479 lines · plain
1==================2S390 Debug Feature3==================4 5files:6 - arch/s390/kernel/debug.c7 - arch/s390/include/asm/debug.h8 9Description:10------------11The goal of this feature is to provide a kernel debug logging API12where log records can be stored efficiently in memory, where each component13(e.g. device drivers) can have one separate debug log.14One purpose of this is to inspect the debug logs after a production system crash15in order to analyze the reason for the crash.16 17If the system still runs but only a subcomponent which uses dbf fails,18it is possible to look at the debug logs on a live system via the Linux19debugfs filesystem.20 21The debug feature may also very useful for kernel and driver development.22 23Design:24-------25Kernel components (e.g. device drivers) can register themselves at the debug26feature with the function call :c:func:`debug_register()`.27This function initializes a28debug log for the caller. For each debug log exists a number of debug areas29where exactly one is active at one time. Each debug area consists of contiguous30pages in memory. In the debug areas there are stored debug entries (log records)31which are written by event- and exception-calls.32 33An event-call writes the specified debug entry to the active debug34area and updates the log pointer for the active area. If the end35of the active debug area is reached, a wrap around is done (ring buffer)36and the next debug entry will be written at the beginning of the active37debug area.38 39An exception-call writes the specified debug entry to the log and40switches to the next debug area. This is done in order to be sure41that the records which describe the origin of the exception are not42overwritten when a wrap around for the current area occurs.43 44The debug areas themselves are also ordered in form of a ring buffer.45When an exception is thrown in the last debug area, the following debug46entries are then written again in the very first area.47 48There are four versions for the event- and exception-calls: One for49logging raw data, one for text, one for numbers (unsigned int and long),50and one for sprintf-like formatted strings.51 52Each debug entry contains the following data:53 54- Timestamp55- Cpu-Number of calling task56- Level of debug entry (0...6)57- Return Address to caller58- Flag, if entry is an exception or not59 60The debug logs can be inspected in a live system through entries in61the debugfs-filesystem. Under the toplevel directory "``s390dbf``" there is62a directory for each registered component, which is named like the63corresponding component. The debugfs normally should be mounted to64``/sys/kernel/debug`` therefore the debug feature can be accessed under65``/sys/kernel/debug/s390dbf``.66 67The content of the directories are files which represent different views68to the debug log. Each component can decide which views should be69used through registering them with the function :c:func:`debug_register_view()`.70Predefined views for hex/ascii and sprintf data are provided.71It is also possible to define other views. The content of72a view can be inspected simply by reading the corresponding debugfs file.73 74All debug logs have an actual debug level (range from 0 to 6).75The default level is 3. Event and Exception functions have a :c:data:`level`76parameter. Only debug entries with a level that is lower or equal77than the actual level are written to the log. This means, when78writing events, high priority log entries should have a low level79value whereas low priority entries should have a high one.80The actual debug level can be changed with the help of the debugfs-filesystem81through writing a number string "x" to the ``level`` debugfs file which is82provided for every debug log. Debugging can be switched off completely83by using "-" on the ``level`` debugfs file.84 85Example::86 87 > echo "-" > /sys/kernel/debug/s390dbf/dasd/level88 89It is also possible to deactivate the debug feature globally for every90debug log. You can change the behavior using 2 sysctl parameters in91``/proc/sys/s390dbf``:92 93There are currently 2 possible triggers, which stop the debug feature94globally. The first possibility is to use the ``debug_active`` sysctl. If95set to 1 the debug feature is running. If ``debug_active`` is set to 0 the96debug feature is turned off.97 98The second trigger which stops the debug feature is a kernel oops.99That prevents the debug feature from overwriting debug information that100happened before the oops. After an oops you can reactivate the debug feature101by piping 1 to ``/proc/sys/s390dbf/debug_active``. Nevertheless, it's not102suggested to use an oopsed kernel in a production environment.103 104If you want to disallow the deactivation of the debug feature, you can use105the ``debug_stoppable`` sysctl. If you set ``debug_stoppable`` to 0 the debug106feature cannot be stopped. If the debug feature is already stopped, it107will stay deactivated.108 109Kernel Interfaces:110------------------111 112.. kernel-doc:: arch/s390/kernel/debug.c113.. kernel-doc:: arch/s390/include/asm/debug.h114 115Predefined views:116-----------------117 118.. code-block:: c119 120 extern struct debug_view debug_hex_ascii_view;121 122 extern struct debug_view debug_sprintf_view;123 124Examples125--------126 127.. code-block:: c128 129 /*130 * hex_ascii-view Example131 */132 133 #include <linux/init.h>134 #include <asm/debug.h>135 136 static debug_info_t *debug_info;137 138 static int init(void)139 {140 /* register 4 debug areas with one page each and 4 byte data field */141 142 debug_info = debug_register("test", 1, 4, 4 );143 debug_register_view(debug_info, &debug_hex_ascii_view);144 145 debug_text_event(debug_info, 4 , "one ");146 debug_int_exception(debug_info, 4, 4711);147 debug_event(debug_info, 3, &debug_info, 4);148 149 return 0;150 }151 152 static void cleanup(void)153 {154 debug_unregister(debug_info);155 }156 157 module_init(init);158 module_exit(cleanup);159 160.. code-block:: c161 162 /*163 * sprintf-view Example164 */165 166 #include <linux/init.h>167 #include <asm/debug.h>168 169 static debug_info_t *debug_info;170 171 static int init(void)172 {173 /* register 4 debug areas with one page each and data field for */174 /* format string pointer + 2 varargs (= 3 * sizeof(long)) */175 176 debug_info = debug_register("test", 1, 4, sizeof(long) * 3);177 debug_register_view(debug_info, &debug_sprintf_view);178 179 debug_sprintf_event(debug_info, 2 , "first event in %s:%i\n",__FILE__,__LINE__);180 debug_sprintf_exception(debug_info, 1, "pointer to debug info: %p\n",&debug_info);181 182 return 0;183 }184 185 static void cleanup(void)186 {187 debug_unregister(debug_info);188 }189 190 module_init(init);191 module_exit(cleanup);192 193Debugfs Interface194-----------------195Views to the debug logs can be investigated through reading the corresponding196debugfs-files:197 198Example::199 200 > ls /sys/kernel/debug/s390dbf/dasd201 flush hex_ascii level pages202 > cat /sys/kernel/debug/s390dbf/dasd/hex_ascii | sort -k2,2 -s203 00 00974733272:680099 2 - 02 0006ad7e 07 ea 4a 90 | ....204 00 00974733272:682210 2 - 02 0006ade6 46 52 45 45 | FREE205 00 00974733272:682213 2 - 02 0006adf6 07 ea 4a 90 | ....206 00 00974733272:682281 1 * 02 0006ab08 41 4c 4c 43 | EXCP207 01 00974733272:682284 2 - 02 0006ab16 45 43 4b 44 | ECKD208 01 00974733272:682287 2 - 02 0006ab28 00 00 00 04 | ....209 01 00974733272:682289 2 - 02 0006ab3e 00 00 00 20 | ...210 01 00974733272:682297 2 - 02 0006ad7e 07 ea 4a 90 | ....211 01 00974733272:684384 2 - 00 0006ade6 46 52 45 45 | FREE212 01 00974733272:684388 2 - 00 0006adf6 07 ea 4a 90 | ....213 214See section about predefined views for explanation of the above output!215 216Changing the debug level217------------------------218 219Example::220 221 222 > cat /sys/kernel/debug/s390dbf/dasd/level223 3224 > echo "5" > /sys/kernel/debug/s390dbf/dasd/level225 > cat /sys/kernel/debug/s390dbf/dasd/level226 5227 228Flushing debug areas229--------------------230Debug areas can be flushed with piping the number of the desired231area (0...n) to the debugfs file "flush". When using "-" all debug areas232are flushed.233 234Examples:235 2361. Flush debug area 0::237 238 > echo "0" > /sys/kernel/debug/s390dbf/dasd/flush239 2402. Flush all debug areas::241 242 > echo "-" > /sys/kernel/debug/s390dbf/dasd/flush243 244Changing the size of debug areas245------------------------------------246It is possible the change the size of debug areas through piping247the number of pages to the debugfs file "pages". The resize request will248also flush the debug areas.249 250Example:251 252Define 4 pages for the debug areas of debug feature "dasd"::253 254 > echo "4" > /sys/kernel/debug/s390dbf/dasd/pages255 256Stopping the debug feature257--------------------------258Example:259 2601. Check if stopping is allowed::261 262 > cat /proc/sys/s390dbf/debug_stoppable263 2642. Stop debug feature::265 266 > echo 0 > /proc/sys/s390dbf/debug_active267 268crash Interface269----------------270The ``crash`` tool since v5.1.0 has a built-in command271``s390dbf`` to display all the debug logs or export them to the file system.272With this tool it is possible273to investigate the debug logs on a live system and with a memory dump after274a system crash.275 276Investigating raw memory277------------------------278One last possibility to investigate the debug logs at a live279system and after a system crash is to look at the raw memory280under VM or at the Service Element.281It is possible to find the anchor of the debug-logs through282the ``debug_area_first`` symbol in the System map. Then one has283to follow the correct pointers of the data-structures defined284in debug.h and find the debug-areas in memory.285Normally modules which use the debug feature will also have286a global variable with the pointer to the debug-logs. Following287this pointer it will also be possible to find the debug logs in288memory.289 290For this method it is recommended to use '16 * x + 4' byte (x = 0..n)291for the length of the data field in :c:func:`debug_register()` in292order to see the debug entries well formatted.293 294 295Predefined Views296----------------297 298There are two predefined views: hex_ascii and sprintf.299The hex_ascii view shows the data field in hex and ascii representation300(e.g. ``45 43 4b 44 | ECKD``).301 302The sprintf view formats the debug entries in the same way as the sprintf303function would do. The sprintf event/exception functions write to the304debug entry a pointer to the format string (size = sizeof(long))305and for each vararg a long value. So e.g. for a debug entry with a format306string plus two varargs one would need to allocate a (3 * sizeof(long))307byte data area in the debug_register() function.308 309IMPORTANT:310 Using "%s" in sprintf event functions is dangerous. You can only311 use "%s" in the sprintf event functions, if the memory for the passed string312 is available as long as the debug feature exists. The reason behind this is313 that due to performance considerations only a pointer to the string is stored314 in the debug feature. If you log a string that is freed afterwards, you will315 get an OOPS when inspecting the debug feature, because then the debug feature316 will access the already freed memory.317 318NOTE:319 If using the sprintf view do NOT use other event/exception functions320 than the sprintf-event and -exception functions.321 322The format of the hex_ascii and sprintf view is as follows:323 324- Number of area325- Timestamp (formatted as seconds and microseconds since 00:00:00 Coordinated326 Universal Time (UTC), January 1, 1970)327- level of debug entry328- Exception flag (* = Exception)329- Cpu-Number of calling task330- Return Address to caller331- data field332 333A typical line of the hex_ascii view will look like the following (first line334is only for explanation and will not be displayed when 'cating' the view)::335 336 area time level exception cpu caller data (hex + ascii)337 --------------------------------------------------------------------------338 00 00964419409:440690 1 - 00 88023fe339 340 341Defining views342--------------343 344Views are specified with the 'debug_view' structure. There are defined345callback functions which are used for reading and writing the debugfs files:346 347.. code-block:: c348 349 struct debug_view {350 char name[DEBUG_MAX_PROCF_LEN];351 debug_prolog_proc_t* prolog_proc;352 debug_header_proc_t* header_proc;353 debug_format_proc_t* format_proc;354 debug_input_proc_t* input_proc;355 void* private_data;356 };357 358where:359 360.. code-block:: c361 362 typedef int (debug_header_proc_t) (debug_info_t* id,363 struct debug_view* view,364 int area,365 debug_entry_t* entry,366 char* out_buf);367 368 typedef int (debug_format_proc_t) (debug_info_t* id,369 struct debug_view* view, char* out_buf,370 const char* in_buf);371 typedef int (debug_prolog_proc_t) (debug_info_t* id,372 struct debug_view* view,373 char* out_buf);374 typedef int (debug_input_proc_t) (debug_info_t* id,375 struct debug_view* view,376 struct file* file, const char* user_buf,377 size_t in_buf_size, loff_t* offset);378 379 380The "private_data" member can be used as pointer to view specific data.381It is not used by the debug feature itself.382 383The output when reading a debugfs file is structured like this::384 385 "prolog_proc output"386 387 "header_proc output 1" "format_proc output 1"388 "header_proc output 2" "format_proc output 2"389 "header_proc output 3" "format_proc output 3"390 ...391 392When a view is read from the debugfs, the Debug Feature calls the393'prolog_proc' once for writing the prolog.394Then 'header_proc' and 'format_proc' are called for each395existing debug entry.396 397The input_proc can be used to implement functionality when it is written to398the view (e.g. like with ``echo "0" > /sys/kernel/debug/s390dbf/dasd/level``).399 400For header_proc there can be used the default function401:c:func:`debug_dflt_header_fn()` which is defined in debug.h.402and which produces the same header output as the predefined views.403E.g::404 405 00 00964419409:440761 2 - 00 88023ec406 407In order to see how to use the callback functions check the implementation408of the default views!409 410Example:411 412.. code-block:: c413 414 #include <asm/debug.h>415 416 #define UNKNOWNSTR "data: %08x"417 418 const char* messages[] =419 {"This error...........\n",420 "That error...........\n",421 "Problem..............\n",422 "Something went wrong.\n",423 "Everything ok........\n",424 NULL425 };426 427 static int debug_test_format_fn(428 debug_info_t *id, struct debug_view *view,429 char *out_buf, const char *in_buf430 )431 {432 int i, rc = 0;433 434 if (id->buf_size >= 4) {435 int msg_nr = *((int*)in_buf);436 if (msg_nr < sizeof(messages) / sizeof(char*) - 1)437 rc += sprintf(out_buf, "%s", messages[msg_nr]);438 else439 rc += sprintf(out_buf, UNKNOWNSTR, msg_nr);440 }441 return rc;442 }443 444 struct debug_view debug_test_view = {445 "myview", /* name of view */446 NULL, /* no prolog */447 &debug_dflt_header_fn, /* default header for each entry */448 &debug_test_format_fn, /* our own format function */449 NULL, /* no input function */450 NULL /* no private data */451 };452 453test:454=====455 456.. code-block:: c457 458 debug_info_t *debug_info;459 int i;460 ...461 debug_info = debug_register("test", 0, 4, 4);462 debug_register_view(debug_info, &debug_test_view);463 for (i = 0; i < 10; i ++)464 debug_int_event(debug_info, 1, i);465 466::467 468 > cat /sys/kernel/debug/s390dbf/test/myview469 00 00964419734:611402 1 - 00 88042ca This error...........470 00 00964419734:611405 1 - 00 88042ca That error...........471 00 00964419734:611408 1 - 00 88042ca Problem..............472 00 00964419734:611411 1 - 00 88042ca Something went wrong.473 00 00964419734:611414 1 - 00 88042ca Everything ok........474 00 00964419734:611417 1 - 00 88042ca data: 00000005475 00 00964419734:611419 1 - 00 88042ca data: 00000006476 00 00964419734:611422 1 - 00 88042ca data: 00000007477 00 00964419734:611425 1 - 00 88042ca data: 00000008478 00 00964419734:611428 1 - 00 88042ca data: 00000009479