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 a series of fields.
56 * We don't really care what they are: the Launcher set them up, and the driver
57 * will look at them during setup.
59 * For us these fields come immediately after that device's descriptor in the
60 * lguest_devices page.
62 * Each field starts with a "type" byte, a "length" byte, then that number of
63 * bytes of configuration information. The device descriptor tells us the
64 * total configuration length so we know when we've reached the last field. */
66 /* type + length bytes */
69 /* This finds the first field of a given type for a device's configuration. */
70 static void *lg_find(struct virtio_device
*vdev
, u8 type
, unsigned int *len
)
72 struct lguest_device_desc
*desc
= to_lgdev(vdev
)->desc
;
75 for (i
= 0; i
< desc
->config_len
; i
+= FHDR_LEN
+ desc
->config
[i
+1]) {
76 if (desc
->config
[i
] == type
) {
77 /* Mark it used, so Host can know we looked at it, and
78 * also so we won't find the same one twice. */
79 desc
->config
[i
] |= 0x80;
80 /* Remember, the second byte is the length. */
81 *len
= desc
->config
[i
+1];
82 /* We return a pointer to the field header. */
83 return desc
->config
+ i
;
87 /* Not found: return NULL for failure. */
91 /* Once they've found a field, getting a copy of it is easy. */
92 static void lg_get(struct virtio_device
*vdev
, void *token
,
93 void *buf
, unsigned len
)
95 /* Check they didn't ask for more than the length of the field! */
96 BUG_ON(len
> ((u8
*)token
)[1]);
97 memcpy(buf
, token
+ FHDR_LEN
, len
);
100 /* Setting the contents is also trivial. */
101 static void lg_set(struct virtio_device
*vdev
, void *token
,
102 const void *buf
, unsigned len
)
104 BUG_ON(len
> ((u8
*)token
)[1]);
105 memcpy(token
+ FHDR_LEN
, buf
, len
);
108 /* The operations to get and set the status word just access the status field
109 * of the device descriptor. */
110 static u8
lg_get_status(struct virtio_device
*vdev
)
112 return to_lgdev(vdev
)->desc
->status
;
115 static void lg_set_status(struct virtio_device
*vdev
, u8 status
)
117 to_lgdev(vdev
)->desc
->status
= status
;
123 * The other piece of infrastructure virtio needs is a "virtqueue": a way of
124 * the Guest device registering buffers for the other side to read from or
125 * write into (ie. send and receive buffers). Each device can have multiple
126 * virtqueues: for example the console driver uses one queue for sending and
127 * another for receiving.
129 * Fortunately for us, a very fast shared-memory-plus-descriptors virtqueue
130 * already exists in virtio_ring.c. We just need to connect it up.
132 * We start with the information we need to keep about each virtqueue.
135 /*D:140 This is the information we remember about each virtqueue. */
136 struct lguest_vq_info
138 /* A copy of the information contained in the device config. */
139 struct lguest_vqconfig config
;
141 /* The address where we mapped the virtio ring, so we can unmap it. */
145 /* When the virtio_ring code wants to prod the Host, it calls us here and we
146 * make a hypercall. We hand the page number of the virtqueue so the Host
147 * knows which virtqueue we're talking about. */
148 static void lg_notify(struct virtqueue
*vq
)
150 /* We store our virtqueue information in the "priv" pointer of the
151 * virtqueue structure. */
152 struct lguest_vq_info
*lvq
= vq
->priv
;
154 hcall(LHCALL_NOTIFY
, lvq
->config
.pfn
<< PAGE_SHIFT
, 0, 0);
157 /* This routine finds the first virtqueue described in the configuration of
158 * this device and sets it up.
160 * This is kind of an ugly duckling. It'd be nicer to have a standard
161 * representation of a virtqueue in the configuration space, but it seems that
162 * everyone wants to do it differently. The KVM coders want the Guest to
163 * allocate its own pages and tell the Host where they are, but for lguest it's
164 * simpler for the Host to simply tell us where the pages are.
166 * So we provide devices with a "find virtqueue and set it up" function. */
167 static struct virtqueue
*lg_find_vq(struct virtio_device
*vdev
,
168 bool (*callback
)(struct virtqueue
*vq
))
170 struct lguest_vq_info
*lvq
;
171 struct virtqueue
*vq
;
176 /* Look for a field of the correct type to mark a virtqueue. Note that
177 * if this succeeds, then the type will be changed so it won't be found
178 * again, and future lg_find_vq() calls will find the next
179 * virtqueue (if any). */
180 token
= vdev
->config
->find(vdev
, VIRTIO_CONFIG_F_VIRTQUEUE
, &len
);
182 return ERR_PTR(-ENOENT
);
184 lvq
= kmalloc(sizeof(*lvq
), GFP_KERNEL
);
186 return ERR_PTR(-ENOMEM
);
188 /* Note: we could use a configuration space inside here, just like we
189 * do for the device. This would allow expansion in future, because
190 * our configuration system is designed to be expansible. But this is
192 if (len
!= sizeof(lvq
->config
)) {
193 dev_err(&vdev
->dev
, "Unexpected virtio config len %u\n", len
);
197 /* Make a copy of the "struct lguest_vqconfig" field. We need a copy
198 * because the config space might not be aligned correctly. */
199 vdev
->config
->get(vdev
, token
, &lvq
->config
, sizeof(lvq
->config
));
201 /* Figure out how many pages the ring will take, and map that memory */
202 lvq
->pages
= lguest_map((unsigned long)lvq
->config
.pfn
<< PAGE_SHIFT
,
203 DIV_ROUND_UP(vring_size(lvq
->config
.num
,
211 /* OK, tell virtio_ring.c to set up a virtqueue now we know its size
212 * and we've got a pointer to its pages. */
213 vq
= vring_new_virtqueue(lvq
->config
.num
, vdev
, lvq
->pages
,
214 lg_notify
, callback
);
220 /* Tell the interrupt for this virtqueue to go to the virtio_ring
221 * interrupt handler. */
222 /* FIXME: We used to have a flag for the Host to tell us we could use
223 * the interrupt as a source of randomness: it'd be nice to have that
225 err
= request_irq(lvq
->config
.irq
, vring_interrupt
, IRQF_SHARED
,
226 vdev
->dev
.bus_id
, vq
);
230 /* Last of all we hook up our 'struct lguest_vq_info" to the
231 * virtqueue's priv pointer. */
236 vring_del_virtqueue(vq
);
238 lguest_unmap(lvq
->pages
);
245 /* Cleaning up a virtqueue is easy */
246 static void lg_del_vq(struct virtqueue
*vq
)
248 struct lguest_vq_info
*lvq
= vq
->priv
;
250 /* Tell virtio_ring.c to free the virtqueue. */
251 vring_del_virtqueue(vq
);
252 /* Unmap the pages containing the ring. */
253 lguest_unmap(lvq
->pages
);
254 /* Free our own queue information. */
258 /* The ops structure which hooks everything together. */
259 static struct virtio_config_ops lguest_config_ops
= {
263 .get_status
= lg_get_status
,
264 .set_status
= lg_set_status
,
265 .find_vq
= lg_find_vq
,
269 /* The root device for the lguest virtio devices. This makes them appear as
270 * /sys/devices/lguest/0,1,2 not /sys/devices/0,1,2. */
271 static struct device lguest_root
= {
276 /*D:120 This is the core of the lguest bus: actually adding a new device.
277 * It's a separate function because it's neater that way, and because an
278 * earlier version of the code supported hotplug and unplug. They were removed
279 * early on because they were never used.
281 * As Andrew Tridgell says, "Untested code is buggy code".
283 * It's worth reading this carefully: we start with a pointer to the new device
284 * descriptor in the "lguest_devices" page. */
285 static void add_lguest_device(struct lguest_device_desc
*d
)
287 struct lguest_device
*ldev
;
289 /* Start with zeroed memory; Linux's device layer seems to count on
291 ldev
= kzalloc(sizeof(*ldev
), GFP_KERNEL
);
293 printk(KERN_EMERG
"Cannot allocate lguest dev %u\n",
298 /* This devices' parent is the lguest/ dir. */
299 ldev
->vdev
.dev
.parent
= &lguest_root
;
300 /* We have a unique device index thanks to the dev_index counter. */
301 ldev
->vdev
.index
= dev_index
++;
302 /* The device type comes straight from the descriptor. There's also a
303 * device vendor field in the virtio_device struct, which we leave as
305 ldev
->vdev
.id
.device
= d
->type
;
306 /* We have a simple set of routines for querying the device's
307 * configuration information and setting its status. */
308 ldev
->vdev
.config
= &lguest_config_ops
;
309 /* And we remember the device's descriptor for lguest_config_ops. */
312 /* register_virtio_device() sets up the generic fields for the struct
313 * virtio_device and calls device_register(). This makes the bus
314 * infrastructure look for a matching driver. */
315 if (register_virtio_device(&ldev
->vdev
) != 0) {
316 printk(KERN_ERR
"Failed to register lguest device %u\n",
322 /*D:110 scan_devices() simply iterates through the device page. The type 0 is
323 * reserved to mean "end of devices". */
324 static void scan_devices(void)
327 struct lguest_device_desc
*d
;
329 /* We start at the page beginning, and skip over each entry. */
330 for (i
= 0; i
< PAGE_SIZE
; i
+= sizeof(*d
) + d
->config_len
) {
331 d
= lguest_devices
+ i
;
333 /* Once we hit a zero, stop. */
337 add_lguest_device(d
);
341 /*D:105 Fairly early in boot, lguest_devices_init() is called to set up the
342 * lguest device infrastructure. We check that we are a Guest by checking
343 * pv_info.name: there are other ways of checking, but this seems most
346 * So we can access the "struct lguest_device_desc"s easily, we map that memory
347 * and store the pointer in the global "lguest_devices". Then we register a
348 * root device from which all our devices will hang (this seems to be the
349 * correct sysfs incantation).
351 * Finally we call scan_devices() which adds all the devices found in the
352 * lguest_devices page. */
353 static int __init
lguest_devices_init(void)
355 if (strcmp(pv_info
.name
, "lguest") != 0)
358 if (device_register(&lguest_root
) != 0)
359 panic("Could not register lguest root");
361 /* Devices are in a single page above top of "normal" mem */
362 lguest_devices
= lguest_map(max_pfn
<<PAGE_SHIFT
, 1);
367 /* We do this after core stuff, but before the drivers. */
368 postcore_initcall(lguest_devices_init
);
370 /*D:150 At this point in the journey we used to now wade through the lguest
371 * devices themselves: net, block and console. Since they're all now virtio
372 * devices rather than lguest-specific, I've decided to ignore them. Mostly,
373 * they're kind of boring. But this does mean you'll never experience the
374 * thrill of reading the forbidden love scene buried deep in the block driver.
376 * "make Launcher" beckons, where we answer questions like "Where do Guests
377 * come from?", and "What do you do when someone asks for optimization?". */