netbsd32_execve.c revision 1.20
1/*	$NetBSD: netbsd32_execve.c,v 1.20 2005/04/01 11:59:36 yamt 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 1.20 2005/04/01 11:59:36 yamt 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(p, &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), p);
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(p, &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_alloc(exec_map, NCARGS, 0,
157	    UVM_KMF_PAGEABLE|UVM_KMF_WAITVA);
158#ifdef DIAGNOSTIC
159	if (argp == (vaddr_t) 0)
160		panic("netbsd32_execve: argp == NULL");
161#endif
162	dp = argp;
163	argc = 0;
164
165	/* copy the fake args list, if there's one, freeing it as we go */
166	if (pack.ep_flags & EXEC_HASARGL) {
167		tmpfap = pack.ep_fa;
168		while (*tmpfap != NULL) {
169			char *cp;
170
171			cp = *tmpfap;
172			while (*cp)
173				*dp++ = *cp++;
174			dp++;
175
176			FREE(*tmpfap, M_EXEC);
177			tmpfap++; argc++;
178		}
179		FREE(pack.ep_fa, M_EXEC);
180		pack.ep_flags &= ~EXEC_HASARGL;
181	}
182
183	/* Now get argv & environment */
184	if (!(cpp = (netbsd32_charp *)SCARG(uap, argp))) {
185		error = EINVAL;
186		goto bad;
187	}
188
189	if (pack.ep_flags & EXEC_SKIPARG)
190		cpp++;
191
192	while (1) {
193		len = argp + ARG_MAX - dp;
194		if ((error = copyin(cpp, &sp, sizeof(sp))) != 0)
195			goto bad;
196		if (!sp)
197			break;
198		if ((error = copyinstr((char *)(u_long)sp, dp,
199				       len, &len)) != 0) {
200			if (error == ENAMETOOLONG)
201				error = E2BIG;
202			goto bad;
203		}
204		dp += len;
205		cpp++;
206		argc++;
207	}
208
209	envc = 0;
210	/* environment need not be there */
211	if ((cpp = (netbsd32_charp *)SCARG(uap, envp)) != NULL ) {
212		while (1) {
213			len = argp + ARG_MAX - dp;
214			if ((error = copyin(cpp, &sp, sizeof(sp))) != 0)
215				goto bad;
216			if (!sp)
217				break;
218			if ((error = copyinstr((char *)(u_long)sp,
219					       dp, len, &len)) != 0) {
220				if (error == ENAMETOOLONG)
221					error = E2BIG;
222				goto bad;
223			}
224			dp += len;
225			cpp++;
226			envc++;
227		}
228	}
229
230	dp = (char *) ALIGN(dp);
231
232	szsigcode = pack.ep_es->es_emul->e_esigcode -
233	    pack.ep_es->es_emul->e_sigcode;
234
235	/* Now check if args & environ fit into new stack */
236	if (pack.ep_flags & EXEC_32)
237		len = ((argc + envc + 2 + pack.ep_es->es_arglen) *
238		    sizeof(int) + sizeof(int) + dp + STACKGAPLEN +
239		    szsigcode + sizeof(struct ps_strings)) - argp;
240	else
241		len = ((argc + envc + 2 + pack.ep_es->es_arglen) *
242		    sizeof(char *) + sizeof(int) + dp + STACKGAPLEN +
243		    szsigcode + sizeof(struct ps_strings)) - argp;
244
245	len = ALIGN(len);	/* make the stack "safely" aligned */
246
247	if (len > pack.ep_ssize) { /* in effect, compare to initial limit */
248		error = ENOMEM;
249		goto bad;
250	}
251
252	/* adjust "active stack depth" for process VSZ */
253	pack.ep_ssize = len;	/* maybe should go elsewhere, but... */
254
255	/*
256	 * Do whatever is necessary to prepare the address space
257	 * for remapping.  Note that this might replace the current
258	 * vmspace with another!
259	 */
260	uvmspace_exec(l, VM_MIN_ADDRESS, (vaddr_t)pack.ep_minsaddr);
261
262	/* Now map address space */
263	vm = p->p_vmspace;
264	vm->vm_taddr = (char *) pack.ep_taddr;
265	vm->vm_tsize = btoc(pack.ep_tsize);
266	vm->vm_daddr = (char *) pack.ep_daddr;
267	vm->vm_dsize = btoc(pack.ep_dsize);
268	vm->vm_ssize = btoc(pack.ep_ssize);
269	vm->vm_maxsaddr = (char *) pack.ep_maxsaddr;
270	vm->vm_minsaddr = (char *) pack.ep_minsaddr;
271
272	/* create the new process's VM space by running the vmcmds */
273#ifdef DIAGNOSTIC
274	if (pack.ep_vmcmds.evs_used == 0)
275		panic("netbsd32_execve: no vmcmds");
276#endif
277	for (i = 0; i < pack.ep_vmcmds.evs_used && !error; i++) {
278		struct exec_vmcmd *vcp;
279
280		vcp = &pack.ep_vmcmds.evs_cmds[i];
281		if (vcp->ev_flags & VMCMD_RELATIVE) {
282#ifdef DIAGNOSTIC
283			if (base_vcp == NULL)
284				panic("netbsd32_execve: relative vmcmd with no base");
285			if (vcp->ev_flags & VMCMD_BASE)
286				panic("netbsd32_execve: illegal base & relative vmcmd");
287#endif
288			vcp->ev_addr += base_vcp->ev_addr;
289		}
290		error = (*vcp->ev_proc)(p, vcp);
291#ifdef DEBUG
292		if (error) {
293			int j;
294
295			for (j = 0; j <= i; j++)
296				printf("vmcmd[%d] = %#lx/%#lx @ %#lx\n", j,
297				       vcp[j-i].ev_addr, vcp[j-i].ev_len,
298				       vcp[j-i].ev_offset);
299		}
300#endif
301		if (vcp->ev_flags & VMCMD_BASE)
302			base_vcp = vcp;
303	}
304
305	/* free the vmspace-creation commands, and release their references */
306	kill_vmcmds(&pack.ep_vmcmds);
307
308	/* if an error happened, deallocate and punt */
309	if (error) {
310#ifdef DEBUG
311		printf("netbsd32_execve: vmcmd %i failed: %d\n", i-1, error);
312#endif
313		goto exec_abort;
314	}
315
316	/* remember information about the process */
317	arginfo.ps_nargvstr = argc;
318	arginfo.ps_nenvstr = envc;
319
320	stack = (char *) (vm->vm_minsaddr - len);
321	/* Now copy argc, args & environ to new stack */
322	error = (*pack.ep_es->es_copyargs)(p, &pack, &arginfo,
323	    &stack, argp);
324	if (error) {
325#ifdef DEBUG
326		printf("netbsd32_execve: copyargs failed\n");
327#endif
328		goto exec_abort;
329	}
330	/* restore the stack back to its original point */
331	stack = (char *) (vm->vm_minsaddr - len);
332
333	/* fill process ps_strings info */
334	p->p_psstr = (struct ps_strings *)(vm->vm_minsaddr -
335	    sizeof(struct ps_strings));
336	p->p_psargv = offsetof(struct ps_strings, ps_argvstr);
337	p->p_psnargv = offsetof(struct ps_strings, ps_nargvstr);
338	p->p_psenv = offsetof(struct ps_strings, ps_envstr);
339	p->p_psnenv = offsetof(struct ps_strings, ps_nenvstr);
340
341	/* copy out the process's ps_strings structure */
342	if (copyout(&arginfo, (char *)p->p_psstr, sizeof(arginfo))) {
343#ifdef DEBUG
344		printf("netbsd32_execve: ps_strings copyout failed\n");
345#endif
346		goto exec_abort;
347	}
348
349	/* copy out the process's signal trapoline code */
350	if (szsigcode) {
351		if (copyout((char *)pack.ep_es->es_emul->e_sigcode,
352		    p->p_sigctx.ps_sigcode = (char *)p->p_psstr - szsigcode,
353		    szsigcode)) {
354#ifdef DEBUG
355			printf("netbsd32_execve: sig trampoline copyout failed\n");
356#endif
357			goto exec_abort;
358		}
359#ifdef PMAP_NEED_PROCWR
360		/* This is code. Let the pmap do what is needed. */
361		pmap_procwr(p, (vaddr_t)p->p_sigctx.ps_sigcode, szsigcode);
362#endif
363	}
364
365	stopprofclock(p);	/* stop profiling */
366	fdcloseexec(p);		/* handle close on exec */
367	execsigs(p);		/* reset catched signals */
368	l->l_ctxlink = NULL;	/* reset ucontext link */
369
370	/* set command name & other accounting info */
371	len = min(nid.ni_cnd.cn_namelen, MAXCOMLEN);
372	memcpy(p->p_comm, nid.ni_cnd.cn_nameptr, len);
373	p->p_comm[len] = 0;
374	p->p_acflag &= ~AFORK;
375
376	/* record proc's vnode, for use by procfs and others */
377        if (p->p_textvp)
378                vrele(p->p_textvp);
379	VREF(pack.ep_vp);
380	p->p_textvp = pack.ep_vp;
381
382	p->p_flag |= P_EXEC;
383	if (p->p_flag & P_PPWAIT) {
384		p->p_flag &= ~P_PPWAIT;
385		wakeup((caddr_t) p->p_pptr);
386	}
387
388	/*
389	 * deal with set[ug]id.
390	 * MNT_NOSUID has already been used to disable s[ug]id.
391	 */
392	if ((p->p_flag & P_TRACED) == 0 &&
393
394	    (((attr.va_mode & S_ISUID) != 0 &&
395	      p->p_ucred->cr_uid != attr.va_uid) ||
396
397	     ((attr.va_mode & S_ISGID) != 0 &&
398	      p->p_ucred->cr_gid != attr.va_gid))) {
399		/*
400		 * Mark the process as SUGID before we do
401		 * anything that might block.
402		 */
403		p_sugid(p);
404
405		p->p_ucred = crcopy(cred);
406#ifdef KTRACE
407		/*
408		 * If process is being ktraced, turn off - unless
409		 * root set it.
410		 */
411		if (p->p_tracep && !(p->p_traceflag & KTRFAC_ROOT))
412			ktrderef(p);
413#endif
414		if (attr.va_mode & S_ISUID)
415			p->p_ucred->cr_uid = attr.va_uid;
416		if (attr.va_mode & S_ISGID)
417			p->p_ucred->cr_gid = attr.va_gid;
418	} else
419		p->p_flag &= ~P_SUGID;
420	p->p_cred->p_svuid = p->p_ucred->cr_uid;
421	p->p_cred->p_svgid = p->p_ucred->cr_gid;
422
423	doexechooks(p);
424
425	uvm_km_free(exec_map, (vaddr_t) argp, NCARGS, UVM_KMF_PAGEABLE);
426
427	PNBUF_PUT(nid.ni_cnd.cn_pnbuf);
428	vn_lock(pack.ep_vp, LK_EXCLUSIVE | LK_RETRY);
429	VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
430	vput(pack.ep_vp);
431
432	/* setup new registers and do misc. setup. */
433	(*pack.ep_es->es_emul->e_setregs)(l, &pack, (u_long) stack);
434	if (pack.ep_es->es_setregs)
435		(*pack.ep_es->es_setregs)(l, &pack, (u_long) stack);
436
437	if (p->p_flag & P_TRACED)
438		psignal(p, SIGTRAP);
439
440	free(pack.ep_hdr, M_EXEC);
441
442	/*
443	 * Call emulation specific exec hook. This can setup setup per-process
444	 * p->p_emuldata or do any other per-process stuff an emulation needs.
445	 *
446	 * If we are executing process of different emulation than the
447	 * original forked process, call e_proc_exit() of the old emulation
448	 * first, then e_proc_exec() of new emulation. If the emulation is
449	 * same, the exec hook code should deallocate any old emulation
450	 * resources held previously by this process.
451	 */
452	if (p->p_emul && p->p_emul->e_proc_exit
453	    && p->p_emul != pack.ep_es->es_emul)
454		(*p->p_emul->e_proc_exit)(p);
455
456	/*
457	 * Call exec hook. Emulation code may NOT store reference to anything
458	 * from &pack.
459	 */
460        if (pack.ep_es->es_emul->e_proc_exec)
461                (*pack.ep_es->es_emul->e_proc_exec)(p, &pack);
462
463	/* update p_emul, the old value is no longer needed */
464	p->p_emul = pack.ep_es->es_emul;
465
466	/* ...and the same for p_execsw */
467	p->p_execsw = pack.ep_es;
468
469#ifdef __HAVE_SYSCALL_INTERN
470	(*p->p_emul->e_syscall_intern)(p);
471#endif
472#ifdef KTRACE
473	if (KTRPOINT(p, KTR_EMUL))
474		ktremul(p);
475#endif
476
477#if defined(LKM) || defined(_LKM)
478	lockmgr(&exec_lock, LK_RELEASE, NULL);
479#endif
480
481	return (EJUSTRETURN);
482
483bad:
484	/* free the vmspace-creation commands, and release their references */
485	kill_vmcmds(&pack.ep_vmcmds);
486	/* kill any opened file descriptor, if necessary */
487	if (pack.ep_flags & EXEC_HASFD) {
488		pack.ep_flags &= ~EXEC_HASFD;
489		(void) fdrelease(p, pack.ep_fd);
490	}
491	/* close and put the exec'd file */
492	vn_lock(pack.ep_vp, LK_EXCLUSIVE | LK_RETRY);
493	VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
494	vput(pack.ep_vp);
495	PNBUF_PUT(nid.ni_cnd.cn_pnbuf);
496	uvm_km_free(exec_map, (vaddr_t) argp, NCARGS, UVM_KMF_PAGEABLE);
497
498freehdr:
499#if defined(LKM) || defined(_LKM)
500	lockmgr(&exec_lock, LK_RELEASE, NULL);
501#endif
502
503	free(pack.ep_hdr, M_EXEC);
504	return error;
505
506exec_abort:
507#if defined(LKM) || defined(_LKM)
508	lockmgr(&exec_lock, LK_RELEASE, NULL);
509#endif
510
511	/*
512	 * the old process doesn't exist anymore.  exit gracefully.
513	 * get rid of the (new) address space we have created, if any, get rid
514	 * of our namei data and vnode, and exit noting failure
515	 */
516	uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
517		VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
518	if (pack.ep_emul_arg)
519		FREE(pack.ep_emul_arg, M_TEMP);
520	PNBUF_PUT(nid.ni_cnd.cn_pnbuf);
521	vn_lock(pack.ep_vp, LK_EXCLUSIVE | LK_RETRY);
522	VOP_CLOSE(pack.ep_vp, FREAD, cred, p);
523	vput(pack.ep_vp);
524	uvm_km_free(exec_map, (vaddr_t) argp, NCARGS, UVM_KMF_PAGEABLE);
525	free(pack.ep_hdr, M_EXEC);
526	exit1(l, W_EXITCODE(error, SIGABRT));
527
528	/* NOTREACHED */
529	return 0;
530}
531