xref: /dragonfly/sys/kern/imgact_resident.c (revision 335b9e93)
1 /*
2  * (MPSAFE)
3  *
4  * Copyright (c) 2003,2004 The DragonFly Project.  All rights reserved.
5  *
6  * This code is derived from software contributed to The DragonFly Project
7  * by Matthew Dillon <dillon@backplane.com>
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in
17  *    the documentation and/or other materials provided with the
18  *    distribution.
19  * 3. Neither the name of The DragonFly Project nor the names of its
20  *    contributors may be used to endorse or promote products derived
21  *    from this software without specific, prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
26  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
27  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
28  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
29  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
30  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
31  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
32  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
33  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
40 #include <sys/sysproto.h>
41 #include <sys/exec.h>
42 #include <sys/imgact.h>
43 #include <sys/imgact_aout.h>
44 #include <sys/mman.h>
45 #include <sys/proc.h>
46 #include <sys/priv.h>
47 #include <sys/resourcevar.h>
48 #include <sys/sysent.h>
49 #include <sys/stat.h>
50 #include <sys/vnode.h>
51 #include <sys/sysctl.h>
52 #include <sys/lock.h>
53 #include <sys/resident.h>
54 #include <sys/malloc.h>
55 
56 #include <vm/vm.h>
57 #include <vm/vm_param.h>
58 #include <vm/pmap.h>
59 #include <vm/vm_map.h>
60 #include <vm/vm_kern.h>
61 #include <vm/vm_extern.h>
62 
63 static int exec_res_id = 0;
64 
65 static TAILQ_HEAD(,vmresident) exec_res_list;
66 
67 static MALLOC_DEFINE(M_EXEC_RES, "vmresident", "resident execs");
68 
69 /* lockmgr lock for protecting the exec_res_list */
70 static struct lock exec_list_lock;
71 
72 static void
73 vm_resident_init(void *__dummy)
74 {
75 	lockinit(&exec_list_lock, "vmres", 0, 0);
76 	TAILQ_INIT(&exec_res_list);
77 }
78 SYSINIT(vmres, SI_BOOT1_LOCK, SI_ORDER_ANY, vm_resident_init, 0);
79 
80 static int
81 fill_xresident(struct vmresident *vr, struct xresident *in, struct thread *td)
82 {
83 	struct stat st;
84 	struct vnode *vrtmp;
85 	int error = 0;
86 
87 	vrtmp = vr->vr_vnode;
88 
89 	in->res_entry_addr = vr->vr_entry_addr;
90 	in->res_id = vr->vr_id;
91 	if (vrtmp) {
92 		char *freepath, *fullpath;
93 		error = vn_fullpath(td->td_proc, vrtmp, &fullpath, &freepath, 0);
94 		if (error != 0) {
95 			/* could not retrieve cached path, return zero'ed string */
96 			bzero(in->res_file, MAXPATHLEN);
97 			error = 0;
98 		} else {
99 			strlcpy(in->res_file, fullpath, sizeof(in->res_file));
100 			kfree(freepath, M_TEMP);
101 		}
102 
103 		/* indicate that we are using the vnode */
104 		error = vget(vrtmp, LK_EXCLUSIVE);
105 		if (error)
106 			goto done;
107 
108 		/* retrieve underlying stat information and release vnode */
109 		error = vn_stat(vrtmp, &st, td->td_ucred);
110 		vput(vrtmp);
111 		if (error)
112 			goto done;
113 
114 		in->res_stat = st;
115 	}
116 
117 done:
118 	if (error)
119 		kprintf("fill_xresident, error = %d\n", error);
120 	return (error);
121 }
122 
123 static int
124 sysctl_vm_resident(SYSCTL_HANDLER_ARGS)
125 {
126 	struct vmresident *vmres;
127 	struct thread *td;
128 	int error;
129 	int count;
130 
131 	/* only super-user should call this sysctl */
132 	td = req->td;
133 	if ((priv_check(td, PRIV_VM_RESIDENT)) != 0)
134 		return EPERM;
135 
136 	error = count = 0;
137 
138 	if (exec_res_id == 0)
139 	    return error;
140 
141 	/* client queried for number of resident binaries */
142 	if (!req->oldptr)
143 	    return SYSCTL_OUT(req, 0, exec_res_id);
144 
145 	lockmgr(&exec_list_lock, LK_SHARED);
146 
147 	TAILQ_FOREACH(vmres, &exec_res_list, vr_link) {
148 		struct xresident xres;
149 		error = fill_xresident(vmres, &xres, td);
150 		if (error != 0)
151 			break;
152 
153 		error = SYSCTL_OUT(req, (void *)&xres,
154 				sizeof(struct xresident));
155 		if (error != 0)
156 			break;
157 	}
158 	lockmgr(&exec_list_lock, LK_RELEASE);
159 
160 	return (error);
161 }
162 SYSCTL_PROC(_vm, OID_AUTO, resident, CTLTYPE_OPAQUE|CTLFLAG_RD, 0, 0,
163   sysctl_vm_resident, "S,xresident", "resident executables (sys/resident.h)");
164 
165 int
166 exec_resident_imgact(struct image_params *imgp)
167 {
168 	struct vmresident *vmres;
169 
170 	/*
171 	 * resident image activator
172 	 */
173 	lockmgr(&exec_list_lock, LK_SHARED);
174 	if ((vmres = imgp->vp->v_resident) == NULL) {
175 	    lockmgr(&exec_list_lock, LK_RELEASE);
176 	    return(-1);
177 	}
178 	atomic_add_int(&vmres->vr_refs, 1);
179 	lockmgr(&exec_list_lock, LK_RELEASE);
180 
181 	/*
182 	 * We want to exec the new vmspace without holding the lock to
183 	 * improve concurrency.
184 	 */
185 	exec_new_vmspace(imgp, vmres->vr_vmspace);
186 	imgp->resident = 1;
187 	imgp->interpreted = 0;
188 	imgp->proc->p_sysent = vmres->vr_sysent;
189 	imgp->entry_addr = vmres->vr_entry_addr;
190 	atomic_subtract_int(&vmres->vr_refs, 1);
191 
192 	return(0);
193 }
194 
195 /*
196  * exec_sys_register(entry)
197  *
198  * Register ourselves for resident execution.  Only root (i.e. a process with
199  * PRIV_VM_RESIDENT credentials) can do this.  This
200  * will snapshot the vmspace and cause future exec's of the specified binary
201  * to use the snapshot directly rather then load & relocate a new copy.
202  *
203  * MPALMOSTSAFE
204  */
205 int
206 sys_exec_sys_register(struct exec_sys_register_args *uap)
207 {
208     struct thread *td = curthread;
209     struct vmresident *vmres;
210     struct vnode *vp;
211     struct proc *p;
212     int error;
213 
214     p = td->td_proc;
215     error = priv_check_cred(td->td_ucred, PRIV_VM_RESIDENT, 0);
216     if (error)
217 	return(error);
218 
219     if ((vp = p->p_textvp) == NULL)
220 	return(ENOENT);
221 
222     lockmgr(&exec_list_lock, LK_EXCLUSIVE);
223 
224     if (vp->v_resident) {
225 	lockmgr(&exec_list_lock, LK_RELEASE);
226 	return(EEXIST);
227     }
228 
229     vhold(vp);
230     vmres = kmalloc(sizeof(*vmres), M_EXEC_RES, M_WAITOK | M_ZERO);
231     vmres->vr_vnode = vp;
232     vmres->vr_sysent = p->p_sysent;
233     vmres->vr_id = ++exec_res_id;
234     vmres->vr_entry_addr = (intptr_t)uap->entry;
235     vmres->vr_vmspace = vmspace_fork(p->p_vmspace, NULL, NULL); /* XXX order */
236     pmap_pinit2(vmspace_pmap(vmres->vr_vmspace));
237     vp->v_resident = vmres;
238 
239     TAILQ_INSERT_TAIL(&exec_res_list, vmres, vr_link);
240     lockmgr(&exec_list_lock, LK_RELEASE);
241 
242     return(0);
243 }
244 
245 /*
246  * exec_sys_unregister(id)
247  *
248  *	Unregister the specified id.  If an id of -1 is used unregister
249  *	the registration associated with the current process.  An id of -2
250  *	unregisters everything.
251  *
252  * MPALMOSTSAFE
253  */
254 int
255 sys_exec_sys_unregister(struct exec_sys_unregister_args *uap)
256 {
257     struct thread *td = curthread;
258     struct vmresident *vmres;
259     struct proc *p;
260     int error;
261     int id;
262     int count;
263 
264     p = td->td_proc;
265     error = priv_check_cred(td->td_ucred, PRIV_VM_RESIDENT, 0);
266     if (error)
267 	return(error);
268 
269     /*
270      * If id is -1, unregister ourselves
271      */
272     lockmgr(&exec_list_lock, LK_EXCLUSIVE);
273 
274     if ((id = uap->id) == -1 && p->p_textvp && p->p_textvp->v_resident)
275 	id = p->p_textvp->v_resident->vr_id;
276 
277     /*
278      * Look for the registration
279      */
280     error = ENOENT;
281     count = 0;
282 
283 restart:
284     TAILQ_FOREACH(vmres, &exec_res_list, vr_link) {
285 	if (id == -2 || vmres->vr_id == id) {
286 	    /*
287 	     * Check race against exec
288 	     */
289 	    if (vmres->vr_refs) {
290 		lockmgr(&exec_list_lock, LK_RELEASE);
291 		tsleep(vmres, 0, "vmres", 1);
292 		lockmgr(&exec_list_lock, LK_EXCLUSIVE);
293 		goto restart;
294 	    }
295 
296 	    /*
297 	     * Remove it
298 	     */
299 	    TAILQ_REMOVE(&exec_res_list, vmres, vr_link);
300 	    if (vmres->vr_vnode) {
301 		vmres->vr_vnode->v_resident = NULL;
302 		vdrop(vmres->vr_vnode);
303 		vmres->vr_vnode = NULL;
304 	    }
305 	    if (vmres->vr_vmspace) {
306 		vmspace_rel(vmres->vr_vmspace);
307 		vmres->vr_vmspace = NULL;
308 	    }
309 	    kfree(vmres, M_EXEC_RES);
310 	    exec_res_id--;
311 	    error = 0;
312 	    ++count;
313 	    goto restart;
314 	}
315     }
316     lockmgr(&exec_list_lock, LK_RELEASE);
317 
318     if (error == 0)
319 	uap->sysmsg_result = count;
320     return(error);
321 }
322 
323