1 /* 2 * Copyright (c) 2014 Roger Pau Monné <roger.pau@citrix.com> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include <sys/param.h> 31 #include <sys/systm.h> 32 #include <sys/uio.h> 33 #include <sys/bus.h> 34 #include <sys/malloc.h> 35 #include <sys/kernel.h> 36 #include <sys/lock.h> 37 #include <sys/mutex.h> 38 #include <sys/rwlock.h> 39 #include <sys/selinfo.h> 40 #include <sys/poll.h> 41 #include <sys/conf.h> 42 #include <sys/fcntl.h> 43 #include <sys/ioccom.h> 44 #include <sys/rman.h> 45 #include <sys/tree.h> 46 #include <sys/module.h> 47 #include <sys/proc.h> 48 49 #include <vm/vm.h> 50 #include <vm/vm_param.h> 51 #include <vm/vm_extern.h> 52 #include <vm/vm_kern.h> 53 #include <vm/vm_page.h> 54 #include <vm/vm_map.h> 55 #include <vm/vm_object.h> 56 #include <vm/vm_pager.h> 57 #include <vm/vm_phys.h> 58 59 #include <machine/md_var.h> 60 61 #include <xen/xen-os.h> 62 #include <xen/hypervisor.h> 63 #include <xen/privcmd.h> 64 #include <xen/error.h> 65 66 MALLOC_DEFINE(M_PRIVCMD, "privcmd_dev", "Xen privcmd user-space device"); 67 68 struct privcmd_map { 69 vm_object_t mem; 70 vm_size_t size; 71 struct resource *pseudo_phys_res; 72 int pseudo_phys_res_id; 73 vm_paddr_t phys_base_addr; 74 boolean_t mapped; 75 int *errs; 76 }; 77 78 static d_ioctl_t privcmd_ioctl; 79 static d_mmap_single_t privcmd_mmap_single; 80 81 static struct cdevsw privcmd_devsw = { 82 .d_version = D_VERSION, 83 .d_ioctl = privcmd_ioctl, 84 .d_mmap_single = privcmd_mmap_single, 85 .d_name = "privcmd", 86 }; 87 88 static int privcmd_pg_ctor(void *handle, vm_ooffset_t size, vm_prot_t prot, 89 vm_ooffset_t foff, struct ucred *cred, u_short *color); 90 static void privcmd_pg_dtor(void *handle); 91 static int privcmd_pg_fault(vm_object_t object, vm_ooffset_t offset, 92 int prot, vm_page_t *mres); 93 94 static struct cdev_pager_ops privcmd_pg_ops = { 95 .cdev_pg_fault = privcmd_pg_fault, 96 .cdev_pg_ctor = privcmd_pg_ctor, 97 .cdev_pg_dtor = privcmd_pg_dtor, 98 }; 99 100 static device_t privcmd_dev = NULL; 101 102 /*------------------------- Privcmd Pager functions --------------------------*/ 103 static int 104 privcmd_pg_ctor(void *handle, vm_ooffset_t size, vm_prot_t prot, 105 vm_ooffset_t foff, struct ucred *cred, u_short *color) 106 { 107 108 return (0); 109 } 110 111 static void 112 privcmd_pg_dtor(void *handle) 113 { 114 struct xen_remove_from_physmap rm = { .domid = DOMID_SELF }; 115 struct privcmd_map *map = handle; 116 int error; 117 vm_size_t i; 118 vm_page_t m; 119 120 /* 121 * Remove the mappings from the used pages. This will remove the 122 * underlying p2m bindings in Xen second stage translation. 123 */ 124 if (map->mapped == true) { 125 VM_OBJECT_WLOCK(map->mem); 126 retry: 127 for (i = 0; i < map->size; i++) { 128 m = vm_page_lookup(map->mem, i); 129 if (m == NULL) 130 continue; 131 if (vm_page_sleep_if_busy(m, "pcmdum")) 132 goto retry; 133 cdev_pager_free_page(map->mem, m); 134 } 135 VM_OBJECT_WUNLOCK(map->mem); 136 137 for (i = 0; i < map->size; i++) { 138 rm.gpfn = atop(map->phys_base_addr) + i; 139 HYPERVISOR_memory_op(XENMEM_remove_from_physmap, &rm); 140 } 141 free(map->errs, M_PRIVCMD); 142 } 143 144 vm_phys_fictitious_unreg_range(map->phys_base_addr, 145 map->phys_base_addr + map->size * PAGE_SIZE); 146 147 error = bus_release_resource(privcmd_dev, SYS_RES_MEMORY, 148 map->pseudo_phys_res_id, map->pseudo_phys_res); 149 KASSERT(error == 0, ("Unable to release memory resource: %d", error)); 150 151 free(map, M_PRIVCMD); 152 } 153 154 static int 155 privcmd_pg_fault(vm_object_t object, vm_ooffset_t offset, 156 int prot, vm_page_t *mres) 157 { 158 struct privcmd_map *map = object->handle; 159 vm_pindex_t pidx; 160 vm_page_t page, oldm; 161 162 if (map->mapped != true) 163 return (VM_PAGER_FAIL); 164 165 pidx = OFF_TO_IDX(offset); 166 if (pidx >= map->size || map->errs[pidx] != 0) 167 return (VM_PAGER_FAIL); 168 169 page = PHYS_TO_VM_PAGE(map->phys_base_addr + offset); 170 if (page == NULL) 171 return (VM_PAGER_FAIL); 172 173 KASSERT((page->flags & PG_FICTITIOUS) != 0, 174 ("not fictitious %p", page)); 175 KASSERT(page->wire_count == 1, ("wire_count not 1 %p", page)); 176 KASSERT(vm_page_busied(page) == 0, ("page %p is busy", page)); 177 178 if (*mres != NULL) { 179 oldm = *mres; 180 vm_page_lock(oldm); 181 vm_page_free(oldm); 182 vm_page_unlock(oldm); 183 *mres = NULL; 184 } 185 186 vm_page_insert(page, object, pidx); 187 page->valid = VM_PAGE_BITS_ALL; 188 vm_page_xbusy(page); 189 *mres = page; 190 return (VM_PAGER_OK); 191 } 192 193 /*----------------------- Privcmd char device methods ------------------------*/ 194 static int 195 privcmd_mmap_single(struct cdev *cdev, vm_ooffset_t *offset, vm_size_t size, 196 vm_object_t *object, int nprot) 197 { 198 struct privcmd_map *map; 199 int error; 200 201 map = malloc(sizeof(*map), M_PRIVCMD, M_WAITOK | M_ZERO); 202 203 map->size = OFF_TO_IDX(size); 204 map->pseudo_phys_res_id = 0; 205 206 map->pseudo_phys_res = bus_alloc_resource(privcmd_dev, SYS_RES_MEMORY, 207 &map->pseudo_phys_res_id, 0, ~0, size, RF_ACTIVE); 208 if (map->pseudo_phys_res == NULL) { 209 free(map, M_PRIVCMD); 210 return (ENOMEM); 211 } 212 213 map->phys_base_addr = rman_get_start(map->pseudo_phys_res); 214 215 error = vm_phys_fictitious_reg_range(map->phys_base_addr, 216 map->phys_base_addr + size, VM_MEMATTR_DEFAULT); 217 if (error) { 218 bus_release_resource(privcmd_dev, SYS_RES_MEMORY, 219 map->pseudo_phys_res_id, map->pseudo_phys_res); 220 free(map, M_PRIVCMD); 221 return (error); 222 } 223 224 map->mem = cdev_pager_allocate(map, OBJT_MGTDEVICE, &privcmd_pg_ops, 225 size, nprot, *offset, NULL); 226 if (map->mem == NULL) { 227 bus_release_resource(privcmd_dev, SYS_RES_MEMORY, 228 map->pseudo_phys_res_id, map->pseudo_phys_res); 229 free(map, M_PRIVCMD); 230 return (ENOMEM); 231 } 232 233 *object = map->mem; 234 235 return (0); 236 } 237 238 static int 239 privcmd_ioctl(struct cdev *dev, unsigned long cmd, caddr_t arg, 240 int mode, struct thread *td) 241 { 242 int error, i; 243 244 switch (cmd) { 245 case IOCTL_PRIVCMD_HYPERCALL: { 246 struct ioctl_privcmd_hypercall *hcall; 247 248 hcall = (struct ioctl_privcmd_hypercall *)arg; 249 250 error = privcmd_hypercall(hcall->op, hcall->arg[0], 251 hcall->arg[1], hcall->arg[2], hcall->arg[3], hcall->arg[4]); 252 if (error >= 0) { 253 hcall->retval = error; 254 error = 0; 255 } else { 256 error = xen_translate_error(error); 257 hcall->retval = 0; 258 } 259 break; 260 } 261 case IOCTL_PRIVCMD_MMAPBATCH: { 262 struct ioctl_privcmd_mmapbatch *mmap; 263 vm_map_t map; 264 vm_map_entry_t entry; 265 vm_object_t mem; 266 vm_pindex_t index; 267 vm_prot_t prot; 268 boolean_t wired; 269 struct xen_add_to_physmap_range add; 270 xen_ulong_t *idxs; 271 xen_pfn_t *gpfns; 272 int *errs; 273 struct privcmd_map *umap; 274 275 mmap = (struct ioctl_privcmd_mmapbatch *)arg; 276 277 if ((mmap->num == 0) || 278 ((mmap->addr & PAGE_MASK) != 0)) { 279 error = EINVAL; 280 break; 281 } 282 283 map = &td->td_proc->p_vmspace->vm_map; 284 error = vm_map_lookup(&map, mmap->addr, VM_PROT_NONE, &entry, 285 &mem, &index, &prot, &wired); 286 if (error != KERN_SUCCESS) { 287 error = EINVAL; 288 break; 289 } 290 if ((entry->start != mmap->addr) || 291 (entry->end != mmap->addr + (mmap->num * PAGE_SIZE))) { 292 vm_map_lookup_done(map, entry); 293 error = EINVAL; 294 break; 295 } 296 vm_map_lookup_done(map, entry); 297 if ((mem->type != OBJT_MGTDEVICE) || 298 (mem->un_pager.devp.ops != &privcmd_pg_ops)) { 299 error = EINVAL; 300 break; 301 } 302 umap = mem->handle; 303 304 add.domid = DOMID_SELF; 305 add.space = XENMAPSPACE_gmfn_foreign; 306 add.size = mmap->num; 307 add.foreign_domid = mmap->dom; 308 309 idxs = malloc(sizeof(*idxs) * mmap->num, M_PRIVCMD, 310 M_WAITOK | M_ZERO); 311 gpfns = malloc(sizeof(*gpfns) * mmap->num, M_PRIVCMD, 312 M_WAITOK | M_ZERO); 313 errs = malloc(sizeof(*errs) * mmap->num, M_PRIVCMD, 314 M_WAITOK | M_ZERO); 315 316 set_xen_guest_handle(add.idxs, idxs); 317 set_xen_guest_handle(add.gpfns, gpfns); 318 set_xen_guest_handle(add.errs, errs); 319 320 error = copyin(&mmap->arr[0], idxs, 321 sizeof(idxs[0]) * mmap->num); 322 if (error != 0) 323 goto mmap_out; 324 325 for (i = 0; i < mmap->num; i++) 326 gpfns[i] = atop(umap->phys_base_addr + i * PAGE_SIZE); 327 328 error = HYPERVISOR_memory_op(XENMEM_add_to_physmap_range, &add); 329 if (error) { 330 error = xen_translate_error(error); 331 goto mmap_out; 332 } 333 334 for (i = 0; i < mmap->num; i++) { 335 if (errs[i] != 0) 336 errs[i] = xen_translate_error(errs[i]); 337 } 338 339 /* 340 * Save errs, so we know which pages have been 341 * successfully mapped. 342 */ 343 umap->errs = errs; 344 umap->mapped = true; 345 346 error = copyout(errs, &mmap->err[0], 347 sizeof(errs[0]) * mmap->num); 348 349 mmap_out: 350 free(idxs, M_PRIVCMD); 351 free(gpfns, M_PRIVCMD); 352 if (!umap->mapped) 353 free(errs, M_PRIVCMD); 354 355 break; 356 } 357 358 default: 359 error = ENOSYS; 360 break; 361 } 362 363 return (error); 364 } 365 366 /*------------------ Private Device Attachment Functions --------------------*/ 367 static void 368 privcmd_identify(driver_t *driver, device_t parent) 369 { 370 371 KASSERT(xen_domain(), 372 ("Trying to attach privcmd device on non Xen domain")); 373 374 if (BUS_ADD_CHILD(parent, 0, "privcmd", 0) == NULL) 375 panic("unable to attach privcmd user-space device"); 376 } 377 378 static int 379 privcmd_probe(device_t dev) 380 { 381 382 privcmd_dev = dev; 383 device_set_desc(dev, "Xen privileged interface user-space device"); 384 return (BUS_PROBE_NOWILDCARD); 385 } 386 387 static int 388 privcmd_attach(device_t dev) 389 { 390 391 make_dev_credf(MAKEDEV_ETERNAL, &privcmd_devsw, 0, NULL, UID_ROOT, 392 GID_WHEEL, 0600, "xen/privcmd"); 393 return (0); 394 } 395 396 /*-------------------- Private Device Attachment Data -----------------------*/ 397 static device_method_t privcmd_methods[] = { 398 DEVMETHOD(device_identify, privcmd_identify), 399 DEVMETHOD(device_probe, privcmd_probe), 400 DEVMETHOD(device_attach, privcmd_attach), 401 402 DEVMETHOD_END 403 }; 404 405 static driver_t privcmd_driver = { 406 "privcmd", 407 privcmd_methods, 408 0, 409 }; 410 411 devclass_t privcmd_devclass; 412 413 DRIVER_MODULE(privcmd, xenpv, privcmd_driver, privcmd_devclass, 0, 0); 414 MODULE_DEPEND(privcmd, xenpv, 1, 1, 1); 415