]>
Commit | Line | Data |
---|---|---|
3285cf4f AP |
1 | /* |
2 | * Copyright (C) 2010 Citrix Ltd. | |
3 | * | |
4 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
5 | * the COPYING file in the top-level directory. | |
6 | * | |
6b620ca3 PB |
7 | * Contributions after 2012-01-13 are licensed under the terms of the |
8 | * GNU GPL, version 2 or (at your option) any later version. | |
3285cf4f AP |
9 | */ |
10 | ||
9ce94e7c AS |
11 | #include <sys/mman.h> |
12 | ||
a2cb15b0 | 13 | #include "hw/pci/pci.h" |
0d09e41a PB |
14 | #include "hw/i386/pc.h" |
15 | #include "hw/xen/xen_common.h" | |
16 | #include "hw/xen/xen_backend.h" | |
39f42439 | 17 | #include "qmp-commands.h" |
3285cf4f | 18 | |
dccfcd0e | 19 | #include "sysemu/char.h" |
1de7afc9 | 20 | #include "qemu/range.h" |
9c17d615 | 21 | #include "sysemu/xen-mapcache.h" |
432d268c | 22 | #include "trace.h" |
022c62cb | 23 | #include "exec/address-spaces.h" |
432d268c | 24 | |
9ce94e7c AS |
25 | #include <xen/hvm/ioreq.h> |
26 | #include <xen/hvm/params.h> | |
8a369e20 | 27 | #include <xen/hvm/e820.h> |
9ce94e7c | 28 | |
04b0de0e | 29 | //#define DEBUG_XEN_HVM |
9ce94e7c | 30 | |
04b0de0e | 31 | #ifdef DEBUG_XEN_HVM |
9ce94e7c AS |
32 | #define DPRINTF(fmt, ...) \ |
33 | do { fprintf(stderr, "xen: " fmt, ## __VA_ARGS__); } while (0) | |
34 | #else | |
35 | #define DPRINTF(fmt, ...) \ | |
36 | do { } while (0) | |
37 | #endif | |
38 | ||
ce76b8a8 | 39 | static MemoryRegion ram_memory, ram_640k, ram_lo, ram_hi; |
c65adf9b | 40 | static MemoryRegion *framebuffer; |
39f42439 | 41 | static bool xen_in_migration; |
ce76b8a8 | 42 | |
9ce94e7c | 43 | /* Compatibility with older version */ |
37f9e258 DS |
44 | |
45 | /* This allows QEMU to build on a system that has Xen 4.5 or earlier | |
46 | * installed. This here (not in hw/xen/xen_common.h) because xen/hvm/ioreq.h | |
47 | * needs to be included before this block and hw/xen/xen_common.h needs to | |
48 | * be included before xen/hvm/ioreq.h | |
49 | */ | |
50 | #ifndef IOREQ_TYPE_VMWARE_PORT | |
51 | #define IOREQ_TYPE_VMWARE_PORT 3 | |
52 | struct vmware_regs { | |
53 | uint32_t esi; | |
54 | uint32_t edi; | |
55 | uint32_t ebx; | |
56 | uint32_t ecx; | |
57 | uint32_t edx; | |
58 | }; | |
59 | typedef struct vmware_regs vmware_regs_t; | |
60 | ||
61 | struct shared_vmport_iopage { | |
62 | struct vmware_regs vcpu_vmport_regs[1]; | |
63 | }; | |
64 | typedef struct shared_vmport_iopage shared_vmport_iopage_t; | |
65 | #endif | |
66 | ||
9ce94e7c AS |
67 | #if __XEN_LATEST_INTERFACE_VERSION__ < 0x0003020a |
68 | static inline uint32_t xen_vcpu_eport(shared_iopage_t *shared_page, int i) | |
69 | { | |
70 | return shared_page->vcpu_iodata[i].vp_eport; | |
71 | } | |
72 | static inline ioreq_t *xen_vcpu_ioreq(shared_iopage_t *shared_page, int vcpu) | |
73 | { | |
74 | return &shared_page->vcpu_iodata[vcpu].vp_ioreq; | |
75 | } | |
76 | # define FMT_ioreq_size PRIx64 | |
77 | #else | |
78 | static inline uint32_t xen_vcpu_eport(shared_iopage_t *shared_page, int i) | |
79 | { | |
80 | return shared_page->vcpu_ioreq[i].vp_eport; | |
81 | } | |
82 | static inline ioreq_t *xen_vcpu_ioreq(shared_iopage_t *shared_page, int vcpu) | |
83 | { | |
84 | return &shared_page->vcpu_ioreq[vcpu]; | |
85 | } | |
86 | # define FMT_ioreq_size "u" | |
87 | #endif | |
88 | ||
89 | #define BUFFER_IO_MAX_DELAY 100 | |
90 | ||
b4dd7802 | 91 | typedef struct XenPhysmap { |
a8170e5e | 92 | hwaddr start_addr; |
b4dd7802 | 93 | ram_addr_t size; |
dc6c4fe8 | 94 | const char *name; |
a8170e5e | 95 | hwaddr phys_offset; |
b4dd7802 AP |
96 | |
97 | QLIST_ENTRY(XenPhysmap) list; | |
98 | } XenPhysmap; | |
99 | ||
9ce94e7c | 100 | typedef struct XenIOState { |
3996e85c | 101 | ioservid_t ioservid; |
9ce94e7c | 102 | shared_iopage_t *shared_page; |
37f9e258 | 103 | shared_vmport_iopage_t *shared_vmport_page; |
9ce94e7c AS |
104 | buffered_iopage_t *buffered_io_page; |
105 | QEMUTimer *buffered_io_timer; | |
37f9e258 | 106 | CPUState **cpu_by_vcpu_id; |
9ce94e7c AS |
107 | /* the evtchn port for polling the notification, */ |
108 | evtchn_port_t *ioreq_local_port; | |
fda1f768 SS |
109 | /* evtchn local port for buffered io */ |
110 | evtchn_port_t bufioreq_local_port; | |
9ce94e7c AS |
111 | /* the evtchn fd for polling */ |
112 | XenEvtchn xce_handle; | |
113 | /* which vcpu we are serving */ | |
114 | int send_vcpu; | |
115 | ||
29321335 | 116 | struct xs_handle *xenstore; |
20581d20 | 117 | MemoryListener memory_listener; |
3996e85c PD |
118 | MemoryListener io_listener; |
119 | DeviceListener device_listener; | |
b4dd7802 | 120 | QLIST_HEAD(, XenPhysmap) physmap; |
a8170e5e | 121 | hwaddr free_phys_offset; |
b4dd7802 | 122 | const XenPhysmap *log_for_dirtybit; |
29321335 | 123 | |
9ce94e7c | 124 | Notifier exit; |
da98c8eb | 125 | Notifier suspend; |
11addd0a | 126 | Notifier wakeup; |
9ce94e7c AS |
127 | } XenIOState; |
128 | ||
41445300 AP |
129 | /* Xen specific function for piix pci */ |
130 | ||
131 | int xen_pci_slot_get_pirq(PCIDevice *pci_dev, int irq_num) | |
132 | { | |
133 | return irq_num + ((pci_dev->devfn >> 3) << 2); | |
134 | } | |
135 | ||
136 | void xen_piix3_set_irq(void *opaque, int irq_num, int level) | |
137 | { | |
138 | xc_hvm_set_pci_intx_level(xen_xc, xen_domid, 0, 0, irq_num >> 2, | |
139 | irq_num & 3, level); | |
140 | } | |
141 | ||
142 | void xen_piix_pci_write_config_client(uint32_t address, uint32_t val, int len) | |
143 | { | |
144 | int i; | |
145 | ||
146 | /* Scan for updates to PCI link routes (0x60-0x63). */ | |
147 | for (i = 0; i < len; i++) { | |
148 | uint8_t v = (val >> (8 * i)) & 0xff; | |
149 | if (v & 0x80) { | |
150 | v = 0; | |
151 | } | |
152 | v &= 0xf; | |
153 | if (((address + i) >= 0x60) && ((address + i) <= 0x63)) { | |
154 | xc_hvm_set_pci_link_route(xen_xc, xen_domid, address + i - 0x60, v); | |
155 | } | |
156 | } | |
157 | } | |
158 | ||
f1dbf015 WL |
159 | void xen_hvm_inject_msi(uint64_t addr, uint32_t data) |
160 | { | |
4c9f8d1b | 161 | xen_xc_hvm_inject_msi(xen_xc, xen_domid, addr, data); |
f1dbf015 WL |
162 | } |
163 | ||
da98c8eb | 164 | static void xen_suspend_notifier(Notifier *notifier, void *data) |
c9622478 | 165 | { |
da98c8eb | 166 | xc_set_hvm_param(xen_xc, xen_domid, HVM_PARAM_ACPI_S_STATE, 3); |
c9622478 AP |
167 | } |
168 | ||
9c11a8ac AP |
169 | /* Xen Interrupt Controller */ |
170 | ||
171 | static void xen_set_irq(void *opaque, int irq, int level) | |
172 | { | |
173 | xc_hvm_set_isa_irq_level(xen_xc, xen_domid, irq, level); | |
174 | } | |
175 | ||
176 | qemu_irq *xen_interrupt_controller_init(void) | |
177 | { | |
178 | return qemu_allocate_irqs(xen_set_irq, NULL, 16); | |
179 | } | |
180 | ||
432d268c JN |
181 | /* Memory Ops */ |
182 | ||
3c2a9669 DS |
183 | static void xen_ram_init(ram_addr_t *below_4g_mem_size, |
184 | ram_addr_t *above_4g_mem_size, | |
185 | ram_addr_t ram_size, MemoryRegion **ram_memory_p) | |
432d268c | 186 | { |
ce76b8a8 | 187 | MemoryRegion *sysmem = get_system_memory(); |
ce76b8a8 | 188 | ram_addr_t block_len; |
c4f5cdc5 DS |
189 | uint64_t user_lowmem = object_property_get_int(qdev_get_machine(), |
190 | PC_MACHINE_MAX_RAM_BELOW_4G, | |
191 | &error_abort); | |
432d268c | 192 | |
a9dd38db | 193 | /* Handle the machine opt max-ram-below-4g. It is basically doing |
c4f5cdc5 DS |
194 | * min(xen limit, user limit). |
195 | */ | |
196 | if (HVM_BELOW_4G_RAM_END <= user_lowmem) { | |
197 | user_lowmem = HVM_BELOW_4G_RAM_END; | |
8a369e20 | 198 | } |
432d268c | 199 | |
c4f5cdc5 DS |
200 | if (ram_size >= user_lowmem) { |
201 | *above_4g_mem_size = ram_size - user_lowmem; | |
202 | *below_4g_mem_size = user_lowmem; | |
432d268c | 203 | } else { |
3c2a9669 DS |
204 | *above_4g_mem_size = 0; |
205 | *below_4g_mem_size = ram_size; | |
432d268c | 206 | } |
c4f5cdc5 DS |
207 | if (!*above_4g_mem_size) { |
208 | block_len = ram_size; | |
209 | } else { | |
210 | /* | |
211 | * Xen does not allocate the memory continuously, it keeps a | |
212 | * hole of the size computed above or passed in. | |
213 | */ | |
214 | block_len = (1ULL << 32) + *above_4g_mem_size; | |
215 | } | |
49946538 HT |
216 | memory_region_init_ram(&ram_memory, NULL, "xen.ram", block_len, |
217 | &error_abort); | |
c4f5cdc5 DS |
218 | *ram_memory_p = &ram_memory; |
219 | vmstate_register_ram_global(&ram_memory); | |
432d268c | 220 | |
2c9b15ca | 221 | memory_region_init_alias(&ram_640k, NULL, "xen.ram.640k", |
ce76b8a8 AK |
222 | &ram_memory, 0, 0xa0000); |
223 | memory_region_add_subregion(sysmem, 0, &ram_640k); | |
8a369e20 AP |
224 | /* Skip of the VGA IO memory space, it will be registered later by the VGA |
225 | * emulated device. | |
226 | * | |
227 | * The area between 0xc0000 and 0x100000 will be used by SeaBIOS to load | |
228 | * the Options ROM, so it is registered here as RAM. | |
229 | */ | |
2c9b15ca | 230 | memory_region_init_alias(&ram_lo, NULL, "xen.ram.lo", |
3c2a9669 DS |
231 | &ram_memory, 0xc0000, |
232 | *below_4g_mem_size - 0xc0000); | |
ce76b8a8 | 233 | memory_region_add_subregion(sysmem, 0xc0000, &ram_lo); |
3c2a9669 | 234 | if (*above_4g_mem_size > 0) { |
2c9b15ca | 235 | memory_region_init_alias(&ram_hi, NULL, "xen.ram.hi", |
ce76b8a8 | 236 | &ram_memory, 0x100000000ULL, |
3c2a9669 | 237 | *above_4g_mem_size); |
ce76b8a8 | 238 | memory_region_add_subregion(sysmem, 0x100000000ULL, &ram_hi); |
432d268c | 239 | } |
432d268c JN |
240 | } |
241 | ||
fce537d4 | 242 | void xen_ram_alloc(ram_addr_t ram_addr, ram_addr_t size, MemoryRegion *mr) |
432d268c JN |
243 | { |
244 | unsigned long nr_pfn; | |
245 | xen_pfn_t *pfn_list; | |
246 | int i; | |
247 | ||
c234572d AP |
248 | if (runstate_check(RUN_STATE_INMIGRATE)) { |
249 | /* RAM already populated in Xen */ | |
250 | fprintf(stderr, "%s: do not alloc "RAM_ADDR_FMT | |
251 | " bytes of ram at "RAM_ADDR_FMT" when runstate is INMIGRATE\n", | |
252 | __func__, size, ram_addr); | |
253 | return; | |
254 | } | |
255 | ||
ce76b8a8 AK |
256 | if (mr == &ram_memory) { |
257 | return; | |
258 | } | |
259 | ||
432d268c JN |
260 | trace_xen_ram_alloc(ram_addr, size); |
261 | ||
262 | nr_pfn = size >> TARGET_PAGE_BITS; | |
7267c094 | 263 | pfn_list = g_malloc(sizeof (*pfn_list) * nr_pfn); |
432d268c JN |
264 | |
265 | for (i = 0; i < nr_pfn; i++) { | |
266 | pfn_list[i] = (ram_addr >> TARGET_PAGE_BITS) + i; | |
267 | } | |
268 | ||
269 | if (xc_domain_populate_physmap_exact(xen_xc, xen_domid, nr_pfn, 0, 0, pfn_list)) { | |
f15fbc4b | 270 | hw_error("xen: failed to populate ram at " RAM_ADDR_FMT, ram_addr); |
432d268c JN |
271 | } |
272 | ||
7267c094 | 273 | g_free(pfn_list); |
432d268c JN |
274 | } |
275 | ||
b4dd7802 | 276 | static XenPhysmap *get_physmapping(XenIOState *state, |
a8170e5e | 277 | hwaddr start_addr, ram_addr_t size) |
b4dd7802 AP |
278 | { |
279 | XenPhysmap *physmap = NULL; | |
280 | ||
281 | start_addr &= TARGET_PAGE_MASK; | |
282 | ||
283 | QLIST_FOREACH(physmap, &state->physmap, list) { | |
284 | if (range_covers_byte(physmap->start_addr, physmap->size, start_addr)) { | |
285 | return physmap; | |
286 | } | |
287 | } | |
288 | return NULL; | |
289 | } | |
290 | ||
a8170e5e | 291 | static hwaddr xen_phys_offset_to_gaddr(hwaddr start_addr, |
cd1ba7de AP |
292 | ram_addr_t size, void *opaque) |
293 | { | |
a8170e5e | 294 | hwaddr addr = start_addr & TARGET_PAGE_MASK; |
cd1ba7de AP |
295 | XenIOState *xen_io_state = opaque; |
296 | XenPhysmap *physmap = NULL; | |
297 | ||
298 | QLIST_FOREACH(physmap, &xen_io_state->physmap, list) { | |
299 | if (range_covers_byte(physmap->phys_offset, physmap->size, addr)) { | |
300 | return physmap->start_addr; | |
301 | } | |
302 | } | |
303 | ||
304 | return start_addr; | |
305 | } | |
306 | ||
b4dd7802 AP |
307 | #if CONFIG_XEN_CTRL_INTERFACE_VERSION >= 340 |
308 | static int xen_add_to_physmap(XenIOState *state, | |
a8170e5e | 309 | hwaddr start_addr, |
b4dd7802 | 310 | ram_addr_t size, |
20581d20 | 311 | MemoryRegion *mr, |
a8170e5e | 312 | hwaddr offset_within_region) |
b4dd7802 AP |
313 | { |
314 | unsigned long i = 0; | |
315 | int rc = 0; | |
316 | XenPhysmap *physmap = NULL; | |
a8170e5e AK |
317 | hwaddr pfn, start_gpfn; |
318 | hwaddr phys_offset = memory_region_get_ram_addr(mr); | |
d1814e08 | 319 | char path[80], value[17]; |
3e1f5086 | 320 | const char *mr_name; |
b4dd7802 AP |
321 | |
322 | if (get_physmapping(state, start_addr, size)) { | |
323 | return 0; | |
324 | } | |
325 | if (size <= 0) { | |
326 | return -1; | |
327 | } | |
328 | ||
ebed8505 SS |
329 | /* Xen can only handle a single dirty log region for now and we want |
330 | * the linear framebuffer to be that region. | |
331 | * Avoid tracking any regions that is not videoram and avoid tracking | |
332 | * the legacy vga region. */ | |
20581d20 AK |
333 | if (mr == framebuffer && start_addr > 0xbffff) { |
334 | goto go_physmap; | |
ebed8505 SS |
335 | } |
336 | return -1; | |
337 | ||
338 | go_physmap: | |
f1b8caf1 SE |
339 | DPRINTF("mapping vram to %"HWADDR_PRIx" - %"HWADDR_PRIx"\n", |
340 | start_addr, start_addr + size); | |
b4dd7802 AP |
341 | |
342 | pfn = phys_offset >> TARGET_PAGE_BITS; | |
343 | start_gpfn = start_addr >> TARGET_PAGE_BITS; | |
344 | for (i = 0; i < size >> TARGET_PAGE_BITS; i++) { | |
345 | unsigned long idx = pfn + i; | |
346 | xen_pfn_t gpfn = start_gpfn + i; | |
347 | ||
348 | rc = xc_domain_add_to_physmap(xen_xc, xen_domid, XENMAPSPACE_gmfn, idx, gpfn); | |
349 | if (rc) { | |
350 | DPRINTF("add_to_physmap MFN %"PRI_xen_pfn" to PFN %" | |
351 | PRI_xen_pfn" failed: %d\n", idx, gpfn, rc); | |
352 | return -rc; | |
353 | } | |
354 | } | |
355 | ||
3e1f5086 PC |
356 | mr_name = memory_region_name(mr); |
357 | ||
7267c094 | 358 | physmap = g_malloc(sizeof (XenPhysmap)); |
b4dd7802 AP |
359 | |
360 | physmap->start_addr = start_addr; | |
361 | physmap->size = size; | |
3e1f5086 | 362 | physmap->name = mr_name; |
b4dd7802 AP |
363 | physmap->phys_offset = phys_offset; |
364 | ||
365 | QLIST_INSERT_HEAD(&state->physmap, physmap, list); | |
366 | ||
367 | xc_domain_pin_memory_cacheattr(xen_xc, xen_domid, | |
368 | start_addr >> TARGET_PAGE_BITS, | |
8b6bb0ad | 369 | (start_addr + size - 1) >> TARGET_PAGE_BITS, |
b4dd7802 | 370 | XEN_DOMCTL_MEM_CACHEATTR_WB); |
d1814e08 SS |
371 | |
372 | snprintf(path, sizeof(path), | |
373 | "/local/domain/0/device-model/%d/physmap/%"PRIx64"/start_addr", | |
374 | xen_domid, (uint64_t)phys_offset); | |
375 | snprintf(value, sizeof(value), "%"PRIx64, (uint64_t)start_addr); | |
376 | if (!xs_write(state->xenstore, 0, path, value, strlen(value))) { | |
377 | return -1; | |
378 | } | |
379 | snprintf(path, sizeof(path), | |
380 | "/local/domain/0/device-model/%d/physmap/%"PRIx64"/size", | |
381 | xen_domid, (uint64_t)phys_offset); | |
382 | snprintf(value, sizeof(value), "%"PRIx64, (uint64_t)size); | |
383 | if (!xs_write(state->xenstore, 0, path, value, strlen(value))) { | |
384 | return -1; | |
385 | } | |
3e1f5086 | 386 | if (mr_name) { |
d1814e08 SS |
387 | snprintf(path, sizeof(path), |
388 | "/local/domain/0/device-model/%d/physmap/%"PRIx64"/name", | |
389 | xen_domid, (uint64_t)phys_offset); | |
3e1f5086 | 390 | if (!xs_write(state->xenstore, 0, path, mr_name, strlen(mr_name))) { |
d1814e08 SS |
391 | return -1; |
392 | } | |
393 | } | |
394 | ||
b4dd7802 AP |
395 | return 0; |
396 | } | |
397 | ||
398 | static int xen_remove_from_physmap(XenIOState *state, | |
a8170e5e | 399 | hwaddr start_addr, |
b4dd7802 AP |
400 | ram_addr_t size) |
401 | { | |
402 | unsigned long i = 0; | |
403 | int rc = 0; | |
404 | XenPhysmap *physmap = NULL; | |
a8170e5e | 405 | hwaddr phys_offset = 0; |
b4dd7802 AP |
406 | |
407 | physmap = get_physmapping(state, start_addr, size); | |
408 | if (physmap == NULL) { | |
409 | return -1; | |
410 | } | |
411 | ||
412 | phys_offset = physmap->phys_offset; | |
413 | size = physmap->size; | |
414 | ||
d18e173a WL |
415 | DPRINTF("unmapping vram to %"HWADDR_PRIx" - %"HWADDR_PRIx", at " |
416 | "%"HWADDR_PRIx"\n", start_addr, start_addr + size, phys_offset); | |
b4dd7802 AP |
417 | |
418 | size >>= TARGET_PAGE_BITS; | |
419 | start_addr >>= TARGET_PAGE_BITS; | |
420 | phys_offset >>= TARGET_PAGE_BITS; | |
421 | for (i = 0; i < size; i++) { | |
643f5932 | 422 | xen_pfn_t idx = start_addr + i; |
b4dd7802 AP |
423 | xen_pfn_t gpfn = phys_offset + i; |
424 | ||
425 | rc = xc_domain_add_to_physmap(xen_xc, xen_domid, XENMAPSPACE_gmfn, idx, gpfn); | |
426 | if (rc) { | |
427 | fprintf(stderr, "add_to_physmap MFN %"PRI_xen_pfn" to PFN %" | |
428 | PRI_xen_pfn" failed: %d\n", idx, gpfn, rc); | |
429 | return -rc; | |
430 | } | |
431 | } | |
432 | ||
433 | QLIST_REMOVE(physmap, list); | |
434 | if (state->log_for_dirtybit == physmap) { | |
435 | state->log_for_dirtybit = NULL; | |
436 | } | |
c5633d99 | 437 | g_free(physmap); |
b4dd7802 AP |
438 | |
439 | return 0; | |
440 | } | |
441 | ||
442 | #else | |
443 | static int xen_add_to_physmap(XenIOState *state, | |
a8170e5e | 444 | hwaddr start_addr, |
b4dd7802 | 445 | ram_addr_t size, |
20581d20 | 446 | MemoryRegion *mr, |
a8170e5e | 447 | hwaddr offset_within_region) |
b4dd7802 AP |
448 | { |
449 | return -ENOSYS; | |
450 | } | |
451 | ||
452 | static int xen_remove_from_physmap(XenIOState *state, | |
a8170e5e | 453 | hwaddr start_addr, |
b4dd7802 AP |
454 | ram_addr_t size) |
455 | { | |
456 | return -ENOSYS; | |
457 | } | |
458 | #endif | |
459 | ||
20581d20 AK |
460 | static void xen_set_memory(struct MemoryListener *listener, |
461 | MemoryRegionSection *section, | |
462 | bool add) | |
b4dd7802 | 463 | { |
20581d20 | 464 | XenIOState *state = container_of(listener, XenIOState, memory_listener); |
a8170e5e | 465 | hwaddr start_addr = section->offset_within_address_space; |
052e87b0 | 466 | ram_addr_t size = int128_get64(section->size); |
20581d20 | 467 | bool log_dirty = memory_region_is_logging(section->mr); |
b4dd7802 AP |
468 | hvmmem_type_t mem_type; |
469 | ||
3996e85c PD |
470 | if (section->mr == &ram_memory) { |
471 | return; | |
472 | } else { | |
473 | if (add) { | |
474 | xen_map_memory_section(xen_xc, xen_domid, state->ioservid, | |
475 | section); | |
476 | } else { | |
477 | xen_unmap_memory_section(xen_xc, xen_domid, state->ioservid, | |
478 | section); | |
479 | } | |
480 | } | |
481 | ||
20581d20 | 482 | if (!memory_region_is_ram(section->mr)) { |
b4dd7802 AP |
483 | return; |
484 | } | |
485 | ||
3996e85c | 486 | if (log_dirty != add) { |
20581d20 AK |
487 | return; |
488 | } | |
489 | ||
490 | trace_xen_client_set_memory(start_addr, size, log_dirty); | |
b4dd7802 AP |
491 | |
492 | start_addr &= TARGET_PAGE_MASK; | |
493 | size = TARGET_PAGE_ALIGN(size); | |
20581d20 AK |
494 | |
495 | if (add) { | |
496 | if (!memory_region_is_rom(section->mr)) { | |
497 | xen_add_to_physmap(state, start_addr, size, | |
498 | section->mr, section->offset_within_region); | |
499 | } else { | |
500 | mem_type = HVMMEM_ram_ro; | |
501 | if (xc_hvm_set_mem_type(xen_xc, xen_domid, mem_type, | |
502 | start_addr >> TARGET_PAGE_BITS, | |
503 | size >> TARGET_PAGE_BITS)) { | |
504 | DPRINTF("xc_hvm_set_mem_type error, addr: "TARGET_FMT_plx"\n", | |
505 | start_addr); | |
506 | } | |
b4dd7802 | 507 | } |
20581d20 | 508 | } else { |
b4dd7802 AP |
509 | if (xen_remove_from_physmap(state, start_addr, size) < 0) { |
510 | DPRINTF("physmapping does not exist at "TARGET_FMT_plx"\n", start_addr); | |
511 | } | |
b4dd7802 AP |
512 | } |
513 | } | |
514 | ||
20581d20 AK |
515 | static void xen_region_add(MemoryListener *listener, |
516 | MemoryRegionSection *section) | |
517 | { | |
dfde4e6e | 518 | memory_region_ref(section->mr); |
20581d20 AK |
519 | xen_set_memory(listener, section, true); |
520 | } | |
521 | ||
522 | static void xen_region_del(MemoryListener *listener, | |
523 | MemoryRegionSection *section) | |
524 | { | |
525 | xen_set_memory(listener, section, false); | |
dfde4e6e | 526 | memory_region_unref(section->mr); |
20581d20 AK |
527 | } |
528 | ||
3996e85c PD |
529 | static void xen_io_add(MemoryListener *listener, |
530 | MemoryRegionSection *section) | |
531 | { | |
532 | XenIOState *state = container_of(listener, XenIOState, io_listener); | |
533 | ||
534 | memory_region_ref(section->mr); | |
535 | ||
536 | xen_map_io_section(xen_xc, xen_domid, state->ioservid, section); | |
537 | } | |
538 | ||
539 | static void xen_io_del(MemoryListener *listener, | |
540 | MemoryRegionSection *section) | |
541 | { | |
542 | XenIOState *state = container_of(listener, XenIOState, io_listener); | |
543 | ||
544 | xen_unmap_io_section(xen_xc, xen_domid, state->ioservid, section); | |
545 | ||
546 | memory_region_unref(section->mr); | |
547 | } | |
548 | ||
549 | static void xen_device_realize(DeviceListener *listener, | |
550 | DeviceState *dev) | |
551 | { | |
552 | XenIOState *state = container_of(listener, XenIOState, device_listener); | |
553 | ||
554 | if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_DEVICE)) { | |
555 | PCIDevice *pci_dev = PCI_DEVICE(dev); | |
556 | ||
557 | xen_map_pcidev(xen_xc, xen_domid, state->ioservid, pci_dev); | |
558 | } | |
559 | } | |
560 | ||
561 | static void xen_device_unrealize(DeviceListener *listener, | |
562 | DeviceState *dev) | |
563 | { | |
564 | XenIOState *state = container_of(listener, XenIOState, device_listener); | |
565 | ||
566 | if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_DEVICE)) { | |
567 | PCIDevice *pci_dev = PCI_DEVICE(dev); | |
568 | ||
569 | xen_unmap_pcidev(xen_xc, xen_domid, state->ioservid, pci_dev); | |
570 | } | |
571 | } | |
572 | ||
b18620cf | 573 | static void xen_sync_dirty_bitmap(XenIOState *state, |
a8170e5e | 574 | hwaddr start_addr, |
b18620cf | 575 | ram_addr_t size) |
b4dd7802 | 576 | { |
a8170e5e | 577 | hwaddr npages = size >> TARGET_PAGE_BITS; |
b4dd7802 AP |
578 | const int width = sizeof(unsigned long) * 8; |
579 | unsigned long bitmap[(npages + width - 1) / width]; | |
580 | int rc, i, j; | |
581 | const XenPhysmap *physmap = NULL; | |
582 | ||
583 | physmap = get_physmapping(state, start_addr, size); | |
584 | if (physmap == NULL) { | |
585 | /* not handled */ | |
b18620cf | 586 | return; |
b4dd7802 AP |
587 | } |
588 | ||
589 | if (state->log_for_dirtybit == NULL) { | |
590 | state->log_for_dirtybit = physmap; | |
591 | } else if (state->log_for_dirtybit != physmap) { | |
b18620cf AP |
592 | /* Only one range for dirty bitmap can be tracked. */ |
593 | return; | |
b4dd7802 | 594 | } |
b4dd7802 AP |
595 | |
596 | rc = xc_hvm_track_dirty_vram(xen_xc, xen_domid, | |
597 | start_addr >> TARGET_PAGE_BITS, npages, | |
598 | bitmap); | |
b18620cf | 599 | if (rc < 0) { |
74bc4151 RPM |
600 | #ifndef ENODATA |
601 | #define ENODATA ENOENT | |
602 | #endif | |
603 | if (errno == ENODATA) { | |
8aba7dc0 AP |
604 | memory_region_set_dirty(framebuffer, 0, size); |
605 | DPRINTF("xen: track_dirty_vram failed (0x" TARGET_FMT_plx | |
b18620cf | 606 | ", 0x" TARGET_FMT_plx "): %s\n", |
74bc4151 | 607 | start_addr, start_addr + size, strerror(errno)); |
b18620cf AP |
608 | } |
609 | return; | |
b4dd7802 AP |
610 | } |
611 | ||
612 | for (i = 0; i < ARRAY_SIZE(bitmap); i++) { | |
613 | unsigned long map = bitmap[i]; | |
614 | while (map != 0) { | |
adf9d70b | 615 | j = ctzl(map); |
b4dd7802 | 616 | map &= ~(1ul << j); |
5a97065b | 617 | memory_region_set_dirty(framebuffer, |
fd4aa979 BS |
618 | (i * width + j) * TARGET_PAGE_SIZE, |
619 | TARGET_PAGE_SIZE); | |
b4dd7802 AP |
620 | }; |
621 | } | |
b4dd7802 AP |
622 | } |
623 | ||
20581d20 AK |
624 | static void xen_log_start(MemoryListener *listener, |
625 | MemoryRegionSection *section) | |
b4dd7802 | 626 | { |
20581d20 | 627 | XenIOState *state = container_of(listener, XenIOState, memory_listener); |
b4dd7802 | 628 | |
b18620cf | 629 | xen_sync_dirty_bitmap(state, section->offset_within_address_space, |
052e87b0 | 630 | int128_get64(section->size)); |
b4dd7802 AP |
631 | } |
632 | ||
20581d20 | 633 | static void xen_log_stop(MemoryListener *listener, MemoryRegionSection *section) |
b4dd7802 | 634 | { |
20581d20 | 635 | XenIOState *state = container_of(listener, XenIOState, memory_listener); |
b4dd7802 AP |
636 | |
637 | state->log_for_dirtybit = NULL; | |
638 | /* Disable dirty bit tracking */ | |
b18620cf | 639 | xc_hvm_track_dirty_vram(xen_xc, xen_domid, 0, 0, NULL); |
b4dd7802 AP |
640 | } |
641 | ||
20581d20 | 642 | static void xen_log_sync(MemoryListener *listener, MemoryRegionSection *section) |
b4dd7802 | 643 | { |
20581d20 | 644 | XenIOState *state = container_of(listener, XenIOState, memory_listener); |
b4dd7802 | 645 | |
b18620cf | 646 | xen_sync_dirty_bitmap(state, section->offset_within_address_space, |
052e87b0 | 647 | int128_get64(section->size)); |
b4dd7802 AP |
648 | } |
649 | ||
20581d20 AK |
650 | static void xen_log_global_start(MemoryListener *listener) |
651 | { | |
39f42439 AP |
652 | if (xen_enabled()) { |
653 | xen_in_migration = true; | |
654 | } | |
20581d20 AK |
655 | } |
656 | ||
657 | static void xen_log_global_stop(MemoryListener *listener) | |
b4dd7802 | 658 | { |
39f42439 | 659 | xen_in_migration = false; |
b4dd7802 AP |
660 | } |
661 | ||
20581d20 AK |
662 | static MemoryListener xen_memory_listener = { |
663 | .region_add = xen_region_add, | |
664 | .region_del = xen_region_del, | |
b4dd7802 AP |
665 | .log_start = xen_log_start, |
666 | .log_stop = xen_log_stop, | |
20581d20 AK |
667 | .log_sync = xen_log_sync, |
668 | .log_global_start = xen_log_global_start, | |
669 | .log_global_stop = xen_log_global_stop, | |
72e22d2f | 670 | .priority = 10, |
b4dd7802 | 671 | }; |
432d268c | 672 | |
3996e85c PD |
673 | static MemoryListener xen_io_listener = { |
674 | .region_add = xen_io_add, | |
675 | .region_del = xen_io_del, | |
676 | .priority = 10, | |
677 | }; | |
678 | ||
679 | static DeviceListener xen_device_listener = { | |
680 | .realize = xen_device_realize, | |
681 | .unrealize = xen_device_unrealize, | |
682 | }; | |
683 | ||
9ce94e7c AS |
684 | /* get the ioreq packets from share mem */ |
685 | static ioreq_t *cpu_get_ioreq_from_shared_memory(XenIOState *state, int vcpu) | |
686 | { | |
687 | ioreq_t *req = xen_vcpu_ioreq(state->shared_page, vcpu); | |
688 | ||
689 | if (req->state != STATE_IOREQ_READY) { | |
690 | DPRINTF("I/O request not ready: " | |
691 | "%x, ptr: %x, port: %"PRIx64", " | |
692 | "data: %"PRIx64", count: %" FMT_ioreq_size ", size: %" FMT_ioreq_size "\n", | |
693 | req->state, req->data_is_ptr, req->addr, | |
694 | req->data, req->count, req->size); | |
695 | return NULL; | |
696 | } | |
697 | ||
698 | xen_rmb(); /* see IOREQ_READY /then/ read contents of ioreq */ | |
699 | ||
700 | req->state = STATE_IOREQ_INPROCESS; | |
701 | return req; | |
702 | } | |
703 | ||
704 | /* use poll to get the port notification */ | |
705 | /* ioreq_vec--out,the */ | |
706 | /* retval--the number of ioreq packet */ | |
707 | static ioreq_t *cpu_get_ioreq(XenIOState *state) | |
708 | { | |
709 | int i; | |
710 | evtchn_port_t port; | |
711 | ||
712 | port = xc_evtchn_pending(state->xce_handle); | |
fda1f768 | 713 | if (port == state->bufioreq_local_port) { |
bc72ad67 AB |
714 | timer_mod(state->buffered_io_timer, |
715 | BUFFER_IO_MAX_DELAY + qemu_clock_get_ms(QEMU_CLOCK_REALTIME)); | |
fda1f768 SS |
716 | return NULL; |
717 | } | |
718 | ||
9ce94e7c | 719 | if (port != -1) { |
1cd25a88 | 720 | for (i = 0; i < max_cpus; i++) { |
9ce94e7c AS |
721 | if (state->ioreq_local_port[i] == port) { |
722 | break; | |
723 | } | |
724 | } | |
725 | ||
1cd25a88 | 726 | if (i == max_cpus) { |
9ce94e7c AS |
727 | hw_error("Fatal error while trying to get io event!\n"); |
728 | } | |
729 | ||
730 | /* unmask the wanted port again */ | |
731 | xc_evtchn_unmask(state->xce_handle, port); | |
732 | ||
733 | /* get the io packet from shared memory */ | |
734 | state->send_vcpu = i; | |
735 | return cpu_get_ioreq_from_shared_memory(state, i); | |
736 | } | |
737 | ||
738 | /* read error or read nothing */ | |
739 | return NULL; | |
740 | } | |
741 | ||
742 | static uint32_t do_inp(pio_addr_t addr, unsigned long size) | |
743 | { | |
744 | switch (size) { | |
745 | case 1: | |
746 | return cpu_inb(addr); | |
747 | case 2: | |
748 | return cpu_inw(addr); | |
749 | case 4: | |
750 | return cpu_inl(addr); | |
751 | default: | |
752 | hw_error("inp: bad size: %04"FMT_pioaddr" %lx", addr, size); | |
753 | } | |
754 | } | |
755 | ||
756 | static void do_outp(pio_addr_t addr, | |
757 | unsigned long size, uint32_t val) | |
758 | { | |
759 | switch (size) { | |
760 | case 1: | |
761 | return cpu_outb(addr, val); | |
762 | case 2: | |
763 | return cpu_outw(addr, val); | |
764 | case 4: | |
765 | return cpu_outl(addr, val); | |
766 | default: | |
767 | hw_error("outp: bad size: %04"FMT_pioaddr" %lx", addr, size); | |
768 | } | |
769 | } | |
770 | ||
a3864829 IJ |
771 | /* |
772 | * Helper functions which read/write an object from/to physical guest | |
773 | * memory, as part of the implementation of an ioreq. | |
774 | * | |
775 | * Equivalent to | |
776 | * cpu_physical_memory_rw(addr + (req->df ? -1 : +1) * req->size * i, | |
777 | * val, req->size, 0/1) | |
778 | * except without the integer overflow problems. | |
779 | */ | |
780 | static void rw_phys_req_item(hwaddr addr, | |
781 | ioreq_t *req, uint32_t i, void *val, int rw) | |
782 | { | |
783 | /* Do everything unsigned so overflow just results in a truncated result | |
784 | * and accesses to undesired parts of guest memory, which is up | |
785 | * to the guest */ | |
786 | hwaddr offset = (hwaddr)req->size * i; | |
787 | if (req->df) { | |
788 | addr -= offset; | |
789 | } else { | |
790 | addr += offset; | |
791 | } | |
792 | cpu_physical_memory_rw(addr, val, req->size, rw); | |
793 | } | |
794 | ||
795 | static inline void read_phys_req_item(hwaddr addr, | |
796 | ioreq_t *req, uint32_t i, void *val) | |
9ce94e7c | 797 | { |
a3864829 IJ |
798 | rw_phys_req_item(addr, req, i, val, 0); |
799 | } | |
800 | static inline void write_phys_req_item(hwaddr addr, | |
801 | ioreq_t *req, uint32_t i, void *val) | |
802 | { | |
803 | rw_phys_req_item(addr, req, i, val, 1); | |
804 | } | |
9ce94e7c | 805 | |
a3864829 IJ |
806 | |
807 | static void cpu_ioreq_pio(ioreq_t *req) | |
808 | { | |
249e7e0f | 809 | uint32_t i; |
9ce94e7c AS |
810 | |
811 | if (req->dir == IOREQ_READ) { | |
812 | if (!req->data_is_ptr) { | |
813 | req->data = do_inp(req->addr, req->size); | |
814 | } else { | |
815 | uint32_t tmp; | |
816 | ||
817 | for (i = 0; i < req->count; i++) { | |
818 | tmp = do_inp(req->addr, req->size); | |
a3864829 | 819 | write_phys_req_item(req->data, req, i, &tmp); |
9ce94e7c AS |
820 | } |
821 | } | |
822 | } else if (req->dir == IOREQ_WRITE) { | |
823 | if (!req->data_is_ptr) { | |
824 | do_outp(req->addr, req->size, req->data); | |
825 | } else { | |
826 | for (i = 0; i < req->count; i++) { | |
827 | uint32_t tmp = 0; | |
828 | ||
a3864829 | 829 | read_phys_req_item(req->data, req, i, &tmp); |
9ce94e7c AS |
830 | do_outp(req->addr, req->size, tmp); |
831 | } | |
832 | } | |
833 | } | |
834 | } | |
835 | ||
836 | static void cpu_ioreq_move(ioreq_t *req) | |
837 | { | |
249e7e0f | 838 | uint32_t i; |
9ce94e7c AS |
839 | |
840 | if (!req->data_is_ptr) { | |
841 | if (req->dir == IOREQ_READ) { | |
842 | for (i = 0; i < req->count; i++) { | |
a3864829 | 843 | read_phys_req_item(req->addr, req, i, &req->data); |
9ce94e7c AS |
844 | } |
845 | } else if (req->dir == IOREQ_WRITE) { | |
846 | for (i = 0; i < req->count; i++) { | |
a3864829 | 847 | write_phys_req_item(req->addr, req, i, &req->data); |
9ce94e7c AS |
848 | } |
849 | } | |
850 | } else { | |
2b734340 | 851 | uint64_t tmp; |
9ce94e7c AS |
852 | |
853 | if (req->dir == IOREQ_READ) { | |
854 | for (i = 0; i < req->count; i++) { | |
a3864829 IJ |
855 | read_phys_req_item(req->addr, req, i, &tmp); |
856 | write_phys_req_item(req->data, req, i, &tmp); | |
9ce94e7c AS |
857 | } |
858 | } else if (req->dir == IOREQ_WRITE) { | |
859 | for (i = 0; i < req->count; i++) { | |
a3864829 IJ |
860 | read_phys_req_item(req->data, req, i, &tmp); |
861 | write_phys_req_item(req->addr, req, i, &tmp); | |
9ce94e7c AS |
862 | } |
863 | } | |
864 | } | |
865 | } | |
866 | ||
37f9e258 DS |
867 | static void regs_to_cpu(vmware_regs_t *vmport_regs, ioreq_t *req) |
868 | { | |
869 | X86CPU *cpu; | |
870 | CPUX86State *env; | |
871 | ||
872 | cpu = X86_CPU(current_cpu); | |
873 | env = &cpu->env; | |
874 | env->regs[R_EAX] = req->data; | |
875 | env->regs[R_EBX] = vmport_regs->ebx; | |
876 | env->regs[R_ECX] = vmport_regs->ecx; | |
877 | env->regs[R_EDX] = vmport_regs->edx; | |
878 | env->regs[R_ESI] = vmport_regs->esi; | |
879 | env->regs[R_EDI] = vmport_regs->edi; | |
880 | } | |
881 | ||
882 | static void regs_from_cpu(vmware_regs_t *vmport_regs) | |
883 | { | |
884 | X86CPU *cpu = X86_CPU(current_cpu); | |
885 | CPUX86State *env = &cpu->env; | |
886 | ||
887 | vmport_regs->ebx = env->regs[R_EBX]; | |
888 | vmport_regs->ecx = env->regs[R_ECX]; | |
889 | vmport_regs->edx = env->regs[R_EDX]; | |
890 | vmport_regs->esi = env->regs[R_ESI]; | |
891 | vmport_regs->edi = env->regs[R_EDI]; | |
892 | } | |
893 | ||
894 | static void handle_vmport_ioreq(XenIOState *state, ioreq_t *req) | |
895 | { | |
896 | vmware_regs_t *vmport_regs; | |
897 | ||
898 | assert(state->shared_vmport_page); | |
899 | vmport_regs = | |
900 | &state->shared_vmport_page->vcpu_vmport_regs[state->send_vcpu]; | |
901 | QEMU_BUILD_BUG_ON(sizeof(*req) < sizeof(*vmport_regs)); | |
902 | ||
903 | current_cpu = state->cpu_by_vcpu_id[state->send_vcpu]; | |
904 | regs_to_cpu(vmport_regs, req); | |
905 | cpu_ioreq_pio(req); | |
906 | regs_from_cpu(vmport_regs); | |
907 | current_cpu = NULL; | |
908 | } | |
909 | ||
910 | static void handle_ioreq(XenIOState *state, ioreq_t *req) | |
9ce94e7c AS |
911 | { |
912 | if (!req->data_is_ptr && (req->dir == IOREQ_WRITE) && | |
913 | (req->size < sizeof (target_ulong))) { | |
914 | req->data &= ((target_ulong) 1 << (8 * req->size)) - 1; | |
915 | } | |
916 | ||
917 | switch (req->type) { | |
918 | case IOREQ_TYPE_PIO: | |
919 | cpu_ioreq_pio(req); | |
920 | break; | |
921 | case IOREQ_TYPE_COPY: | |
922 | cpu_ioreq_move(req); | |
923 | break; | |
37f9e258 DS |
924 | case IOREQ_TYPE_VMWARE_PORT: |
925 | handle_vmport_ioreq(state, req); | |
926 | break; | |
9ce94e7c AS |
927 | case IOREQ_TYPE_TIMEOFFSET: |
928 | break; | |
929 | case IOREQ_TYPE_INVALIDATE: | |
e41d7c69 | 930 | xen_invalidate_map_cache(); |
9ce94e7c | 931 | break; |
3996e85c PD |
932 | case IOREQ_TYPE_PCI_CONFIG: { |
933 | uint32_t sbdf = req->addr >> 32; | |
934 | uint32_t val; | |
935 | ||
936 | /* Fake a write to port 0xCF8 so that | |
937 | * the config space access will target the | |
938 | * correct device model. | |
939 | */ | |
940 | val = (1u << 31) | | |
941 | ((req->addr & 0x0f00) << 16) | | |
942 | ((sbdf & 0xffff) << 8) | | |
943 | (req->addr & 0xfc); | |
944 | do_outp(0xcf8, 4, val); | |
945 | ||
946 | /* Now issue the config space access via | |
947 | * port 0xCFC | |
948 | */ | |
949 | req->addr = 0xcfc | (req->addr & 0x03); | |
950 | cpu_ioreq_pio(req); | |
951 | break; | |
952 | } | |
9ce94e7c AS |
953 | default: |
954 | hw_error("Invalid ioreq type 0x%x\n", req->type); | |
955 | } | |
956 | } | |
957 | ||
fda1f768 | 958 | static int handle_buffered_iopage(XenIOState *state) |
9ce94e7c AS |
959 | { |
960 | buf_ioreq_t *buf_req = NULL; | |
961 | ioreq_t req; | |
962 | int qw; | |
963 | ||
964 | if (!state->buffered_io_page) { | |
fda1f768 | 965 | return 0; |
9ce94e7c AS |
966 | } |
967 | ||
fda1f768 SS |
968 | memset(&req, 0x00, sizeof(req)); |
969 | ||
9ce94e7c AS |
970 | while (state->buffered_io_page->read_pointer != state->buffered_io_page->write_pointer) { |
971 | buf_req = &state->buffered_io_page->buf_ioreq[ | |
972 | state->buffered_io_page->read_pointer % IOREQ_BUFFER_SLOT_NUM]; | |
973 | req.size = 1UL << buf_req->size; | |
974 | req.count = 1; | |
975 | req.addr = buf_req->addr; | |
976 | req.data = buf_req->data; | |
977 | req.state = STATE_IOREQ_READY; | |
978 | req.dir = buf_req->dir; | |
979 | req.df = 1; | |
980 | req.type = buf_req->type; | |
981 | req.data_is_ptr = 0; | |
982 | qw = (req.size == 8); | |
983 | if (qw) { | |
984 | buf_req = &state->buffered_io_page->buf_ioreq[ | |
985 | (state->buffered_io_page->read_pointer + 1) % IOREQ_BUFFER_SLOT_NUM]; | |
986 | req.data |= ((uint64_t)buf_req->data) << 32; | |
987 | } | |
988 | ||
37f9e258 | 989 | handle_ioreq(state, &req); |
9ce94e7c AS |
990 | |
991 | xen_mb(); | |
992 | state->buffered_io_page->read_pointer += qw ? 2 : 1; | |
993 | } | |
fda1f768 SS |
994 | |
995 | return req.count; | |
9ce94e7c AS |
996 | } |
997 | ||
998 | static void handle_buffered_io(void *opaque) | |
999 | { | |
1000 | XenIOState *state = opaque; | |
1001 | ||
fda1f768 | 1002 | if (handle_buffered_iopage(state)) { |
bc72ad67 AB |
1003 | timer_mod(state->buffered_io_timer, |
1004 | BUFFER_IO_MAX_DELAY + qemu_clock_get_ms(QEMU_CLOCK_REALTIME)); | |
fda1f768 | 1005 | } else { |
bc72ad67 | 1006 | timer_del(state->buffered_io_timer); |
fda1f768 SS |
1007 | xc_evtchn_unmask(state->xce_handle, state->bufioreq_local_port); |
1008 | } | |
9ce94e7c AS |
1009 | } |
1010 | ||
1011 | static void cpu_handle_ioreq(void *opaque) | |
1012 | { | |
1013 | XenIOState *state = opaque; | |
1014 | ioreq_t *req = cpu_get_ioreq(state); | |
1015 | ||
1016 | handle_buffered_iopage(state); | |
1017 | if (req) { | |
37f9e258 | 1018 | handle_ioreq(state, req); |
9ce94e7c AS |
1019 | |
1020 | if (req->state != STATE_IOREQ_INPROCESS) { | |
1021 | fprintf(stderr, "Badness in I/O request ... not in service?!: " | |
1022 | "%x, ptr: %x, port: %"PRIx64", " | |
37f9e258 DS |
1023 | "data: %"PRIx64", count: %" FMT_ioreq_size |
1024 | ", size: %" FMT_ioreq_size | |
1025 | ", type: %"FMT_ioreq_size"\n", | |
9ce94e7c | 1026 | req->state, req->data_is_ptr, req->addr, |
37f9e258 | 1027 | req->data, req->count, req->size, req->type); |
180640ea | 1028 | destroy_hvm_domain(false); |
9ce94e7c AS |
1029 | return; |
1030 | } | |
1031 | ||
1032 | xen_wmb(); /* Update ioreq contents /then/ update state. */ | |
1033 | ||
1034 | /* | |
1035 | * We do this before we send the response so that the tools | |
1036 | * have the opportunity to pick up on the reset before the | |
1037 | * guest resumes and does a hlt with interrupts disabled which | |
1038 | * causes Xen to powerdown the domain. | |
1039 | */ | |
1354869c | 1040 | if (runstate_is_running()) { |
9ce94e7c | 1041 | if (qemu_shutdown_requested_get()) { |
180640ea | 1042 | destroy_hvm_domain(false); |
9ce94e7c AS |
1043 | } |
1044 | if (qemu_reset_requested_get()) { | |
e063eb1f | 1045 | qemu_system_reset(VMRESET_REPORT); |
180640ea | 1046 | destroy_hvm_domain(true); |
9ce94e7c AS |
1047 | } |
1048 | } | |
1049 | ||
1050 | req->state = STATE_IORESP_READY; | |
1051 | xc_evtchn_notify(state->xce_handle, state->ioreq_local_port[state->send_vcpu]); | |
1052 | } | |
1053 | } | |
1054 | ||
1055 | static void xen_main_loop_prepare(XenIOState *state) | |
1056 | { | |
1057 | int evtchn_fd = -1; | |
1058 | ||
1059 | if (state->xce_handle != XC_HANDLER_INITIAL_VALUE) { | |
1060 | evtchn_fd = xc_evtchn_fd(state->xce_handle); | |
1061 | } | |
1062 | ||
bc72ad67 | 1063 | state->buffered_io_timer = timer_new_ms(QEMU_CLOCK_REALTIME, handle_buffered_io, |
9ce94e7c | 1064 | state); |
9ce94e7c AS |
1065 | |
1066 | if (evtchn_fd != -1) { | |
37f9e258 DS |
1067 | CPUState *cpu_state; |
1068 | ||
1069 | DPRINTF("%s: Init cpu_by_vcpu_id\n", __func__); | |
1070 | CPU_FOREACH(cpu_state) { | |
1071 | DPRINTF("%s: cpu_by_vcpu_id[%d]=%p\n", | |
1072 | __func__, cpu_state->cpu_index, cpu_state); | |
1073 | state->cpu_by_vcpu_id[cpu_state->cpu_index] = cpu_state; | |
1074 | } | |
9ce94e7c AS |
1075 | qemu_set_fd_handler(evtchn_fd, cpu_handle_ioreq, NULL, state); |
1076 | } | |
1077 | } | |
1078 | ||
1079 | ||
1dfb4dd9 LC |
1080 | static void xen_hvm_change_state_handler(void *opaque, int running, |
1081 | RunState rstate) | |
9ce94e7c | 1082 | { |
3996e85c PD |
1083 | XenIOState *state = opaque; |
1084 | ||
9ce94e7c | 1085 | if (running) { |
3996e85c | 1086 | xen_main_loop_prepare(state); |
9ce94e7c | 1087 | } |
3996e85c PD |
1088 | |
1089 | xen_set_ioreq_server_state(xen_xc, xen_domid, | |
1090 | state->ioservid, | |
1091 | (rstate == RUN_STATE_RUNNING)); | |
9ce94e7c AS |
1092 | } |
1093 | ||
9e8dd451 | 1094 | static void xen_exit_notifier(Notifier *n, void *data) |
9ce94e7c AS |
1095 | { |
1096 | XenIOState *state = container_of(n, XenIOState, exit); | |
1097 | ||
1098 | xc_evtchn_close(state->xce_handle); | |
29321335 | 1099 | xs_daemon_close(state->xenstore); |
9ce94e7c AS |
1100 | } |
1101 | ||
d1814e08 SS |
1102 | static void xen_read_physmap(XenIOState *state) |
1103 | { | |
1104 | XenPhysmap *physmap = NULL; | |
1105 | unsigned int len, num, i; | |
1106 | char path[80], *value = NULL; | |
1107 | char **entries = NULL; | |
1108 | ||
1109 | snprintf(path, sizeof(path), | |
1110 | "/local/domain/0/device-model/%d/physmap", xen_domid); | |
1111 | entries = xs_directory(state->xenstore, 0, path, &num); | |
1112 | if (entries == NULL) | |
1113 | return; | |
1114 | ||
1115 | for (i = 0; i < num; i++) { | |
1116 | physmap = g_malloc(sizeof (XenPhysmap)); | |
1117 | physmap->phys_offset = strtoull(entries[i], NULL, 16); | |
1118 | snprintf(path, sizeof(path), | |
1119 | "/local/domain/0/device-model/%d/physmap/%s/start_addr", | |
1120 | xen_domid, entries[i]); | |
1121 | value = xs_read(state->xenstore, 0, path, &len); | |
1122 | if (value == NULL) { | |
c5633d99 | 1123 | g_free(physmap); |
d1814e08 SS |
1124 | continue; |
1125 | } | |
1126 | physmap->start_addr = strtoull(value, NULL, 16); | |
1127 | free(value); | |
1128 | ||
1129 | snprintf(path, sizeof(path), | |
1130 | "/local/domain/0/device-model/%d/physmap/%s/size", | |
1131 | xen_domid, entries[i]); | |
1132 | value = xs_read(state->xenstore, 0, path, &len); | |
1133 | if (value == NULL) { | |
c5633d99 | 1134 | g_free(physmap); |
d1814e08 SS |
1135 | continue; |
1136 | } | |
1137 | physmap->size = strtoull(value, NULL, 16); | |
1138 | free(value); | |
1139 | ||
1140 | snprintf(path, sizeof(path), | |
1141 | "/local/domain/0/device-model/%d/physmap/%s/name", | |
1142 | xen_domid, entries[i]); | |
1143 | physmap->name = xs_read(state->xenstore, 0, path, &len); | |
1144 | ||
1145 | QLIST_INSERT_HEAD(&state->physmap, physmap, list); | |
1146 | } | |
1147 | free(entries); | |
d1814e08 SS |
1148 | } |
1149 | ||
11addd0a LJ |
1150 | static void xen_wakeup_notifier(Notifier *notifier, void *data) |
1151 | { | |
1152 | xc_set_hvm_param(xen_xc, xen_domid, HVM_PARAM_ACPI_S_STATE, 0); | |
1153 | } | |
1154 | ||
6c5b0c0a | 1155 | /* return 0 means OK, or -1 means critical issue -- will exit(1) */ |
3c2a9669 DS |
1156 | int xen_hvm_init(ram_addr_t *below_4g_mem_size, ram_addr_t *above_4g_mem_size, |
1157 | MemoryRegion **ram_memory) | |
29d3ccde | 1158 | { |
9ce94e7c | 1159 | int i, rc; |
3996e85c PD |
1160 | xen_pfn_t ioreq_pfn; |
1161 | xen_pfn_t bufioreq_pfn; | |
1162 | evtchn_port_t bufioreq_evtchn; | |
9ce94e7c AS |
1163 | XenIOState *state; |
1164 | ||
7267c094 | 1165 | state = g_malloc0(sizeof (XenIOState)); |
9ce94e7c AS |
1166 | |
1167 | state->xce_handle = xen_xc_evtchn_open(NULL, 0); | |
1168 | if (state->xce_handle == XC_HANDLER_INITIAL_VALUE) { | |
1169 | perror("xen: event channel open"); | |
6c5b0c0a | 1170 | return -1; |
9ce94e7c AS |
1171 | } |
1172 | ||
29321335 AP |
1173 | state->xenstore = xs_daemon_open(); |
1174 | if (state->xenstore == NULL) { | |
1175 | perror("xen: xenstore open"); | |
6c5b0c0a | 1176 | return -1; |
29321335 AP |
1177 | } |
1178 | ||
3996e85c PD |
1179 | rc = xen_create_ioreq_server(xen_xc, xen_domid, &state->ioservid); |
1180 | if (rc < 0) { | |
1181 | perror("xen: ioreq server create"); | |
1182 | return -1; | |
1183 | } | |
1184 | ||
9ce94e7c AS |
1185 | state->exit.notify = xen_exit_notifier; |
1186 | qemu_add_exit_notifier(&state->exit); | |
1187 | ||
da98c8eb GH |
1188 | state->suspend.notify = xen_suspend_notifier; |
1189 | qemu_register_suspend_notifier(&state->suspend); | |
1190 | ||
11addd0a LJ |
1191 | state->wakeup.notify = xen_wakeup_notifier; |
1192 | qemu_register_wakeup_notifier(&state->wakeup); | |
1193 | ||
3996e85c PD |
1194 | rc = xen_get_ioreq_server_info(xen_xc, xen_domid, state->ioservid, |
1195 | &ioreq_pfn, &bufioreq_pfn, | |
1196 | &bufioreq_evtchn); | |
1197 | if (rc < 0) { | |
1198 | hw_error("failed to get ioreq server info: error %d handle=" XC_INTERFACE_FMT, | |
1199 | errno, xen_xc); | |
1200 | } | |
1201 | ||
9ce94e7c | 1202 | DPRINTF("shared page at pfn %lx\n", ioreq_pfn); |
3996e85c PD |
1203 | DPRINTF("buffered io page at pfn %lx\n", bufioreq_pfn); |
1204 | DPRINTF("buffered io evtchn is %x\n", bufioreq_evtchn); | |
1205 | ||
9ce94e7c AS |
1206 | state->shared_page = xc_map_foreign_range(xen_xc, xen_domid, XC_PAGE_SIZE, |
1207 | PROT_READ|PROT_WRITE, ioreq_pfn); | |
1208 | if (state->shared_page == NULL) { | |
1209 | hw_error("map shared IO page returned error %d handle=" XC_INTERFACE_FMT, | |
1210 | errno, xen_xc); | |
1211 | } | |
1212 | ||
37f9e258 DS |
1213 | rc = xen_get_vmport_regs_pfn(xen_xc, xen_domid, &ioreq_pfn); |
1214 | if (!rc) { | |
1215 | DPRINTF("shared vmport page at pfn %lx\n", ioreq_pfn); | |
1216 | state->shared_vmport_page = | |
1217 | xc_map_foreign_range(xen_xc, xen_domid, XC_PAGE_SIZE, | |
1218 | PROT_READ|PROT_WRITE, ioreq_pfn); | |
1219 | if (state->shared_vmport_page == NULL) { | |
1220 | hw_error("map shared vmport IO page returned error %d handle=" | |
1221 | XC_INTERFACE_FMT, errno, xen_xc); | |
1222 | } | |
1223 | } else if (rc != -ENOSYS) { | |
1224 | hw_error("get vmport regs pfn returned error %d, rc=%d", errno, rc); | |
1225 | } | |
1226 | ||
3996e85c PD |
1227 | state->buffered_io_page = xc_map_foreign_range(xen_xc, xen_domid, |
1228 | XC_PAGE_SIZE, | |
1229 | PROT_READ|PROT_WRITE, | |
1230 | bufioreq_pfn); | |
9ce94e7c AS |
1231 | if (state->buffered_io_page == NULL) { |
1232 | hw_error("map buffered IO page returned error %d", errno); | |
1233 | } | |
1234 | ||
37f9e258 DS |
1235 | /* Note: cpus is empty at this point in init */ |
1236 | state->cpu_by_vcpu_id = g_malloc0(max_cpus * sizeof(CPUState *)); | |
1237 | ||
3996e85c PD |
1238 | rc = xen_set_ioreq_server_state(xen_xc, xen_domid, state->ioservid, true); |
1239 | if (rc < 0) { | |
1240 | hw_error("failed to enable ioreq server info: error %d handle=" XC_INTERFACE_FMT, | |
1241 | errno, xen_xc); | |
1242 | } | |
1243 | ||
1cd25a88 | 1244 | state->ioreq_local_port = g_malloc0(max_cpus * sizeof (evtchn_port_t)); |
9ce94e7c AS |
1245 | |
1246 | /* FIXME: how about if we overflow the page here? */ | |
1cd25a88 | 1247 | for (i = 0; i < max_cpus; i++) { |
9ce94e7c AS |
1248 | rc = xc_evtchn_bind_interdomain(state->xce_handle, xen_domid, |
1249 | xen_vcpu_eport(state->shared_page, i)); | |
1250 | if (rc == -1) { | |
3996e85c | 1251 | fprintf(stderr, "shared evtchn %d bind error %d\n", i, errno); |
9ce94e7c AS |
1252 | return -1; |
1253 | } | |
1254 | state->ioreq_local_port[i] = rc; | |
1255 | } | |
1256 | ||
fda1f768 | 1257 | rc = xc_evtchn_bind_interdomain(state->xce_handle, xen_domid, |
3996e85c | 1258 | bufioreq_evtchn); |
fda1f768 | 1259 | if (rc == -1) { |
3996e85c | 1260 | fprintf(stderr, "buffered evtchn bind error %d\n", errno); |
fda1f768 SS |
1261 | return -1; |
1262 | } | |
1263 | state->bufioreq_local_port = rc; | |
1264 | ||
432d268c | 1265 | /* Init RAM management */ |
cd1ba7de | 1266 | xen_map_cache_init(xen_phys_offset_to_gaddr, state); |
3c2a9669 | 1267 | xen_ram_init(below_4g_mem_size, above_4g_mem_size, ram_size, ram_memory); |
432d268c | 1268 | |
fb4bb2b5 | 1269 | qemu_add_vm_change_state_handler(xen_hvm_change_state_handler, state); |
9ce94e7c | 1270 | |
20581d20 | 1271 | state->memory_listener = xen_memory_listener; |
b4dd7802 | 1272 | QLIST_INIT(&state->physmap); |
f6790af6 | 1273 | memory_listener_register(&state->memory_listener, &address_space_memory); |
b4dd7802 AP |
1274 | state->log_for_dirtybit = NULL; |
1275 | ||
3996e85c PD |
1276 | state->io_listener = xen_io_listener; |
1277 | memory_listener_register(&state->io_listener, &address_space_io); | |
1278 | ||
1279 | state->device_listener = xen_device_listener; | |
1280 | device_listener_register(&state->device_listener); | |
1281 | ||
ad35a7da SS |
1282 | /* Initialize backend core & drivers */ |
1283 | if (xen_be_init() != 0) { | |
1284 | fprintf(stderr, "%s: xen backend core setup failed\n", __FUNCTION__); | |
6c5b0c0a | 1285 | return -1; |
ad35a7da SS |
1286 | } |
1287 | xen_be_register("console", &xen_console_ops); | |
37cdfcf1 | 1288 | xen_be_register("vkbd", &xen_kbdmouse_ops); |
ad35a7da | 1289 | xen_be_register("qdisk", &xen_blkdev_ops); |
d1814e08 | 1290 | xen_read_physmap(state); |
ad35a7da | 1291 | |
29d3ccde AP |
1292 | return 0; |
1293 | } | |
9ce94e7c | 1294 | |
180640ea | 1295 | void destroy_hvm_domain(bool reboot) |
9ce94e7c AS |
1296 | { |
1297 | XenXC xc_handle; | |
1298 | int sts; | |
1299 | ||
1300 | xc_handle = xen_xc_interface_open(0, 0, 0); | |
1301 | if (xc_handle == XC_HANDLER_INITIAL_VALUE) { | |
1302 | fprintf(stderr, "Cannot acquire xenctrl handle\n"); | |
1303 | } else { | |
180640ea JB |
1304 | sts = xc_domain_shutdown(xc_handle, xen_domid, |
1305 | reboot ? SHUTDOWN_reboot : SHUTDOWN_poweroff); | |
9ce94e7c | 1306 | if (sts != 0) { |
180640ea JB |
1307 | fprintf(stderr, "xc_domain_shutdown failed to issue %s, " |
1308 | "sts %d, %s\n", reboot ? "reboot" : "poweroff", | |
1309 | sts, strerror(errno)); | |
9ce94e7c | 1310 | } else { |
180640ea JB |
1311 | fprintf(stderr, "Issued domain %d %s\n", xen_domid, |
1312 | reboot ? "reboot" : "poweroff"); | |
9ce94e7c AS |
1313 | } |
1314 | xc_interface_close(xc_handle); | |
1315 | } | |
1316 | } | |
c65adf9b AK |
1317 | |
1318 | void xen_register_framebuffer(MemoryRegion *mr) | |
1319 | { | |
1320 | framebuffer = mr; | |
1321 | } | |
eaab4d60 AK |
1322 | |
1323 | void xen_shutdown_fatal_error(const char *fmt, ...) | |
1324 | { | |
1325 | va_list ap; | |
1326 | ||
1327 | va_start(ap, fmt); | |
1328 | vfprintf(stderr, fmt, ap); | |
1329 | va_end(ap); | |
1330 | fprintf(stderr, "Will destroy the domain.\n"); | |
1331 | /* destroy the domain */ | |
1332 | qemu_system_shutdown_request(); | |
1333 | } | |
910b38e4 AP |
1334 | |
1335 | void xen_modified_memory(ram_addr_t start, ram_addr_t length) | |
1336 | { | |
1337 | if (unlikely(xen_in_migration)) { | |
1338 | int rc; | |
1339 | ram_addr_t start_pfn, nb_pages; | |
1340 | ||
1341 | if (length == 0) { | |
1342 | length = TARGET_PAGE_SIZE; | |
1343 | } | |
1344 | start_pfn = start >> TARGET_PAGE_BITS; | |
1345 | nb_pages = ((start + length + TARGET_PAGE_SIZE - 1) >> TARGET_PAGE_BITS) | |
1346 | - start_pfn; | |
1347 | rc = xc_hvm_modified_memory(xen_xc, xen_domid, start_pfn, nb_pages); | |
1348 | if (rc) { | |
1349 | fprintf(stderr, | |
1350 | "%s failed for "RAM_ADDR_FMT" ("RAM_ADDR_FMT"): %i, %s\n", | |
1351 | __func__, start, nb_pages, rc, strerror(-rc)); | |
1352 | } | |
1353 | } | |
1354 | } | |
04b0de0e WL |
1355 | |
1356 | void qmp_xen_set_global_dirty_log(bool enable, Error **errp) | |
1357 | { | |
1358 | if (enable) { | |
1359 | memory_global_dirty_log_start(); | |
1360 | } else { | |
1361 | memory_global_dirty_log_stop(); | |
1362 | } | |
1363 | } |