procfs_mem.c revision 16308
1/*
2 * Copyright (c) 1993 Jan-Simon Pendry
3 * Copyright (c) 1993 Sean Eric Fagan
4 * Copyright (c) 1993
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Jan-Simon Pendry and Sean Eric Fagan.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
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 the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 *    must display the following acknowledgement:
20 *	This product includes software developed by the University of
21 *	California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 *    may be used to endorse or promote products derived from this software
24 *    without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 *	@(#)procfs_mem.c	8.4 (Berkeley) 1/21/94
39 *
40 *	$Id: procfs_mem.c,v 1.17 1996/01/25 06:05:38 peter Exp $
41 */
42
43/*
44 * This is a lightly hacked and merged version
45 * of sef's pread/pwrite functions
46 */
47
48#include <sys/param.h>
49#include <sys/systm.h>
50#include <sys/time.h>
51#include <sys/kernel.h>
52#include <sys/proc.h>
53#include <sys/vnode.h>
54#include <miscfs/procfs/procfs.h>
55#include <vm/vm.h>
56#include <vm/vm_param.h>
57#include <vm/vm_prot.h>
58#include <vm/lock.h>
59#include <vm/pmap.h>
60#include <vm/vm_map.h>
61#include <vm/vm_kern.h>
62#include <vm/vm_object.h>
63#include <vm/vm_page.h>
64#include <vm/vm_extern.h>
65#include <sys/user.h>
66
67static int	procfs_rwmem __P((struct proc *p, struct uio *uio));
68
69static int
70procfs_rwmem(p, uio)
71	struct proc *p;
72	struct uio *uio;
73{
74	int error;
75	int writing;
76	struct vmspace *vm;
77
78	/*
79	 * if the vmspace is in the midst of being deallocated or the
80	 * process is exiting, don't try to grab anything.  The page table
81	 * usage in that process can be messed up.
82	 */
83	vm = p->p_vmspace;
84	if ((p->p_flag & P_WEXIT) || (vm->vm_refcnt < 1))
85		return EFAULT;
86	++vm->vm_refcnt;
87
88	writing = uio->uio_rw == UIO_WRITE;
89
90	/*
91	 * Only map in one page at a time.  We don't have to, but it
92	 * makes things easier.  This way is trivial - right?
93	 */
94	do {
95		vm_map_t map, tmap;
96		vm_object_t object;
97		vm_offset_t kva = 0;
98		vm_offset_t uva;
99		int page_offset;		/* offset into page */
100		vm_offset_t pageno;		/* page number */
101		vm_map_entry_t out_entry;
102		vm_prot_t out_prot;
103		vm_page_t m;
104		boolean_t wired, single_use;
105		vm_pindex_t pindex;
106		u_int len;
107		int fix_prot;
108
109		uva = (vm_offset_t) uio->uio_offset;
110
111		/*
112		 * Get the page number of this segment.
113		 */
114		pageno = trunc_page(uva);
115		page_offset = uva - pageno;
116
117		/*
118		 * How many bytes to copy
119		 */
120		len = min(PAGE_SIZE - page_offset, uio->uio_resid);
121
122		if (uva >= VM_MAXUSER_ADDRESS) {
123			if (writing || (uva >= (VM_MAXUSER_ADDRESS + UPAGES * PAGE_SIZE))) {
124				error = 0;
125				break;
126			}
127
128			/* we are reading the "U area", force it into core */
129			PHOLD(p);
130
131			/* sanity check */
132			if (!(p->p_flag & P_INMEM)) {
133				/* aiee! */
134				error = EFAULT;
135				break;
136			}
137
138			/* populate the ptrace/procfs area */
139			p->p_addr->u_kproc.kp_proc = *p;
140			fill_eproc (p, &p->p_addr->u_kproc.kp_eproc);
141
142			/* locate the in-core address */
143			kva = (u_int)p->p_addr + uva - VM_MAXUSER_ADDRESS;
144
145			/* transfer it */
146			error = uiomove((caddr_t)kva, len, uio);
147
148			/* let the pages go */
149			PRELE(p);
150
151			continue;
152		}
153
154
155		/*
156		 * The map we want...
157		 */
158		map = &vm->vm_map;
159
160		/*
161		 * Check the permissions for the area we're interested
162		 * in.
163		 */
164		fix_prot = 0;
165		if (writing)
166			fix_prot = !vm_map_check_protection(map, pageno,
167					pageno + PAGE_SIZE, VM_PROT_WRITE);
168
169		if (fix_prot) {
170			/*
171			 * If the page is not writable, we make it so.
172			 * XXX It is possible that a page may *not* be
173			 * read/executable, if a process changes that!
174			 * We will assume, for now, that a page is either
175			 * VM_PROT_ALL, or VM_PROT_READ|VM_PROT_EXECUTE.
176			 */
177			error = vm_map_protect(map, pageno,
178					pageno + PAGE_SIZE, VM_PROT_ALL, 0);
179			if (error)
180				break;
181		}
182
183		/*
184		 * Now we need to get the page.  out_entry, out_prot, wired,
185		 * and single_use aren't used.  One would think the vm code
186		 * would be a *bit* nicer...  We use tmap because
187		 * vm_map_lookup() can change the map argument.
188		 */
189		tmap = map;
190		error = vm_map_lookup(&tmap, pageno,
191				      writing ? VM_PROT_WRITE : VM_PROT_READ,
192				      &out_entry, &object, &pindex, &out_prot,
193				      &wired, &single_use);
194		/*
195		 * We're done with tmap now.
196		 */
197		if (!error)
198			vm_map_lookup_done(tmap, out_entry);
199
200		/*
201		 * Fault the page in...
202		 */
203		if (!error && writing && object->backing_object) {
204			m = vm_page_lookup(object, pindex);
205			if (m == 0)
206				error = vm_fault(map, pageno,
207							VM_PROT_WRITE, FALSE);
208		}
209
210		/* Find space in kernel_map for the page we're interested in */
211		if (!error)
212			error = vm_map_find(kernel_map, object,
213				IDX_TO_OFF(pindex), &kva, PAGE_SIZE, 1,
214				VM_PROT_ALL, VM_PROT_ALL, 0);
215
216		if (!error) {
217			/*
218			 * Neither vm_map_lookup() nor vm_map_find() appear
219			 * to add a reference count to the object, so we do
220			 * that here and now.
221			 */
222			vm_object_reference(object);
223
224			/*
225			 * Mark the page we just found as pageable.
226			 */
227			error = vm_map_pageable(kernel_map, kva,
228				kva + PAGE_SIZE, 0);
229
230			/*
231			 * Now do the i/o move.
232			 */
233			if (!error)
234				error = uiomove((caddr_t)(kva + page_offset),
235						len, uio);
236
237			vm_map_remove(kernel_map, kva, kva + PAGE_SIZE);
238		}
239		if (fix_prot)
240			vm_map_protect(map, pageno, pageno + PAGE_SIZE,
241					VM_PROT_READ|VM_PROT_EXECUTE, 0);
242	} while (error == 0 && uio->uio_resid > 0);
243
244	vmspace_free(vm);
245	return (error);
246}
247
248/*
249 * Copy data in and out of the target process.
250 * We do this by mapping the process's page into
251 * the kernel and then doing a uiomove direct
252 * from the kernel address space.
253 */
254int
255procfs_domem(curp, p, pfs, uio)
256	struct proc *curp;
257	struct proc *p;
258	struct pfsnode *pfs;
259	struct uio *uio;
260{
261	int error;
262
263	if (uio->uio_resid == 0)
264		return (0);
265
266	error = procfs_rwmem(p, uio);
267
268	return (error);
269}
270
271/*
272 * Given process (p), find the vnode from which
273 * it's text segment is being executed.
274 *
275 * It would be nice to grab this information from
276 * the VM system, however, there is no sure-fire
277 * way of doing that.  Instead, fork(), exec() and
278 * wait() all maintain the p_textvp field in the
279 * process proc structure which contains a held
280 * reference to the exec'ed vnode.
281 */
282struct vnode *
283procfs_findtextvp(p)
284	struct proc *p;
285{
286	return (p->p_textvp);
287}
288