]> git.proxmox.com Git - mirror_qemu.git/blame - hw/virtio-pci.c
qemu/virtio: MSI-X support in virtio PCI
[mirror_qemu.git] / hw / virtio-pci.c
CommitLineData
53c25cea
PB
1/*
2 * Virtio PCI Bindings
3 *
4 * Copyright IBM, Corp. 2007
5 * Copyright (c) 2009 CodeSourcery
6 *
7 * Authors:
8 * Anthony Liguori <aliguori@us.ibm.com>
9 * Paul Brook <paul@codesourcery.com>
10 *
11 * This work is licensed under the terms of the GNU GPL, version 2. See
12 * the COPYING file in the top-level directory.
13 *
14 */
15
16#include <inttypes.h>
17
18#include "virtio.h"
19#include "pci.h"
1ad2134f 20//#include "sysemu.h"
aba800a3 21#include "msix.h"
53c25cea
PB
22
23/* from Linux's linux/virtio_pci.h */
24
25/* A 32-bit r/o bitmask of the features supported by the host */
26#define VIRTIO_PCI_HOST_FEATURES 0
27
28/* A 32-bit r/w bitmask of features activated by the guest */
29#define VIRTIO_PCI_GUEST_FEATURES 4
30
31/* A 32-bit r/w PFN for the currently selected queue */
32#define VIRTIO_PCI_QUEUE_PFN 8
33
34/* A 16-bit r/o queue size for the currently selected queue */
35#define VIRTIO_PCI_QUEUE_NUM 12
36
37/* A 16-bit r/w queue selector */
38#define VIRTIO_PCI_QUEUE_SEL 14
39
40/* A 16-bit r/w queue notifier */
41#define VIRTIO_PCI_QUEUE_NOTIFY 16
42
43/* An 8-bit device status register. */
44#define VIRTIO_PCI_STATUS 18
45
46/* An 8-bit r/o interrupt status register. Reading the value will return the
47 * current contents of the ISR and will also clear it. This is effectively
48 * a read-and-acknowledge. */
49#define VIRTIO_PCI_ISR 19
50
aba800a3
MT
51/* MSI-X registers: only enabled if MSI-X is enabled. */
52/* A 16-bit vector for configuration changes. */
53#define VIRTIO_MSI_CONFIG_VECTOR 20
54/* A 16-bit vector for selected queue notifications. */
55#define VIRTIO_MSI_QUEUE_VECTOR 22
56
57/* Config space size */
58#define VIRTIO_PCI_CONFIG_NOMSI 20
59#define VIRTIO_PCI_CONFIG_MSI 24
60#define VIRTIO_PCI_REGION_SIZE(dev) (msix_present(dev) ? \
61 VIRTIO_PCI_CONFIG_MSI : \
62 VIRTIO_PCI_CONFIG_NOMSI)
63
64/* The remaining space is defined by each driver as the per-driver
65 * configuration space */
66#define VIRTIO_PCI_CONFIG(dev) (msix_enabled(dev) ? \
67 VIRTIO_PCI_CONFIG_MSI : \
68 VIRTIO_PCI_CONFIG_NOMSI)
53c25cea
PB
69
70/* Virtio ABI version, if we increment this, we break the guest driver. */
71#define VIRTIO_PCI_ABI_VERSION 0
72
73/* How many bits to shift physical queue address written to QUEUE_PFN.
74 * 12 is historical, and due to x86 page size. */
75#define VIRTIO_PCI_QUEUE_ADDR_SHIFT 12
76
77/* QEMU doesn't strictly need write barriers since everything runs in
78 * lock-step. We'll leave the calls to wmb() in though to make it obvious for
79 * KVM or if kqemu gets SMP support.
80 */
81#define wmb() do { } while (0)
82
83/* PCI bindings. */
84
85typedef struct {
86 PCIDevice pci_dev;
87 VirtIODevice *vdev;
88 uint32_t addr;
89
90 uint16_t vendor;
91 uint16_t device;
92 uint16_t subvendor;
93 uint16_t class_code;
94 uint8_t pif;
95} VirtIOPCIProxy;
96
97/* virtio device */
98
7055e687 99static void virtio_pci_notify(void *opaque, uint16_t vector)
53c25cea
PB
100{
101 VirtIOPCIProxy *proxy = opaque;
aba800a3
MT
102 if (msix_enabled(&proxy->pci_dev))
103 msix_notify(&proxy->pci_dev, vector);
104 else
105 qemu_set_irq(proxy->pci_dev.irq[0], proxy->vdev->isr & 1);
53c25cea
PB
106}
107
7055e687
MT
108static void virtio_pci_reset(void *opaque)
109{
110 VirtIOPCIProxy *proxy = opaque;
111 virtio_reset(proxy->vdev);
aba800a3 112 msix_reset(&proxy->pci_dev);
7055e687
MT
113}
114
53c25cea
PB
115static void virtio_ioport_write(void *opaque, uint32_t addr, uint32_t val)
116{
117 VirtIOPCIProxy *proxy = opaque;
118 VirtIODevice *vdev = proxy->vdev;
119 target_phys_addr_t pa;
120
53c25cea
PB
121 switch (addr) {
122 case VIRTIO_PCI_GUEST_FEATURES:
123 /* Guest does not negotiate properly? We have to assume nothing. */
124 if (val & (1 << VIRTIO_F_BAD_FEATURE)) {
125 if (vdev->bad_features)
126 val = vdev->bad_features(vdev);
127 else
128 val = 0;
129 }
130 if (vdev->set_features)
131 vdev->set_features(vdev, val);
132 vdev->features = val;
133 break;
134 case VIRTIO_PCI_QUEUE_PFN:
135 pa = (target_phys_addr_t)val << VIRTIO_PCI_QUEUE_ADDR_SHIFT;
7055e687
MT
136 if (pa == 0)
137 virtio_pci_reset(proxy);
138 else
139 virtio_queue_set_addr(vdev, vdev->queue_sel, pa);
53c25cea
PB
140 break;
141 case VIRTIO_PCI_QUEUE_SEL:
142 if (val < VIRTIO_PCI_QUEUE_MAX)
143 vdev->queue_sel = val;
144 break;
145 case VIRTIO_PCI_QUEUE_NOTIFY:
146 virtio_queue_notify(vdev, val);
147 break;
148 case VIRTIO_PCI_STATUS:
149 vdev->status = val & 0xFF;
150 if (vdev->status == 0)
7055e687 151 virtio_pci_reset(proxy);
53c25cea 152 break;
aba800a3
MT
153 case VIRTIO_MSI_CONFIG_VECTOR:
154 msix_vector_unuse(&proxy->pci_dev, vdev->config_vector);
155 /* Make it possible for guest to discover an error took place. */
156 if (msix_vector_use(&proxy->pci_dev, val) < 0)
157 val = VIRTIO_NO_VECTOR;
158 vdev->config_vector = val;
159 break;
160 case VIRTIO_MSI_QUEUE_VECTOR:
161 msix_vector_unuse(&proxy->pci_dev,
162 virtio_queue_vector(vdev, vdev->queue_sel));
163 /* Make it possible for guest to discover an error took place. */
164 if (msix_vector_use(&proxy->pci_dev, val) < 0)
165 val = VIRTIO_NO_VECTOR;
166 virtio_queue_set_vector(vdev, vdev->queue_sel, val);
167 break;
168 default:
169 fprintf(stderr, "%s: unexpected address 0x%x value 0x%x\n",
170 __func__, addr, val);
171 break;
53c25cea
PB
172 }
173}
174
aba800a3 175static uint32_t virtio_ioport_read(VirtIOPCIProxy *proxy, uint32_t addr)
53c25cea 176{
53c25cea
PB
177 VirtIODevice *vdev = proxy->vdev;
178 uint32_t ret = 0xFFFFFFFF;
179
53c25cea
PB
180 switch (addr) {
181 case VIRTIO_PCI_HOST_FEATURES:
182 ret = vdev->get_features(vdev);
efeea6d0
MM
183 ret |= (1 << VIRTIO_F_NOTIFY_ON_EMPTY);
184 ret |= (1 << VIRTIO_RING_F_INDIRECT_DESC);
185 ret |= (1 << VIRTIO_F_BAD_FEATURE);
53c25cea
PB
186 break;
187 case VIRTIO_PCI_GUEST_FEATURES:
188 ret = vdev->features;
189 break;
190 case VIRTIO_PCI_QUEUE_PFN:
191 ret = virtio_queue_get_addr(vdev, vdev->queue_sel)
192 >> VIRTIO_PCI_QUEUE_ADDR_SHIFT;
193 break;
194 case VIRTIO_PCI_QUEUE_NUM:
195 ret = virtio_queue_get_num(vdev, vdev->queue_sel);
196 break;
197 case VIRTIO_PCI_QUEUE_SEL:
198 ret = vdev->queue_sel;
199 break;
200 case VIRTIO_PCI_STATUS:
201 ret = vdev->status;
202 break;
203 case VIRTIO_PCI_ISR:
204 /* reading from the ISR also clears it. */
205 ret = vdev->isr;
206 vdev->isr = 0;
7055e687 207 qemu_set_irq(proxy->pci_dev.irq[0], 0);
53c25cea 208 break;
aba800a3
MT
209 case VIRTIO_MSI_CONFIG_VECTOR:
210 ret = vdev->config_vector;
211 break;
212 case VIRTIO_MSI_QUEUE_VECTOR:
213 ret = virtio_queue_vector(vdev, vdev->queue_sel);
214 break;
53c25cea
PB
215 default:
216 break;
217 }
218
219 return ret;
220}
221
222static uint32_t virtio_pci_config_readb(void *opaque, uint32_t addr)
223{
224 VirtIOPCIProxy *proxy = opaque;
aba800a3
MT
225 uint32_t config = VIRTIO_PCI_CONFIG(&proxy->pci_dev);
226 addr -= proxy->addr;
227 if (addr < config)
228 return virtio_ioport_read(proxy, addr);
229 addr -= config;
53c25cea
PB
230 return virtio_config_readb(proxy->vdev, addr);
231}
232
233static uint32_t virtio_pci_config_readw(void *opaque, uint32_t addr)
234{
235 VirtIOPCIProxy *proxy = opaque;
aba800a3
MT
236 uint32_t config = VIRTIO_PCI_CONFIG(&proxy->pci_dev);
237 addr -= proxy->addr;
238 if (addr < config)
239 return virtio_ioport_read(proxy, addr);
240 addr -= config;
53c25cea
PB
241 return virtio_config_readw(proxy->vdev, addr);
242}
243
244static uint32_t virtio_pci_config_readl(void *opaque, uint32_t addr)
245{
246 VirtIOPCIProxy *proxy = opaque;
aba800a3
MT
247 uint32_t config = VIRTIO_PCI_CONFIG(&proxy->pci_dev);
248 addr -= proxy->addr;
249 if (addr < config)
250 return virtio_ioport_read(proxy, addr);
251 addr -= config;
53c25cea
PB
252 return virtio_config_readl(proxy->vdev, addr);
253}
254
255static void virtio_pci_config_writeb(void *opaque, uint32_t addr, uint32_t val)
256{
257 VirtIOPCIProxy *proxy = opaque;
aba800a3
MT
258 uint32_t config = VIRTIO_PCI_CONFIG(&proxy->pci_dev);
259 addr -= proxy->addr;
260 if (addr < config) {
261 virtio_ioport_write(proxy, addr, val);
262 return;
263 }
264 addr -= config;
53c25cea
PB
265 virtio_config_writeb(proxy->vdev, addr, val);
266}
267
268static void virtio_pci_config_writew(void *opaque, uint32_t addr, uint32_t val)
269{
270 VirtIOPCIProxy *proxy = opaque;
aba800a3
MT
271 uint32_t config = VIRTIO_PCI_CONFIG(&proxy->pci_dev);
272 addr -= proxy->addr;
273 if (addr < config) {
274 virtio_ioport_write(proxy, addr, val);
275 return;
276 }
277 addr -= config;
53c25cea
PB
278 virtio_config_writew(proxy->vdev, addr, val);
279}
280
281static void virtio_pci_config_writel(void *opaque, uint32_t addr, uint32_t val)
282{
283 VirtIOPCIProxy *proxy = opaque;
aba800a3
MT
284 uint32_t config = VIRTIO_PCI_CONFIG(&proxy->pci_dev);
285 addr -= proxy->addr;
286 if (addr < config) {
287 virtio_ioport_write(proxy, addr, val);
288 return;
289 }
290 addr -= config;
53c25cea
PB
291 virtio_config_writel(proxy->vdev, addr, val);
292}
293
294static void virtio_map(PCIDevice *pci_dev, int region_num,
295 uint32_t addr, uint32_t size, int type)
296{
297 VirtIOPCIProxy *proxy = container_of(pci_dev, VirtIOPCIProxy, pci_dev);
298 VirtIODevice *vdev = proxy->vdev;
aba800a3 299 unsigned config_len = VIRTIO_PCI_REGION_SIZE(pci_dev) + vdev->config_len;
53c25cea
PB
300
301 proxy->addr = addr;
53c25cea 302
aba800a3
MT
303 register_ioport_write(addr, config_len, 1, virtio_pci_config_writeb, proxy);
304 register_ioport_write(addr, config_len, 2, virtio_pci_config_writew, proxy);
305 register_ioport_write(addr, config_len, 4, virtio_pci_config_writel, proxy);
306 register_ioport_read(addr, config_len, 1, virtio_pci_config_readb, proxy);
307 register_ioport_read(addr, config_len, 2, virtio_pci_config_readw, proxy);
308 register_ioport_read(addr, config_len, 4, virtio_pci_config_readl, proxy);
53c25cea 309
aba800a3 310 if (vdev->config_len)
53c25cea 311 vdev->get_config(vdev, vdev->config);
aba800a3
MT
312}
313
314static void virtio_write_config(PCIDevice *pci_dev, uint32_t address,
315 uint32_t val, int len)
316{
317 pci_default_write_config(pci_dev, address, val, len);
318 msix_write_config(pci_dev, address, val, len);
53c25cea
PB
319}
320
321static const VirtIOBindings virtio_pci_bindings = {
7055e687 322 .notify = virtio_pci_notify
53c25cea
PB
323};
324
325static void virtio_init_pci(VirtIOPCIProxy *proxy, VirtIODevice *vdev,
326 uint16_t vendor, uint16_t device,
327 uint16_t class_code, uint8_t pif)
328{
329 uint8_t *config;
330 uint32_t size;
331
332 proxy->vdev = vdev;
333
334 config = proxy->pci_dev.config;
335 pci_config_set_vendor_id(config, vendor);
336 pci_config_set_device_id(config, device);
337
338 config[0x08] = VIRTIO_PCI_ABI_VERSION;
339
340 config[0x09] = pif;
341 pci_config_set_class(config, class_code);
342 config[PCI_HEADER_TYPE] = PCI_HEADER_TYPE_NORMAL;
343
344 config[0x2c] = vendor & 0xFF;
345 config[0x2d] = (vendor >> 8) & 0xFF;
346 config[0x2e] = vdev->device_id & 0xFF;
347 config[0x2f] = (vdev->device_id >> 8) & 0xFF;
348
349 config[0x3d] = 1;
350
aba800a3
MT
351 if (vdev->nvectors && !msix_init(&proxy->pci_dev, vdev->nvectors, 1, 0)) {
352 pci_register_bar(&proxy->pci_dev, 1,
353 msix_bar_size(&proxy->pci_dev),
354 PCI_ADDRESS_SPACE_MEM,
355 msix_mmio_map);
356 proxy->pci_dev.config_write = virtio_write_config;
357 proxy->pci_dev.unregister = msix_uninit;
358 } else
359 vdev->nvectors = 0;
360
361 size = VIRTIO_PCI_REGION_SIZE(&proxy->pci_dev) + vdev->config_len;
53c25cea
PB
362 if (size & (size-1))
363 size = 1 << qemu_fls(size);
364
28c2c264 365 pci_register_bar(&proxy->pci_dev, 0, size, PCI_ADDRESS_SPACE_IO,
53c25cea
PB
366 virtio_map);
367
7055e687
MT
368 qemu_register_reset(virtio_pci_reset, 0, proxy);
369
53c25cea
PB
370 virtio_bind_device(vdev, &virtio_pci_bindings, proxy);
371}
372
373static void virtio_blk_init_pci(PCIDevice *pci_dev)
374{
375 VirtIOPCIProxy *proxy = DO_UPCAST(VirtIOPCIProxy, pci_dev, pci_dev);
376 VirtIODevice *vdev;
377
378 vdev = virtio_blk_init(&pci_dev->qdev);
379 virtio_init_pci(proxy, vdev,
380 PCI_VENDOR_ID_REDHAT_QUMRANET,
381 PCI_DEVICE_ID_VIRTIO_BLOCK,
382 PCI_CLASS_STORAGE_OTHER,
383 0x00);
384}
385
386static void virtio_console_init_pci(PCIDevice *pci_dev)
387{
388 VirtIOPCIProxy *proxy = DO_UPCAST(VirtIOPCIProxy, pci_dev, pci_dev);
389 VirtIODevice *vdev;
390
391 vdev = virtio_console_init(&pci_dev->qdev);
392 virtio_init_pci(proxy, vdev,
393 PCI_VENDOR_ID_REDHAT_QUMRANET,
394 PCI_DEVICE_ID_VIRTIO_CONSOLE,
395 PCI_CLASS_DISPLAY_OTHER,
396 0x00);
397}
398
399static void virtio_net_init_pci(PCIDevice *pci_dev)
400{
401 VirtIOPCIProxy *proxy = DO_UPCAST(VirtIOPCIProxy, pci_dev, pci_dev);
402 VirtIODevice *vdev;
403
404 vdev = virtio_net_init(&pci_dev->qdev);
405 virtio_init_pci(proxy, vdev,
406 PCI_VENDOR_ID_REDHAT_QUMRANET,
407 PCI_DEVICE_ID_VIRTIO_NET,
408 PCI_CLASS_NETWORK_ETHERNET,
409 0x00);
410}
411
412static void virtio_balloon_init_pci(PCIDevice *pci_dev)
413{
414 VirtIOPCIProxy *proxy = DO_UPCAST(VirtIOPCIProxy, pci_dev, pci_dev);
415 VirtIODevice *vdev;
416
417 vdev = virtio_balloon_init(&pci_dev->qdev);
418 virtio_init_pci(proxy, vdev,
419 PCI_VENDOR_ID_REDHAT_QUMRANET,
420 PCI_DEVICE_ID_VIRTIO_BALLOON,
421 PCI_CLASS_MEMORY_RAM,
422 0x00);
423}
424
425static void virtio_pci_register_devices(void)
426{
427 pci_qdev_register("virtio-blk-pci", sizeof(VirtIOPCIProxy),
428 virtio_blk_init_pci);
429 pci_qdev_register("virtio-net-pci", sizeof(VirtIOPCIProxy),
430 virtio_net_init_pci);
431 pci_qdev_register("virtio-console-pci", sizeof(VirtIOPCIProxy),
432 virtio_console_init_pci);
433 pci_qdev_register("virtio-balloon-pci", sizeof(VirtIOPCIProxy),
434 virtio_balloon_init_pci);
435}
436
437device_init(virtio_pci_register_devices)