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
  37 #include <sys/param.h>
  38 #include <sys/systm.h>
  39 #include <sys/kernel.h>
  40 #include <sys/sysproto.h>
  41 #include <sys/exec.h>
  42 #include <sys/imgact.h>
  43 #include <sys/imgact_aout.h>
  44 #include <sys/mman.h>
  45 #include <sys/proc.h>
  46 #include <sys/priv.h>
  47 #include <sys/resourcevar.h>
  48 #include <sys/sysent.h>
  49 #include <sys/stat.h>
  50 #include <sys/vnode.h>
  51 #include <sys/inflate.h>
  52 #include <sys/sysctl.h>
  53 #include <sys/lock.h>
  54 #include <sys/resident.h>
  55
  56 #include <vm/vm.h>
  57 #include <vm/vm_param.h>
  58 #include <vm/pmap.h>
  59 #include <vm/vm_map.h>
  60 #include <vm/vm_kern.h>
  61 #include <vm/vm_extern.h>
  62
  63 #include <sys/sysref2.h>
  64
  65 static int exec_res_id = 0;
  66
  67 static TAILQ_HEAD(,vmresident) exec_res_list;
  68
  69 static MALLOC_DEFINE(M_EXEC_RES, "vmresident", "resident execs");
  70
  71 /* lockmgr lock for protecting the exec_res_list */
  72 static struct lock exec_list_lock;
  73
  74 static void
  75 vm_resident_init(void *__dummy)
  76 {
  77         lockinit(&exec_list_lock, "vmres", 0, 0);
  78         TAILQ_INIT(&exec_res_list);
  79 }
  80 SYSINIT(vmres, SI_BOOT1_LOCK, SI_ORDER_ANY, vm_resident_init, 0);
  81
  82 static int
  83 fill_xresident(struct vmresident *vr, struct xresident *in, struct thread *td)
  84 {
  85         struct stat st;
  86         struct vnode *vrtmp;
  87         int error = 0;
  88
  89         vrtmp = vr->vr_vnode;
  90
  91         in->res_entry_addr = vr->vr_entry_addr;
  92         in->res_id = vr->vr_id;
  93         if (vrtmp) {
  94                 char *freepath, *fullpath;
  95                 error = vn_fullpath(td->td_proc, vrtmp, &fullpath, &freepath, 0);
  96                 if (error != 0) {
  97                         /* could not retrieve cached path, return zero'ed string */
  98                         bzero(in->res_file, MAXPATHLEN);
  99                         error = 0;
 100                 } else {
 101                         strlcpy(in->res_file, fullpath, sizeof(in->res_file));
 102                         kfree(freepath, M_TEMP);
 103                 }
 104
 105                 /* indicate that we are using the vnode */
 106                 error = vget(vrtmp, LK_EXCLUSIVE);
 107                 if (error)
 108                         goto done;
 109
 110                 /* retrieve underlying stat information and release vnode */
 111                 error = vn_stat(vrtmp, &st, td->td_ucred);
 112                 vput(vrtmp);
 113                 if (error)
 114                         goto done;
 115
 116                 in->res_stat = st;
 117         }
 118
 119 done:
 120         if (error)
 121                 kprintf("fill_xresident, error = %d\n", error);
 122         return (error);
 123 }
 124
 125 static int
 126 sysctl_vm_resident(SYSCTL_HANDLER_ARGS)
 127 {
 128         struct vmresident *vmres;
 129         struct thread *td;
 130         int error;
 131         int count;
 132
 133         /* only super-user should call this sysctl */
 134         td = req->td;
 135         if ((priv_check(td, PRIV_VM_RESIDENT)) != 0)
 136                 return EPERM;
 137
 138         error = count = 0;
 139
 140         if (exec_res_id == 0)
 141             return error;
 142
 143         /* client queried for number of resident binaries */
 144         if (!req->oldptr)
 145             return SYSCTL_OUT(req, 0, exec_res_id);
 146
 147         lockmgr(&exec_list_lock, LK_SHARED);
 148
 149         TAILQ_FOREACH(vmres, &exec_res_list, vr_link) {
 150                 struct xresident xres;
 151                 error = fill_xresident(vmres, &xres, td);
 152                 if (error != 0)
 153                         break;
 154
 155                 error = SYSCTL_OUT(req, (void *)&xres,
 156                                 sizeof(struct xresident));
 157                 if (error != 0)
 158                         break;
 159         }
 160         lockmgr(&exec_list_lock, LK_RELEASE);
 161
 162         return (error);
 163 }
 164 SYSCTL_PROC(_vm, OID_AUTO, resident, CTLTYPE_OPAQUE|CTLFLAG_RD, 0, 0,
 165   sysctl_vm_resident, "S,xresident", "resident executables (sys/resident.h)");
 166
 167 int
 168 exec_resident_imgact(struct image_params *imgp)
 169 {
 170         struct vmresident *vmres;
 171
 172         /*
 173          * resident image activator
 174          */
 175         lockmgr(&exec_list_lock, LK_SHARED);
 176         if ((vmres = imgp->vp->v_resident) == NULL) {
 177             lockmgr(&exec_list_lock, LK_RELEASE);
 178             return(-1);
 179         }
 180         atomic_add_int(&vmres->vr_refs, 1);
 181         lockmgr(&exec_list_lock, LK_RELEASE);
 182
 183         /*
 184          * We want to exec the new vmspace without holding the lock to
 185          * improve concurrency.
 186          */
 187         exec_new_vmspace(imgp, vmres->vr_vmspace);
 188         imgp->resident = 1;
 189         imgp->interpreted = 0;
 190         imgp->proc->p_sysent = vmres->vr_sysent;
 191         imgp->entry_addr = vmres->vr_entry_addr;
 192         atomic_subtract_int(&vmres->vr_refs, 1);
 193
 194         return(0);
 195 }
 196
 197 /*
 198  * exec_sys_register(entry)
 199  *
 200  * Register ourselves for resident execution.  Only root (i.e. a process with
 201  * PRIV_VM_RESIDENT credentials) can do this.  This
 202  * will snapshot the vmspace and cause future exec's of the specified binary
 203  * to use the snapshot directly rather then load & relocate a new copy.
 204  *
 205  * MPALMOSTSAFE
 206  */
 207 int
 208 sys_exec_sys_register(struct exec_sys_register_args *uap)
 209 {
 210     struct thread *td = curthread;
 211     struct vmresident *vmres;
 212     struct vnode *vp;
 213     struct proc *p;
 214     int error;
 215
 216     p = td->td_proc;
 217     error = priv_check_cred(td->td_ucred, PRIV_VM_RESIDENT, 0);
 218     if (error)
 219         return(error);
 220
 221     if ((vp = p->p_textvp) == NULL)
 222         return(ENOENT);
 223
 224     lockmgr(&exec_list_lock, LK_EXCLUSIVE);
 225
 226     if (vp->v_resident) {
 227         lockmgr(&exec_list_lock, LK_RELEASE);
 228         return(EEXIST);
 229     }
 230
 231     vhold(vp);
 232     vmres = kmalloc(sizeof(*vmres), M_EXEC_RES, M_WAITOK | M_ZERO);
 233     vmres->vr_vnode = vp;
 234     vmres->vr_sysent = p->p_sysent;
 235     vmres->vr_id = ++exec_res_id;
 236     vmres->vr_entry_addr = (intptr_t)uap->entry;
 237     vmres->vr_vmspace = vmspace_fork(p->p_vmspace); /* XXX order */
 238     pmap_pinit2(vmspace_pmap(vmres->vr_vmspace));
 239     vp->v_resident = vmres;
 240
 241     TAILQ_INSERT_TAIL(&exec_res_list, vmres, vr_link);
 242     lockmgr(&exec_list_lock, LK_RELEASE);
 243
 244     return(0);
 245 }
 246
 247 /*
 248  * exec_sys_unregister(id)
 249  *
 250  *      Unregister the specified id.  If an id of -1 is used unregister
 251  *      the registration associated with the current process.  An id of -2
 252  *      unregisters everything.
 253  *
 254  * MPALMOSTSAFE
 255  */
 256 int
 257 sys_exec_sys_unregister(struct exec_sys_unregister_args *uap)
 258 {
 259     struct thread *td = curthread;
 260     struct vmresident *vmres;
 261     struct proc *p;
 262     int error;
 263     int id;
 264     int count;
 265
 266     p = td->td_proc;
 267     error = priv_check_cred(td->td_ucred, PRIV_VM_RESIDENT, 0);
 268     if (error)
 269         return(error);
 270
 271     /*
 272      * If id is -1, unregister ourselves
 273      */
 274     lockmgr(&exec_list_lock, LK_EXCLUSIVE);
 275
 276     if ((id = uap->id) == -1 && p->p_textvp && p->p_textvp->v_resident)
 277         id = p->p_textvp->v_resident->vr_id;
 278
 279     /*
 280      * Look for the registration
 281      */
 282     error = ENOENT;
 283     count = 0;
 284
 285 restart:
 286     TAILQ_FOREACH(vmres, &exec_res_list, vr_link) {
 287         if (id == -2 || vmres->vr_id == id) {
 288             /*
 289              * Check race against exec
 290              */
 291             if (vmres->vr_refs) {
 292                 lockmgr(&exec_list_lock, LK_RELEASE);
 293                 tsleep(vmres, 0, "vmres", 1);
 294                 lockmgr(&exec_list_lock, LK_EXCLUSIVE);
 295                 goto restart;
 296             }
 297
 298             /*
 299              * Remove it
 300              */
 301             TAILQ_REMOVE(&exec_res_list, vmres, vr_link);
 302             if (vmres->vr_vnode) {
 303                 vmres->vr_vnode->v_resident = NULL;
 304                 vdrop(vmres->vr_vnode);
 305                 vmres->vr_vnode = NULL;
 306             }
 307             if (vmres->vr_vmspace) {
 308                 vmspace_free(vmres->vr_vmspace);
 309                 vmres->vr_vmspace = NULL;
 310             }
 311             kfree(vmres, M_EXEC_RES);
 312             exec_res_id--;
 313             error = 0;
 314             ++count;
 315             goto restart;
 316         }
 317     }
 318     lockmgr(&exec_list_lock, LK_RELEASE);
 319
 320     if (error == 0)
 321         uap->sysmsg_result = count;
 322     return(error);
 323 }
 324