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/sysmsg.h>
  41 #include <sys/exec.h>
  42 #include <sys/imgact.h>
  43 #include <sys/mman.h>
  44 #include <sys/proc.h>
  45 #include <sys/caps.h>
  46 #include <sys/resourcevar.h>
  47 #include <sys/sysent.h>
  48 #include <sys/stat.h>
  49 #include <sys/vnode.h>
  50 #include <sys/sysctl.h>
  51 #include <sys/lock.h>
  52 #include <sys/resident.h>
  53 #include <sys/malloc.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         error = caps_priv_check_td(td, SYSCAP_NOVM_RESIDENT);
 133         if (error)
 134                 return error;
 135
 136         count = 0;
 137
 138         if (exec_res_id == 0)
 139             return error;
 140
 141         /* client queried for number of resident binaries */
 142         if (!req->oldptr)
 143             return SYSCTL_OUT(req, 0, exec_res_id);
 144
 145         lockmgr(&exec_list_lock, LK_SHARED);
 146
 147         TAILQ_FOREACH(vmres, &exec_res_list, vr_link) {
 148                 struct xresident xres;
 149                 error = fill_xresident(vmres, &xres, td);
 150                 if (error != 0)
 151                         break;
 152
 153                 error = SYSCTL_OUT(req, (void *)&xres,
 154                                 sizeof(struct xresident));
 155                 if (error != 0)
 156                         break;
 157         }
 158         lockmgr(&exec_list_lock, LK_RELEASE);
 159
 160         return (error);
 161 }
 162 SYSCTL_PROC(_vm, OID_AUTO, resident, CTLTYPE_OPAQUE|CTLFLAG_RD, 0, 0,
 163   sysctl_vm_resident, "S,xresident", "resident executables (sys/resident.h)");
 164
 165 int
 166 exec_resident_imgact(struct image_params *imgp)
 167 {
 168         struct vmresident *vmres;
 169
 170         /*
 171          * resident image activator
 172          */
 173         lockmgr(&exec_list_lock, LK_SHARED);
 174         if ((vmres = imgp->vp->v_resident) == NULL) {
 175             lockmgr(&exec_list_lock, LK_RELEASE);
 176             return(-1);
 177         }
 178         atomic_add_int(&vmres->vr_refs, 1);
 179         lockmgr(&exec_list_lock, LK_RELEASE);
 180
 181         /*
 182          * We want to exec the new vmspace without holding the lock to
 183          * improve concurrency.
 184          */
 185         exec_new_vmspace(imgp, vmres->vr_vmspace);
 186         imgp->resident = 1;
 187         imgp->interpreted = 0;
 188         imgp->proc->p_sysent = vmres->vr_sysent;
 189         imgp->entry_addr = vmres->vr_entry_addr;
 190         atomic_subtract_int(&vmres->vr_refs, 1);
 191
 192         return(0);
 193 }
 194
 195 /*
 196  * exec_sys_register(entry)
 197  *
 198  * Register ourselves for resident execution.  Only root (i.e. a process with
 199  * PRIV_VM_RESIDENT credentials) can do this.  This
 200  * will snapshot the vmspace and cause future exec's of the specified binary
 201  * to use the snapshot directly rather then load & relocate a new copy.
 202  *
 203  * MPALMOSTSAFE
 204  */
 205 int
 206 sys_exec_sys_register(struct sysmsg *sysmsg,
 207                       const struct exec_sys_register_args *uap)
 208 {
 209     struct thread *td = curthread;
 210     struct vmresident *vmres;
 211     struct vnode *vp;
 212     struct proc *p;
 213     int error;
 214
 215     p = td->td_proc;
 216     error = caps_priv_check_td(td, SYSCAP_NOVM_RESIDENT);
 217     if (error)
 218         return(error);
 219
 220     if ((vp = p->p_textvp) == NULL)
 221         return(ENOENT);
 222
 223     lockmgr(&exec_list_lock, LK_EXCLUSIVE);
 224
 225     if (vp->v_resident) {
 226         lockmgr(&exec_list_lock, LK_RELEASE);
 227         return(EEXIST);
 228     }
 229
 230     vhold(vp);
 231     vmres = kmalloc(sizeof(*vmres), M_EXEC_RES, M_WAITOK | M_ZERO);
 232     vmres->vr_vnode = vp;
 233     vmres->vr_sysent = p->p_sysent;
 234     vmres->vr_id = ++exec_res_id;
 235     vmres->vr_entry_addr = (intptr_t)uap->entry;
 236     vmres->vr_vmspace = vmspace_fork(p->p_vmspace, NULL, NULL); /* XXX order */
 237     pmap_pinit2(vmspace_pmap(vmres->vr_vmspace));
 238     vp->v_resident = vmres;
 239
 240     TAILQ_INSERT_TAIL(&exec_res_list, vmres, vr_link);
 241     lockmgr(&exec_list_lock, LK_RELEASE);
 242
 243     return(0);
 244 }
 245
 246 /*
 247  * exec_sys_unregister(id)
 248  *
 249  *      Unregister the specified id.  If an id of -1 is used unregister
 250  *      the registration associated with the current process.  An id of -2
 251  *      unregisters everything.
 252  *
 253  * MPALMOSTSAFE
 254  */
 255 int
 256 sys_exec_sys_unregister(struct sysmsg *sysmsg,
 257                         const 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 = caps_priv_check_td(td, SYSCAP_NOVM_RESIDENT);
 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_rel(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         sysmsg->sysmsg_result = count;
 322     return(error);
 323 }
 324