1 /*
2  * arch/s390/appldata/appldata_base.c
3  *
4  * Base infrastructure for Linux-z/VM Monitor Stream, Stage 1.
5  * Exports appldata_register_ops() and appldata_unregister_ops() for the
6  * data gathering modules.
7  *
8  * Copyright IBM Corp. 2003, 2009
9  *
10  * Author: Gerald Schaefer <gerald.schaefer@de.ibm.com>
11  */
12 
13 #define KMSG_COMPONENT	"appldata"
14 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
15 
16 #include <linux/module.h>
17 #include <linux/init.h>
18 #include <linux/slab.h>
19 #include <linux/errno.h>
20 #include <linux/interrupt.h>
21 #include <linux/proc_fs.h>
22 #include <linux/mm.h>
23 #include <linux/swap.h>
24 #include <linux/pagemap.h>
25 #include <linux/sysctl.h>
26 #include <linux/notifier.h>
27 #include <linux/cpu.h>
28 #include <linux/workqueue.h>
29 #include <linux/suspend.h>
30 #include <linux/platform_device.h>
31 #include <asm/appldata.h>
32 #include <asm/timer.h>
33 #include <asm/uaccess.h>
34 #include <asm/io.h>
35 #include <asm/smp.h>
36 
37 #include "appldata.h"
38 
39 
40 #define APPLDATA_CPU_INTERVAL	10000		/* default (CPU) time for
41 						   sampling interval in
42 						   milliseconds */
43 
44 #define TOD_MICRO	0x01000			/* nr. of TOD clock units
45 						   for 1 microsecond */
46 
47 static struct platform_device *appldata_pdev;
48 
49 /*
50  * /proc entries (sysctl)
51  */
52 static const char appldata_proc_name[APPLDATA_PROC_NAME_LENGTH] = "appldata";
53 static int appldata_timer_handler(ctl_table *ctl, int write, struct file *filp,
54 				  void __user *buffer, size_t *lenp, loff_t *ppos);
55 static int appldata_interval_handler(ctl_table *ctl, int write,
56 					 struct file *filp,
57 					 void __user *buffer,
58 					 size_t *lenp, loff_t *ppos);
59 
60 static struct ctl_table_header *appldata_sysctl_header;
61 static struct ctl_table appldata_table[] = {
62 	{
63 		.procname	= "timer",
64 		.mode		= S_IRUGO | S_IWUSR,
65 		.proc_handler	= &appldata_timer_handler,
66 	},
67 	{
68 		.procname	= "interval",
69 		.mode		= S_IRUGO | S_IWUSR,
70 		.proc_handler	= &appldata_interval_handler,
71 	},
72 	{ },
73 };
74 
75 static struct ctl_table appldata_dir_table[] = {
76 	{
77 		.procname	= appldata_proc_name,
78 		.maxlen		= 0,
79 		.mode		= S_IRUGO | S_IXUGO,
80 		.child		= appldata_table,
81 	},
82 	{ },
83 };
84 
85 /*
86  * Timer
87  */
88 static DEFINE_PER_CPU(struct vtimer_list, appldata_timer);
89 static atomic_t appldata_expire_count = ATOMIC_INIT(0);
90 
91 static DEFINE_SPINLOCK(appldata_timer_lock);
92 static int appldata_interval = APPLDATA_CPU_INTERVAL;
93 static int appldata_timer_active;
94 static int appldata_timer_suspended = 0;
95 
96 /*
97  * Work queue
98  */
99 static struct workqueue_struct *appldata_wq;
100 static void appldata_work_fn(struct work_struct *work);
101 static DECLARE_WORK(appldata_work, appldata_work_fn);
102 
103 
104 /*
105  * Ops list
106  */
107 static DEFINE_MUTEX(appldata_ops_mutex);
108 static LIST_HEAD(appldata_ops_list);
109 
110 
111 /*************************** timer, work, DIAG *******************************/
112 /*
113  * appldata_timer_function()
114  *
115  * schedule work and reschedule timer
116  */
117 static void appldata_timer_function(unsigned long data)
118 {
119 	if (atomic_dec_and_test(&appldata_expire_count)) {
120 		atomic_set(&appldata_expire_count, num_online_cpus());
121 		queue_work(appldata_wq, (struct work_struct *) data);
122 	}
123 }
124 
125 /*
126  * appldata_work_fn()
127  *
128  * call data gathering function for each (active) module
129  */
130 static void appldata_work_fn(struct work_struct *work)
131 {
132 	struct list_head *lh;
133 	struct appldata_ops *ops;
134 	int i;
135 
136 	i = 0;
137 	get_online_cpus();
138 	mutex_lock(&appldata_ops_mutex);
139 	list_for_each(lh, &appldata_ops_list) {
140 		ops = list_entry(lh, struct appldata_ops, list);
141 		if (ops->active == 1) {
142 			ops->callback(ops->data);
143 		}
144 	}
145 	mutex_unlock(&appldata_ops_mutex);
146 	put_online_cpus();
147 }
148 
149 /*
150  * appldata_diag()
151  *
152  * prepare parameter list, issue DIAG 0xDC
153  */
154 int appldata_diag(char record_nr, u16 function, unsigned long buffer,
155 			u16 length, char *mod_lvl)
156 {
157 	struct appldata_product_id id = {
158 		.prod_nr    = {0xD3, 0xC9, 0xD5, 0xE4,
159 			       0xE7, 0xD2, 0xD9},	/* "LINUXKR" */
160 		.prod_fn    = 0xD5D3,			/* "NL" */
161 		.version_nr = 0xF2F6,			/* "26" */
162 		.release_nr = 0xF0F1,			/* "01" */
163 	};
164 
165 	id.record_nr = record_nr;
166 	id.mod_lvl = (mod_lvl[0]) << 8 | mod_lvl[1];
167 	return appldata_asm(&id, function, (void *) buffer, length);
168 }
169 /************************ timer, work, DIAG <END> ****************************/
170 
171 
172 /****************************** /proc stuff **********************************/
173 
174 /*
175  * appldata_mod_vtimer_wrap()
176  *
177  * wrapper function for mod_virt_timer(), because smp_call_function_single()
178  * accepts only one parameter.
179  */
180 static void __appldata_mod_vtimer_wrap(void *p) {
181 	struct {
182 		struct vtimer_list *timer;
183 		u64    expires;
184 	} *args = p;
185 	mod_virt_timer_periodic(args->timer, args->expires);
186 }
187 
188 #define APPLDATA_ADD_TIMER	0
189 #define APPLDATA_DEL_TIMER	1
190 #define APPLDATA_MOD_TIMER	2
191 
192 /*
193  * __appldata_vtimer_setup()
194  *
195  * Add, delete or modify virtual timers on all online cpus.
196  * The caller needs to get the appldata_timer_lock spinlock.
197  */
198 static void
199 __appldata_vtimer_setup(int cmd)
200 {
201 	u64 per_cpu_interval;
202 	int i;
203 
204 	switch (cmd) {
205 	case APPLDATA_ADD_TIMER:
206 		if (appldata_timer_active)
207 			break;
208 		per_cpu_interval = (u64) (appldata_interval*1000 /
209 					  num_online_cpus()) * TOD_MICRO;
210 		for_each_online_cpu(i) {
211 			per_cpu(appldata_timer, i).expires = per_cpu_interval;
212 			smp_call_function_single(i, add_virt_timer_periodic,
213 						 &per_cpu(appldata_timer, i),
214 						 1);
215 		}
216 		appldata_timer_active = 1;
217 		break;
218 	case APPLDATA_DEL_TIMER:
219 		for_each_online_cpu(i)
220 			del_virt_timer(&per_cpu(appldata_timer, i));
221 		if (!appldata_timer_active)
222 			break;
223 		appldata_timer_active = 0;
224 		atomic_set(&appldata_expire_count, num_online_cpus());
225 		break;
226 	case APPLDATA_MOD_TIMER:
227 		per_cpu_interval = (u64) (appldata_interval*1000 /
228 					  num_online_cpus()) * TOD_MICRO;
229 		if (!appldata_timer_active)
230 			break;
231 		for_each_online_cpu(i) {
232 			struct {
233 				struct vtimer_list *timer;
234 				u64    expires;
235 			} args;
236 			args.timer = &per_cpu(appldata_timer, i);
237 			args.expires = per_cpu_interval;
238 			smp_call_function_single(i, __appldata_mod_vtimer_wrap,
239 						 &args, 1);
240 		}
241 	}
242 }
243 
244 /*
245  * appldata_timer_handler()
246  *
247  * Start/Stop timer, show status of timer (0 = not active, 1 = active)
248  */
249 static int
250 appldata_timer_handler(ctl_table *ctl, int write, struct file *filp,
251 			   void __user *buffer, size_t *lenp, loff_t *ppos)
252 {
253 	int len;
254 	char buf[2];
255 
256 	if (!*lenp || *ppos) {
257 		*lenp = 0;
258 		return 0;
259 	}
260 	if (!write) {
261 		len = sprintf(buf, appldata_timer_active ? "1\n" : "0\n");
262 		if (len > *lenp)
263 			len = *lenp;
264 		if (copy_to_user(buffer, buf, len))
265 			return -EFAULT;
266 		goto out;
267 	}
268 	len = *lenp;
269 	if (copy_from_user(buf, buffer, len > sizeof(buf) ? sizeof(buf) : len))
270 		return -EFAULT;
271 	get_online_cpus();
272 	spin_lock(&appldata_timer_lock);
273 	if (buf[0] == '1')
274 		__appldata_vtimer_setup(APPLDATA_ADD_TIMER);
275 	else if (buf[0] == '0')
276 		__appldata_vtimer_setup(APPLDATA_DEL_TIMER);
277 	spin_unlock(&appldata_timer_lock);
278 	put_online_cpus();
279 out:
280 	*lenp = len;
281 	*ppos += len;
282 	return 0;
283 }
284 
285 /*
286  * appldata_interval_handler()
287  *
288  * Set (CPU) timer interval for collection of data (in milliseconds), show
289  * current timer interval.
290  */
291 static int
292 appldata_interval_handler(ctl_table *ctl, int write, struct file *filp,
293 			   void __user *buffer, size_t *lenp, loff_t *ppos)
294 {
295 	int len, interval;
296 	char buf[16];
297 
298 	if (!*lenp || *ppos) {
299 		*lenp = 0;
300 		return 0;
301 	}
302 	if (!write) {
303 		len = sprintf(buf, "%i\n", appldata_interval);
304 		if (len > *lenp)
305 			len = *lenp;
306 		if (copy_to_user(buffer, buf, len))
307 			return -EFAULT;
308 		goto out;
309 	}
310 	len = *lenp;
311 	if (copy_from_user(buf, buffer, len > sizeof(buf) ? sizeof(buf) : len)) {
312 		return -EFAULT;
313 	}
314 	interval = 0;
315 	sscanf(buf, "%i", &interval);
316 	if (interval <= 0)
317 		return -EINVAL;
318 
319 	get_online_cpus();
320 	spin_lock(&appldata_timer_lock);
321 	appldata_interval = interval;
322 	__appldata_vtimer_setup(APPLDATA_MOD_TIMER);
323 	spin_unlock(&appldata_timer_lock);
324 	put_online_cpus();
325 out:
326 	*lenp = len;
327 	*ppos += len;
328 	return 0;
329 }
330 
331 /*
332  * appldata_generic_handler()
333  *
334  * Generic start/stop monitoring and DIAG, show status of
335  * monitoring (0 = not in process, 1 = in process)
336  */
337 static int
338 appldata_generic_handler(ctl_table *ctl, int write, struct file *filp,
339 			   void __user *buffer, size_t *lenp, loff_t *ppos)
340 {
341 	struct appldata_ops *ops = NULL, *tmp_ops;
342 	int rc, len, found;
343 	char buf[2];
344 	struct list_head *lh;
345 
346 	found = 0;
347 	mutex_lock(&appldata_ops_mutex);
348 	list_for_each(lh, &appldata_ops_list) {
349 		tmp_ops = list_entry(lh, struct appldata_ops, list);
350 		if (&tmp_ops->ctl_table[2] == ctl) {
351 			found = 1;
352 		}
353 	}
354 	if (!found) {
355 		mutex_unlock(&appldata_ops_mutex);
356 		return -ENODEV;
357 	}
358 	ops = ctl->data;
359 	if (!try_module_get(ops->owner)) {	// protect this function
360 		mutex_unlock(&appldata_ops_mutex);
361 		return -ENODEV;
362 	}
363 	mutex_unlock(&appldata_ops_mutex);
364 
365 	if (!*lenp || *ppos) {
366 		*lenp = 0;
367 		module_put(ops->owner);
368 		return 0;
369 	}
370 	if (!write) {
371 		len = sprintf(buf, ops->active ? "1\n" : "0\n");
372 		if (len > *lenp)
373 			len = *lenp;
374 		if (copy_to_user(buffer, buf, len)) {
375 			module_put(ops->owner);
376 			return -EFAULT;
377 		}
378 		goto out;
379 	}
380 	len = *lenp;
381 	if (copy_from_user(buf, buffer,
382 			   len > sizeof(buf) ? sizeof(buf) : len)) {
383 		module_put(ops->owner);
384 		return -EFAULT;
385 	}
386 
387 	mutex_lock(&appldata_ops_mutex);
388 	if ((buf[0] == '1') && (ops->active == 0)) {
389 		// protect work queue callback
390 		if (!try_module_get(ops->owner)) {
391 			mutex_unlock(&appldata_ops_mutex);
392 			module_put(ops->owner);
393 			return -ENODEV;
394 		}
395 		ops->callback(ops->data);	// init record
396 		rc = appldata_diag(ops->record_nr,
397 					APPLDATA_START_INTERVAL_REC,
398 					(unsigned long) ops->data, ops->size,
399 					ops->mod_lvl);
400 		if (rc != 0) {
401 			pr_err("Starting the data collection for %s "
402 			       "failed with rc=%d\n", ops->name, rc);
403 			module_put(ops->owner);
404 		} else
405 			ops->active = 1;
406 	} else if ((buf[0] == '0') && (ops->active == 1)) {
407 		ops->active = 0;
408 		rc = appldata_diag(ops->record_nr, APPLDATA_STOP_REC,
409 				(unsigned long) ops->data, ops->size,
410 				ops->mod_lvl);
411 		if (rc != 0)
412 			pr_err("Stopping the data collection for %s "
413 			       "failed with rc=%d\n", ops->name, rc);
414 		module_put(ops->owner);
415 	}
416 	mutex_unlock(&appldata_ops_mutex);
417 out:
418 	*lenp = len;
419 	*ppos += len;
420 	module_put(ops->owner);
421 	return 0;
422 }
423 
424 /*************************** /proc stuff <END> *******************************/
425 
426 
427 /************************* module-ops management *****************************/
428 /*
429  * appldata_register_ops()
430  *
431  * update ops list, register /proc/sys entries
432  */
433 int appldata_register_ops(struct appldata_ops *ops)
434 {
435 	if (ops->size > APPLDATA_MAX_REC_SIZE)
436 		return -EINVAL;
437 
438 	ops->ctl_table = kzalloc(4 * sizeof(struct ctl_table), GFP_KERNEL);
439 	if (!ops->ctl_table)
440 		return -ENOMEM;
441 
442 	mutex_lock(&appldata_ops_mutex);
443 	list_add(&ops->list, &appldata_ops_list);
444 	mutex_unlock(&appldata_ops_mutex);
445 
446 	ops->ctl_table[0].procname = appldata_proc_name;
447 	ops->ctl_table[0].maxlen   = 0;
448 	ops->ctl_table[0].mode     = S_IRUGO | S_IXUGO;
449 	ops->ctl_table[0].child    = &ops->ctl_table[2];
450 
451 	ops->ctl_table[2].procname = ops->name;
452 	ops->ctl_table[2].mode     = S_IRUGO | S_IWUSR;
453 	ops->ctl_table[2].proc_handler = appldata_generic_handler;
454 	ops->ctl_table[2].data = ops;
455 
456 	ops->sysctl_header = register_sysctl_table(ops->ctl_table);
457 	if (!ops->sysctl_header)
458 		goto out;
459 	return 0;
460 out:
461 	mutex_lock(&appldata_ops_mutex);
462 	list_del(&ops->list);
463 	mutex_unlock(&appldata_ops_mutex);
464 	kfree(ops->ctl_table);
465 	return -ENOMEM;
466 }
467 
468 /*
469  * appldata_unregister_ops()
470  *
471  * update ops list, unregister /proc entries, stop DIAG if necessary
472  */
473 void appldata_unregister_ops(struct appldata_ops *ops)
474 {
475 	mutex_lock(&appldata_ops_mutex);
476 	list_del(&ops->list);
477 	mutex_unlock(&appldata_ops_mutex);
478 	unregister_sysctl_table(ops->sysctl_header);
479 	kfree(ops->ctl_table);
480 }
481 /********************** module-ops management <END> **************************/
482 
483 
484 /**************************** suspend / resume *******************************/
485 static int appldata_freeze(struct device *dev)
486 {
487 	struct appldata_ops *ops;
488 	int rc;
489 	struct list_head *lh;
490 
491 	get_online_cpus();
492 	spin_lock(&appldata_timer_lock);
493 	if (appldata_timer_active) {
494 		__appldata_vtimer_setup(APPLDATA_DEL_TIMER);
495 		appldata_timer_suspended = 1;
496 	}
497 	spin_unlock(&appldata_timer_lock);
498 	put_online_cpus();
499 
500 	mutex_lock(&appldata_ops_mutex);
501 	list_for_each(lh, &appldata_ops_list) {
502 		ops = list_entry(lh, struct appldata_ops, list);
503 		if (ops->active == 1) {
504 			rc = appldata_diag(ops->record_nr, APPLDATA_STOP_REC,
505 					(unsigned long) ops->data, ops->size,
506 					ops->mod_lvl);
507 			if (rc != 0)
508 				pr_err("Stopping the data collection for %s "
509 				       "failed with rc=%d\n", ops->name, rc);
510 		}
511 	}
512 	mutex_unlock(&appldata_ops_mutex);
513 	return 0;
514 }
515 
516 static int appldata_restore(struct device *dev)
517 {
518 	struct appldata_ops *ops;
519 	int rc;
520 	struct list_head *lh;
521 
522 	get_online_cpus();
523 	spin_lock(&appldata_timer_lock);
524 	if (appldata_timer_suspended) {
525 		__appldata_vtimer_setup(APPLDATA_ADD_TIMER);
526 		appldata_timer_suspended = 0;
527 	}
528 	spin_unlock(&appldata_timer_lock);
529 	put_online_cpus();
530 
531 	mutex_lock(&appldata_ops_mutex);
532 	list_for_each(lh, &appldata_ops_list) {
533 		ops = list_entry(lh, struct appldata_ops, list);
534 		if (ops->active == 1) {
535 			ops->callback(ops->data);	// init record
536 			rc = appldata_diag(ops->record_nr,
537 					APPLDATA_START_INTERVAL_REC,
538 					(unsigned long) ops->data, ops->size,
539 					ops->mod_lvl);
540 			if (rc != 0) {
541 				pr_err("Starting the data collection for %s "
542 				       "failed with rc=%d\n", ops->name, rc);
543 			}
544 		}
545 	}
546 	mutex_unlock(&appldata_ops_mutex);
547 	return 0;
548 }
549 
550 static int appldata_thaw(struct device *dev)
551 {
552 	return appldata_restore(dev);
553 }
554 
555 static struct dev_pm_ops appldata_pm_ops = {
556 	.freeze		= appldata_freeze,
557 	.thaw		= appldata_thaw,
558 	.restore	= appldata_restore,
559 };
560 
561 static struct platform_driver appldata_pdrv = {
562 	.driver = {
563 		.name	= "appldata",
564 		.owner	= THIS_MODULE,
565 		.pm	= &appldata_pm_ops,
566 	},
567 };
568 /************************* suspend / resume <END> ****************************/
569 
570 
571 /******************************* init / exit *********************************/
572 
573 static void __cpuinit appldata_online_cpu(int cpu)
574 {
575 	init_virt_timer(&per_cpu(appldata_timer, cpu));
576 	per_cpu(appldata_timer, cpu).function = appldata_timer_function;
577 	per_cpu(appldata_timer, cpu).data = (unsigned long)
578 		&appldata_work;
579 	atomic_inc(&appldata_expire_count);
580 	spin_lock(&appldata_timer_lock);
581 	__appldata_vtimer_setup(APPLDATA_MOD_TIMER);
582 	spin_unlock(&appldata_timer_lock);
583 }
584 
585 static void __cpuinit appldata_offline_cpu(int cpu)
586 {
587 	del_virt_timer(&per_cpu(appldata_timer, cpu));
588 	if (atomic_dec_and_test(&appldata_expire_count)) {
589 		atomic_set(&appldata_expire_count, num_online_cpus());
590 		queue_work(appldata_wq, &appldata_work);
591 	}
592 	spin_lock(&appldata_timer_lock);
593 	__appldata_vtimer_setup(APPLDATA_MOD_TIMER);
594 	spin_unlock(&appldata_timer_lock);
595 }
596 
597 static int __cpuinit appldata_cpu_notify(struct notifier_block *self,
598 					 unsigned long action,
599 					 void *hcpu)
600 {
601 	switch (action) {
602 	case CPU_ONLINE:
603 	case CPU_ONLINE_FROZEN:
604 		appldata_online_cpu((long) hcpu);
605 		break;
606 	case CPU_DEAD:
607 	case CPU_DEAD_FROZEN:
608 		appldata_offline_cpu((long) hcpu);
609 		break;
610 	default:
611 		break;
612 	}
613 	return NOTIFY_OK;
614 }
615 
616 static struct notifier_block __cpuinitdata appldata_nb = {
617 	.notifier_call = appldata_cpu_notify,
618 };
619 
620 /*
621  * appldata_init()
622  *
623  * init timer, register /proc entries
624  */
625 static int __init appldata_init(void)
626 {
627 	int i, rc;
628 
629 	rc = platform_driver_register(&appldata_pdrv);
630 	if (rc)
631 		return rc;
632 
633 	appldata_pdev = platform_device_register_simple("appldata", -1, NULL,
634 							0);
635 	if (IS_ERR(appldata_pdev)) {
636 		rc = PTR_ERR(appldata_pdev);
637 		goto out_driver;
638 	}
639 	appldata_wq = create_singlethread_workqueue("appldata");
640 	if (!appldata_wq) {
641 		rc = -ENOMEM;
642 		goto out_device;
643 	}
644 
645 	get_online_cpus();
646 	for_each_online_cpu(i)
647 		appldata_online_cpu(i);
648 	put_online_cpus();
649 
650 	/* Register cpu hotplug notifier */
651 	register_hotcpu_notifier(&appldata_nb);
652 
653 	appldata_sysctl_header = register_sysctl_table(appldata_dir_table);
654 	return 0;
655 
656 out_device:
657 	platform_device_unregister(appldata_pdev);
658 out_driver:
659 	platform_driver_unregister(&appldata_pdrv);
660 	return rc;
661 }
662 
663 __initcall(appldata_init);
664 
665 /**************************** init / exit <END> ******************************/
666 
667 EXPORT_SYMBOL_GPL(appldata_register_ops);
668 EXPORT_SYMBOL_GPL(appldata_unregister_ops);
669 EXPORT_SYMBOL_GPL(appldata_diag);
670 
671 #ifdef CONFIG_SWAP
672 EXPORT_SYMBOL_GPL(si_swapinfo);
673 #endif
674 EXPORT_SYMBOL_GPL(nr_threads);
675 EXPORT_SYMBOL_GPL(nr_running);
676 EXPORT_SYMBOL_GPL(nr_iowait);
677