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>
11 /*L:055 When something happens, the Waker process needs a way to stop the
12 * kernel running the Guest and return to the Launcher. So the Waker writes
13 * LHREQ_BREAK and the value "1" to /dev/lguest to do this. Once the Launcher
14 * has done whatever needs attention, it writes LHREQ_BREAK and "0" to release
16 static int break_guest_out(struct lg_cpu *cpu, const unsigned long __user*input)
20 /* Fetch whether they're turning break on or off. */
21 if (get_user(on, input) != 0)
26 /* Pop it out of the Guest (may be running on different CPU) */
27 wake_up_process(cpu->tsk);
28 /* Wait for them to reset it */
29 return wait_event_interruptible(cpu->break_wq, !cpu->break_out);
32 wake_up(&cpu->break_wq);
37 /*L:050 Sending an interrupt is done by writing LHREQ_IRQ and an interrupt
38 * number to /dev/lguest. */
39 static int user_send_irq(struct lg_cpu *cpu, const unsigned long __user *input)
43 if (get_user(irq, input) != 0)
45 if (irq >= LGUEST_IRQS)
47 /* Next time the Guest runs, the core code will see if it can deliver
49 set_bit(irq, cpu->irqs_pending);
53 /*L:040 Once our Guest is initialized, the Launcher makes it run by reading
54 * from /dev/lguest. */
55 static ssize_t read(struct file *file, char __user *user, size_t size,loff_t*o)
57 struct lguest *lg = file->private_data;
59 unsigned int cpu_id = *o;
61 /* You must write LHREQ_INITIALIZE first! */
65 /* Watch out for arbitrary vcpu indexes! */
66 if (cpu_id >= lg->nr_cpus)
69 cpu = &lg->cpus[cpu_id];
71 /* If you're not the task which owns the Guest, go away. */
72 if (current != cpu->tsk)
75 /* If the guest is already dead, we indicate why */
79 /* lg->dead either contains an error code, or a string. */
81 return PTR_ERR(lg->dead);
83 /* We can only return as much as the buffer they read with. */
84 len = min(size, strlen(lg->dead)+1);
85 if (copy_to_user(user, lg->dead, len) != 0)
90 /* If we returned from read() last time because the Guest notified,
92 if (cpu->pending_notify)
93 cpu->pending_notify = 0;
95 /* Run the Guest until something interesting happens. */
96 return run_guest(cpu, (unsigned long __user *)user);
99 static int lg_cpu_start(struct lg_cpu *cpu, unsigned id, unsigned long start_ip)
105 cpu->lg = container_of((cpu - id), struct lguest, cpus[0]);
109 /* We need a complete page for the Guest registers: they are accessible
110 * to the Guest and we can only grant it access to whole pages. */
111 cpu->regs_page = get_zeroed_page(GFP_KERNEL);
115 /* We actually put the registers at the bottom of the page. */
116 cpu->regs = (void *)cpu->regs_page + PAGE_SIZE - sizeof(*cpu->regs);
118 /* Now we initialize the Guest's registers, handing it the start
120 lguest_arch_setup_regs(cpu, start_ip);
122 /* Initialize the queue for the waker to wait on */
123 init_waitqueue_head(&cpu->break_wq);
125 /* We keep a pointer to the Launcher task (ie. current task) for when
126 * other Guests want to wake this one (inter-Guest I/O). */
129 /* We need to keep a pointer to the Launcher's memory map, because if
130 * the Launcher dies we need to clean it up. If we don't keep a
131 * reference, it is destroyed before close() is called. */
132 cpu->mm = get_task_mm(cpu->tsk);
134 /* We remember which CPU's pages this Guest used last, for optimization
135 * when the same Guest runs on the same CPU twice. */
136 cpu->last_pages = NULL;
141 /*L:020 The initialization write supplies 4 pointer sized (32 or 64 bit)
142 * values (in addition to the LHREQ_INITIALIZE value). These are:
144 * base: The start of the Guest-physical memory inside the Launcher memory.
146 * pfnlimit: The highest (Guest-physical) page number the Guest should be
147 * allowed to access. The Guest memory lives inside the Launcher, so it sets
148 * this to ensure the Guest can only reach its own memory.
150 * pgdir: The (Guest-physical) address of the top of the initial Guest
151 * pagetables (which are set up by the Launcher).
153 * start: The first instruction to execute ("eip" in x86-speak).
155 static int initialize(struct file *file, const unsigned long __user *input)
157 /* "struct lguest" contains everything we (the Host) know about a
161 unsigned long args[4];
163 /* We grab the Big Lguest lock, which protects against multiple
164 * simultaneous initializations. */
165 mutex_lock(&lguest_lock);
166 /* You can't initialize twice! Close the device and start again... */
167 if (file->private_data) {
172 if (copy_from_user(args, input, sizeof(args)) != 0) {
177 lg = kzalloc(sizeof(*lg), GFP_KERNEL);
183 /* Populate the easy fields of our "struct lguest" */
184 lg->mem_base = (void __user *)(long)args[0];
185 lg->pfn_limit = args[1];
187 /* This is the first cpu */
188 err = lg_cpu_start(&lg->cpus[0], 0, args[3]);
192 /* Initialize the Guest's shadow page tables, using the toplevel
193 * address the Launcher gave us. This allocates memory, so can
195 err = init_guest_pagetable(lg, args[2]);
199 /* We keep our "struct lguest" in the file's private_data. */
200 file->private_data = lg;
202 mutex_unlock(&lguest_lock);
204 /* And because this is a write() call, we return the length used. */
208 /* FIXME: This should be in free_vcpu */
209 free_page(lg->cpus[0].regs_page);
213 mutex_unlock(&lguest_lock);
217 /*L:010 The first operation the Launcher does must be a write. All writes
218 * start with an unsigned long number: for the first write this must be
219 * LHREQ_INITIALIZE to set up the Guest. After that the Launcher can use
220 * writes of other values to send interrupts. */
221 static ssize_t write(struct file *file, const char __user *in,
222 size_t size, loff_t *off)
224 /* Once the guest is initialized, we hold the "struct lguest" in the
225 * file private data. */
226 struct lguest *lg = file->private_data;
227 const unsigned long __user *input = (const unsigned long __user *)in;
229 struct lg_cpu *uninitialized_var(cpu);
230 unsigned int cpu_id = *off;
232 if (get_user(req, input) != 0)
236 /* If you haven't initialized, you must do that first. */
237 if (req != LHREQ_INITIALIZE) {
238 if (!lg || (cpu_id >= lg->nr_cpus))
240 cpu = &lg->cpus[cpu_id];
245 /* Once the Guest is dead, all you can do is read() why it died. */
249 /* If you're not the task which owns the Guest, you can only break */
250 if (lg && current != cpu->tsk && req != LHREQ_BREAK)
254 case LHREQ_INITIALIZE:
255 return initialize(file, input);
257 return user_send_irq(cpu, input);
259 return break_guest_out(cpu, input);
265 /*L:060 The final piece of interface code is the close() routine. It reverses
266 * everything done in initialize(). This is usually called because the
269 * Note that the close routine returns 0 or a negative error number: it can't
270 * really fail, but it can whine. I blame Sun for this wart, and K&R C for
271 * letting them do it. :*/
272 static int close(struct inode *inode, struct file *file)
274 struct lguest *lg = file->private_data;
277 /* If we never successfully initialized, there's nothing to clean up */
281 /* We need the big lock, to protect from inter-guest I/O and other
282 * Launchers initializing guests. */
283 mutex_lock(&lguest_lock);
285 /* Free up the shadow page tables for the Guest. */
286 free_guest_pagetable(lg);
288 for (i = 0; i < lg->nr_cpus; i++) {
289 /* Cancels the hrtimer set via LHCALL_SET_CLOCKEVENT. */
290 hrtimer_cancel(&lg->cpus[i].hrt);
291 /* We can free up the register page we allocated. */
292 free_page(lg->cpus[i].regs_page);
293 /* Now all the memory cleanups are done, it's safe to release
294 * the Launcher's memory management structure. */
295 mmput(lg->cpus[i].mm);
297 /* If lg->dead doesn't contain an error code it will be NULL or a
298 * kmalloc()ed string, either of which is ok to hand to kfree(). */
299 if (!IS_ERR(lg->dead))
301 /* We clear the entire structure, which also marks it as free for the
303 memset(lg, 0, sizeof(*lg));
304 /* Release lock and exit. */
305 mutex_unlock(&lguest_lock);
311 * Welcome to our journey through the Launcher!
313 * The Launcher is the Host userspace program which sets up, runs and services
314 * the Guest. In fact, many comments in the Drivers which refer to "the Host"
315 * doing things are inaccurate: the Launcher does all the device handling for
316 * the Guest, but the Guest can't know that.
318 * Just to confuse you: to the Host kernel, the Launcher *is* the Guest and we
319 * shall see more of that later.
321 * We begin our understanding with the Host kernel interface which the Launcher
322 * uses: reading and writing a character device called /dev/lguest. All the
323 * work happens in the read(), write() and close() routines: */
324 static struct file_operations lguest_fops = {
325 .owner = THIS_MODULE,
331 /* This is a textbook example of a "misc" character device. Populate a "struct
332 * miscdevice" and register it with misc_register(). */
333 static struct miscdevice lguest_dev = {
334 .minor = MISC_DYNAMIC_MINOR,
336 .fops = &lguest_fops,
339 int __init lguest_device_init(void)
341 return misc_register(&lguest_dev);
344 void __exit lguest_device_remove(void)
346 misc_deregister(&lguest_dev);