1 /*P:200 This contains all the /dev/lguest code, whereby the userspace launcher
2 * controls and communicates with the Guest. For example, the first write will
3 * tell us the Guest's memory layout, pagetable, entry point and kernel address
4 * offset. A read will run the Guest until something happens, such as a signal
5 * or the Guest doing a NOTIFY out to the Launcher. :*/
6 #include <linux/uaccess.h>
7 #include <linux/miscdevice.h>
9 #include <linux/sched.h>
12 /*L:055 When something happens, the Waker process needs a way to stop the
13 * kernel running the Guest and return to the Launcher. So the Waker writes
14 * LHREQ_BREAK and the value "1" to /dev/lguest to do this. Once the Launcher
15 * has done whatever needs attention, it writes LHREQ_BREAK and "0" to release
17 static int break_guest_out(struct lg_cpu
*cpu
, const unsigned long __user
*input
)
21 /* Fetch whether they're turning break on or off. */
22 if (get_user(on
, input
) != 0)
27 /* Pop it out of the Guest (may be running on different CPU) */
28 wake_up_process(cpu
->tsk
);
29 /* Wait for them to reset it */
30 return wait_event_interruptible(cpu
->break_wq
, !cpu
->break_out
);
33 wake_up(&cpu
->break_wq
);
38 /*L:050 Sending an interrupt is done by writing LHREQ_IRQ and an interrupt
39 * number to /dev/lguest. */
40 static int user_send_irq(struct lg_cpu
*cpu
, const unsigned long __user
*input
)
44 if (get_user(irq
, input
) != 0)
46 if (irq
>= LGUEST_IRQS
)
48 /* Next time the Guest runs, the core code will see if it can deliver
50 set_bit(irq
, cpu
->irqs_pending
);
54 /*L:040 Once our Guest is initialized, the Launcher makes it run by reading
55 * from /dev/lguest. */
56 static ssize_t
read(struct file
*file
, char __user
*user
, size_t size
,loff_t
*o
)
58 struct lguest
*lg
= file
->private_data
;
60 unsigned int cpu_id
= *o
;
62 /* You must write LHREQ_INITIALIZE first! */
66 /* Watch out for arbitrary vcpu indexes! */
67 if (cpu_id
>= lg
->nr_cpus
)
70 cpu
= &lg
->cpus
[cpu_id
];
72 /* If you're not the task which owns the Guest, go away. */
73 if (current
!= cpu
->tsk
)
76 /* If the Guest is already dead, we indicate why */
80 /* lg->dead either contains an error code, or a string. */
82 return PTR_ERR(lg
->dead
);
84 /* We can only return as much as the buffer they read with. */
85 len
= min(size
, strlen(lg
->dead
)+1);
86 if (copy_to_user(user
, lg
->dead
, len
) != 0)
91 /* If we returned from read() last time because the Guest sent I/O,
93 if (cpu
->pending_notify
)
94 cpu
->pending_notify
= 0;
96 /* Run the Guest until something interesting happens. */
97 return run_guest(cpu
, (unsigned long __user
*)user
);
100 /*L:025 This actually initializes a CPU. For the moment, a Guest is only
101 * uniprocessor, so "id" is always 0. */
102 static int lg_cpu_start(struct lg_cpu
*cpu
, unsigned id
, unsigned long start_ip
)
104 /* We have a limited number the number of CPUs in the lguest struct. */
105 if (id
>= ARRAY_SIZE(cpu
->lg
->cpus
))
108 /* Set up this CPU's id, and pointer back to the lguest struct. */
110 cpu
->lg
= container_of((cpu
- id
), struct lguest
, cpus
[0]);
113 /* Each CPU has a timer it can set. */
116 /* We need a complete page for the Guest registers: they are accessible
117 * to the Guest and we can only grant it access to whole pages. */
118 cpu
->regs_page
= get_zeroed_page(GFP_KERNEL
);
122 /* We actually put the registers at the bottom of the page. */
123 cpu
->regs
= (void *)cpu
->regs_page
+ PAGE_SIZE
- sizeof(*cpu
->regs
);
125 /* Now we initialize the Guest's registers, handing it the start
127 lguest_arch_setup_regs(cpu
, start_ip
);
129 /* Initialize the queue for the Waker to wait on */
130 init_waitqueue_head(&cpu
->break_wq
);
132 /* We keep a pointer to the Launcher task (ie. current task) for when
133 * other Guests want to wake this one (eg. console input). */
136 /* We need to keep a pointer to the Launcher's memory map, because if
137 * the Launcher dies we need to clean it up. If we don't keep a
138 * reference, it is destroyed before close() is called. */
139 cpu
->mm
= get_task_mm(cpu
->tsk
);
141 /* We remember which CPU's pages this Guest used last, for optimization
142 * when the same Guest runs on the same CPU twice. */
143 cpu
->last_pages
= NULL
;
145 /* No error == success. */
149 /*L:020 The initialization write supplies 3 pointer sized (32 or 64 bit)
150 * values (in addition to the LHREQ_INITIALIZE value). These are:
152 * base: The start of the Guest-physical memory inside the Launcher memory.
154 * pfnlimit: The highest (Guest-physical) page number the Guest should be
155 * allowed to access. The Guest memory lives inside the Launcher, so it sets
156 * this to ensure the Guest can only reach its own memory.
158 * start: The first instruction to execute ("eip" in x86-speak).
160 static int initialize(struct file
*file
, const unsigned long __user
*input
)
162 /* "struct lguest" contains everything we (the Host) know about a
166 unsigned long args
[3];
168 /* We grab the Big Lguest lock, which protects against multiple
169 * simultaneous initializations. */
170 mutex_lock(&lguest_lock
);
171 /* You can't initialize twice! Close the device and start again... */
172 if (file
->private_data
) {
177 if (copy_from_user(args
, input
, sizeof(args
)) != 0) {
182 lg
= kzalloc(sizeof(*lg
), GFP_KERNEL
);
188 /* Populate the easy fields of our "struct lguest" */
189 lg
->mem_base
= (void __user
*)args
[0];
190 lg
->pfn_limit
= args
[1];
192 /* This is the first cpu (cpu 0) and it will start booting at args[2] */
193 err
= lg_cpu_start(&lg
->cpus
[0], 0, args
[2]);
197 /* Initialize the Guest's shadow page tables, using the toplevel
198 * address the Launcher gave us. This allocates memory, so can fail. */
199 err
= init_guest_pagetable(lg
);
203 /* We keep our "struct lguest" in the file's private_data. */
204 file
->private_data
= lg
;
206 mutex_unlock(&lguest_lock
);
208 /* And because this is a write() call, we return the length used. */
212 /* FIXME: This should be in free_vcpu */
213 free_page(lg
->cpus
[0].regs_page
);
217 mutex_unlock(&lguest_lock
);
221 /*L:010 The first operation the Launcher does must be a write. All writes
222 * start with an unsigned long number: for the first write this must be
223 * LHREQ_INITIALIZE to set up the Guest. After that the Launcher can use
224 * writes of other values to send interrupts.
226 * Note that we overload the "offset" in the /dev/lguest file to indicate what
227 * CPU number we're dealing with. Currently this is always 0, since we only
228 * support uniprocessor Guests, but you can see the beginnings of SMP support
230 static ssize_t
write(struct file
*file
, const char __user
*in
,
231 size_t size
, loff_t
*off
)
233 /* Once the Guest is initialized, we hold the "struct lguest" in the
234 * file private data. */
235 struct lguest
*lg
= file
->private_data
;
236 const unsigned long __user
*input
= (const unsigned long __user
*)in
;
238 struct lg_cpu
*uninitialized_var(cpu
);
239 unsigned int cpu_id
= *off
;
241 /* The first value tells us what this request is. */
242 if (get_user(req
, input
) != 0)
246 /* If you haven't initialized, you must do that first. */
247 if (req
!= LHREQ_INITIALIZE
) {
248 if (!lg
|| (cpu_id
>= lg
->nr_cpus
))
250 cpu
= &lg
->cpus
[cpu_id
];
252 /* Once the Guest is dead, you can only read() why it died. */
256 /* If you're not the task which owns the Guest, all you can do
257 * is break the Launcher out of running the Guest. */
258 if (current
!= cpu
->tsk
&& req
!= LHREQ_BREAK
)
263 case LHREQ_INITIALIZE
:
264 return initialize(file
, input
);
266 return user_send_irq(cpu
, input
);
268 return break_guest_out(cpu
, input
);
274 /*L:060 The final piece of interface code is the close() routine. It reverses
275 * everything done in initialize(). This is usually called because the
278 * Note that the close routine returns 0 or a negative error number: it can't
279 * really fail, but it can whine. I blame Sun for this wart, and K&R C for
280 * letting them do it. :*/
281 static int close(struct inode
*inode
, struct file
*file
)
283 struct lguest
*lg
= file
->private_data
;
286 /* If we never successfully initialized, there's nothing to clean up */
290 /* We need the big lock, to protect from inter-guest I/O and other
291 * Launchers initializing guests. */
292 mutex_lock(&lguest_lock
);
294 /* Free up the shadow page tables for the Guest. */
295 free_guest_pagetable(lg
);
297 for (i
= 0; i
< lg
->nr_cpus
; i
++) {
298 /* Cancels the hrtimer set via LHCALL_SET_CLOCKEVENT. */
299 hrtimer_cancel(&lg
->cpus
[i
].hrt
);
300 /* We can free up the register page we allocated. */
301 free_page(lg
->cpus
[i
].regs_page
);
302 /* Now all the memory cleanups are done, it's safe to release
303 * the Launcher's memory management structure. */
304 mmput(lg
->cpus
[i
].mm
);
306 /* If lg->dead doesn't contain an error code it will be NULL or a
307 * kmalloc()ed string, either of which is ok to hand to kfree(). */
308 if (!IS_ERR(lg
->dead
))
310 /* Free the memory allocated to the lguest_struct */
312 /* Release lock and exit. */
313 mutex_unlock(&lguest_lock
);
319 * Welcome to our journey through the Launcher!
321 * The Launcher is the Host userspace program which sets up, runs and services
322 * the Guest. In fact, many comments in the Drivers which refer to "the Host"
323 * doing things are inaccurate: the Launcher does all the device handling for
324 * the Guest, but the Guest can't know that.
326 * Just to confuse you: to the Host kernel, the Launcher *is* the Guest and we
327 * shall see more of that later.
329 * We begin our understanding with the Host kernel interface which the Launcher
330 * uses: reading and writing a character device called /dev/lguest. All the
331 * work happens in the read(), write() and close() routines: */
332 static struct file_operations lguest_fops
= {
333 .owner
= THIS_MODULE
,
339 /* This is a textbook example of a "misc" character device. Populate a "struct
340 * miscdevice" and register it with misc_register(). */
341 static struct miscdevice lguest_dev
= {
342 .minor
= MISC_DYNAMIC_MINOR
,
344 .fops
= &lguest_fops
,
347 int __init
lguest_device_init(void)
349 return misc_register(&lguest_dev
);
352 void __exit
lguest_device_remove(void)
354 misc_deregister(&lguest_dev
);