[PATCH] ppc64: replace schedule_timeout() with msleep_interruptible()
[linux-2.6/linux-mips.git] / arch / ppc64 / kernel / rtasd.c
blobe26b0420b6ddea490c8a6fb8f81a0bd9ac351422
1 /*
2 * Copyright (C) 2001 Anton Blanchard <anton@au.ibm.com>, IBM
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
9 * Communication to userspace based on kernel/printk.c
12 #include <linux/types.h>
13 #include <linux/errno.h>
14 #include <linux/sched.h>
15 #include <linux/kernel.h>
16 #include <linux/poll.h>
17 #include <linux/proc_fs.h>
18 #include <linux/init.h>
19 #include <linux/vmalloc.h>
20 #include <linux/spinlock.h>
21 #include <linux/cpu.h>
22 #include <linux/delay.h>
24 #include <asm/uaccess.h>
25 #include <asm/io.h>
26 #include <asm/rtas.h>
27 #include <asm/prom.h>
28 #include <asm/nvram.h>
29 #include <asm/atomic.h>
30 #include <asm/systemcfg.h>
32 #if 0
33 #define DEBUG(A...) printk(KERN_ERR A)
34 #else
35 #define DEBUG(A...)
36 #endif
38 static DEFINE_SPINLOCK(rtasd_log_lock);
40 DECLARE_WAIT_QUEUE_HEAD(rtas_log_wait);
42 static char *rtas_log_buf;
43 static unsigned long rtas_log_start;
44 static unsigned long rtas_log_size;
46 static int surveillance_timeout = -1;
47 static unsigned int rtas_event_scan_rate;
48 static unsigned int rtas_error_log_max;
49 static unsigned int rtas_error_log_buffer_max;
51 static int full_rtas_msgs = 0;
53 extern int no_logging;
55 volatile int error_log_cnt = 0;
58 * Since we use 32 bit RTAS, the physical address of this must be below
59 * 4G or else bad things happen. Allocate this in the kernel data and
60 * make it big enough.
62 static unsigned char logdata[RTAS_ERROR_LOG_MAX];
64 static int get_eventscan_parms(void);
66 static char *rtas_type[] = {
67 "Unknown", "Retry", "TCE Error", "Internal Device Failure",
68 "Timeout", "Data Parity", "Address Parity", "Cache Parity",
69 "Address Invalid", "ECC Uncorrected", "ECC Corrupted",
72 static char *rtas_event_type(int type)
74 if ((type > 0) && (type < 11))
75 return rtas_type[type];
77 switch (type) {
78 case RTAS_TYPE_EPOW:
79 return "EPOW";
80 case RTAS_TYPE_PLATFORM:
81 return "Platform Error";
82 case RTAS_TYPE_IO:
83 return "I/O Event";
84 case RTAS_TYPE_INFO:
85 return "Platform Information Event";
86 case RTAS_TYPE_DEALLOC:
87 return "Resource Deallocation Event";
88 case RTAS_TYPE_DUMP:
89 return "Dump Notification Event";
92 return rtas_type[0];
95 /* To see this info, grep RTAS /var/log/messages and each entry
96 * will be collected together with obvious begin/end.
97 * There will be a unique identifier on the begin and end lines.
98 * This will persist across reboots.
100 * format of error logs returned from RTAS:
101 * bytes (size) : contents
102 * --------------------------------------------------------
103 * 0-7 (8) : rtas_error_log
104 * 8-47 (40) : extended info
105 * 48-51 (4) : vendor id
106 * 52-1023 (vendor specific) : location code and debug data
108 static void printk_log_rtas(char *buf, int len)
111 int i,j,n = 0;
112 int perline = 16;
113 char buffer[64];
114 char * str = "RTAS event";
116 if (full_rtas_msgs) {
117 printk(RTAS_DEBUG "%d -------- %s begin --------\n",
118 error_log_cnt, str);
121 * Print perline bytes on each line, each line will start
122 * with RTAS and a changing number, so syslogd will
123 * print lines that are otherwise the same. Separate every
124 * 4 bytes with a space.
126 for (i = 0; i < len; i++) {
127 j = i % perline;
128 if (j == 0) {
129 memset(buffer, 0, sizeof(buffer));
130 n = sprintf(buffer, "RTAS %d:", i/perline);
133 if ((i % 4) == 0)
134 n += sprintf(buffer+n, " ");
136 n += sprintf(buffer+n, "%02x", (unsigned char)buf[i]);
138 if (j == (perline-1))
139 printk(KERN_DEBUG "%s\n", buffer);
141 if ((i % perline) != 0)
142 printk(KERN_DEBUG "%s\n", buffer);
144 printk(RTAS_DEBUG "%d -------- %s end ----------\n",
145 error_log_cnt, str);
146 } else {
147 struct rtas_error_log *errlog = (struct rtas_error_log *)buf;
149 printk(RTAS_DEBUG "event: %d, Type: %s, Severity: %d\n",
150 error_log_cnt, rtas_event_type(errlog->type),
151 errlog->severity);
155 static int log_rtas_len(char * buf)
157 int len;
158 struct rtas_error_log *err;
160 /* rtas fixed header */
161 len = 8;
162 err = (struct rtas_error_log *)buf;
163 if (err->extended_log_length) {
165 /* extended header */
166 len += err->extended_log_length;
169 if (rtas_error_log_max == 0) {
170 get_eventscan_parms();
172 if (len > rtas_error_log_max)
173 len = rtas_error_log_max;
175 return len;
179 * First write to nvram, if fatal error, that is the only
180 * place we log the info. The error will be picked up
181 * on the next reboot by rtasd. If not fatal, run the
182 * method for the type of error. Currently, only RTAS
183 * errors have methods implemented, but in the future
184 * there might be a need to store data in nvram before a
185 * call to panic().
187 * XXX We write to nvram periodically, to indicate error has
188 * been written and sync'd, but there is a possibility
189 * that if we don't shutdown correctly, a duplicate error
190 * record will be created on next reboot.
192 void pSeries_log_error(char *buf, unsigned int err_type, int fatal)
194 unsigned long offset;
195 unsigned long s;
196 int len = 0;
198 DEBUG("logging event\n");
199 if (buf == NULL)
200 return;
202 spin_lock_irqsave(&rtasd_log_lock, s);
204 /* get length and increase count */
205 switch (err_type & ERR_TYPE_MASK) {
206 case ERR_TYPE_RTAS_LOG:
207 len = log_rtas_len(buf);
208 if (!(err_type & ERR_FLAG_BOOT))
209 error_log_cnt++;
210 break;
211 case ERR_TYPE_KERNEL_PANIC:
212 default:
213 spin_unlock_irqrestore(&rtasd_log_lock, s);
214 return;
217 /* Write error to NVRAM */
218 if (!no_logging && !(err_type & ERR_FLAG_BOOT))
219 nvram_write_error_log(buf, len, err_type);
222 * rtas errors can occur during boot, and we do want to capture
223 * those somewhere, even if nvram isn't ready (why not?), and even
224 * if rtasd isn't ready. Put them into the boot log, at least.
226 if ((err_type & ERR_TYPE_MASK) == ERR_TYPE_RTAS_LOG)
227 printk_log_rtas(buf, len);
229 /* Check to see if we need to or have stopped logging */
230 if (fatal || no_logging) {
231 no_logging = 1;
232 spin_unlock_irqrestore(&rtasd_log_lock, s);
233 return;
236 /* call type specific method for error */
237 switch (err_type & ERR_TYPE_MASK) {
238 case ERR_TYPE_RTAS_LOG:
239 offset = rtas_error_log_buffer_max *
240 ((rtas_log_start+rtas_log_size) & LOG_NUMBER_MASK);
242 /* First copy over sequence number */
243 memcpy(&rtas_log_buf[offset], (void *) &error_log_cnt, sizeof(int));
245 /* Second copy over error log data */
246 offset += sizeof(int);
247 memcpy(&rtas_log_buf[offset], buf, len);
249 if (rtas_log_size < LOG_NUMBER)
250 rtas_log_size += 1;
251 else
252 rtas_log_start += 1;
254 spin_unlock_irqrestore(&rtasd_log_lock, s);
255 wake_up_interruptible(&rtas_log_wait);
256 break;
257 case ERR_TYPE_KERNEL_PANIC:
258 default:
259 spin_unlock_irqrestore(&rtasd_log_lock, s);
260 return;
266 static int rtas_log_open(struct inode * inode, struct file * file)
268 return 0;
271 static int rtas_log_release(struct inode * inode, struct file * file)
273 return 0;
276 /* This will check if all events are logged, if they are then, we
277 * know that we can safely clear the events in NVRAM.
278 * Next we'll sit and wait for something else to log.
280 static ssize_t rtas_log_read(struct file * file, char __user * buf,
281 size_t count, loff_t *ppos)
283 int error;
284 char *tmp;
285 unsigned long s;
286 unsigned long offset;
288 if (!buf || count < rtas_error_log_buffer_max)
289 return -EINVAL;
291 count = rtas_error_log_buffer_max;
293 if (!access_ok(VERIFY_WRITE, buf, count))
294 return -EFAULT;
296 tmp = kmalloc(count, GFP_KERNEL);
297 if (!tmp)
298 return -ENOMEM;
301 spin_lock_irqsave(&rtasd_log_lock, s);
302 /* if it's 0, then we know we got the last one (the one in NVRAM) */
303 if (rtas_log_size == 0 && !no_logging)
304 nvram_clear_error_log();
305 spin_unlock_irqrestore(&rtasd_log_lock, s);
308 error = wait_event_interruptible(rtas_log_wait, rtas_log_size);
309 if (error)
310 goto out;
312 spin_lock_irqsave(&rtasd_log_lock, s);
313 offset = rtas_error_log_buffer_max * (rtas_log_start & LOG_NUMBER_MASK);
314 memcpy(tmp, &rtas_log_buf[offset], count);
316 rtas_log_start += 1;
317 rtas_log_size -= 1;
318 spin_unlock_irqrestore(&rtasd_log_lock, s);
320 error = copy_to_user(buf, tmp, count) ? -EFAULT : count;
321 out:
322 kfree(tmp);
323 return error;
326 static unsigned int rtas_log_poll(struct file *file, poll_table * wait)
328 poll_wait(file, &rtas_log_wait, wait);
329 if (rtas_log_size)
330 return POLLIN | POLLRDNORM;
331 return 0;
334 struct file_operations proc_rtas_log_operations = {
335 .read = rtas_log_read,
336 .poll = rtas_log_poll,
337 .open = rtas_log_open,
338 .release = rtas_log_release,
341 static int enable_surveillance(int timeout)
343 int error;
345 error = rtas_set_indicator(SURVEILLANCE_TOKEN, 0, timeout);
347 if (error == 0)
348 return 0;
350 if (error == -EINVAL) {
351 printk(KERN_INFO "rtasd: surveillance not supported\n");
352 return 0;
355 printk(KERN_ERR "rtasd: could not update surveillance\n");
356 return -1;
359 static int get_eventscan_parms(void)
361 struct device_node *node;
362 int *ip;
364 node = of_find_node_by_path("/rtas");
366 ip = (int *)get_property(node, "rtas-event-scan-rate", NULL);
367 if (ip == NULL) {
368 printk(KERN_ERR "rtasd: no rtas-event-scan-rate\n");
369 of_node_put(node);
370 return -1;
372 rtas_event_scan_rate = *ip;
373 DEBUG("rtas-event-scan-rate %d\n", rtas_event_scan_rate);
375 /* Make room for the sequence number */
376 rtas_error_log_max = rtas_get_error_log_max();
377 rtas_error_log_buffer_max = rtas_error_log_max + sizeof(int);
379 of_node_put(node);
381 return 0;
384 static void do_event_scan(int event_scan)
386 int error;
387 do {
388 memset(logdata, 0, rtas_error_log_max);
389 error = rtas_call(event_scan, 4, 1, NULL,
390 RTAS_EVENT_SCAN_ALL_EVENTS, 0,
391 __pa(logdata), rtas_error_log_max);
392 if (error == -1) {
393 printk(KERN_ERR "event-scan failed\n");
394 break;
397 if (error == 0)
398 pSeries_log_error(logdata, ERR_TYPE_RTAS_LOG, 0);
400 } while(error == 0);
403 static void do_event_scan_all_cpus(long delay)
405 int cpu;
407 lock_cpu_hotplug();
408 cpu = first_cpu(cpu_online_map);
409 for (;;) {
410 set_cpus_allowed(current, cpumask_of_cpu(cpu));
411 do_event_scan(rtas_token("event-scan"));
412 set_cpus_allowed(current, CPU_MASK_ALL);
414 /* Drop hotplug lock, and sleep for the specified delay */
415 unlock_cpu_hotplug();
416 msleep_interruptible(delay);
417 lock_cpu_hotplug();
419 cpu = next_cpu(cpu, cpu_online_map);
420 if (cpu == NR_CPUS)
421 break;
423 unlock_cpu_hotplug();
426 static int rtasd(void *unused)
428 unsigned int err_type;
429 int event_scan = rtas_token("event-scan");
430 int rc;
432 daemonize("rtasd");
434 if (event_scan == RTAS_UNKNOWN_SERVICE || get_eventscan_parms() == -1)
435 goto error;
437 rtas_log_buf = vmalloc(rtas_error_log_buffer_max*LOG_NUMBER);
438 if (!rtas_log_buf) {
439 printk(KERN_ERR "rtasd: no memory\n");
440 goto error;
443 printk(KERN_INFO "RTAS daemon started\n");
445 DEBUG("will sleep for %d milliseconds\n", (30000/rtas_event_scan_rate));
447 /* See if we have any error stored in NVRAM */
448 memset(logdata, 0, rtas_error_log_max);
450 rc = nvram_read_error_log(logdata, rtas_error_log_max, &err_type);
452 /* We can use rtas_log_buf now */
453 no_logging = 0;
455 if (!rc) {
456 if (err_type != ERR_FLAG_ALREADY_LOGGED) {
457 pSeries_log_error(logdata, err_type | ERR_FLAG_BOOT, 0);
461 /* First pass. */
462 do_event_scan_all_cpus(1000);
464 if (surveillance_timeout != -1) {
465 DEBUG("enabling surveillance\n");
466 enable_surveillance(surveillance_timeout);
467 DEBUG("surveillance enabled\n");
470 /* Delay should be at least one second since some
471 * machines have problems if we call event-scan too
472 * quickly. */
473 for (;;)
474 do_event_scan_all_cpus(30000/rtas_event_scan_rate);
476 error:
477 /* Should delete proc entries */
478 return -EINVAL;
481 static int __init rtas_init(void)
483 struct proc_dir_entry *entry;
485 /* No RTAS, only warn if we are on a pSeries box */
486 if (rtas_token("event-scan") == RTAS_UNKNOWN_SERVICE) {
487 if (systemcfg->platform & PLATFORM_PSERIES)
488 printk(KERN_INFO "rtasd: no event-scan on system\n");
489 return 1;
492 entry = create_proc_entry("ppc64/rtas/error_log", S_IRUSR, NULL);
493 if (entry)
494 entry->proc_fops = &proc_rtas_log_operations;
495 else
496 printk(KERN_ERR "Failed to create error_log proc entry\n");
498 if (kernel_thread(rtasd, NULL, CLONE_FS) < 0)
499 printk(KERN_ERR "Failed to start RTAS daemon\n");
501 return 0;
504 static int __init surveillance_setup(char *str)
506 int i;
508 if (get_option(&str,&i)) {
509 if (i >= 0 && i <= 255)
510 surveillance_timeout = i;
513 return 1;
516 static int __init rtasmsgs_setup(char *str)
518 if (strcmp(str, "on") == 0)
519 full_rtas_msgs = 1;
520 else if (strcmp(str, "off") == 0)
521 full_rtas_msgs = 0;
523 return 1;
525 __initcall(rtas_init);
526 __setup("surveillance=", surveillance_setup);
527 __setup("rtasmsgs=", rtasmsgs_setup);