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 <linux/atomic.h>
31 #include <asm/machdep.h>
32 #include <asm/topology.h>
35 static DEFINE_SPINLOCK(rtasd_log_lock
);
37 static DECLARE_WAIT_QUEUE_HEAD(rtas_log_wait
);
39 static char *rtas_log_buf
;
40 static unsigned long rtas_log_start
;
41 static unsigned long rtas_log_size
;
43 static int surveillance_timeout
= -1;
45 static unsigned int rtas_error_log_max
;
46 static unsigned int rtas_error_log_buffer_max
;
48 /* RTAS service tokens */
49 static unsigned int event_scan
;
50 static unsigned int rtas_event_scan_rate
;
52 static int full_rtas_msgs
= 0;
54 /* Stop logging to nvram after first fatal error */
55 static int logging_enabled
; /* Until we initialize everything,
56 * make sure we don't try logging
58 static int error_log_cnt
;
61 * Since we use 32 bit RTAS, the physical address of this must be below
62 * 4G or else bad things happen. Allocate this in the kernel data and
65 static unsigned char logdata
[RTAS_ERROR_LOG_MAX
];
67 static char *rtas_type
[] = {
68 "Unknown", "Retry", "TCE Error", "Internal Device Failure",
69 "Timeout", "Data Parity", "Address Parity", "Cache Parity",
70 "Address Invalid", "ECC Uncorrected", "ECC Corrupted",
73 static char *rtas_event_type(int type
)
75 if ((type
> 0) && (type
< 11))
76 return rtas_type
[type
];
81 case RTAS_TYPE_PLATFORM
:
82 return "Platform Error";
86 return "Platform Information Event";
87 case RTAS_TYPE_DEALLOC
:
88 return "Resource Deallocation Event";
90 return "Dump Notification Event";
92 return "Platform Resource Reassignment Event";
98 /* To see this info, grep RTAS /var/log/messages and each entry
99 * will be collected together with obvious begin/end.
100 * There will be a unique identifier on the begin and end lines.
101 * This will persist across reboots.
103 * format of error logs returned from RTAS:
104 * bytes (size) : contents
105 * --------------------------------------------------------
106 * 0-7 (8) : rtas_error_log
107 * 8-47 (40) : extended info
108 * 48-51 (4) : vendor id
109 * 52-1023 (vendor specific) : location code and debug data
111 static void printk_log_rtas(char *buf
, int len
)
117 char * str
= "RTAS event";
119 if (full_rtas_msgs
) {
120 printk(RTAS_DEBUG
"%d -------- %s begin --------\n",
124 * Print perline bytes on each line, each line will start
125 * with RTAS and a changing number, so syslogd will
126 * print lines that are otherwise the same. Separate every
127 * 4 bytes with a space.
129 for (i
= 0; i
< len
; i
++) {
132 memset(buffer
, 0, sizeof(buffer
));
133 n
= sprintf(buffer
, "RTAS %d:", i
/perline
);
137 n
+= sprintf(buffer
+n
, " ");
139 n
+= sprintf(buffer
+n
, "%02x", (unsigned char)buf
[i
]);
141 if (j
== (perline
-1))
142 printk(KERN_DEBUG
"%s\n", buffer
);
144 if ((i
% perline
) != 0)
145 printk(KERN_DEBUG
"%s\n", buffer
);
147 printk(RTAS_DEBUG
"%d -------- %s end ----------\n",
150 struct rtas_error_log
*errlog
= (struct rtas_error_log
*)buf
;
152 printk(RTAS_DEBUG
"event: %d, Type: %s, Severity: %d\n",
153 error_log_cnt
, rtas_event_type(rtas_error_type(errlog
)),
154 rtas_error_severity(errlog
));
158 static int log_rtas_len(char * buf
)
161 struct rtas_error_log
*err
;
162 uint32_t extended_log_length
;
164 /* rtas fixed header */
166 err
= (struct rtas_error_log
*)buf
;
167 extended_log_length
= rtas_error_extended_log_length(err
);
168 if (rtas_error_extended(err
) && extended_log_length
) {
170 /* extended header */
171 len
+= extended_log_length
;
174 if (rtas_error_log_max
== 0)
175 rtas_error_log_max
= rtas_get_error_log_max();
177 if (len
> rtas_error_log_max
)
178 len
= rtas_error_log_max
;
184 * First write to nvram, if fatal error, that is the only
185 * place we log the info. The error will be picked up
186 * on the next reboot by rtasd. If not fatal, run the
187 * method for the type of error. Currently, only RTAS
188 * errors have methods implemented, but in the future
189 * there might be a need to store data in nvram before a
192 * XXX We write to nvram periodically, to indicate error has
193 * been written and sync'd, but there is a possibility
194 * that if we don't shutdown correctly, a duplicate error
195 * record will be created on next reboot.
197 void pSeries_log_error(char *buf
, unsigned int err_type
, int fatal
)
199 unsigned long offset
;
203 pr_debug("rtasd: logging event\n");
207 spin_lock_irqsave(&rtasd_log_lock
, s
);
209 /* get length and increase count */
210 switch (err_type
& ERR_TYPE_MASK
) {
211 case ERR_TYPE_RTAS_LOG
:
212 len
= log_rtas_len(buf
);
213 if (!(err_type
& ERR_FLAG_BOOT
))
216 case ERR_TYPE_KERNEL_PANIC
:
218 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
219 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
224 /* Write error to NVRAM */
225 if (logging_enabled
&& !(err_type
& ERR_FLAG_BOOT
))
226 nvram_write_error_log(buf
, len
, err_type
, error_log_cnt
);
227 #endif /* CONFIG_PPC64 */
230 * rtas errors can occur during boot, and we do want to capture
231 * those somewhere, even if nvram isn't ready (why not?), and even
232 * if rtasd isn't ready. Put them into the boot log, at least.
234 if ((err_type
& ERR_TYPE_MASK
) == ERR_TYPE_RTAS_LOG
)
235 printk_log_rtas(buf
, len
);
237 /* Check to see if we need to or have stopped logging */
238 if (fatal
|| !logging_enabled
) {
240 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
241 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
245 /* call type specific method for error */
246 switch (err_type
& ERR_TYPE_MASK
) {
247 case ERR_TYPE_RTAS_LOG
:
248 offset
= rtas_error_log_buffer_max
*
249 ((rtas_log_start
+rtas_log_size
) & LOG_NUMBER_MASK
);
251 /* First copy over sequence number */
252 memcpy(&rtas_log_buf
[offset
], (void *) &error_log_cnt
, sizeof(int));
254 /* Second copy over error log data */
255 offset
+= sizeof(int);
256 memcpy(&rtas_log_buf
[offset
], buf
, len
);
258 if (rtas_log_size
< LOG_NUMBER
)
263 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
264 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
265 wake_up_interruptible(&rtas_log_wait
);
267 case ERR_TYPE_KERNEL_PANIC
:
269 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
270 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
275 #ifdef CONFIG_PPC_PSERIES
276 static s32 prrn_update_scope
;
278 static void prrn_work_fn(struct work_struct
*work
)
281 * For PRRN, we must pass the negative of the scope value in
284 pseries_devicetree_update(-prrn_update_scope
);
287 static DECLARE_WORK(prrn_work
, prrn_work_fn
);
289 void prrn_schedule_update(u32 scope
)
291 flush_work(&prrn_work
);
292 prrn_update_scope
= scope
;
293 schedule_work(&prrn_work
);
296 static void handle_rtas_event(const struct rtas_error_log
*log
)
298 if (rtas_error_type(log
) != RTAS_TYPE_PRRN
|| !prrn_is_enabled())
301 /* For PRRN Events the extended log length is used to denote
302 * the scope for calling rtas update-nodes.
304 prrn_schedule_update(rtas_error_extended_log_length(log
));
309 static void handle_rtas_event(const struct rtas_error_log
*log
)
316 static int rtas_log_open(struct inode
* inode
, struct file
* file
)
321 static int rtas_log_release(struct inode
* inode
, struct file
* file
)
326 /* This will check if all events are logged, if they are then, we
327 * know that we can safely clear the events in NVRAM.
328 * Next we'll sit and wait for something else to log.
330 static ssize_t
rtas_log_read(struct file
* file
, char __user
* buf
,
331 size_t count
, loff_t
*ppos
)
336 unsigned long offset
;
338 if (!buf
|| count
< rtas_error_log_buffer_max
)
341 count
= rtas_error_log_buffer_max
;
343 if (!access_ok(VERIFY_WRITE
, buf
, count
))
346 tmp
= kmalloc(count
, GFP_KERNEL
);
350 spin_lock_irqsave(&rtasd_log_lock
, s
);
352 /* if it's 0, then we know we got the last one (the one in NVRAM) */
353 while (rtas_log_size
== 0) {
354 if (file
->f_flags
& O_NONBLOCK
) {
355 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
360 if (!logging_enabled
) {
361 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
366 nvram_clear_error_log();
367 #endif /* CONFIG_PPC64 */
369 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
370 error
= wait_event_interruptible(rtas_log_wait
, rtas_log_size
);
373 spin_lock_irqsave(&rtasd_log_lock
, s
);
376 offset
= rtas_error_log_buffer_max
* (rtas_log_start
& LOG_NUMBER_MASK
);
377 memcpy(tmp
, &rtas_log_buf
[offset
], count
);
381 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
383 error
= copy_to_user(buf
, tmp
, count
) ? -EFAULT
: count
;
389 static unsigned int rtas_log_poll(struct file
*file
, poll_table
* wait
)
391 poll_wait(file
, &rtas_log_wait
, wait
);
393 return POLLIN
| POLLRDNORM
;
397 static const struct file_operations proc_rtas_log_operations
= {
398 .read
= rtas_log_read
,
399 .poll
= rtas_log_poll
,
400 .open
= rtas_log_open
,
401 .release
= rtas_log_release
,
402 .llseek
= noop_llseek
,
405 static int enable_surveillance(int timeout
)
409 error
= rtas_set_indicator(SURVEILLANCE_TOKEN
, 0, timeout
);
414 if (error
== -EINVAL
) {
415 printk(KERN_DEBUG
"rtasd: surveillance not supported\n");
419 printk(KERN_ERR
"rtasd: could not update surveillance\n");
423 static void do_event_scan(void)
427 memset(logdata
, 0, rtas_error_log_max
);
428 error
= rtas_call(event_scan
, 4, 1, NULL
,
429 RTAS_EVENT_SCAN_ALL_EVENTS
, 0,
430 __pa(logdata
), rtas_error_log_max
);
432 printk(KERN_ERR
"event-scan failed\n");
437 pSeries_log_error(logdata
, ERR_TYPE_RTAS_LOG
, 0);
438 handle_rtas_event((struct rtas_error_log
*)logdata
);
444 static void rtas_event_scan(struct work_struct
*w
);
445 DECLARE_DELAYED_WORK(event_scan_work
, rtas_event_scan
);
448 * Delay should be at least one second since some machines have problems if
449 * we call event-scan too quickly.
451 static unsigned long event_scan_delay
= 1*HZ
;
452 static int first_pass
= 1;
454 static void rtas_event_scan(struct work_struct
*w
)
462 /* raw_ OK because just using CPU as starting point. */
463 cpu
= cpumask_next(raw_smp_processor_id(), cpu_online_mask
);
464 if (cpu
>= nr_cpu_ids
) {
465 cpu
= cpumask_first(cpu_online_mask
);
469 event_scan_delay
= 30*HZ
/rtas_event_scan_rate
;
471 if (surveillance_timeout
!= -1) {
472 pr_debug("rtasd: enabling surveillance\n");
473 enable_surveillance(surveillance_timeout
);
474 pr_debug("rtasd: surveillance enabled\n");
479 schedule_delayed_work_on(cpu
, &event_scan_work
,
480 __round_jiffies_relative(event_scan_delay
, cpu
));
486 static void retreive_nvram_error_log(void)
488 unsigned int err_type
;
491 /* See if we have any error stored in NVRAM */
492 memset(logdata
, 0, rtas_error_log_max
);
493 rc
= nvram_read_error_log(logdata
, rtas_error_log_max
,
494 &err_type
, &error_log_cnt
);
495 /* We can use rtas_log_buf now */
498 if (err_type
!= ERR_FLAG_ALREADY_LOGGED
) {
499 pSeries_log_error(logdata
, err_type
| ERR_FLAG_BOOT
, 0);
503 #else /* CONFIG_PPC64 */
504 static void retreive_nvram_error_log(void)
507 #endif /* CONFIG_PPC64 */
509 static void start_event_scan(void)
511 printk(KERN_DEBUG
"RTAS daemon started\n");
512 pr_debug("rtasd: will sleep for %d milliseconds\n",
513 (30000 / rtas_event_scan_rate
));
515 /* Retrieve errors from nvram if any */
516 retreive_nvram_error_log();
518 schedule_delayed_work_on(cpumask_first(cpu_online_mask
),
519 &event_scan_work
, event_scan_delay
);
522 /* Cancel the rtas event scan work */
523 void rtas_cancel_event_scan(void)
525 cancel_delayed_work_sync(&event_scan_work
);
527 EXPORT_SYMBOL_GPL(rtas_cancel_event_scan
);
529 static int __init
rtas_init(void)
531 struct proc_dir_entry
*entry
;
533 if (!machine_is(pseries
) && !machine_is(chrp
))
537 event_scan
= rtas_token("event-scan");
538 if (event_scan
== RTAS_UNKNOWN_SERVICE
) {
539 printk(KERN_INFO
"rtasd: No event-scan on system\n");
543 rtas_event_scan_rate
= rtas_token("rtas-event-scan-rate");
544 if (rtas_event_scan_rate
== RTAS_UNKNOWN_SERVICE
) {
545 printk(KERN_ERR
"rtasd: no rtas-event-scan-rate on system\n");
549 if (!rtas_event_scan_rate
) {
550 /* Broken firmware: take a rate of zero to mean don't scan */
551 printk(KERN_DEBUG
"rtasd: scan rate is 0, not scanning\n");
555 /* Make room for the sequence number */
556 rtas_error_log_max
= rtas_get_error_log_max();
557 rtas_error_log_buffer_max
= rtas_error_log_max
+ sizeof(int);
559 rtas_log_buf
= vmalloc(rtas_error_log_buffer_max
*LOG_NUMBER
);
561 printk(KERN_ERR
"rtasd: no memory\n");
565 entry
= proc_create("powerpc/rtas/error_log", S_IRUSR
, NULL
,
566 &proc_rtas_log_operations
);
568 printk(KERN_ERR
"Failed to create error_log proc entry\n");
574 __initcall(rtas_init
);
576 static int __init
surveillance_setup(char *str
)
580 /* We only do surveillance on pseries */
581 if (!machine_is(pseries
))
584 if (get_option(&str
,&i
)) {
585 if (i
>= 0 && i
<= 255)
586 surveillance_timeout
= i
;
591 __setup("surveillance=", surveillance_setup
);
593 static int __init
rtasmsgs_setup(char *str
)
595 if (strcmp(str
, "on") == 0)
597 else if (strcmp(str
, "off") == 0)
602 __setup("rtasmsgs=", rtasmsgs_setup
);