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>
23 #include <linux/slab.h>
25 #include <asm/uaccess.h>
29 #include <asm/nvram.h>
30 #include <asm/atomic.h>
31 #include <asm/machdep.h>
34 static DEFINE_SPINLOCK(rtasd_log_lock
);
36 static DECLARE_WAIT_QUEUE_HEAD(rtas_log_wait
);
38 static char *rtas_log_buf
;
39 static unsigned long rtas_log_start
;
40 static unsigned long rtas_log_size
;
42 static int surveillance_timeout
= -1;
44 static unsigned int rtas_error_log_max
;
45 static unsigned int rtas_error_log_buffer_max
;
47 /* RTAS service tokens */
48 static unsigned int event_scan
;
49 static unsigned int rtas_event_scan_rate
;
51 static int full_rtas_msgs
= 0;
53 /* Stop logging to nvram after first fatal error */
54 static int logging_enabled
; /* Until we initialize everything,
55 * make sure we don't try logging
57 static int error_log_cnt
;
60 * Since we use 32 bit RTAS, the physical address of this must be below
61 * 4G or else bad things happen. Allocate this in the kernel data and
64 static unsigned char logdata
[RTAS_ERROR_LOG_MAX
];
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
];
80 case RTAS_TYPE_PLATFORM
:
81 return "Platform Error";
85 return "Platform Information Event";
86 case RTAS_TYPE_DEALLOC
:
87 return "Resource Deallocation Event";
89 return "Dump Notification Event";
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
)
114 char * str
= "RTAS event";
116 if (full_rtas_msgs
) {
117 printk(RTAS_DEBUG
"%d -------- %s begin --------\n",
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
++) {
129 memset(buffer
, 0, sizeof(buffer
));
130 n
= sprintf(buffer
, "RTAS %d:", i
/perline
);
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",
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
),
155 static int log_rtas_len(char * buf
)
158 struct rtas_error_log
*err
;
160 /* rtas fixed header */
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 rtas_error_log_max
= rtas_get_error_log_max();
172 if (len
> rtas_error_log_max
)
173 len
= rtas_error_log_max
;
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
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
;
198 pr_debug("rtasd: logging event\n");
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
))
211 case ERR_TYPE_KERNEL_PANIC
:
213 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
214 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
219 /* Write error to NVRAM */
220 if (logging_enabled
&& !(err_type
& ERR_FLAG_BOOT
))
221 nvram_write_error_log(buf
, len
, err_type
, error_log_cnt
);
222 #endif /* CONFIG_PPC64 */
225 * rtas errors can occur during boot, and we do want to capture
226 * those somewhere, even if nvram isn't ready (why not?), and even
227 * if rtasd isn't ready. Put them into the boot log, at least.
229 if ((err_type
& ERR_TYPE_MASK
) == ERR_TYPE_RTAS_LOG
)
230 printk_log_rtas(buf
, len
);
232 /* Check to see if we need to or have stopped logging */
233 if (fatal
|| !logging_enabled
) {
235 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
236 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
240 /* call type specific method for error */
241 switch (err_type
& ERR_TYPE_MASK
) {
242 case ERR_TYPE_RTAS_LOG
:
243 offset
= rtas_error_log_buffer_max
*
244 ((rtas_log_start
+rtas_log_size
) & LOG_NUMBER_MASK
);
246 /* First copy over sequence number */
247 memcpy(&rtas_log_buf
[offset
], (void *) &error_log_cnt
, sizeof(int));
249 /* Second copy over error log data */
250 offset
+= sizeof(int);
251 memcpy(&rtas_log_buf
[offset
], buf
, len
);
253 if (rtas_log_size
< LOG_NUMBER
)
258 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
259 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
260 wake_up_interruptible(&rtas_log_wait
);
262 case ERR_TYPE_KERNEL_PANIC
:
264 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
265 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
271 static int rtas_log_open(struct inode
* inode
, struct file
* file
)
276 static int rtas_log_release(struct inode
* inode
, struct file
* file
)
281 /* This will check if all events are logged, if they are then, we
282 * know that we can safely clear the events in NVRAM.
283 * Next we'll sit and wait for something else to log.
285 static ssize_t
rtas_log_read(struct file
* file
, char __user
* buf
,
286 size_t count
, loff_t
*ppos
)
291 unsigned long offset
;
293 if (!buf
|| count
< rtas_error_log_buffer_max
)
296 count
= rtas_error_log_buffer_max
;
298 if (!access_ok(VERIFY_WRITE
, buf
, count
))
301 tmp
= kmalloc(count
, GFP_KERNEL
);
305 spin_lock_irqsave(&rtasd_log_lock
, s
);
307 /* if it's 0, then we know we got the last one (the one in NVRAM) */
308 while (rtas_log_size
== 0) {
309 if (file
->f_flags
& O_NONBLOCK
) {
310 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
315 if (!logging_enabled
) {
316 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
321 nvram_clear_error_log();
322 #endif /* CONFIG_PPC64 */
324 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
325 error
= wait_event_interruptible(rtas_log_wait
, rtas_log_size
);
328 spin_lock_irqsave(&rtasd_log_lock
, s
);
331 offset
= rtas_error_log_buffer_max
* (rtas_log_start
& LOG_NUMBER_MASK
);
332 memcpy(tmp
, &rtas_log_buf
[offset
], count
);
336 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
338 error
= copy_to_user(buf
, tmp
, count
) ? -EFAULT
: count
;
344 static unsigned int rtas_log_poll(struct file
*file
, poll_table
* wait
)
346 poll_wait(file
, &rtas_log_wait
, wait
);
348 return POLLIN
| POLLRDNORM
;
352 static const struct file_operations proc_rtas_log_operations
= {
353 .read
= rtas_log_read
,
354 .poll
= rtas_log_poll
,
355 .open
= rtas_log_open
,
356 .release
= rtas_log_release
,
359 static int enable_surveillance(int timeout
)
363 error
= rtas_set_indicator(SURVEILLANCE_TOKEN
, 0, timeout
);
368 if (error
== -EINVAL
) {
369 printk(KERN_DEBUG
"rtasd: surveillance not supported\n");
373 printk(KERN_ERR
"rtasd: could not update surveillance\n");
377 static void do_event_scan(void)
381 memset(logdata
, 0, rtas_error_log_max
);
382 error
= rtas_call(event_scan
, 4, 1, NULL
,
383 RTAS_EVENT_SCAN_ALL_EVENTS
, 0,
384 __pa(logdata
), rtas_error_log_max
);
386 printk(KERN_ERR
"event-scan failed\n");
391 pSeries_log_error(logdata
, ERR_TYPE_RTAS_LOG
, 0);
396 static void rtas_event_scan(struct work_struct
*w
);
397 DECLARE_DELAYED_WORK(event_scan_work
, rtas_event_scan
);
400 * Delay should be at least one second since some machines have problems if
401 * we call event-scan too quickly.
403 static unsigned long event_scan_delay
= 1*HZ
;
404 static int first_pass
= 1;
406 static void rtas_event_scan(struct work_struct
*w
)
414 cpu
= cpumask_next(smp_processor_id(), cpu_online_mask
);
415 if (cpu
>= nr_cpu_ids
) {
416 cpu
= cpumask_first(cpu_online_mask
);
420 event_scan_delay
= 30*HZ
/rtas_event_scan_rate
;
422 if (surveillance_timeout
!= -1) {
423 pr_debug("rtasd: enabling surveillance\n");
424 enable_surveillance(surveillance_timeout
);
425 pr_debug("rtasd: surveillance enabled\n");
430 schedule_delayed_work_on(cpu
, &event_scan_work
,
431 __round_jiffies_relative(event_scan_delay
, cpu
));
437 static void retreive_nvram_error_log(void)
439 unsigned int err_type
;
442 /* See if we have any error stored in NVRAM */
443 memset(logdata
, 0, rtas_error_log_max
);
444 rc
= nvram_read_error_log(logdata
, rtas_error_log_max
,
445 &err_type
, &error_log_cnt
);
446 /* We can use rtas_log_buf now */
449 if (err_type
!= ERR_FLAG_ALREADY_LOGGED
) {
450 pSeries_log_error(logdata
, err_type
| ERR_FLAG_BOOT
, 0);
454 #else /* CONFIG_PPC64 */
455 static void retreive_nvram_error_log(void)
458 #endif /* CONFIG_PPC64 */
460 static void start_event_scan(void)
462 printk(KERN_DEBUG
"RTAS daemon started\n");
463 pr_debug("rtasd: will sleep for %d milliseconds\n",
464 (30000 / rtas_event_scan_rate
));
466 /* Retreive errors from nvram if any */
467 retreive_nvram_error_log();
469 schedule_delayed_work_on(cpumask_first(cpu_online_mask
),
470 &event_scan_work
, event_scan_delay
);
473 static int __init
rtas_init(void)
475 struct proc_dir_entry
*entry
;
477 if (!machine_is(pseries
) && !machine_is(chrp
))
481 event_scan
= rtas_token("event-scan");
482 if (event_scan
== RTAS_UNKNOWN_SERVICE
) {
483 printk(KERN_INFO
"rtasd: No event-scan on system\n");
487 rtas_event_scan_rate
= rtas_token("rtas-event-scan-rate");
488 if (rtas_event_scan_rate
== RTAS_UNKNOWN_SERVICE
) {
489 printk(KERN_ERR
"rtasd: no rtas-event-scan-rate on system\n");
493 /* Make room for the sequence number */
494 rtas_error_log_max
= rtas_get_error_log_max();
495 rtas_error_log_buffer_max
= rtas_error_log_max
+ sizeof(int);
497 rtas_log_buf
= vmalloc(rtas_error_log_buffer_max
*LOG_NUMBER
);
499 printk(KERN_ERR
"rtasd: no memory\n");
503 entry
= proc_create("powerpc/rtas/error_log", S_IRUSR
, NULL
,
504 &proc_rtas_log_operations
);
506 printk(KERN_ERR
"Failed to create error_log proc entry\n");
512 __initcall(rtas_init
);
514 static int __init
surveillance_setup(char *str
)
518 /* We only do surveillance on pseries */
519 if (!machine_is(pseries
))
522 if (get_option(&str
,&i
)) {
523 if (i
>= 0 && i
<= 255)
524 surveillance_timeout
= i
;
529 __setup("surveillance=", surveillance_setup
);
531 static int __init
rtasmsgs_setup(char *str
)
533 if (strcmp(str
, "on") == 0)
535 else if (strcmp(str
, "off") == 0)
540 __setup("rtasmsgs=", rtasmsgs_setup
);