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;
42 static unsigned int rtas_error_log_max
;
43 static unsigned int rtas_error_log_buffer_max
;
45 /* RTAS service tokens */
46 static unsigned int event_scan
;
47 static unsigned int rtas_event_scan_rate
;
49 static int full_rtas_msgs
= 0;
51 /* Stop logging to nvram after first fatal error */
52 static int logging_enabled
; /* Until we initialize everything,
53 * make sure we don't try logging
55 static int error_log_cnt
;
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
62 static unsigned char logdata
[RTAS_ERROR_LOG_MAX
];
64 static char *rtas_type
[] = {
65 "Unknown", "Retry", "TCE Error", "Internal Device Failure",
66 "Timeout", "Data Parity", "Address Parity", "Cache Parity",
67 "Address Invalid", "ECC Uncorrected", "ECC Corrupted",
70 static char *rtas_event_type(int type
)
72 if ((type
> 0) && (type
< 11))
73 return rtas_type
[type
];
78 case RTAS_TYPE_PLATFORM
:
79 return "Platform Error";
83 return "Platform Information Event";
84 case RTAS_TYPE_DEALLOC
:
85 return "Resource Deallocation Event";
87 return "Dump Notification Event";
93 /* To see this info, grep RTAS /var/log/messages and each entry
94 * will be collected together with obvious begin/end.
95 * There will be a unique identifier on the begin and end lines.
96 * This will persist across reboots.
98 * format of error logs returned from RTAS:
99 * bytes (size) : contents
100 * --------------------------------------------------------
101 * 0-7 (8) : rtas_error_log
102 * 8-47 (40) : extended info
103 * 48-51 (4) : vendor id
104 * 52-1023 (vendor specific) : location code and debug data
106 static void printk_log_rtas(char *buf
, int len
)
112 char * str
= "RTAS event";
114 if (full_rtas_msgs
) {
115 printk(RTAS_DEBUG
"%d -------- %s begin --------\n",
119 * Print perline bytes on each line, each line will start
120 * with RTAS and a changing number, so syslogd will
121 * print lines that are otherwise the same. Separate every
122 * 4 bytes with a space.
124 for (i
= 0; i
< len
; i
++) {
127 memset(buffer
, 0, sizeof(buffer
));
128 n
= sprintf(buffer
, "RTAS %d:", i
/perline
);
132 n
+= sprintf(buffer
+n
, " ");
134 n
+= sprintf(buffer
+n
, "%02x", (unsigned char)buf
[i
]);
136 if (j
== (perline
-1))
137 printk(KERN_DEBUG
"%s\n", buffer
);
139 if ((i
% perline
) != 0)
140 printk(KERN_DEBUG
"%s\n", buffer
);
142 printk(RTAS_DEBUG
"%d -------- %s end ----------\n",
145 struct rtas_error_log
*errlog
= (struct rtas_error_log
*)buf
;
147 printk(RTAS_DEBUG
"event: %d, Type: %s, Severity: %d\n",
148 error_log_cnt
, rtas_event_type(errlog
->type
),
153 static int log_rtas_len(char * buf
)
156 struct rtas_error_log
*err
;
158 /* rtas fixed header */
160 err
= (struct rtas_error_log
*)buf
;
161 if (err
->extended_log_length
) {
163 /* extended header */
164 len
+= err
->extended_log_length
;
167 if (rtas_error_log_max
== 0)
168 rtas_error_log_max
= rtas_get_error_log_max();
170 if (len
> rtas_error_log_max
)
171 len
= rtas_error_log_max
;
177 * First write to nvram, if fatal error, that is the only
178 * place we log the info. The error will be picked up
179 * on the next reboot by rtasd. If not fatal, run the
180 * method for the type of error. Currently, only RTAS
181 * errors have methods implemented, but in the future
182 * there might be a need to store data in nvram before a
185 * XXX We write to nvram periodically, to indicate error has
186 * been written and sync'd, but there is a possibility
187 * that if we don't shutdown correctly, a duplicate error
188 * record will be created on next reboot.
190 void pSeries_log_error(char *buf
, unsigned int err_type
, int fatal
)
192 unsigned long offset
;
196 pr_debug("rtasd: logging event\n");
200 spin_lock_irqsave(&rtasd_log_lock
, s
);
202 /* get length and increase count */
203 switch (err_type
& ERR_TYPE_MASK
) {
204 case ERR_TYPE_RTAS_LOG
:
205 len
= log_rtas_len(buf
);
206 if (!(err_type
& ERR_FLAG_BOOT
))
209 case ERR_TYPE_KERNEL_PANIC
:
211 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
212 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
216 /* Write error to NVRAM */
217 if (logging_enabled
&& !(err_type
& ERR_FLAG_BOOT
))
218 nvram_write_error_log(buf
, len
, err_type
, error_log_cnt
);
221 * rtas errors can occur during boot, and we do want to capture
222 * those somewhere, even if nvram isn't ready (why not?), and even
223 * if rtasd isn't ready. Put them into the boot log, at least.
225 if ((err_type
& ERR_TYPE_MASK
) == ERR_TYPE_RTAS_LOG
)
226 printk_log_rtas(buf
, len
);
228 /* Check to see if we need to or have stopped logging */
229 if (fatal
|| !logging_enabled
) {
231 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
232 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
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
)
254 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
255 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
256 wake_up_interruptible(&rtas_log_wait
);
258 case ERR_TYPE_KERNEL_PANIC
:
260 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
261 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
268 static int rtas_log_open(struct inode
* inode
, struct file
* file
)
273 static int rtas_log_release(struct inode
* inode
, struct file
* file
)
278 /* This will check if all events are logged, if they are then, we
279 * know that we can safely clear the events in NVRAM.
280 * Next we'll sit and wait for something else to log.
282 static ssize_t
rtas_log_read(struct file
* file
, char __user
* buf
,
283 size_t count
, loff_t
*ppos
)
288 unsigned long offset
;
290 if (!buf
|| count
< rtas_error_log_buffer_max
)
293 count
= rtas_error_log_buffer_max
;
295 if (!access_ok(VERIFY_WRITE
, buf
, count
))
298 tmp
= kmalloc(count
, GFP_KERNEL
);
302 spin_lock_irqsave(&rtasd_log_lock
, s
);
303 /* if it's 0, then we know we got the last one (the one in NVRAM) */
304 while (rtas_log_size
== 0) {
305 if (file
->f_flags
& O_NONBLOCK
) {
306 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
311 if (!logging_enabled
) {
312 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
316 nvram_clear_error_log();
318 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
319 error
= wait_event_interruptible(rtas_log_wait
, rtas_log_size
);
322 spin_lock_irqsave(&rtasd_log_lock
, s
);
325 offset
= rtas_error_log_buffer_max
* (rtas_log_start
& LOG_NUMBER_MASK
);
326 memcpy(tmp
, &rtas_log_buf
[offset
], count
);
330 spin_unlock_irqrestore(&rtasd_log_lock
, s
);
332 error
= copy_to_user(buf
, tmp
, count
) ? -EFAULT
: count
;
338 static unsigned int rtas_log_poll(struct file
*file
, poll_table
* wait
)
340 poll_wait(file
, &rtas_log_wait
, wait
);
342 return POLLIN
| POLLRDNORM
;
346 static const struct file_operations proc_rtas_log_operations
= {
347 .read
= rtas_log_read
,
348 .poll
= rtas_log_poll
,
349 .open
= rtas_log_open
,
350 .release
= rtas_log_release
,
353 static int enable_surveillance(int timeout
)
357 error
= rtas_set_indicator(SURVEILLANCE_TOKEN
, 0, timeout
);
362 if (error
== -EINVAL
) {
363 printk(KERN_DEBUG
"rtasd: surveillance not supported\n");
367 printk(KERN_ERR
"rtasd: could not update surveillance\n");
371 static void do_event_scan(void)
375 memset(logdata
, 0, rtas_error_log_max
);
376 error
= rtas_call(event_scan
, 4, 1, NULL
,
377 RTAS_EVENT_SCAN_ALL_EVENTS
, 0,
378 __pa(logdata
), rtas_error_log_max
);
380 printk(KERN_ERR
"event-scan failed\n");
385 pSeries_log_error(logdata
, ERR_TYPE_RTAS_LOG
, 0);
390 static void rtas_event_scan(struct work_struct
*w
);
391 DECLARE_DELAYED_WORK(event_scan_work
, rtas_event_scan
);
394 * Delay should be at least one second since some machines have problems if
395 * we call event-scan too quickly.
397 static unsigned long event_scan_delay
= 1*HZ
;
398 static int first_pass
= 1;
400 static void rtas_event_scan(struct work_struct
*w
)
408 cpu
= next_cpu(smp_processor_id(), cpu_online_map
);
409 if (cpu
== NR_CPUS
) {
410 cpu
= first_cpu(cpu_online_map
);
414 event_scan_delay
= 30*HZ
/rtas_event_scan_rate
;
416 if (surveillance_timeout
!= -1) {
417 pr_debug("rtasd: enabling surveillance\n");
418 enable_surveillance(surveillance_timeout
);
419 pr_debug("rtasd: surveillance enabled\n");
424 schedule_delayed_work_on(cpu
, &event_scan_work
,
425 __round_jiffies_relative(event_scan_delay
, cpu
));
430 static void start_event_scan(void)
432 unsigned int err_type
;
435 printk(KERN_DEBUG
"RTAS daemon started\n");
436 pr_debug("rtasd: will sleep for %d milliseconds\n",
437 (30000 / rtas_event_scan_rate
));
439 /* See if we have any error stored in NVRAM */
440 memset(logdata
, 0, rtas_error_log_max
);
441 rc
= nvram_read_error_log(logdata
, rtas_error_log_max
,
442 &err_type
, &error_log_cnt
);
443 /* We can use rtas_log_buf now */
447 if (err_type
!= ERR_FLAG_ALREADY_LOGGED
) {
448 pSeries_log_error(logdata
, err_type
| ERR_FLAG_BOOT
, 0);
452 schedule_delayed_work_on(first_cpu(cpu_online_map
), &event_scan_work
,
456 static int __init
rtas_init(void)
458 struct proc_dir_entry
*entry
;
460 if (!machine_is(pseries
))
464 event_scan
= rtas_token("event-scan");
465 if (event_scan
== RTAS_UNKNOWN_SERVICE
) {
466 printk(KERN_DEBUG
"rtasd: no event-scan on system\n");
470 rtas_event_scan_rate
= rtas_token("rtas-event-scan-rate");
471 if (rtas_event_scan_rate
== RTAS_UNKNOWN_SERVICE
) {
472 printk(KERN_ERR
"rtasd: no rtas-event-scan-rate on system\n");
476 /* Make room for the sequence number */
477 rtas_error_log_max
= rtas_get_error_log_max();
478 rtas_error_log_buffer_max
= rtas_error_log_max
+ sizeof(int);
480 rtas_log_buf
= vmalloc(rtas_error_log_buffer_max
*LOG_NUMBER
);
482 printk(KERN_ERR
"rtasd: no memory\n");
486 entry
= proc_create("ppc64/rtas/error_log", S_IRUSR
, NULL
,
487 &proc_rtas_log_operations
);
489 printk(KERN_ERR
"Failed to create error_log proc entry\n");
496 static int __init
surveillance_setup(char *str
)
500 if (get_option(&str
,&i
)) {
501 if (i
>= 0 && i
<= 255)
502 surveillance_timeout
= i
;
508 static int __init
rtasmsgs_setup(char *str
)
510 if (strcmp(str
, "on") == 0)
512 else if (strcmp(str
, "off") == 0)
517 __initcall(rtas_init
);
518 __setup("surveillance=", surveillance_setup
);
519 __setup("rtasmsgs=", rtasmsgs_setup
);