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. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)procfs_mem.c 8.5 (Berkeley) 6/15/94 35 * 36 * $FreeBSD: src/sys/miscfs/procfs/procfs_mem.c,v 1.46.2.3 2002/01/22 17:22:59 nectar Exp $ 37 */ 38 39 /* 40 * This is a lightly hacked and merged version 41 * of sef's pread/pwrite functions 42 */ 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/proc.h> 47 #include <sys/priv.h> 48 #include <sys/vnode.h> 49 #include <vfs/procfs/procfs.h> 50 #include <vm/vm.h> 51 #include <vm/vm_param.h> 52 #include <sys/lock.h> 53 #include <vm/pmap.h> 54 #include <vm/vm_extern.h> 55 #include <vm/vm_map.h> 56 #include <vm/vm_kern.h> 57 #include <vm/vm_object.h> 58 #include <vm/vm_page.h> 59 #include <sys/user.h> 60 #include <sys/ptrace.h> 61 62 #include <machine/vmm.h> 63 64 #include <sys/thread2.h> 65 66 static int procfs_rwmem (struct proc *curp, 67 struct proc *p, struct uio *uio); 68 69 /* 70 * p->p_token is held on entry. 71 */ 72 static int 73 procfs_rwmem(struct proc *curp, struct proc *p, struct uio *uio) 74 { 75 int error; 76 int writing; 77 struct vmspace *vm; 78 vm_map_t map; 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 allocated or deallocated, 85 * or the process is exiting, don't try to grab anything. The 86 * page table usage in that process may be messed up. 87 */ 88 vm = p->p_vmspace; 89 if (p->p_stat == SIDL || p->p_stat == SZOMB) 90 return EFAULT; 91 if ((p->p_flags & (P_WEXIT | P_INEXEC)) || vmspace_getrefs(vm) < 0) 92 return EFAULT; 93 94 /* 95 * The map we want... 96 */ 97 vmspace_hold(vm); 98 map = &vm->vm_map; 99 100 writing = (uio->uio_rw == UIO_WRITE); 101 reqprot = VM_PROT_READ; 102 if (writing) 103 reqprot |= VM_PROT_WRITE | VM_PROT_OVERRIDE_WRITE; 104 105 kva = kmem_alloc_pageable(&kernel_map, PAGE_SIZE, VM_SUBSYS_PROC); 106 107 /* 108 * Only map in one page at a time. We don't have to, but it 109 * makes things easier. This way is trivial - right? 110 */ 111 do { 112 vm_offset_t uva; 113 vm_offset_t page_offset; /* offset into page */ 114 size_t len; 115 vm_page_t m; 116 int busy; 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 * If the target process is running in VMM mode 128 * translate the address into a GPA (Guest Physical 129 * Address) via the EPT before doing the lookup. 130 */ 131 if (p->p_vmm) { 132 register_t gpa; 133 vmm_vm_get_gpa(p, &gpa, (register_t) pageno); 134 pageno = (vm_offset_t)gpa; 135 } 136 137 /* 138 * How many bytes to copy 139 */ 140 len = szmin(PAGE_SIZE - page_offset, uio->uio_resid); 141 142 /* 143 * Fault the page on behalf of the process 144 * 145 * XXX busied page on write fault can deadlock against our 146 * uiomove. 147 */ 148 m = vm_fault_page(map, pageno, reqprot, 149 VM_FAULT_NORMAL, 150 &error, &busy); 151 if (error) { 152 KKASSERT(m == NULL); 153 error = EFAULT; 154 break; 155 } 156 157 /* 158 * Cleanup pmap then create a temporary KVA mapping and 159 * do the I/O. We can switch between cpus so don't bother 160 * synchronizing across all cores. 161 */ 162 pmap_kenter_quick(kva, VM_PAGE_TO_PHYS(m)); 163 error = uiomove((caddr_t)(kva + page_offset), len, uio); 164 pmap_kremove_quick(kva); 165 166 /* 167 * Release the page and we are done 168 */ 169 if (busy) 170 vm_page_wakeup(m); 171 else 172 vm_page_unhold(m); 173 } while (error == 0 && uio->uio_resid > 0); 174 175 vmspace_drop(vm); 176 kmem_free(&kernel_map, kva, PAGE_SIZE); 177 178 return (error); 179 } 180 181 /* 182 * Copy data in and out of the target process. 183 * We do this by mapping the process's page into 184 * the kernel and then doing a uiomove direct 185 * from the kernel address space. 186 * 187 * lp->lwp_proc->p_token is held on entry. 188 */ 189 int 190 procfs_domem(struct proc *curp, struct lwp *lp, struct pfsnode *pfs, 191 struct uio *uio) 192 { 193 struct proc *p = lp->lwp_proc; 194 int error; 195 196 if (uio->uio_resid == 0) 197 return (0); 198 199 if ((p->p_flags & P_INEXEC) != 0) { 200 /* 201 * Can't trace a process that's currently exec'ing. 202 */ 203 error = EAGAIN; 204 } else if (!CHECKIO(curp, p) || p_trespass(curp->p_ucred, p->p_ucred)) { 205 /* 206 * Can't trace processes outside our jail 207 */ 208 error = EPERM; 209 } else { 210 error = procfs_rwmem(curp, p, uio); 211 } 212 return(error); 213 } 214 215 /* 216 * Given process (p), find the vnode from which 217 * its text segment is being executed. 218 * 219 * It would be nice to grab this information from 220 * the VM system, however, there is no sure-fire 221 * way of doing that. Instead, fork(), exec() and 222 * wait() all maintain the p_textvp field in the 223 * process proc structure which contains a held 224 * reference to the exec'ed vnode. 225 * 226 * XXX - Currently, this is not not used, as the 227 * /proc/pid/file object exposes an information leak 228 * that shouldn't happen. Using a mount option would 229 * make it configurable on a per-system (or, at least, 230 * per-mount) basis; however, that's not really best. 231 * The best way to do it, I think, would be as an 232 * ioctl; this would restrict it to the uid running 233 * program, or root, which seems a reasonable compromise. 234 * However, the number of applications for this is 235 * minimal, if it can't be seen in the filesytem space, 236 * and doint it as an ioctl makes it somewhat less 237 * useful due to the, well, inelegance. 238 * 239 */ 240 struct vnode * 241 procfs_findtextvp(struct proc *p) 242 { 243 return (p->p_textvp); 244 } 245