sys/vfs/procfs/procfs_mem.c

   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)) || vmspace_getrefs(vm) < 0)
  94                 return EFAULT;
  95
  96         /*
  97          * The map we want...
  98          */
  99         vmspace_hold(vm);
 100         map = &vm->vm_map;
 101
 102         writing = (uio->uio_rw == UIO_WRITE);
 103         reqprot = VM_PROT_READ;
 104         if (writing)
 105                 reqprot |= VM_PROT_WRITE | VM_PROT_OVERRIDE_WRITE;
 106
 107         kva = kmem_alloc_pageable(&kernel_map, PAGE_SIZE, VM_SUBSYS_PROC);
 108
 109         /*
 110          * Only map in one page at a time.  We don't have to, but it
 111          * makes things easier.  This way is trivial - right?
 112          */
 113         do {
 114                 vm_offset_t uva;
 115                 vm_offset_t page_offset;        /* offset into page */
 116                 size_t len;
 117                 vm_page_t m;
 118                 int busy;
 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                  * XXX busied page on write fault can deadlock against our
 148                  *     uiomove.
 149                  */
 150                 m = vm_fault_page(map, pageno, reqprot,
 151                                   VM_FAULT_NORMAL,
 152                                   &error, &busy);
 153                 if (error) {
 154                         KKASSERT(m == NULL);
 155                         error = EFAULT;
 156                         break;
 157                 }
 158
 159                 /*
 160                  * Cleanup tmap then create a temporary KVA mapping and
 161                  * do the I/O.  We can switch between cpus so don't bother
 162                  * synchronizing across all cores.
 163                  */
 164                 pmap_kenter_quick(kva, VM_PAGE_TO_PHYS(m));
 165                 error = uiomove((caddr_t)(kva + page_offset), len, uio);
 166                 pmap_kremove_quick(kva);
 167
 168                 /*
 169                  * Release the page and we are done
 170                  */
 171                 if (busy)
 172                         vm_page_wakeup(m);
 173                 else
 174                         vm_page_unhold(m);
 175         } while (error == 0 && uio->uio_resid > 0);
 176
 177         vmspace_drop(vm);
 178         kmem_free(&kernel_map, kva, PAGE_SIZE);
 179
 180         return (error);
 181 }
 182
 183 /*
 184  * Copy data in and out of the target process.
 185  * We do this by mapping the process's page into
 186  * the kernel and then doing a uiomove direct
 187  * from the kernel address space.
 188  *
 189  * lp->lwp_proc->p_token is held on entry.
 190  */
 191 int
 192 procfs_domem(struct proc *curp, struct lwp *lp, struct pfsnode *pfs,
 193              struct uio *uio)
 194 {
 195         struct proc *p = lp->lwp_proc;
 196         int error;
 197
 198         if (uio->uio_resid == 0)
 199                 return (0);
 200
 201         if ((p->p_flags & P_INEXEC) != 0) {
 202                 /*
 203                  * Can't trace a process that's currently exec'ing.
 204                  */
 205                 error = EAGAIN;
 206         } else if (!CHECKIO(curp, p) || p_trespass(curp->p_ucred, p->p_ucred)) {
 207                 /*
 208                  * Can't trace processes outside our jail
 209                  */
 210                 error = EPERM;
 211         } else {
 212                 error = procfs_rwmem(curp, p, uio);
 213         }
 214         return(error);
 215 }
 216
 217 /*
 218  * Given process (p), find the vnode from which
 219  * its text segment is being executed.
 220  *
 221  * It would be nice to grab this information from
 222  * the VM system, however, there is no sure-fire
 223  * way of doing that.  Instead, fork(), exec() and
 224  * wait() all maintain the p_textvp field in the
 225  * process proc structure which contains a held
 226  * reference to the exec'ed vnode.
 227  *
 228  * XXX - Currently, this is not not used, as the
 229  * /proc/pid/file object exposes an information leak
 230  * that shouldn't happen.  Using a mount option would
 231  * make it configurable on a per-system (or, at least,
 232  * per-mount) basis; however, that's not really best.
 233  * The best way to do it, I think, would be as an
 234  * ioctl; this would restrict it to the uid running
 235  * program, or root, which seems a reasonable compromise.
 236  * However, the number of applications for this is
 237  * minimal, if it can't be seen in the filesytem space,
 238  * and doint it as an ioctl makes it somewhat less
 239  * useful due to the, well, inelegance.
 240  *
 241  */
 242 struct vnode *
 243 procfs_findtextvp(struct proc *p)
 244 {
 245         return (p->p_textvp);
 246 }