Documentation/s390/s390dbf.txt

   1 S390 Debug Feature
   2 ==================
   3
   4 files: arch/s390/kernel/debug.c
   5        include/asm-s390/debug.h
   6
   7 Description:
   8 ------------
   9 The goal of this feature is to provide a kernel debug logging API
  10 where log records can be stored efficiently in memory, where each component
  11 (e.g. device drivers) can have one separate debug log.
  12 One purpose of this is to inspect the debug logs after a production system crash
  13 in order to analyze the reason for the crash.
  14 If the system still runs but only a subcomponent which uses dbf failes,
  15 it is possible to look at the debug logs on a live system via the Linux proc
  16 filesystem.
  17 The debug feature may also very useful for kernel and driver development.
  18
  19 Design:
  20 -------
  21 Kernel components (e.g. device drivers) can register themselves at the debug
  22 feature with the function call debug_register(). This function initializes a
  23 debug log for the caller. For each debug log exists a number of debug areas
  24 where exactly one is active at one time.  Each debug area consists of contiguous
  25 pages in memory. In the debug areas there are stored debug entries (log records)
  26 which are written by event- and exception-calls.
  27
  28 An event-call writes the specified debug entry to the active debug
  29 area and updates the log pointer for the active area. If the end
  30 of the active debug area is reached, a wrap around is done (ring buffer)
  31 and the next debug entry will be written at the beginning of the active
  32 debug area.
  33
  34 An exception-call writes the specified debug entry to the log and
  35 switches to the next debug area. This is done in order to be sure
  36 that the records which describe the origin of the exception are not
  37 overwritten when a wrap around for the current area occurs.
  38
  39 The debug areas itselve are also ordered in form of a ring buffer.
  40 When an exception is thrown in the last debug area, the following debug
  41 entries are then written again in the very first area.
  42
  43 There are three versions for the event- and exception-calls: One for
  44 logging raw data, one for text and one for numbers.
  45
  46 Each debug entry contains the following data:
  47
  48 - Timestamp
  49 - Cpu-Number of calling task
  50 - Level of debug entry (0...6)
  51 - Return Address to caller
  52 - Flag, if entry is an exception or not
  53
  54 The debug logs can be inspected in a live system through entries in
  55 the proc-filesystem. Under the path /proc/s390dbf there is
  56 a directory for each registered component, which is named like the
  57 corresponding component.
  58
  59 The content of the directories are files which represent different views
  60 to the debug log. Each component can decide which views should be
  61 used through registering them with the function debug_register_view().
  62 Predefined views for hex/ascii, sprintf and raw binary data are provided.
  63 It is also possible to define other views. The content of
  64 a view can be inspected simply by reading the corresponding proc file.
  65
  66 All debug logs have an an actual debug level (range from 0 to 6).
  67 The default level is 3. Event and Exception functions have a 'level'
  68 parameter. Only debug entries with a level that is lower or equal
  69 than the actual level are written to the log. This means, when
  70 writing events, high priority log entries should have a low level
  71 value whereas low priority entries should have a high one.
  72 The actual debug level can be changed with the help of the proc-filesystem
  73 through writing a number string "x" to the 'level' proc file which is
  74 provided for every debug log. Debugging can be switched off completely
  75 by using "-" on the 'level' proc file.
  76
  77 Example:
  78
  79 > echo "-" > /proc/s390dbf/dasd/level
  80
  81 It is also possible to deactivate the debug feature globally for every
  82 debug log. You can change the behavior using  2 sysctl parameters in
  83 /proc/sys/s390dbf:
  84 There are currently 2 possible triggers, which stop the  debug feature
  85 globally. The first possbility is to use the "debug_active" sysctl. If
  86 set to 1 the debug feature is running. If "debug_active" is set to 0 the
  87 debug feature is turned off.
  88 The second trigger which stops the debug feature is an kernel oops.
  89 That prevents the debug feature from overwriting debug information that
  90 happened before the oops. After an oops you can reactivate the debug feature
  91 by piping 1 to /proc/sys/s390dbf/debug_active. Nevertheless, its not
  92 suggested to use an oopsed kernel in an production environment.
  93 If you want to disallow the deactivation of the debug feature, you can use
  94 the "debug_stoppable" sysctl. If you set "debug_stoppable" to 0 the debug
  95 feature cannot be stopped. If the debug feature is already stopped, it
  96 will stay deactivated.
  97
  98 Kernel Interfaces:
  99 ------------------
 100
 101 ----------------------------------------------------------------------------
 102 debug_info_t *debug_register(char *name, int pages_index, int nr_areas,
 103                              int buf_size);
 104
 105 Parameter:    name:        Name of debug log (e.g. used for proc entry)
 106               pages_index: 2^pages_index pages will be allocated per area
 107               nr_areas:    number of debug areas
 108               buf_size:    size of data area in each debug entry
 109
 110 Return Value: Handle for generated debug area
 111               NULL if register failed
 112
 113 Description:  Allocates memory for a debug log
 114               Must not be called within an interrupt handler
 115
 116 ---------------------------------------------------------------------------
 117 void debug_unregister (debug_info_t * id);
 118
 119 Parameter:     id:   handle for debug log
 120
 121 Return Value:  none
 122
 123 Description:   frees memory for a debug log
 124                Must not be called within an interrupt handler
 125
 126 ---------------------------------------------------------------------------
 127 void debug_set_level (debug_info_t * id, int new_level);
 128
 129 Parameter:     id:        handle for debug log
 130                new_level: new debug level
 131
 132 Return Value:  none
 133
 134 Description:   Sets new actual debug level if new_level is valid.
 135
 136 ---------------------------------------------------------------------------
 137 +void debug_stop_all(void);
 138
 139 Parameter:     none
 140
 141 Return Value:  none
 142
 143 Description:   stops the debug feature if stopping is allowed. Currently
 144                used in case of a kernel oops.
 145
 146 ---------------------------------------------------------------------------
 147 debug_entry_t* debug_event (debug_info_t* id, int level, void* data,
 148                             int length);
 149
 150 Parameter:     id:     handle for debug log
 151                level:  debug level
 152                data:   pointer to data for debug entry
 153                length: length of data in bytes
 154
 155 Return Value:  Address of written debug entry
 156
 157 Description:   writes debug entry to active debug area (if level <= actual
 158                debug level)
 159
 160 ---------------------------------------------------------------------------
 161 debug_entry_t* debug_int_event (debug_info_t * id, int level,
 162                                 unsigned int data);
 163 debug_entry_t* debug_long_event(debug_info_t * id, int level,
 164                                 unsigned long data);
 165
 166 Parameter:     id:     handle for debug log
 167                level:  debug level
 168                data:   integer value for debug entry
 169
 170 Return Value:  Address of written debug entry
 171
 172 Description:   writes debug entry to active debug area (if level <= actual
 173                debug level)
 174
 175 ---------------------------------------------------------------------------
 176 debug_entry_t* debug_text_event (debug_info_t * id, int level,
 177                                  const char* data);
 178
 179 Parameter:     id:     handle for debug log
 180                level:  debug level
 181                data:   string for debug entry
 182
 183 Return Value:  Address of written debug entry
 184
 185 Description:   writes debug entry in ascii format to active debug area
 186                (if level <= actual debug level)
 187
 188 ---------------------------------------------------------------------------
 189 debug_entry_t* debug_sprintf_event (debug_info_t * id, int level,
 190                                     char* string,...);
 191
 192 Parameter:     id:    handle for debug log
 193                level: debug level
 194                string: format string for debug entry
 195                ...: varargs used as in sprintf()
 196
 197 Return Value:  Address of written debug entry
 198
 199 Description:   writes debug entry with format string and varargs (longs) to
 200                active debug area (if level $<=$ actual debug level).
 201                floats and long long datatypes cannot be used as varargs.
 202
 203 ---------------------------------------------------------------------------
 204
 205 debug_entry_t* debug_exception (debug_info_t* id, int level, void* data,
 206                                 int length);
 207
 208 Parameter:     id:     handle for debug log
 209                level:  debug level
 210                data:   pointer to data for debug entry
 211                length: length of data in bytes
 212
 213 Return Value:  Address of written debug entry
 214
 215 Description:   writes debug entry to active debug area (if level <= actual
 216                debug level) and switches to next debug area
 217
 218 ---------------------------------------------------------------------------
 219 debug_entry_t* debug_int_exception (debug_info_t * id, int level,
 220                                     unsigned int data);
 221 debug_entry_t* debug_long_exception(debug_info_t * id, int level,
 222                                     unsigned long data);
 223
 224 Parameter:     id:     handle for debug log
 225                level:  debug level
 226                data:   integer value for debug entry
 227
 228 Return Value:  Address of written debug entry
 229
 230 Description:   writes debug entry to active debug area (if level <= actual
 231                debug level) and switches to next debug area
 232
 233 ---------------------------------------------------------------------------
 234 debug_entry_t* debug_text_exception (debug_info_t * id, int level,
 235                                      const char* data);
 236
 237 Parameter:     id:     handle for debug log
 238                level:  debug level
 239                data:   string for debug entry
 240
 241 Return Value:  Address of written debug entry
 242
 243 Description:   writes debug entry in ascii format to active debug area
 244                (if level <= actual debug level) and switches to next debug
 245                area
 246
 247 ---------------------------------------------------------------------------
 248 debug_entry_t* debug_sprintf_exception (debug_info_t * id, int level,
 249                                         char* string,...);
 250
 251 Parameter:     id:    handle for debug log
 252                level: debug level
 253                string: format string for debug entry
 254                ...: varargs used as in sprintf()
 255
 256 Return Value:  Address of written debug entry
 257
 258 Description:   writes debug entry with format string and varargs (longs) to
 259                active debug area (if level $<=$ actual debug level) and
 260                switches to next debug area.
 261                floats and long long datatypes cannot be used as varargs.
 262
 263 ---------------------------------------------------------------------------
 264
 265 int debug_register_view (debug_info_t * id, struct debug_view *view);
 266
 267 Parameter:     id:    handle for debug log
 268                view:  pointer to debug view struct
 269
 270 Return Value:  0  : ok
 271                < 0: Error
 272
 273 Description:   registers new debug view and creates proc dir entry
 274
 275 ---------------------------------------------------------------------------
 276 int debug_unregister_view (debug_info_t * id, struct debug_view *view);
 277
 278 Parameter:     id:    handle for debug log
 279                view:  pointer to debug view struct
 280
 281 Return Value:  0  : ok
 282                < 0: Error
 283
 284 Description:   unregisters debug view and removes proc dir entry
 285
 286
 287
 288 Predefined views:
 289 -----------------
 290
 291 extern struct debug_view debug_hex_ascii_view;
 292 extern struct debug_view debug_raw_view;
 293 extern struct debug_view debug_sprintf_view;
 294
 295 Examples
 296 --------
 297
 298 /*
 299  * hex_ascii- + raw-view Example
 300  */
 301
 302 #include <linux/init.h>
 303 #include <asm/debug.h>
 304
 305 static debug_info_t* debug_info;
 306
 307 static int init(void)
 308 {
 309     /* register 4 debug areas with one page each and 4 byte data field */
 310
 311     debug_info = debug_register ("test", 0, 4, 4 );
 312     debug_register_view(debug_info,&debug_hex_ascii_view);
 313     debug_register_view(debug_info,&debug_raw_view);
 314
 315     debug_text_event(debug_info, 4 , "one ");
 316     debug_int_exception(debug_info, 4, 4711);
 317     debug_event(debug_info, 3, &debug_info, 4);
 318
 319     return 0;
 320 }
 321
 322 static void cleanup(void)
 323 {
 324     debug_unregister (debug_info);
 325 }
 326
 327 module_init(init);
 328 module_exit(cleanup);
 329
 330 ---------------------------------------------------------------------------
 331
 332 /*
 333  * sprintf-view Example
 334  */
 335
 336 #include <linux/init.h>
 337 #include <asm/debug.h>
 338
 339 static debug_info_t* debug_info;
 340
 341 static int init(void)
 342 {
 343     /* register 4 debug areas with one page each and data field for */
 344     /* format string pointer + 2 varargs (= 3 * sizeof(long))       */
 345
 346     debug_info = debug_register ("test", 0, 4, sizeof(long) * 3);
 347     debug_register_view(debug_info,&debug_sprintf_view);
 348
 349     debug_sprintf_event(debug_info, 2 , "first event in %s:%i\n",__FILE__,__LINE__);
 350     debug_sprintf_exception(debug_info, 1, "pointer to debug info: %p\n",&debug_info);
 351
 352     return 0;
 353 }
 354
 355 static void cleanup(void)
 356 {
 357     debug_unregister (debug_info);
 358 }
 359
 360 module_init(init);
 361 module_exit(cleanup);
 362
 363
 364
 365 ProcFS Interface
 366 ----------------
 367 Views to the debug logs can be investigated through reading the corresponding
 368 proc-files:
 369
 370 Example:
 371
 372 > ls /proc/s390dbf/dasd
 373 flush  hex_ascii  level      raw
 374 > cat /proc/s390dbf/dasd/hex_ascii | sort +1
 375 00 00974733272:680099 2 - 02 0006ad7e  07 ea 4a 90 | ....
 376 00 00974733272:682210 2 - 02 0006ade6  46 52 45 45 | FREE
 377 00 00974733272:682213 2 - 02 0006adf6  07 ea 4a 90 | ....
 378 00 00974733272:682281 1 * 02 0006ab08  41 4c 4c 43 | EXCP
 379 01 00974733272:682284 2 - 02 0006ab16  45 43 4b 44 | ECKD
 380 01 00974733272:682287 2 - 02 0006ab28  00 00 00 04 | ....
 381 01 00974733272:682289 2 - 02 0006ab3e  00 00 00 20 | ...
 382 01 00974733272:682297 2 - 02 0006ad7e  07 ea 4a 90 | ....
 383 01 00974733272:684384 2 - 00 0006ade6  46 52 45 45 | FREE
 384 01 00974733272:684388 2 - 00 0006adf6  07 ea 4a 90 | ....
 385
 386 See section about predefined views for explanation of the above output!
 387
 388 Changing the debug level
 389 ------------------------
 390
 391 Example:
 392
 393
 394 > cat /proc/s390dbf/dasd/level
 395 3
 396 > echo "5" > /proc/s390dbf/dasd/level
 397 > cat /proc/s390dbf/dasd/level
 398 5
 399
 400 Flushing debug areas
 401 --------------------
 402 Debug areas can be flushed with piping the number of the desired
 403 area (0...n) to the proc file "flush". When using "-" all debug areas
 404 are flushed.
 405
 406 Examples:
 407
 408 1. Flush debug area 0:
 409 > echo "0" > /proc/s390dbf/dasd/flush
 410
 411 2. Flush all debug areas:
 412 > echo "-" > /proc/s390dbf/dasd/flush
 413
 414 Stooping the debug feature
 415 --------------------------
 416 Example:
 417
 418 1. Check if stopping is allowed
 419 > cat /proc/sys/s390dbf/debug_stoppable
 420 2. Stop debug feature
 421 > echo 0 > /proc/sys/s390dbf/debug_active
 422
 423 lcrash Interface
 424 ----------------
 425 It is planned that the dump analysis tool lcrash gets an additional command
 426 's390dbf' to display all the debug logs. With this tool it will be possible
 427 to investigate the debug logs on a live system and with a memory dump after
 428 a system crash.
 429
 430 Investigating raw memory
 431 ------------------------
 432 One last possibility to investigate the debug logs at a live
 433 system and after a system crash is to look at the raw memory
 434 under VM or at the Service Element.
 435 It is possible to find the anker of the debug-logs through
 436 the 'debug_area_first' symbol in the System map. Then one has
 437 to follow the correct pointers of the data-structures defined
 438 in debug.h and find the debug-areas in memory.
 439 Normally modules which use the debug feature will also have
 440 a global variable with the pointer to the debug-logs. Following
 441 this pointer it will also be possible to find the debug logs in
 442 memory.
 443
 444 For this method it is recommended to use '16 * x + 4' byte (x = 0..n)
 445 for the length of the data field in debug_register() in
 446 order to see the debug entries well formatted.
 447
 448
 449 Predefined Views
 450 ----------------
 451
 452 There are three predefined views: hex_ascii, raw and sprintf.
 453 The hex_ascii view shows the data field in hex and ascii representation
 454 (e.g. '45 43 4b 44 | ECKD').
 455 The raw view returns a bytestream as the debug areas are stored in memory.
 456
 457 The sprintf view formats the debug entries in the same way as the sprintf
 458 function would do. The sprintf event/expection fuctions write to the
 459 debug entry a pointer to the format string (size = sizeof(long))
 460 and for each vararg a long value. So e.g. for a debug entry with a format
 461 string plus two varargs one would need to allocate a (3 * sizeof(long))
 462 byte data area in the debug_register() function.
 463
 464
 465 NOTE: If using the sprintf view do NOT use other event/exception functions
 466 than the sprintf-event and -exception functions.
 467
 468 The format of the hex_ascii and sprintf view is as follows:
 469 - Number of area
 470 - Timestamp (formatted as seconds and microseconds since 00:00:00 Coordinated
 471   Universal Time (UTC), January 1, 1970)
 472 - level of debug entry
 473 - Exception flag (* = Exception)
 474 - Cpu-Number of calling task
 475 - Return Address to caller
 476 - data field
 477
 478 The format of the raw view is:
 479 - Header as described in debug.h
 480 - datafield
 481
 482 A typical line of the hex_ascii view will look like the following (first line
 483 is only for explanation and will not be displayed when 'cating' the view):
 484
 485 area  time           level exception cpu caller    data (hex + ascii)
 486 --------------------------------------------------------------------------
 487 00    00964419409:440690 1 -         00  88023fe
 488
 489
 490 Defining views
 491 --------------
 492
 493 Views are specified with the 'debug_view' structure. There are defined
 494 callback functions which are used for reading and writing the proc files:
 495
 496 struct debug_view {
 497         char name[DEBUG_MAX_PROCF_LEN];
 498         debug_prolog_proc_t* prolog_proc;
 499         debug_header_proc_t* header_proc;
 500         debug_format_proc_t* format_proc;
 501         debug_input_proc_t*  input_proc;
 502         void*                private_data;
 503 };
 504
 505 where
 506
 507 typedef int (debug_header_proc_t) (debug_info_t* id,
 508                                    struct debug_view* view,
 509                                    int area,
 510                                    debug_entry_t* entry,
 511                                    char* out_buf);
 512
 513 typedef int (debug_format_proc_t) (debug_info_t* id,
 514                                    struct debug_view* view, char* out_buf,
 515                                    const char* in_buf);
 516 typedef int (debug_prolog_proc_t) (debug_info_t* id,
 517                                    struct debug_view* view,
 518                                    char* out_buf);
 519 typedef int (debug_input_proc_t) (debug_info_t* id,
 520                                   struct debug_view* view,
 521                                   struct file* file, const char* user_buf,
 522                                   size_t in_buf_size, loff_t* offset);
 523
 524
 525 The "private_data" member can be used as pointer to view specific data.
 526 It is not used by the debug feature itself.
 527
 528 The output when reading a debug-proc file is structured like this:
 529
 530 "prolog_proc output"
 531
 532 "header_proc output 1"  "format_proc output 1"
 533 "header_proc output 2"  "format_proc output 2"
 534 "header_proc output 3"  "format_proc output 3"
 535 ...
 536
 537 When a view is read from the proc fs, the Debug Feature calls the
 538 'prolog_proc' once for writing the prolog.
 539 Then 'header_proc' and 'format_proc' are called for each
 540 existing debug entry.
 541
 542 The input_proc can be used to implement functionality when it is written to
 543 the view (e.g. like with 'echo "0" > /proc/s390dbf/dasd/level).
 544
 545 For header_proc there can be used the default function
 546 debug_dflt_header_fn() which is defined in in debug.h.
 547 and which produces the same header output as the predefined views.
 548 E.g:
 549 00 00964419409:440761 2 - 00 88023ec
 550
 551 In order to see how to use the callback functions check the implementation
 552 of the default views!
 553
 554 Example
 555
 556 #include <asm/debug.h>
 557
 558 #define UNKNOWNSTR "data: %08x"
 559
 560 const char* messages[] =
 561 {"This error...........\n",
 562  "That error...........\n",
 563  "Problem..............\n",
 564  "Something went wrong.\n",
 565  "Everything ok........\n",
 566  NULL
 567 };
 568
 569 static int debug_test_format_fn(
 570    debug_info_t * id, struct debug_view *view,
 571    char *out_buf, const char *in_buf
 572 )
 573 {
 574   int i, rc = 0;
 575
 576   if(id->buf_size >= 4) {
 577      int msg_nr = *((int*)in_buf);
 578      if(msg_nr < sizeof(messages)/sizeof(char*) - 1)
 579         rc += sprintf(out_buf, "%s", messages[msg_nr]);
 580      else
 581         rc += sprintf(out_buf, UNKNOWNSTR, msg_nr);
 582   }
 583  out:
 584    return rc;
 585 }
 586
 587 struct debug_view debug_test_view = {
 588   "myview",                 /* name of view */
 589   NULL,                     /* no prolog */
 590   &debug_dflt_header_fn,    /* default header for each entry */
 591   &debug_test_format_fn,    /* our own format function */
 592   NULL,                     /* no input function */
 593   NULL                      /* no private data */
 594 };
 595
 596 =====
 597 test:
 598 =====
 599 debug_info_t *debug_info;
 600 ...
 601 debug_info = debug_register ("test", 0, 4, 4 ));
 602 debug_register_view(debug_info, &debug_test_view);
 603 for(i = 0; i < 10; i ++) debug_int_event(debug_info, 1, i);
 604
 605 > cat /proc/s390dbf/test/myview
 606 00 00964419734:611402 1 - 00 88042ca   This error...........
 607 00 00964419734:611405 1 - 00 88042ca   That error...........
 608 00 00964419734:611408 1 - 00 88042ca   Problem..............
 609 00 00964419734:611411 1 - 00 88042ca   Something went wrong.
 610 00 00964419734:611414 1 - 00 88042ca   Everything ok........
 611 00 00964419734:611417 1 - 00 88042ca   data: 00000005
 612 00 00964419734:611419 1 - 00 88042ca   data: 00000006
 613 00 00964419734:611422 1 - 00 88042ca   data: 00000007
 614 00 00964419734:611425 1 - 00 88042ca   data: 00000008
 615 00 00964419734:611428 1 - 00 88042ca   data: 00000009