1 # SPDX-License-Identifier: GPL-2.0-only
4 default "$(CC_VERSION_TEXT)"
6 This is used in unclear ways:
8 - Re-run Kconfig when the compiler is updated
9 The 'default' property references the environment variable,
10 CC_VERSION_TEXT so it is recorded in include/config/auto.conf.cmd.
11 When the compiler is updated, Kconfig will be invoked.
13 - Ensure full rebuild when the compiler is updated
14 include/linux/compiler-version.h contains this option in the comment
15 line so fixdep adds include/config/CC_VERSION_TEXT into the
16 auto-generated dependency. When the compiler is updated, syncconfig
17 will touch it and then every file will be rebuilt.
20 def_bool $(success,test "$(cc-name)" = GCC)
24 default $(cc-version) if CC_IS_GCC
28 def_bool $(success,test "$(cc-name)" = Clang)
32 default $(cc-version) if CC_IS_CLANG
36 def_bool $(success,test "$(as-name)" = GNU)
39 def_bool $(success,test "$(as-name)" = LLVM)
43 # Use clang version if this is the integrated assembler
44 default CLANG_VERSION if AS_IS_LLVM
48 def_bool $(success,test "$(ld-name)" = BFD)
52 default $(ld-version) if LD_IS_BFD
56 def_bool $(success,test "$(ld-name)" = LLD)
60 default $(ld-version) if LD_IS_LLD
65 default $(success,$(srctree)/scripts/cc-can-link.sh $(CC) $(CLANG_FLAGS) $(m64-flag)) if 64BIT
66 default $(success,$(srctree)/scripts/cc-can-link.sh $(CC) $(CLANG_FLAGS) $(m32-flag))
68 config CC_CAN_LINK_STATIC
70 default $(success,$(srctree)/scripts/cc-can-link.sh $(CC) $(CLANG_FLAGS) $(m64-flag) -static) if 64BIT
71 default $(success,$(srctree)/scripts/cc-can-link.sh $(CC) $(CLANG_FLAGS) $(m32-flag) -static)
73 config CC_HAS_ASM_GOTO
74 def_bool $(success,$(srctree)/scripts/gcc-goto.sh $(CC))
76 config CC_HAS_ASM_GOTO_OUTPUT
77 depends on CC_HAS_ASM_GOTO
78 def_bool $(success,echo 'int foo(int x) { asm goto ("": "=r"(x) ::: bar); return x; bar: return 0; }' | $(CC) -x c - -c -o /dev/null)
80 config CC_HAS_ASM_GOTO_TIED_OUTPUT
81 depends on CC_HAS_ASM_GOTO_OUTPUT
82 # Detect buggy gcc and clang, fixed in gcc-11 clang-14.
83 def_bool $(success,echo 'int foo(int *x) { asm goto (".long (%l[bar]) - .": "+m"(*x) ::: bar); return *x; bar: return 0; }' | $CC -x c - -c -o /dev/null)
85 config TOOLS_SUPPORT_RELR
86 def_bool $(success,env "CC=$(CC)" "LD=$(LD)" "NM=$(NM)" "OBJCOPY=$(OBJCOPY)" $(srctree)/scripts/tools-support-relr.sh)
88 config CC_HAS_ASM_INLINE
89 def_bool $(success,echo 'void foo(void) { asm inline (""); }' | $(CC) -x c - -c -o /dev/null)
91 config CC_HAS_NO_PROFILE_FN_ATTR
92 def_bool $(success,echo '__attribute__((no_profile_instrument_function)) int x();' | $(CC) -x c - -c -o /dev/null -Werror)
96 default $(shell,$(srctree)/scripts/pahole-version.sh $(PAHOLE))
104 config BUILDTIME_TABLE_SORT
107 config THREAD_INFO_IN_TASK
110 Select this to move thread_info off the stack into task_struct. To
111 make this work, an arch will need to remove all thread_info fields
112 except flags and fix any runtime bugs.
114 One subtle change that will be needed is to use try_get_task_stack()
115 and put_task_stack() in save_thread_stack_tsk() and get_wchan().
124 depends on BROKEN || !SMP
127 config INIT_ENV_ARG_LIMIT
132 Maximum of each of the number of arguments and environment
133 variables passed to init from the kernel command line.
136 bool "Compile also drivers which will not load"
139 Some drivers can be compiled on a different platform than they are
140 intended to be run on. Despite they cannot be loaded there (or even
141 when they load they cannot be used due to missing HW support),
142 developers still, opposing to distributors, might want to build such
143 drivers to compile-test them.
145 If you are a developer and want to build everything available, say Y
146 here. If you are a user/distributor, say N here to exclude useless
147 drivers to be distributed.
150 bool "Compile the kernel with warnings as errors"
153 A kernel build should not cause any compiler warnings, and this
154 enables the '-Werror' flag to enforce that rule by default.
156 However, if you have a new (or very old) compiler with odd and
157 unusual warnings, or you have some architecture with problems,
158 you may need to disable this config option in order to
159 successfully build the kernel.
163 config UAPI_HEADER_TEST
164 bool "Compile test UAPI headers"
165 depends on HEADERS_INSTALL && CC_CAN_LINK
167 Compile test headers exported to user-space to ensure they are
168 self-contained, i.e. compilable as standalone units.
170 If you are a developer or tester and want to ensure the exported
171 headers are self-contained, say Y here. Otherwise, choose N.
174 string "Local version - append to kernel release"
176 Append an extra string to the end of your kernel version.
177 This will show up when you type uname, for example.
178 The string you set here will be appended after the contents of
179 any files with a filename matching localversion* in your
180 object and source tree, in that order. Your total string can
181 be a maximum of 64 characters.
183 config LOCALVERSION_AUTO
184 bool "Automatically append version information to the version string"
186 depends on !COMPILE_TEST
188 This will try to automatically determine if the current tree is a
189 release tree by looking for git tags that belong to the current
190 top of tree revision.
192 A string of the format -gxxxxxxxx will be added to the localversion
193 if a git-based tree is found. The string generated by this will be
194 appended after any matching localversion* files, and after the value
195 set in CONFIG_LOCALVERSION.
197 (The actual string used here is the first eight characters produced
198 by running the command:
200 $ git rev-parse --verify HEAD
202 which is done within the script "scripts/setlocalversion".)
205 string "Build ID Salt"
208 The build ID is used to link binaries and their debug info. Setting
209 this option will use the value in the calculation of the build id.
210 This is mostly useful for distributions which want to ensure the
211 build is unique between builds. It's safe to leave the default.
213 config HAVE_KERNEL_GZIP
216 config HAVE_KERNEL_BZIP2
219 config HAVE_KERNEL_LZMA
222 config HAVE_KERNEL_XZ
225 config HAVE_KERNEL_LZO
228 config HAVE_KERNEL_LZ4
231 config HAVE_KERNEL_ZSTD
234 config HAVE_KERNEL_UNCOMPRESSED
238 prompt "Kernel compression mode"
240 depends on HAVE_KERNEL_GZIP || HAVE_KERNEL_BZIP2 || HAVE_KERNEL_LZMA || HAVE_KERNEL_XZ || HAVE_KERNEL_LZO || HAVE_KERNEL_LZ4 || HAVE_KERNEL_ZSTD || HAVE_KERNEL_UNCOMPRESSED
242 The linux kernel is a kind of self-extracting executable.
243 Several compression algorithms are available, which differ
244 in efficiency, compression and decompression speed.
245 Compression speed is only relevant when building a kernel.
246 Decompression speed is relevant at each boot.
248 If you have any problems with bzip2 or lzma compressed
249 kernels, mail me (Alain Knaff) <alain@knaff.lu>. (An older
250 version of this functionality (bzip2 only), for 2.4, was
251 supplied by Christian Ludwig)
253 High compression options are mostly useful for users, who
254 are low on disk space (embedded systems), but for whom ram
257 If in doubt, select 'gzip'
261 depends on HAVE_KERNEL_GZIP
263 The old and tried gzip compression. It provides a good balance
264 between compression ratio and decompression speed.
268 depends on HAVE_KERNEL_BZIP2
270 Its compression ratio and speed is intermediate.
271 Decompression speed is slowest among the choices. The kernel
272 size is about 10% smaller with bzip2, in comparison to gzip.
273 Bzip2 uses a large amount of memory. For modern kernels you
274 will need at least 8MB RAM or more for booting.
278 depends on HAVE_KERNEL_LZMA
280 This compression algorithm's ratio is best. Decompression speed
281 is between gzip and bzip2. Compression is slowest.
282 The kernel size is about 33% smaller with LZMA in comparison to gzip.
286 depends on HAVE_KERNEL_XZ
288 XZ uses the LZMA2 algorithm and instruction set specific
289 BCJ filters which can improve compression ratio of executable
290 code. The size of the kernel is about 30% smaller with XZ in
291 comparison to gzip. On architectures for which there is a BCJ
292 filter (i386, x86_64, ARM, IA-64, PowerPC, and SPARC), XZ
293 will create a few percent smaller kernel than plain LZMA.
295 The speed is about the same as with LZMA: The decompression
296 speed of XZ is better than that of bzip2 but worse than gzip
297 and LZO. Compression is slow.
301 depends on HAVE_KERNEL_LZO
303 Its compression ratio is the poorest among the choices. The kernel
304 size is about 10% bigger than gzip; however its speed
305 (both compression and decompression) is the fastest.
309 depends on HAVE_KERNEL_LZ4
311 LZ4 is an LZ77-type compressor with a fixed, byte-oriented encoding.
312 A preliminary version of LZ4 de/compression tool is available at
313 <https://code.google.com/p/lz4/>.
315 Its compression ratio is worse than LZO. The size of the kernel
316 is about 8% bigger than LZO. But the decompression speed is
321 depends on HAVE_KERNEL_ZSTD
323 ZSTD is a compression algorithm targeting intermediate compression
324 with fast decompression speed. It will compress better than GZIP and
325 decompress around the same speed as LZO, but slower than LZ4. You
326 will need at least 192 KB RAM or more for booting. The zstd command
327 line tool is required for compression.
329 config KERNEL_UNCOMPRESSED
331 depends on HAVE_KERNEL_UNCOMPRESSED
333 Produce uncompressed kernel image. This option is usually not what
334 you want. It is useful for debugging the kernel in slow simulation
335 environments, where decompressing and moving the kernel is awfully
336 slow. This option allows early boot code to skip the decompressor
337 and jump right at uncompressed kernel image.
342 string "Default init path"
345 This option determines the default init for the system if no init=
346 option is passed on the kernel command line. If the requested path is
347 not present, we will still then move on to attempting further
348 locations (e.g. /sbin/init, etc). If this is empty, we will just use
349 the fallback list when init= is not passed.
351 config DEFAULT_HOSTNAME
352 string "Default hostname"
355 This option determines the default system hostname before userspace
356 calls sethostname(2). The kernel traditionally uses "(none)" here,
357 but you may wish to use a different default here to make a minimal
358 system more usable with less configuration.
361 # For some reason microblaze and nios2 hard code SWAP=n. Hopefully we can
362 # add proper SWAP support to them, in which case this can be remove.
367 config VERSION_SIGNATURE
368 string "Arbitrary version signature"
370 This string will be created in a file, /proc/version_signature. It
371 is useful in determining arbitrary data about your kernel. For instance,
372 if you have several kernels of the same version, but need to keep track
373 of a revision of the same kernel, but not affect it's ability to load
374 compatible modules, this is the easiest way to do that.
377 bool "Support for paging of anonymous memory (swap)"
378 depends on MMU && BLOCK && !ARCH_NO_SWAP
381 This option allows you to choose whether you want to have support
382 for so called swap devices or swap files in your kernel that are
383 used to provide more virtual memory than the actual RAM present
384 in your computer. If unsure say Y.
389 Inter Process Communication is a suite of library functions and
390 system calls which let processes (running programs) synchronize and
391 exchange information. It is generally considered to be a good thing,
392 and some programs won't run unless you say Y here. In particular, if
393 you want to run the DOS emulator dosemu under Linux (read the
394 DOSEMU-HOWTO, available from <http://www.tldp.org/docs.html#howto>),
395 you'll need to say Y here.
397 You can find documentation about IPC with "info ipc" and also in
398 section 6.4 of the Linux Programmer's Guide, available from
399 <http://www.tldp.org/guides.html>.
401 config SYSVIPC_SYSCTL
408 bool "POSIX Message Queues"
411 POSIX variant of message queues is a part of IPC. In POSIX message
412 queues every message has a priority which decides about succession
413 of receiving it by a process. If you want to compile and run
414 programs written e.g. for Solaris with use of its POSIX message
415 queues (functions mq_*) say Y here.
417 POSIX message queues are visible as a filesystem called 'mqueue'
418 and can be mounted somewhere if you want to do filesystem
419 operations on message queues.
423 config POSIX_MQUEUE_SYSCTL
425 depends on POSIX_MQUEUE
430 bool "General notification queue"
434 This is a general notification queue for the kernel to pass events to
435 userspace by splicing them into pipes. It can be used in conjunction
436 with watches for key/keyring change notifications and device
439 See Documentation/watch_queue.rst
441 config CROSS_MEMORY_ATTACH
442 bool "Enable process_vm_readv/writev syscalls"
446 Enabling this option adds the system calls process_vm_readv and
447 process_vm_writev which allow a process with the correct privileges
448 to directly read from or write to another process' address space.
449 See the man page for more details.
452 bool "uselib syscall"
453 def_bool ALPHA || M68K || SPARC || X86_32 || IA32_EMULATION
455 This option enables the uselib syscall, a system call used in the
456 dynamic linker from libc5 and earlier. glibc does not use this
457 system call. If you intend to run programs built on libc5 or
458 earlier, you may need to enable this syscall. Current systems
459 running glibc can safely disable this.
462 bool "Auditing support"
465 Enable auditing infrastructure that can be used with another
466 kernel subsystem, such as SELinux (which requires this for
467 logging of avc messages output). System call auditing is included
468 on architectures which support it.
470 config HAVE_ARCH_AUDITSYSCALL
475 depends on AUDIT && HAVE_ARCH_AUDITSYSCALL
478 source "kernel/irq/Kconfig"
479 source "kernel/time/Kconfig"
480 source "kernel/bpf/Kconfig"
481 source "kernel/Kconfig.preempt"
483 menu "CPU/Task time and stats accounting"
485 config VIRT_CPU_ACCOUNTING
489 prompt "Cputime accounting"
490 default TICK_CPU_ACCOUNTING if !PPC64
491 default VIRT_CPU_ACCOUNTING_NATIVE if PPC64
493 # Kind of a stub config for the pure tick based cputime accounting
494 config TICK_CPU_ACCOUNTING
495 bool "Simple tick based cputime accounting"
496 depends on !S390 && !NO_HZ_FULL
498 This is the basic tick based cputime accounting that maintains
499 statistics about user, system and idle time spent on per jiffies
504 config VIRT_CPU_ACCOUNTING_NATIVE
505 bool "Deterministic task and CPU time accounting"
506 depends on HAVE_VIRT_CPU_ACCOUNTING && !NO_HZ_FULL
507 select VIRT_CPU_ACCOUNTING
509 Select this option to enable more accurate task and CPU time
510 accounting. This is done by reading a CPU counter on each
511 kernel entry and exit and on transitions within the kernel
512 between system, softirq and hardirq state, so there is a
513 small performance impact. In the case of s390 or IBM POWER > 5,
514 this also enables accounting of stolen time on logically-partitioned
517 config VIRT_CPU_ACCOUNTING_GEN
518 bool "Full dynticks CPU time accounting"
519 depends on HAVE_CONTEXT_TRACKING
520 depends on HAVE_VIRT_CPU_ACCOUNTING_GEN
521 depends on GENERIC_CLOCKEVENTS
522 select VIRT_CPU_ACCOUNTING
523 select CONTEXT_TRACKING
525 Select this option to enable task and CPU time accounting on full
526 dynticks systems. This accounting is implemented by watching every
527 kernel-user boundaries using the context tracking subsystem.
528 The accounting is thus performed at the expense of some significant
531 For now this is only useful if you are working on the full
532 dynticks subsystem development.
538 config IRQ_TIME_ACCOUNTING
539 bool "Fine granularity task level IRQ time accounting"
540 depends on HAVE_IRQ_TIME_ACCOUNTING && !VIRT_CPU_ACCOUNTING_NATIVE
542 Select this option to enable fine granularity task irq time
543 accounting. This is done by reading a timestamp on each
544 transitions between softirq and hardirq state, so there can be a
545 small performance impact.
547 If in doubt, say N here.
549 config HAVE_SCHED_AVG_IRQ
551 depends on IRQ_TIME_ACCOUNTING || PARAVIRT_TIME_ACCOUNTING
554 config SCHED_THERMAL_PRESSURE
556 default y if ARM && ARM_CPU_TOPOLOGY
559 depends on CPU_FREQ_THERMAL
561 Select this option to enable thermal pressure accounting in the
562 scheduler. Thermal pressure is the value conveyed to the scheduler
563 that reflects the reduction in CPU compute capacity resulted from
564 thermal throttling. Thermal throttling occurs when the performance of
565 a CPU is capped due to high operating temperatures.
567 If selected, the scheduler will be able to balance tasks accordingly,
568 i.e. put less load on throttled CPUs than on non/less throttled ones.
570 This requires the architecture to implement
571 arch_set_thermal_pressure() and arch_scale_thermal_pressure().
573 config BSD_PROCESS_ACCT
574 bool "BSD Process Accounting"
577 If you say Y here, a user level program will be able to instruct the
578 kernel (via a special system call) to write process accounting
579 information to a file: whenever a process exits, information about
580 that process will be appended to the file by the kernel. The
581 information includes things such as creation time, owning user,
582 command name, memory usage, controlling terminal etc. (the complete
583 list is in the struct acct in <file:include/linux/acct.h>). It is
584 up to the user level program to do useful things with this
585 information. This is generally a good idea, so say Y.
587 config BSD_PROCESS_ACCT_V3
588 bool "BSD Process Accounting version 3 file format"
589 depends on BSD_PROCESS_ACCT
592 If you say Y here, the process accounting information is written
593 in a new file format that also logs the process IDs of each
594 process and its parent. Note that this file format is incompatible
595 with previous v0/v1/v2 file formats, so you will need updated tools
596 for processing it. A preliminary version of these tools is available
597 at <http://www.gnu.org/software/acct/>.
600 bool "Export task/process statistics through netlink"
605 Export selected statistics for tasks/processes through the
606 generic netlink interface. Unlike BSD process accounting, the
607 statistics are available during the lifetime of tasks/processes as
608 responses to commands. Like BSD accounting, they are sent to user
613 config TASK_DELAY_ACCT
614 bool "Enable per-task delay accounting"
618 Collect information on time spent by a task waiting for system
619 resources like cpu, synchronous block I/O completion and swapping
620 in pages. Such statistics can help in setting a task's priorities
621 relative to other tasks for cpu, io, rss limits etc.
626 bool "Enable extended accounting over taskstats"
629 Collect extended task accounting data and send the data
630 to userland for processing over the taskstats interface.
634 config TASK_IO_ACCOUNTING
635 bool "Enable per-task storage I/O accounting"
636 depends on TASK_XACCT
638 Collect information on the number of bytes of storage I/O which this
644 bool "Pressure stall information tracking"
646 Collect metrics that indicate how overcommitted the CPU, memory,
647 and IO capacity are in the system.
649 If you say Y here, the kernel will create /proc/pressure/ with the
650 pressure statistics files cpu, memory, and io. These will indicate
651 the share of walltime in which some or all tasks in the system are
652 delayed due to contention of the respective resource.
654 In kernels with cgroup support, cgroups (cgroup2 only) will
655 have cpu.pressure, memory.pressure, and io.pressure files,
656 which aggregate pressure stalls for the grouped tasks only.
658 For more details see Documentation/accounting/psi.rst.
662 config PSI_DEFAULT_DISABLED
663 bool "Require boot parameter to enable pressure stall information tracking"
667 If set, pressure stall information tracking will be disabled
668 per default but can be enabled through passing psi=1 on the
669 kernel commandline during boot.
671 This feature adds some code to the task wakeup and sleep
672 paths of the scheduler. The overhead is too low to affect
673 common scheduling-intense workloads in practice (such as
674 webservers, memcache), but it does show up in artificial
675 scheduler stress tests, such as hackbench.
677 If you are paranoid and not sure what the kernel will be
682 endmenu # "CPU/Task time and stats accounting"
686 depends on SMP || COMPILE_TEST
689 Make sure that CPUs running critical tasks are not disturbed by
690 any source of "noise" such as unbound workqueues, timers, kthreads...
691 Unbound jobs get offloaded to housekeeping CPUs. This is driven by
692 the "isolcpus=" boot parameter.
696 source "kernel/rcu/Kconfig"
703 tristate "Kernel .config support"
705 This option enables the complete Linux kernel ".config" file
706 contents to be saved in the kernel. It provides documentation
707 of which kernel options are used in a running kernel or in an
708 on-disk kernel. This information can be extracted from the kernel
709 image file with the script scripts/extract-ikconfig and used as
710 input to rebuild the current kernel or to build another kernel.
711 It can also be extracted from a running kernel by reading
712 /proc/config.gz if enabled (below).
715 bool "Enable access to .config through /proc/config.gz"
716 depends on IKCONFIG && PROC_FS
718 This option enables access to the kernel configuration file
719 through /proc/config.gz.
722 tristate "Enable kernel headers through /sys/kernel/kheaders.tar.xz"
725 This option enables access to the in-kernel headers that are generated during
726 the build process. These can be used to build eBPF tracing programs,
727 or similar programs. If you build the headers as a module, a module called
728 kheaders.ko is built which can be loaded on-demand to get access to headers.
731 int "Kernel log buffer size (16 => 64KB, 17 => 128KB)"
732 range 12 25 if !H8300
737 Select the minimal kernel log buffer size as a power of 2.
738 The final size is affected by LOG_CPU_MAX_BUF_SHIFT config
739 parameter, see below. Any higher size also might be forced
740 by "log_buf_len" boot parameter.
750 config LOG_CPU_MAX_BUF_SHIFT
751 int "CPU kernel log buffer size contribution (13 => 8 KB, 17 => 128KB)"
754 default 12 if !BASE_SMALL
755 default 0 if BASE_SMALL
758 This option allows to increase the default ring buffer size
759 according to the number of CPUs. The value defines the contribution
760 of each CPU as a power of 2. The used space is typically only few
761 lines however it might be much more when problems are reported,
764 The increased size means that a new buffer has to be allocated and
765 the original static one is unused. It makes sense only on systems
766 with more CPUs. Therefore this value is used only when the sum of
767 contributions is greater than the half of the default kernel ring
768 buffer as defined by LOG_BUF_SHIFT. The default values are set
769 so that more than 16 CPUs are needed to trigger the allocation.
771 Also this option is ignored when "log_buf_len" kernel parameter is
772 used as it forces an exact (power of two) size of the ring buffer.
774 The number of possible CPUs is used for this computation ignoring
775 hotplugging making the computation optimal for the worst case
776 scenario while allowing a simple algorithm to be used from bootup.
778 Examples shift values and their meaning:
779 17 => 128 KB for each CPU
780 16 => 64 KB for each CPU
781 15 => 32 KB for each CPU
782 14 => 16 KB for each CPU
783 13 => 8 KB for each CPU
784 12 => 4 KB for each CPU
786 config PRINTK_SAFE_LOG_BUF_SHIFT
787 int "Temporary per-CPU printk log buffer size (12 => 4KB, 13 => 8KB)"
792 Select the size of an alternate printk per-CPU buffer where messages
793 printed from usafe contexts are temporary stored. One example would
794 be NMI messages, another one - printk recursion. The messages are
795 copied to the main log buffer in a safe context to avoid a deadlock.
796 The value defines the size as a power of 2.
798 Those messages are rare and limited. The largest one is when
799 a backtrace is printed. It usually fits into 4KB. Select
800 8KB if you want to be on the safe side.
803 17 => 128 KB for each CPU
804 16 => 64 KB for each CPU
805 15 => 32 KB for each CPU
806 14 => 16 KB for each CPU
807 13 => 8 KB for each CPU
808 12 => 4 KB for each CPU
811 bool "Printk indexing debugfs interface"
812 depends on PRINTK && DEBUG_FS
814 Add support for indexing of all printk formats known at compile time
815 at <debugfs>/printk/index/<module>.
817 This can be used as part of maintaining daemons which monitor
818 /dev/kmsg, as it permits auditing the printk formats present in a
819 kernel, allowing detection of cases where monitored printks are
820 changed or no longer present.
822 There is no additional runtime cost to printk with this enabled.
825 # Architectures with an unreliable sched_clock() should select this:
827 config HAVE_UNSTABLE_SCHED_CLOCK
830 config GENERIC_SCHED_CLOCK
833 menu "Scheduler features"
836 bool "Enable utilization clamping for RT/FAIR tasks"
837 depends on CPU_FREQ_GOV_SCHEDUTIL
839 This feature enables the scheduler to track the clamped utilization
840 of each CPU based on RUNNABLE tasks scheduled on that CPU.
842 With this option, the user can specify the min and max CPU
843 utilization allowed for RUNNABLE tasks. The max utilization defines
844 the maximum frequency a task should use while the min utilization
845 defines the minimum frequency it should use.
847 Both min and max utilization clamp values are hints to the scheduler,
848 aiming at improving its frequency selection policy, but they do not
849 enforce or grant any specific bandwidth for tasks.
853 config UCLAMP_BUCKETS_COUNT
854 int "Number of supported utilization clamp buckets"
857 depends on UCLAMP_TASK
859 Defines the number of clamp buckets to use. The range of each bucket
860 will be SCHED_CAPACITY_SCALE/UCLAMP_BUCKETS_COUNT. The higher the
861 number of clamp buckets the finer their granularity and the higher
862 the precision of clamping aggregation and tracking at run-time.
864 For example, with the minimum configuration value we will have 5
865 clamp buckets tracking 20% utilization each. A 25% boosted tasks will
866 be refcounted in the [20..39]% bucket and will set the bucket clamp
867 effective value to 25%.
868 If a second 30% boosted task should be co-scheduled on the same CPU,
869 that task will be refcounted in the same bucket of the first task and
870 it will boost the bucket clamp effective value to 30%.
871 The clamp effective value of a bucket is reset to its nominal value
872 (20% in the example above) when there are no more tasks refcounted in
875 An additional boost/capping margin can be added to some tasks. In the
876 example above the 25% task will be boosted to 30% until it exits the
877 CPU. If that should be considered not acceptable on certain systems,
878 it's always possible to reduce the margin by increasing the number of
879 clamp buckets to trade off used memory for run-time tracking
882 If in doubt, use the default value.
887 # For architectures that want to enable the support for NUMA-affine scheduler
890 config ARCH_SUPPORTS_NUMA_BALANCING
894 # For architectures that prefer to flush all TLBs after a number of pages
895 # are unmapped instead of sending one IPI per page to flush. The architecture
896 # must provide guarantees on what happens if a clean TLB cache entry is
897 # written after the unmap. Details are in mm/rmap.c near the check for
898 # should_defer_flush. The architecture should also consider if the full flush
899 # and the refill costs are offset by the savings of sending fewer IPIs.
900 config ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
904 def_bool !$(cc-option,$(m64-flag) -D__SIZEOF_INT128__=0) && 64BIT
907 # For architectures that know their GCC __int128 support is sound
909 config ARCH_SUPPORTS_INT128
912 # For architectures that (ab)use NUMA to represent different memory regions
913 # all cpu-local but of different latencies, such as SuperH.
915 config ARCH_WANT_NUMA_VARIABLE_LOCALITY
918 config NUMA_BALANCING
919 bool "Memory placement aware NUMA scheduler"
920 depends on ARCH_SUPPORTS_NUMA_BALANCING
921 depends on !ARCH_WANT_NUMA_VARIABLE_LOCALITY
922 depends on SMP && NUMA && MIGRATION
924 This option adds support for automatic NUMA aware memory/task placement.
925 The mechanism is quite primitive and is based on migrating memory when
926 it has references to the node the task is running on.
928 This system will be inactive on UMA systems.
930 config NUMA_BALANCING_DEFAULT_ENABLED
931 bool "Automatically enable NUMA aware memory/task placement"
933 depends on NUMA_BALANCING
935 If set, automatic NUMA balancing will be enabled if running on a NUMA
939 bool "Control Group support"
942 This option adds support for grouping sets of processes together, for
943 use with process control subsystems such as Cpusets, CFS, memory
944 controls or device isolation.
946 - Documentation/scheduler/sched-design-CFS.rst (CFS)
947 - Documentation/admin-guide/cgroup-v1/ (features for grouping, isolation
948 and resource control)
958 bool "Memory controller"
962 Provides control over the memory footprint of tasks in a cgroup.
966 depends on MEMCG && SWAP
971 depends on MEMCG && !SLOB
979 Generic block IO controller cgroup interface. This is the common
980 cgroup interface which should be used by various IO controlling
983 Currently, CFQ IO scheduler uses it to recognize task groups and
984 control disk bandwidth allocation (proportional time slice allocation)
985 to such task groups. It is also used by bio throttling logic in
986 block layer to implement upper limit in IO rates on a device.
988 This option only enables generic Block IO controller infrastructure.
989 One needs to also enable actual IO controlling logic/policy. For
990 enabling proportional weight division of disk bandwidth in CFQ, set
991 CONFIG_BFQ_GROUP_IOSCHED=y; for enabling throttling policy, set
992 CONFIG_BLK_DEV_THROTTLING=y.
994 See Documentation/admin-guide/cgroup-v1/blkio-controller.rst for more information.
996 config CGROUP_WRITEBACK
998 depends on MEMCG && BLK_CGROUP
1001 menuconfig CGROUP_SCHED
1002 bool "CPU controller"
1005 This feature lets CPU scheduler recognize task groups and control CPU
1006 bandwidth allocation to such task groups. It uses cgroups to group
1010 config FAIR_GROUP_SCHED
1011 bool "Group scheduling for SCHED_OTHER"
1012 depends on CGROUP_SCHED
1013 default CGROUP_SCHED
1015 config CFS_BANDWIDTH
1016 bool "CPU bandwidth provisioning for FAIR_GROUP_SCHED"
1017 depends on FAIR_GROUP_SCHED
1020 This option allows users to define CPU bandwidth rates (limits) for
1021 tasks running within the fair group scheduler. Groups with no limit
1022 set are considered to be unconstrained and will run with no
1024 See Documentation/scheduler/sched-bwc.rst for more information.
1026 config RT_GROUP_SCHED
1027 bool "Group scheduling for SCHED_RR/FIFO"
1028 depends on CGROUP_SCHED
1031 This feature lets you explicitly allocate real CPU bandwidth
1032 to task groups. If enabled, it will also make it impossible to
1033 schedule realtime tasks for non-root users until you allocate
1034 realtime bandwidth for them.
1035 See Documentation/scheduler/sched-rt-group.rst for more information.
1039 config UCLAMP_TASK_GROUP
1040 bool "Utilization clamping per group of tasks"
1041 depends on CGROUP_SCHED
1042 depends on UCLAMP_TASK
1045 This feature enables the scheduler to track the clamped utilization
1046 of each CPU based on RUNNABLE tasks currently scheduled on that CPU.
1048 When this option is enabled, the user can specify a min and max
1049 CPU bandwidth which is allowed for each single task in a group.
1050 The max bandwidth allows to clamp the maximum frequency a task
1051 can use, while the min bandwidth allows to define a minimum
1052 frequency a task will always use.
1054 When task group based utilization clamping is enabled, an eventually
1055 specified task-specific clamp value is constrained by the cgroup
1056 specified clamp value. Both minimum and maximum task clamping cannot
1057 be bigger than the corresponding clamping defined at task group level.
1062 bool "PIDs controller"
1064 Provides enforcement of process number limits in the scope of a
1065 cgroup. Any attempt to fork more processes than is allowed in the
1066 cgroup will fail. PIDs are fundamentally a global resource because it
1067 is fairly trivial to reach PID exhaustion before you reach even a
1068 conservative kmemcg limit. As a result, it is possible to grind a
1069 system to halt without being limited by other cgroup policies. The
1070 PIDs controller is designed to stop this from happening.
1072 It should be noted that organisational operations (such as attaching
1073 to a cgroup hierarchy) will *not* be blocked by the PIDs controller,
1074 since the PIDs limit only affects a process's ability to fork, not to
1078 bool "RDMA controller"
1080 Provides enforcement of RDMA resources defined by IB stack.
1081 It is fairly easy for consumers to exhaust RDMA resources, which
1082 can result into resource unavailability to other consumers.
1083 RDMA controller is designed to stop this from happening.
1084 Attaching processes with active RDMA resources to the cgroup
1085 hierarchy is allowed even if can cross the hierarchy's limit.
1087 config CGROUP_FREEZER
1088 bool "Freezer controller"
1090 Provides a way to freeze and unfreeze all tasks in a
1093 This option affects the ORIGINAL cgroup interface. The cgroup2 memory
1094 controller includes important in-kernel memory consumers per default.
1096 If you're using cgroup2, say N.
1098 config CGROUP_HUGETLB
1099 bool "HugeTLB controller"
1100 depends on HUGETLB_PAGE
1104 Provides a cgroup controller for HugeTLB pages.
1105 When you enable this, you can put a per cgroup limit on HugeTLB usage.
1106 The limit is enforced during page fault. Since HugeTLB doesn't
1107 support page reclaim, enforcing the limit at page fault time implies
1108 that, the application will get SIGBUS signal if it tries to access
1109 HugeTLB pages beyond its limit. This requires the application to know
1110 beforehand how much HugeTLB pages it would require for its use. The
1111 control group is tracked in the third page lru pointer. This means
1112 that we cannot use the controller with huge page less than 3 pages.
1115 bool "Cpuset controller"
1118 This option will let you create and manage CPUSETs which
1119 allow dynamically partitioning a system into sets of CPUs and
1120 Memory Nodes and assigning tasks to run only within those sets.
1121 This is primarily useful on large SMP or NUMA systems.
1125 config PROC_PID_CPUSET
1126 bool "Include legacy /proc/<pid>/cpuset file"
1130 config CGROUP_DEVICE
1131 bool "Device controller"
1133 Provides a cgroup controller implementing whitelists for
1134 devices which a process in the cgroup can mknod or open.
1136 config CGROUP_CPUACCT
1137 bool "Simple CPU accounting controller"
1139 Provides a simple controller for monitoring the
1140 total CPU consumed by the tasks in a cgroup.
1143 bool "Perf controller"
1144 depends on PERF_EVENTS
1146 This option extends the perf per-cpu mode to restrict monitoring
1147 to threads which belong to the cgroup specified and run on the
1148 designated cpu. Or this can be used to have cgroup ID in samples
1149 so that it can monitor performance events among cgroups.
1154 bool "Support for eBPF programs attached to cgroups"
1155 depends on BPF_SYSCALL
1156 select SOCK_CGROUP_DATA
1158 Allow attaching eBPF programs to a cgroup using the bpf(2)
1159 syscall command BPF_PROG_ATTACH.
1161 In which context these programs are accessed depends on the type
1162 of attachment. For instance, programs that are attached using
1163 BPF_CGROUP_INET_INGRESS will be executed on the ingress path of
1167 bool "Misc resource controller"
1170 Provides a controller for miscellaneous resources on a host.
1172 Miscellaneous scalar resources are the resources on the host system
1173 which cannot be abstracted like the other cgroups. This controller
1174 tracks and limits the miscellaneous resources used by a process
1175 attached to a cgroup hierarchy.
1177 For more information, please check misc cgroup section in
1178 /Documentation/admin-guide/cgroup-v2.rst.
1181 bool "Debug controller"
1183 depends on DEBUG_KERNEL
1185 This option enables a simple controller that exports
1186 debugging information about the cgroups framework. This
1187 controller is for control cgroup debugging only. Its
1188 interfaces are not stable.
1192 config SOCK_CGROUP_DATA
1198 menuconfig NAMESPACES
1199 bool "Namespaces support" if EXPERT
1200 depends on MULTIUSER
1203 Provides the way to make tasks work with different objects using
1204 the same id. For example same IPC id may refer to different objects
1205 or same user id or pid may refer to different tasks when used in
1206 different namespaces.
1211 bool "UTS namespace"
1214 In this namespace tasks see different info provided with the
1218 bool "TIME namespace"
1219 depends on GENERIC_VDSO_TIME_NS
1222 In this namespace boottime and monotonic clocks can be set.
1223 The time will keep going with the same pace.
1226 bool "IPC namespace"
1227 depends on (SYSVIPC || POSIX_MQUEUE)
1230 In this namespace tasks work with IPC ids which correspond to
1231 different IPC objects in different namespaces.
1234 bool "User namespace"
1237 This allows containers, i.e. vservers, to use user namespaces
1238 to provide different user info for different servers.
1240 When user namespaces are enabled in the kernel it is
1241 recommended that the MEMCG option also be enabled and that
1242 user-space use the memory control groups to limit the amount
1243 of memory a memory unprivileged users can use.
1248 bool "PID Namespaces"
1251 Support process id namespaces. This allows having multiple
1252 processes with the same pid as long as they are in different
1253 pid namespaces. This is a building block of containers.
1256 bool "Network namespace"
1260 Allow user space to create what appear to be multiple instances
1261 of the network stack.
1265 config CHECKPOINT_RESTORE
1266 bool "Checkpoint/restore support"
1267 select PROC_CHILDREN
1271 Enables additional kernel features in a sake of checkpoint/restore.
1272 In particular it adds auxiliary prctl codes to setup process text,
1273 data and heap segment sizes, and a few additional /proc filesystem
1276 If unsure, say N here.
1278 config SCHED_AUTOGROUP
1279 bool "Automatic process group scheduling"
1282 select FAIR_GROUP_SCHED
1284 This option optimizes the scheduler for common desktop workloads by
1285 automatically creating and populating task groups. This separation
1286 of workloads isolates aggressive CPU burners (like build jobs) from
1287 desktop applications. Task group autogeneration is currently based
1290 config SYSFS_DEPRECATED
1291 bool "Enable deprecated sysfs features to support old userspace tools"
1295 This option adds code that switches the layout of the "block" class
1296 devices, to not show up in /sys/class/block/, but only in
1299 This switch is only active when the sysfs.deprecated=1 boot option is
1300 passed or the SYSFS_DEPRECATED_V2 option is set.
1302 This option allows new kernels to run on old distributions and tools,
1303 which might get confused by /sys/class/block/. Since 2007/2008 all
1304 major distributions and tools handle this just fine.
1306 Recent distributions and userspace tools after 2009/2010 depend on
1307 the existence of /sys/class/block/, and will not work with this
1310 Only if you are using a new kernel on an old distribution, you might
1313 config SYSFS_DEPRECATED_V2
1314 bool "Enable deprecated sysfs features by default"
1317 depends on SYSFS_DEPRECATED
1319 Enable deprecated sysfs by default.
1321 See the CONFIG_SYSFS_DEPRECATED option for more details about this
1324 Only if you are using a new kernel on an old distribution, you might
1325 need to say Y here. Even then, odds are you would not need it
1326 enabled, you can always pass the boot option if absolutely necessary.
1329 bool "Kernel->user space relay support (formerly relayfs)"
1332 This option enables support for relay interface support in
1333 certain file systems (such as debugfs).
1334 It is designed to provide an efficient mechanism for tools and
1335 facilities to relay large amounts of data from kernel space to
1340 config BLK_DEV_INITRD
1341 bool "Initial RAM filesystem and RAM disk (initramfs/initrd) support"
1343 The initial RAM filesystem is a ramfs which is loaded by the
1344 boot loader (loadlin or lilo) and that is mounted as root
1345 before the normal boot procedure. It is typically used to
1346 load modules needed to mount the "real" root file system,
1347 etc. See <file:Documentation/admin-guide/initrd.rst> for details.
1349 If RAM disk support (BLK_DEV_RAM) is also included, this
1350 also enables initial RAM disk (initrd) support and adds
1351 15 Kbytes (more on some other architectures) to the kernel size.
1357 source "usr/Kconfig"
1362 bool "Boot config support"
1363 select BLK_DEV_INITRD
1365 Extra boot config allows system admin to pass a config file as
1366 complemental extension of kernel cmdline when booting.
1367 The boot config file must be attached at the end of initramfs
1368 with checksum, size and magic word.
1369 See <file:Documentation/admin-guide/bootconfig.rst> for details.
1374 prompt "Compiler optimization level"
1375 default CC_OPTIMIZE_FOR_PERFORMANCE
1377 config CC_OPTIMIZE_FOR_PERFORMANCE
1378 bool "Optimize for performance (-O2)"
1380 This is the default optimization level for the kernel, building
1381 with the "-O2" compiler flag for best performance and most
1382 helpful compile-time warnings.
1384 config CC_OPTIMIZE_FOR_PERFORMANCE_O3
1385 bool "Optimize more for performance (-O3)"
1388 Choosing this option will pass "-O3" to your compiler to optimize
1389 the kernel yet more for performance.
1391 config CC_OPTIMIZE_FOR_SIZE
1392 bool "Optimize for size (-Os)"
1394 Choosing this option will pass "-Os" to your compiler resulting
1395 in a smaller kernel.
1399 config HAVE_LD_DEAD_CODE_DATA_ELIMINATION
1402 This requires that the arch annotates or otherwise protects
1403 its external entry points from being discarded. Linker scripts
1404 must also merge .text.*, .data.*, and .bss.* correctly into
1405 output sections. Care must be taken not to pull in unrelated
1406 sections (e.g., '.text.init'). Typically '.' in section names
1407 is used to distinguish them from label names / C identifiers.
1409 config LD_DEAD_CODE_DATA_ELIMINATION
1410 bool "Dead code and data elimination (EXPERIMENTAL)"
1411 depends on HAVE_LD_DEAD_CODE_DATA_ELIMINATION
1413 depends on $(cc-option,-ffunction-sections -fdata-sections)
1414 depends on $(ld-option,--gc-sections)
1416 Enable this if you want to do dead code and data elimination with
1417 the linker by compiling with -ffunction-sections -fdata-sections,
1418 and linking with --gc-sections.
1420 This can reduce on disk and in-memory size of the kernel
1421 code and static data, particularly for small configs and
1422 on small systems. This has the possibility of introducing
1423 silently broken kernel if the required annotations are not
1424 present. This option is not well tested yet, so use at your
1427 config LD_ORPHAN_WARN
1429 depends on ARCH_WANT_LD_ORPHAN_WARN
1430 depends on !LD_IS_LLD || LLD_VERSION >= 110000
1431 depends on $(ld-option,--orphan-handling=warn)
1439 config SYSCTL_EXCEPTION_TRACE
1442 Enable support for /proc/sys/debug/exception-trace.
1444 config SYSCTL_ARCH_UNALIGN_NO_WARN
1447 Enable support for /proc/sys/kernel/ignore-unaligned-usertrap
1448 Allows arch to define/use @no_unaligned_warning to possibly warn
1449 about unaligned access emulation going on under the hood.
1451 config SYSCTL_ARCH_UNALIGN_ALLOW
1454 Enable support for /proc/sys/kernel/unaligned-trap
1455 Allows arches to define/use @unaligned_enabled to runtime toggle
1456 the unaligned access emulation.
1457 see arch/parisc/kernel/unaligned.c for reference
1459 config HAVE_PCSPKR_PLATFORM
1462 # interpreter that classic socket filters depend on
1467 bool "Configure standard kernel features (expert users)"
1468 # Unhide debug options, to make the on-by-default options visible
1471 This option allows certain base kernel options and settings
1472 to be disabled or tweaked. This is for specialized
1473 environments which can tolerate a "non-standard" kernel.
1474 Only use this if you really know what you are doing.
1477 bool "Enable 16-bit UID system calls" if EXPERT
1478 depends on HAVE_UID16 && MULTIUSER
1481 This enables the legacy 16-bit UID syscall wrappers.
1484 bool "Multiple users, groups and capabilities support" if EXPERT
1487 This option enables support for non-root users, groups and
1490 If you say N here, all processes will run with UID 0, GID 0, and all
1491 possible capabilities. Saying N here also compiles out support for
1492 system calls related to UIDs, GIDs, and capabilities, such as setuid,
1495 If unsure, say Y here.
1497 config SGETMASK_SYSCALL
1498 bool "sgetmask/ssetmask syscalls support" if EXPERT
1499 def_bool PARISC || M68K || PPC || MIPS || X86 || SPARC || MICROBLAZE || SUPERH
1501 sys_sgetmask and sys_ssetmask are obsolete system calls
1502 no longer supported in libc but still enabled by default in some
1505 If unsure, leave the default option here.
1507 config SYSFS_SYSCALL
1508 bool "Sysfs syscall support" if EXPERT
1511 sys_sysfs is an obsolete system call no longer supported in libc.
1512 Note that disabling this option is more secure but might break
1513 compatibility with some systems.
1515 If unsure say Y here.
1518 bool "open by fhandle syscalls" if EXPERT
1522 If you say Y here, a user level program will be able to map
1523 file names to handle and then later use the handle for
1524 different file system operations. This is useful in implementing
1525 userspace file servers, which now track files using handles instead
1526 of names. The handle would remain the same even if file names
1527 get renamed. Enables open_by_handle_at(2) and name_to_handle_at(2)
1531 bool "Posix Clocks & timers" if EXPERT
1534 This includes native support for POSIX timers to the kernel.
1535 Some embedded systems have no use for them and therefore they
1536 can be configured out to reduce the size of the kernel image.
1538 When this option is disabled, the following syscalls won't be
1539 available: timer_create, timer_gettime: timer_getoverrun,
1540 timer_settime, timer_delete, clock_adjtime, getitimer,
1541 setitimer, alarm. Furthermore, the clock_settime, clock_gettime,
1542 clock_getres and clock_nanosleep syscalls will be limited to
1543 CLOCK_REALTIME, CLOCK_MONOTONIC and CLOCK_BOOTTIME only.
1549 bool "Enable support for printk" if EXPERT
1552 This option enables normal printk support. Removing it
1553 eliminates most of the message strings from the kernel image
1554 and makes the kernel more or less silent. As this makes it
1555 very difficult to diagnose system problems, saying N here is
1556 strongly discouraged.
1559 bool "BUG() support" if EXPERT
1562 Disabling this option eliminates support for BUG and WARN, reducing
1563 the size of your kernel image and potentially quietly ignoring
1564 numerous fatal conditions. You should only consider disabling this
1565 option for embedded systems with no facilities for reporting errors.
1571 bool "Enable ELF core dumps" if EXPERT
1573 Enable support for generating core dumps. Disabling saves about 4k.
1576 config PCSPKR_PLATFORM
1577 bool "Enable PC-Speaker support" if EXPERT
1578 depends on HAVE_PCSPKR_PLATFORM
1582 This option allows to disable the internal PC-Speaker
1583 support, saving some memory.
1587 bool "Enable full-sized data structures for core" if EXPERT
1589 Disabling this option reduces the size of miscellaneous core
1590 kernel data structures. This saves memory on small machines,
1591 but may reduce performance.
1594 bool "Enable futex support" if EXPERT
1598 Disabling this option will cause the kernel to be built without
1599 support for "fast userspace mutexes". The resulting kernel may not
1600 run glibc-based applications correctly.
1604 depends on FUTEX && RT_MUTEXES
1607 config HAVE_FUTEX_CMPXCHG
1611 Architectures should select this if futex_atomic_cmpxchg_inatomic()
1612 is implemented and always working. This removes a couple of runtime
1616 bool "Enable eventpoll support" if EXPERT
1619 Disabling this option will cause the kernel to be built without
1620 support for epoll family of system calls.
1623 bool "Enable signalfd() system call" if EXPERT
1626 Enable the signalfd() system call that allows to receive signals
1627 on a file descriptor.
1632 bool "Enable timerfd() system call" if EXPERT
1635 Enable the timerfd() system call that allows to receive timer
1636 events on a file descriptor.
1641 bool "Enable eventfd() system call" if EXPERT
1644 Enable the eventfd() system call that allows to receive both
1645 kernel notification (ie. KAIO) or userspace notifications.
1650 bool "Use full shmem filesystem" if EXPERT
1654 The shmem is an internal filesystem used to manage shared memory.
1655 It is backed by swap and manages resource limits. It is also exported
1656 to userspace as tmpfs if TMPFS is enabled. Disabling this
1657 option replaces shmem and tmpfs with the much simpler ramfs code,
1658 which may be appropriate on small systems without swap.
1661 bool "Enable AIO support" if EXPERT
1664 This option enables POSIX asynchronous I/O which may by used
1665 by some high performance threaded applications. Disabling
1666 this option saves about 7k.
1669 bool "Enable IO uring support" if EXPERT
1673 This option enables support for the io_uring interface, enabling
1674 applications to submit and complete IO through submission and
1675 completion rings that are shared between the kernel and application.
1677 config ADVISE_SYSCALLS
1678 bool "Enable madvise/fadvise syscalls" if EXPERT
1681 This option enables the madvise and fadvise syscalls, used by
1682 applications to advise the kernel about their future memory or file
1683 usage, improving performance. If building an embedded system where no
1684 applications use these syscalls, you can disable this option to save
1687 config HAVE_ARCH_USERFAULTFD_WP
1690 Arch has userfaultfd write protection support
1692 config HAVE_ARCH_USERFAULTFD_MINOR
1695 Arch has userfaultfd minor fault support
1698 bool "Enable membarrier() system call" if EXPERT
1701 Enable the membarrier() system call that allows issuing memory
1702 barriers across all running threads, which can be used to distribute
1703 the cost of user-space memory barriers asymmetrically by transforming
1704 pairs of memory barriers into pairs consisting of membarrier() and a
1710 bool "Load all symbols for debugging/ksymoops" if EXPERT
1713 Say Y here to let the kernel print out symbolic crash information and
1714 symbolic stack backtraces. This increases the size of the kernel
1715 somewhat, as all symbols have to be loaded into the kernel image.
1718 bool "Include all symbols in kallsyms"
1719 depends on DEBUG_KERNEL && KALLSYMS
1721 Normally kallsyms only contains the symbols of functions for nicer
1722 OOPS messages and backtraces (i.e., symbols from the text and inittext
1723 sections). This is sufficient for most cases. And only in very rare
1724 cases (e.g., when a debugger is used) all symbols are required (e.g.,
1725 names of variables from the data sections, etc).
1727 This option makes sure that all symbols are loaded into the kernel
1728 image (i.e., symbols from all sections) in cost of increased kernel
1729 size (depending on the kernel configuration, it may be 300KiB or
1730 something like this).
1732 Say N unless you really need all symbols.
1734 config KALLSYMS_ABSOLUTE_PERCPU
1737 default X86_64 && SMP
1739 config KALLSYMS_BASE_RELATIVE
1744 Instead of emitting them as absolute values in the native word size,
1745 emit the symbol references in the kallsyms table as 32-bit entries,
1746 each containing a relative value in the range [base, base + U32_MAX]
1747 or, when KALLSYMS_ABSOLUTE_PERCPU is in effect, each containing either
1748 an absolute value in the range [0, S32_MAX] or a relative value in the
1749 range [base, base + S32_MAX], where base is the lowest relative symbol
1750 address encountered in the image.
1752 On 64-bit builds, this reduces the size of the address table by 50%,
1753 but more importantly, it results in entries whose values are build
1754 time constants, and no relocation pass is required at runtime to fix
1755 up the entries based on the runtime load address of the kernel.
1757 # end of the "standard kernel features (expert users)" menu
1759 # syscall, maps, verifier
1762 bool "Enable userfaultfd() system call"
1765 Enable the userfaultfd() system call that allows to intercept and
1766 handle page faults in userland.
1768 config ARCH_HAS_MEMBARRIER_CALLBACKS
1771 config ARCH_HAS_MEMBARRIER_SYNC_CORE
1775 bool "Enable kcmp() system call" if EXPERT
1777 Enable the kernel resource comparison system call. It provides
1778 user-space with the ability to compare two processes to see if they
1779 share a common resource, such as a file descriptor or even virtual
1785 bool "Enable rseq() system call" if EXPERT
1787 depends on HAVE_RSEQ
1790 Enable the restartable sequences system call. It provides a
1791 user-space cache for the current CPU number value, which
1792 speeds up getting the current CPU number from user-space,
1793 as well as an ABI to speed up user-space operations on
1800 bool "Enabled debugging of rseq() system call" if EXPERT
1801 depends on RSEQ && DEBUG_KERNEL
1803 Enable extra debugging checks for the rseq system call.
1808 bool "Embedded system"
1811 This option should be enabled if compiling the kernel for
1812 an embedded system so certain expert options are available
1815 config HAVE_PERF_EVENTS
1818 See tools/perf/design.txt for details.
1820 config PERF_USE_VMALLOC
1823 See tools/perf/design.txt for details
1826 bool "PC/104 support" if EXPERT
1828 Expose PC/104 form factor device drivers and options available for
1829 selection and configuration. Enable this option if your target
1830 machine has a PC/104 bus.
1832 menu "Kernel Performance Events And Counters"
1835 bool "Kernel performance events and counters"
1836 default y if PROFILING
1837 depends on HAVE_PERF_EVENTS
1841 Enable kernel support for various performance events provided
1842 by software and hardware.
1844 Software events are supported either built-in or via the
1845 use of generic tracepoints.
1847 Most modern CPUs support performance events via performance
1848 counter registers. These registers count the number of certain
1849 types of hw events: such as instructions executed, cachemisses
1850 suffered, or branches mis-predicted - without slowing down the
1851 kernel or applications. These registers can also trigger interrupts
1852 when a threshold number of events have passed - and can thus be
1853 used to profile the code that runs on that CPU.
1855 The Linux Performance Event subsystem provides an abstraction of
1856 these software and hardware event capabilities, available via a
1857 system call and used by the "perf" utility in tools/perf/. It
1858 provides per task and per CPU counters, and it provides event
1859 capabilities on top of those.
1863 config DEBUG_PERF_USE_VMALLOC
1865 bool "Debug: use vmalloc to back perf mmap() buffers"
1866 depends on PERF_EVENTS && DEBUG_KERNEL && !PPC
1867 select PERF_USE_VMALLOC
1869 Use vmalloc memory to back perf mmap() buffers.
1871 Mostly useful for debugging the vmalloc code on platforms
1872 that don't require it.
1878 config VM_EVENT_COUNTERS
1880 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1882 VM event counters are needed for event counts to be shown.
1883 This option allows the disabling of the VM event counters
1884 on EXPERT systems. /proc/vmstat will only show page counts
1885 if VM event counters are disabled.
1889 bool "Enable SLUB debugging support" if EXPERT
1890 depends on SLUB && SYSFS
1892 SLUB has extensive debug support features. Disabling these can
1893 result in significant savings in code size. This also disables
1894 SLUB sysfs support. /sys/slab will not exist and there will be
1895 no support for cache validation etc.
1898 bool "Disable heap randomization"
1901 Randomizing heap placement makes heap exploits harder, but it
1902 also breaks ancient binaries (including anything libc5 based).
1903 This option changes the bootup default to heap randomization
1904 disabled, and can be overridden at runtime by setting
1905 /proc/sys/kernel/randomize_va_space to 2.
1907 On non-ancient distros (post-2000 ones) N is usually a safe choice.
1910 prompt "Choose SLAB allocator"
1913 This option allows to select a slab allocator.
1917 select HAVE_HARDENED_USERCOPY_ALLOCATOR
1919 The regular slab allocator that is established and known to work
1920 well in all environments. It organizes cache hot objects in
1921 per cpu and per node queues.
1924 bool "SLUB (Unqueued Allocator)"
1925 select HAVE_HARDENED_USERCOPY_ALLOCATOR
1927 SLUB is a slab allocator that minimizes cache line usage
1928 instead of managing queues of cached objects (SLAB approach).
1929 Per cpu caching is realized using slabs of objects instead
1930 of queues of objects. SLUB can use memory efficiently
1931 and has enhanced diagnostics. SLUB is the default choice for
1936 bool "SLOB (Simple Allocator)"
1938 SLOB replaces the stock allocator with a drastically simpler
1939 allocator. SLOB is generally more space efficient but
1940 does not perform as well on large systems.
1944 config SLAB_MERGE_DEFAULT
1945 bool "Allow slab caches to be merged"
1948 For reduced kernel memory fragmentation, slab caches can be
1949 merged when they share the same size and other characteristics.
1950 This carries a risk of kernel heap overflows being able to
1951 overwrite objects from merged caches (and more easily control
1952 cache layout), which makes such heap attacks easier to exploit
1953 by attackers. By keeping caches unmerged, these kinds of exploits
1954 can usually only damage objects in the same cache. To disable
1955 merging at runtime, "slab_nomerge" can be passed on the kernel
1958 config SLAB_FREELIST_RANDOM
1959 bool "Randomize slab freelist"
1960 depends on SLAB || SLUB
1962 Randomizes the freelist order used on creating new pages. This
1963 security feature reduces the predictability of the kernel slab
1964 allocator against heap overflows.
1966 config SLAB_FREELIST_HARDENED
1967 bool "Harden slab freelist metadata"
1968 depends on SLAB || SLUB
1970 Many kernel heap attacks try to target slab cache metadata and
1971 other infrastructure. This options makes minor performance
1972 sacrifices to harden the kernel slab allocator against common
1973 freelist exploit methods. Some slab implementations have more
1974 sanity-checking than others. This option is most effective with
1977 config SHUFFLE_PAGE_ALLOCATOR
1978 bool "Page allocator randomization"
1979 default SLAB_FREELIST_RANDOM && ACPI_NUMA
1981 Randomization of the page allocator improves the average
1982 utilization of a direct-mapped memory-side-cache. See section
1983 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
1984 6.2a specification for an example of how a platform advertises
1985 the presence of a memory-side-cache. There are also incidental
1986 security benefits as it reduces the predictability of page
1987 allocations to compliment SLAB_FREELIST_RANDOM, but the
1988 default granularity of shuffling on the "MAX_ORDER - 1" i.e,
1989 10th order of pages is selected based on cache utilization
1992 While the randomization improves cache utilization it may
1993 negatively impact workloads on platforms without a cache. For
1994 this reason, by default, the randomization is enabled only
1995 after runtime detection of a direct-mapped memory-side-cache.
1996 Otherwise, the randomization may be force enabled with the
1997 'page_alloc.shuffle' kernel command line parameter.
2001 config SLUB_CPU_PARTIAL
2003 depends on SLUB && SMP
2004 bool "SLUB per cpu partial cache"
2006 Per cpu partial caches accelerate objects allocation and freeing
2007 that is local to a processor at the price of more indeterminism
2008 in the latency of the free. On overflow these caches will be cleared
2009 which requires the taking of locks that may cause latency spikes.
2010 Typically one would choose no for a realtime system.
2012 config MMAP_ALLOW_UNINITIALIZED
2013 bool "Allow mmapped anonymous memory to be uninitialized"
2014 depends on EXPERT && !MMU
2017 Normally, and according to the Linux spec, anonymous memory obtained
2018 from mmap() has its contents cleared before it is passed to
2019 userspace. Enabling this config option allows you to request that
2020 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
2021 providing a huge performance boost. If this option is not enabled,
2022 then the flag will be ignored.
2024 This is taken advantage of by uClibc's malloc(), and also by
2025 ELF-FDPIC binfmt's brk and stack allocator.
2027 Because of the obvious security issues, this option should only be
2028 enabled on embedded devices where you control what is run in
2029 userspace. Since that isn't generally a problem on no-MMU systems,
2030 it is normally safe to say Y here.
2032 See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
2034 config SYSTEM_DATA_VERIFICATION
2036 select SYSTEM_TRUSTED_KEYRING
2040 select ASYMMETRIC_KEY_TYPE
2041 select ASYMMETRIC_PUBLIC_KEY_SUBTYPE
2044 select X509_CERTIFICATE_PARSER
2045 select PKCS7_MESSAGE_PARSER
2047 Provide PKCS#7 message verification using the contents of the system
2048 trusted keyring to provide public keys. This then can be used for
2049 module verification, kexec image verification and firmware blob
2053 bool "Profiling support"
2055 Say Y here to enable the extended profiling support mechanisms used
2059 # Place an empty function call at each tracepoint site. Can be
2060 # dynamically changed for a probe function.
2065 endmenu # General setup
2067 source "arch/Kconfig"
2074 default 0 if BASE_FULL
2075 default 1 if !BASE_FULL
2077 config MODULE_SIG_FORMAT
2079 select SYSTEM_DATA_VERIFICATION
2082 bool "Enable loadable module support"
2085 Kernel modules are small pieces of compiled code which can
2086 be inserted in the running kernel, rather than being
2087 permanently built into the kernel. You use the "modprobe"
2088 tool to add (and sometimes remove) them. If you say Y here,
2089 many parts of the kernel can be built as modules (by
2090 answering M instead of Y where indicated): this is most
2091 useful for infrequently used options which are not required
2092 for booting. For more information, see the man pages for
2093 modprobe, lsmod, modinfo, insmod and rmmod.
2095 If you say Y here, you will need to run "make
2096 modules_install" to put the modules under /lib/modules/
2097 where modprobe can find them (you may need to be root to do
2104 config MODULE_FORCE_LOAD
2105 bool "Forced module loading"
2108 Allow loading of modules without version information (ie. modprobe
2109 --force). Forced module loading sets the 'F' (forced) taint flag and
2110 is usually a really bad idea.
2112 config MODULE_UNLOAD
2113 bool "Module unloading"
2115 Without this option you will not be able to unload any
2116 modules (note that some modules may not be unloadable
2117 anyway), which makes your kernel smaller, faster
2118 and simpler. If unsure, say Y.
2120 config MODULE_FORCE_UNLOAD
2121 bool "Forced module unloading"
2122 depends on MODULE_UNLOAD
2124 This option allows you to force a module to unload, even if the
2125 kernel believes it is unsafe: the kernel will remove the module
2126 without waiting for anyone to stop using it (using the -f option to
2127 rmmod). This is mainly for kernel developers and desperate users.
2131 bool "Module versioning support"
2133 Usually, you have to use modules compiled with your kernel.
2134 Saying Y here makes it sometimes possible to use modules
2135 compiled for different kernels, by adding enough information
2136 to the modules to (hopefully) spot any changes which would
2137 make them incompatible with the kernel you are running. If
2140 config ASM_MODVERSIONS
2142 default HAVE_ASM_MODVERSIONS && MODVERSIONS
2144 This enables module versioning for exported symbols also from
2145 assembly. This can be enabled only when the target architecture
2148 config MODULE_REL_CRCS
2150 depends on MODVERSIONS
2152 config MODULE_SRCVERSION_ALL
2153 bool "Source checksum for all modules"
2155 Modules which contain a MODULE_VERSION get an extra "srcversion"
2156 field inserted into their modinfo section, which contains a
2157 sum of the source files which made it. This helps maintainers
2158 see exactly which source was used to build a module (since
2159 others sometimes change the module source without updating
2160 the version). With this option, such a "srcversion" field
2161 will be created for all modules. If unsure, say N.
2164 bool "Module signature verification"
2165 select MODULE_SIG_FORMAT
2167 Check modules for valid signatures upon load: the signature
2168 is simply appended to the module. For more information see
2169 <file:Documentation/admin-guide/module-signing.rst>.
2171 Note that this option adds the OpenSSL development packages as a
2172 kernel build dependency so that the signing tool can use its crypto
2175 You should enable this option if you wish to use either
2176 CONFIG_SECURITY_LOCKDOWN_LSM or lockdown functionality imposed via
2177 another LSM - otherwise unsigned modules will be loadable regardless
2178 of the lockdown policy.
2180 !!!WARNING!!! If you enable this option, you MUST make sure that the
2181 module DOES NOT get stripped after being signed. This includes the
2182 debuginfo strip done by some packagers (such as rpmbuild) and
2183 inclusion into an initramfs that wants the module size reduced.
2185 config MODULE_SIG_FORCE
2186 bool "Require modules to be validly signed"
2187 depends on MODULE_SIG
2189 Reject unsigned modules or signed modules for which we don't have a
2190 key. Without this, such modules will simply taint the kernel.
2192 config MODULE_SIG_ALL
2193 bool "Automatically sign all modules"
2195 depends on MODULE_SIG || IMA_APPRAISE_MODSIG
2197 Sign all modules during make modules_install. Without this option,
2198 modules must be signed manually, using the scripts/sign-file tool.
2200 comment "Do not forget to sign required modules with scripts/sign-file"
2201 depends on MODULE_SIG_FORCE && !MODULE_SIG_ALL
2204 prompt "Which hash algorithm should modules be signed with?"
2205 depends on MODULE_SIG || IMA_APPRAISE_MODSIG
2207 This determines which sort of hashing algorithm will be used during
2208 signature generation. This algorithm _must_ be built into the kernel
2209 directly so that signature verification can take place. It is not
2210 possible to load a signed module containing the algorithm to check
2211 the signature on that module.
2213 config MODULE_SIG_SHA1
2214 bool "Sign modules with SHA-1"
2217 config MODULE_SIG_SHA224
2218 bool "Sign modules with SHA-224"
2219 select CRYPTO_SHA256
2221 config MODULE_SIG_SHA256
2222 bool "Sign modules with SHA-256"
2223 select CRYPTO_SHA256
2225 config MODULE_SIG_SHA384
2226 bool "Sign modules with SHA-384"
2227 select CRYPTO_SHA512
2229 config MODULE_SIG_SHA512
2230 bool "Sign modules with SHA-512"
2231 select CRYPTO_SHA512
2235 config MODULE_SIG_HASH
2237 depends on MODULE_SIG || IMA_APPRAISE_MODSIG
2238 default "sha1" if MODULE_SIG_SHA1
2239 default "sha224" if MODULE_SIG_SHA224
2240 default "sha256" if MODULE_SIG_SHA256
2241 default "sha384" if MODULE_SIG_SHA384
2242 default "sha512" if MODULE_SIG_SHA512
2245 prompt "Module compression mode"
2247 This option allows you to choose the algorithm which will be used to
2248 compress modules when 'make modules_install' is run. (or, you can
2249 choose to not compress modules at all.)
2251 External modules will also be compressed in the same way during the
2254 For modules inside an initrd or initramfs, it's more efficient to
2255 compress the whole initrd or initramfs instead.
2257 This is fully compatible with signed modules.
2259 Please note that the tool used to load modules needs to support the
2260 corresponding algorithm. module-init-tools MAY support gzip, and kmod
2261 MAY support gzip, xz and zstd.
2263 Your build system needs to provide the appropriate compression tool
2264 to compress the modules.
2266 If in doubt, select 'None'.
2268 config MODULE_COMPRESS_NONE
2271 Do not compress modules. The installed modules are suffixed
2274 config MODULE_COMPRESS_GZIP
2277 Compress modules with GZIP. The installed modules are suffixed
2280 config MODULE_COMPRESS_XZ
2283 Compress modules with XZ. The installed modules are suffixed
2286 config MODULE_COMPRESS_ZSTD
2289 Compress modules with ZSTD. The installed modules are suffixed
2294 config MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS
2295 bool "Allow loading of modules with missing namespace imports"
2297 Symbols exported with EXPORT_SYMBOL_NS*() are considered exported in
2298 a namespace. A module that makes use of a symbol exported with such a
2299 namespace is required to import the namespace via MODULE_IMPORT_NS().
2300 There is no technical reason to enforce correct namespace imports,
2301 but it creates consistency between symbols defining namespaces and
2302 users importing namespaces they make use of. This option relaxes this
2303 requirement and lifts the enforcement when loading a module.
2307 config MODPROBE_PATH
2308 string "Path to modprobe binary"
2309 default "/sbin/modprobe"
2311 When kernel code requests a module, it does so by calling
2312 the "modprobe" userspace utility. This option allows you to
2313 set the path where that binary is found. This can be changed
2314 at runtime via the sysctl file
2315 /proc/sys/kernel/modprobe. Setting this to the empty string
2316 removes the kernel's ability to request modules (but
2317 userspace can still load modules explicitly).
2319 config TRIM_UNUSED_KSYMS
2320 bool "Trim unused exported kernel symbols" if EXPERT
2321 depends on !COMPILE_TEST
2323 The kernel and some modules make many symbols available for
2324 other modules to use via EXPORT_SYMBOL() and variants. Depending
2325 on the set of modules being selected in your kernel configuration,
2326 many of those exported symbols might never be used.
2328 This option allows for unused exported symbols to be dropped from
2329 the build. In turn, this provides the compiler more opportunities
2330 (especially when using LTO) for optimizing the code and reducing
2331 binary size. This might have some security advantages as well.
2333 If unsure, or if you need to build out-of-tree modules, say N.
2335 config UNUSED_KSYMS_WHITELIST
2336 string "Whitelist of symbols to keep in ksymtab"
2337 depends on TRIM_UNUSED_KSYMS
2339 By default, all unused exported symbols will be un-exported from the
2340 build when TRIM_UNUSED_KSYMS is selected.
2342 UNUSED_KSYMS_WHITELIST allows to whitelist symbols that must be kept
2343 exported at all times, even in absence of in-tree users. The value to
2344 set here is the path to a text file containing the list of symbols,
2345 one per line. The path can be absolute, or relative to the kernel
2350 config MODULES_TREE_LOOKUP
2352 depends on PERF_EVENTS || TRACING || CFI_CLANG
2354 config INIT_ALL_POSSIBLE
2357 Back when each arch used to define their own cpu_online_mask and
2358 cpu_possible_mask, some of them chose to initialize cpu_possible_mask
2359 with all 1s, and others with all 0s. When they were centralised,
2360 it was better to provide this option than to break all the archs
2361 and have several arch maintainers pursuing me down dark alleys.
2363 source "block/Kconfig"
2365 config PREEMPT_NOTIFIERS
2375 Build a simple ASN.1 grammar compiler that produces a bytecode output
2376 that can be interpreted by the ASN.1 stream decoder and used to
2377 inform it as to what tags are to be expected in a stream and what
2378 functions to call on what tags.
2380 source "kernel/Kconfig.locks"
2382 config ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
2385 config ARCH_HAS_SYNC_CORE_BEFORE_USERMODE
2388 # It may be useful for an architecture to override the definitions of the
2389 # SYSCALL_DEFINE() and __SYSCALL_DEFINEx() macros in <linux/syscalls.h>
2390 # and the COMPAT_ variants in <linux/compat.h>, in particular to use a
2391 # different calling convention for syscalls. They can also override the
2392 # macros for not-implemented syscalls in kernel/sys_ni.c and
2393 # kernel/time/posix-stubs.c. All these overrides need to be available in
2394 # <asm/syscall_wrapper.h>.
2395 config ARCH_HAS_SYSCALL_WRAPPER
projects / mirror_ubuntu-jammy-kernel.git / blob