1 menu "printk and dmesg options"
4 bool "Show timing information on printks"
7 Selecting this option causes time stamps of the printk()
8 messages to be added to the output of the syslog() system
9 call and at the console.
11 The timestamp is always recorded internally, and exported
12 to /dev/kmsg. This flag just specifies if the timestamp should
13 be included, not that the timestamp is recorded.
15 The behavior is also controlled by the kernel command line
16 parameter printk.time=1. See Documentation/admin-guide/kernel-parameters.rst
18 config CONSOLE_LOGLEVEL_DEFAULT
19 int "Default console loglevel (1-15)"
23 Default loglevel to determine what will be printed on the console.
25 Setting a default here is equivalent to passing in loglevel=<x> in
26 the kernel bootargs. loglevel=<x> continues to override whatever
27 value is specified here as well.
29 Note: This does not affect the log level of un-prefixed printk()
30 usage in the kernel. That is controlled by the MESSAGE_LOGLEVEL_DEFAULT
33 config MESSAGE_LOGLEVEL_DEFAULT
34 int "Default message log level (1-7)"
38 Default log level for printk statements with no specified priority.
40 This was hard-coded to KERN_WARNING since at least 2.6.10 but folks
41 that are auditing their logs closely may want to set it to a lower
44 Note: This does not affect what message level gets printed on the console
45 by default. To change that, use loglevel=<x> in the kernel bootargs,
46 or pick a different CONSOLE_LOGLEVEL_DEFAULT configuration value.
48 config BOOT_PRINTK_DELAY
49 bool "Delay each boot printk message by N milliseconds"
50 depends on DEBUG_KERNEL && PRINTK && GENERIC_CALIBRATE_DELAY
52 This build option allows you to read kernel boot messages
53 by inserting a short delay after each one. The delay is
54 specified in milliseconds on the kernel command line,
57 It is likely that you would also need to use "lpj=M" to preset
58 the "loops per jiffie" value.
59 See a previous boot log for the "lpj" value to use for your
60 system, and then set "lpj=M" before setting "boot_delay=N".
61 NOTE: Using this option may adversely affect SMP systems.
62 I.e., processors other than the first one may not boot up.
63 BOOT_PRINTK_DELAY also may cause LOCKUP_DETECTOR to detect
64 what it believes to be lockup conditions.
67 bool "Enable dynamic printk() support"
73 Compiles debug level messages into the kernel, which would not
74 otherwise be available at runtime. These messages can then be
75 enabled/disabled based on various levels of scope - per source file,
76 function, module, format string, and line number. This mechanism
77 implicitly compiles in all pr_debug() and dev_dbg() calls, which
78 enlarges the kernel text size by about 2%.
80 If a source file is compiled with DEBUG flag set, any
81 pr_debug() calls in it are enabled by default, but can be
82 disabled at runtime as below. Note that DEBUG flag is
83 turned on by many CONFIG_*DEBUG* options.
87 Dynamic debugging is controlled via the 'dynamic_debug/control' file,
88 which is contained in the 'debugfs' filesystem. Thus, the debugfs
89 filesystem must first be mounted before making use of this feature.
90 We refer the control file as: <debugfs>/dynamic_debug/control. This
91 file contains a list of the debug statements that can be enabled. The
92 format for each line of the file is:
94 filename:lineno [module]function flags format
96 filename : source file of the debug statement
97 lineno : line number of the debug statement
98 module : module that contains the debug statement
99 function : function that contains the debug statement
100 flags : '=p' means the line is turned 'on' for printing
101 format : the format used for the debug statement
105 nullarbor:~ # cat <debugfs>/dynamic_debug/control
106 # filename:lineno [module]function flags format
107 fs/aio.c:222 [aio]__put_ioctx =_ "__put_ioctx:\040freeing\040%p\012"
108 fs/aio.c:248 [aio]ioctx_alloc =_ "ENOMEM:\040nr_events\040too\040high\012"
109 fs/aio.c:1770 [aio]sys_io_cancel =_ "calling\040cancel\012"
113 // enable the message at line 1603 of file svcsock.c
114 nullarbor:~ # echo -n 'file svcsock.c line 1603 +p' >
115 <debugfs>/dynamic_debug/control
117 // enable all the messages in file svcsock.c
118 nullarbor:~ # echo -n 'file svcsock.c +p' >
119 <debugfs>/dynamic_debug/control
121 // enable all the messages in the NFS server module
122 nullarbor:~ # echo -n 'module nfsd +p' >
123 <debugfs>/dynamic_debug/control
125 // enable all 12 messages in the function svc_process()
126 nullarbor:~ # echo -n 'func svc_process +p' >
127 <debugfs>/dynamic_debug/control
129 // disable all 12 messages in the function svc_process()
130 nullarbor:~ # echo -n 'func svc_process -p' >
131 <debugfs>/dynamic_debug/control
133 See Documentation/dynamic-debug-howto.txt for additional information.
135 endmenu # "printk and dmesg options"
137 menu "Compile-time checks and compiler options"
140 bool "Compile the kernel with debug info"
141 depends on DEBUG_KERNEL && !COMPILE_TEST
143 If you say Y here the resulting kernel image will include
144 debugging info resulting in a larger kernel image.
145 This adds debug symbols to the kernel and modules (gcc -g), and
146 is needed if you intend to use kernel crashdump or binary object
147 tools like crash, kgdb, LKCD, gdb, etc on the kernel.
148 Say Y here only if you plan to debug the kernel.
152 config DEBUG_INFO_REDUCED
153 bool "Reduce debugging information"
154 depends on DEBUG_INFO
156 If you say Y here gcc is instructed to generate less debugging
157 information for structure types. This means that tools that
158 need full debugging information (like kgdb or systemtap) won't
159 be happy. But if you merely need debugging information to
160 resolve line numbers there is no loss. Advantage is that
161 build directory object sizes shrink dramatically over a full
162 DEBUG_INFO build and compile times are reduced too.
163 Only works with newer gcc versions.
165 config DEBUG_INFO_SPLIT
166 bool "Produce split debuginfo in .dwo files"
167 depends on DEBUG_INFO
169 Generate debug info into separate .dwo files. This significantly
170 reduces the build directory size for builds with DEBUG_INFO,
171 because it stores the information only once on disk in .dwo
172 files instead of multiple times in object files and executables.
173 In addition the debug information is also compressed.
175 Requires recent gcc (4.7+) and recent gdb/binutils.
176 Any tool that packages or reads debug information would need
177 to know about the .dwo files and include them.
178 Incompatible with older versions of ccache.
180 config DEBUG_INFO_DWARF4
181 bool "Generate dwarf4 debuginfo"
182 depends on DEBUG_INFO
184 Generate dwarf4 debug info. This requires recent versions
185 of gcc and gdb. It makes the debug information larger.
186 But it significantly improves the success of resolving
187 variables in gdb on optimized code.
190 bool "Provide GDB scripts for kernel debugging"
191 depends on DEBUG_INFO
193 This creates the required links to GDB helper scripts in the
194 build directory. If you load vmlinux into gdb, the helper
195 scripts will be automatically imported by gdb as well, and
196 additional functions are available to analyze a Linux kernel
197 instance. See Documentation/dev-tools/gdb-kernel-debugging.rst
200 config ENABLE_WARN_DEPRECATED
201 bool "Enable __deprecated logic"
204 Enable the __deprecated logic in the kernel build.
205 Disable this to suppress the "warning: 'foo' is deprecated
206 (declared at kernel/power/somefile.c:1234)" messages.
208 config ENABLE_MUST_CHECK
209 bool "Enable __must_check logic"
212 Enable the __must_check logic in the kernel build. Disable this to
213 suppress the "warning: ignoring return value of 'foo', declared with
214 attribute warn_unused_result" messages.
217 int "Warn for stack frames larger than (needs gcc 4.4)"
220 default 2048 if GCC_PLUGIN_LATENT_ENTROPY
221 default 1024 if !64BIT
222 default 2048 if 64BIT
224 Tell gcc to warn at build time for stack frames larger than this.
225 Setting this too low will cause a lot of warnings.
226 Setting it to 0 disables the warning.
229 config STRIP_ASM_SYMS
230 bool "Strip assembler-generated symbols during link"
233 Strip internal assembler-generated symbols during a link (symbols
234 that look like '.Lxxx') so they don't pollute the output of
235 get_wchan() and suchlike.
238 bool "Generate readable assembler code"
239 depends on DEBUG_KERNEL
241 Disable some compiler optimizations that tend to generate human unreadable
242 assembler output. This may make the kernel slightly slower, but it helps
243 to keep kernel developers who have to stare a lot at assembler listings
246 config UNUSED_SYMBOLS
247 bool "Enable unused/obsolete exported symbols"
250 Unused but exported symbols make the kernel needlessly bigger. For
251 that reason most of these unused exports will soon be removed. This
252 option is provided temporarily to provide a transition period in case
253 some external kernel module needs one of these symbols anyway. If you
254 encounter such a case in your module, consider if you are actually
255 using the right API. (rationale: since nobody in the kernel is using
256 this in a module, there is a pretty good chance it's actually the
257 wrong interface to use). If you really need the symbol, please send a
258 mail to the linux kernel mailing list mentioning the symbol and why
259 you really need it, and what the merge plan to the mainline kernel for
263 bool "Track page owner"
264 depends on DEBUG_KERNEL && STACKTRACE_SUPPORT
268 select PAGE_EXTENSION
270 This keeps track of what call chain is the owner of a page, may
271 help to find bare alloc_page(s) leaks. Even if you include this
272 feature on your build, it is disabled in default. You should pass
273 "page_owner=on" to boot parameter in order to enable it. Eats
274 a fair amount of memory if enabled. See tools/vm/page_owner_sort.c
275 for user-space helper.
280 bool "Debug Filesystem"
283 debugfs is a virtual file system that kernel developers use to put
284 debugging files into. Enable this option to be able to read and
285 write to these files.
287 For detailed documentation on the debugfs API, see
288 Documentation/DocBook/filesystems.
293 bool "Run 'make headers_check' when building vmlinux"
296 This option will extract the user-visible kernel headers whenever
297 building the kernel, and will run basic sanity checks on them to
298 ensure that exported files do not attempt to include files which
299 were not exported, etc.
301 If you're making modifications to header files which are
302 relevant for userspace, say 'Y', and check the headers
303 exported to $(INSTALL_HDR_PATH) (usually 'usr/include' in
304 your build tree), to make sure they're suitable.
306 config DEBUG_SECTION_MISMATCH
307 bool "Enable full Section mismatch analysis"
309 The section mismatch analysis checks if there are illegal
310 references from one section to another section.
311 During linktime or runtime, some sections are dropped;
312 any use of code/data previously in these sections would
313 most likely result in an oops.
314 In the code, functions and variables are annotated with
315 __init,, etc. (see the full list in include/linux/init.h),
316 which results in the code/data being placed in specific sections.
317 The section mismatch analysis is always performed after a full
318 kernel build, and enabling this option causes the following
319 additional steps to occur:
320 - Add the option -fno-inline-functions-called-once to gcc commands.
321 When inlining a function annotated with __init in a non-init
322 function, we would lose the section information and thus
323 the analysis would not catch the illegal reference.
324 This option tells gcc to inline less (but it does result in
326 - Run the section mismatch analysis for each module/built-in.o file.
327 When we run the section mismatch analysis on vmlinux.o, we
328 lose valuable information about where the mismatch was
330 Running the analysis for each module/built-in.o file
331 tells where the mismatch happens much closer to the
332 source. The drawback is that the same mismatch is
333 reported at least twice.
334 - Enable verbose reporting from modpost in order to help resolve
335 the section mismatches that are reported.
337 config SECTION_MISMATCH_WARN_ONLY
338 bool "Make section mismatch errors non-fatal"
341 If you say N here, the build process will fail if there are any
342 section mismatch, instead of just throwing warnings.
347 # Select this config option from the architecture Kconfig, if it
348 # is preferred to always offer frame pointers as a config
349 # option on the architecture (regardless of KERNEL_DEBUG):
351 config ARCH_WANT_FRAME_POINTERS
356 bool "Compile the kernel with frame pointers"
357 depends on DEBUG_KERNEL && \
358 (CRIS || M68K || FRV || UML || \
359 AVR32 || SUPERH || BLACKFIN || MN10300 || METAG) || \
360 ARCH_WANT_FRAME_POINTERS
361 default y if (DEBUG_INFO && UML) || ARCH_WANT_FRAME_POINTERS
363 If you say Y here the resulting kernel image will be slightly
364 larger and slower, but it gives very useful debugging information
365 in case of kernel bugs. (precise oopses/stacktraces/warnings)
367 config STACK_VALIDATION
368 bool "Compile-time stack metadata validation"
369 depends on HAVE_STACK_VALIDATION
372 Add compile-time checks to validate stack metadata, including frame
373 pointers (if CONFIG_FRAME_POINTER is enabled). This helps ensure
374 that runtime stack traces are more reliable.
376 For more information, see
377 tools/objtool/Documentation/stack-validation.txt.
379 config DEBUG_FORCE_WEAK_PER_CPU
380 bool "Force weak per-cpu definitions"
381 depends on DEBUG_KERNEL
383 s390 and alpha require percpu variables in modules to be
384 defined weak to work around addressing range issue which
385 puts the following two restrictions on percpu variable
388 1. percpu symbols must be unique whether static or not
389 2. percpu variables can't be defined inside a function
391 To ensure that generic code follows the above rules, this
392 option forces all percpu variables to be defined as weak.
394 endmenu # "Compiler options"
397 bool "Magic SysRq key"
400 If you say Y here, you will have some control over the system even
401 if the system crashes for example during kernel debugging (e.g., you
402 will be able to flush the buffer cache to disk, reboot the system
403 immediately or dump some status information). This is accomplished
404 by pressing various keys while holding SysRq (Alt+PrintScreen). It
405 also works on a serial console (on PC hardware at least), if you
406 send a BREAK and then within 5 seconds a command keypress. The
407 keys are documented in <file:Documentation/sysrq.txt>. Don't say Y
408 unless you really know what this hack does.
410 config MAGIC_SYSRQ_DEFAULT_ENABLE
411 hex "Enable magic SysRq key functions by default"
412 depends on MAGIC_SYSRQ
415 Specifies which SysRq key functions are enabled by default.
416 This may be set to 1 or 0 to enable or disable them all, or
417 to a bitmask as described in Documentation/sysrq.txt.
420 bool "Kernel debugging"
422 Say Y here if you are developing drivers or trying to debug and
423 identify kernel problems.
425 menu "Memory Debugging"
427 source mm/Kconfig.debug
430 bool "Debug object operations"
431 depends on DEBUG_KERNEL
433 If you say Y here, additional code will be inserted into the
434 kernel to track the life time of various objects and validate
435 the operations on those objects.
437 config DEBUG_OBJECTS_SELFTEST
438 bool "Debug objects selftest"
439 depends on DEBUG_OBJECTS
441 This enables the selftest of the object debug code.
443 config DEBUG_OBJECTS_FREE
444 bool "Debug objects in freed memory"
445 depends on DEBUG_OBJECTS
447 This enables checks whether a k/v free operation frees an area
448 which contains an object which has not been deactivated
449 properly. This can make kmalloc/kfree-intensive workloads
452 config DEBUG_OBJECTS_TIMERS
453 bool "Debug timer objects"
454 depends on DEBUG_OBJECTS
456 If you say Y here, additional code will be inserted into the
457 timer routines to track the life time of timer objects and
458 validate the timer operations.
460 config DEBUG_OBJECTS_WORK
461 bool "Debug work objects"
462 depends on DEBUG_OBJECTS
464 If you say Y here, additional code will be inserted into the
465 work queue routines to track the life time of work objects and
466 validate the work operations.
468 config DEBUG_OBJECTS_RCU_HEAD
469 bool "Debug RCU callbacks objects"
470 depends on DEBUG_OBJECTS
472 Enable this to turn on debugging of RCU list heads (call_rcu() usage).
474 config DEBUG_OBJECTS_PERCPU_COUNTER
475 bool "Debug percpu counter objects"
476 depends on DEBUG_OBJECTS
478 If you say Y here, additional code will be inserted into the
479 percpu counter routines to track the life time of percpu counter
480 objects and validate the percpu counter operations.
482 config DEBUG_OBJECTS_ENABLE_DEFAULT
483 int "debug_objects bootup default value (0-1)"
486 depends on DEBUG_OBJECTS
488 Debug objects boot parameter default value
491 bool "Debug slab memory allocations"
492 depends on DEBUG_KERNEL && SLAB && !KMEMCHECK
494 Say Y here to have the kernel do limited verification on memory
495 allocation as well as poisoning memory on free to catch use of freed
496 memory. This can make kmalloc/kfree-intensive workloads much slower.
498 config DEBUG_SLAB_LEAK
499 bool "Memory leak debugging"
500 depends on DEBUG_SLAB
503 bool "SLUB debugging on by default"
504 depends on SLUB && SLUB_DEBUG && !KMEMCHECK
507 Boot with debugging on by default. SLUB boots by default with
508 the runtime debug capabilities switched off. Enabling this is
509 equivalent to specifying the "slub_debug" parameter on boot.
510 There is no support for more fine grained debug control like
511 possible with slub_debug=xxx. SLUB debugging may be switched
512 off in a kernel built with CONFIG_SLUB_DEBUG_ON by specifying
517 bool "Enable SLUB performance statistics"
518 depends on SLUB && SYSFS
520 SLUB statistics are useful to debug SLUBs allocation behavior in
521 order find ways to optimize the allocator. This should never be
522 enabled for production use since keeping statistics slows down
523 the allocator by a few percentage points. The slabinfo command
524 supports the determination of the most active slabs to figure
525 out which slabs are relevant to a particular load.
526 Try running: slabinfo -DA
528 config HAVE_DEBUG_KMEMLEAK
531 config DEBUG_KMEMLEAK
532 bool "Kernel memory leak detector"
533 depends on DEBUG_KERNEL && HAVE_DEBUG_KMEMLEAK
535 select STACKTRACE if STACKTRACE_SUPPORT
539 Say Y here if you want to enable the memory leak
540 detector. The memory allocation/freeing is traced in a way
541 similar to the Boehm's conservative garbage collector, the
542 difference being that the orphan objects are not freed but
543 only shown in /sys/kernel/debug/kmemleak. Enabling this
544 feature will introduce an overhead to memory
545 allocations. See Documentation/dev-tools/kmemleak.rst for more
548 Enabling DEBUG_SLAB or SLUB_DEBUG may increase the chances
549 of finding leaks due to the slab objects poisoning.
551 In order to access the kmemleak file, debugfs needs to be
552 mounted (usually at /sys/kernel/debug).
554 config DEBUG_KMEMLEAK_EARLY_LOG_SIZE
555 int "Maximum kmemleak early log entries"
556 depends on DEBUG_KMEMLEAK
560 Kmemleak must track all the memory allocations to avoid
561 reporting false positives. Since memory may be allocated or
562 freed before kmemleak is initialised, an early log buffer is
563 used to store these actions. If kmemleak reports "early log
564 buffer exceeded", please increase this value.
566 config DEBUG_KMEMLEAK_TEST
567 tristate "Simple test for the kernel memory leak detector"
568 depends on DEBUG_KMEMLEAK && m
570 This option enables a module that explicitly leaks memory.
574 config DEBUG_KMEMLEAK_DEFAULT_OFF
575 bool "Default kmemleak to off"
576 depends on DEBUG_KMEMLEAK
578 Say Y here to disable kmemleak by default. It can then be enabled
579 on the command line via kmemleak=on.
581 config DEBUG_STACK_USAGE
582 bool "Stack utilization instrumentation"
583 depends on DEBUG_KERNEL && !IA64
585 Enables the display of the minimum amount of free stack which each
586 task has ever had available in the sysrq-T and sysrq-P debug output.
588 This option will slow down process creation somewhat.
592 depends on DEBUG_KERNEL
594 Enable this to turn on extended checks in the virtual-memory system
595 that may impact performance.
599 config DEBUG_VM_VMACACHE
600 bool "Debug VMA caching"
603 Enable this to turn on VMA caching debug information. Doing so
604 can cause significant overhead, so only enable it in non-production
610 bool "Debug VM red-black trees"
613 Enable VM red-black tree debugging information and extra validations.
617 config DEBUG_VM_PGFLAGS
618 bool "Debug page-flags operations"
621 Enables extra validation on page flags operations.
626 bool "Debug VM translations"
627 depends on DEBUG_KERNEL && X86
629 Enable some costly sanity checks in virtual to page code. This can
630 catch mistakes with virt_to_page() and friends.
634 config DEBUG_NOMMU_REGIONS
635 bool "Debug the global anon/private NOMMU mapping region tree"
636 depends on DEBUG_KERNEL && !MMU
638 This option causes the global tree of anonymous and private mapping
639 regions to be regularly checked for invalid topology.
641 config DEBUG_MEMORY_INIT
642 bool "Debug memory initialisation" if EXPERT
645 Enable this for additional checks during memory initialisation.
646 The sanity checks verify aspects of the VM such as the memory model
647 and other information provided by the architecture. Verbose
648 information will be printed at KERN_DEBUG loglevel depending
649 on the mminit_loglevel= command-line option.
653 config MEMORY_NOTIFIER_ERROR_INJECT
654 tristate "Memory hotplug notifier error injection module"
655 depends on MEMORY_HOTPLUG_SPARSE && NOTIFIER_ERROR_INJECTION
657 This option provides the ability to inject artificial errors to
658 memory hotplug notifier chain callbacks. It is controlled through
659 debugfs interface under /sys/kernel/debug/notifier-error-inject/memory
661 If the notifier call chain should be failed with some events
662 notified, write the error code to "actions/<notifier event>/error".
664 Example: Inject memory hotplug offline error (-12 == -ENOMEM)
666 # cd /sys/kernel/debug/notifier-error-inject/memory
667 # echo -12 > actions/MEM_GOING_OFFLINE/error
668 # echo offline > /sys/devices/system/memory/memoryXXX/state
669 bash: echo: write error: Cannot allocate memory
671 To compile this code as a module, choose M here: the module will
672 be called memory-notifier-error-inject.
676 config DEBUG_PER_CPU_MAPS
677 bool "Debug access to per_cpu maps"
678 depends on DEBUG_KERNEL
681 Say Y to verify that the per_cpu map being accessed has
682 been set up. This adds a fair amount of code to kernel memory
683 and decreases performance.
688 bool "Highmem debugging"
689 depends on DEBUG_KERNEL && HIGHMEM
691 This option enables additional error checking for high memory
692 systems. Disable for production systems.
694 config HAVE_DEBUG_STACKOVERFLOW
697 config DEBUG_STACKOVERFLOW
698 bool "Check for stack overflows"
699 depends on DEBUG_KERNEL && HAVE_DEBUG_STACKOVERFLOW
701 Say Y here if you want to check for overflows of kernel, IRQ
702 and exception stacks (if your architecture uses them). This
703 option will show detailed messages if free stack space drops
704 below a certain limit.
706 These kinds of bugs usually occur when call-chains in the
707 kernel get too deep, especially when interrupts are
710 Use this in cases where you see apparently random memory
711 corruption, especially if it appears in 'struct thread_info'
713 If in doubt, say "N".
715 source "lib/Kconfig.kmemcheck"
717 source "lib/Kconfig.kasan"
719 endmenu # "Memory Debugging"
724 KCOV does not have any arch-specific code, but currently it is enabled
725 only for x86_64. KCOV requires testing on other archs, and most likely
726 disabling of instrumentation for some early boot code.
729 bool "Code coverage for fuzzing"
730 depends on ARCH_HAS_KCOV
732 select GCC_PLUGINS if !COMPILE_TEST
733 select GCC_PLUGIN_SANCOV if !COMPILE_TEST
735 KCOV exposes kernel code coverage information in a form suitable
736 for coverage-guided fuzzing (randomized testing).
738 If RANDOMIZE_BASE is enabled, PC values will not be stable across
739 different machines and across reboots. If you need stable PC values,
740 disable RANDOMIZE_BASE.
742 For more details, see Documentation/dev-tools/kcov.rst.
744 config KCOV_INSTRUMENT_ALL
745 bool "Instrument all code by default"
749 If you are doing generic system call fuzzing (like e.g. syzkaller),
750 then you will want to instrument the whole kernel and you should
751 say y here. If you are doing more targeted fuzzing (like e.g.
752 filesystem fuzzing with AFL) then you will want to enable coverage
753 for more specific subsets of files, and should say n here.
756 bool "Debug shared IRQ handlers"
757 depends on DEBUG_KERNEL
759 Enable this to generate a spurious interrupt as soon as a shared
760 interrupt handler is registered, and just before one is deregistered.
761 Drivers ought to be able to handle interrupts coming in at those
762 points; some don't and need to be caught.
764 menu "Debug Lockups and Hangs"
766 config LOCKUP_DETECTOR
767 bool "Detect Hard and Soft Lockups"
768 depends on DEBUG_KERNEL && !S390
770 Say Y here to enable the kernel to act as a watchdog to detect
771 hard and soft lockups.
773 Softlockups are bugs that cause the kernel to loop in kernel
774 mode for more than 20 seconds, without giving other tasks a
775 chance to run. The current stack trace is displayed upon
776 detection and the system will stay locked up.
778 Hardlockups are bugs that cause the CPU to loop in kernel mode
779 for more than 10 seconds, without letting other interrupts have a
780 chance to run. The current stack trace is displayed upon detection
781 and the system will stay locked up.
783 The overhead should be minimal. A periodic hrtimer runs to
784 generate interrupts and kick the watchdog task every 4 seconds.
785 An NMI is generated every 10 seconds or so to check for hardlockups.
787 The frequency of hrtimer and NMI events and the soft and hard lockup
788 thresholds can be controlled through the sysctl watchdog_thresh.
790 config HARDLOCKUP_DETECTOR
792 depends on LOCKUP_DETECTOR && !HAVE_NMI_WATCHDOG
793 depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI
795 config BOOTPARAM_HARDLOCKUP_PANIC
796 bool "Panic (Reboot) On Hard Lockups"
797 depends on HARDLOCKUP_DETECTOR
799 Say Y here to enable the kernel to panic on "hard lockups",
800 which are bugs that cause the kernel to loop in kernel
801 mode with interrupts disabled for more than 10 seconds (configurable
802 using the watchdog_thresh sysctl).
806 config BOOTPARAM_HARDLOCKUP_PANIC_VALUE
808 depends on HARDLOCKUP_DETECTOR
810 default 0 if !BOOTPARAM_HARDLOCKUP_PANIC
811 default 1 if BOOTPARAM_HARDLOCKUP_PANIC
813 config BOOTPARAM_SOFTLOCKUP_PANIC
814 bool "Panic (Reboot) On Soft Lockups"
815 depends on LOCKUP_DETECTOR
817 Say Y here to enable the kernel to panic on "soft lockups",
818 which are bugs that cause the kernel to loop in kernel
819 mode for more than 20 seconds (configurable using the watchdog_thresh
820 sysctl), without giving other tasks a chance to run.
822 The panic can be used in combination with panic_timeout,
823 to cause the system to reboot automatically after a
824 lockup has been detected. This feature is useful for
825 high-availability systems that have uptime guarantees and
826 where a lockup must be resolved ASAP.
830 config BOOTPARAM_SOFTLOCKUP_PANIC_VALUE
832 depends on LOCKUP_DETECTOR
834 default 0 if !BOOTPARAM_SOFTLOCKUP_PANIC
835 default 1 if BOOTPARAM_SOFTLOCKUP_PANIC
837 config DETECT_HUNG_TASK
838 bool "Detect Hung Tasks"
839 depends on DEBUG_KERNEL
840 default LOCKUP_DETECTOR
842 Say Y here to enable the kernel to detect "hung tasks",
843 which are bugs that cause the task to be stuck in
844 uninterruptible "D" state indefinitely.
846 When a hung task is detected, the kernel will print the
847 current stack trace (which you should report), but the
848 task will stay in uninterruptible state. If lockdep is
849 enabled then all held locks will also be reported. This
850 feature has negligible overhead.
852 config DEFAULT_HUNG_TASK_TIMEOUT
853 int "Default timeout for hung task detection (in seconds)"
854 depends on DETECT_HUNG_TASK
857 This option controls the default timeout (in seconds) used
858 to determine when a task has become non-responsive and should
861 It can be adjusted at runtime via the kernel.hung_task_timeout_secs
862 sysctl or by writing a value to
863 /proc/sys/kernel/hung_task_timeout_secs.
865 A timeout of 0 disables the check. The default is two minutes.
866 Keeping the default should be fine in most cases.
868 config BOOTPARAM_HUNG_TASK_PANIC
869 bool "Panic (Reboot) On Hung Tasks"
870 depends on DETECT_HUNG_TASK
872 Say Y here to enable the kernel to panic on "hung tasks",
873 which are bugs that cause the kernel to leave a task stuck
874 in uninterruptible "D" state.
876 The panic can be used in combination with panic_timeout,
877 to cause the system to reboot automatically after a
878 hung task has been detected. This feature is useful for
879 high-availability systems that have uptime guarantees and
880 where a hung tasks must be resolved ASAP.
884 config BOOTPARAM_HUNG_TASK_PANIC_VALUE
886 depends on DETECT_HUNG_TASK
888 default 0 if !BOOTPARAM_HUNG_TASK_PANIC
889 default 1 if BOOTPARAM_HUNG_TASK_PANIC
892 bool "Detect Workqueue Stalls"
893 depends on DEBUG_KERNEL
895 Say Y here to enable stall detection on workqueues. If a
896 worker pool doesn't make forward progress on a pending work
897 item for over a given amount of time, 30s by default, a
898 warning message is printed along with dump of workqueue
899 state. This can be configured through kernel parameter
900 "workqueue.watchdog_thresh" and its sysfs counterpart.
902 endmenu # "Debug lockups and hangs"
907 Say Y here to enable the kernel to panic when it oopses. This
908 has the same effect as setting oops=panic on the kernel command
911 This feature is useful to ensure that the kernel does not do
912 anything erroneous after an oops which could result in data
913 corruption or other issues.
917 config PANIC_ON_OOPS_VALUE
920 default 0 if !PANIC_ON_OOPS
921 default 1 if PANIC_ON_OOPS
927 Set the timeout value (in seconds) until a reboot occurs when the
928 the kernel panics. If n = 0, then we wait forever. A timeout
929 value n > 0 will wait n seconds before rebooting, while a timeout
930 value n < 0 will reboot immediately.
933 bool "Collect scheduler debugging info"
934 depends on DEBUG_KERNEL && PROC_FS
937 If you say Y here, the /proc/sched_debug file will be provided
938 that can help debug the scheduler. The runtime overhead of this
946 bool "Collect scheduler statistics"
947 depends on DEBUG_KERNEL && PROC_FS
950 If you say Y here, additional code will be inserted into the
951 scheduler and related routines to collect statistics about
952 scheduler behavior and provide them in /proc/schedstat. These
953 stats may be useful for both tuning and debugging the scheduler
954 If you aren't debugging the scheduler or trying to tune a specific
955 application, you can say N to avoid the very slight overhead
958 config SCHED_STACK_END_CHECK
959 bool "Detect stack corruption on calls to schedule()"
960 depends on DEBUG_KERNEL
963 This option checks for a stack overrun on calls to schedule().
964 If the stack end location is found to be over written always panic as
965 the content of the corrupted region can no longer be trusted.
966 This is to ensure no erroneous behaviour occurs which could result in
967 data corruption or a sporadic crash at a later stage once the region
968 is examined. The runtime overhead introduced is minimal.
970 config DEBUG_TIMEKEEPING
971 bool "Enable extra timekeeping sanity checking"
973 This option will enable additional timekeeping sanity checks
974 which may be helpful when diagnosing issues where timekeeping
975 problems are suspected.
977 This may include checks in the timekeeping hotpaths, so this
978 option may have a (very small) performance impact to some
984 bool "Collect kernel timers statistics"
985 depends on DEBUG_KERNEL && PROC_FS
987 If you say Y here, additional code will be inserted into the
988 timer routines to collect statistics about kernel timers being
989 reprogrammed. The statistics can be read from /proc/timer_stats.
990 The statistics collection is started by writing 1 to /proc/timer_stats,
991 writing 0 stops it. This feature is useful to collect information
992 about timer usage patterns in kernel and userspace. This feature
993 is lightweight if enabled in the kernel config but not activated
994 (it defaults to deactivated on bootup and will only be activated
995 if some application like powertop activates it explicitly).
998 bool "Debug preemptible kernel"
999 depends on DEBUG_KERNEL && PREEMPT && TRACE_IRQFLAGS_SUPPORT
1002 If you say Y here then the kernel will use a debug variant of the
1003 commonly used smp_processor_id() function and will print warnings
1004 if kernel code uses it in a preemption-unsafe way. Also, the kernel
1005 will detect preemption count underflows.
1007 menu "Lock Debugging (spinlocks, mutexes, etc...)"
1009 config DEBUG_RT_MUTEXES
1010 bool "RT Mutex debugging, deadlock detection"
1011 depends on DEBUG_KERNEL && RT_MUTEXES
1013 This allows rt mutex semantics violations and rt mutex related
1014 deadlocks (lockups) to be detected and reported automatically.
1016 config DEBUG_SPINLOCK
1017 bool "Spinlock and rw-lock debugging: basic checks"
1018 depends on DEBUG_KERNEL
1019 select UNINLINE_SPIN_UNLOCK
1021 Say Y here and build SMP to catch missing spinlock initialization
1022 and certain other kinds of spinlock errors commonly made. This is
1023 best used in conjunction with the NMI watchdog so that spinlock
1024 deadlocks are also debuggable.
1026 config DEBUG_MUTEXES
1027 bool "Mutex debugging: basic checks"
1028 depends on DEBUG_KERNEL
1030 This feature allows mutex semantics violations to be detected and
1033 config DEBUG_WW_MUTEX_SLOWPATH
1034 bool "Wait/wound mutex debugging: Slowpath testing"
1035 depends on DEBUG_KERNEL && TRACE_IRQFLAGS_SUPPORT && STACKTRACE_SUPPORT && LOCKDEP_SUPPORT
1036 select DEBUG_LOCK_ALLOC
1037 select DEBUG_SPINLOCK
1038 select DEBUG_MUTEXES
1040 This feature enables slowpath testing for w/w mutex users by
1041 injecting additional -EDEADLK wound/backoff cases. Together with
1042 the full mutex checks enabled with (CONFIG_PROVE_LOCKING) this
1043 will test all possible w/w mutex interface abuse with the
1044 exception of simply not acquiring all the required locks.
1045 Note that this feature can introduce significant overhead, so
1046 it really should not be enabled in a production or distro kernel,
1047 even a debug kernel. If you are a driver writer, enable it. If
1048 you are a distro, do not.
1050 config DEBUG_LOCK_ALLOC
1051 bool "Lock debugging: detect incorrect freeing of live locks"
1052 depends on DEBUG_KERNEL && TRACE_IRQFLAGS_SUPPORT && STACKTRACE_SUPPORT && LOCKDEP_SUPPORT
1053 select DEBUG_SPINLOCK
1054 select DEBUG_MUTEXES
1057 This feature will check whether any held lock (spinlock, rwlock,
1058 mutex or rwsem) is incorrectly freed by the kernel, via any of the
1059 memory-freeing routines (kfree(), kmem_cache_free(), free_pages(),
1060 vfree(), etc.), whether a live lock is incorrectly reinitialized via
1061 spin_lock_init()/mutex_init()/etc., or whether there is any lock
1062 held during task exit.
1064 config PROVE_LOCKING
1065 bool "Lock debugging: prove locking correctness"
1066 depends on DEBUG_KERNEL && TRACE_IRQFLAGS_SUPPORT && STACKTRACE_SUPPORT && LOCKDEP_SUPPORT
1068 select DEBUG_SPINLOCK
1069 select DEBUG_MUTEXES
1070 select DEBUG_LOCK_ALLOC
1071 select TRACE_IRQFLAGS
1074 This feature enables the kernel to prove that all locking
1075 that occurs in the kernel runtime is mathematically
1076 correct: that under no circumstance could an arbitrary (and
1077 not yet triggered) combination of observed locking
1078 sequences (on an arbitrary number of CPUs, running an
1079 arbitrary number of tasks and interrupt contexts) cause a
1082 In short, this feature enables the kernel to report locking
1083 related deadlocks before they actually occur.
1085 The proof does not depend on how hard and complex a
1086 deadlock scenario would be to trigger: how many
1087 participant CPUs, tasks and irq-contexts would be needed
1088 for it to trigger. The proof also does not depend on
1089 timing: if a race and a resulting deadlock is possible
1090 theoretically (no matter how unlikely the race scenario
1091 is), it will be proven so and will immediately be
1092 reported by the kernel (once the event is observed that
1093 makes the deadlock theoretically possible).
1095 If a deadlock is impossible (i.e. the locking rules, as
1096 observed by the kernel, are mathematically correct), the
1097 kernel reports nothing.
1099 NOTE: this feature can also be enabled for rwlocks, mutexes
1100 and rwsems - in which case all dependencies between these
1101 different locking variants are observed and mapped too, and
1102 the proof of observed correctness is also maintained for an
1103 arbitrary combination of these separate locking variants.
1105 For more details, see Documentation/locking/lockdep-design.txt.
1107 config PROVE_LOCKING_SMALL
1112 depends on DEBUG_KERNEL && TRACE_IRQFLAGS_SUPPORT && STACKTRACE_SUPPORT && LOCKDEP_SUPPORT
1114 select FRAME_POINTER if !MIPS && !PPC && !ARM_UNWIND && !S390 && !MICROBLAZE && !ARC && !SCORE
1119 bool "Lock usage statistics"
1120 depends on DEBUG_KERNEL && TRACE_IRQFLAGS_SUPPORT && STACKTRACE_SUPPORT && LOCKDEP_SUPPORT
1122 select DEBUG_SPINLOCK
1123 select DEBUG_MUTEXES
1124 select DEBUG_LOCK_ALLOC
1127 This feature enables tracking lock contention points
1129 For more details, see Documentation/locking/lockstat.txt
1131 This also enables lock events required by "perf lock",
1133 If you want to use "perf lock", you also need to turn on
1134 CONFIG_EVENT_TRACING.
1136 CONFIG_LOCK_STAT defines "contended" and "acquired" lock events.
1137 (CONFIG_LOCKDEP defines "acquire" and "release" events.)
1139 config DEBUG_LOCKDEP
1140 bool "Lock dependency engine debugging"
1141 depends on DEBUG_KERNEL && LOCKDEP
1143 If you say Y here, the lock dependency engine will do
1144 additional runtime checks to debug itself, at the price
1145 of more runtime overhead.
1147 config DEBUG_ATOMIC_SLEEP
1148 bool "Sleep inside atomic section checking"
1149 select PREEMPT_COUNT
1150 depends on DEBUG_KERNEL
1152 If you say Y here, various routines which may sleep will become very
1153 noisy if they are called inside atomic sections: when a spinlock is
1154 held, inside an rcu read side critical section, inside preempt disabled
1155 sections, inside an interrupt, etc...
1157 config DEBUG_LOCKING_API_SELFTESTS
1158 bool "Locking API boot-time self-tests"
1159 depends on DEBUG_KERNEL
1161 Say Y here if you want the kernel to run a short self-test during
1162 bootup. The self-test checks whether common types of locking bugs
1163 are detected by debugging mechanisms or not. (if you disable
1164 lock debugging then those bugs wont be detected of course.)
1165 The following locking APIs are covered: spinlocks, rwlocks,
1168 config LOCK_TORTURE_TEST
1169 tristate "torture tests for locking"
1170 depends on DEBUG_KERNEL
1174 This option provides a kernel module that runs torture tests
1175 on kernel locking primitives. The kernel module may be built
1176 after the fact on the running kernel to be tested, if desired.
1178 Say Y here if you want kernel locking-primitive torture tests
1179 to be built into the kernel.
1180 Say M if you want these torture tests to build as a module.
1181 Say N if you are unsure.
1183 endmenu # lock debugging
1185 config TRACE_IRQFLAGS
1188 Enables hooks to interrupt enabling and disabling for
1189 either tracing or lock debugging.
1192 bool "Stack backtrace support"
1193 depends on STACKTRACE_SUPPORT
1195 This option causes the kernel to create a /proc/pid/stack for
1196 every process, showing its current stack trace.
1197 It is also used by various kernel debugging features that require
1198 stack trace generation.
1200 config DEBUG_KOBJECT
1201 bool "kobject debugging"
1202 depends on DEBUG_KERNEL
1204 If you say Y here, some extra kobject debugging messages will be sent
1207 config DEBUG_KOBJECT_RELEASE
1208 bool "kobject release debugging"
1209 depends on DEBUG_OBJECTS_TIMERS
1211 kobjects are reference counted objects. This means that their
1212 last reference count put is not predictable, and the kobject can
1213 live on past the point at which a driver decides to drop it's
1214 initial reference to the kobject gained on allocation. An
1215 example of this would be a struct device which has just been
1218 However, some buggy drivers assume that after such an operation,
1219 the memory backing the kobject can be immediately freed. This
1220 goes completely against the principles of a refcounted object.
1222 If you say Y here, the kernel will delay the release of kobjects
1223 on the last reference count to improve the visibility of this
1224 kind of kobject release bug.
1226 config HAVE_DEBUG_BUGVERBOSE
1229 config DEBUG_BUGVERBOSE
1230 bool "Verbose BUG() reporting (adds 70K)" if DEBUG_KERNEL && EXPERT
1231 depends on BUG && (GENERIC_BUG || HAVE_DEBUG_BUGVERBOSE)
1234 Say Y here to make BUG() panics output the file name and line number
1235 of the BUG call as well as the EIP and oops trace. This aids
1236 debugging but costs about 70-100K of memory.
1239 bool "Debug linked list manipulation"
1240 depends on DEBUG_KERNEL || BUG_ON_DATA_CORRUPTION
1242 Enable this to turn on extended checks in the linked-list
1247 config DEBUG_PI_LIST
1248 bool "Debug priority linked list manipulation"
1249 depends on DEBUG_KERNEL
1251 Enable this to turn on extended checks in the priority-ordered
1252 linked-list (plist) walking routines. This checks the entire
1253 list multiple times during each manipulation.
1258 bool "Debug SG table operations"
1259 depends on DEBUG_KERNEL
1261 Enable this to turn on checks on scatter-gather tables. This can
1262 help find problems with drivers that do not properly initialize
1267 config DEBUG_NOTIFIERS
1268 bool "Debug notifier call chains"
1269 depends on DEBUG_KERNEL
1271 Enable this to turn on sanity checking for notifier call chains.
1272 This is most useful for kernel developers to make sure that
1273 modules properly unregister themselves from notifier chains.
1274 This is a relatively cheap check but if you care about maximum
1277 config DEBUG_CREDENTIALS
1278 bool "Debug credential management"
1279 depends on DEBUG_KERNEL
1281 Enable this to turn on some debug checking for credential
1282 management. The additional code keeps track of the number of
1283 pointers from task_structs to any given cred struct, and checks to
1284 see that this number never exceeds the usage count of the cred
1287 Furthermore, if SELinux is enabled, this also checks that the
1288 security pointer in the cred struct is never seen to be invalid.
1292 menu "RCU Debugging"
1295 def_bool PROVE_LOCKING
1297 config PROVE_RCU_REPEATEDLY
1298 bool "RCU debugging: don't disable PROVE_RCU on first splat"
1299 depends on PROVE_RCU
1302 By itself, PROVE_RCU will disable checking upon issuing the
1303 first warning (or "splat"). This feature prevents such
1304 disabling, allowing multiple RCU-lockdep warnings to be printed
1307 Say Y to allow multiple RCU-lockdep warnings per boot.
1309 Say N if you are unsure.
1311 config SPARSE_RCU_POINTER
1312 bool "RCU debugging: sparse-based checks for pointer usage"
1315 This feature enables the __rcu sparse annotation for
1316 RCU-protected pointers. This annotation will cause sparse
1317 to flag any non-RCU used of annotated pointers. This can be
1318 helpful when debugging RCU usage. Please note that this feature
1319 is not intended to enforce code cleanliness; it is instead merely
1322 Say Y to make sparse flag questionable use of RCU-protected pointers
1324 Say N if you are unsure.
1330 config RCU_PERF_TEST
1331 tristate "performance tests for RCU"
1332 depends on DEBUG_KERNEL
1338 This option provides a kernel module that runs performance
1339 tests on the RCU infrastructure. The kernel module may be built
1340 after the fact on the running kernel to be tested, if desired.
1342 Say Y here if you want RCU performance tests to be built into
1344 Say M if you want the RCU performance tests to build as a module.
1345 Say N if you are unsure.
1347 config RCU_TORTURE_TEST
1348 tristate "torture tests for RCU"
1349 depends on DEBUG_KERNEL
1355 This option provides a kernel module that runs torture tests
1356 on the RCU infrastructure. The kernel module may be built
1357 after the fact on the running kernel to be tested, if desired.
1359 Say Y here if you want RCU torture tests to be built into
1361 Say M if you want the RCU torture tests to build as a module.
1362 Say N if you are unsure.
1364 config RCU_TORTURE_TEST_SLOW_PREINIT
1365 bool "Slow down RCU grace-period pre-initialization to expose races"
1366 depends on RCU_TORTURE_TEST
1368 This option delays grace-period pre-initialization (the
1369 propagation of CPU-hotplug changes up the rcu_node combining
1370 tree) for a few jiffies between initializing each pair of
1371 consecutive rcu_node structures. This helps to expose races
1372 involving grace-period pre-initialization, in other words, it
1373 makes your kernel less stable. It can also greatly increase
1374 grace-period latency, especially on systems with large numbers
1375 of CPUs. This is useful when torture-testing RCU, but in
1376 almost no other circumstance.
1378 Say Y here if you want your system to crash and hang more often.
1379 Say N if you want a sane system.
1381 config RCU_TORTURE_TEST_SLOW_PREINIT_DELAY
1382 int "How much to slow down RCU grace-period pre-initialization"
1385 depends on RCU_TORTURE_TEST_SLOW_PREINIT
1387 This option specifies the number of jiffies to wait between
1388 each rcu_node structure pre-initialization step.
1390 config RCU_TORTURE_TEST_SLOW_INIT
1391 bool "Slow down RCU grace-period initialization to expose races"
1392 depends on RCU_TORTURE_TEST
1394 This option delays grace-period initialization for a few
1395 jiffies between initializing each pair of consecutive
1396 rcu_node structures. This helps to expose races involving
1397 grace-period initialization, in other words, it makes your
1398 kernel less stable. It can also greatly increase grace-period
1399 latency, especially on systems with large numbers of CPUs.
1400 This is useful when torture-testing RCU, but in almost no
1403 Say Y here if you want your system to crash and hang more often.
1404 Say N if you want a sane system.
1406 config RCU_TORTURE_TEST_SLOW_INIT_DELAY
1407 int "How much to slow down RCU grace-period initialization"
1410 depends on RCU_TORTURE_TEST_SLOW_INIT
1412 This option specifies the number of jiffies to wait between
1413 each rcu_node structure initialization.
1415 config RCU_TORTURE_TEST_SLOW_CLEANUP
1416 bool "Slow down RCU grace-period cleanup to expose races"
1417 depends on RCU_TORTURE_TEST
1419 This option delays grace-period cleanup for a few jiffies
1420 between cleaning up each pair of consecutive rcu_node
1421 structures. This helps to expose races involving grace-period
1422 cleanup, in other words, it makes your kernel less stable.
1423 It can also greatly increase grace-period latency, especially
1424 on systems with large numbers of CPUs. This is useful when
1425 torture-testing RCU, but in almost no other circumstance.
1427 Say Y here if you want your system to crash and hang more often.
1428 Say N if you want a sane system.
1430 config RCU_TORTURE_TEST_SLOW_CLEANUP_DELAY
1431 int "How much to slow down RCU grace-period cleanup"
1434 depends on RCU_TORTURE_TEST_SLOW_CLEANUP
1436 This option specifies the number of jiffies to wait between
1437 each rcu_node structure cleanup operation.
1439 config RCU_CPU_STALL_TIMEOUT
1440 int "RCU CPU stall timeout in seconds"
1441 depends on RCU_STALL_COMMON
1445 If a given RCU grace period extends more than the specified
1446 number of seconds, a CPU stall warning is printed. If the
1447 RCU grace period persists, additional CPU stall warnings are
1448 printed at more widely spaced intervals.
1451 bool "Enable tracing for RCU"
1452 depends on DEBUG_KERNEL
1455 This option provides tracing in RCU which presents stats
1456 in debugfs for debugging RCU implementation. It also enables
1457 additional tracepoints for ftrace-style event tracing.
1459 Say Y here if you want to enable RCU tracing
1460 Say N if you are unsure.
1462 config RCU_EQS_DEBUG
1463 bool "Provide debugging asserts for adding NO_HZ support to an arch"
1464 depends on DEBUG_KERNEL
1466 This option provides consistency checks in RCU's handling of
1467 NO_HZ. These checks have proven quite helpful in detecting
1468 bugs in arch-specific NO_HZ code.
1470 Say N here if you need ultimate kernel/user switch latencies
1471 Say Y if you are unsure
1473 endmenu # "RCU Debugging"
1475 config DEBUG_WQ_FORCE_RR_CPU
1476 bool "Force round-robin CPU selection for unbound work items"
1477 depends on DEBUG_KERNEL
1480 Workqueue used to implicitly guarantee that work items queued
1481 without explicit CPU specified are put on the local CPU. This
1482 guarantee is no longer true and while local CPU is still
1483 preferred work items may be put on foreign CPUs. Kernel
1484 parameter "workqueue.debug_force_rr_cpu" is added to force
1485 round-robin CPU selection to flush out usages which depend on the
1486 now broken guarantee. This config option enables the debug
1487 feature by default. When enabled, memory and cache locality will
1490 config DEBUG_BLOCK_EXT_DEVT
1491 bool "Force extended block device numbers and spread them"
1492 depends on DEBUG_KERNEL
1496 BIG FAT WARNING: ENABLING THIS OPTION MIGHT BREAK BOOTING ON
1497 SOME DISTRIBUTIONS. DO NOT ENABLE THIS UNLESS YOU KNOW WHAT
1498 YOU ARE DOING. Distros, please enable this and fix whatever
1501 Conventionally, block device numbers are allocated from
1502 predetermined contiguous area. However, extended block area
1503 may introduce non-contiguous block device numbers. This
1504 option forces most block device numbers to be allocated from
1505 the extended space and spreads them to discover kernel or
1506 userland code paths which assume predetermined contiguous
1507 device number allocation.
1509 Note that turning on this debug option shuffles all the
1510 device numbers for all IDE and SCSI devices including libata
1511 ones, so root partition specified using device number
1512 directly (via rdev or root=MAJ:MIN) won't work anymore.
1513 Textual device names (root=/dev/sdXn) will continue to work.
1515 Say N if you are unsure.
1517 config CPU_HOTPLUG_STATE_CONTROL
1518 bool "Enable CPU hotplug state control"
1519 depends on DEBUG_KERNEL
1520 depends on HOTPLUG_CPU
1523 Allows to write steps between "offline" and "online" to the CPUs
1524 sysfs target file so states can be stepped granular. This is a debug
1525 option for now as the hotplug machinery cannot be stopped and
1526 restarted at arbitrary points yet.
1528 Say N if your are unsure.
1530 config NOTIFIER_ERROR_INJECTION
1531 tristate "Notifier error injection"
1532 depends on DEBUG_KERNEL
1535 This option provides the ability to inject artificial errors to
1536 specified notifier chain callbacks. It is useful to test the error
1537 handling of notifier call chain failures.
1541 config PM_NOTIFIER_ERROR_INJECT
1542 tristate "PM notifier error injection module"
1543 depends on PM && NOTIFIER_ERROR_INJECTION
1544 default m if PM_DEBUG
1546 This option provides the ability to inject artificial errors to
1547 PM notifier chain callbacks. It is controlled through debugfs
1548 interface /sys/kernel/debug/notifier-error-inject/pm
1550 If the notifier call chain should be failed with some events
1551 notified, write the error code to "actions/<notifier event>/error".
1553 Example: Inject PM suspend error (-12 = -ENOMEM)
1555 # cd /sys/kernel/debug/notifier-error-inject/pm/
1556 # echo -12 > actions/PM_SUSPEND_PREPARE/error
1557 # echo mem > /sys/power/state
1558 bash: echo: write error: Cannot allocate memory
1560 To compile this code as a module, choose M here: the module will
1561 be called pm-notifier-error-inject.
1565 config OF_RECONFIG_NOTIFIER_ERROR_INJECT
1566 tristate "OF reconfig notifier error injection module"
1567 depends on OF_DYNAMIC && NOTIFIER_ERROR_INJECTION
1569 This option provides the ability to inject artificial errors to
1570 OF reconfig notifier chain callbacks. It is controlled
1571 through debugfs interface under
1572 /sys/kernel/debug/notifier-error-inject/OF-reconfig/
1574 If the notifier call chain should be failed with some events
1575 notified, write the error code to "actions/<notifier event>/error".
1577 To compile this code as a module, choose M here: the module will
1578 be called of-reconfig-notifier-error-inject.
1582 config NETDEV_NOTIFIER_ERROR_INJECT
1583 tristate "Netdev notifier error injection module"
1584 depends on NET && NOTIFIER_ERROR_INJECTION
1586 This option provides the ability to inject artificial errors to
1587 netdevice notifier chain callbacks. It is controlled through debugfs
1588 interface /sys/kernel/debug/notifier-error-inject/netdev
1590 If the notifier call chain should be failed with some events
1591 notified, write the error code to "actions/<notifier event>/error".
1593 Example: Inject netdevice mtu change error (-22 = -EINVAL)
1595 # cd /sys/kernel/debug/notifier-error-inject/netdev
1596 # echo -22 > actions/NETDEV_CHANGEMTU/error
1597 # ip link set eth0 mtu 1024
1598 RTNETLINK answers: Invalid argument
1600 To compile this code as a module, choose M here: the module will
1601 be called netdev-notifier-error-inject.
1605 config FAULT_INJECTION
1606 bool "Fault-injection framework"
1607 depends on DEBUG_KERNEL
1609 Provide fault-injection framework.
1610 For more details, see Documentation/fault-injection/.
1613 bool "Fault-injection capability for kmalloc"
1614 depends on FAULT_INJECTION
1615 depends on SLAB || SLUB
1617 Provide fault-injection capability for kmalloc.
1619 config FAIL_PAGE_ALLOC
1620 bool "Fault-injection capabilitiy for alloc_pages()"
1621 depends on FAULT_INJECTION
1623 Provide fault-injection capability for alloc_pages().
1625 config FAIL_MAKE_REQUEST
1626 bool "Fault-injection capability for disk IO"
1627 depends on FAULT_INJECTION && BLOCK
1629 Provide fault-injection capability for disk IO.
1631 config FAIL_IO_TIMEOUT
1632 bool "Fault-injection capability for faking disk interrupts"
1633 depends on FAULT_INJECTION && BLOCK
1635 Provide fault-injection capability on end IO handling. This
1636 will make the block layer "forget" an interrupt as configured,
1637 thus exercising the error handling.
1639 Only works with drivers that use the generic timeout handling,
1640 for others it wont do anything.
1642 config FAIL_MMC_REQUEST
1643 bool "Fault-injection capability for MMC IO"
1644 depends on FAULT_INJECTION_DEBUG_FS && MMC
1646 Provide fault-injection capability for MMC IO.
1647 This will make the mmc core return data errors. This is
1648 useful to test the error handling in the mmc block device
1649 and to test how the mmc host driver handles retries from
1653 bool "Fault-injection capability for futexes"
1655 depends on FAULT_INJECTION && FUTEX
1657 Provide fault-injection capability for futexes.
1659 config FAULT_INJECTION_DEBUG_FS
1660 bool "Debugfs entries for fault-injection capabilities"
1661 depends on FAULT_INJECTION && SYSFS && DEBUG_FS
1663 Enable configuration of fault-injection capabilities via debugfs.
1665 config FAULT_INJECTION_STACKTRACE_FILTER
1666 bool "stacktrace filter for fault-injection capabilities"
1667 depends on FAULT_INJECTION_DEBUG_FS && STACKTRACE_SUPPORT
1670 select FRAME_POINTER if !MIPS && !PPC && !S390 && !MICROBLAZE && !ARM_UNWIND && !ARC && !SCORE
1672 Provide stacktrace filter for fault-injection capabilities
1675 bool "Latency measuring infrastructure"
1676 depends on DEBUG_KERNEL
1677 depends on STACKTRACE_SUPPORT
1679 select FRAME_POINTER if !MIPS && !PPC && !S390 && !MICROBLAZE && !ARM_UNWIND && !ARC
1686 Enable this option if you want to use the LatencyTOP tool
1687 to find out which userspace is blocking on what kernel operations.
1689 source kernel/trace/Kconfig
1691 menu "Runtime Testing"
1694 tristate "Linux Kernel Dump Test Tool Module"
1699 This module enables testing of the different dumping mechanisms by
1700 inducing system failures at predefined crash points.
1701 If you don't need it: say N
1702 Choose M here to compile this code as a module. The module will be
1705 Documentation on how to use the module can be found in
1706 Documentation/fault-injection/provoke-crashes.txt
1708 config TEST_LIST_SORT
1709 bool "Linked list sorting test"
1710 depends on DEBUG_KERNEL
1712 Enable this to turn on 'list_sort()' function test. This test is
1713 executed only once during system boot, so affects only boot time.
1717 config KPROBES_SANITY_TEST
1718 bool "Kprobes sanity tests"
1719 depends on DEBUG_KERNEL
1723 This option provides for testing basic kprobes functionality on
1724 boot. A sample kprobe, jprobe and kretprobe are inserted and
1725 verified for functionality.
1727 Say N if you are unsure.
1729 config BACKTRACE_SELF_TEST
1730 tristate "Self test for the backtrace code"
1731 depends on DEBUG_KERNEL
1734 This option provides a kernel module that can be used to test
1735 the kernel stack backtrace code. This option is not useful
1736 for distributions or general kernels, but only for kernel
1737 developers working on architecture code.
1739 Note that if you want to also test saved backtraces, you will
1740 have to enable STACKTRACE as well.
1742 Say N if you are unsure.
1745 tristate "Red-Black tree test"
1746 depends on DEBUG_KERNEL
1748 A benchmark measuring the performance of the rbtree library.
1749 Also includes rbtree invariant checks.
1751 config INTERVAL_TREE_TEST
1752 tristate "Interval tree test"
1753 depends on m && DEBUG_KERNEL
1754 select INTERVAL_TREE
1756 A benchmark measuring the performance of the interval tree library
1759 tristate "Per cpu operations test"
1760 depends on m && DEBUG_KERNEL
1762 Enable this option to build test module which validates per-cpu
1767 config ATOMIC64_SELFTEST
1768 bool "Perform an atomic64_t self-test at boot"
1770 Enable this option to test the atomic64_t functions at boot.
1774 config ASYNC_RAID6_TEST
1775 tristate "Self test for hardware accelerated raid6 recovery"
1776 depends on ASYNC_RAID6_RECOV
1779 This is a one-shot self test that permutes through the
1780 recovery of all the possible two disk failure scenarios for a
1781 N-disk array. Recovery is performed with the asynchronous
1782 raid6 recovery routines, and will optionally use an offload
1783 engine if one is available.
1788 tristate "Test functions located in the hexdump module at runtime"
1790 config TEST_STRING_HELPERS
1791 tristate "Test functions located in the string_helpers module at runtime"
1794 tristate "Test kstrto*() family of functions at runtime"
1797 tristate "Test printf() family of functions at runtime"
1800 tristate "Test bitmap_*() family of functions at runtime"
1803 Enable this option to test the bitmap functions at boot.
1808 tristate "Test functions located in the uuid module at runtime"
1810 config TEST_RHASHTABLE
1811 tristate "Perform selftest on resizable hash table"
1814 Enable this option to test the rhashtable functions at boot.
1819 tristate "Perform selftest on hash functions"
1822 Enable this option to test the kernel's integer (<linux/hash,h>)
1823 and string (<linux/stringhash.h>) hash functions on boot
1826 This is intended to help people writing architecture-specific
1827 optimized versions. If unsure, say N.
1829 endmenu # runtime tests
1831 config PROVIDE_OHCI1394_DMA_INIT
1832 bool "Remote debugging over FireWire early on boot"
1833 depends on PCI && X86
1835 If you want to debug problems which hang or crash the kernel early
1836 on boot and the crashing machine has a FireWire port, you can use
1837 this feature to remotely access the memory of the crashed machine
1838 over FireWire. This employs remote DMA as part of the OHCI1394
1839 specification which is now the standard for FireWire controllers.
1841 With remote DMA, you can monitor the printk buffer remotely using
1842 firescope and access all memory below 4GB using fireproxy from gdb.
1843 Even controlling a kernel debugger is possible using remote DMA.
1847 If ohci1394_dma=early is used as boot parameter, it will initialize
1848 all OHCI1394 controllers which are found in the PCI config space.
1850 As all changes to the FireWire bus such as enabling and disabling
1851 devices cause a bus reset and thereby disable remote DMA for all
1852 devices, be sure to have the cable plugged and FireWire enabled on
1853 the debugging host before booting the debug target for debugging.
1855 This code (~1k) is freed after boot. By then, the firewire stack
1856 in charge of the OHCI-1394 controllers should be used instead.
1858 See Documentation/debugging-via-ohci1394.txt for more information.
1860 config DMA_API_DEBUG
1861 bool "Enable debugging of DMA-API usage"
1862 depends on HAVE_DMA_API_DEBUG
1864 Enable this option to debug the use of the DMA API by device drivers.
1865 With this option you will be able to detect common bugs in device
1866 drivers like double-freeing of DMA mappings or freeing mappings that
1867 were never allocated.
1869 This also attempts to catch cases where a page owned by DMA is
1870 accessed by the cpu in a way that could cause data corruption. For
1871 example, this enables cow_user_page() to check that the source page is
1874 This option causes a performance degradation. Use only if you want to
1875 debug device drivers and dma interactions.
1880 tristate "Test module loading with 'hello world' module"
1884 This builds the "test_module" module that emits "Hello, world"
1885 on printk when loaded. It is designed to be used for basic
1886 evaluation of the module loading subsystem (for example when
1887 validating module verification). It lacks any extra dependencies,
1888 and will not normally be loaded by the system unless explicitly
1893 config TEST_USER_COPY
1894 tristate "Test user/kernel boundary protections"
1898 This builds the "test_user_copy" module that runs sanity checks
1899 on the copy_to/from_user infrastructure, making sure basic
1900 user/kernel boundary testing is working. If it fails to load,
1901 a regression has been detected in the user/kernel memory boundary
1907 tristate "Test BPF filter functionality"
1911 This builds the "test_bpf" module that runs various test vectors
1912 against the BPF interpreter or BPF JIT compiler depending on the
1913 current setting. This is in particular useful for BPF JIT compiler
1914 development, but also to run regression tests against changes in
1915 the interpreter code. It also enables test stubs for eBPF maps and
1916 verifier used by user space verifier testsuite.
1920 config TEST_FIRMWARE
1921 tristate "Test firmware loading via userspace interface"
1923 depends on FW_LOADER
1925 This builds the "test_firmware" module that creates a userspace
1926 interface for testing firmware loading. This can be used to
1927 control the triggering of firmware loading without needing an
1928 actual firmware-using device. The contents can be rechecked by
1934 tristate "udelay test driver"
1937 This builds the "udelay_test" module that helps to make sure
1938 that udelay() is working properly.
1944 depends on HAVE_MEMBLOCK
1946 This option adds a kernel parameter 'memtest', which allows memtest
1948 memtest=0, mean disabled; -- default
1949 memtest=1, mean do 1 test pattern;
1951 memtest=17, mean do 17 test patterns.
1952 If you are unsure how to answer this question, answer N.
1954 config TEST_STATIC_KEYS
1955 tristate "Test static keys"
1959 Test the static key interfaces.
1963 config BUG_ON_DATA_CORRUPTION
1964 bool "Trigger a BUG when data corruption is detected"
1967 Select this option if the kernel should BUG when it encounters
1968 data corruption in kernel memory structures when they get checked
1973 source "samples/Kconfig"
1975 source "lib/Kconfig.kgdb"
1977 source "lib/Kconfig.ubsan"
1979 config ARCH_HAS_DEVMEM_IS_ALLOWED
1982 config STRICT_DEVMEM
1983 bool "Filter access to /dev/mem"
1984 depends on MMU && DEVMEM
1985 depends on ARCH_HAS_DEVMEM_IS_ALLOWED
1986 default y if TILE || PPC
1988 If this option is disabled, you allow userspace (root) access to all
1989 of memory, including kernel and userspace memory. Accidental
1990 access to this is obviously disastrous, but specific access can
1991 be used by people debugging the kernel. Note that with PAT support
1992 enabled, even in this case there are restrictions on /dev/mem
1993 use due to the cache aliasing requirements.
1995 If this option is switched on, and IO_STRICT_DEVMEM=n, the /dev/mem
1996 file only allows userspace access to PCI space and the BIOS code and
1997 data regions. This is sufficient for dosemu and X and all common
2002 config IO_STRICT_DEVMEM
2003 bool "Filter I/O access to /dev/mem"
2004 depends on STRICT_DEVMEM
2006 If this option is disabled, you allow userspace (root) access to all
2007 io-memory regardless of whether a driver is actively using that
2008 range. Accidental access to this is obviously disastrous, but
2009 specific access can be used by people debugging kernel drivers.
2011 If this option is switched on, the /dev/mem file only allows
2012 userspace access to *idle* io-memory ranges (see /proc/iomem) This
2013 may break traditional users of /dev/mem (dosemu, legacy X, etc...)
2014 if the driver using a given range cannot be disabled.