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 (C) 2003,2006 IBM Corporation, IBM Deutschland Entwicklung GmbH.
9  *
10  * Author: Gerald Schaefer <gerald.schaefer@de.ibm.com>
11  */
12 
13 #include <linux/module.h>
14 #include <linux/init.h>
15 #include <linux/slab.h>
16 #include <linux/errno.h>
17 #include <linux/interrupt.h>
18 #include <linux/proc_fs.h>
19 #include <linux/mm.h>
20 #include <linux/swap.h>
21 #include <linux/pagemap.h>
22 #include <linux/sysctl.h>
23 #include <linux/notifier.h>
24 #include <linux/cpu.h>
25 #include <linux/workqueue.h>
26 #include <asm/appldata.h>
27 #include <asm/timer.h>
28 #include <asm/uaccess.h>
29 #include <asm/io.h>
30 #include <asm/smp.h>
31 
32 #include "appldata.h"
33 
34 
35 #define MY_PRINT_NAME	"appldata"		/* for debug messages, etc. */
36 #define APPLDATA_CPU_INTERVAL	10000		/* default (CPU) time for
37 						   sampling interval in
38 						   milliseconds */
39 
40 #define TOD_MICRO	0x01000			/* nr. of TOD clock units
41 						   for 1 microsecond */
42 /*
43  * /proc entries (sysctl)
44  */
45 static const char appldata_proc_name[APPLDATA_PROC_NAME_LENGTH] = "appldata";
46 static int appldata_timer_handler(ctl_table *ctl, int write, struct file *filp,
47 				  void __user *buffer, size_t *lenp, loff_t *ppos);
48 static int appldata_interval_handler(ctl_table *ctl, int write,
49 					 struct file *filp,
50 					 void __user *buffer,
51 					 size_t *lenp, loff_t *ppos);
52 
53 static struct ctl_table_header *appldata_sysctl_header;
54 static struct ctl_table appldata_table[] = {
55 	{
56 		.procname	= "timer",
57 		.mode		= S_IRUGO | S_IWUSR,
58 		.proc_handler	= &appldata_timer_handler,
59 	},
60 	{
61 		.procname	= "interval",
62 		.mode		= S_IRUGO | S_IWUSR,
63 		.proc_handler	= &appldata_interval_handler,
64 	},
65 	{ },
66 };
67 
68 static struct ctl_table appldata_dir_table[] = {
69 	{
70 		.procname	= appldata_proc_name,
71 		.maxlen		= 0,
72 		.mode		= S_IRUGO | S_IXUGO,
73 		.child		= appldata_table,
74 	},
75 	{ },
76 };
77 
78 /*
79  * Timer
80  */
81 static DEFINE_PER_CPU(struct vtimer_list, appldata_timer);
82 static atomic_t appldata_expire_count = ATOMIC_INIT(0);
83 
84 static DEFINE_SPINLOCK(appldata_timer_lock);
85 static int appldata_interval = APPLDATA_CPU_INTERVAL;
86 static int appldata_timer_active;
87 
88 /*
89  * Work queue
90  */
91 static struct workqueue_struct *appldata_wq;
92 static void appldata_work_fn(struct work_struct *work);
93 static DECLARE_WORK(appldata_work, appldata_work_fn);
94 
95 
96 /*
97  * Ops list
98  */
99 static DEFINE_SPINLOCK(appldata_ops_lock);
100 static LIST_HEAD(appldata_ops_list);
101 
102 
103 /*************************** timer, work, DIAG *******************************/
104 /*
105  * appldata_timer_function()
106  *
107  * schedule work and reschedule timer
108  */
109 static void appldata_timer_function(unsigned long data)
110 {
111 	P_DEBUG("   -= Timer =-\n");
112 	P_DEBUG("CPU: %i, expire_count: %i\n", smp_processor_id(),
113 		atomic_read(&appldata_expire_count));
114 	if (atomic_dec_and_test(&appldata_expire_count)) {
115 		atomic_set(&appldata_expire_count, num_online_cpus());
116 		queue_work(appldata_wq, (struct work_struct *) data);
117 	}
118 }
119 
120 /*
121  * appldata_work_fn()
122  *
123  * call data gathering function for each (active) module
124  */
125 static void appldata_work_fn(struct work_struct *work)
126 {
127 	struct list_head *lh;
128 	struct appldata_ops *ops;
129 	int i;
130 
131 	P_DEBUG("  -= Work Queue =-\n");
132 	i = 0;
133 	spin_lock(&appldata_ops_lock);
134 	list_for_each(lh, &appldata_ops_list) {
135 		ops = list_entry(lh, struct appldata_ops, list);
136 		P_DEBUG("list_for_each loop: %i) active = %u, name = %s\n",
137 			++i, ops->active, ops->name);
138 		if (ops->active == 1) {
139 			ops->callback(ops->data);
140 		}
141 	}
142 	spin_unlock(&appldata_ops_lock);
143 }
144 
145 /*
146  * appldata_diag()
147  *
148  * prepare parameter list, issue DIAG 0xDC
149  */
150 int appldata_diag(char record_nr, u16 function, unsigned long buffer,
151 			u16 length, char *mod_lvl)
152 {
153 	struct appldata_product_id id = {
154 		.prod_nr    = {0xD3, 0xC9, 0xD5, 0xE4,
155 			       0xE7, 0xD2, 0xD9},	/* "LINUXKR" */
156 		.prod_fn    = 0xD5D3,			/* "NL" */
157 		.version_nr = 0xF2F6,			/* "26" */
158 		.release_nr = 0xF0F1,			/* "01" */
159 	};
160 
161 	id.record_nr = record_nr;
162 	id.mod_lvl = (mod_lvl[0]) << 8 | mod_lvl[1];
163 	return appldata_asm(&id, function, (void *) buffer, length);
164 }
165 /************************ timer, work, DIAG <END> ****************************/
166 
167 
168 /****************************** /proc stuff **********************************/
169 
170 /*
171  * appldata_mod_vtimer_wrap()
172  *
173  * wrapper function for mod_virt_timer(), because smp_call_function_single()
174  * accepts only one parameter.
175  */
176 static void __appldata_mod_vtimer_wrap(void *p) {
177 	struct {
178 		struct vtimer_list *timer;
179 		u64    expires;
180 	} *args = p;
181 	mod_virt_timer(args->timer, args->expires);
182 }
183 
184 #define APPLDATA_ADD_TIMER	0
185 #define APPLDATA_DEL_TIMER	1
186 #define APPLDATA_MOD_TIMER	2
187 
188 /*
189  * __appldata_vtimer_setup()
190  *
191  * Add, delete or modify virtual timers on all online cpus.
192  * The caller needs to get the appldata_timer_lock spinlock.
193  */
194 static void
195 __appldata_vtimer_setup(int cmd)
196 {
197 	u64 per_cpu_interval;
198 	int i;
199 
200 	switch (cmd) {
201 	case APPLDATA_ADD_TIMER:
202 		if (appldata_timer_active)
203 			break;
204 		per_cpu_interval = (u64) (appldata_interval*1000 /
205 					  num_online_cpus()) * TOD_MICRO;
206 		for_each_online_cpu(i) {
207 			per_cpu(appldata_timer, i).expires = per_cpu_interval;
208 			smp_call_function_single(i, add_virt_timer_periodic,
209 						 &per_cpu(appldata_timer, i),
210 						 0, 1);
211 		}
212 		appldata_timer_active = 1;
213 		P_INFO("Monitoring timer started.\n");
214 		break;
215 	case APPLDATA_DEL_TIMER:
216 		for_each_online_cpu(i)
217 			del_virt_timer(&per_cpu(appldata_timer, i));
218 		if (!appldata_timer_active)
219 			break;
220 		appldata_timer_active = 0;
221 		atomic_set(&appldata_expire_count, num_online_cpus());
222 		P_INFO("Monitoring timer stopped.\n");
223 		break;
224 	case APPLDATA_MOD_TIMER:
225 		per_cpu_interval = (u64) (appldata_interval*1000 /
226 					  num_online_cpus()) * TOD_MICRO;
227 		if (!appldata_timer_active)
228 			break;
229 		for_each_online_cpu(i) {
230 			struct {
231 				struct vtimer_list *timer;
232 				u64    expires;
233 			} args;
234 			args.timer = &per_cpu(appldata_timer, i);
235 			args.expires = per_cpu_interval;
236 			smp_call_function_single(i, __appldata_mod_vtimer_wrap,
237 						 &args, 0, 1);
238 		}
239 	}
240 }
241 
242 /*
243  * appldata_timer_handler()
244  *
245  * Start/Stop timer, show status of timer (0 = not active, 1 = active)
246  */
247 static int
248 appldata_timer_handler(ctl_table *ctl, int write, struct file *filp,
249 			   void __user *buffer, size_t *lenp, loff_t *ppos)
250 {
251 	int len;
252 	char buf[2];
253 
254 	if (!*lenp || *ppos) {
255 		*lenp = 0;
256 		return 0;
257 	}
258 	if (!write) {
259 		len = sprintf(buf, appldata_timer_active ? "1\n" : "0\n");
260 		if (len > *lenp)
261 			len = *lenp;
262 		if (copy_to_user(buffer, buf, len))
263 			return -EFAULT;
264 		goto out;
265 	}
266 	len = *lenp;
267 	if (copy_from_user(buf, buffer, len > sizeof(buf) ? sizeof(buf) : len))
268 		return -EFAULT;
269 	spin_lock(&appldata_timer_lock);
270 	if (buf[0] == '1')
271 		__appldata_vtimer_setup(APPLDATA_ADD_TIMER);
272 	else if (buf[0] == '0')
273 		__appldata_vtimer_setup(APPLDATA_DEL_TIMER);
274 	spin_unlock(&appldata_timer_lock);
275 out:
276 	*lenp = len;
277 	*ppos += len;
278 	return 0;
279 }
280 
281 /*
282  * appldata_interval_handler()
283  *
284  * Set (CPU) timer interval for collection of data (in milliseconds), show
285  * current timer interval.
286  */
287 static int
288 appldata_interval_handler(ctl_table *ctl, int write, struct file *filp,
289 			   void __user *buffer, size_t *lenp, loff_t *ppos)
290 {
291 	int len, interval;
292 	char buf[16];
293 
294 	if (!*lenp || *ppos) {
295 		*lenp = 0;
296 		return 0;
297 	}
298 	if (!write) {
299 		len = sprintf(buf, "%i\n", appldata_interval);
300 		if (len > *lenp)
301 			len = *lenp;
302 		if (copy_to_user(buffer, buf, len))
303 			return -EFAULT;
304 		goto out;
305 	}
306 	len = *lenp;
307 	if (copy_from_user(buf, buffer, len > sizeof(buf) ? sizeof(buf) : len)) {
308 		return -EFAULT;
309 	}
310 	interval = 0;
311 	sscanf(buf, "%i", &interval);
312 	if (interval <= 0) {
313 		P_ERROR("Timer CPU interval has to be > 0!\n");
314 		return -EINVAL;
315 	}
316 
317 	spin_lock(&appldata_timer_lock);
318 	appldata_interval = interval;
319 	__appldata_vtimer_setup(APPLDATA_MOD_TIMER);
320 	spin_unlock(&appldata_timer_lock);
321 
322 	P_INFO("Monitoring CPU interval set to %u milliseconds.\n",
323 		 interval);
324 out:
325 	*lenp = len;
326 	*ppos += len;
327 	return 0;
328 }
329 
330 /*
331  * appldata_generic_handler()
332  *
333  * Generic start/stop monitoring and DIAG, show status of
334  * monitoring (0 = not in process, 1 = in process)
335  */
336 static int
337 appldata_generic_handler(ctl_table *ctl, int write, struct file *filp,
338 			   void __user *buffer, size_t *lenp, loff_t *ppos)
339 {
340 	struct appldata_ops *ops = NULL, *tmp_ops;
341 	int rc, len, found;
342 	char buf[2];
343 	struct list_head *lh;
344 
345 	found = 0;
346 	spin_lock(&appldata_ops_lock);
347 	list_for_each(lh, &appldata_ops_list) {
348 		tmp_ops = list_entry(lh, struct appldata_ops, list);
349 		if (&tmp_ops->ctl_table[2] == ctl) {
350 			found = 1;
351 		}
352 	}
353 	if (!found) {
354 		spin_unlock(&appldata_ops_lock);
355 		return -ENODEV;
356 	}
357 	ops = ctl->data;
358 	if (!try_module_get(ops->owner)) {	// protect this function
359 		spin_unlock(&appldata_ops_lock);
360 		return -ENODEV;
361 	}
362 	spin_unlock(&appldata_ops_lock);
363 
364 	if (!*lenp || *ppos) {
365 		*lenp = 0;
366 		module_put(ops->owner);
367 		return 0;
368 	}
369 	if (!write) {
370 		len = sprintf(buf, ops->active ? "1\n" : "0\n");
371 		if (len > *lenp)
372 			len = *lenp;
373 		if (copy_to_user(buffer, buf, len)) {
374 			module_put(ops->owner);
375 			return -EFAULT;
376 		}
377 		goto out;
378 	}
379 	len = *lenp;
380 	if (copy_from_user(buf, buffer,
381 			   len > sizeof(buf) ? sizeof(buf) : len)) {
382 		module_put(ops->owner);
383 		return -EFAULT;
384 	}
385 
386 	spin_lock(&appldata_ops_lock);
387 	if ((buf[0] == '1') && (ops->active == 0)) {
388 		// protect work queue callback
389 		if (!try_module_get(ops->owner)) {
390 			spin_unlock(&appldata_ops_lock);
391 			module_put(ops->owner);
392 			return -ENODEV;
393 		}
394 		ops->callback(ops->data);	// init record
395 		rc = appldata_diag(ops->record_nr,
396 					APPLDATA_START_INTERVAL_REC,
397 					(unsigned long) ops->data, ops->size,
398 					ops->mod_lvl);
399 		if (rc != 0) {
400 			P_ERROR("START DIAG 0xDC for %s failed, "
401 				"return code: %d\n", ops->name, rc);
402 			module_put(ops->owner);
403 		} else {
404 			P_INFO("Monitoring %s data enabled, "
405 				"DIAG 0xDC started.\n", ops->name);
406 			ops->active = 1;
407 		}
408 	} else if ((buf[0] == '0') && (ops->active == 1)) {
409 		ops->active = 0;
410 		rc = appldata_diag(ops->record_nr, APPLDATA_STOP_REC,
411 				(unsigned long) ops->data, ops->size,
412 				ops->mod_lvl);
413 		if (rc != 0) {
414 			P_ERROR("STOP DIAG 0xDC for %s failed, "
415 				"return code: %d\n", ops->name, rc);
416 		} else {
417 			P_INFO("Monitoring %s data disabled, "
418 				"DIAG 0xDC stopped.\n", ops->name);
419 		}
420 		module_put(ops->owner);
421 	}
422 	spin_unlock(&appldata_ops_lock);
423 out:
424 	*lenp = len;
425 	*ppos += len;
426 	module_put(ops->owner);
427 	return 0;
428 }
429 
430 /*************************** /proc stuff <END> *******************************/
431 
432 
433 /************************* module-ops management *****************************/
434 /*
435  * appldata_register_ops()
436  *
437  * update ops list, register /proc/sys entries
438  */
439 int appldata_register_ops(struct appldata_ops *ops)
440 {
441 	if ((ops->size > APPLDATA_MAX_REC_SIZE) || (ops->size < 0))
442 		return -EINVAL;
443 
444 	ops->ctl_table = kzalloc(4 * sizeof(struct ctl_table), GFP_KERNEL);
445 	if (!ops->ctl_table)
446 		return -ENOMEM;
447 
448 	spin_lock(&appldata_ops_lock);
449 	list_add(&ops->list, &appldata_ops_list);
450 	spin_unlock(&appldata_ops_lock);
451 
452 	ops->ctl_table[0].procname = appldata_proc_name;
453 	ops->ctl_table[0].maxlen   = 0;
454 	ops->ctl_table[0].mode     = S_IRUGO | S_IXUGO;
455 	ops->ctl_table[0].child    = &ops->ctl_table[2];
456 
457 	ops->ctl_table[2].procname = ops->name;
458 	ops->ctl_table[2].mode     = S_IRUGO | S_IWUSR;
459 	ops->ctl_table[2].proc_handler = appldata_generic_handler;
460 	ops->ctl_table[2].data = ops;
461 
462 	ops->sysctl_header = register_sysctl_table(ops->ctl_table);
463 	if (!ops->sysctl_header)
464 		goto out;
465 	P_INFO("%s-ops registered!\n", ops->name);
466 	return 0;
467 out:
468 	spin_lock(&appldata_ops_lock);
469 	list_del(&ops->list);
470 	spin_unlock(&appldata_ops_lock);
471 	kfree(ops->ctl_table);
472 	return -ENOMEM;
473 }
474 
475 /*
476  * appldata_unregister_ops()
477  *
478  * update ops list, unregister /proc entries, stop DIAG if necessary
479  */
480 void appldata_unregister_ops(struct appldata_ops *ops)
481 {
482 	spin_lock(&appldata_ops_lock);
483 	list_del(&ops->list);
484 	spin_unlock(&appldata_ops_lock);
485 	unregister_sysctl_table(ops->sysctl_header);
486 	kfree(ops->ctl_table);
487 	P_INFO("%s-ops unregistered!\n", ops->name);
488 }
489 /********************** module-ops management <END> **************************/
490 
491 
492 /******************************* init / exit *********************************/
493 
494 static void __cpuinit appldata_online_cpu(int cpu)
495 {
496 	init_virt_timer(&per_cpu(appldata_timer, cpu));
497 	per_cpu(appldata_timer, cpu).function = appldata_timer_function;
498 	per_cpu(appldata_timer, cpu).data = (unsigned long)
499 		&appldata_work;
500 	atomic_inc(&appldata_expire_count);
501 	spin_lock(&appldata_timer_lock);
502 	__appldata_vtimer_setup(APPLDATA_MOD_TIMER);
503 	spin_unlock(&appldata_timer_lock);
504 }
505 
506 static void __cpuinit appldata_offline_cpu(int cpu)
507 {
508 	del_virt_timer(&per_cpu(appldata_timer, cpu));
509 	if (atomic_dec_and_test(&appldata_expire_count)) {
510 		atomic_set(&appldata_expire_count, num_online_cpus());
511 		queue_work(appldata_wq, &appldata_work);
512 	}
513 	spin_lock(&appldata_timer_lock);
514 	__appldata_vtimer_setup(APPLDATA_MOD_TIMER);
515 	spin_unlock(&appldata_timer_lock);
516 }
517 
518 static int __cpuinit appldata_cpu_notify(struct notifier_block *self,
519 					 unsigned long action,
520 					 void *hcpu)
521 {
522 	switch (action) {
523 	case CPU_ONLINE:
524 	case CPU_ONLINE_FROZEN:
525 		appldata_online_cpu((long) hcpu);
526 		break;
527 	case CPU_DEAD:
528 	case CPU_DEAD_FROZEN:
529 		appldata_offline_cpu((long) hcpu);
530 		break;
531 	default:
532 		break;
533 	}
534 	return NOTIFY_OK;
535 }
536 
537 static struct notifier_block __cpuinitdata appldata_nb = {
538 	.notifier_call = appldata_cpu_notify,
539 };
540 
541 /*
542  * appldata_init()
543  *
544  * init timer, register /proc entries
545  */
546 static int __init appldata_init(void)
547 {
548 	int i;
549 
550 	P_DEBUG("sizeof(parameter_list) = %lu\n",
551 		sizeof(struct appldata_parameter_list));
552 
553 	appldata_wq = create_singlethread_workqueue("appldata");
554 	if (!appldata_wq) {
555 		P_ERROR("Could not create work queue\n");
556 		return -ENOMEM;
557 	}
558 
559 	for_each_online_cpu(i)
560 		appldata_online_cpu(i);
561 
562 	/* Register cpu hotplug notifier */
563 	register_hotcpu_notifier(&appldata_nb);
564 
565 	appldata_sysctl_header = register_sysctl_table(appldata_dir_table);
566 
567 	P_DEBUG("Base interface initialized.\n");
568 	return 0;
569 }
570 
571 __initcall(appldata_init);
572 
573 /**************************** init / exit <END> ******************************/
574 
575 EXPORT_SYMBOL_GPL(appldata_register_ops);
576 EXPORT_SYMBOL_GPL(appldata_unregister_ops);
577 EXPORT_SYMBOL_GPL(appldata_diag);
578 
579 EXPORT_SYMBOL_GPL(si_swapinfo);
580 EXPORT_SYMBOL_GPL(nr_threads);
581 EXPORT_SYMBOL_GPL(nr_running);
582 EXPORT_SYMBOL_GPL(nr_iowait);
583