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/workqueue.h>
24 #include <asm/uaccess.h>
28 #include <asm/nvram.h>
29 #include <asm/atomic.h>
30 #include <asm/machdep.h>
33 static DEFINE_SPINLOCK(rtasd_log_lock);
35 static DECLARE_WAIT_QUEUE_HEAD(rtas_log_wait);
37 static char *rtas_log_buf;
38 static unsigned long rtas_log_start;
39 static unsigned long rtas_log_size;
41 static int surveillance_timeout = -1;
43 static unsigned int rtas_error_log_max;
44 static unsigned int rtas_error_log_buffer_max;
46 /* RTAS service tokens */
47 static unsigned int event_scan;
48 static unsigned int rtas_event_scan_rate;
50 static int full_rtas_msgs = 0;
52 /* Stop logging to nvram after first fatal error */
53 static int logging_enabled; /* Until we initialize everything,
54 * make sure we don't try logging
56 static int error_log_cnt;
59 * Since we use 32 bit RTAS, the physical address of this must be below
60 * 4G or else bad things happen. Allocate this in the kernel data and
63 static unsigned char logdata[RTAS_ERROR_LOG_MAX];
65 static char *rtas_type[] = {
66 "Unknown", "Retry", "TCE Error", "Internal Device Failure",
67 "Timeout", "Data Parity", "Address Parity", "Cache Parity",
68 "Address Invalid", "ECC Uncorrected", "ECC Corrupted",
71 static char *rtas_event_type(int type)
73 if ((type > 0) && (type < 11))
74 return rtas_type[type];
79 case RTAS_TYPE_PLATFORM:
80 return "Platform Error";
84 return "Platform Information Event";
85 case RTAS_TYPE_DEALLOC:
86 return "Resource Deallocation Event";
88 return "Dump Notification Event";
94 /* To see this info, grep RTAS /var/log/messages and each entry
95 * will be collected together with obvious begin/end.
96 * There will be a unique identifier on the begin and end lines.
97 * This will persist across reboots.
99 * format of error logs returned from RTAS:
100 * bytes (size) : contents
101 * --------------------------------------------------------
102 * 0-7 (8) : rtas_error_log
103 * 8-47 (40) : extended info
104 * 48-51 (4) : vendor id
105 * 52-1023 (vendor specific) : location code and debug data
107 static void printk_log_rtas(char *buf, int len)
113 char * str = "RTAS event";
115 if (full_rtas_msgs) {
116 printk(RTAS_DEBUG "%d -------- %s begin --------\n",
120 * Print perline bytes on each line, each line will start
121 * with RTAS and a changing number, so syslogd will
122 * print lines that are otherwise the same. Separate every
123 * 4 bytes with a space.
125 for (i = 0; i < len; i++) {
128 memset(buffer, 0, sizeof(buffer));
129 n = sprintf(buffer, "RTAS %d:", i/perline);
133 n += sprintf(buffer+n, " ");
135 n += sprintf(buffer+n, "%02x", (unsigned char)buf[i]);
137 if (j == (perline-1))
138 printk(KERN_DEBUG "%s\n", buffer);
140 if ((i % perline) != 0)
141 printk(KERN_DEBUG "%s\n", buffer);
143 printk(RTAS_DEBUG "%d -------- %s end ----------\n",
146 struct rtas_error_log *errlog = (struct rtas_error_log *)buf;
148 printk(RTAS_DEBUG "event: %d, Type: %s, Severity: %d\n",
149 error_log_cnt, rtas_event_type(errlog->type),
154 static int log_rtas_len(char * buf)
157 struct rtas_error_log *err;
159 /* rtas fixed header */
161 err = (struct rtas_error_log *)buf;
162 if (err->extended_log_length) {
164 /* extended header */
165 len += err->extended_log_length;
168 if (rtas_error_log_max == 0)
169 rtas_error_log_max = rtas_get_error_log_max();
171 if (len > rtas_error_log_max)
172 len = rtas_error_log_max;
178 * First write to nvram, if fatal error, that is the only
179 * place we log the info. The error will be picked up
180 * on the next reboot by rtasd. If not fatal, run the
181 * method for the type of error. Currently, only RTAS
182 * errors have methods implemented, but in the future
183 * there might be a need to store data in nvram before a
186 * XXX We write to nvram periodically, to indicate error has
187 * been written and sync'd, but there is a possibility
188 * that if we don't shutdown correctly, a duplicate error
189 * record will be created on next reboot.
191 void pSeries_log_error(char *buf, unsigned int err_type, int fatal)
193 unsigned long offset;
197 pr_debug("rtasd: logging event\n");
201 spin_lock_irqsave(&rtasd_log_lock, s);
203 /* get length and increase count */
204 switch (err_type & ERR_TYPE_MASK) {
205 case ERR_TYPE_RTAS_LOG:
206 len = log_rtas_len(buf);
207 if (!(err_type & ERR_FLAG_BOOT))
210 case ERR_TYPE_KERNEL_PANIC:
212 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
213 spin_unlock_irqrestore(&rtasd_log_lock, s);
218 /* Write error to NVRAM */
219 if (logging_enabled && !(err_type & ERR_FLAG_BOOT))
220 nvram_write_error_log(buf, len, err_type, error_log_cnt);
221 #endif /* CONFIG_PPC64 */
224 * rtas errors can occur during boot, and we do want to capture
225 * those somewhere, even if nvram isn't ready (why not?), and even
226 * if rtasd isn't ready. Put them into the boot log, at least.
228 if ((err_type & ERR_TYPE_MASK) == ERR_TYPE_RTAS_LOG)
229 printk_log_rtas(buf, len);
231 /* Check to see if we need to or have stopped logging */
232 if (fatal || !logging_enabled) {
234 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
235 spin_unlock_irqrestore(&rtasd_log_lock, s);
239 /* call type specific method for error */
240 switch (err_type & ERR_TYPE_MASK) {
241 case ERR_TYPE_RTAS_LOG:
242 offset = rtas_error_log_buffer_max *
243 ((rtas_log_start+rtas_log_size) & LOG_NUMBER_MASK);
245 /* First copy over sequence number */
246 memcpy(&rtas_log_buf[offset], (void *) &error_log_cnt, sizeof(int));
248 /* Second copy over error log data */
249 offset += sizeof(int);
250 memcpy(&rtas_log_buf[offset], buf, len);
252 if (rtas_log_size < LOG_NUMBER)
257 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
258 spin_unlock_irqrestore(&rtasd_log_lock, s);
259 wake_up_interruptible(&rtas_log_wait);
261 case ERR_TYPE_KERNEL_PANIC:
263 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
264 spin_unlock_irqrestore(&rtasd_log_lock, s);
270 static int rtas_log_open(struct inode * inode, struct file * file)
275 static int rtas_log_release(struct inode * inode, struct file * file)
280 /* This will check if all events are logged, if they are then, we
281 * know that we can safely clear the events in NVRAM.
282 * Next we'll sit and wait for something else to log.
284 static ssize_t rtas_log_read(struct file * file, char __user * buf,
285 size_t count, loff_t *ppos)
290 unsigned long offset;
292 if (!buf || count < rtas_error_log_buffer_max)
295 count = rtas_error_log_buffer_max;
297 if (!access_ok(VERIFY_WRITE, buf, count))
300 tmp = kmalloc(count, GFP_KERNEL);
304 spin_lock_irqsave(&rtasd_log_lock, s);
306 /* if it's 0, then we know we got the last one (the one in NVRAM) */
307 while (rtas_log_size == 0) {
308 if (file->f_flags & O_NONBLOCK) {
309 spin_unlock_irqrestore(&rtasd_log_lock, s);
314 if (!logging_enabled) {
315 spin_unlock_irqrestore(&rtasd_log_lock, s);
320 nvram_clear_error_log();
321 #endif /* CONFIG_PPC64 */
323 spin_unlock_irqrestore(&rtasd_log_lock, s);
324 error = wait_event_interruptible(rtas_log_wait, rtas_log_size);
327 spin_lock_irqsave(&rtasd_log_lock, s);
330 offset = rtas_error_log_buffer_max * (rtas_log_start & LOG_NUMBER_MASK);
331 memcpy(tmp, &rtas_log_buf[offset], count);
335 spin_unlock_irqrestore(&rtasd_log_lock, s);
337 error = copy_to_user(buf, tmp, count) ? -EFAULT : count;
343 static unsigned int rtas_log_poll(struct file *file, poll_table * wait)
345 poll_wait(file, &rtas_log_wait, wait);
347 return POLLIN | POLLRDNORM;
351 static const struct file_operations proc_rtas_log_operations = {
352 .read = rtas_log_read,
353 .poll = rtas_log_poll,
354 .open = rtas_log_open,
355 .release = rtas_log_release,
358 static int enable_surveillance(int timeout)
362 error = rtas_set_indicator(SURVEILLANCE_TOKEN, 0, timeout);
367 if (error == -EINVAL) {
368 printk(KERN_DEBUG "rtasd: surveillance not supported\n");
372 printk(KERN_ERR "rtasd: could not update surveillance\n");
376 static void do_event_scan(void)
380 memset(logdata, 0, rtas_error_log_max);
381 error = rtas_call(event_scan, 4, 1, NULL,
382 RTAS_EVENT_SCAN_ALL_EVENTS, 0,
383 __pa(logdata), rtas_error_log_max);
385 printk(KERN_ERR "event-scan failed\n");
390 pSeries_log_error(logdata, ERR_TYPE_RTAS_LOG, 0);
395 static void rtas_event_scan(struct work_struct *w);
396 DECLARE_DELAYED_WORK(event_scan_work, rtas_event_scan);
399 * Delay should be at least one second since some machines have problems if
400 * we call event-scan too quickly.
402 static unsigned long event_scan_delay = 1*HZ;
403 static int first_pass = 1;
405 static void rtas_event_scan(struct work_struct *w)
413 cpu = next_cpu(smp_processor_id(), cpu_online_map);
414 if (cpu == NR_CPUS) {
415 cpu = first_cpu(cpu_online_map);
419 event_scan_delay = 30*HZ/rtas_event_scan_rate;
421 if (surveillance_timeout != -1) {
422 pr_debug("rtasd: enabling surveillance\n");
423 enable_surveillance(surveillance_timeout);
424 pr_debug("rtasd: surveillance enabled\n");
429 schedule_delayed_work_on(cpu, &event_scan_work,
430 __round_jiffies_relative(event_scan_delay, cpu));
436 static void retreive_nvram_error_log(void)
438 unsigned int err_type ;
441 /* See if we have any error stored in NVRAM */
442 memset(logdata, 0, rtas_error_log_max);
443 rc = nvram_read_error_log(logdata, rtas_error_log_max,
444 &err_type, &error_log_cnt);
445 /* We can use rtas_log_buf now */
448 if (err_type != ERR_FLAG_ALREADY_LOGGED) {
449 pSeries_log_error(logdata, err_type | ERR_FLAG_BOOT, 0);
453 #else /* CONFIG_PPC64 */
454 static void retreive_nvram_error_log(void)
457 #endif /* CONFIG_PPC64 */
459 static void start_event_scan(void)
461 printk(KERN_DEBUG "RTAS daemon started\n");
462 pr_debug("rtasd: will sleep for %d milliseconds\n",
463 (30000 / rtas_event_scan_rate));
465 /* Retreive errors from nvram if any */
466 retreive_nvram_error_log();
468 schedule_delayed_work_on(first_cpu(cpu_online_map), &event_scan_work,
472 static int __init rtas_init(void)
474 struct proc_dir_entry *entry;
476 if (!machine_is(pseries) && !machine_is(chrp))
480 event_scan = rtas_token("event-scan");
481 if (event_scan == RTAS_UNKNOWN_SERVICE) {
482 printk(KERN_INFO "rtasd: No event-scan on system\n");
486 rtas_event_scan_rate = rtas_token("rtas-event-scan-rate");
487 if (rtas_event_scan_rate == RTAS_UNKNOWN_SERVICE) {
488 printk(KERN_ERR "rtasd: no rtas-event-scan-rate on system\n");
492 /* Make room for the sequence number */
493 rtas_error_log_max = rtas_get_error_log_max();
494 rtas_error_log_buffer_max = rtas_error_log_max + sizeof(int);
496 rtas_log_buf = vmalloc(rtas_error_log_buffer_max*LOG_NUMBER);
498 printk(KERN_ERR "rtasd: no memory\n");
502 entry = proc_create("powerpc/rtas/error_log", S_IRUSR, NULL,
503 &proc_rtas_log_operations);
505 printk(KERN_ERR "Failed to create error_log proc entry\n");
511 __initcall(rtas_init);
513 static int __init surveillance_setup(char *str)
517 /* We only do surveillance on pseries */
518 if (!machine_is(pseries))
521 if (get_option(&str,&i)) {
522 if (i >= 0 && i <= 255)
523 surveillance_timeout = i;
528 __setup("surveillance=", surveillance_setup);
530 static int __init rtasmsgs_setup(char *str)
532 if (strcmp(str, "on") == 0)
534 else if (strcmp(str, "off") == 0)
539 __setup("rtasmsgs=", rtasmsgs_setup);