hw/i386/xen/xen_platform.c

   1 /*
   2  * XEN platform pci device, formerly known as the event channel device
   3  *
   4  * Copyright (c) 2003-2004 Intel Corp.
   5  * Copyright (c) 2006 XenSource
   6  *
   7  * Permission is hereby granted, free of charge, to any person obtaining a copy
   8  * of this software and associated documentation files (the "Software"), to deal
   9  * in the Software without restriction, including without limitation the rights
  10  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  11  * copies of the Software, and to permit persons to whom the Software is
  12  * furnished to do so, subject to the following conditions:
  13  *
  14  * The above copyright notice and this permission notice shall be included in
  15  * all copies or substantial portions of the Software.
  16  *
  17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  18  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  21  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  22  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  23  * THE SOFTWARE.
  24  */
  25
  26 #include "qemu/osdep.h"
  27 #include "qapi/error.h"
  28 #include "hw/ide.h"
  29 #include "hw/pci/pci.h"
  30 #include "hw/irq.h"
  31 #include "hw/xen/xen_common.h"
  32 #include "migration/vmstate.h"
  33 #include "hw/xen/xen-legacy-backend.h"
  34 #include "trace.h"
  35 #include "exec/address-spaces.h"
  36 #include "sysemu/xen.h"
  37 #include "sysemu/block-backend.h"
  38 #include "qemu/error-report.h"
  39 #include "qemu/module.h"
  40
  41 #include <xenguest.h>
  42 #include "qom/object.h"
  43
  44 //#define DEBUG_PLATFORM
  45
  46 #ifdef DEBUG_PLATFORM
  47 #define DPRINTF(fmt, ...) do { \
  48     fprintf(stderr, "xen_platform: " fmt, ## __VA_ARGS__); \
  49 } while (0)
  50 #else
  51 #define DPRINTF(fmt, ...) do { } while (0)
  52 #endif
  53
  54 #define PFFLAG_ROM_LOCK 1 /* Sets whether ROM memory area is RW or RO */
  55
  56 struct PCIXenPlatformState {
  57     /*< private >*/
  58     PCIDevice parent_obj;
  59     /*< public >*/
  60
  61     MemoryRegion fixed_io;
  62     MemoryRegion bar;
  63     MemoryRegion mmio_bar;
  64     uint8_t flags; /* used only for version_id == 2 */
  65     int drivers_blacklisted;
  66     uint16_t driver_product_version;
  67
  68     /* Log from guest drivers */
  69     char log_buffer[4096];
  70     int log_buffer_off;
  71 };
  72
  73 #define TYPE_XEN_PLATFORM "xen-platform"
  74 OBJECT_DECLARE_SIMPLE_TYPE(PCIXenPlatformState, XEN_PLATFORM)
  75
  76 #define XEN_PLATFORM_IOPORT 0x10
  77
  78 /* Send bytes to syslog */
  79 static void log_writeb(PCIXenPlatformState *s, char val)
  80 {
  81     if (val == '\n' || s->log_buffer_off == sizeof(s->log_buffer) - 1) {
  82         /* Flush buffer */
  83         s->log_buffer[s->log_buffer_off] = 0;
  84         trace_xen_platform_log(s->log_buffer);
  85         s->log_buffer_off = 0;
  86     } else {
  87         s->log_buffer[s->log_buffer_off++] = val;
  88     }
  89 }
  90
  91 /*
  92  * Unplug device flags.
  93  *
  94  * The logic got a little confused at some point in the past but this is
  95  * what they do now.
  96  *
  97  * bit 0: Unplug all IDE and SCSI disks.
  98  * bit 1: Unplug all NICs.
  99  * bit 2: Unplug IDE disks except primary master. This is overridden if
 100  *        bit 0 is also present in the mask.
 101  * bit 3: Unplug all NVMe disks.
 102  *
 103  */
 104 #define _UNPLUG_IDE_SCSI_DISKS 0
 105 #define UNPLUG_IDE_SCSI_DISKS (1u << _UNPLUG_IDE_SCSI_DISKS)
 106
 107 #define _UNPLUG_ALL_NICS 1
 108 #define UNPLUG_ALL_NICS (1u << _UNPLUG_ALL_NICS)
 109
 110 #define _UNPLUG_AUX_IDE_DISKS 2
 111 #define UNPLUG_AUX_IDE_DISKS (1u << _UNPLUG_AUX_IDE_DISKS)
 112
 113 #define _UNPLUG_NVME_DISKS 3
 114 #define UNPLUG_NVME_DISKS (1u << _UNPLUG_NVME_DISKS)
 115
 116 static void unplug_nic(PCIBus *b, PCIDevice *d, void *o)
 117 {
 118     /* We have to ignore passthrough devices */
 119     if (pci_get_word(d->config + PCI_CLASS_DEVICE) ==
 120             PCI_CLASS_NETWORK_ETHERNET
 121             && strcmp(d->name, "xen-pci-passthrough") != 0) {
 122         object_unparent(OBJECT(d));
 123     }
 124 }
 125
 126 /* Remove the peer of the NIC device. Normally, this would be a tap device. */
 127 static void del_nic_peer(NICState *nic, void *opaque)
 128 {
 129     NetClientState *nc;
 130
 131     nc = qemu_get_queue(nic);
 132     if (nc->peer)
 133         qemu_del_net_client(nc->peer);
 134 }
 135
 136 static void pci_unplug_nics(PCIBus *bus)
 137 {
 138     qemu_foreach_nic(del_nic_peer, NULL);
 139     pci_for_each_device(bus, 0, unplug_nic, NULL);
 140 }
 141
 142 static void unplug_disks(PCIBus *b, PCIDevice *d, void *opaque)
 143 {
 144     uint32_t flags = *(uint32_t *)opaque;
 145     bool aux = (flags & UNPLUG_AUX_IDE_DISKS) &&
 146         !(flags & UNPLUG_IDE_SCSI_DISKS);
 147
 148     /* We have to ignore passthrough devices */
 149     if (!strcmp(d->name, "xen-pci-passthrough")) {
 150         return;
 151     }
 152
 153     switch (pci_get_word(d->config + PCI_CLASS_DEVICE)) {
 154     case PCI_CLASS_STORAGE_IDE:
 155         pci_piix3_xen_ide_unplug(DEVICE(d), aux);
 156         break;
 157
 158     case PCI_CLASS_STORAGE_SCSI:
 159         if (!aux) {
 160             object_unparent(OBJECT(d));
 161         }
 162         break;
 163
 164     case PCI_CLASS_STORAGE_EXPRESS:
 165         if (flags & UNPLUG_NVME_DISKS) {
 166             object_unparent(OBJECT(d));
 167         }
 168
 169     default:
 170         break;
 171     }
 172 }
 173
 174 static void pci_unplug_disks(PCIBus *bus, uint32_t flags)
 175 {
 176     pci_for_each_device(bus, 0, unplug_disks, &flags);
 177 }
 178
 179 static void platform_fixed_ioport_writew(void *opaque, uint32_t addr, uint32_t val)
 180 {
 181     PCIXenPlatformState *s = opaque;
 182
 183     switch (addr) {
 184     case 0: {
 185         PCIDevice *pci_dev = PCI_DEVICE(s);
 186         /* Unplug devices. See comment above flag definitions */
 187         if (val & (UNPLUG_IDE_SCSI_DISKS | UNPLUG_AUX_IDE_DISKS |
 188                    UNPLUG_NVME_DISKS)) {
 189             DPRINTF("unplug disks\n");
 190             pci_unplug_disks(pci_get_bus(pci_dev), val);
 191         }
 192         if (val & UNPLUG_ALL_NICS) {
 193             DPRINTF("unplug nics\n");
 194             pci_unplug_nics(pci_get_bus(pci_dev));
 195         }
 196         break;
 197     }
 198     case 2:
 199         switch (val) {
 200         case 1:
 201             DPRINTF("Citrix Windows PV drivers loaded in guest\n");
 202             break;
 203         case 0:
 204             DPRINTF("Guest claimed to be running PV product 0?\n");
 205             break;
 206         default:
 207             DPRINTF("Unknown PV product %d loaded in guest\n", val);
 208             break;
 209         }
 210         s->driver_product_version = val;
 211         break;
 212     }
 213 }
 214
 215 static void platform_fixed_ioport_writel(void *opaque, uint32_t addr,
 216                                          uint32_t val)
 217 {
 218     switch (addr) {
 219     case 0:
 220         /* PV driver version */
 221         break;
 222     }
 223 }
 224
 225 static void platform_fixed_ioport_writeb(void *opaque, uint32_t addr, uint32_t val)
 226 {
 227     PCIXenPlatformState *s = opaque;
 228
 229     switch (addr) {
 230     case 0: /* Platform flags */ {
 231         hvmmem_type_t mem_type = (val & PFFLAG_ROM_LOCK) ?
 232             HVMMEM_ram_ro : HVMMEM_ram_rw;
 233         if (xen_set_mem_type(xen_domid, mem_type, 0xc0, 0x40)) {
 234             DPRINTF("unable to change ro/rw state of ROM memory area!\n");
 235         } else {
 236             s->flags = val & PFFLAG_ROM_LOCK;
 237             DPRINTF("changed ro/rw state of ROM memory area. now is %s state.\n",
 238                     (mem_type == HVMMEM_ram_ro ? "ro":"rw"));
 239         }
 240         break;
 241     }
 242     case 2:
 243         log_writeb(s, val);
 244         break;
 245     }
 246 }
 247
 248 static uint32_t platform_fixed_ioport_readw(void *opaque, uint32_t addr)
 249 {
 250     PCIXenPlatformState *s = opaque;
 251
 252     switch (addr) {
 253     case 0:
 254         if (s->drivers_blacklisted) {
 255             /* The drivers will recognise this magic number and refuse
 256              * to do anything. */
 257             return 0xd249;
 258         } else {
 259             /* Magic value so that you can identify the interface. */
 260             return 0x49d2;
 261         }
 262     default:
 263         return 0xffff;
 264     }
 265 }
 266
 267 static uint32_t platform_fixed_ioport_readb(void *opaque, uint32_t addr)
 268 {
 269     PCIXenPlatformState *s = opaque;
 270
 271     switch (addr) {
 272     case 0:
 273         /* Platform flags */
 274         return s->flags;
 275     case 2:
 276         /* Version number */
 277         return 1;
 278     default:
 279         return 0xff;
 280     }
 281 }
 282
 283 static void platform_fixed_ioport_reset(void *opaque)
 284 {
 285     PCIXenPlatformState *s = opaque;
 286
 287     platform_fixed_ioport_writeb(s, 0, 0);
 288 }
 289
 290 static uint64_t platform_fixed_ioport_read(void *opaque,
 291                                            hwaddr addr,
 292                                            unsigned size)
 293 {
 294     switch (size) {
 295     case 1:
 296         return platform_fixed_ioport_readb(opaque, addr);
 297     case 2:
 298         return platform_fixed_ioport_readw(opaque, addr);
 299     default:
 300         return -1;
 301     }
 302 }
 303
 304 static void platform_fixed_ioport_write(void *opaque, hwaddr addr,
 305
 306                                         uint64_t val, unsigned size)
 307 {
 308     switch (size) {
 309     case 1:
 310         platform_fixed_ioport_writeb(opaque, addr, val);
 311         break;
 312     case 2:
 313         platform_fixed_ioport_writew(opaque, addr, val);
 314         break;
 315     case 4:
 316         platform_fixed_ioport_writel(opaque, addr, val);
 317         break;
 318     }
 319 }
 320
 321
 322 static const MemoryRegionOps platform_fixed_io_ops = {
 323     .read = platform_fixed_ioport_read,
 324     .write = platform_fixed_ioport_write,
 325     .valid = {
 326         .unaligned = true,
 327     },
 328     .impl = {
 329         .min_access_size = 1,
 330         .max_access_size = 4,
 331         .unaligned = true,
 332     },
 333     .endianness = DEVICE_LITTLE_ENDIAN,
 334 };
 335
 336 static void platform_fixed_ioport_init(PCIXenPlatformState* s)
 337 {
 338     memory_region_init_io(&s->fixed_io, OBJECT(s), &platform_fixed_io_ops, s,
 339                           "xen-fixed", 16);
 340     memory_region_add_subregion(get_system_io(), XEN_PLATFORM_IOPORT,
 341                                 &s->fixed_io);
 342 }
 343
 344 /* Xen Platform PCI Device */
 345
 346 static uint64_t xen_platform_ioport_readb(void *opaque, hwaddr addr,
 347                                           unsigned int size)
 348 {
 349     if (addr == 0) {
 350         return platform_fixed_ioport_readb(opaque, 0);
 351     } else {
 352         return ~0u;
 353     }
 354 }
 355
 356 static void xen_platform_ioport_writeb(void *opaque, hwaddr addr,
 357                                        uint64_t val, unsigned int size)
 358 {
 359     PCIXenPlatformState *s = opaque;
 360     PCIDevice *pci_dev = PCI_DEVICE(s);
 361
 362     switch (addr) {
 363     case 0: /* Platform flags */
 364         platform_fixed_ioport_writeb(opaque, 0, (uint32_t)val);
 365         break;
 366     case 4:
 367         if (val == 1) {
 368             /*
 369              * SUSE unplug for Xenlinux
 370              * xen-kmp used this since xen-3.0.4, instead the official protocol
 371              * from xen-3.3+ It did an unconditional "outl(1, (ioaddr + 4));"
 372              * Pre VMDP 1.7 used 4 and 8 depending on how VMDP was configured.
 373              * If VMDP was to control both disk and LAN it would use 4.
 374              * If it controlled just disk or just LAN, it would use 8 below.
 375              */
 376             pci_unplug_disks(pci_get_bus(pci_dev), UNPLUG_IDE_SCSI_DISKS);
 377             pci_unplug_nics(pci_get_bus(pci_dev));
 378         }
 379         break;
 380     case 8:
 381         switch (val) {
 382         case 1:
 383             pci_unplug_disks(pci_get_bus(pci_dev), UNPLUG_IDE_SCSI_DISKS);
 384             break;
 385         case 2:
 386             pci_unplug_nics(pci_get_bus(pci_dev));
 387             break;
 388         default:
 389             log_writeb(s, (uint32_t)val);
 390             break;
 391         }
 392         break;
 393     default:
 394         break;
 395     }
 396 }
 397
 398 static const MemoryRegionOps xen_pci_io_ops = {
 399     .read  = xen_platform_ioport_readb,
 400     .write = xen_platform_ioport_writeb,
 401     .impl.min_access_size = 1,
 402     .impl.max_access_size = 1,
 403 };
 404
 405 static void platform_ioport_bar_setup(PCIXenPlatformState *d)
 406 {
 407     memory_region_init_io(&d->bar, OBJECT(d), &xen_pci_io_ops, d,
 408                           "xen-pci", 0x100);
 409 }
 410
 411 static uint64_t platform_mmio_read(void *opaque, hwaddr addr,
 412                                    unsigned size)
 413 {
 414     DPRINTF("Warning: attempted read from physical address "
 415             "0x" TARGET_FMT_plx " in xen platform mmio space\n", addr);
 416
 417     return 0;
 418 }
 419
 420 static void platform_mmio_write(void *opaque, hwaddr addr,
 421                                 uint64_t val, unsigned size)
 422 {
 423     DPRINTF("Warning: attempted write of 0x%"PRIx64" to physical "
 424             "address 0x" TARGET_FMT_plx " in xen platform mmio space\n",
 425             val, addr);
 426 }
 427
 428 static const MemoryRegionOps platform_mmio_handler = {
 429     .read = &platform_mmio_read,
 430     .write = &platform_mmio_write,
 431     .endianness = DEVICE_NATIVE_ENDIAN,
 432 };
 433
 434 static void platform_mmio_setup(PCIXenPlatformState *d)
 435 {
 436     memory_region_init_io(&d->mmio_bar, OBJECT(d), &platform_mmio_handler, d,
 437                           "xen-mmio", 0x1000000);
 438 }
 439
 440 static int xen_platform_post_load(void *opaque, int version_id)
 441 {
 442     PCIXenPlatformState *s = opaque;
 443
 444     platform_fixed_ioport_writeb(s, 0, s->flags);
 445
 446     return 0;
 447 }
 448
 449 static const VMStateDescription vmstate_xen_platform = {
 450     .name = "platform",
 451     .version_id = 4,
 452     .minimum_version_id = 4,
 453     .post_load = xen_platform_post_load,
 454     .fields = (VMStateField[]) {
 455         VMSTATE_PCI_DEVICE(parent_obj, PCIXenPlatformState),
 456         VMSTATE_UINT8(flags, PCIXenPlatformState),
 457         VMSTATE_END_OF_LIST()
 458     }
 459 };
 460
 461 static void xen_platform_realize(PCIDevice *dev, Error **errp)
 462 {
 463     PCIXenPlatformState *d = XEN_PLATFORM(dev);
 464     uint8_t *pci_conf;
 465
 466     /* Device will crash on reset if xen is not initialized */
 467     if (!xen_enabled()) {
 468         error_setg(errp, "xen-platform device requires the Xen accelerator");
 469         return;
 470     }
 471
 472     pci_conf = dev->config;
 473
 474     pci_set_word(pci_conf + PCI_COMMAND, PCI_COMMAND_IO | PCI_COMMAND_MEMORY);
 475
 476     pci_config_set_prog_interface(pci_conf, 0);
 477
 478     pci_conf[PCI_INTERRUPT_PIN] = 1;
 479
 480     platform_ioport_bar_setup(d);
 481     pci_register_bar(dev, 0, PCI_BASE_ADDRESS_SPACE_IO, &d->bar);
 482
 483     /* reserve 16MB mmio address for share memory*/
 484     platform_mmio_setup(d);
 485     pci_register_bar(dev, 1, PCI_BASE_ADDRESS_MEM_PREFETCH,
 486                      &d->mmio_bar);
 487
 488     platform_fixed_ioport_init(d);
 489 }
 490
 491 static void platform_reset(DeviceState *dev)
 492 {
 493     PCIXenPlatformState *s = XEN_PLATFORM(dev);
 494
 495     platform_fixed_ioport_reset(s);
 496 }
 497
 498 static void xen_platform_class_init(ObjectClass *klass, void *data)
 499 {
 500     DeviceClass *dc = DEVICE_CLASS(klass);
 501     PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
 502
 503     k->realize = xen_platform_realize;
 504     k->vendor_id = PCI_VENDOR_ID_XEN;
 505     k->device_id = PCI_DEVICE_ID_XEN_PLATFORM;
 506     k->class_id = PCI_CLASS_OTHERS << 8 | 0x80;
 507     k->subsystem_vendor_id = PCI_VENDOR_ID_XEN;
 508     k->subsystem_id = PCI_DEVICE_ID_XEN_PLATFORM;
 509     k->revision = 1;
 510     set_bit(DEVICE_CATEGORY_MISC, dc->categories);
 511     dc->desc = "XEN platform pci device";
 512     dc->reset = platform_reset;
 513     dc->vmsd = &vmstate_xen_platform;
 514 }
 515
 516 static const TypeInfo xen_platform_info = {
 517     .name          = TYPE_XEN_PLATFORM,
 518     .parent        = TYPE_PCI_DEVICE,
 519     .instance_size = sizeof(PCIXenPlatformState),
 520     .class_init    = xen_platform_class_init,
 521     .interfaces = (InterfaceInfo[]) {
 522         { INTERFACE_CONVENTIONAL_PCI_DEVICE },
 523         { },
 524     },
 525 };
 526
 527 static void xen_platform_register_types(void)
 528 {
 529     type_register_static(&xen_platform_info);
 530 }
 531
 532 type_init(xen_platform_register_types)