procfs_mem.c revision 33108
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 8.5 (Berkeley) 6/15/94 39 * 40 * $Id: procfs_mem.c,v 1.28 1998/01/22 17:30:01 dyson Exp $ 41 */ 42 43/* 44 * This is a lightly hacked and merged version 45 * of sef's pread/pwrite functions 46 */ 47 48#include "opt_diagnostic.h" 49 50#include <sys/param.h> 51#include <sys/systm.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 <sys/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 vm_map_t map; 78 vm_object_t object = NULL; 79 vm_offset_t pageno = 0; /* page number */ 80 vm_prot_t reqprot; 81 vm_offset_t kva; 82 83 /* 84 * if the vmspace is in the midst of being deallocated or the 85 * process is exiting, don't try to grab anything. The page table 86 * usage in that process can be messed up. 87 */ 88 vm = p->p_vmspace; 89 if ((p->p_flag & P_WEXIT) || (vm->vm_refcnt < 1)) 90 return EFAULT; 91 ++vm->vm_refcnt; 92 /* 93 * The map we want... 94 */ 95 map = &vm->vm_map; 96 97 writing = uio->uio_rw == UIO_WRITE; 98 reqprot = writing ? (VM_PROT_WRITE | VM_PROT_OVERRIDE_WRITE) : VM_PROT_READ; 99 100 kva = kmem_alloc_pageable(kernel_map, PAGE_SIZE); 101 102 /* 103 * Only map in one page at a time. We don't have to, but it 104 * makes things easier. This way is trivial - right? 105 */ 106 do { 107 vm_map_t tmap; 108 vm_offset_t uva; 109 int page_offset; /* offset into page */ 110 vm_map_entry_t out_entry; 111 vm_prot_t out_prot; 112 boolean_t wired; 113 vm_pindex_t pindex; 114 u_int len; 115 vm_page_t m; 116 117 object = NULL; 118 119 uva = (vm_offset_t) uio->uio_offset; 120 121 /* 122 * Get the page number of this segment. 123 */ 124 pageno = trunc_page(uva); 125 page_offset = uva - pageno; 126 127 /* 128 * How many bytes to copy 129 */ 130 len = min(PAGE_SIZE - page_offset, uio->uio_resid); 131 132 if (uva >= VM_MAXUSER_ADDRESS) { 133 vm_offset_t tkva; 134 135 if (writing || (uva >= (VM_MAXUSER_ADDRESS + UPAGES * PAGE_SIZE))) { 136 error = 0; 137 break; 138 } 139 140 /* we are reading the "U area", force it into core */ 141 PHOLD(p); 142 143 /* sanity check */ 144 if (!(p->p_flag & P_INMEM)) { 145 /* aiee! */ 146 PRELE(p); 147 error = EFAULT; 148 break; 149 } 150 151 /* populate the ptrace/procfs area */ 152 p->p_addr->u_kproc.kp_proc = *p; 153 fill_eproc (p, &p->p_addr->u_kproc.kp_eproc); 154 155 /* locate the in-core address */ 156 tkva = (u_int)p->p_addr + uva - VM_MAXUSER_ADDRESS; 157 158 /* transfer it */ 159 error = uiomove((caddr_t)tkva, len, uio); 160 161 /* let the pages go */ 162 PRELE(p); 163 164 continue; 165 } 166 167 /* 168 * Fault the page on behalf of the process 169 */ 170 error = vm_fault(map, pageno, reqprot, FALSE); 171 if (error) { 172 error = EFAULT; 173 break; 174 } 175 176 /* 177 * Now we need to get the page. out_entry, out_prot, wired, 178 * and single_use aren't used. One would think the vm code 179 * would be a *bit* nicer... We use tmap because 180 * vm_map_lookup() can change the map argument. 181 */ 182 tmap = map; 183 error = vm_map_lookup(&tmap, pageno, reqprot, 184 &out_entry, &object, &pindex, &out_prot, 185 &wired); 186 187 if (error) { 188 error = EFAULT; 189 190 /* 191 * Make sure that there is no residue in 'object' from 192 * an error return on vm_map_lookup. 193 */ 194 object = NULL; 195 196 break; 197 } 198 199 m = vm_page_lookup(object, pindex); 200 201 /* Allow fallback to backing objects if we are reading */ 202 203 while (m == NULL && !writing && object->backing_object) { 204 205 pindex += OFF_TO_IDX(object->backing_object_offset); 206 object = object->backing_object; 207 208 m = vm_page_lookup(object, pindex); 209 } 210 211 if (m == NULL) { 212 error = EFAULT; 213 214 /* 215 * Make sure that there is no residue in 'object' from 216 * an error return on vm_map_lookup. 217 */ 218 object = NULL; 219 220 vm_map_lookup_done(tmap, out_entry); 221 222 break; 223 } 224 225 /* 226 * Wire the page into memory 227 */ 228 vm_page_wire(m); 229 230 /* 231 * We're done with tmap now. 232 * But reference the object first, so that we won't loose 233 * it. 234 */ 235 vm_object_reference(object); 236 vm_map_lookup_done(tmap, out_entry); 237 238 pmap_kenter(kva, VM_PAGE_TO_PHYS(m)); 239 240 /* 241 * Now do the i/o move. 242 */ 243 error = uiomove((caddr_t)(kva + page_offset), len, uio); 244 245 pmap_kremove(kva); 246 247 /* 248 * release the page and the object 249 */ 250 vm_page_unwire(m); 251 vm_object_deallocate(object); 252 253 object = NULL; 254 255 } while (error == 0 && uio->uio_resid > 0); 256 257 if (object) 258 vm_object_deallocate(object); 259 260 kmem_free(kernel_map, kva, PAGE_SIZE); 261 vmspace_free(vm); 262 return (error); 263} 264 265/* 266 * Copy data in and out of the target process. 267 * We do this by mapping the process's page into 268 * the kernel and then doing a uiomove direct 269 * from the kernel address space. 270 */ 271int 272procfs_domem(curp, p, pfs, uio) 273 struct proc *curp; 274 struct proc *p; 275 struct pfsnode *pfs; 276 struct uio *uio; 277{ 278 279 if (uio->uio_resid == 0) 280 return (0); 281 282 /* 283 * XXX 284 * We need to check for KMEM_GROUP because ps is sgid kmem; 285 * not allowing it here causes ps to not work properly. Arguably, 286 * this is a bug with what ps does. We only need to do this 287 * for Pmem nodes, and only if it's reading. This is still not 288 * good, as it may still be possible to grab illicit data if 289 * a process somehow gets to be KMEM_GROUP. Note that this also 290 * means that KMEM_GROUP can't change without editing procfs.h! 291 * All in all, quite yucky. 292 */ 293 294 if (!CHECKIO(curp, p) && 295 !(curp->p_cred->pc_ucred->cr_gid == KMEM_GROUP && 296 uio->uio_rw == UIO_READ)) 297 return EPERM; 298 299 return (procfs_rwmem(p, uio)); 300} 301 302/* 303 * Given process (p), find the vnode from which 304 * it's text segment is being executed. 305 * 306 * It would be nice to grab this information from 307 * the VM system, however, there is no sure-fire 308 * way of doing that. Instead, fork(), exec() and 309 * wait() all maintain the p_textvp field in the 310 * process proc structure which contains a held 311 * reference to the exec'ed vnode. 312 */ 313struct vnode * 314procfs_findtextvp(p) 315 struct proc *p; 316{ 317 318 return (p->p_textvp); 319} 320