map, NCARGS);
157#ifdef DIAGNOSTIC
158	if (argp == (vaddr_t) 0)
159		panic("netbsd32_execve: argp == NULL");
160#endif
161	dp = argp;
162	argc = 0;
163
164	/* copy the fake args list, if there's one, freeing it as we go */
165	if (pack.ep_flags & EXEC_HASARGL) {
166		tmpfap = pack.ep_fa;
167		while (*tmpfap != NULL) {
168			char *cp;
169
170			cp = *tmpfap;
171			while (*cp)
172				*dp++ = *cp++;
173			dp++;
174
175			FREE(*tmpfap, M_EXEC);
176			tmpfap++; argc++;
177		}
178		FREE(pack.ep_fa, M_EXEC);
179		pack.ep_flags &= ~EXEC_HASARGL;
180	}
181
182	/* Now get argv & environment */
183	if (!(cpp = (netbsd32_charp *)SCARG(uap, argp))) {
184		error = EINVAL;
185		goto bad;
186	}
187
188	if (pack.ep_flags & EXEC_SKIPARG)
189		cpp++;
190
191	while (1) {
192		len = argp + ARG_MAX - dp;
193		if ((error = copyin(cpp, &sp, sizeof(sp))) != 0)
194			goto bad;
195		if (!sp)
196			break;
197		if ((error = copyinstr((char *)(u_long)sp, dp,
198				       len, &len)) != 0) {
199			if (error == ENAMETOOLONG)
200				error = E2BIG;
201			goto bad;
202		}
203		dp += len;
204		cpp++;
205		argc++;
206	}
207
208	envc = 0;
209	/* environment need not be there */
210	if ((cpp = (netbsd32_charp *)SCARG(uap, envp)) != NULL ) {
211		while (1) {
212			len = argp + ARG_MAX - dp;
213			if ((error = copyin(cpp, &sp, sizeof(sp))) != 0)
214				goto bad;
215			if (!sp)
216				break;
217			if ((error = copyinstr((char *)(u_long)sp,
218					       dp, len, &len)) != 0) {
219				if (error == ENAMETOOLONG)
220					error = E2BIG;
221				goto bad;
222			}
223			dp += len;
224			cpp++;
225			envc++;
226		}
227	}
228
229	dp = (char *) ALIGN(dp);
230
231	szsigcode = pack.ep_es->es_emul->e_esigcode -
232	    pack.ep_es->es_emul->e_sigcode;
233
234	/* Now check if args & environ fit into new stack */
235	if (pack.ep_flags & EXEC_32)
236		len = ((argc + envc + 2 + pack.ep_es->es_arglen) *
237		    sizeof(int) + sizeof(int) + dp + STACKGAPLEN +
238		    szsigcode + sizeof(struct ps_strings)) - argp;
239	else
240		len = ((argc + envc + 2 + pack.ep_es->es_arglen) *
241		    sizeof(char *) + sizeof(int) + dp + STACKGAPLEN +
242		    szsigcode + sizeof(struct ps_strings)) - argp;
243
244	len = ALIGN(len);	/* make the stack "safely" aligned */
245
246	if (len > pack.ep_ssize) { /* in effect, compare to initial limit */
247		error = ENOMEM;
248		goto bad;
249	}
250
251	/* adjust "active stack depth" for process VSZ */
252	pack.ep_ssize = len;	/* maybe should go elsewhere, but... */
253
254	/*
255	 * Do whatever is necessary to prepare the address space
256	 * for remapping.  Note that this might replace the current
257	 * vmspace with another!
258	 */
259	uvmspace_exec(l, VM_MIN_ADDRESS, (vaddr_t)pack.ep_minsaddr);
260
261	/* Now map address space */
262	vm = p->p_vmspace;
263	vm->vm_taddr = (char *) pack.ep_taddr;
264	vm->vm_tsize = btoc(pack.ep_tsize);
265	vm->vm_daddr = (char *) pack.ep_daddr;
266	vm->vm_dsize = btoc(pack.ep_dsize);
267	vm->vm_ssize = btoc(pack.ep_ssize);
268	vm->vm_maxsaddr = (char *) pack.ep_maxsaddr;
269	vm->vm_minsaddr = (char *) pack.ep_minsaddr;
270
271	/* create the new process's VM space by running the vmcmds */
272#ifdef DIAGNOSTIC
273	if (pack.ep_vmcmds.evs_used == 0)
274		panic("netbsd32_execve: no vmcmds");
275#endif
276	for (i = 0; i < pack.ep_vmcmds.evs_used && !error; i++) {
277		struct exec_vmcmd *vcp;
278
279		vcp = &pack.ep_vmcmds.evs_cmds[i];
280		if (vcp->ev_flags & VMCMD_RELATIVE) {
281#ifdef DIAGNOSTIC
282			if (base_vcp == NULL)
283				panic("netbsd32_execve: relative vmcmd with no base");
284			if (vcp->ev_flags & VMCMD_BASE)
285				panic("netbsd32_execve: illegal base & relative vmcmd");
286#endif
287			vcp->ev_addr += base_vcp->ev_addr;
288		}
289		error = (*vcp->ev_proc)(l, vcp);
290#ifdef DEBUG
291		if (error) {
292			int j;
293
294			for (j = 0; j <= i; j++)
295				printf("vmcmd[%d] = %#lx/%#lx @ %#lx\n", j,
296				       vcp[j-i].ev_addr, vcp[j-i].ev_len,
297				       vcp[j-i].ev_offset);
298		}
299#endif
300		if (vcp->ev_flags & VMCMD_BASE)
301			base_vcp = vcp;
302	}
303
304	/* free the vmspace-creation commands, and release their references */
305	kill_vmcmds(&pack.ep_vmcmds);
306
307	/* if an error happened, deallocate and punt */
308	if (error) {
309#ifdef DEBUG
310		printf("netbsd32_execve: vmcmd %i failed: %d\n", i-1, error);
311#endif
312		goto exec_abort;
313	}
314
315	/* remember information about the process */
316	arginfo.ps_nargvstr = argc;
317	arginfo.ps_nenvstr = envc;
318
319	stack = (char *) (vm->vm_minsaddr - len);
320	/* Now copy argc, args & environ to new stack */
321	error = (*pack.ep_es->es_copyargs)(l, &pack, &arginfo,
322	    &stack, argp);
323	if (error) {
324#ifdef DEBUG
325		printf("netbsd32_execve: copyargs failed\n");
326#endif
327		goto exec_abort;
328	}
329	/* restore the stack back to its original point */
330	stack = (char *) (vm->vm_minsaddr - len);
331
332	/* fill process ps_strings info */
333	p->p_psstr = (struct ps_strings *)(vm->vm_minsaddr -
334	    sizeof(struct ps_strings));
335	p->p_psargv = offsetof(struct ps_strings, ps_argvstr);
336	p->p_psnargv = offsetof(struct ps_strings, ps_nargvstr);
337	p->p_psenv = offsetof(struct ps_strings, ps_envstr);
338	p->p_psnenv = offsetof(struct ps_strings, ps_nenvstr);
339
340	/* copy out the process's ps_strings structure */
341	if (copyout(&arginfo, (char *)p->p_psstr, sizeof(arginfo))) {
342#ifdef DEBUG
343		printf("netbsd32_execve: ps_strings copyout failed\n");
344#endif
345		goto exec_abort;
346	}
347
348	/* copy out the process's signal trapoline code */
349	if (szsigcode) {
350		if (copyout((char *)pack.ep_es->es_emul->e_sigcode,
351		    p->p_sigctx.ps_sigcode = (char *)p->p_psstr - szsigcode,
352		    szsigcode)) {
353#ifdef DEBUG
354			printf("netbsd32_execve: sig trampoline copyout failed\n");
355#endif
356			goto exec_abort;
357		}
358#ifdef PMAP_NEED_PROCWR
359		/* This is code. Let the pmap do what is needed. */
360		pmap_procwr(p, (vaddr_t)p->p_sigctx.ps_sigcode, szsigcode);
361#endif
362	}
363
364	stopprofclock(p);	/* stop profiling */
365	fdcloseexec(l);		/* handle close on exec */
366	execsigs(p);		/* reset catched signals */
367	l->l_ctxlink = NULL;	/* reset ucontext link */
368
369	/* set command name & other accounting info */
370	len = min(nid.ni_cnd.cn_namelen, MAXCOMLEN);
371	memcpy(p->p_comm, nid.ni_cnd.cn_nameptr, len);
372	p->p_comm[len] = 0;
373	p->p_acflag &= ~AFORK;
374
375	/* record proc's vnode, for use by procfs and others */
376        if (p->p_textvp)
377                vrele(p->p_textvp);
378	VREF(pack.ep_vp);
379	p->p_textvp = pack.ep_vp;
380
381	p->p_flag |= P_EXEC;
382	if (p->p_flag & P_PPWAIT) {
383		p->p_flag &= ~P_PPWAIT;
384		wakeup((caddr_t) p->p_pptr);
385	}
386
387	/*
388	 * deal with set[ug]id.
389	 * MNT_NOSUID has already been used to disable s[ug]id.
390	 */
391	if ((p->p_flag & P_TRACED) == 0 &&
392
393	    (((attr.va_mode & S_ISUID) != 0 &&
394	      p->p_ucred->cr_uid != attr.va_uid) ||
395
396	     ((attr.va_mode & S_ISGID) != 0 &&
397	      p->p_ucred->cr_gid != attr.va_gid))) {
398		/*
399		 * Mark the process as SUGID before we do
400		 * anything that might block.
401		 */
402		p_sugid(p);
403
404		p->p_ucred = crcopy(cred);
405#ifdef KTRACE
406		/*
407		 * If process is being ktraced, turn off - unless
408		 * root set it.
409		 */
410		if (p->p_tracep && !(p->p_traceflag & KTRFAC_ROOT))
411			ktrderef(p);
412#endif
413		if (attr.va_mode & S_ISUID)
414			p->p_ucred->cr_uid = attr.va_uid;
415		if (attr.va_mode & S_ISGID)
416			p->p_ucred->cr_gid = attr.va_gid;
417	} else
418		p->p_flag &= ~P_SUGID;
419	p->p_cred->p_svuid = p->p_ucred->cr_uid;
420	p->p_cred->p_svgid = p->p_ucred->cr_gid;
421
422	doexechooks(p);
423
424	uvm_km_free_wakeup(exec_map, (vaddr_t) argp, NCARGS);
425
426	PNBUF_PUT(nid.ni_cnd.cn_pnbuf);
427	vn_lock(pack.ep_vp, LK_EXCLUSIVE | LK_RETRY);
428	VOP_CLOSE(pack.ep_vp, FREAD, cred, l);
429	vput(pack.ep_vp);
430
431	/* setup new registers and do misc. setup. */
432	(*pack.ep_es->es_emul->e_setregs)(l, &pack, (u_long) stack);
433	if (pack.ep_es->es_setregs)
434		(*pack.ep_es->es_setregs)(l, &pack, (u_long) stack);
435
436	if (p->p_flag & P_TRACED)
437		psignal(p, SIGTRAP);
438
439	free(pack.ep_hdr, M_EXEC);
440
441	/*
442	 * Call emulation specific exec hook. This can setup setup per-process
443	 * p->p_emuldata or do any other per-process stuff an emulation needs.
444	 *
445	 * If we are executing process of different emulation than the
446	 * original forked process, call e_proc_exit() of the old emulation
447	 * first, then e_proc_exec() of new emulation. If the emulation is
448	 * same, the exec hook code should deallocate any old emulation
449	 * resources held previously by this process.
450	 */
451	if (p->p_emul && p->p_emul->e_proc_exit
452	    && p->p_emul != pack.ep_es->es_emul)
453		(*p->p_emul->e_proc_exit)(p);
454
455	/*
456	 * Call exec hook. Emulation code may NOT store reference to anything
457	 * from &pack.
458	 */
459        if (pack.ep_es->es_emul->e_proc_exec)
460                (*pack.ep_es->es_emul->e_proc_exec)(p, &pack);
461
462	/* update p_emul, the old value is no longer needed */
463	p->p_emul = pack.ep_es->es_emul;
464
465#ifdef KTRACE
466	if (KTRPOINT(p, KTR_EMUL))
467		ktremul(l);
468#endif
469
470#if defined(LKM) || defined(_LKM)
471	lockmgr(&exec_lock, LK_RELEASE, NULL);
472#endif
473
474	return (EJUSTRETURN);
475
476bad:
477	/* free the vmspace-creation commands, and release their references */
478	kill_vmcmds(&pack.ep_vmcmds);
479	/* kill any opened file descriptor, if necessary */
480	if (pack.ep_flags & EXEC_HASFD) {
481		pack.ep_flags &= ~EXEC_HASFD;
482		(void) fdrelease(l, pack.ep_fd);
483	}
484	/* close and put the exec'd file */
485	vn_lock(pack.ep_vp, LK_EXCLUSIVE | LK_RETRY);
486	VOP_CLOSE(pack.ep_vp, FREAD, cred, l);
487	vput(pack.ep_vp);
488	PNBUF_PUT(nid.ni_cnd.cn_pnbuf);
489	uvm_km_free_wakeup(exec_map, (vaddr_t) argp, NCARGS);
490
491freehdr:
492#if defined(LKM) || defined(_LKM)
493	lockmgr(&exec_lock, LK_RELEASE, NULL);
494#endif
495
496	free(pack.ep_hdr, M_EXEC);
497	return error;
498
499exec_abort:
500#if defined(LKM) || defined(_LKM)
501	lockmgr(&exec_lock, LK_RELEASE, NULL);
502#endif
503
504	/*
505	 * the old process doesn't exist anymore.  exit gracefully.
506	 * get rid of the (new) address space we have created, if any, get rid
507	 * of our namei data and vnode, and exit noting failure
508	 */
509	uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
510		VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
511	if (pack.ep_emul_arg)
512		FREE(pack.ep_emul_arg, M_TEMP);
513	PNBUF_PUT(nid.ni_cnd.cn_pnbuf);
514	vn_lock(pack.ep_vp, LK_EXCLUSIVE | LK_RETRY);
515	VOP_CLOSE(pack.ep_vp, FREAD, cred, l);
516	vput(pack.ep_vp);
517	uvm_km_free_wakeup(exec_map, (vaddr_t) argp, NCARGS);
518	free(pack.ep_hdr, M_EXEC);
519	exit1(l, W_EXITCODE(error, SIGABRT));
520
521	/* NOTREACHED */
522	return 0;
523}
524