aboutsummaryrefslogtreecommitdiffstats
path: root/arch/powerpc/platforms/pseries
diff options
context:
space:
mode:
authorBenjamin Herrenschmidt <benh@kernel.crashing.org>2009-09-24 15:30:05 -0400
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>2009-10-30 02:20:53 -0400
commit3d541c4b7f6efd55a98189afd1b2f1c9d048c1b3 (patch)
tree37ea005412feedefe836afd0752051b0c93f4f71 /arch/powerpc/platforms/pseries
parent188917e183cf9ad0374b571006d0fc6d48a7f447 (diff)
powerpc/chrp: Use the same RTAS daemon as pSeries
The CHRP code has some fishy timer based code to scan the RTAS event log, which uses a 1KB stack buffer and doesn't even use the results. The pSeries code as a nicer daemon that allows userspace to read the event log and basically uses the same RTAS interface This patch moves rtasd.c out of platform/pseries and makes it usable by CHRP, after removing the old crufty event log mechanism in there. The nvram logging part of the daemon is still only available on 64-bit since the underlying nvram management routines aren't currently shared. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Diffstat (limited to 'arch/powerpc/platforms/pseries')
-rw-r--r--arch/powerpc/platforms/pseries/Kconfig1
-rw-r--r--arch/powerpc/platforms/pseries/Makefile2
-rw-r--r--arch/powerpc/platforms/pseries/rtasd.c519
3 files changed, 2 insertions, 520 deletions
diff --git a/arch/powerpc/platforms/pseries/Kconfig b/arch/powerpc/platforms/pseries/Kconfig
index f0e6f28427bd..26a24bd92623 100644
--- a/arch/powerpc/platforms/pseries/Kconfig
+++ b/arch/powerpc/platforms/pseries/Kconfig
@@ -4,6 +4,7 @@ config PPC_PSERIES
4 select MPIC 4 select MPIC
5 select PPC_I8259 5 select PPC_I8259
6 select PPC_RTAS 6 select PPC_RTAS
7 select PPC_RTAS_DAEMON
7 select RTAS_ERROR_LOGGING 8 select RTAS_ERROR_LOGGING
8 select PPC_UDBG_16550 9 select PPC_UDBG_16550
9 select PPC_NATIVE 10 select PPC_NATIVE
diff --git a/arch/powerpc/platforms/pseries/Makefile b/arch/powerpc/platforms/pseries/Makefile
index 790c0b872d4f..4b1c422b8145 100644
--- a/arch/powerpc/platforms/pseries/Makefile
+++ b/arch/powerpc/platforms/pseries/Makefile
@@ -7,7 +7,7 @@ EXTRA_CFLAGS += -DDEBUG
7endif 7endif
8 8
9obj-y := lpar.o hvCall.o nvram.o reconfig.o \ 9obj-y := lpar.o hvCall.o nvram.o reconfig.o \
10 setup.o iommu.o ras.o rtasd.o \ 10 setup.o iommu.o ras.o \
11 firmware.o power.o 11 firmware.o power.o
12obj-$(CONFIG_SMP) += smp.o 12obj-$(CONFIG_SMP) += smp.o
13obj-$(CONFIG_XICS) += xics.o 13obj-$(CONFIG_XICS) += xics.o
diff --git a/arch/powerpc/platforms/pseries/rtasd.c b/arch/powerpc/platforms/pseries/rtasd.c
deleted file mode 100644
index b3cbac855924..000000000000
--- a/arch/powerpc/platforms/pseries/rtasd.c
+++ /dev/null
@@ -1,519 +0,0 @@
1/*
2 * Copyright (C) 2001 Anton Blanchard <anton@au.ibm.com>, IBM
3 *
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.
8 *
9 * Communication to userspace based on kernel/printk.c
10 */
11
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
24#include <asm/uaccess.h>
25#include <asm/io.h>
26#include <asm/rtas.h>
27#include <asm/prom.h>
28#include <asm/nvram.h>
29#include <asm/atomic.h>
30#include <asm/machdep.h>
31
32
33static DEFINE_SPINLOCK(rtasd_log_lock);
34
35static DECLARE_WAIT_QUEUE_HEAD(rtas_log_wait);
36
37static char *rtas_log_buf;
38static unsigned long rtas_log_start;
39static unsigned long rtas_log_size;
40
41static int surveillance_timeout = -1;
42static unsigned int rtas_error_log_max;
43static unsigned int rtas_error_log_buffer_max;
44
45/* RTAS service tokens */
46static unsigned int event_scan;
47static unsigned int rtas_event_scan_rate;
48
49static int full_rtas_msgs = 0;
50
51/* Stop logging to nvram after first fatal error */
52static int logging_enabled; /* Until we initialize everything,
53 * make sure we don't try logging
54 * anything */
55static int error_log_cnt;
56
57/*
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
60 * make it big enough.
61 */
62static unsigned char logdata[RTAS_ERROR_LOG_MAX];
63
64static 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",
68};
69
70static char *rtas_event_type(int type)
71{
72 if ((type > 0) && (type < 11))
73 return rtas_type[type];
74
75 switch (type) {
76 case RTAS_TYPE_EPOW:
77 return "EPOW";
78 case RTAS_TYPE_PLATFORM:
79 return "Platform Error";
80 case RTAS_TYPE_IO:
81 return "I/O Event";
82 case RTAS_TYPE_INFO:
83 return "Platform Information Event";
84 case RTAS_TYPE_DEALLOC:
85 return "Resource Deallocation Event";
86 case RTAS_TYPE_DUMP:
87 return "Dump Notification Event";
88 }
89
90 return rtas_type[0];
91}
92
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.
97 *
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
105 */
106static void printk_log_rtas(char *buf, int len)
107{
108
109 int i,j,n = 0;
110 int perline = 16;
111 char buffer[64];
112 char * str = "RTAS event";
113
114 if (full_rtas_msgs) {
115 printk(RTAS_DEBUG "%d -------- %s begin --------\n",
116 error_log_cnt, str);
117
118 /*
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.
123 */
124 for (i = 0; i < len; i++) {
125 j = i % perline;
126 if (j == 0) {
127 memset(buffer, 0, sizeof(buffer));
128 n = sprintf(buffer, "RTAS %d:", i/perline);
129 }
130
131 if ((i % 4) == 0)
132 n += sprintf(buffer+n, " ");
133
134 n += sprintf(buffer+n, "%02x", (unsigned char)buf[i]);
135
136 if (j == (perline-1))
137 printk(KERN_DEBUG "%s\n", buffer);
138 }
139 if ((i % perline) != 0)
140 printk(KERN_DEBUG "%s\n", buffer);
141
142 printk(RTAS_DEBUG "%d -------- %s end ----------\n",
143 error_log_cnt, str);
144 } else {
145 struct rtas_error_log *errlog = (struct rtas_error_log *)buf;
146
147 printk(RTAS_DEBUG "event: %d, Type: %s, Severity: %d\n",
148 error_log_cnt, rtas_event_type(errlog->type),
149 errlog->severity);
150 }
151}
152
153static int log_rtas_len(char * buf)
154{
155 int len;
156 struct rtas_error_log *err;
157
158 /* rtas fixed header */
159 len = 8;
160 err = (struct rtas_error_log *)buf;
161 if (err->extended_log_length) {
162
163 /* extended header */
164 len += err->extended_log_length;
165 }
166
167 if (rtas_error_log_max == 0)
168 rtas_error_log_max = rtas_get_error_log_max();
169
170 if (len > rtas_error_log_max)
171 len = rtas_error_log_max;
172
173 return len;
174}
175
176/*
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
183 * call to panic().
184 *
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.
189 */
190void pSeries_log_error(char *buf, unsigned int err_type, int fatal)
191{
192 unsigned long offset;
193 unsigned long s;
194 int len = 0;
195
196 pr_debug("rtasd: logging event\n");
197 if (buf == NULL)
198 return;
199
200 spin_lock_irqsave(&rtasd_log_lock, s);
201
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))
207 error_log_cnt++;
208 break;
209 case ERR_TYPE_KERNEL_PANIC:
210 default:
211 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
212 spin_unlock_irqrestore(&rtasd_log_lock, s);
213 return;
214 }
215
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);
219
220 /*
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.
224 */
225 if ((err_type & ERR_TYPE_MASK) == ERR_TYPE_RTAS_LOG)
226 printk_log_rtas(buf, len);
227
228 /* Check to see if we need to or have stopped logging */
229 if (fatal || !logging_enabled) {
230 logging_enabled = 0;
231 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
232 spin_unlock_irqrestore(&rtasd_log_lock, s);
233 return;
234 }
235
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);
241
242 /* First copy over sequence number */
243 memcpy(&rtas_log_buf[offset], (void *) &error_log_cnt, sizeof(int));
244
245 /* Second copy over error log data */
246 offset += sizeof(int);
247 memcpy(&rtas_log_buf[offset], buf, len);
248
249 if (rtas_log_size < LOG_NUMBER)
250 rtas_log_size += 1;
251 else
252 rtas_log_start += 1;
253
254 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
255 spin_unlock_irqrestore(&rtasd_log_lock, s);
256 wake_up_interruptible(&rtas_log_wait);
257 break;
258 case ERR_TYPE_KERNEL_PANIC:
259 default:
260 WARN_ON_ONCE(!irqs_disabled()); /* @@@ DEBUG @@@ */
261 spin_unlock_irqrestore(&rtasd_log_lock, s);
262 return;
263 }
264
265}
266
267
268static int rtas_log_open(struct inode * inode, struct file * file)
269{
270 return 0;
271}
272
273static int rtas_log_release(struct inode * inode, struct file * file)
274{
275 return 0;
276}
277
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.
281 */
282static ssize_t rtas_log_read(struct file * file, char __user * buf,
283 size_t count, loff_t *ppos)
284{
285 int error;
286 char *tmp;
287 unsigned long s;
288 unsigned long offset;
289
290 if (!buf || count < rtas_error_log_buffer_max)
291 return -EINVAL;
292
293 count = rtas_error_log_buffer_max;
294
295 if (!access_ok(VERIFY_WRITE, buf, count))
296 return -EFAULT;
297
298 tmp = kmalloc(count, GFP_KERNEL);
299 if (!tmp)
300 return -ENOMEM;
301
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);
307 error = -EAGAIN;
308 goto out;
309 }
310
311 if (!logging_enabled) {
312 spin_unlock_irqrestore(&rtasd_log_lock, s);
313 error = -ENODATA;
314 goto out;
315 }
316 nvram_clear_error_log();
317
318 spin_unlock_irqrestore(&rtasd_log_lock, s);
319 error = wait_event_interruptible(rtas_log_wait, rtas_log_size);
320 if (error)
321 goto out;
322 spin_lock_irqsave(&rtasd_log_lock, s);
323 }
324
325 offset = rtas_error_log_buffer_max * (rtas_log_start & LOG_NUMBER_MASK);
326 memcpy(tmp, &rtas_log_buf[offset], count);
327
328 rtas_log_start += 1;
329 rtas_log_size -= 1;
330 spin_unlock_irqrestore(&rtasd_log_lock, s);
331
332 error = copy_to_user(buf, tmp, count) ? -EFAULT : count;
333out:
334 kfree(tmp);
335 return error;
336}
337
338static unsigned int rtas_log_poll(struct file *file, poll_table * wait)
339{
340 poll_wait(file, &rtas_log_wait, wait);
341 if (rtas_log_size)
342 return POLLIN | POLLRDNORM;
343 return 0;
344}
345
346static 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,
351};
352
353static int enable_surveillance(int timeout)
354{
355 int error;
356
357 error = rtas_set_indicator(SURVEILLANCE_TOKEN, 0, timeout);
358
359 if (error == 0)
360 return 0;
361
362 if (error == -EINVAL) {
363 printk(KERN_DEBUG "rtasd: surveillance not supported\n");
364 return 0;
365 }
366
367 printk(KERN_ERR "rtasd: could not update surveillance\n");
368 return -1;
369}
370
371static void do_event_scan(void)
372{
373 int error;
374 do {
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);
379 if (error == -1) {
380 printk(KERN_ERR "event-scan failed\n");
381 break;
382 }
383
384 if (error == 0)
385 pSeries_log_error(logdata, ERR_TYPE_RTAS_LOG, 0);
386
387 } while(error == 0);
388}
389
390static void rtas_event_scan(struct work_struct *w);
391DECLARE_DELAYED_WORK(event_scan_work, rtas_event_scan);
392
393/*
394 * Delay should be at least one second since some machines have problems if
395 * we call event-scan too quickly.
396 */
397static unsigned long event_scan_delay = 1*HZ;
398static int first_pass = 1;
399
400static void rtas_event_scan(struct work_struct *w)
401{
402 unsigned int cpu;
403
404 do_event_scan();
405
406 get_online_cpus();
407
408 cpu = next_cpu(smp_processor_id(), cpu_online_map);
409 if (cpu == NR_CPUS) {
410 cpu = first_cpu(cpu_online_map);
411
412 if (first_pass) {
413 first_pass = 0;
414 event_scan_delay = 30*HZ/rtas_event_scan_rate;
415
416 if (surveillance_timeout != -1) {
417 pr_debug("rtasd: enabling surveillance\n");
418 enable_surveillance(surveillance_timeout);
419 pr_debug("rtasd: surveillance enabled\n");
420 }
421 }
422 }
423
424 schedule_delayed_work_on(cpu, &event_scan_work,
425 __round_jiffies_relative(event_scan_delay, cpu));
426
427 put_online_cpus();
428}
429
430static void start_event_scan(void)
431{
432 unsigned int err_type;
433 int rc;
434
435 printk(KERN_DEBUG "RTAS daemon started\n");
436 pr_debug("rtasd: will sleep for %d milliseconds\n",
437 (30000 / rtas_event_scan_rate));
438
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 */
444 logging_enabled = 1;
445
446 if (!rc) {
447 if (err_type != ERR_FLAG_ALREADY_LOGGED) {
448 pSeries_log_error(logdata, err_type | ERR_FLAG_BOOT, 0);
449 }
450 }
451
452 schedule_delayed_work_on(first_cpu(cpu_online_map), &event_scan_work,
453 event_scan_delay);
454}
455
456static int __init rtas_init(void)
457{
458 struct proc_dir_entry *entry;
459
460 if (!machine_is(pseries))
461 return 0;
462
463 /* No RTAS */
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");
467 return -ENODEV;
468 }
469
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");
473 return -ENODEV;
474 }
475
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);
479
480 rtas_log_buf = vmalloc(rtas_error_log_buffer_max*LOG_NUMBER);
481 if (!rtas_log_buf) {
482 printk(KERN_ERR "rtasd: no memory\n");
483 return -ENOMEM;
484 }
485
486 entry = proc_create("ppc64/rtas/error_log", S_IRUSR, NULL,
487 &proc_rtas_log_operations);
488 if (!entry)
489 printk(KERN_ERR "Failed to create error_log proc entry\n");
490
491 start_event_scan();
492
493 return 0;
494}
495
496static int __init surveillance_setup(char *str)
497{
498 int i;
499
500 if (get_option(&str,&i)) {
501 if (i >= 0 && i <= 255)
502 surveillance_timeout = i;
503 }
504
505 return 1;
506}
507
508static int __init rtasmsgs_setup(char *str)
509{
510 if (strcmp(str, "on") == 0)
511 full_rtas_msgs = 1;
512 else if (strcmp(str, "off") == 0)
513 full_rtas_msgs = 0;
514
515 return 1;
516}
517__initcall(rtas_init);
518__setup("surveillance=", surveillance_setup);
519__setup("rtasmsgs=", rtasmsgs_setup);
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-03 11:55:20 -0500