sys/kern/imgact_resident.c

   1 /*
   2  * (MPSAFE)
   3  *
   4  * Copyright (c) 2003,2004 The DragonFly Project.  All rights reserved.
   5  *
   6  * This code is derived from software contributed to The DragonFly Project
   7  * by Matthew Dillon <dillon@backplane.com>
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  *
  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
  17  *    the documentation and/or other materials provided with the
  18  *    distribution.
  19  * 3. Neither the name of The DragonFly Project nor the names of its
  20  *    contributors may be used to endorse or promote products derived
  21  *    from this software without specific, prior written permission.
  22  *
  23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  24  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  25  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  26  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  27  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  28  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  29  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  30  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  31  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  32  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  33  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  34  * SUCH DAMAGE.
  35  *
  36  * $DragonFly: src/sys/kern/imgact_resident.c,v 1.17 2007/04/30 07:18:53 dillon Exp $
  37  */
  38
  39 #include <sys/param.h>
  40 #include <sys/systm.h>
  41 #include <sys/kernel.h>
  42 #include <sys/sysproto.h>
  43 #include <sys/exec.h>
  44 #include <sys/imgact.h>
  45 #include <sys/imgact_aout.h>
  46 #include <sys/mman.h>
  47 #include <sys/proc.h>
  48 #include <sys/priv.h>
  49 #include <sys/resourcevar.h>
  50 #include <sys/sysent.h>
  51 #include <sys/systm.h>
  52 #include <sys/stat.h>
  53 #include <sys/vnode.h>
  54 #include <sys/inflate.h>
  55 #include <sys/sysctl.h>
  56 #include <sys/lock.h>
  57 #include <sys/resident.h>
  58
  59 #include <vm/vm.h>
  60 #include <vm/vm_param.h>
  61 #include <vm/pmap.h>
  62 #include <vm/vm_map.h>
  63 #include <vm/vm_kern.h>
  64 #include <vm/vm_extern.h>
  65
  66 #include <sys/sysref2.h>
  67 #include <sys/mplock2.h>
  68
  69 static int exec_res_id = 0;
  70
  71 static TAILQ_HEAD(,vmresident) exec_res_list;
  72
  73 static MALLOC_DEFINE(M_EXEC_RES, "vmresident", "resident execs");
  74
  75 /* lockmgr lock for protecting the exec_res_list */
  76 static struct lock exec_list_lock;
  77
  78 static void
  79 vm_resident_init(void *__dummy)
  80 {
  81         lockinit(&exec_list_lock, "vmres", 0, 0);
  82         TAILQ_INIT(&exec_res_list);
  83 }
  84 SYSINIT(vmres, SI_BOOT1_LOCK, SI_ORDER_ANY, vm_resident_init, 0);
  85
  86 static int
  87 fill_xresident(struct vmresident *vr, struct xresident *in, struct thread *td)
  88 {
  89         struct stat st;
  90         struct vnode *vrtmp;
  91         int error = 0;
  92
  93         vrtmp = vr->vr_vnode;
  94
  95         in->res_entry_addr = vr->vr_entry_addr;
  96         in->res_id = vr->vr_id;
  97         if (vrtmp) {
  98                 char *freepath, *fullpath;
  99                 error = vn_fullpath(td->td_proc, vrtmp, &fullpath, &freepath, 0);
 100                 if (error != 0) {
 101                         /* could not retrieve cached path, return zero'ed string */
 102                         bzero(in->res_file, MAXPATHLEN);
 103                         error = 0;
 104                 } else {
 105                         strlcpy(in->res_file, fullpath, sizeof(in->res_file));
 106                         kfree(freepath, M_TEMP);
 107                 }
 108
 109                 /* indicate that we are using the vnode */
 110                 error = vget(vrtmp, LK_EXCLUSIVE);
 111                 if (error)
 112                         goto done;
 113
 114                 /* retrieve underlying stat information and release vnode */
 115                 error = vn_stat(vrtmp, &st, td->td_ucred);
 116                 vput(vrtmp);
 117                 if (error)
 118                         goto done;
 119
 120                 in->res_stat = st;
 121         }
 122
 123 done:
 124         if (error)
 125                 kprintf("fill_xresident, error = %d\n", error);
 126         return (error);
 127 }
 128
 129 static int
 130 sysctl_vm_resident(SYSCTL_HANDLER_ARGS)
 131 {
 132         struct vmresident *vmres;
 133         struct thread *td;
 134         int error;
 135         int count;
 136
 137         /* only super-user should call this sysctl */
 138         td = req->td;
 139         if ((priv_check(td, PRIV_VM_RESIDENT)) != 0)
 140                 return EPERM;
 141
 142         error = count = 0;
 143
 144         if (exec_res_id == 0)
 145             return error;
 146
 147         /* client queried for number of resident binaries */
 148         if (!req->oldptr)
 149             return SYSCTL_OUT(req, 0, exec_res_id);
 150
 151         lockmgr(&exec_list_lock, LK_SHARED);
 152
 153         TAILQ_FOREACH(vmres, &exec_res_list, vr_link) {
 154                 struct xresident xres;
 155                 error = fill_xresident(vmres, &xres, td);
 156                 if (error != 0)
 157                         break;
 158
 159                 error = SYSCTL_OUT(req, (void *)&xres,
 160                                 sizeof(struct xresident));
 161                 if (error != 0)
 162                         break;
 163         }
 164         lockmgr(&exec_list_lock, LK_RELEASE);
 165
 166         return (error);
 167 }
 168 SYSCTL_PROC(_vm, OID_AUTO, resident, CTLTYPE_OPAQUE|CTLFLAG_RD, 0, 0,
 169   sysctl_vm_resident, "S,xresident", "resident executables (sys/resident.h)");
 170
 171 int
 172 exec_resident_imgact(struct image_params *imgp)
 173 {
 174         struct vmresident *vmres;
 175
 176         /*
 177          * resident image activator
 178          */
 179         lockmgr(&exec_list_lock, LK_SHARED);
 180         if ((vmres = imgp->vp->v_resident) == NULL) {
 181             lockmgr(&exec_list_lock, LK_RELEASE);
 182             return(-1);
 183         }
 184         atomic_add_int(&vmres->vr_refs, 1);
 185         lockmgr(&exec_list_lock, LK_RELEASE);
 186
 187         /*
 188          * We want to exec the new vmspace without holding the lock to
 189          * improve concurrency.
 190          */
 191         exec_new_vmspace(imgp, vmres->vr_vmspace);
 192         imgp->resident = 1;
 193         imgp->interpreted = 0;
 194         imgp->proc->p_sysent = vmres->vr_sysent;
 195         imgp->entry_addr = vmres->vr_entry_addr;
 196         atomic_subtract_int(&vmres->vr_refs, 1);
 197
 198         return(0);
 199 }
 200
 201 /*
 202  * exec_sys_register(entry)
 203  *
 204  * Register ourselves for resident execution.  Only root (i.e. a process with
 205  * PRIV_VM_RESIDENT credentials) can do this.  This
 206  * will snapshot the vmspace and cause future exec's of the specified binary
 207  * to use the snapshot directly rather then load & relocate a new copy.
 208  *
 209  * MPALMOSTSAFE
 210  */
 211 int
 212 sys_exec_sys_register(struct exec_sys_register_args *uap)
 213 {
 214     struct thread *td = curthread;
 215     struct vmresident *vmres;
 216     struct vnode *vp;
 217     struct proc *p;
 218     int error;
 219
 220     p = td->td_proc;
 221     error = priv_check_cred(td->td_ucred, PRIV_VM_RESIDENT, 0);
 222     if (error)
 223         return(error);
 224
 225     if ((vp = p->p_textvp) == NULL)
 226         return(ENOENT);
 227
 228     lockmgr(&exec_list_lock, LK_EXCLUSIVE);
 229
 230     if (vp->v_resident) {
 231         lockmgr(&exec_list_lock, LK_RELEASE);
 232         return(EEXIST);
 233     }
 234
 235     vhold(vp);
 236     vmres = kmalloc(sizeof(*vmres), M_EXEC_RES, M_WAITOK | M_ZERO);
 237     vmres->vr_vnode = vp;
 238     vmres->vr_sysent = p->p_sysent;
 239     vmres->vr_id = ++exec_res_id;
 240     vmres->vr_entry_addr = (intptr_t)uap->entry;
 241     vmres->vr_vmspace = vmspace_fork(p->p_vmspace); /* XXX order */
 242     pmap_pinit2(vmspace_pmap(vmres->vr_vmspace));
 243     vp->v_resident = vmres;
 244
 245     TAILQ_INSERT_TAIL(&exec_res_list, vmres, vr_link);
 246     lockmgr(&exec_list_lock, LK_RELEASE);
 247
 248     return(0);
 249 }
 250
 251 /*
 252  * exec_sys_unregister(id)
 253  *
 254  *      Unregister the specified id.  If an id of -1 is used unregister
 255  *      the registration associated with the current process.  An id of -2
 256  *      unregisters everything.
 257  *
 258  * MPALMOSTSAFE
 259  */
 260 int
 261 sys_exec_sys_unregister(struct exec_sys_unregister_args *uap)
 262 {
 263     struct thread *td = curthread;
 264     struct vmresident *vmres;
 265     struct proc *p;
 266     int error;
 267     int id;
 268     int count;
 269
 270     p = td->td_proc;
 271     error = priv_check_cred(td->td_ucred, PRIV_VM_RESIDENT, 0);
 272     if (error)
 273         return(error);
 274
 275     /*
 276      * If id is -1, unregister ourselves
 277      */
 278     lockmgr(&exec_list_lock, LK_EXCLUSIVE);
 279
 280     if ((id = uap->id) == -1 && p->p_textvp && p->p_textvp->v_resident)
 281         id = p->p_textvp->v_resident->vr_id;
 282
 283     /*
 284      * Look for the registration
 285      */
 286     error = ENOENT;
 287     count = 0;
 288
 289 restart:
 290     TAILQ_FOREACH(vmres, &exec_res_list, vr_link) {
 291         if (id == -2 || vmres->vr_id == id) {
 292             /*
 293              * Check race against exec
 294              */
 295             if (vmres->vr_refs) {
 296                 lockmgr(&exec_list_lock, LK_RELEASE);
 297                 tsleep(vmres, 0, "vmres", 1);
 298                 lockmgr(&exec_list_lock, LK_EXCLUSIVE);
 299                 goto restart;
 300             }
 301
 302             /*
 303              * Remove it
 304              */
 305             TAILQ_REMOVE(&exec_res_list, vmres, vr_link);
 306             if (vmres->vr_vnode) {
 307                 vmres->vr_vnode->v_resident = NULL;
 308                 vdrop(vmres->vr_vnode);
 309                 vmres->vr_vnode = NULL;
 310             }
 311             if (vmres->vr_vmspace) {
 312                 sysref_put(&vmres->vr_vmspace->vm_sysref);
 313                 vmres->vr_vmspace = NULL;
 314             }
 315             kfree(vmres, M_EXEC_RES);
 316             exec_res_id--;
 317             error = 0;
 318             ++count;
 319             goto restart;
 320         }
 321     }
 322     lockmgr(&exec_list_lock, LK_RELEASE);
 323
 324     if (error == 0)
 325         uap->sysmsg_result = count;
 326     return(error);
 327 }
 328