diff options
Diffstat (limited to 'drivers/lguest/lguest_device.c')
-rw-r--r-- | drivers/lguest/lguest_device.c | 160 |
1 files changed, 102 insertions, 58 deletions
diff --git a/drivers/lguest/lguest_device.c b/drivers/lguest/lguest_device.c index e082cdac88b4..b6200bc39b58 100644 --- a/drivers/lguest/lguest_device.c +++ b/drivers/lguest/lguest_device.c | |||
@@ -1,10 +1,12 @@ | |||
1 | /*P:050 Lguest guests use a very simple method to describe devices. It's a | 1 | /*P:050 |
2 | * Lguest guests use a very simple method to describe devices. It's a | ||
2 | * series of device descriptors contained just above the top of normal Guest | 3 | * series of device descriptors contained just above the top of normal Guest |
3 | * memory. | 4 | * memory. |
4 | * | 5 | * |
5 | * We use the standard "virtio" device infrastructure, which provides us with a | 6 | * We use the standard "virtio" device infrastructure, which provides us with a |
6 | * console, a network and a block driver. Each one expects some configuration | 7 | * console, a network and a block driver. Each one expects some configuration |
7 | * information and a "virtqueue" or two to send and receive data. :*/ | 8 | * information and a "virtqueue" or two to send and receive data. |
9 | :*/ | ||
8 | #include <linux/init.h> | 10 | #include <linux/init.h> |
9 | #include <linux/bootmem.h> | 11 | #include <linux/bootmem.h> |
10 | #include <linux/lguest_launcher.h> | 12 | #include <linux/lguest_launcher.h> |
@@ -20,8 +22,10 @@ | |||
20 | /* The pointer to our (page) of device descriptions. */ | 22 | /* The pointer to our (page) of device descriptions. */ |
21 | static void *lguest_devices; | 23 | static void *lguest_devices; |
22 | 24 | ||
23 | /* For Guests, device memory can be used as normal memory, so we cast away the | 25 | /* |
24 | * __iomem to quieten sparse. */ | 26 | * For Guests, device memory can be used as normal memory, so we cast away the |
27 | * __iomem to quieten sparse. | ||
28 | */ | ||
25 | static inline void *lguest_map(unsigned long phys_addr, unsigned long pages) | 29 | static inline void *lguest_map(unsigned long phys_addr, unsigned long pages) |
26 | { | 30 | { |
27 | return (__force void *)ioremap_cache(phys_addr, PAGE_SIZE*pages); | 31 | return (__force void *)ioremap_cache(phys_addr, PAGE_SIZE*pages); |
@@ -32,8 +36,10 @@ static inline void lguest_unmap(void *addr) | |||
32 | iounmap((__force void __iomem *)addr); | 36 | iounmap((__force void __iomem *)addr); |
33 | } | 37 | } |
34 | 38 | ||
35 | /*D:100 Each lguest device is just a virtio device plus a pointer to its entry | 39 | /*D:100 |
36 | * in the lguest_devices page. */ | 40 | * Each lguest device is just a virtio device plus a pointer to its entry |
41 | * in the lguest_devices page. | ||
42 | */ | ||
37 | struct lguest_device { | 43 | struct lguest_device { |
38 | struct virtio_device vdev; | 44 | struct virtio_device vdev; |
39 | 45 | ||
@@ -41,9 +47,11 @@ struct lguest_device { | |||
41 | struct lguest_device_desc *desc; | 47 | struct lguest_device_desc *desc; |
42 | }; | 48 | }; |
43 | 49 | ||
44 | /* Since the virtio infrastructure hands us a pointer to the virtio_device all | 50 | /* |
51 | * Since the virtio infrastructure hands us a pointer to the virtio_device all | ||
45 | * the time, it helps to have a curt macro to get a pointer to the struct | 52 | * the time, it helps to have a curt macro to get a pointer to the struct |
46 | * lguest_device it's enclosed in. */ | 53 | * lguest_device it's enclosed in. |
54 | */ | ||
47 | #define to_lgdev(vd) container_of(vd, struct lguest_device, vdev) | 55 | #define to_lgdev(vd) container_of(vd, struct lguest_device, vdev) |
48 | 56 | ||
49 | /*D:130 | 57 | /*D:130 |
@@ -55,7 +63,8 @@ struct lguest_device { | |||
55 | * the driver will look at them during setup. | 63 | * the driver will look at them during setup. |
56 | * | 64 | * |
57 | * A convenient routine to return the device's virtqueue config array: | 65 | * A convenient routine to return the device's virtqueue config array: |
58 | * immediately after the descriptor. */ | 66 | * immediately after the descriptor. |
67 | */ | ||
59 | static struct lguest_vqconfig *lg_vq(const struct lguest_device_desc *desc) | 68 | static struct lguest_vqconfig *lg_vq(const struct lguest_device_desc *desc) |
60 | { | 69 | { |
61 | return (void *)(desc + 1); | 70 | return (void *)(desc + 1); |
@@ -98,10 +107,12 @@ static u32 lg_get_features(struct virtio_device *vdev) | |||
98 | return features; | 107 | return features; |
99 | } | 108 | } |
100 | 109 | ||
101 | /* The virtio core takes the features the Host offers, and copies the | 110 | /* |
102 | * ones supported by the driver into the vdev->features array. Once | 111 | * The virtio core takes the features the Host offers, and copies the ones |
103 | * that's all sorted out, this routine is called so we can tell the | 112 | * supported by the driver into the vdev->features array. Once that's all |
104 | * Host which features we understand and accept. */ | 113 | * sorted out, this routine is called so we can tell the Host which features we |
114 | * understand and accept. | ||
115 | */ | ||
105 | static void lg_finalize_features(struct virtio_device *vdev) | 116 | static void lg_finalize_features(struct virtio_device *vdev) |
106 | { | 117 | { |
107 | unsigned int i, bits; | 118 | unsigned int i, bits; |
@@ -112,10 +123,11 @@ static void lg_finalize_features(struct virtio_device *vdev) | |||
112 | /* Give virtio_ring a chance to accept features. */ | 123 | /* Give virtio_ring a chance to accept features. */ |
113 | vring_transport_features(vdev); | 124 | vring_transport_features(vdev); |
114 | 125 | ||
115 | /* The vdev->feature array is a Linux bitmask: this isn't the | 126 | /* |
116 | * same as a the simple array of bits used by lguest devices | 127 | * The vdev->feature array is a Linux bitmask: this isn't the same as a |
117 | * for features. So we do this slow, manual conversion which is | 128 | * the simple array of bits used by lguest devices for features. So we |
118 | * completely general. */ | 129 | * do this slow, manual conversion which is completely general. |
130 | */ | ||
119 | memset(out_features, 0, desc->feature_len); | 131 | memset(out_features, 0, desc->feature_len); |
120 | bits = min_t(unsigned, desc->feature_len, sizeof(vdev->features)) * 8; | 132 | bits = min_t(unsigned, desc->feature_len, sizeof(vdev->features)) * 8; |
121 | for (i = 0; i < bits; i++) { | 133 | for (i = 0; i < bits; i++) { |
@@ -146,15 +158,19 @@ static void lg_set(struct virtio_device *vdev, unsigned int offset, | |||
146 | memcpy(lg_config(desc) + offset, buf, len); | 158 | memcpy(lg_config(desc) + offset, buf, len); |
147 | } | 159 | } |
148 | 160 | ||
149 | /* The operations to get and set the status word just access the status field | 161 | /* |
150 | * of the device descriptor. */ | 162 | * The operations to get and set the status word just access the status field |
163 | * of the device descriptor. | ||
164 | */ | ||
151 | static u8 lg_get_status(struct virtio_device *vdev) | 165 | static u8 lg_get_status(struct virtio_device *vdev) |
152 | { | 166 | { |
153 | return to_lgdev(vdev)->desc->status; | 167 | return to_lgdev(vdev)->desc->status; |
154 | } | 168 | } |
155 | 169 | ||
156 | /* To notify on status updates, we (ab)use the NOTIFY hypercall, with the | 170 | /* |
157 | * descriptor address of the device. A zero status means "reset". */ | 171 | * To notify on status updates, we (ab)use the NOTIFY hypercall, with the |
172 | * descriptor address of the device. A zero status means "reset". | ||
173 | */ | ||
158 | static void set_status(struct virtio_device *vdev, u8 status) | 174 | static void set_status(struct virtio_device *vdev, u8 status) |
159 | { | 175 | { |
160 | unsigned long offset = (void *)to_lgdev(vdev)->desc - lguest_devices; | 176 | unsigned long offset = (void *)to_lgdev(vdev)->desc - lguest_devices; |
@@ -191,8 +207,7 @@ static void lg_reset(struct virtio_device *vdev) | |||
191 | */ | 207 | */ |
192 | 208 | ||
193 | /*D:140 This is the information we remember about each virtqueue. */ | 209 | /*D:140 This is the information we remember about each virtqueue. */ |
194 | struct lguest_vq_info | 210 | struct lguest_vq_info { |
195 | { | ||
196 | /* A copy of the information contained in the device config. */ | 211 | /* A copy of the information contained in the device config. */ |
197 | struct lguest_vqconfig config; | 212 | struct lguest_vqconfig config; |
198 | 213 | ||
@@ -200,13 +215,17 @@ struct lguest_vq_info | |||
200 | void *pages; | 215 | void *pages; |
201 | }; | 216 | }; |
202 | 217 | ||
203 | /* When the virtio_ring code wants to prod the Host, it calls us here and we | 218 | /* |
219 | * When the virtio_ring code wants to prod the Host, it calls us here and we | ||
204 | * make a hypercall. We hand the physical address of the virtqueue so the Host | 220 | * make a hypercall. We hand the physical address of the virtqueue so the Host |
205 | * knows which virtqueue we're talking about. */ | 221 | * knows which virtqueue we're talking about. |
222 | */ | ||
206 | static void lg_notify(struct virtqueue *vq) | 223 | static void lg_notify(struct virtqueue *vq) |
207 | { | 224 | { |
208 | /* We store our virtqueue information in the "priv" pointer of the | 225 | /* |
209 | * virtqueue structure. */ | 226 | * We store our virtqueue information in the "priv" pointer of the |
227 | * virtqueue structure. | ||
228 | */ | ||
210 | struct lguest_vq_info *lvq = vq->priv; | 229 | struct lguest_vq_info *lvq = vq->priv; |
211 | 230 | ||
212 | kvm_hypercall1(LHCALL_NOTIFY, lvq->config.pfn << PAGE_SHIFT); | 231 | kvm_hypercall1(LHCALL_NOTIFY, lvq->config.pfn << PAGE_SHIFT); |
@@ -215,7 +234,8 @@ static void lg_notify(struct virtqueue *vq) | |||
215 | /* An extern declaration inside a C file is bad form. Don't do it. */ | 234 | /* An extern declaration inside a C file is bad form. Don't do it. */ |
216 | extern void lguest_setup_irq(unsigned int irq); | 235 | extern void lguest_setup_irq(unsigned int irq); |
217 | 236 | ||
218 | /* This routine finds the first virtqueue described in the configuration of | 237 | /* |
238 | * This routine finds the Nth virtqueue described in the configuration of | ||
219 | * this device and sets it up. | 239 | * this device and sets it up. |
220 | * | 240 | * |
221 | * This is kind of an ugly duckling. It'd be nicer to have a standard | 241 | * This is kind of an ugly duckling. It'd be nicer to have a standard |
@@ -223,9 +243,7 @@ extern void lguest_setup_irq(unsigned int irq); | |||
223 | * everyone wants to do it differently. The KVM coders want the Guest to | 243 | * everyone wants to do it differently. The KVM coders want the Guest to |
224 | * allocate its own pages and tell the Host where they are, but for lguest it's | 244 | * allocate its own pages and tell the Host where they are, but for lguest it's |
225 | * simpler for the Host to simply tell us where the pages are. | 245 | * simpler for the Host to simply tell us where the pages are. |
226 | * | 246 | */ |
227 | * So we provide drivers with a "find the Nth virtqueue and set it up" | ||
228 | * function. */ | ||
229 | static struct virtqueue *lg_find_vq(struct virtio_device *vdev, | 247 | static struct virtqueue *lg_find_vq(struct virtio_device *vdev, |
230 | unsigned index, | 248 | unsigned index, |
231 | void (*callback)(struct virtqueue *vq), | 249 | void (*callback)(struct virtqueue *vq), |
@@ -244,9 +262,11 @@ static struct virtqueue *lg_find_vq(struct virtio_device *vdev, | |||
244 | if (!lvq) | 262 | if (!lvq) |
245 | return ERR_PTR(-ENOMEM); | 263 | return ERR_PTR(-ENOMEM); |
246 | 264 | ||
247 | /* Make a copy of the "struct lguest_vqconfig" entry, which sits after | 265 | /* |
266 | * Make a copy of the "struct lguest_vqconfig" entry, which sits after | ||
248 | * the descriptor. We need a copy because the config space might not | 267 | * the descriptor. We need a copy because the config space might not |
249 | * be aligned correctly. */ | 268 | * be aligned correctly. |
269 | */ | ||
250 | memcpy(&lvq->config, lg_vq(ldev->desc)+index, sizeof(lvq->config)); | 270 | memcpy(&lvq->config, lg_vq(ldev->desc)+index, sizeof(lvq->config)); |
251 | 271 | ||
252 | printk("Mapping virtqueue %i addr %lx\n", index, | 272 | printk("Mapping virtqueue %i addr %lx\n", index, |
@@ -261,8 +281,10 @@ static struct virtqueue *lg_find_vq(struct virtio_device *vdev, | |||
261 | goto free_lvq; | 281 | goto free_lvq; |
262 | } | 282 | } |
263 | 283 | ||
264 | /* OK, tell virtio_ring.c to set up a virtqueue now we know its size | 284 | /* |
265 | * and we've got a pointer to its pages. */ | 285 | * OK, tell virtio_ring.c to set up a virtqueue now we know its size |
286 | * and we've got a pointer to its pages. | ||
287 | */ | ||
266 | vq = vring_new_virtqueue(lvq->config.num, LGUEST_VRING_ALIGN, | 288 | vq = vring_new_virtqueue(lvq->config.num, LGUEST_VRING_ALIGN, |
267 | vdev, lvq->pages, lg_notify, callback, name); | 289 | vdev, lvq->pages, lg_notify, callback, name); |
268 | if (!vq) { | 290 | if (!vq) { |
@@ -273,18 +295,23 @@ static struct virtqueue *lg_find_vq(struct virtio_device *vdev, | |||
273 | /* Make sure the interrupt is allocated. */ | 295 | /* Make sure the interrupt is allocated. */ |
274 | lguest_setup_irq(lvq->config.irq); | 296 | lguest_setup_irq(lvq->config.irq); |
275 | 297 | ||
276 | /* Tell the interrupt for this virtqueue to go to the virtio_ring | 298 | /* |
277 | * interrupt handler. */ | 299 | * Tell the interrupt for this virtqueue to go to the virtio_ring |
278 | /* FIXME: We used to have a flag for the Host to tell us we could use | 300 | * interrupt handler. |
301 | * | ||
302 | * FIXME: We used to have a flag for the Host to tell us we could use | ||
279 | * the interrupt as a source of randomness: it'd be nice to have that | 303 | * the interrupt as a source of randomness: it'd be nice to have that |
280 | * back.. */ | 304 | * back. |
305 | */ | ||
281 | err = request_irq(lvq->config.irq, vring_interrupt, IRQF_SHARED, | 306 | err = request_irq(lvq->config.irq, vring_interrupt, IRQF_SHARED, |
282 | dev_name(&vdev->dev), vq); | 307 | dev_name(&vdev->dev), vq); |
283 | if (err) | 308 | if (err) |
284 | goto destroy_vring; | 309 | goto destroy_vring; |
285 | 310 | ||
286 | /* Last of all we hook up our 'struct lguest_vq_info" to the | 311 | /* |
287 | * virtqueue's priv pointer. */ | 312 | * Last of all we hook up our 'struct lguest_vq_info" to the |
313 | * virtqueue's priv pointer. | ||
314 | */ | ||
288 | vq->priv = lvq; | 315 | vq->priv = lvq; |
289 | return vq; | 316 | return vq; |
290 | 317 | ||
@@ -358,11 +385,14 @@ static struct virtio_config_ops lguest_config_ops = { | |||
358 | .del_vqs = lg_del_vqs, | 385 | .del_vqs = lg_del_vqs, |
359 | }; | 386 | }; |
360 | 387 | ||
361 | /* The root device for the lguest virtio devices. This makes them appear as | 388 | /* |
362 | * /sys/devices/lguest/0,1,2 not /sys/devices/0,1,2. */ | 389 | * The root device for the lguest virtio devices. This makes them appear as |
390 | * /sys/devices/lguest/0,1,2 not /sys/devices/0,1,2. | ||
391 | */ | ||
363 | static struct device *lguest_root; | 392 | static struct device *lguest_root; |
364 | 393 | ||
365 | /*D:120 This is the core of the lguest bus: actually adding a new device. | 394 | /*D:120 |
395 | * This is the core of the lguest bus: actually adding a new device. | ||
366 | * It's a separate function because it's neater that way, and because an | 396 | * It's a separate function because it's neater that way, and because an |
367 | * earlier version of the code supported hotplug and unplug. They were removed | 397 | * earlier version of the code supported hotplug and unplug. They were removed |
368 | * early on because they were never used. | 398 | * early on because they were never used. |
@@ -371,14 +401,14 @@ static struct device *lguest_root; | |||
371 | * | 401 | * |
372 | * It's worth reading this carefully: we start with a pointer to the new device | 402 | * It's worth reading this carefully: we start with a pointer to the new device |
373 | * descriptor in the "lguest_devices" page, and the offset into the device | 403 | * descriptor in the "lguest_devices" page, and the offset into the device |
374 | * descriptor page so we can uniquely identify it if things go badly wrong. */ | 404 | * descriptor page so we can uniquely identify it if things go badly wrong. |
405 | */ | ||
375 | static void add_lguest_device(struct lguest_device_desc *d, | 406 | static void add_lguest_device(struct lguest_device_desc *d, |
376 | unsigned int offset) | 407 | unsigned int offset) |
377 | { | 408 | { |
378 | struct lguest_device *ldev; | 409 | struct lguest_device *ldev; |
379 | 410 | ||
380 | /* Start with zeroed memory; Linux's device layer seems to count on | 411 | /* Start with zeroed memory; Linux's device layer counts on it. */ |
381 | * it. */ | ||
382 | ldev = kzalloc(sizeof(*ldev), GFP_KERNEL); | 412 | ldev = kzalloc(sizeof(*ldev), GFP_KERNEL); |
383 | if (!ldev) { | 413 | if (!ldev) { |
384 | printk(KERN_EMERG "Cannot allocate lguest dev %u type %u\n", | 414 | printk(KERN_EMERG "Cannot allocate lguest dev %u type %u\n", |
@@ -388,17 +418,25 @@ static void add_lguest_device(struct lguest_device_desc *d, | |||
388 | 418 | ||
389 | /* This devices' parent is the lguest/ dir. */ | 419 | /* This devices' parent is the lguest/ dir. */ |
390 | ldev->vdev.dev.parent = lguest_root; | 420 | ldev->vdev.dev.parent = lguest_root; |
391 | /* We have a unique device index thanks to the dev_index counter. */ | 421 | /* |
422 | * The device type comes straight from the descriptor. There's also a | ||
423 | * device vendor field in the virtio_device struct, which we leave as | ||
424 | * 0. | ||
425 | */ | ||
392 | ldev->vdev.id.device = d->type; | 426 | ldev->vdev.id.device = d->type; |
393 | /* We have a simple set of routines for querying the device's | 427 | /* |
394 | * configuration information and setting its status. */ | 428 | * We have a simple set of routines for querying the device's |
429 | * configuration information and setting its status. | ||
430 | */ | ||
395 | ldev->vdev.config = &lguest_config_ops; | 431 | ldev->vdev.config = &lguest_config_ops; |
396 | /* And we remember the device's descriptor for lguest_config_ops. */ | 432 | /* And we remember the device's descriptor for lguest_config_ops. */ |
397 | ldev->desc = d; | 433 | ldev->desc = d; |
398 | 434 | ||
399 | /* register_virtio_device() sets up the generic fields for the struct | 435 | /* |
436 | * register_virtio_device() sets up the generic fields for the struct | ||
400 | * virtio_device and calls device_register(). This makes the bus | 437 | * virtio_device and calls device_register(). This makes the bus |
401 | * infrastructure look for a matching driver. */ | 438 | * infrastructure look for a matching driver. |
439 | */ | ||
402 | if (register_virtio_device(&ldev->vdev) != 0) { | 440 | if (register_virtio_device(&ldev->vdev) != 0) { |
403 | printk(KERN_ERR "Failed to register lguest dev %u type %u\n", | 441 | printk(KERN_ERR "Failed to register lguest dev %u type %u\n", |
404 | offset, d->type); | 442 | offset, d->type); |
@@ -406,8 +444,10 @@ static void add_lguest_device(struct lguest_device_desc *d, | |||
406 | } | 444 | } |
407 | } | 445 | } |
408 | 446 | ||
409 | /*D:110 scan_devices() simply iterates through the device page. The type 0 is | 447 | /*D:110 |
410 | * reserved to mean "end of devices". */ | 448 | * scan_devices() simply iterates through the device page. The type 0 is |
449 | * reserved to mean "end of devices". | ||
450 | */ | ||
411 | static void scan_devices(void) | 451 | static void scan_devices(void) |
412 | { | 452 | { |
413 | unsigned int i; | 453 | unsigned int i; |
@@ -426,7 +466,8 @@ static void scan_devices(void) | |||
426 | } | 466 | } |
427 | } | 467 | } |
428 | 468 | ||
429 | /*D:105 Fairly early in boot, lguest_devices_init() is called to set up the | 469 | /*D:105 |
470 | * Fairly early in boot, lguest_devices_init() is called to set up the | ||
430 | * lguest device infrastructure. We check that we are a Guest by checking | 471 | * lguest device infrastructure. We check that we are a Guest by checking |
431 | * pv_info.name: there are other ways of checking, but this seems most | 472 | * pv_info.name: there are other ways of checking, but this seems most |
432 | * obvious to me. | 473 | * obvious to me. |
@@ -437,7 +478,8 @@ static void scan_devices(void) | |||
437 | * correct sysfs incantation). | 478 | * correct sysfs incantation). |
438 | * | 479 | * |
439 | * Finally we call scan_devices() which adds all the devices found in the | 480 | * Finally we call scan_devices() which adds all the devices found in the |
440 | * lguest_devices page. */ | 481 | * lguest_devices page. |
482 | */ | ||
441 | static int __init lguest_devices_init(void) | 483 | static int __init lguest_devices_init(void) |
442 | { | 484 | { |
443 | if (strcmp(pv_info.name, "lguest") != 0) | 485 | if (strcmp(pv_info.name, "lguest") != 0) |
@@ -456,11 +498,13 @@ static int __init lguest_devices_init(void) | |||
456 | /* We do this after core stuff, but before the drivers. */ | 498 | /* We do this after core stuff, but before the drivers. */ |
457 | postcore_initcall(lguest_devices_init); | 499 | postcore_initcall(lguest_devices_init); |
458 | 500 | ||
459 | /*D:150 At this point in the journey we used to now wade through the lguest | 501 | /*D:150 |
502 | * At this point in the journey we used to now wade through the lguest | ||
460 | * devices themselves: net, block and console. Since they're all now virtio | 503 | * devices themselves: net, block and console. Since they're all now virtio |
461 | * devices rather than lguest-specific, I've decided to ignore them. Mostly, | 504 | * devices rather than lguest-specific, I've decided to ignore them. Mostly, |
462 | * they're kind of boring. But this does mean you'll never experience the | 505 | * they're kind of boring. But this does mean you'll never experience the |
463 | * thrill of reading the forbidden love scene buried deep in the block driver. | 506 | * thrill of reading the forbidden love scene buried deep in the block driver. |
464 | * | 507 | * |
465 | * "make Launcher" beckons, where we answer questions like "Where do Guests | 508 | * "make Launcher" beckons, where we answer questions like "Where do Guests |
466 | * come from?", and "What do you do when someone asks for optimization?". */ | 509 | * come from?", and "What do you do when someone asks for optimization?". |
510 | */ | ||