1 /*P:050 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
5 * 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 * information and a "virtqueue" mechanism to send and receive data. :*/
8 #include <linux/init.h>
9 #include <linux/bootmem.h>
10 #include <linux/lguest_launcher.h>
11 #include <linux/virtio.h>
12 #include <linux/virtio_config.h>
13 #include <linux/interrupt.h>
14 #include <linux/virtio_ring.h>
15 #include <linux/err.h>
17 #include <asm/paravirt.h>
18 #include <asm/lguest_hcall.h>
20 /* The pointer to our (page) of device descriptions. */
21 static void *lguest_devices;
23 /* Unique numbering for lguest devices. */
24 static unsigned int dev_index;
26 /* For Guests, device memory can be used as normal memory, so we cast away the
27 * __iomem to quieten sparse. */
28 static inline void *lguest_map(unsigned long phys_addr, unsigned long pages)
30 return (__force void *)ioremap(phys_addr, PAGE_SIZE*pages);
33 static inline void lguest_unmap(void *addr)
35 iounmap((__force void __iomem *)addr);
38 /*D:100 Each lguest device is just a virtio device plus a pointer to its entry
39 * in the lguest_devices page. */
40 struct lguest_device {
41 struct virtio_device vdev;
43 /* The entry in the lguest_devices page for this device. */
44 struct lguest_device_desc *desc;
47 /* Since the virtio infrastructure hands us a pointer to the virtio_device all
48 * the time, it helps to have a curt macro to get a pointer to the struct
49 * lguest_device it's enclosed in. */
50 #define to_lgdev(vdev) container_of(vdev, struct lguest_device, vdev)
53 * Device configurations
55 * The configuration information for a device consists of one or more
56 * virtqueues, a feature bitmaks, and some configuration bytes. The
57 * configuration bytes don't really matter to us: the Launcher set them up, and
58 * the driver will look at them during setup.
60 * A convenient routine to return the device's virtqueue config array:
61 * immediately after the descriptor. */
62 static struct lguest_vqconfig *lg_vq(const struct lguest_device_desc *desc)
64 return (void *)(desc + 1);
67 /* The features come immediately after the virtqueues. */
68 static u8 *lg_features(const struct lguest_device_desc *desc)
70 return (void *)(lg_vq(desc) + desc->num_vq);
73 /* The config space comes after the two feature bitmasks. */
74 static u8 *lg_config(const struct lguest_device_desc *desc)
76 return lg_features(desc) + desc->feature_len * 2;
79 /* The total size of the config page used by this device (incl. desc) */
80 static unsigned desc_size(const struct lguest_device_desc *desc)
83 + desc->num_vq * sizeof(struct lguest_vqconfig)
84 + desc->feature_len * 2
88 /* This tests (and acknowleges) a feature bit. */
89 static bool lg_feature(struct virtio_device *vdev, unsigned fbit)
91 struct lguest_device_desc *desc = to_lgdev(vdev)->desc;
94 /* Obviously if they ask for a feature off the end of our feature
95 * bitmap, it's not set. */
96 if (fbit / 8 > desc->feature_len)
99 /* The feature bitmap comes after the virtqueues. */
100 features = lg_features(desc);
101 if (!(features[fbit / 8] & (1 << (fbit % 8))))
104 /* We set the matching bit in the other half of the bitmap to tell the
105 * Host we want to use this feature. We don't use this yet, but we
106 * could in future. */
107 features[desc->feature_len + fbit / 8] |= (1 << (fbit % 8));
111 /* Once they've found a field, getting a copy of it is easy. */
112 static void lg_get(struct virtio_device *vdev, unsigned int offset,
113 void *buf, unsigned len)
115 struct lguest_device_desc *desc = to_lgdev(vdev)->desc;
117 /* Check they didn't ask for more than the length of the config! */
118 BUG_ON(offset + len > desc->config_len);
119 memcpy(buf, lg_config(desc) + offset, len);
122 /* Setting the contents is also trivial. */
123 static void lg_set(struct virtio_device *vdev, unsigned int offset,
124 const void *buf, unsigned len)
126 struct lguest_device_desc *desc = to_lgdev(vdev)->desc;
128 /* Check they didn't ask for more than the length of the config! */
129 BUG_ON(offset + len > desc->config_len);
130 memcpy(lg_config(desc) + offset, buf, len);
133 /* The operations to get and set the status word just access the status field
134 * of the device descriptor. */
135 static u8 lg_get_status(struct virtio_device *vdev)
137 return to_lgdev(vdev)->desc->status;
140 static void lg_set_status(struct virtio_device *vdev, u8 status)
142 to_lgdev(vdev)->desc->status = status;
148 * The other piece of infrastructure virtio needs is a "virtqueue": a way of
149 * the Guest device registering buffers for the other side to read from or
150 * write into (ie. send and receive buffers). Each device can have multiple
151 * virtqueues: for example the console driver uses one queue for sending and
152 * another for receiving.
154 * Fortunately for us, a very fast shared-memory-plus-descriptors virtqueue
155 * already exists in virtio_ring.c. We just need to connect it up.
157 * We start with the information we need to keep about each virtqueue.
160 /*D:140 This is the information we remember about each virtqueue. */
161 struct lguest_vq_info
163 /* A copy of the information contained in the device config. */
164 struct lguest_vqconfig config;
166 /* The address where we mapped the virtio ring, so we can unmap it. */
170 /* When the virtio_ring code wants to prod the Host, it calls us here and we
171 * make a hypercall. We hand the page number of the virtqueue so the Host
172 * knows which virtqueue we're talking about. */
173 static void lg_notify(struct virtqueue *vq)
175 /* We store our virtqueue information in the "priv" pointer of the
176 * virtqueue structure. */
177 struct lguest_vq_info *lvq = vq->priv;
179 hcall(LHCALL_NOTIFY, lvq->config.pfn << PAGE_SHIFT, 0, 0);
182 /* This routine finds the first virtqueue described in the configuration of
183 * this device and sets it up.
185 * This is kind of an ugly duckling. It'd be nicer to have a standard
186 * representation of a virtqueue in the configuration space, but it seems that
187 * everyone wants to do it differently. The KVM coders want the Guest to
188 * allocate its own pages and tell the Host where they are, but for lguest it's
189 * simpler for the Host to simply tell us where the pages are.
191 * So we provide devices with a "find virtqueue and set it up" function. */
192 static struct virtqueue *lg_find_vq(struct virtio_device *vdev,
194 void (*callback)(struct virtqueue *vq))
196 struct lguest_device *ldev = to_lgdev(vdev);
197 struct lguest_vq_info *lvq;
198 struct virtqueue *vq;
201 /* We must have this many virtqueues. */
202 if (index >= ldev->desc->num_vq)
203 return ERR_PTR(-ENOENT);
205 lvq = kmalloc(sizeof(*lvq), GFP_KERNEL);
207 return ERR_PTR(-ENOMEM);
209 /* Make a copy of the "struct lguest_vqconfig" entry, which sits after
210 * the descriptor. We need a copy because the config space might not
211 * be aligned correctly. */
212 memcpy(&lvq->config, lg_vq(ldev->desc)+index, sizeof(lvq->config));
214 printk("Mapping virtqueue %i addr %lx\n", index,
215 (unsigned long)lvq->config.pfn << PAGE_SHIFT);
216 /* Figure out how many pages the ring will take, and map that memory */
217 lvq->pages = lguest_map((unsigned long)lvq->config.pfn << PAGE_SHIFT,
218 DIV_ROUND_UP(vring_size(lvq->config.num,
226 /* OK, tell virtio_ring.c to set up a virtqueue now we know its size
227 * and we've got a pointer to its pages. */
228 vq = vring_new_virtqueue(lvq->config.num, vdev, lvq->pages,
229 lg_notify, callback);
235 /* Tell the interrupt for this virtqueue to go to the virtio_ring
236 * interrupt handler. */
237 /* FIXME: We used to have a flag for the Host to tell us we could use
238 * the interrupt as a source of randomness: it'd be nice to have that
240 err = request_irq(lvq->config.irq, vring_interrupt, IRQF_SHARED,
241 vdev->dev.bus_id, vq);
245 /* Last of all we hook up our 'struct lguest_vq_info" to the
246 * virtqueue's priv pointer. */
251 vring_del_virtqueue(vq);
253 lguest_unmap(lvq->pages);
260 /* Cleaning up a virtqueue is easy */
261 static void lg_del_vq(struct virtqueue *vq)
263 struct lguest_vq_info *lvq = vq->priv;
265 /* Release the interrupt */
266 free_irq(lvq->config.irq, vq);
267 /* Tell virtio_ring.c to free the virtqueue. */
268 vring_del_virtqueue(vq);
269 /* Unmap the pages containing the ring. */
270 lguest_unmap(lvq->pages);
271 /* Free our own queue information. */
275 /* The ops structure which hooks everything together. */
276 static struct virtio_config_ops lguest_config_ops = {
277 .feature = lg_feature,
280 .get_status = lg_get_status,
281 .set_status = lg_set_status,
282 .find_vq = lg_find_vq,
286 /* The root device for the lguest virtio devices. This makes them appear as
287 * /sys/devices/lguest/0,1,2 not /sys/devices/0,1,2. */
288 static struct device lguest_root = {
293 /*D:120 This is the core of the lguest bus: actually adding a new device.
294 * It's a separate function because it's neater that way, and because an
295 * earlier version of the code supported hotplug and unplug. They were removed
296 * early on because they were never used.
298 * As Andrew Tridgell says, "Untested code is buggy code".
300 * It's worth reading this carefully: we start with a pointer to the new device
301 * descriptor in the "lguest_devices" page. */
302 static void add_lguest_device(struct lguest_device_desc *d)
304 struct lguest_device *ldev;
306 /* Start with zeroed memory; Linux's device layer seems to count on
308 ldev = kzalloc(sizeof(*ldev), GFP_KERNEL);
310 printk(KERN_EMERG "Cannot allocate lguest dev %u\n",
315 /* This devices' parent is the lguest/ dir. */
316 ldev->vdev.dev.parent = &lguest_root;
317 /* We have a unique device index thanks to the dev_index counter. */
318 ldev->vdev.index = dev_index++;
319 /* The device type comes straight from the descriptor. There's also a
320 * device vendor field in the virtio_device struct, which we leave as
322 ldev->vdev.id.device = d->type;
323 /* We have a simple set of routines for querying the device's
324 * configuration information and setting its status. */
325 ldev->vdev.config = &lguest_config_ops;
326 /* And we remember the device's descriptor for lguest_config_ops. */
329 /* register_virtio_device() sets up the generic fields for the struct
330 * virtio_device and calls device_register(). This makes the bus
331 * infrastructure look for a matching driver. */
332 if (register_virtio_device(&ldev->vdev) != 0) {
333 printk(KERN_ERR "Failed to register lguest device %u\n",
339 /*D:110 scan_devices() simply iterates through the device page. The type 0 is
340 * reserved to mean "end of devices". */
341 static void scan_devices(void)
344 struct lguest_device_desc *d;
346 /* We start at the page beginning, and skip over each entry. */
347 for (i = 0; i < PAGE_SIZE; i += desc_size(d)) {
348 d = lguest_devices + i;
350 /* Once we hit a zero, stop. */
354 printk("Device at %i has size %u\n", i, desc_size(d));
355 add_lguest_device(d);
359 /*D:105 Fairly early in boot, lguest_devices_init() is called to set up the
360 * lguest device infrastructure. We check that we are a Guest by checking
361 * pv_info.name: there are other ways of checking, but this seems most
364 * So we can access the "struct lguest_device_desc"s easily, we map that memory
365 * and store the pointer in the global "lguest_devices". Then we register a
366 * root device from which all our devices will hang (this seems to be the
367 * correct sysfs incantation).
369 * Finally we call scan_devices() which adds all the devices found in the
370 * lguest_devices page. */
371 static int __init lguest_devices_init(void)
373 if (strcmp(pv_info.name, "lguest") != 0)
376 if (device_register(&lguest_root) != 0)
377 panic("Could not register lguest root");
379 /* Devices are in a single page above top of "normal" mem */
380 lguest_devices = lguest_map(max_pfn<<PAGE_SHIFT, 1);
385 /* We do this after core stuff, but before the drivers. */
386 postcore_initcall(lguest_devices_init);
388 /*D:150 At this point in the journey we used to now wade through the lguest
389 * devices themselves: net, block and console. Since they're all now virtio
390 * devices rather than lguest-specific, I've decided to ignore them. Mostly,
391 * they're kind of boring. But this does mean you'll never experience the
392 * thrill of reading the forbidden love scene buried deep in the block driver.
394 * "make Launcher" beckons, where we answer questions like "Where do Guests
395 * come from?", and "What do you do when someone asks for optimization?". */