1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
|
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/vdev.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>
#include <sys/fm/fs/zfs.h>
#include <sys/fm/protocol.h>
#include <sys/fm/util.h>
#ifdef _KERNEL
/* Including sys/bus.h is just too hard, so I declare what I need here. */
extern void devctl_notify(const char *__system, const char *__subsystem,
const char *__type, const char *__data);
#endif
/*
* This general routine is responsible for generating all the different ZFS
* ereports. The payload is dependent on the class, and which arguments are
* supplied to the function:
*
* EREPORT POOL VDEV IO
* block X X X
* data X X
* device X X
* pool X
*
* If we are in a loading state, all errors are chained together by the same
* SPA-wide ENA (Error Numeric Association).
*
* For isolated I/O requests, we get the ENA from the zio_t. The propagation
* gets very complicated due to RAID-Z, gang blocks, and vdev caching. We want
* to chain together all ereports associated with a logical piece of data. For
* read I/Os, there are basically three 'types' of I/O, which form a roughly
* layered diagram:
*
* +---------------+
* | Aggregate I/O | No associated logical data or device
* +---------------+
* |
* V
* +---------------+ Reads associated with a piece of logical data.
* | Read I/O | This includes reads on behalf of RAID-Z,
* +---------------+ mirrors, gang blocks, retries, etc.
* |
* V
* +---------------+ Reads associated with a particular device, but
* | Physical I/O | no logical data. Issued as part of vdev caching
* +---------------+ and I/O aggregation.
*
* Note that 'physical I/O' here is not the same terminology as used in the rest
* of ZIO. Typically, 'physical I/O' simply means that there is no attached
* blockpointer. But I/O with no associated block pointer can still be related
* to a logical piece of data (i.e. RAID-Z requests).
*
* Purely physical I/O always have unique ENAs. They are not related to a
* particular piece of logical data, and therefore cannot be chained together.
* We still generate an ereport, but the DE doesn't correlate it with any
* logical piece of data. When such an I/O fails, the delegated I/O requests
* will issue a retry, which will trigger the 'real' ereport with the correct
* ENA.
*
* We keep track of the ENA for a ZIO chain through the 'io_logical' member.
* When a new logical I/O is issued, we set this to point to itself. Child I/Os
* then inherit this pointer, so that when it is first set subsequent failures
* will use the same ENA. For vdev cache fill and queue aggregation I/O,
* this pointer is set to NULL, and no ereport will be generated (since it
* doesn't actually correspond to any particular device or piece of data,
* and the caller will always retry without caching or queueing anyway).
*/
void
zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
uint64_t stateoroffset, uint64_t size)
{
#ifdef _KERNEL
char buf[1024];
struct sbuf sb;
struct timespec ts;
int state;
/*
* If we are doing a spa_tryimport(), ignore errors.
*/
if (spa->spa_load_state == SPA_LOAD_TRYIMPORT)
return;
/*
* If we are in the middle of opening a pool, and the previous attempt
* failed, don't bother logging any new ereports - we're just going to
* get the same diagnosis anyway.
*/
if (spa->spa_load_state != SPA_LOAD_NONE &&
spa->spa_last_open_failed)
return;
if (zio != NULL) {
/*
* If this is not a read or write zio, ignore the error. This
* can occur if the DKIOCFLUSHWRITECACHE ioctl fails.
*/
if (zio->io_type != ZIO_TYPE_READ &&
zio->io_type != ZIO_TYPE_WRITE)
return;
/*
* Ignore any errors from speculative I/Os, as failure is an
* expected result.
*/
if (zio->io_flags & ZIO_FLAG_SPECULATIVE)
return;
/*
* If the vdev has already been marked as failing due to a
* failed probe, then ignore any subsequent I/O errors, as the
* DE will automatically fault the vdev on the first such
* failure.
*/
if (vd != NULL &&
(!vdev_readable(vd) || !vdev_writeable(vd)) &&
strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) != 0)
return;
}
nanotime(&ts);
sbuf_new(&sb, buf, sizeof(buf), SBUF_FIXEDLEN);
sbuf_printf(&sb, "time=%ju.%ld", (uintmax_t)ts.tv_sec, ts.tv_nsec);
/*
* Serialize ereport generation
*/
mutex_enter(&spa->spa_errlist_lock);
#if 0
/*
* Determine the ENA to use for this event. If we are in a loading
* state, use a SPA-wide ENA. Otherwise, if we are in an I/O state, use
* a root zio-wide ENA. Otherwise, simply use a unique ENA.
*/
if (spa->spa_load_state != SPA_LOAD_NONE) {
#if 0
if (spa->spa_ena == 0)
spa->spa_ena = fm_ena_generate(0, FM_ENA_FMT1);
#endif
ena = spa->spa_ena;
} else if (zio != NULL && zio->io_logical != NULL) {
#if 0
if (zio->io_logical->io_ena == 0)
zio->io_logical->io_ena =
fm_ena_generate(0, FM_ENA_FMT1);
#endif
ena = zio->io_logical->io_ena;
} else {
#if 0
ena = fm_ena_generate(0, FM_ENA_FMT1);
#else
ena = 0;
#endif
}
#endif
/*
* Construct the full class, detector, and other standard FMA fields.
*/
sbuf_printf(&sb, " ereport_version=%u", FM_EREPORT_VERSION);
sbuf_printf(&sb, " class=%s.%s", ZFS_ERROR_CLASS, subclass);
sbuf_printf(&sb, " zfs_scheme_version=%u", FM_ZFS_SCHEME_VERSION);
/*
* Construct the per-ereport payload, depending on which parameters are
* passed in.
*/
/*
* If we are importing a faulted pool, then we treat it like an open,
* not an import. Otherwise, the DE will ignore all faults during
* import, since the default behavior is to mark the devices as
* persistently unavailable, not leave them in the faulted state.
*/
state = spa->spa_import_faulted ? SPA_LOAD_OPEN : spa->spa_load_state;
/*
* Generic payload members common to all ereports.
*/
sbuf_printf(&sb, " %s=%s", FM_EREPORT_PAYLOAD_ZFS_POOL, spa_name(spa));
sbuf_printf(&sb, " %s=%ju", FM_EREPORT_PAYLOAD_ZFS_POOL_GUID,
spa_guid(spa));
sbuf_printf(&sb, " %s=%d", FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, state);
if (spa != NULL) {
sbuf_printf(&sb, " %s=%s", FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE,
spa_get_failmode(spa) == ZIO_FAILURE_MODE_WAIT ?
FM_EREPORT_FAILMODE_WAIT :
spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE ?
FM_EREPORT_FAILMODE_CONTINUE : FM_EREPORT_FAILMODE_PANIC);
}
if (vd != NULL) {
vdev_t *pvd = vd->vdev_parent;
sbuf_printf(&sb, " %s=%ju", FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
vd->vdev_guid);
sbuf_printf(&sb, " %s=%s", FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
vd->vdev_ops->vdev_op_type);
if (vd->vdev_path)
sbuf_printf(&sb, " %s=%s",
FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH, vd->vdev_path);
if (vd->vdev_devid)
sbuf_printf(&sb, " %s=%s",
FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID, vd->vdev_devid);
if (pvd != NULL) {
sbuf_printf(&sb, " %s=%ju",
FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID, pvd->vdev_guid);
sbuf_printf(&sb, " %s=%s",
FM_EREPORT_PAYLOAD_ZFS_PARENT_TYPE,
pvd->vdev_ops->vdev_op_type);
if (pvd->vdev_path)
sbuf_printf(&sb, " %s=%s",
FM_EREPORT_PAYLOAD_ZFS_PARENT_PATH,
pvd->vdev_path);
if (pvd->vdev_devid)
sbuf_printf(&sb, " %s=%s",
FM_EREPORT_PAYLOAD_ZFS_PARENT_DEVID,
pvd->vdev_devid);
}
}
if (zio != NULL) {
/*
* Payload common to all I/Os.
*/
sbuf_printf(&sb, " %s=%u", FM_EREPORT_PAYLOAD_ZFS_ZIO_ERR,
zio->io_error);
/*
* If the 'size' parameter is non-zero, it indicates this is a
* RAID-Z or other I/O where the physical offset and length are
* provided for us, instead of within the zio_t.
*/
if (vd != NULL) {
if (size) {
sbuf_printf(&sb, " %s=%ju",
FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
stateoroffset);
sbuf_printf(&sb, " %s=%ju",
FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE, size);
} else {
sbuf_printf(&sb, " %s=%ju",
FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
zio->io_offset);
sbuf_printf(&sb, " %s=%ju",
FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
zio->io_size);
}
}
/*
* Payload for I/Os with corresponding logical information.
*/
if (zio->io_logical != NULL) {
sbuf_printf(&sb, " %s=%ju",
FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJECT,
zio->io_logical->io_bookmark.zb_object);
sbuf_printf(&sb, " %s=%ju",
FM_EREPORT_PAYLOAD_ZFS_ZIO_LEVEL,
zio->io_logical->io_bookmark.zb_level);
sbuf_printf(&sb, " %s=%ju",
FM_EREPORT_PAYLOAD_ZFS_ZIO_BLKID,
zio->io_logical->io_bookmark.zb_blkid);
}
} else if (vd != NULL) {
/*
* If we have a vdev but no zio, this is a device fault, and the
* 'stateoroffset' parameter indicates the previous state of the
* vdev.
*/
sbuf_printf(&sb, " %s=%ju", FM_EREPORT_PAYLOAD_ZFS_PREV_STATE,
stateoroffset);
}
mutex_exit(&spa->spa_errlist_lock);
sbuf_finish(&sb);
devctl_notify("ZFS", spa->spa_name, subclass, sbuf_data(&sb));
if (sbuf_overflowed(&sb))
printf("ZFS WARNING: sbuf overflowed\n");
sbuf_delete(&sb);
#endif
}
static void
zfs_post_common(spa_t *spa, vdev_t *vd, const char *name)
{
#ifdef _KERNEL
char buf[1024];
char class[64];
struct sbuf sb;
struct timespec ts;
nanotime(&ts);
sbuf_new(&sb, buf, sizeof(buf), SBUF_FIXEDLEN);
sbuf_printf(&sb, "time=%ju.%ld", (uintmax_t)ts.tv_sec, ts.tv_nsec);
snprintf(class, sizeof(class), "%s.%s.%s", FM_RSRC_RESOURCE,
ZFS_ERROR_CLASS, name);
sbuf_printf(&sb, " %s=%hhu", FM_VERSION, FM_RSRC_VERSION);
sbuf_printf(&sb, " %s=%s", FM_CLASS, class);
sbuf_printf(&sb, " %s=%ju", FM_EREPORT_PAYLOAD_ZFS_POOL_GUID,
spa_guid(spa));
if (vd)
sbuf_printf(&sb, " %s=%ju", FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
vd->vdev_guid);
sbuf_finish(&sb);
ZFS_LOG(1, "%s", sbuf_data(&sb));
devctl_notify("ZFS", spa->spa_name, class, sbuf_data(&sb));
if (sbuf_overflowed(&sb))
printf("ZFS WARNING: sbuf overflowed\n");
sbuf_delete(&sb);
#endif
}
/*
* The 'resource.fs.zfs.removed' event is an internal signal that the given vdev
* has been removed from the system. This will cause the DE to ignore any
* recent I/O errors, inferring that they are due to the asynchronous device
* removal.
*/
void
zfs_post_remove(spa_t *spa, vdev_t *vd)
{
zfs_post_common(spa, vd, FM_RESOURCE_REMOVED);
}
/*
* The 'resource.fs.zfs.autoreplace' event is an internal signal that the pool
* has the 'autoreplace' property set, and therefore any broken vdevs will be
* handled by higher level logic, and no vdev fault should be generated.
*/
void
zfs_post_autoreplace(spa_t *spa, vdev_t *vd)
{
zfs_post_common(spa, vd, FM_RESOURCE_AUTOREPLACE);
}
|