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