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 * $FreeBSD: src/sys/miscfs/procfs/procfs_mem.c,v 1.46.2.3 2002/01/22 17:22:59 nectar Exp $ 41 * $DragonFly: src/sys/vfs/procfs/procfs_mem.c,v 1.11 2004/10/12 19:29:31 dillon Exp $ 42 */ 43 44 /* 45 * This is a lightly hacked and merged version 46 * of sef's pread/pwrite functions 47 */ 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/proc.h> 52 #include <sys/vnode.h> 53 #include <vfs/procfs/procfs.h> 54 #include <vm/vm.h> 55 #include <vm/vm_param.h> 56 #include <sys/lock.h> 57 #include <vm/pmap.h> 58 #include <vm/vm_extern.h> 59 #include <vm/vm_map.h> 60 #include <vm/vm_kern.h> 61 #include <vm/vm_object.h> 62 #include <vm/vm_page.h> 63 #include <sys/user.h> 64 #include <sys/ptrace.h> 65 66 #include <sys/thread2.h> 67 68 static int procfs_rwmem (struct proc *curp, 69 struct proc *p, struct uio *uio); 70 71 static int 72 procfs_rwmem(struct proc *curp, struct proc *p, struct uio *uio) 73 { 74 int error; 75 int writing; 76 struct vmspace *vm; 77 vm_map_t map; 78 vm_offset_t pageno = 0; /* page number */ 79 vm_prot_t reqprot; 80 vm_offset_t kva; 81 82 /* 83 * if the vmspace is in the midst of being deallocated or the 84 * process is exiting, don't try to grab anything. The page table 85 * usage in that process can be messed up. 86 */ 87 vm = p->p_vmspace; 88 if ((p->p_flag & P_WEXIT) || (vm->vm_refcnt < 1)) 89 return EFAULT; 90 ++vm->vm_refcnt; 91 /* 92 * The map we want... 93 */ 94 map = &vm->vm_map; 95 96 writing = uio->uio_rw == UIO_WRITE; 97 reqprot = writing ? (VM_PROT_WRITE | VM_PROT_OVERRIDE_WRITE) : VM_PROT_READ; 98 99 kva = kmem_alloc_pageable(kernel_map, PAGE_SIZE); 100 101 /* 102 * Only map in one page at a time. We don't have to, but it 103 * makes things easier. This way is trivial - right? 104 */ 105 do { 106 vm_map_t tmap; 107 vm_offset_t uva; 108 int page_offset; /* offset into page */ 109 vm_map_entry_t out_entry; 110 vm_prot_t out_prot; 111 boolean_t wired; 112 vm_pindex_t pindex; 113 vm_object_t object; 114 vm_object_t nobject; 115 u_int len; 116 vm_page_t m; 117 118 uva = (vm_offset_t) uio->uio_offset; 119 120 /* 121 * Get the page number of this segment. 122 */ 123 pageno = trunc_page(uva); 124 page_offset = uva - pageno; 125 126 /* 127 * How many bytes to copy 128 */ 129 len = min(PAGE_SIZE - page_offset, uio->uio_resid); 130 131 /* 132 * Fault the page on behalf of the process 133 */ 134 error = vm_fault(map, pageno, reqprot, VM_FAULT_NORMAL); 135 if (error) { 136 error = EFAULT; 137 break; 138 } 139 140 /* 141 * Now we need to get the page. out_entry, out_prot, wired, 142 * and single_use aren't used. One would think the vm code 143 * would be a *bit* nicer... We use tmap because 144 * vm_map_lookup() can change the map argument. 145 */ 146 tmap = map; 147 error = vm_map_lookup(&tmap, pageno, reqprot, 148 &out_entry, &object, &pindex, &out_prot, 149 &wired); 150 151 if (error) { 152 error = EFAULT; 153 break; 154 } 155 156 /* 157 * spl protection is required to avoid interrupt freeing 158 * races, reference the object to avoid it being ripped 159 * out from under us if we block. 160 */ 161 crit_enter(); 162 vm_object_reference(object); 163 again: 164 m = vm_page_lookup(object, pindex); 165 166 /* 167 * Allow fallback to backing objects if we are reading 168 */ 169 while (m == NULL && !writing && object->backing_object) { 170 pindex += OFF_TO_IDX(object->backing_object_offset); 171 nobject = object->backing_object; 172 vm_object_reference(nobject); 173 vm_object_deallocate(object); 174 object = nobject; 175 m = vm_page_lookup(object, pindex); 176 } 177 178 /* 179 * Wait for any I/O's to complete, then hold the page 180 * so we can release the spl. 181 */ 182 if (m) { 183 if (vm_page_sleep_busy(m, FALSE, "rwmem")) 184 goto again; 185 vm_page_hold(m); 186 } 187 crit_exit(); 188 189 /* 190 * We no longer need the object. If we do not have a page 191 * then cleanup. 192 */ 193 vm_object_deallocate(object); 194 if (m == NULL) { 195 vm_map_lookup_done(tmap, out_entry, 0); 196 error = EFAULT; 197 break; 198 } 199 200 /* 201 * Cleanup tmap then create a temporary KVA mapping and 202 * do the I/O. 203 */ 204 vm_map_lookup_done(tmap, out_entry, 0); 205 pmap_kenter(kva, VM_PAGE_TO_PHYS(m)); 206 error = uiomove((caddr_t)(kva + page_offset), len, uio); 207 pmap_kremove(kva); 208 209 /* 210 * release the page and we are done 211 */ 212 crit_enter(); 213 vm_page_unhold(m); 214 crit_exit(); 215 } while (error == 0 && uio->uio_resid > 0); 216 217 kmem_free(kernel_map, kva, PAGE_SIZE); 218 vmspace_free(vm); 219 return (error); 220 } 221 222 /* 223 * Copy data in and out of the target process. 224 * We do this by mapping the process's page into 225 * the kernel and then doing a uiomove direct 226 * from the kernel address space. 227 */ 228 int 229 procfs_domem(struct proc *curp, struct proc *p, struct pfsnode *pfs, 230 struct uio *uio) 231 { 232 if (uio->uio_resid == 0) 233 return (0); 234 235 /* Can't trace a process that's currently exec'ing. */ 236 if ((p->p_flag & P_INEXEC) != 0) 237 return EAGAIN; 238 if (!CHECKIO(curp, p) || p_trespass(curp->p_ucred, p->p_ucred)) 239 return EPERM; 240 241 return (procfs_rwmem(curp, p, uio)); 242 } 243 244 /* 245 * Given process (p), find the vnode from which 246 * its text segment is being executed. 247 * 248 * It would be nice to grab this information from 249 * the VM system, however, there is no sure-fire 250 * way of doing that. Instead, fork(), exec() and 251 * wait() all maintain the p_textvp field in the 252 * process proc structure which contains a held 253 * reference to the exec'ed vnode. 254 * 255 * XXX - Currently, this is not not used, as the 256 * /proc/pid/file object exposes an information leak 257 * that shouldn't happen. Using a mount option would 258 * make it configurable on a per-system (or, at least, 259 * per-mount) basis; however, that's not really best. 260 * The best way to do it, I think, would be as an 261 * ioctl; this would restrict it to the uid running 262 * program, or root, which seems a reasonable compromise. 263 * However, the number of applications for this is 264 * minimal, if it can't be seen in the filesytem space, 265 * and doint it as an ioctl makes it somewhat less 266 * useful due to the, well, inelegance. 267 * 268 */ 269 struct vnode * 270 procfs_findtextvp(struct proc *p) 271 { 272 return (p->p_textvp); 273 } 274