| 1 | /* $NetBSD: xbd_xenbus.c,v 1.75 2015/10/25 07:51:16 maxv Exp $ */ |
| 2 | |
| 3 | /* |
| 4 | * Copyright (c) 2006 Manuel Bouyer. |
| 5 | * |
| 6 | * Redistribution and use in source and binary forms, with or without |
| 7 | * modification, are permitted provided that the following conditions |
| 8 | * are met: |
| 9 | * 1. Redistributions of source code must retain the above copyright |
| 10 | * notice, this list of conditions and the following disclaimer. |
| 11 | * 2. Redistributions in binary form must reproduce the above copyright |
| 12 | * notice, this list of conditions and the following disclaimer in the |
| 13 | * documentation and/or other materials provided with the distribution. |
| 14 | * |
| 15 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR |
| 16 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
| 17 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
| 18 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, |
| 19 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
| 20 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 21 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 22 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 23 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
| 24 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 25 | * |
| 26 | */ |
| 27 | |
| 28 | /* |
| 29 | * The file contains the xbd frontend code required for block-level |
| 30 | * communications (similar to hard disks) between two Xen domains. |
| 31 | * |
| 32 | * We are not supposed to receive solicitations spontaneously from backend. The |
| 33 | * protocol is therefore fairly simple and uses only one ring to communicate |
| 34 | * with backend: frontend posts requests to the ring then wait for their |
| 35 | * replies asynchronously. |
| 36 | * |
| 37 | * xbd follows NetBSD's disk(9) convention. At any time, a LWP can schedule |
| 38 | * an operation request for the device (be it open(), read(), write(), ...). |
| 39 | * Calls are typically processed that way: |
| 40 | * - initiate request: xbdread/write/open/ioctl/.. |
| 41 | * - depending on operation, it is handled directly by disk(9) subsystem or |
| 42 | * goes through physio(9) first. |
| 43 | * - the request is ultimately processed by xbd_diskstart() that prepares the |
| 44 | * xbd requests, post them in the ring I/O queue, then signal the backend. |
| 45 | * |
| 46 | * When a response is available in the queue, the backend signals the frontend |
| 47 | * via its event channel. This triggers xbd_handler(), which will link back |
| 48 | * the response to its request through the request ID, and mark the I/O as |
| 49 | * completed. |
| 50 | */ |
| 51 | |
| 52 | #include <sys/cdefs.h> |
| 53 | __KERNEL_RCSID(0, "$NetBSD: xbd_xenbus.c,v 1.75 2015/10/25 07:51:16 maxv Exp $" ); |
| 54 | |
| 55 | #include "opt_xen.h" |
| 56 | |
| 57 | |
| 58 | #include <sys/param.h> |
| 59 | #include <sys/buf.h> |
| 60 | #include <sys/bufq.h> |
| 61 | #include <sys/device.h> |
| 62 | #include <sys/disk.h> |
| 63 | #include <sys/disklabel.h> |
| 64 | #include <sys/conf.h> |
| 65 | #include <sys/fcntl.h> |
| 66 | #include <sys/kernel.h> |
| 67 | #include <sys/proc.h> |
| 68 | #include <sys/systm.h> |
| 69 | #include <sys/stat.h> |
| 70 | #include <sys/vnode.h> |
| 71 | |
| 72 | #include <dev/dkvar.h> |
| 73 | |
| 74 | #include <uvm/uvm.h> |
| 75 | |
| 76 | #include <xen/hypervisor.h> |
| 77 | #include <xen/evtchn.h> |
| 78 | #include <xen/granttables.h> |
| 79 | #include <xen/xen-public/io/blkif.h> |
| 80 | #include <xen/xen-public/io/protocols.h> |
| 81 | |
| 82 | #include <xen/xenbus.h> |
| 83 | #include "locators.h" |
| 84 | |
| 85 | #undef XBD_DEBUG |
| 86 | #ifdef XBD_DEBUG |
| 87 | #define DPRINTF(x) printf x; |
| 88 | #else |
| 89 | #define DPRINTF(x) |
| 90 | #endif |
| 91 | |
| 92 | #define GRANT_INVALID_REF -1 |
| 93 | |
| 94 | #define XBD_RING_SIZE __CONST_RING_SIZE(blkif, PAGE_SIZE) |
| 95 | #define XBD_MAX_XFER (PAGE_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST) |
| 96 | |
| 97 | #define XEN_BSHIFT 9 /* log2(XEN_BSIZE) */ |
| 98 | #define XEN_BSIZE (1 << XEN_BSHIFT) |
| 99 | |
| 100 | struct xbd_req { |
| 101 | SLIST_ENTRY(xbd_req) req_next; |
| 102 | uint16_t req_id; /* ID passed to backend */ |
| 103 | union { |
| 104 | struct { |
| 105 | grant_ref_t req_gntref[BLKIF_MAX_SEGMENTS_PER_REQUEST]; |
| 106 | int req_nr_segments; /* number of segments in this request */ |
| 107 | struct buf *req_bp; /* buffer associated with this request */ |
| 108 | void *req_data; /* pointer to the data buffer */ |
| 109 | } req_rw; |
| 110 | struct { |
| 111 | int s_error; |
| 112 | volatile int s_done; |
| 113 | } req_sync; |
| 114 | } u; |
| 115 | }; |
| 116 | #define req_gntref u.req_rw.req_gntref |
| 117 | #define req_nr_segments u.req_rw.req_nr_segments |
| 118 | #define req_bp u.req_rw.req_bp |
| 119 | #define req_data u.req_rw.req_data |
| 120 | #define req_sync u.req_sync |
| 121 | |
| 122 | struct xbd_xenbus_softc { |
| 123 | struct dk_softc sc_dksc; /* Must be first in this struct */ |
| 124 | struct xenbus_device *sc_xbusd; |
| 125 | |
| 126 | blkif_front_ring_t sc_ring; |
| 127 | |
| 128 | unsigned int sc_evtchn; |
| 129 | |
| 130 | grant_ref_t sc_ring_gntref; |
| 131 | |
| 132 | struct xbd_req sc_reqs[XBD_RING_SIZE]; |
| 133 | SLIST_HEAD(,xbd_req) sc_xbdreq_head; /* list of free requests */ |
| 134 | bool sc_xbdreq_wait; /* special waiting on xbd_req */ |
| 135 | |
| 136 | int sc_backend_status; /* our status with backend */ |
| 137 | #define BLKIF_STATE_DISCONNECTED 0 |
| 138 | #define BLKIF_STATE_CONNECTED 1 |
| 139 | #define BLKIF_STATE_SUSPENDED 2 |
| 140 | |
| 141 | int sc_shutdown; |
| 142 | #define BLKIF_SHUTDOWN_RUN 0 /* no shutdown */ |
| 143 | #define BLKIF_SHUTDOWN_REMOTE 1 /* backend-initiated shutdown in progress */ |
| 144 | #define BLKIF_SHUTDOWN_LOCAL 2 /* locally-initiated shutdown in progress */ |
| 145 | |
| 146 | uint64_t sc_sectors; /* number of sectors for this device */ |
| 147 | u_long sc_secsize; /* sector size */ |
| 148 | uint64_t sc_xbdsize; /* size of disk in DEV_BSIZE */ |
| 149 | u_long sc_info; /* VDISK_* */ |
| 150 | u_long sc_handle; /* from backend */ |
| 151 | int sc_cache_flush; /* backend supports BLKIF_OP_FLUSH_DISKCACHE */ |
| 152 | }; |
| 153 | |
| 154 | #if 0 |
| 155 | /* too big to be on stack */ |
| 156 | static multicall_entry_t rq_mcl[XBD_RING_SIZE+1]; |
| 157 | static paddr_t rq_pages[XBD_RING_SIZE]; |
| 158 | #endif |
| 159 | |
| 160 | static int xbd_xenbus_match(device_t, cfdata_t, void *); |
| 161 | static void xbd_xenbus_attach(device_t, device_t, void *); |
| 162 | static int xbd_xenbus_detach(device_t, int); |
| 163 | |
| 164 | static bool xbd_xenbus_suspend(device_t, const pmf_qual_t *); |
| 165 | static bool xbd_xenbus_resume(device_t, const pmf_qual_t *); |
| 166 | |
| 167 | static int xbd_handler(void *); |
| 168 | static int xbd_diskstart(device_t, struct buf *); |
| 169 | static void xbd_backend_changed(void *, XenbusState); |
| 170 | static void xbd_connect(struct xbd_xenbus_softc *); |
| 171 | |
| 172 | static int xbd_map_align(struct xbd_req *); |
| 173 | static void xbd_unmap_align(struct xbd_req *); |
| 174 | |
| 175 | static void xbdminphys(struct buf *); |
| 176 | |
| 177 | CFATTACH_DECL3_NEW(xbd, sizeof(struct xbd_xenbus_softc), |
| 178 | xbd_xenbus_match, xbd_xenbus_attach, xbd_xenbus_detach, NULL, NULL, NULL, |
| 179 | DVF_DETACH_SHUTDOWN); |
| 180 | |
| 181 | dev_type_open(xbdopen); |
| 182 | dev_type_close(xbdclose); |
| 183 | dev_type_read(xbdread); |
| 184 | dev_type_write(xbdwrite); |
| 185 | dev_type_ioctl(xbdioctl); |
| 186 | dev_type_strategy(xbdstrategy); |
| 187 | dev_type_dump(xbddump); |
| 188 | dev_type_size(xbdsize); |
| 189 | |
| 190 | const struct bdevsw xbd_bdevsw = { |
| 191 | .d_open = xbdopen, |
| 192 | .d_close = xbdclose, |
| 193 | .d_strategy = xbdstrategy, |
| 194 | .d_ioctl = xbdioctl, |
| 195 | .d_dump = xbddump, |
| 196 | .d_psize = xbdsize, |
| 197 | .d_discard = nodiscard, |
| 198 | .d_flag = D_DISK |
| 199 | }; |
| 200 | |
| 201 | const struct cdevsw xbd_cdevsw = { |
| 202 | .d_open = xbdopen, |
| 203 | .d_close = xbdclose, |
| 204 | .d_read = xbdread, |
| 205 | .d_write = xbdwrite, |
| 206 | .d_ioctl = xbdioctl, |
| 207 | .d_stop = nostop, |
| 208 | .d_tty = notty, |
| 209 | .d_poll = nopoll, |
| 210 | .d_mmap = nommap, |
| 211 | .d_kqfilter = nokqfilter, |
| 212 | .d_discard = nodiscard, |
| 213 | .d_flag = D_DISK |
| 214 | }; |
| 215 | |
| 216 | extern struct cfdriver xbd_cd; |
| 217 | |
| 218 | static struct dkdriver xbddkdriver = { |
| 219 | .d_strategy = xbdstrategy, |
| 220 | .d_minphys = xbdminphys, |
| 221 | .d_open = xbdopen, |
| 222 | .d_close = xbdclose, |
| 223 | .d_diskstart = xbd_diskstart, |
| 224 | }; |
| 225 | |
| 226 | static int |
| 227 | xbd_xenbus_match(device_t parent, cfdata_t match, void *aux) |
| 228 | { |
| 229 | struct xenbusdev_attach_args *xa = aux; |
| 230 | |
| 231 | if (strcmp(xa->xa_type, "vbd" ) != 0) |
| 232 | return 0; |
| 233 | |
| 234 | if (match->cf_loc[XENBUSCF_ID] != XENBUSCF_ID_DEFAULT && |
| 235 | match->cf_loc[XENBUSCF_ID] != xa->xa_id) |
| 236 | return 0; |
| 237 | |
| 238 | return 1; |
| 239 | } |
| 240 | |
| 241 | static void |
| 242 | xbd_xenbus_attach(device_t parent, device_t self, void *aux) |
| 243 | { |
| 244 | struct xbd_xenbus_softc *sc = device_private(self); |
| 245 | struct xenbusdev_attach_args *xa = aux; |
| 246 | blkif_sring_t *ring; |
| 247 | RING_IDX i; |
| 248 | #ifdef XBD_DEBUG |
| 249 | char **dir, *val; |
| 250 | int dir_n = 0; |
| 251 | char id_str[20]; |
| 252 | int err; |
| 253 | #endif |
| 254 | |
| 255 | config_pending_incr(self); |
| 256 | aprint_normal(": Xen Virtual Block Device Interface\n" ); |
| 257 | |
| 258 | dk_init(&sc->sc_dksc, self, DKTYPE_ESDI); |
| 259 | disk_init(&sc->sc_dksc.sc_dkdev, device_xname(self), &xbddkdriver); |
| 260 | |
| 261 | #ifdef XBD_DEBUG |
| 262 | printf("path: %s\n" , xa->xa_xbusd->xbusd_path); |
| 263 | snprintf(id_str, sizeof(id_str), "%d" , xa->xa_id); |
| 264 | err = xenbus_directory(NULL, "device/vbd" , id_str, &dir_n, &dir); |
| 265 | if (err) { |
| 266 | aprint_error_dev(self, "xenbus_directory err %d\n" , err); |
| 267 | } else { |
| 268 | printf("%s/\n" , xa->xa_xbusd->xbusd_path); |
| 269 | for (i = 0; i < dir_n; i++) { |
| 270 | printf("\t/%s" , dir[i]); |
| 271 | err = xenbus_read(NULL, xa->xa_xbusd->xbusd_path, |
| 272 | dir[i], NULL, &val); |
| 273 | if (err) { |
| 274 | aprint_error_dev(self, "xenbus_read err %d\n" , |
| 275 | err); |
| 276 | } else { |
| 277 | printf(" = %s\n" , val); |
| 278 | free(val, M_DEVBUF); |
| 279 | } |
| 280 | } |
| 281 | } |
| 282 | #endif /* XBD_DEBUG */ |
| 283 | sc->sc_xbusd = xa->xa_xbusd; |
| 284 | sc->sc_xbusd->xbusd_otherend_changed = xbd_backend_changed; |
| 285 | |
| 286 | /* initialize free requests list */ |
| 287 | SLIST_INIT(&sc->sc_xbdreq_head); |
| 288 | for (i = 0; i < XBD_RING_SIZE; i++) { |
| 289 | sc->sc_reqs[i].req_id = i; |
| 290 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, &sc->sc_reqs[i], |
| 291 | req_next); |
| 292 | } |
| 293 | |
| 294 | sc->sc_backend_status = BLKIF_STATE_DISCONNECTED; |
| 295 | sc->sc_shutdown = BLKIF_SHUTDOWN_REMOTE; |
| 296 | |
| 297 | ring = (void *)uvm_km_alloc(kernel_map, PAGE_SIZE, 0, UVM_KMF_WIRED); |
| 298 | if (ring == NULL) |
| 299 | panic("%s: can't alloc ring" , device_xname(self)); |
| 300 | sc->sc_ring.sring = ring; |
| 301 | |
| 302 | /* resume shared structures and tell backend that we are ready */ |
| 303 | if (xbd_xenbus_resume(self, PMF_Q_NONE) == false) { |
| 304 | uvm_km_free(kernel_map, (vaddr_t)ring, PAGE_SIZE, |
| 305 | UVM_KMF_WIRED); |
| 306 | return; |
| 307 | } |
| 308 | |
| 309 | if (!pmf_device_register(self, xbd_xenbus_suspend, xbd_xenbus_resume)) |
| 310 | aprint_error_dev(self, "couldn't establish power handler\n" ); |
| 311 | |
| 312 | } |
| 313 | |
| 314 | static int |
| 315 | xbd_xenbus_detach(device_t dev, int flags) |
| 316 | { |
| 317 | struct xbd_xenbus_softc *sc = device_private(dev); |
| 318 | int bmaj, cmaj, i, mn, rc, s; |
| 319 | |
| 320 | rc = disk_begindetach(&sc->sc_dksc.sc_dkdev, NULL, dev, flags); |
| 321 | if (rc != 0) |
| 322 | return rc; |
| 323 | |
| 324 | s = splbio(); |
| 325 | DPRINTF(("%s: xbd_detach\n" , device_xname(dev))); |
| 326 | if (sc->sc_shutdown == BLKIF_SHUTDOWN_RUN) { |
| 327 | sc->sc_shutdown = BLKIF_SHUTDOWN_LOCAL; |
| 328 | /* wait for requests to complete */ |
| 329 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && |
| 330 | sc->sc_dksc.sc_dkdev.dk_stats->io_busy > 0) |
| 331 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach" , hz/2); |
| 332 | |
| 333 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateClosing); |
| 334 | } |
| 335 | if ((flags & DETACH_FORCE) == 0) { |
| 336 | /* xbd_xenbus_detach already in progress */ |
| 337 | wakeup(xbd_xenbus_detach); |
| 338 | splx(s); |
| 339 | return EALREADY; |
| 340 | } |
| 341 | while (xenbus_read_driver_state(sc->sc_xbusd->xbusd_otherend) |
| 342 | != XenbusStateClosed) |
| 343 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach2" , hz/2); |
| 344 | splx(s); |
| 345 | |
| 346 | /* locate the major number */ |
| 347 | bmaj = bdevsw_lookup_major(&xbd_bdevsw); |
| 348 | cmaj = cdevsw_lookup_major(&xbd_cdevsw); |
| 349 | |
| 350 | /* Nuke the vnodes for any open instances. */ |
| 351 | for (i = 0; i < MAXPARTITIONS; i++) { |
| 352 | mn = DISKMINOR(device_unit(dev), i); |
| 353 | vdevgone(bmaj, mn, mn, VBLK); |
| 354 | vdevgone(cmaj, mn, mn, VCHR); |
| 355 | } |
| 356 | if (sc->sc_backend_status == BLKIF_STATE_CONNECTED) { |
| 357 | /* Delete all of our wedges. */ |
| 358 | dkwedge_delall(&sc->sc_dksc.sc_dkdev); |
| 359 | |
| 360 | /* Kill off any queued buffers. */ |
| 361 | dk_drain(&sc->sc_dksc); |
| 362 | bufq_free(sc->sc_dksc.sc_bufq); |
| 363 | |
| 364 | /* detach disk */ |
| 365 | disk_detach(&sc->sc_dksc.sc_dkdev); |
| 366 | disk_destroy(&sc->sc_dksc.sc_dkdev); |
| 367 | dk_detach(&sc->sc_dksc); |
| 368 | } |
| 369 | |
| 370 | hypervisor_mask_event(sc->sc_evtchn); |
| 371 | event_remove_handler(sc->sc_evtchn, &xbd_handler, sc); |
| 372 | while (xengnt_status(sc->sc_ring_gntref)) { |
| 373 | tsleep(xbd_xenbus_detach, PRIBIO, "xbd_ref" , hz/2); |
| 374 | } |
| 375 | xengnt_revoke_access(sc->sc_ring_gntref); |
| 376 | uvm_km_free(kernel_map, (vaddr_t)sc->sc_ring.sring, |
| 377 | PAGE_SIZE, UVM_KMF_WIRED); |
| 378 | |
| 379 | pmf_device_deregister(dev); |
| 380 | |
| 381 | return 0; |
| 382 | } |
| 383 | |
| 384 | static bool |
| 385 | xbd_xenbus_suspend(device_t dev, const pmf_qual_t *qual) { |
| 386 | |
| 387 | int s; |
| 388 | struct xbd_xenbus_softc *sc; |
| 389 | |
| 390 | sc = device_private(dev); |
| 391 | |
| 392 | s = splbio(); |
| 393 | /* wait for requests to complete, then suspend device */ |
| 394 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && |
| 395 | sc->sc_dksc.sc_dkdev.dk_stats->io_busy > 0) |
| 396 | tsleep(xbd_xenbus_suspend, PRIBIO, "xbdsuspend" , hz/2); |
| 397 | |
| 398 | hypervisor_mask_event(sc->sc_evtchn); |
| 399 | sc->sc_backend_status = BLKIF_STATE_SUSPENDED; |
| 400 | event_remove_handler(sc->sc_evtchn, xbd_handler, sc); |
| 401 | |
| 402 | splx(s); |
| 403 | |
| 404 | xenbus_device_suspend(sc->sc_xbusd); |
| 405 | aprint_verbose_dev(dev, "removed event channel %d\n" , sc->sc_evtchn); |
| 406 | |
| 407 | return true; |
| 408 | } |
| 409 | |
| 410 | static bool |
| 411 | xbd_xenbus_resume(device_t dev, const pmf_qual_t *qual) |
| 412 | { |
| 413 | struct xbd_xenbus_softc *sc; |
| 414 | struct xenbus_transaction *xbt; |
| 415 | int error; |
| 416 | blkif_sring_t *ring; |
| 417 | paddr_t ma; |
| 418 | const char *errmsg; |
| 419 | |
| 420 | sc = device_private(dev); |
| 421 | |
| 422 | if (sc->sc_backend_status == BLKIF_STATE_SUSPENDED) { |
| 423 | /* |
| 424 | * Device was suspended, so ensure that access associated to |
| 425 | * the block I/O ring is revoked. |
| 426 | */ |
| 427 | xengnt_revoke_access(sc->sc_ring_gntref); |
| 428 | } |
| 429 | sc->sc_ring_gntref = GRANT_INVALID_REF; |
| 430 | |
| 431 | /* Initialize ring */ |
| 432 | ring = sc->sc_ring.sring; |
| 433 | memset(ring, 0, PAGE_SIZE); |
| 434 | SHARED_RING_INIT(ring); |
| 435 | FRONT_RING_INIT(&sc->sc_ring, ring, PAGE_SIZE); |
| 436 | |
| 437 | /* |
| 438 | * get MA address of the ring, and use it to set up the grant entry |
| 439 | * for the block device |
| 440 | */ |
| 441 | (void)pmap_extract_ma(pmap_kernel(), (vaddr_t)ring, &ma); |
| 442 | error = xenbus_grant_ring(sc->sc_xbusd, ma, &sc->sc_ring_gntref); |
| 443 | if (error) |
| 444 | goto abort_resume; |
| 445 | |
| 446 | error = xenbus_alloc_evtchn(sc->sc_xbusd, &sc->sc_evtchn); |
| 447 | if (error) |
| 448 | goto abort_resume; |
| 449 | |
| 450 | aprint_verbose_dev(dev, "using event channel %d\n" , |
| 451 | sc->sc_evtchn); |
| 452 | event_set_handler(sc->sc_evtchn, &xbd_handler, sc, |
| 453 | IPL_BIO, device_xname(dev)); |
| 454 | |
| 455 | again: |
| 456 | xbt = xenbus_transaction_start(); |
| 457 | if (xbt == NULL) |
| 458 | return false; |
| 459 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, |
| 460 | "ring-ref" ,"%u" , sc->sc_ring_gntref); |
| 461 | if (error) { |
| 462 | errmsg = "writing ring-ref" ; |
| 463 | goto abort_transaction; |
| 464 | } |
| 465 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, |
| 466 | "event-channel" , "%u" , sc->sc_evtchn); |
| 467 | if (error) { |
| 468 | errmsg = "writing event channel" ; |
| 469 | goto abort_transaction; |
| 470 | } |
| 471 | error = xenbus_printf(xbt, sc->sc_xbusd->xbusd_path, |
| 472 | "protocol" , "%s" , XEN_IO_PROTO_ABI_NATIVE); |
| 473 | if (error) { |
| 474 | errmsg = "writing protocol" ; |
| 475 | goto abort_transaction; |
| 476 | } |
| 477 | error = xenbus_transaction_end(xbt, 0); |
| 478 | if (error == EAGAIN) |
| 479 | goto again; |
| 480 | if (error != 0) { |
| 481 | xenbus_dev_fatal(sc->sc_xbusd, error, |
| 482 | "completing transaction" ); |
| 483 | return false; |
| 484 | } |
| 485 | |
| 486 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateInitialised); |
| 487 | |
| 488 | if (sc->sc_backend_status == BLKIF_STATE_SUSPENDED) { |
| 489 | /* |
| 490 | * device was suspended, softc structures are |
| 491 | * already initialized - we use a shortcut |
| 492 | */ |
| 493 | sc->sc_backend_status = BLKIF_STATE_CONNECTED; |
| 494 | xenbus_device_resume(sc->sc_xbusd); |
| 495 | hypervisor_enable_event(sc->sc_evtchn); |
| 496 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateConnected); |
| 497 | } |
| 498 | |
| 499 | return true; |
| 500 | |
| 501 | abort_resume: |
| 502 | xenbus_dev_fatal(sc->sc_xbusd, error, "resuming device" ); |
| 503 | return false; |
| 504 | |
| 505 | abort_transaction: |
| 506 | xenbus_transaction_end(xbt, 1); |
| 507 | xenbus_dev_fatal(sc->sc_xbusd, error, "%s" , errmsg); |
| 508 | return false; |
| 509 | } |
| 510 | |
| 511 | static void xbd_backend_changed(void *arg, XenbusState new_state) |
| 512 | { |
| 513 | struct xbd_xenbus_softc *sc = device_private((device_t)arg); |
| 514 | struct disk_geom *dg; |
| 515 | |
| 516 | char buf[9]; |
| 517 | int s; |
| 518 | DPRINTF(("%s: new backend state %d\n" , |
| 519 | device_xname(sc->sc_dksc.sc_dev), new_state)); |
| 520 | |
| 521 | switch (new_state) { |
| 522 | case XenbusStateUnknown: |
| 523 | case XenbusStateInitialising: |
| 524 | case XenbusStateInitWait: |
| 525 | case XenbusStateInitialised: |
| 526 | break; |
| 527 | case XenbusStateClosing: |
| 528 | s = splbio(); |
| 529 | if (sc->sc_shutdown == BLKIF_SHUTDOWN_RUN) |
| 530 | sc->sc_shutdown = BLKIF_SHUTDOWN_REMOTE; |
| 531 | /* wait for requests to complete */ |
| 532 | while (sc->sc_backend_status == BLKIF_STATE_CONNECTED && |
| 533 | sc->sc_dksc.sc_dkdev.dk_stats->io_busy > 0) |
| 534 | tsleep(xbd_xenbus_detach, PRIBIO, "xbddetach" , hz/2); |
| 535 | splx(s); |
| 536 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateClosed); |
| 537 | break; |
| 538 | case XenbusStateConnected: |
| 539 | /* |
| 540 | * note that xbd_backend_changed() can only be called by |
| 541 | * the xenbus thread. |
| 542 | */ |
| 543 | |
| 544 | if (sc->sc_backend_status == BLKIF_STATE_CONNECTED || |
| 545 | sc->sc_backend_status == BLKIF_STATE_SUSPENDED) |
| 546 | /* already connected */ |
| 547 | return; |
| 548 | |
| 549 | xbd_connect(sc); |
| 550 | sc->sc_shutdown = BLKIF_SHUTDOWN_RUN; |
| 551 | hypervisor_enable_event(sc->sc_evtchn); |
| 552 | |
| 553 | sc->sc_xbdsize = |
| 554 | sc->sc_sectors * (uint64_t)sc->sc_secsize / DEV_BSIZE; |
| 555 | dg = &sc->sc_dksc.sc_dkdev.dk_geom; |
| 556 | memset(dg, 0, sizeof(*dg)); |
| 557 | |
| 558 | dg->dg_secperunit = sc->sc_xbdsize; |
| 559 | dg->dg_secsize = DEV_BSIZE; |
| 560 | dg->dg_ntracks = 1; |
| 561 | // XXX: Ok to hard-code DEV_BSIZE? |
| 562 | dg->dg_nsectors = 1024 * (1024 / dg->dg_secsize); |
| 563 | dg->dg_ncylinders = dg->dg_secperunit / dg->dg_nsectors; |
| 564 | |
| 565 | bufq_alloc(&sc->sc_dksc.sc_bufq, "fcfs" , 0); |
| 566 | dk_attach(&sc->sc_dksc); |
| 567 | disk_attach(&sc->sc_dksc.sc_dkdev); |
| 568 | |
| 569 | sc->sc_backend_status = BLKIF_STATE_CONNECTED; |
| 570 | |
| 571 | /* try to read the disklabel */ |
| 572 | dk_getdisklabel(&sc->sc_dksc, 0 /* XXX ? */); |
| 573 | format_bytes(buf, sizeof(buf), sc->sc_sectors * sc->sc_secsize); |
| 574 | aprint_verbose_dev(sc->sc_dksc.sc_dev, |
| 575 | "%s, %d bytes/sect x %" PRIu64 " sectors\n" , |
| 576 | buf, (int)dg->dg_secsize, sc->sc_xbdsize); |
| 577 | /* Discover wedges on this disk. */ |
| 578 | dkwedge_discover(&sc->sc_dksc.sc_dkdev); |
| 579 | |
| 580 | disk_set_info(sc->sc_dksc.sc_dev, &sc->sc_dksc.sc_dkdev, NULL); |
| 581 | |
| 582 | /* the disk should be working now */ |
| 583 | config_pending_decr(sc->sc_dksc.sc_dev); |
| 584 | break; |
| 585 | default: |
| 586 | panic("bad backend state %d" , new_state); |
| 587 | } |
| 588 | } |
| 589 | |
| 590 | static void |
| 591 | xbd_connect(struct xbd_xenbus_softc *sc) |
| 592 | { |
| 593 | int err; |
| 594 | unsigned long long sectors; |
| 595 | u_long cache_flush; |
| 596 | |
| 597 | err = xenbus_read_ul(NULL, |
| 598 | sc->sc_xbusd->xbusd_path, "virtual-device" , &sc->sc_handle, 10); |
| 599 | if (err) |
| 600 | panic("%s: can't read number from %s/virtual-device\n" , |
| 601 | device_xname(sc->sc_dksc.sc_dev), |
| 602 | sc->sc_xbusd->xbusd_otherend); |
| 603 | err = xenbus_read_ull(NULL, |
| 604 | sc->sc_xbusd->xbusd_otherend, "sectors" , §ors, 10); |
| 605 | if (err) |
| 606 | panic("%s: can't read number from %s/sectors\n" , |
| 607 | device_xname(sc->sc_dksc.sc_dev), |
| 608 | sc->sc_xbusd->xbusd_otherend); |
| 609 | sc->sc_sectors = sectors; |
| 610 | |
| 611 | err = xenbus_read_ul(NULL, |
| 612 | sc->sc_xbusd->xbusd_otherend, "info" , &sc->sc_info, 10); |
| 613 | if (err) |
| 614 | panic("%s: can't read number from %s/info\n" , |
| 615 | device_xname(sc->sc_dksc.sc_dev), |
| 616 | sc->sc_xbusd->xbusd_otherend); |
| 617 | err = xenbus_read_ul(NULL, |
| 618 | sc->sc_xbusd->xbusd_otherend, "sector-size" , &sc->sc_secsize, 10); |
| 619 | if (err) |
| 620 | panic("%s: can't read number from %s/sector-size\n" , |
| 621 | device_xname(sc->sc_dksc.sc_dev), |
| 622 | sc->sc_xbusd->xbusd_otherend); |
| 623 | |
| 624 | err = xenbus_read_ul(NULL, sc->sc_xbusd->xbusd_otherend, |
| 625 | "feature-flush-cache" , &cache_flush, 10); |
| 626 | if (err) |
| 627 | cache_flush = 0; |
| 628 | if (cache_flush > 0) |
| 629 | sc->sc_cache_flush = 1; |
| 630 | else |
| 631 | sc->sc_cache_flush = 0; |
| 632 | |
| 633 | xenbus_switch_state(sc->sc_xbusd, NULL, XenbusStateConnected); |
| 634 | } |
| 635 | |
| 636 | static int |
| 637 | xbd_handler(void *arg) |
| 638 | { |
| 639 | struct xbd_xenbus_softc *sc = arg; |
| 640 | struct buf *bp; |
| 641 | RING_IDX resp_prod, i; |
| 642 | int more_to_do; |
| 643 | int seg; |
| 644 | |
| 645 | DPRINTF(("xbd_handler(%s)\n" , device_xname(sc->sc_dksc.sc_dev))); |
| 646 | |
| 647 | if (__predict_false(sc->sc_backend_status != BLKIF_STATE_CONNECTED)) |
| 648 | return 0; |
| 649 | again: |
| 650 | resp_prod = sc->sc_ring.sring->rsp_prod; |
| 651 | xen_rmb(); /* ensure we see replies up to resp_prod */ |
| 652 | for (i = sc->sc_ring.rsp_cons; i != resp_prod; i++) { |
| 653 | blkif_response_t *rep = RING_GET_RESPONSE(&sc->sc_ring, i); |
| 654 | struct xbd_req *xbdreq = &sc->sc_reqs[rep->id]; |
| 655 | bp = xbdreq->req_bp; |
| 656 | DPRINTF(("xbd_handler(%p): b_bcount = %ld\n" , |
| 657 | xbdreq->req_bp, (long)bp->b_bcount)); |
| 658 | if (rep->operation == BLKIF_OP_FLUSH_DISKCACHE) { |
| 659 | xbdreq->req_sync.s_error = rep->status; |
| 660 | xbdreq->req_sync.s_done = 1; |
| 661 | wakeup(xbdreq); |
| 662 | /* caller will free the req */ |
| 663 | continue; |
| 664 | } |
| 665 | for (seg = xbdreq->req_nr_segments - 1; seg >= 0; seg--) { |
| 666 | if (__predict_false( |
| 667 | xengnt_status(xbdreq->req_gntref[seg]))) { |
| 668 | aprint_verbose_dev(sc->sc_dksc.sc_dev, |
| 669 | "grant still used by backend\n" ); |
| 670 | sc->sc_ring.rsp_cons = i; |
| 671 | xbdreq->req_nr_segments = seg + 1; |
| 672 | goto done; |
| 673 | } |
| 674 | xengnt_revoke_access(xbdreq->req_gntref[seg]); |
| 675 | xbdreq->req_nr_segments--; |
| 676 | } |
| 677 | if (rep->operation != BLKIF_OP_READ && |
| 678 | rep->operation != BLKIF_OP_WRITE) { |
| 679 | aprint_error_dev(sc->sc_dksc.sc_dev, |
| 680 | "bad operation %d from backend\n" , rep->operation); |
| 681 | bp->b_error = EIO; |
| 682 | bp->b_resid = bp->b_bcount; |
| 683 | goto next; |
| 684 | } |
| 685 | if (rep->status != BLKIF_RSP_OKAY) { |
| 686 | bp->b_error = EIO; |
| 687 | bp->b_resid = bp->b_bcount; |
| 688 | goto next; |
| 689 | } |
| 690 | /* b_resid was set in dk_start */ |
| 691 | next: |
| 692 | if (bp->b_data != xbdreq->req_data) |
| 693 | xbd_unmap_align(xbdreq); |
| 694 | |
| 695 | dk_done(&sc->sc_dksc, bp); |
| 696 | |
| 697 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, xbdreq, req_next); |
| 698 | } |
| 699 | done: |
| 700 | xen_rmb(); |
| 701 | sc->sc_ring.rsp_cons = i; |
| 702 | |
| 703 | RING_FINAL_CHECK_FOR_RESPONSES(&sc->sc_ring, more_to_do); |
| 704 | if (more_to_do) |
| 705 | goto again; |
| 706 | |
| 707 | if (sc->sc_xbdreq_wait) |
| 708 | wakeup(&sc->sc_xbdreq_wait); |
| 709 | else |
| 710 | dk_start(&sc->sc_dksc, NULL); |
| 711 | return 1; |
| 712 | } |
| 713 | |
| 714 | static void |
| 715 | xbdminphys(struct buf *bp) |
| 716 | { |
| 717 | if (bp->b_bcount > XBD_MAX_XFER) { |
| 718 | bp->b_bcount = XBD_MAX_XFER; |
| 719 | } |
| 720 | minphys(bp); |
| 721 | } |
| 722 | |
| 723 | int |
| 724 | xbdopen(dev_t dev, int flags, int fmt, struct lwp *l) |
| 725 | { |
| 726 | struct xbd_xenbus_softc *sc; |
| 727 | |
| 728 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
| 729 | if (sc == NULL) |
| 730 | return (ENXIO); |
| 731 | if ((flags & FWRITE) && (sc->sc_info & VDISK_READONLY)) |
| 732 | return EROFS; |
| 733 | |
| 734 | DPRINTF(("xbdopen(0x%04x, %d)\n" , dev, flags)); |
| 735 | return dk_open(&sc->sc_dksc, dev, flags, fmt, l); |
| 736 | } |
| 737 | |
| 738 | int |
| 739 | xbdclose(dev_t dev, int flags, int fmt, struct lwp *l) |
| 740 | { |
| 741 | struct xbd_xenbus_softc *sc; |
| 742 | |
| 743 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
| 744 | |
| 745 | DPRINTF(("xbdclose(%d, %d)\n" , dev, flags)); |
| 746 | return dk_close(&sc->sc_dksc, dev, flags, fmt, l); |
| 747 | } |
| 748 | |
| 749 | void |
| 750 | xbdstrategy(struct buf *bp) |
| 751 | { |
| 752 | struct xbd_xenbus_softc *sc; |
| 753 | |
| 754 | sc = device_lookup_private(&xbd_cd, DISKUNIT(bp->b_dev)); |
| 755 | |
| 756 | DPRINTF(("xbdstrategy(%p): b_bcount = %ld\n" , bp, |
| 757 | (long)bp->b_bcount)); |
| 758 | |
| 759 | if (sc == NULL || sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) { |
| 760 | bp->b_error = EIO; |
| 761 | biodone(bp); |
| 762 | return; |
| 763 | } |
| 764 | if (__predict_false((sc->sc_info & VDISK_READONLY) && |
| 765 | (bp->b_flags & B_READ) == 0)) { |
| 766 | bp->b_error = EROFS; |
| 767 | biodone(bp); |
| 768 | return; |
| 769 | } |
| 770 | |
| 771 | dk_strategy(&sc->sc_dksc, bp); |
| 772 | return; |
| 773 | } |
| 774 | |
| 775 | int |
| 776 | xbdsize(dev_t dev) |
| 777 | { |
| 778 | struct xbd_xenbus_softc *sc; |
| 779 | |
| 780 | DPRINTF(("xbdsize(%d)\n" , dev)); |
| 781 | |
| 782 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
| 783 | if (sc == NULL || sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) |
| 784 | return -1; |
| 785 | return dk_size(&sc->sc_dksc, dev); |
| 786 | } |
| 787 | |
| 788 | int |
| 789 | xbdread(dev_t dev, struct uio *uio, int flags) |
| 790 | { |
| 791 | struct xbd_xenbus_softc *sc = |
| 792 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
| 793 | struct dk_softc *dksc = &sc->sc_dksc; |
| 794 | |
| 795 | if (!DK_ATTACHED(dksc)) |
| 796 | return ENXIO; |
| 797 | return physio(xbdstrategy, NULL, dev, B_READ, xbdminphys, uio); |
| 798 | } |
| 799 | |
| 800 | int |
| 801 | xbdwrite(dev_t dev, struct uio *uio, int flags) |
| 802 | { |
| 803 | struct xbd_xenbus_softc *sc = |
| 804 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
| 805 | struct dk_softc *dksc = &sc->sc_dksc; |
| 806 | |
| 807 | if (!DK_ATTACHED(dksc)) |
| 808 | return ENXIO; |
| 809 | if (__predict_false(sc->sc_info & VDISK_READONLY)) |
| 810 | return EROFS; |
| 811 | return physio(xbdstrategy, NULL, dev, B_WRITE, xbdminphys, uio); |
| 812 | } |
| 813 | |
| 814 | int |
| 815 | xbdioctl(dev_t dev, u_long cmd, void *data, int flag, struct lwp *l) |
| 816 | { |
| 817 | struct xbd_xenbus_softc *sc = |
| 818 | device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
| 819 | struct dk_softc *dksc; |
| 820 | int error; |
| 821 | int s; |
| 822 | struct xbd_req *xbdreq; |
| 823 | blkif_request_t *req; |
| 824 | int notify; |
| 825 | |
| 826 | DPRINTF(("xbdioctl(%d, %08lx, %p, %d, %p)\n" , |
| 827 | dev, cmd, data, flag, l)); |
| 828 | dksc = &sc->sc_dksc; |
| 829 | |
| 830 | switch (cmd) { |
| 831 | case DIOCSSTRATEGY: |
| 832 | error = EOPNOTSUPP; |
| 833 | break; |
| 834 | case DIOCCACHESYNC: |
| 835 | if (sc->sc_cache_flush <= 0) { |
| 836 | if (sc->sc_cache_flush == 0) { |
| 837 | aprint_error_dev(sc->sc_dksc.sc_dev, |
| 838 | "WARNING: cache flush not supported " |
| 839 | "by backend\n" ); |
| 840 | sc->sc_cache_flush = -1; |
| 841 | } |
| 842 | return EOPNOTSUPP; |
| 843 | } |
| 844 | |
| 845 | s = splbio(); |
| 846 | |
| 847 | while (RING_FULL(&sc->sc_ring)) { |
| 848 | sc->sc_xbdreq_wait = 1; |
| 849 | tsleep(&sc->sc_xbdreq_wait, PRIBIO, "xbdreq" , 0); |
| 850 | } |
| 851 | sc->sc_xbdreq_wait = 0; |
| 852 | |
| 853 | xbdreq = SLIST_FIRST(&sc->sc_xbdreq_head); |
| 854 | if (__predict_false(xbdreq == NULL)) { |
| 855 | DPRINTF(("xbdioctl: no req\n" )); |
| 856 | error = ENOMEM; |
| 857 | } else { |
| 858 | SLIST_REMOVE_HEAD(&sc->sc_xbdreq_head, req_next); |
| 859 | req = RING_GET_REQUEST(&sc->sc_ring, |
| 860 | sc->sc_ring.req_prod_pvt); |
| 861 | req->id = xbdreq->req_id; |
| 862 | req->operation = BLKIF_OP_FLUSH_DISKCACHE; |
| 863 | req->handle = sc->sc_handle; |
| 864 | xbdreq->req_sync.s_done = 0; |
| 865 | sc->sc_ring.req_prod_pvt++; |
| 866 | RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->sc_ring, |
| 867 | notify); |
| 868 | if (notify) |
| 869 | hypervisor_notify_via_evtchn(sc->sc_evtchn); |
| 870 | /* request sent, no wait for completion */ |
| 871 | while (xbdreq->req_sync.s_done == 0) { |
| 872 | tsleep(xbdreq, PRIBIO, "xbdsync" , 0); |
| 873 | } |
| 874 | if (xbdreq->req_sync.s_error == BLKIF_RSP_EOPNOTSUPP) |
| 875 | error = EOPNOTSUPP; |
| 876 | else if (xbdreq->req_sync.s_error == BLKIF_RSP_OKAY) |
| 877 | error = 0; |
| 878 | else |
| 879 | error = EIO; |
| 880 | SLIST_INSERT_HEAD(&sc->sc_xbdreq_head, xbdreq, |
| 881 | req_next); |
| 882 | } |
| 883 | splx(s); |
| 884 | break; |
| 885 | |
| 886 | default: |
| 887 | error = dk_ioctl(dksc, dev, cmd, data, flag, l); |
| 888 | break; |
| 889 | } |
| 890 | |
| 891 | return error; |
| 892 | } |
| 893 | |
| 894 | int |
| 895 | xbddump(dev_t dev, daddr_t blkno, void *va, size_t size) |
| 896 | { |
| 897 | struct xbd_xenbus_softc *sc; |
| 898 | |
| 899 | sc = device_lookup_private(&xbd_cd, DISKUNIT(dev)); |
| 900 | if (sc == NULL) |
| 901 | return (ENXIO); |
| 902 | |
| 903 | DPRINTF(("xbddump(%d, %" PRId64 ", %p, %lu)\n" , dev, blkno, va, |
| 904 | (unsigned long)size)); |
| 905 | return dk_dump(&sc->sc_dksc, dev, blkno, va, size); |
| 906 | } |
| 907 | |
| 908 | static int |
| 909 | xbd_diskstart(device_t self, struct buf *bp) |
| 910 | { |
| 911 | struct xbd_xenbus_softc *sc = device_private(self); |
| 912 | struct xbd_req *xbdreq; |
| 913 | blkif_request_t *req; |
| 914 | size_t bcount, off; |
| 915 | paddr_t ma; |
| 916 | vaddr_t va; |
| 917 | int nsects, nbytes, seg; |
| 918 | int notify, error = 0; |
| 919 | |
| 920 | DPRINTF(("xbd_diskstart(%p): b_bcount = %ld\n" , |
| 921 | bp, (long)bp->b_bcount)); |
| 922 | |
| 923 | if (sc->sc_shutdown != BLKIF_SHUTDOWN_RUN) { |
| 924 | error = EIO; |
| 925 | goto err; |
| 926 | } |
| 927 | |
| 928 | if (bp->b_rawblkno < 0 || bp->b_rawblkno > sc->sc_xbdsize) { |
| 929 | /* invalid block number */ |
| 930 | error = EINVAL; |
| 931 | goto err; |
| 932 | } |
| 933 | |
| 934 | if (__predict_false( |
| 935 | sc->sc_backend_status == BLKIF_STATE_SUSPENDED)) { |
| 936 | /* device is suspended, do not consume buffer */ |
| 937 | DPRINTF(("%s: (xbd_diskstart) device suspended\n" , |
| 938 | sc->sc_dksc.sc_xname)); |
| 939 | error = EAGAIN; |
| 940 | goto out; |
| 941 | } |
| 942 | |
| 943 | if (RING_FULL(&sc->sc_ring) || sc->sc_xbdreq_wait) { |
| 944 | DPRINTF(("xbd_diskstart: ring_full\n" )); |
| 945 | error = EAGAIN; |
| 946 | goto out; |
| 947 | } |
| 948 | |
| 949 | xbdreq = SLIST_FIRST(&sc->sc_xbdreq_head); |
| 950 | if (__predict_false(xbdreq == NULL)) { |
| 951 | DPRINTF(("xbd_diskstart: no req\n" )); |
| 952 | error = EAGAIN; |
| 953 | goto out; |
| 954 | } |
| 955 | |
| 956 | xbdreq->req_bp = bp; |
| 957 | xbdreq->req_data = bp->b_data; |
| 958 | if ((vaddr_t)bp->b_data & (XEN_BSIZE - 1)) { |
| 959 | if (__predict_false(xbd_map_align(xbdreq) != 0)) { |
| 960 | DPRINTF(("xbd_diskstart: no align\n" )); |
| 961 | error = EAGAIN; |
| 962 | goto out; |
| 963 | } |
| 964 | } |
| 965 | |
| 966 | SLIST_REMOVE_HEAD(&sc->sc_xbdreq_head, req_next); |
| 967 | req = RING_GET_REQUEST(&sc->sc_ring, sc->sc_ring.req_prod_pvt); |
| 968 | req->id = xbdreq->req_id; |
| 969 | req->operation = |
| 970 | bp->b_flags & B_READ ? BLKIF_OP_READ : BLKIF_OP_WRITE; |
| 971 | req->sector_number = bp->b_rawblkno; |
| 972 | req->handle = sc->sc_handle; |
| 973 | |
| 974 | va = (vaddr_t)xbdreq->req_data & ~PAGE_MASK; |
| 975 | off = (vaddr_t)xbdreq->req_data & PAGE_MASK; |
| 976 | bcount = bp->b_bcount; |
| 977 | bp->b_resid = 0; |
| 978 | for (seg = 0; bcount > 0;) { |
| 979 | pmap_extract_ma(pmap_kernel(), va, &ma); |
| 980 | KASSERT((ma & (XEN_BSIZE - 1)) == 0); |
| 981 | if (bcount > PAGE_SIZE - off) |
| 982 | nbytes = PAGE_SIZE - off; |
| 983 | else |
| 984 | nbytes = bcount; |
| 985 | nsects = nbytes >> XEN_BSHIFT; |
| 986 | req->seg[seg].first_sect = off >> XEN_BSHIFT; |
| 987 | req->seg[seg].last_sect = |
| 988 | (off >> XEN_BSHIFT) + nsects - 1; |
| 989 | KASSERT(req->seg[seg].first_sect <= |
| 990 | req->seg[seg].last_sect); |
| 991 | KASSERT(req->seg[seg].last_sect < 8); |
| 992 | if (__predict_false(xengnt_grant_access( |
| 993 | sc->sc_xbusd->xbusd_otherend_id, ma, |
| 994 | (bp->b_flags & B_READ) == 0, |
| 995 | &xbdreq->req_gntref[seg]))) |
| 996 | panic("xbd_diskstart: xengnt_grant_access" ); /* XXX XXX !!! */ |
| 997 | req->seg[seg].gref = xbdreq->req_gntref[seg]; |
| 998 | seg++; |
| 999 | KASSERT(seg <= BLKIF_MAX_SEGMENTS_PER_REQUEST); |
| 1000 | va += PAGE_SIZE; |
| 1001 | off = 0; |
| 1002 | bcount -= nbytes; |
| 1003 | } |
| 1004 | xbdreq->req_nr_segments = req->nr_segments = seg; |
| 1005 | sc->sc_ring.req_prod_pvt++; |
| 1006 | |
| 1007 | out: |
| 1008 | RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&sc->sc_ring, notify); |
| 1009 | if (notify) |
| 1010 | hypervisor_notify_via_evtchn(sc->sc_evtchn); |
| 1011 | err: |
| 1012 | return error; |
| 1013 | } |
| 1014 | |
| 1015 | static int |
| 1016 | xbd_map_align(struct xbd_req *req) |
| 1017 | { |
| 1018 | int s = splvm(); |
| 1019 | int rc; |
| 1020 | |
| 1021 | rc = uvm_km_kmem_alloc(kmem_va_arena, |
| 1022 | req->req_bp->b_bcount, (VM_NOSLEEP | VM_INSTANTFIT), |
| 1023 | (vmem_addr_t *)&req->req_data); |
| 1024 | splx(s); |
| 1025 | if (__predict_false(rc != 0)) |
| 1026 | return ENOMEM; |
| 1027 | if ((req->req_bp->b_flags & B_READ) == 0) |
| 1028 | memcpy(req->req_data, req->req_bp->b_data, |
| 1029 | req->req_bp->b_bcount); |
| 1030 | return 0; |
| 1031 | } |
| 1032 | |
| 1033 | static void |
| 1034 | xbd_unmap_align(struct xbd_req *req) |
| 1035 | { |
| 1036 | int s; |
| 1037 | if (req->req_bp->b_flags & B_READ) |
| 1038 | memcpy(req->req_bp->b_data, req->req_data, |
| 1039 | req->req_bp->b_bcount); |
| 1040 | s = splvm(); |
| 1041 | uvm_km_kmem_free(kmem_va_arena, (vaddr_t)req->req_data, req->req_bp->b_bcount); |
| 1042 | splx(s); |
| 1043 | } |
| 1044 | |