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 * $DragonFly: src/sys/vfs/procfs/procfs_mem.c,v 1.16 2007/04/29 18:25:40 dillon Exp $ 38 */ 39 40 /* 41 * This is a lightly hacked and merged version 42 * of sef's pread/pwrite functions 43 */ 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/proc.h> 48 #include <sys/priv.h> 49 #include <sys/vnode.h> 50 #include <vfs/procfs/procfs.h> 51 #include <vm/vm.h> 52 #include <vm/vm_param.h> 53 #include <sys/lock.h> 54 #include <vm/pmap.h> 55 #include <vm/vm_extern.h> 56 #include <vm/vm_map.h> 57 #include <vm/vm_kern.h> 58 #include <vm/vm_object.h> 59 #include <vm/vm_page.h> 60 #include <sys/user.h> 61 #include <sys/ptrace.h> 62 63 #include <machine/vmm.h> 64 65 #include <sys/thread2.h> 66 #include <sys/sysref2.h> 67 68 static int procfs_rwmem (struct proc *curp, 69 struct proc *p, struct uio *uio); 70 71 /* 72 * p->p_token is held on entry. 73 */ 74 static int 75 procfs_rwmem(struct proc *curp, struct proc *p, struct uio *uio) 76 { 77 int error; 78 int writing; 79 struct vmspace *vm; 80 vm_map_t map; 81 vm_offset_t pageno = 0; /* page number */ 82 vm_prot_t reqprot; 83 vm_offset_t kva; 84 85 /* 86 * if the vmspace is in the midst of being allocated or deallocated, 87 * or the process is exiting, don't try to grab anything. The 88 * page table usage in that process may be messed up. 89 */ 90 vm = p->p_vmspace; 91 if (p->p_stat == SIDL || p->p_stat == SZOMB) 92 return EFAULT; 93 if ((p->p_flags & (P_WEXIT | P_INEXEC)) || 94 sysref_isinactive(&vm->vm_sysref)) 95 return EFAULT; 96 97 /* 98 * The map we want... 99 */ 100 vmspace_hold(vm); 101 map = &vm->vm_map; 102 103 writing = (uio->uio_rw == UIO_WRITE); 104 reqprot = VM_PROT_READ; 105 if (writing) 106 reqprot |= VM_PROT_WRITE | VM_PROT_OVERRIDE_WRITE; 107 108 kva = kmem_alloc_pageable(&kernel_map, PAGE_SIZE); 109 110 /* 111 * Only map in one page at a time. We don't have to, but it 112 * makes things easier. This way is trivial - right? 113 */ 114 do { 115 vm_offset_t uva; 116 vm_offset_t page_offset; /* offset into page */ 117 size_t len; 118 vm_page_t m; 119 120 uva = (vm_offset_t) uio->uio_offset; 121 122 /* 123 * Get the page number of this segment. 124 */ 125 pageno = trunc_page(uva); 126 page_offset = uva - pageno; 127 128 /* 129 * If the target process is running in VMM mode 130 * translate the address into a GPA (Guest Physical 131 * Address) via the EPT before doing the lookup. 132 */ 133 if (p->p_vmm) { 134 register_t gpa; 135 vmm_vm_get_gpa(p, &gpa, (register_t) pageno); 136 pageno = (vm_offset_t)gpa; 137 } 138 139 /* 140 * How many bytes to copy 141 */ 142 len = szmin(PAGE_SIZE - page_offset, uio->uio_resid); 143 144 /* 145 * Fault the page on behalf of the process 146 */ 147 m = vm_fault_page(map, pageno, reqprot, 148 VM_FAULT_NORMAL, &error); 149 if (error) { 150 KKASSERT(m == NULL); 151 error = EFAULT; 152 break; 153 } 154 155 /* 156 * Cleanup tmap then create a temporary KVA mapping and 157 * do the I/O. We can switch between cpus so don't bother 158 * synchronizing across all cores. 159 */ 160 pmap_kenter_quick(kva, VM_PAGE_TO_PHYS(m)); 161 error = uiomove((caddr_t)(kva + page_offset), len, uio); 162 pmap_kremove_quick(kva); 163 164 /* 165 * release the page and we are done 166 */ 167 vm_page_unhold(m); 168 } while (error == 0 && uio->uio_resid > 0); 169 170 vmspace_drop(vm); 171 kmem_free(&kernel_map, kva, PAGE_SIZE); 172 173 return (error); 174 } 175 176 /* 177 * Copy data in and out of the target process. 178 * We do this by mapping the process's page into 179 * the kernel and then doing a uiomove direct 180 * from the kernel address space. 181 * 182 * lp->lwp_proc->p_token is held on entry. 183 */ 184 int 185 procfs_domem(struct proc *curp, struct lwp *lp, struct pfsnode *pfs, 186 struct uio *uio) 187 { 188 struct proc *p = lp->lwp_proc; 189 int error; 190 191 if (uio->uio_resid == 0) 192 return (0); 193 194 if ((p->p_flags & P_INEXEC) != 0) { 195 /* 196 * Can't trace a process that's currently exec'ing. 197 */ 198 error = EAGAIN; 199 } else if (!CHECKIO(curp, p) || p_trespass(curp->p_ucred, p->p_ucred)) { 200 /* 201 * Can't trace processes outside our jail 202 */ 203 error = EPERM; 204 } else { 205 error = procfs_rwmem(curp, p, uio); 206 } 207 return(error); 208 } 209 210 /* 211 * Given process (p), find the vnode from which 212 * its text segment is being executed. 213 * 214 * It would be nice to grab this information from 215 * the VM system, however, there is no sure-fire 216 * way of doing that. Instead, fork(), exec() and 217 * wait() all maintain the p_textvp field in the 218 * process proc structure which contains a held 219 * reference to the exec'ed vnode. 220 * 221 * XXX - Currently, this is not not used, as the 222 * /proc/pid/file object exposes an information leak 223 * that shouldn't happen. Using a mount option would 224 * make it configurable on a per-system (or, at least, 225 * per-mount) basis; however, that's not really best. 226 * The best way to do it, I think, would be as an 227 * ioctl; this would restrict it to the uid running 228 * program, or root, which seems a reasonable compromise. 229 * However, the number of applications for this is 230 * minimal, if it can't be seen in the filesytem space, 231 * and doint it as an ioctl makes it somewhat less 232 * useful due to the, well, inelegance. 233 * 234 */ 235 struct vnode * 236 procfs_findtextvp(struct proc *p) 237 { 238 return (p->p_textvp); 239 } 240