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>
23 #include <asm/uaccess.h>
27 #include <asm/nvram.h>
28 #include <asm/atomic.h>
31 #define DEBUG(A...) printk(KERN_ERR A)
36 static spinlock_t log_lock = SPIN_LOCK_UNLOCKED;
38 DECLARE_WAIT_QUEUE_HEAD(rtas_log_wait);
40 static char *rtas_log_buf;
41 static unsigned long rtas_log_start;
42 static unsigned long rtas_log_size;
44 static int surveillance_timeout = -1;
45 static unsigned int rtas_event_scan_rate;
46 static unsigned int rtas_error_log_max;
47 static unsigned int rtas_error_log_buffer_max;
49 extern volatile int no_more_logging;
51 volatile int error_log_cnt = 0;
54 * Since we use 32 bit RTAS, the physical address of this must be below
55 * 4G or else bad things happen. Allocate this in the kernel data and
58 static unsigned char logdata[RTAS_ERROR_LOG_MAX];
60 /* To see this info, grep RTAS /var/log/messages and each entry
61 * will be collected together with obvious begin/end.
62 * There will be a unique identifier on the begin and end lines.
63 * This will persist across reboots.
65 * format of error logs returned from RTAS:
66 * bytes (size) : contents
67 * --------------------------------------------------------
68 * 0-7 (8) : rtas_error_log
69 * 8-47 (40) : extended info
70 * 48-51 (4) : vendor id
71 * 52-1023 (vendor specific) : location code and debug data
73 static void printk_log_rtas(char *buf, int len)
79 char * str = "RTAS event";
81 printk(RTAS_DEBUG "%d -------- %s begin --------\n", error_log_cnt, str);
84 * Print perline bytes on each line, each line will start
85 * with RTAS and a changing number, so syslogd will
86 * print lines that are otherwise the same. Separate every
87 * 4 bytes with a space.
89 for (i=0; i < len; i++) {
92 memset(buffer, 0, sizeof(buffer));
93 n = sprintf(buffer, "RTAS %d:", i/perline);
97 n += sprintf(buffer+n, " ");
99 n += sprintf(buffer+n, "%02x", (unsigned char)buf[i]);
101 if (j == (perline-1))
102 printk(KERN_DEBUG "%s\n", buffer);
104 if ((i % perline) != 0)
105 printk(KERN_DEBUG "%s\n", buffer);
107 printk(RTAS_DEBUG "%d -------- %s end ----------\n", error_log_cnt, str);
110 static int log_rtas_len(char * buf)
113 struct rtas_error_log *err;
115 /* rtas fixed header */
117 err = (struct rtas_error_log *)buf;
118 if (err->extended_log_length) {
120 /* extended header */
121 len += err->extended_log_length;
124 if (len > rtas_error_log_max)
125 len = rtas_error_log_max;
131 * First write to nvram, if fatal error, that is the only
132 * place we log the info. The error will be picked up
133 * on the next reboot by rtasd. If not fatal, run the
134 * method for the type of error. Currently, only RTAS
135 * errors have methods implemented, but in the future
136 * there might be a need to store data in nvram before a
139 * XXX We write to nvram periodically, to indicate error has
140 * been written and sync'd, but there is a possibility
141 * that if we don't shutdown correctly, a duplicate error
142 * record will be created on next reboot.
144 void pSeries_log_error(char *buf, unsigned int err_type, int fatal)
146 unsigned long offset;
150 DEBUG("logging event\n");
155 spin_lock_irqsave(&log_lock, s);
157 /* get length and increase count */
158 switch (err_type & ERR_TYPE_MASK) {
159 case ERR_TYPE_RTAS_LOG:
160 len = log_rtas_len(buf);
161 if (!(err_type & ERR_FLAG_BOOT))
164 case ERR_TYPE_KERNEL_PANIC:
166 spin_unlock_irqrestore(&log_lock, s);
170 /* Write error to NVRAM */
171 if (!no_more_logging && !(err_type & ERR_FLAG_BOOT))
172 nvram_write_error_log(buf, len, err_type);
174 /* Check to see if we need to or have stopped logging */
175 if (fatal || no_more_logging) {
177 spin_unlock_irqrestore(&log_lock, s);
181 /* call type specific method for error */
182 switch (err_type & ERR_TYPE_MASK) {
183 case ERR_TYPE_RTAS_LOG:
184 /* put into syslog and error_log file */
185 printk_log_rtas(buf, len);
187 offset = rtas_error_log_buffer_max *
188 ((rtas_log_start+rtas_log_size) & LOG_NUMBER_MASK);
190 /* First copy over sequence number */
191 memcpy(&rtas_log_buf[offset], (void *) &error_log_cnt, sizeof(int));
193 /* Second copy over error log data */
194 offset += sizeof(int);
195 memcpy(&rtas_log_buf[offset], buf, len);
197 if (rtas_log_size < LOG_NUMBER)
202 spin_unlock_irqrestore(&log_lock, s);
203 wake_up_interruptible(&rtas_log_wait);
205 case ERR_TYPE_KERNEL_PANIC:
207 spin_unlock_irqrestore(&log_lock, s);
214 static int rtas_log_open(struct inode * inode, struct file * file)
219 static int rtas_log_release(struct inode * inode, struct file * file)
224 /* This will check if all events are logged, if they are then, we
225 * know that we can safely clear the events in NVRAM.
226 * Next we'll sit and wait for something else to log.
228 static ssize_t rtas_log_read(struct file * file, char * buf,
229 size_t count, loff_t *ppos)
234 unsigned long offset;
236 if (!buf || count < rtas_error_log_buffer_max)
239 count = rtas_error_log_buffer_max;
241 error = verify_area(VERIFY_WRITE, buf, count);
245 tmp = kmalloc(count, GFP_KERNEL);
250 spin_lock_irqsave(&log_lock, s);
251 /* if it's 0, then we know we got the last one (the one in NVRAM) */
252 if (rtas_log_size == 0 && !no_more_logging)
253 nvram_clear_error_log();
254 spin_unlock_irqrestore(&log_lock, s);
257 error = wait_event_interruptible(rtas_log_wait, rtas_log_size);
261 spin_lock_irqsave(&log_lock, s);
262 offset = rtas_error_log_buffer_max * (rtas_log_start & LOG_NUMBER_MASK);
263 memcpy(tmp, &rtas_log_buf[offset], count);
267 spin_unlock_irqrestore(&log_lock, s);
269 error = copy_to_user(buf, tmp, count) ? -EFAULT : count;
275 static unsigned int rtas_log_poll(struct file *file, poll_table * wait)
277 poll_wait(file, &rtas_log_wait, wait);
279 return POLLIN | POLLRDNORM;
283 struct file_operations proc_rtas_log_operations = {
284 .read = rtas_log_read,
285 .poll = rtas_log_poll,
286 .open = rtas_log_open,
287 .release = rtas_log_release,
290 static int enable_surveillance(int timeout)
294 error = rtas_call(rtas_token("set-indicator"), 3, 1, NULL,
295 SURVEILLANCE_TOKEN, 0, timeout);
298 printk(KERN_ERR "rtasd: could not enable surveillance\n");
305 static int get_eventscan_parms(void)
307 struct device_node *node;
310 node = of_find_node_by_path("/rtas");
312 ip = (int *)get_property(node, "rtas-event-scan-rate", NULL);
314 printk(KERN_ERR "rtasd: no rtas-event-scan-rate\n");
318 rtas_event_scan_rate = *ip;
319 DEBUG("rtas-event-scan-rate %d\n", rtas_event_scan_rate);
321 ip = (int *)get_property(node, "rtas-error-log-max", NULL);
323 printk(KERN_ERR "rtasd: no rtas-error-log-max\n");
327 rtas_error_log_max = *ip;
328 DEBUG("rtas-error-log-max %d\n", rtas_error_log_max);
330 if (rtas_error_log_max > RTAS_ERROR_LOG_MAX) {
331 printk(KERN_ERR "rtasd: truncated error log from %d to %d bytes\n", rtas_error_log_max, RTAS_ERROR_LOG_MAX);
332 rtas_error_log_max = RTAS_ERROR_LOG_MAX;
335 /* Make room for the sequence number */
336 rtas_error_log_buffer_max = rtas_error_log_max + sizeof(int);
343 static void do_event_scan(int event_scan)
347 memset(logdata, 0, rtas_error_log_max);
348 error = rtas_call(event_scan, 4, 1, NULL,
349 RTAS_EVENT_SCAN_ALL_EVENTS, 0,
350 __pa(logdata), rtas_error_log_max);
352 printk(KERN_ERR "event-scan failed\n");
357 pSeries_log_error(logdata, ERR_TYPE_RTAS_LOG, 0);
362 static int rtasd(void *unused)
364 unsigned int err_type;
366 int event_scan = rtas_token("event-scan");
371 if (event_scan == RTAS_UNKNOWN_SERVICE || get_eventscan_parms() == -1)
374 rtas_log_buf = vmalloc(rtas_error_log_buffer_max*LOG_NUMBER);
376 printk(KERN_ERR "rtasd: no memory\n");
380 /* We can use rtas_log_buf now */
383 printk(KERN_ERR "RTAS daemon started\n");
385 DEBUG("will sleep for %d jiffies\n", (HZ*60/rtas_event_scan_rate) / 2);
387 /* See if we have any error stored in NVRAM */
388 memset(logdata, 0, rtas_error_log_max);
390 rc = nvram_read_error_log(logdata, rtas_error_log_max, &err_type);
392 if (err_type != ERR_FLAG_ALREADY_LOGGED) {
393 pSeries_log_error(logdata, err_type | ERR_FLAG_BOOT, 0);
399 for_each_online_cpu(cpu) {
400 DEBUG("scheduling on %d\n", cpu);
401 set_cpus_allowed(current, cpumask_of_cpu(cpu));
402 DEBUG("watchdog scheduled on cpu %d\n", smp_processor_id());
404 do_event_scan(event_scan);
405 set_current_state(TASK_INTERRUPTIBLE);
406 schedule_timeout(HZ);
408 unlock_cpu_hotplug();
410 if (surveillance_timeout != -1) {
411 DEBUG("enabling surveillance\n");
412 enable_surveillance(surveillance_timeout);
413 DEBUG("surveillance enabled\n");
417 cpu = first_cpu(cpu_online_map);
419 set_cpus_allowed(current, cpumask_of_cpu(cpu));
420 do_event_scan(event_scan);
421 set_cpus_allowed(current, CPU_MASK_ALL);
423 /* Drop hotplug lock, and sleep for a bit (at least
424 * one second since some machines have problems if we
425 * call event-scan too quickly). */
426 unlock_cpu_hotplug();
427 set_current_state(TASK_INTERRUPTIBLE);
428 schedule_timeout((HZ*60/rtas_event_scan_rate) / 2);
431 cpu = next_cpu(cpu, cpu_online_map);
433 cpu = first_cpu(cpu_online_map);
437 /* Should delete proc entries */
441 static int __init rtas_init(void)
443 struct proc_dir_entry *entry;
445 /* No RTAS, only warn if we are on a pSeries box */
446 if (rtas_token("event-scan") == RTAS_UNKNOWN_SERVICE) {
447 if (systemcfg->platform & PLATFORM_PSERIES);
448 printk(KERN_ERR "rtasd: no RTAS on system\n");
452 entry = create_proc_entry("ppc64/rtas/error_log", S_IRUSR, NULL);
454 entry->proc_fops = &proc_rtas_log_operations;
456 printk(KERN_ERR "Failed to create error_log proc entry\n");
458 if (kernel_thread(rtasd, 0, CLONE_FS) < 0)
459 printk(KERN_ERR "Failed to start RTAS daemon\n");
464 static int __init surveillance_setup(char *str)
468 if (get_option(&str,&i)) {
469 if (i >= 0 && i <= 255)
470 surveillance_timeout = i;
476 __initcall(rtas_init);
477 __setup("surveillance=", surveillance_setup);