1S390 Debug Feature 2================== 3 4files: arch/s390/kernel/debug.c 5 arch/s390/include/asm/debug.h 6 7Description: 8------------ 9The goal of this feature is to provide a kernel debug logging API 10where log records can be stored efficiently in memory, where each component 11(e.g. device drivers) can have one separate debug log. 12One purpose of this is to inspect the debug logs after a production system crash 13in order to analyze the reason for the crash. 14If the system still runs but only a subcomponent which uses dbf fails, 15it is possible to look at the debug logs on a live system via the Linux 16debugfs filesystem. 17The debug feature may also very useful for kernel and driver development. 18 19Design: 20------- 21Kernel components (e.g. device drivers) can register themselves at the debug 22feature with the function call debug_register(). This function initializes a 23debug log for the caller. For each debug log exists a number of debug areas 24where exactly one is active at one time. Each debug area consists of contiguous 25pages in memory. In the debug areas there are stored debug entries (log records) 26which are written by event- and exception-calls. 27 28An event-call writes the specified debug entry to the active debug 29area and updates the log pointer for the active area. If the end 30of the active debug area is reached, a wrap around is done (ring buffer) 31and the next debug entry will be written at the beginning of the active 32debug area. 33 34An exception-call writes the specified debug entry to the log and 35switches to the next debug area. This is done in order to be sure 36that the records which describe the origin of the exception are not 37overwritten when a wrap around for the current area occurs. 38 39The debug areas themselves are also ordered in form of a ring buffer. 40When an exception is thrown in the last debug area, the following debug 41entries are then written again in the very first area. 42 43There are three versions for the event- and exception-calls: One for 44logging raw data, one for text and one for numbers. 45 46Each 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 54The debug logs can be inspected in a live system through entries in 55the debugfs-filesystem. Under the toplevel directory "s390dbf" there is 56a directory for each registered component, which is named like the 57corresponding component. The debugfs normally should be mounted to 58/sys/kernel/debug therefore the debug feature can be accessed under 59/sys/kernel/debug/s390dbf. 60 61The content of the directories are files which represent different views 62to the debug log. Each component can decide which views should be 63used through registering them with the function debug_register_view(). 64Predefined views for hex/ascii, sprintf and raw binary data are provided. 65It is also possible to define other views. The content of 66a view can be inspected simply by reading the corresponding debugfs file. 67 68All debug logs have an actual debug level (range from 0 to 6). 69The default level is 3. Event and Exception functions have a 'level' 70parameter. Only debug entries with a level that is lower or equal 71than the actual level are written to the log. This means, when 72writing events, high priority log entries should have a low level 73value whereas low priority entries should have a high one. 74The actual debug level can be changed with the help of the debugfs-filesystem 75through writing a number string "x" to the 'level' debugfs file which is 76provided for every debug log. Debugging can be switched off completely 77by using "-" on the 'level' debugfs file. 78 79Example: 80 81> echo "-" > /sys/kernel/debug/s390dbf/dasd/level 82 83It is also possible to deactivate the debug feature globally for every 84debug log. You can change the behavior using 2 sysctl parameters in 85/proc/sys/s390dbf: 86There are currently 2 possible triggers, which stop the debug feature 87globally. The first possibility is to use the "debug_active" sysctl. If 88set to 1 the debug feature is running. If "debug_active" is set to 0 the 89debug feature is turned off. 90The second trigger which stops the debug feature is a kernel oops. 91That prevents the debug feature from overwriting debug information that 92happened before the oops. After an oops you can reactivate the debug feature 93by piping 1 to /proc/sys/s390dbf/debug_active. Nevertheless, its not 94suggested to use an oopsed kernel in a production environment. 95If you want to disallow the deactivation of the debug feature, you can use 96the "debug_stoppable" sysctl. If you set "debug_stoppable" to 0 the debug 97feature cannot be stopped. If the debug feature is already stopped, it 98will stay deactivated. 99 100Kernel Interfaces: 101------------------ 102 103---------------------------------------------------------------------------- 104debug_info_t *debug_register(char *name, int pages, int nr_areas, 105 int buf_size); 106 107Parameter: name: Name of debug log (e.g. used for debugfs entry) 108 pages: number of pages, which will be allocated per area 109 nr_areas: number of debug areas 110 buf_size: size of data area in each debug entry 111 112Return Value: Handle for generated debug area 113 NULL if register failed 114 115Description: Allocates memory for a debug log 116 Must not be called within an interrupt handler 117 118---------------------------------------------------------------------------- 119debug_info_t *debug_register_mode(char *name, int pages, int nr_areas, 120 int buf_size, mode_t mode, uid_t uid, 121 gid_t gid); 122 123Parameter: name: Name of debug log (e.g. used for debugfs entry) 124 pages: Number of pages, which will be allocated per area 125 nr_areas: Number of debug areas 126 buf_size: Size of data area in each debug entry 127 mode: File mode for debugfs files. E.g. S_IRWXUGO 128 uid: User ID for debugfs files. Currently only 0 is 129 supported. 130 gid: Group ID for debugfs files. Currently only 0 is 131 supported. 132 133Return Value: Handle for generated debug area 134 NULL if register failed 135 136Description: Allocates memory for a debug log 137 Must not be called within an interrupt handler 138 139--------------------------------------------------------------------------- 140void debug_unregister (debug_info_t * id); 141 142Parameter: id: handle for debug log 143 144Return Value: none 145 146Description: frees memory for a debug log and removes all registered debug 147 views. 148 Must not be called within an interrupt handler 149 150--------------------------------------------------------------------------- 151void debug_set_level (debug_info_t * id, int new_level); 152 153Parameter: id: handle for debug log 154 new_level: new debug level 155 156Return Value: none 157 158Description: Sets new actual debug level if new_level is valid. 159 160--------------------------------------------------------------------------- 161bool debug_level_enabled (debug_info_t * id, int level); 162 163Parameter: id: handle for debug log 164 level: debug level 165 166Return Value: True if level is less or equal to the current debug level. 167 168Description: Returns true if debug events for the specified level would be 169 logged. Otherwise returns false. 170--------------------------------------------------------------------------- 171void debug_stop_all(void); 172 173Parameter: none 174 175Return Value: none 176 177Description: stops the debug feature if stopping is allowed. Currently 178 used in case of a kernel oops. 179 180--------------------------------------------------------------------------- 181debug_entry_t* debug_event (debug_info_t* id, int level, void* data, 182 int length); 183 184Parameter: id: handle for debug log 185 level: debug level 186 data: pointer to data for debug entry 187 length: length of data in bytes 188 189Return Value: Address of written debug entry 190 191Description: writes debug entry to active debug area (if level <= actual 192 debug level) 193 194--------------------------------------------------------------------------- 195debug_entry_t* debug_int_event (debug_info_t * id, int level, 196 unsigned int data); 197debug_entry_t* debug_long_event(debug_info_t * id, int level, 198 unsigned long data); 199 200Parameter: id: handle for debug log 201 level: debug level 202 data: integer value for debug entry 203 204Return Value: Address of written debug entry 205 206Description: writes debug entry to active debug area (if level <= actual 207 debug level) 208 209--------------------------------------------------------------------------- 210debug_entry_t* debug_text_event (debug_info_t * id, int level, 211 const char* data); 212 213Parameter: id: handle for debug log 214 level: debug level 215 data: string for debug entry 216 217Return Value: Address of written debug entry 218 219Description: writes debug entry in ascii format to active debug area 220 (if level <= actual debug level) 221 222--------------------------------------------------------------------------- 223debug_entry_t* debug_sprintf_event (debug_info_t * id, int level, 224 char* string,...); 225 226Parameter: id: handle for debug log 227 level: debug level 228 string: format string for debug entry 229 ...: varargs used as in sprintf() 230 231Return Value: Address of written debug entry 232 233Description: writes debug entry with format string and varargs (longs) to 234 active debug area (if level $<=$ actual debug level). 235 floats and long long datatypes cannot be used as varargs. 236 237--------------------------------------------------------------------------- 238 239debug_entry_t* debug_exception (debug_info_t* id, int level, void* data, 240 int length); 241 242Parameter: id: handle for debug log 243 level: debug level 244 data: pointer to data for debug entry 245 length: length of data in bytes 246 247Return Value: Address of written debug entry 248 249Description: writes debug entry to active debug area (if level <= actual 250 debug level) and switches to next debug area 251 252--------------------------------------------------------------------------- 253debug_entry_t* debug_int_exception (debug_info_t * id, int level, 254 unsigned int data); 255debug_entry_t* debug_long_exception(debug_info_t * id, int level, 256 unsigned long data); 257 258Parameter: id: handle for debug log 259 level: debug level 260 data: integer value for debug entry 261 262Return Value: Address of written debug entry 263 264Description: writes debug entry to active debug area (if level <= actual 265 debug level) and switches to next debug area 266 267--------------------------------------------------------------------------- 268debug_entry_t* debug_text_exception (debug_info_t * id, int level, 269 const char* data); 270 271Parameter: id: handle for debug log 272 level: debug level 273 data: string for debug entry 274 275Return Value: Address of written debug entry 276 277Description: writes debug entry in ascii format to active debug area 278 (if level <= actual debug level) and switches to next debug 279 area 280 281--------------------------------------------------------------------------- 282debug_entry_t* debug_sprintf_exception (debug_info_t * id, int level, 283 char* string,...); 284 285Parameter: id: handle for debug log 286 level: debug level 287 string: format string for debug entry 288 ...: varargs used as in sprintf() 289 290Return Value: Address of written debug entry 291 292Description: writes debug entry with format string and varargs (longs) to 293 active debug area (if level $<=$ actual debug level) and 294 switches to next debug area. 295 floats and long long datatypes cannot be used as varargs. 296 297--------------------------------------------------------------------------- 298 299int debug_register_view (debug_info_t * id, struct debug_view *view); 300 301Parameter: id: handle for debug log 302 view: pointer to debug view struct 303 304Return Value: 0 : ok 305 < 0: Error 306 307Description: registers new debug view and creates debugfs dir entry 308 309--------------------------------------------------------------------------- 310int debug_unregister_view (debug_info_t * id, struct debug_view *view); 311 312Parameter: id: handle for debug log 313 view: pointer to debug view struct 314 315Return Value: 0 : ok 316 < 0: Error 317 318Description: unregisters debug view and removes debugfs dir entry 319 320 321 322Predefined views: 323----------------- 324 325extern struct debug_view debug_hex_ascii_view; 326extern struct debug_view debug_raw_view; 327extern struct debug_view debug_sprintf_view; 328 329Examples 330-------- 331 332/* 333 * hex_ascii- + raw-view Example 334 */ 335 336#include <linux/init.h> 337#include <asm/debug.h> 338 339static debug_info_t* debug_info; 340 341static int init(void) 342{ 343 /* register 4 debug areas with one page each and 4 byte data field */ 344 345 debug_info = debug_register ("test", 1, 4, 4 ); 346 debug_register_view(debug_info,&debug_hex_ascii_view); 347 debug_register_view(debug_info,&debug_raw_view); 348 349 debug_text_event(debug_info, 4 , "one "); 350 debug_int_exception(debug_info, 4, 4711); 351 debug_event(debug_info, 3, &debug_info, 4); 352 353 return 0; 354} 355 356static void cleanup(void) 357{ 358 debug_unregister (debug_info); 359} 360 361module_init(init); 362module_exit(cleanup); 363 364--------------------------------------------------------------------------- 365 366/* 367 * sprintf-view Example 368 */ 369 370#include <linux/init.h> 371#include <asm/debug.h> 372 373static debug_info_t* debug_info; 374 375static int init(void) 376{ 377 /* register 4 debug areas with one page each and data field for */ 378 /* format string pointer + 2 varargs (= 3 * sizeof(long)) */ 379 380 debug_info = debug_register ("test", 1, 4, sizeof(long) * 3); 381 debug_register_view(debug_info,&debug_sprintf_view); 382 383 debug_sprintf_event(debug_info, 2 , "first event in %s:%i\n",__FILE__,__LINE__); 384 debug_sprintf_exception(debug_info, 1, "pointer to debug info: %p\n",&debug_info); 385 386 return 0; 387} 388 389static void cleanup(void) 390{ 391 debug_unregister (debug_info); 392} 393 394module_init(init); 395module_exit(cleanup); 396 397 398 399Debugfs Interface 400---------------- 401Views to the debug logs can be investigated through reading the corresponding 402debugfs-files: 403 404Example: 405 406> ls /sys/kernel/debug/s390dbf/dasd 407flush hex_ascii level pages raw 408> cat /sys/kernel/debug/s390dbf/dasd/hex_ascii | sort +1 40900 00974733272:680099 2 - 02 0006ad7e 07 ea 4a 90 | .... 41000 00974733272:682210 2 - 02 0006ade6 46 52 45 45 | FREE 41100 00974733272:682213 2 - 02 0006adf6 07 ea 4a 90 | .... 41200 00974733272:682281 1 * 02 0006ab08 41 4c 4c 43 | EXCP 41301 00974733272:682284 2 - 02 0006ab16 45 43 4b 44 | ECKD 41401 00974733272:682287 2 - 02 0006ab28 00 00 00 04 | .... 41501 00974733272:682289 2 - 02 0006ab3e 00 00 00 20 | ... 41601 00974733272:682297 2 - 02 0006ad7e 07 ea 4a 90 | .... 41701 00974733272:684384 2 - 00 0006ade6 46 52 45 45 | FREE 41801 00974733272:684388 2 - 00 0006adf6 07 ea 4a 90 | .... 419 420See section about predefined views for explanation of the above output! 421 422Changing the debug level 423------------------------ 424 425Example: 426 427 428> cat /sys/kernel/debug/s390dbf/dasd/level 4293 430> echo "5" > /sys/kernel/debug/s390dbf/dasd/level 431> cat /sys/kernel/debug/s390dbf/dasd/level 4325 433 434Flushing debug areas 435-------------------- 436Debug areas can be flushed with piping the number of the desired 437area (0...n) to the debugfs file "flush". When using "-" all debug areas 438are flushed. 439 440Examples: 441 4421. Flush debug area 0: 443> echo "0" > /sys/kernel/debug/s390dbf/dasd/flush 444 4452. Flush all debug areas: 446> echo "-" > /sys/kernel/debug/s390dbf/dasd/flush 447 448Changing the size of debug areas 449------------------------------------ 450It is possible the change the size of debug areas through piping 451the number of pages to the debugfs file "pages". The resize request will 452also flush the debug areas. 453 454Example: 455 456Define 4 pages for the debug areas of debug feature "dasd": 457> echo "4" > /sys/kernel/debug/s390dbf/dasd/pages 458 459Stooping the debug feature 460-------------------------- 461Example: 462 4631. Check if stopping is allowed 464> cat /proc/sys/s390dbf/debug_stoppable 4652. Stop debug feature 466> echo 0 > /proc/sys/s390dbf/debug_active 467 468lcrash Interface 469---------------- 470It is planned that the dump analysis tool lcrash gets an additional command 471's390dbf' to display all the debug logs. With this tool it will be possible 472to investigate the debug logs on a live system and with a memory dump after 473a system crash. 474 475Investigating raw memory 476------------------------ 477One last possibility to investigate the debug logs at a live 478system and after a system crash is to look at the raw memory 479under VM or at the Service Element. 480It is possible to find the anker of the debug-logs through 481the 'debug_area_first' symbol in the System map. Then one has 482to follow the correct pointers of the data-structures defined 483in debug.h and find the debug-areas in memory. 484Normally modules which use the debug feature will also have 485a global variable with the pointer to the debug-logs. Following 486this pointer it will also be possible to find the debug logs in 487memory. 488 489For this method it is recommended to use '16 * x + 4' byte (x = 0..n) 490for the length of the data field in debug_register() in 491order to see the debug entries well formatted. 492 493 494Predefined Views 495---------------- 496 497There are three predefined views: hex_ascii, raw and sprintf. 498The hex_ascii view shows the data field in hex and ascii representation 499(e.g. '45 43 4b 44 | ECKD'). 500The raw view returns a bytestream as the debug areas are stored in memory. 501 502The sprintf view formats the debug entries in the same way as the sprintf 503function would do. The sprintf event/exception functions write to the 504debug entry a pointer to the format string (size = sizeof(long)) 505and for each vararg a long value. So e.g. for a debug entry with a format 506string plus two varargs one would need to allocate a (3 * sizeof(long)) 507byte data area in the debug_register() function. 508 509IMPORTANT: Using "%s" in sprintf event functions is dangerous. You can only 510use "%s" in the sprintf event functions, if the memory for the passed string is 511available as long as the debug feature exists. The reason behind this is that 512due to performance considerations only a pointer to the string is stored in 513the debug feature. If you log a string that is freed afterwards, you will get 514an OOPS when inspecting the debug feature, because then the debug feature will 515access the already freed memory. 516 517NOTE: If using the sprintf view do NOT use other event/exception functions 518than the sprintf-event and -exception functions. 519 520The format of the hex_ascii and sprintf view is as follows: 521- Number of area 522- Timestamp (formatted as seconds and microseconds since 00:00:00 Coordinated 523 Universal Time (UTC), January 1, 1970) 524- level of debug entry 525- Exception flag (* = Exception) 526- Cpu-Number of calling task 527- Return Address to caller 528- data field 529 530The format of the raw view is: 531- Header as described in debug.h 532- datafield 533 534A typical line of the hex_ascii view will look like the following (first line 535is only for explanation and will not be displayed when 'cating' the view): 536 537area time level exception cpu caller data (hex + ascii) 538-------------------------------------------------------------------------- 53900 00964419409:440690 1 - 00 88023fe 540 541 542Defining views 543-------------- 544 545Views are specified with the 'debug_view' structure. There are defined 546callback functions which are used for reading and writing the debugfs files: 547 548struct debug_view { 549 char name[DEBUG_MAX_PROCF_LEN]; 550 debug_prolog_proc_t* prolog_proc; 551 debug_header_proc_t* header_proc; 552 debug_format_proc_t* format_proc; 553 debug_input_proc_t* input_proc; 554 void* private_data; 555}; 556 557where 558 559typedef int (debug_header_proc_t) (debug_info_t* id, 560 struct debug_view* view, 561 int area, 562 debug_entry_t* entry, 563 char* out_buf); 564 565typedef int (debug_format_proc_t) (debug_info_t* id, 566 struct debug_view* view, char* out_buf, 567 const char* in_buf); 568typedef int (debug_prolog_proc_t) (debug_info_t* id, 569 struct debug_view* view, 570 char* out_buf); 571typedef int (debug_input_proc_t) (debug_info_t* id, 572 struct debug_view* view, 573 struct file* file, const char* user_buf, 574 size_t in_buf_size, loff_t* offset); 575 576 577The "private_data" member can be used as pointer to view specific data. 578It is not used by the debug feature itself. 579 580The output when reading a debugfs file is structured like this: 581 582"prolog_proc output" 583 584"header_proc output 1" "format_proc output 1" 585"header_proc output 2" "format_proc output 2" 586"header_proc output 3" "format_proc output 3" 587... 588 589When a view is read from the debugfs, the Debug Feature calls the 590'prolog_proc' once for writing the prolog. 591Then 'header_proc' and 'format_proc' are called for each 592existing debug entry. 593 594The input_proc can be used to implement functionality when it is written to 595the view (e.g. like with 'echo "0" > /sys/kernel/debug/s390dbf/dasd/level). 596 597For header_proc there can be used the default function 598debug_dflt_header_fn() which is defined in debug.h. 599and which produces the same header output as the predefined views. 600E.g: 60100 00964419409:440761 2 - 00 88023ec 602 603In order to see how to use the callback functions check the implementation 604of the default views! 605 606Example 607 608#include <asm/debug.h> 609 610#define UNKNOWNSTR "data: %08x" 611 612const char* messages[] = 613{"This error...........\n", 614 "That error...........\n", 615 "Problem..............\n", 616 "Something went wrong.\n", 617 "Everything ok........\n", 618 NULL 619}; 620 621static int debug_test_format_fn( 622 debug_info_t * id, struct debug_view *view, 623 char *out_buf, const char *in_buf 624) 625{ 626 int i, rc = 0; 627 628 if(id->buf_size >= 4) { 629 int msg_nr = *((int*)in_buf); 630 if(msg_nr < sizeof(messages)/sizeof(char*) - 1) 631 rc += sprintf(out_buf, "%s", messages[msg_nr]); 632 else 633 rc += sprintf(out_buf, UNKNOWNSTR, msg_nr); 634 } 635 out: 636 return rc; 637} 638 639struct debug_view debug_test_view = { 640 "myview", /* name of view */ 641 NULL, /* no prolog */ 642 &debug_dflt_header_fn, /* default header for each entry */ 643 &debug_test_format_fn, /* our own format function */ 644 NULL, /* no input function */ 645 NULL /* no private data */ 646}; 647 648===== 649test: 650===== 651debug_info_t *debug_info; 652... 653debug_info = debug_register ("test", 0, 4, 4 )); 654debug_register_view(debug_info, &debug_test_view); 655for(i = 0; i < 10; i ++) debug_int_event(debug_info, 1, i); 656 657> cat /sys/kernel/debug/s390dbf/test/myview 65800 00964419734:611402 1 - 00 88042ca This error........... 65900 00964419734:611405 1 - 00 88042ca That error........... 66000 00964419734:611408 1 - 00 88042ca Problem.............. 66100 00964419734:611411 1 - 00 88042ca Something went wrong. 66200 00964419734:611414 1 - 00 88042ca Everything ok........ 66300 00964419734:611417 1 - 00 88042ca data: 00000005 66400 00964419734:611419 1 - 00 88042ca data: 00000006 66500 00964419734:611422 1 - 00 88042ca data: 00000007 66600 00964419734:611425 1 - 00 88042ca data: 00000008 66700 00964419734:611428 1 - 00 88042ca data: 00000009 668