1 # SPDX-License-Identifier: GPL-2.0-only
6 default "/lib/modules/$(shell,uname -r)/.config"
7 default "/etc/kernel-config"
8 default "/boot/config-$(shell,uname -r)"
10 default "arch/$(ARCH)/defconfig"
13 def_bool $(success,$(CC) --version | head -n 1 | grep -q gcc)
17 default $(shell,$(srctree)/scripts/gcc-version.sh $(CC)) if CC_IS_GCC
21 def_bool $(success,$(CC) --version | head -n 1 | grep -q clang)
25 default $(shell,$(srctree)/scripts/clang-version.sh $(CC))
28 def_bool $(success,$(srctree)/scripts/cc-can-link.sh $(CC))
30 config CC_HAS_ASM_GOTO
31 def_bool $(success,$(srctree)/scripts/gcc-goto.sh $(CC))
33 config TOOLS_SUPPORT_RELR
34 def_bool $(success,env "CC=$(CC)" "LD=$(LD)" "NM=$(NM)" "OBJCOPY=$(OBJCOPY)" $(srctree)/scripts/tools-support-relr.sh)
36 config CC_HAS_ASM_INLINE
37 def_bool $(success,echo 'void foo(void) { asm inline (""); }' | $(CC) -x c - -c -o /dev/null)
39 config CC_HAS_WARN_MAYBE_UNINITIALIZED
40 def_bool $(cc-option,-Wmaybe-uninitialized)
42 GCC >= 4.7 supports this option.
44 config CC_DISABLE_WARN_MAYBE_UNINITIALIZED
46 depends on CC_HAS_WARN_MAYBE_UNINITIALIZED
47 default CC_IS_GCC && GCC_VERSION < 40900 # unreliable for GCC < 4.9
49 GCC's -Wmaybe-uninitialized is not reliable by definition.
50 Lots of false positive warnings are produced in some cases.
52 If this option is enabled, -Wno-maybe-uninitialzed is passed
53 to the compiler to suppress maybe-uninitialized warnings.
61 config BUILDTIME_EXTABLE_SORT
64 config THREAD_INFO_IN_TASK
67 Select this to move thread_info off the stack into task_struct. To
68 make this work, an arch will need to remove all thread_info fields
69 except flags and fix any runtime bugs.
71 One subtle change that will be needed is to use try_get_task_stack()
72 and put_task_stack() in save_thread_stack_tsk() and get_wchan().
81 depends on BROKEN || !SMP
84 config INIT_ENV_ARG_LIMIT
89 Maximum of each of the number of arguments and environment
90 variables passed to init from the kernel command line.
93 bool "Compile also drivers which will not load"
97 Some drivers can be compiled on a different platform than they are
98 intended to be run on. Despite they cannot be loaded there (or even
99 when they load they cannot be used due to missing HW support),
100 developers still, opposing to distributors, might want to build such
101 drivers to compile-test them.
103 If you are a developer and want to build everything available, say Y
104 here. If you are a user/distributor, say N here to exclude useless
105 drivers to be distributed.
108 bool "Compile test headers that should be standalone compilable"
110 Compile test headers listed in header-test-y target to ensure they are
111 self-contained, i.e. compilable as standalone units.
113 If you are a developer or tester and want to ensure the requested
114 headers are self-contained, say Y here. Otherwise, choose N.
116 config KERNEL_HEADER_TEST
117 bool "Compile test kernel headers"
118 depends on HEADER_TEST
120 Headers in include/ are used to build external moduls.
121 Compile test them to ensure they are self-contained, i.e.
122 compilable as standalone units.
124 If you are a developer or tester and want to ensure the headers
125 in include/ are self-contained, say Y here. Otherwise, choose N.
127 config UAPI_HEADER_TEST
128 bool "Compile test UAPI headers"
129 depends on HEADER_TEST && HEADERS_INSTALL && CC_CAN_LINK
131 Compile test headers exported to user-space to ensure they are
132 self-contained, i.e. compilable as standalone units.
134 If you are a developer or tester and want to ensure the exported
135 headers are self-contained, say Y here. Otherwise, choose N.
138 string "Local version - append to kernel release"
140 Append an extra string to the end of your kernel version.
141 This will show up when you type uname, for example.
142 The string you set here will be appended after the contents of
143 any files with a filename matching localversion* in your
144 object and source tree, in that order. Your total string can
145 be a maximum of 64 characters.
147 config LOCALVERSION_AUTO
148 bool "Automatically append version information to the version string"
150 depends on !COMPILE_TEST
152 This will try to automatically determine if the current tree is a
153 release tree by looking for git tags that belong to the current
154 top of tree revision.
156 A string of the format -gxxxxxxxx will be added to the localversion
157 if a git-based tree is found. The string generated by this will be
158 appended after any matching localversion* files, and after the value
159 set in CONFIG_LOCALVERSION.
161 (The actual string used here is the first eight characters produced
162 by running the command:
164 $ git rev-parse --verify HEAD
166 which is done within the script "scripts/setlocalversion".)
169 string "Build ID Salt"
172 The build ID is used to link binaries and their debug info. Setting
173 this option will use the value in the calculation of the build id.
174 This is mostly useful for distributions which want to ensure the
175 build is unique between builds. It's safe to leave the default.
177 config HAVE_KERNEL_GZIP
180 config HAVE_KERNEL_BZIP2
183 config HAVE_KERNEL_LZMA
186 config HAVE_KERNEL_XZ
189 config HAVE_KERNEL_LZO
192 config HAVE_KERNEL_LZ4
195 config HAVE_KERNEL_UNCOMPRESSED
199 prompt "Kernel compression mode"
201 depends on HAVE_KERNEL_GZIP || HAVE_KERNEL_BZIP2 || HAVE_KERNEL_LZMA || HAVE_KERNEL_XZ || HAVE_KERNEL_LZO || HAVE_KERNEL_LZ4 || HAVE_KERNEL_UNCOMPRESSED
203 The linux kernel is a kind of self-extracting executable.
204 Several compression algorithms are available, which differ
205 in efficiency, compression and decompression speed.
206 Compression speed is only relevant when building a kernel.
207 Decompression speed is relevant at each boot.
209 If you have any problems with bzip2 or lzma compressed
210 kernels, mail me (Alain Knaff) <alain@knaff.lu>. (An older
211 version of this functionality (bzip2 only), for 2.4, was
212 supplied by Christian Ludwig)
214 High compression options are mostly useful for users, who
215 are low on disk space (embedded systems), but for whom ram
218 If in doubt, select 'gzip'
222 depends on HAVE_KERNEL_GZIP
224 The old and tried gzip compression. It provides a good balance
225 between compression ratio and decompression speed.
229 depends on HAVE_KERNEL_BZIP2
231 Its compression ratio and speed is intermediate.
232 Decompression speed is slowest among the choices. The kernel
233 size is about 10% smaller with bzip2, in comparison to gzip.
234 Bzip2 uses a large amount of memory. For modern kernels you
235 will need at least 8MB RAM or more for booting.
239 depends on HAVE_KERNEL_LZMA
241 This compression algorithm's ratio is best. Decompression speed
242 is between gzip and bzip2. Compression is slowest.
243 The kernel size is about 33% smaller with LZMA in comparison to gzip.
247 depends on HAVE_KERNEL_XZ
249 XZ uses the LZMA2 algorithm and instruction set specific
250 BCJ filters which can improve compression ratio of executable
251 code. The size of the kernel is about 30% smaller with XZ in
252 comparison to gzip. On architectures for which there is a BCJ
253 filter (i386, x86_64, ARM, IA-64, PowerPC, and SPARC), XZ
254 will create a few percent smaller kernel than plain LZMA.
256 The speed is about the same as with LZMA: The decompression
257 speed of XZ is better than that of bzip2 but worse than gzip
258 and LZO. Compression is slow.
262 depends on HAVE_KERNEL_LZO
264 Its compression ratio is the poorest among the choices. The kernel
265 size is about 10% bigger than gzip; however its speed
266 (both compression and decompression) is the fastest.
270 depends on HAVE_KERNEL_LZ4
272 LZ4 is an LZ77-type compressor with a fixed, byte-oriented encoding.
273 A preliminary version of LZ4 de/compression tool is available at
274 <https://code.google.com/p/lz4/>.
276 Its compression ratio is worse than LZO. The size of the kernel
277 is about 8% bigger than LZO. But the decompression speed is
280 config KERNEL_UNCOMPRESSED
282 depends on HAVE_KERNEL_UNCOMPRESSED
284 Produce uncompressed kernel image. This option is usually not what
285 you want. It is useful for debugging the kernel in slow simulation
286 environments, where decompressing and moving the kernel is awfully
287 slow. This option allows early boot code to skip the decompressor
288 and jump right at uncompressed kernel image.
292 config DEFAULT_HOSTNAME
293 string "Default hostname"
296 This option determines the default system hostname before userspace
297 calls sethostname(2). The kernel traditionally uses "(none)" here,
298 but you may wish to use a different default here to make a minimal
299 system more usable with less configuration.
302 # For some reason microblaze and nios2 hard code SWAP=n. Hopefully we can
303 # add proper SWAP support to them, in which case this can be remove.
309 bool "Support for paging of anonymous memory (swap)"
310 depends on MMU && BLOCK && !ARCH_NO_SWAP
313 This option allows you to choose whether you want to have support
314 for so called swap devices or swap files in your kernel that are
315 used to provide more virtual memory than the actual RAM present
316 in your computer. If unsure say Y.
321 Inter Process Communication is a suite of library functions and
322 system calls which let processes (running programs) synchronize and
323 exchange information. It is generally considered to be a good thing,
324 and some programs won't run unless you say Y here. In particular, if
325 you want to run the DOS emulator dosemu under Linux (read the
326 DOSEMU-HOWTO, available from <http://www.tldp.org/docs.html#howto>),
327 you'll need to say Y here.
329 You can find documentation about IPC with "info ipc" and also in
330 section 6.4 of the Linux Programmer's Guide, available from
331 <http://www.tldp.org/guides.html>.
333 config SYSVIPC_SYSCTL
340 bool "POSIX Message Queues"
343 POSIX variant of message queues is a part of IPC. In POSIX message
344 queues every message has a priority which decides about succession
345 of receiving it by a process. If you want to compile and run
346 programs written e.g. for Solaris with use of its POSIX message
347 queues (functions mq_*) say Y here.
349 POSIX message queues are visible as a filesystem called 'mqueue'
350 and can be mounted somewhere if you want to do filesystem
351 operations on message queues.
355 config POSIX_MQUEUE_SYSCTL
357 depends on POSIX_MQUEUE
361 config CROSS_MEMORY_ATTACH
362 bool "Enable process_vm_readv/writev syscalls"
366 Enabling this option adds the system calls process_vm_readv and
367 process_vm_writev which allow a process with the correct privileges
368 to directly read from or write to another process' address space.
369 See the man page for more details.
372 bool "uselib syscall"
373 def_bool ALPHA || M68K || SPARC || X86_32 || IA32_EMULATION
375 This option enables the uselib syscall, a system call used in the
376 dynamic linker from libc5 and earlier. glibc does not use this
377 system call. If you intend to run programs built on libc5 or
378 earlier, you may need to enable this syscall. Current systems
379 running glibc can safely disable this.
382 bool "Auditing support"
385 Enable auditing infrastructure that can be used with another
386 kernel subsystem, such as SELinux (which requires this for
387 logging of avc messages output). System call auditing is included
388 on architectures which support it.
390 config HAVE_ARCH_AUDITSYSCALL
395 depends on AUDIT && HAVE_ARCH_AUDITSYSCALL
398 source "kernel/irq/Kconfig"
399 source "kernel/time/Kconfig"
400 source "kernel/Kconfig.preempt"
402 menu "CPU/Task time and stats accounting"
404 config VIRT_CPU_ACCOUNTING
408 prompt "Cputime accounting"
409 default TICK_CPU_ACCOUNTING if !PPC64
410 default VIRT_CPU_ACCOUNTING_NATIVE if PPC64
412 # Kind of a stub config for the pure tick based cputime accounting
413 config TICK_CPU_ACCOUNTING
414 bool "Simple tick based cputime accounting"
415 depends on !S390 && !NO_HZ_FULL
417 This is the basic tick based cputime accounting that maintains
418 statistics about user, system and idle time spent on per jiffies
423 config VIRT_CPU_ACCOUNTING_NATIVE
424 bool "Deterministic task and CPU time accounting"
425 depends on HAVE_VIRT_CPU_ACCOUNTING && !NO_HZ_FULL
426 select VIRT_CPU_ACCOUNTING
428 Select this option to enable more accurate task and CPU time
429 accounting. This is done by reading a CPU counter on each
430 kernel entry and exit and on transitions within the kernel
431 between system, softirq and hardirq state, so there is a
432 small performance impact. In the case of s390 or IBM POWER > 5,
433 this also enables accounting of stolen time on logically-partitioned
436 config VIRT_CPU_ACCOUNTING_GEN
437 bool "Full dynticks CPU time accounting"
438 depends on HAVE_CONTEXT_TRACKING
439 depends on HAVE_VIRT_CPU_ACCOUNTING_GEN
440 depends on GENERIC_CLOCKEVENTS
441 select VIRT_CPU_ACCOUNTING
442 select CONTEXT_TRACKING
444 Select this option to enable task and CPU time accounting on full
445 dynticks systems. This accounting is implemented by watching every
446 kernel-user boundaries using the context tracking subsystem.
447 The accounting is thus performed at the expense of some significant
450 For now this is only useful if you are working on the full
451 dynticks subsystem development.
457 config IRQ_TIME_ACCOUNTING
458 bool "Fine granularity task level IRQ time accounting"
459 depends on HAVE_IRQ_TIME_ACCOUNTING && !VIRT_CPU_ACCOUNTING_NATIVE
461 Select this option to enable fine granularity task irq time
462 accounting. This is done by reading a timestamp on each
463 transitions between softirq and hardirq state, so there can be a
464 small performance impact.
466 If in doubt, say N here.
468 config HAVE_SCHED_AVG_IRQ
470 depends on IRQ_TIME_ACCOUNTING || PARAVIRT_TIME_ACCOUNTING
473 config BSD_PROCESS_ACCT
474 bool "BSD Process Accounting"
477 If you say Y here, a user level program will be able to instruct the
478 kernel (via a special system call) to write process accounting
479 information to a file: whenever a process exits, information about
480 that process will be appended to the file by the kernel. The
481 information includes things such as creation time, owning user,
482 command name, memory usage, controlling terminal etc. (the complete
483 list is in the struct acct in <file:include/linux/acct.h>). It is
484 up to the user level program to do useful things with this
485 information. This is generally a good idea, so say Y.
487 config BSD_PROCESS_ACCT_V3
488 bool "BSD Process Accounting version 3 file format"
489 depends on BSD_PROCESS_ACCT
492 If you say Y here, the process accounting information is written
493 in a new file format that also logs the process IDs of each
494 process and its parent. Note that this file format is incompatible
495 with previous v0/v1/v2 file formats, so you will need updated tools
496 for processing it. A preliminary version of these tools is available
497 at <http://www.gnu.org/software/acct/>.
500 bool "Export task/process statistics through netlink"
505 Export selected statistics for tasks/processes through the
506 generic netlink interface. Unlike BSD process accounting, the
507 statistics are available during the lifetime of tasks/processes as
508 responses to commands. Like BSD accounting, they are sent to user
513 config TASK_DELAY_ACCT
514 bool "Enable per-task delay accounting"
518 Collect information on time spent by a task waiting for system
519 resources like cpu, synchronous block I/O completion and swapping
520 in pages. Such statistics can help in setting a task's priorities
521 relative to other tasks for cpu, io, rss limits etc.
526 bool "Enable extended accounting over taskstats"
529 Collect extended task accounting data and send the data
530 to userland for processing over the taskstats interface.
534 config TASK_IO_ACCOUNTING
535 bool "Enable per-task storage I/O accounting"
536 depends on TASK_XACCT
538 Collect information on the number of bytes of storage I/O which this
544 bool "Pressure stall information tracking"
546 Collect metrics that indicate how overcommitted the CPU, memory,
547 and IO capacity are in the system.
549 If you say Y here, the kernel will create /proc/pressure/ with the
550 pressure statistics files cpu, memory, and io. These will indicate
551 the share of walltime in which some or all tasks in the system are
552 delayed due to contention of the respective resource.
554 In kernels with cgroup support, cgroups (cgroup2 only) will
555 have cpu.pressure, memory.pressure, and io.pressure files,
556 which aggregate pressure stalls for the grouped tasks only.
558 For more details see Documentation/accounting/psi.rst.
562 config PSI_DEFAULT_DISABLED
563 bool "Require boot parameter to enable pressure stall information tracking"
567 If set, pressure stall information tracking will be disabled
568 per default but can be enabled through passing psi=1 on the
569 kernel commandline during boot.
571 This feature adds some code to the task wakeup and sleep
572 paths of the scheduler. The overhead is too low to affect
573 common scheduling-intense workloads in practice (such as
574 webservers, memcache), but it does show up in artificial
575 scheduler stress tests, such as hackbench.
577 If you are paranoid and not sure what the kernel will be
582 endmenu # "CPU/Task time and stats accounting"
586 depends on SMP || COMPILE_TEST
589 Make sure that CPUs running critical tasks are not disturbed by
590 any source of "noise" such as unbound workqueues, timers, kthreads...
591 Unbound jobs get offloaded to housekeeping CPUs. This is driven by
592 the "isolcpus=" boot parameter.
596 source "kernel/rcu/Kconfig"
603 tristate "Kernel .config support"
605 This option enables the complete Linux kernel ".config" file
606 contents to be saved in the kernel. It provides documentation
607 of which kernel options are used in a running kernel or in an
608 on-disk kernel. This information can be extracted from the kernel
609 image file with the script scripts/extract-ikconfig and used as
610 input to rebuild the current kernel or to build another kernel.
611 It can also be extracted from a running kernel by reading
612 /proc/config.gz if enabled (below).
615 bool "Enable access to .config through /proc/config.gz"
616 depends on IKCONFIG && PROC_FS
618 This option enables access to the kernel configuration file
619 through /proc/config.gz.
622 tristate "Enable kernel headers through /sys/kernel/kheaders.tar.xz"
625 This option enables access to the in-kernel headers that are generated during
626 the build process. These can be used to build eBPF tracing programs,
627 or similar programs. If you build the headers as a module, a module called
628 kheaders.ko is built which can be loaded on-demand to get access to headers.
631 int "Kernel log buffer size (16 => 64KB, 17 => 128KB)"
636 Select the minimal kernel log buffer size as a power of 2.
637 The final size is affected by LOG_CPU_MAX_BUF_SHIFT config
638 parameter, see below. Any higher size also might be forced
639 by "log_buf_len" boot parameter.
649 config LOG_CPU_MAX_BUF_SHIFT
650 int "CPU kernel log buffer size contribution (13 => 8 KB, 17 => 128KB)"
653 default 12 if !BASE_SMALL
654 default 0 if BASE_SMALL
657 This option allows to increase the default ring buffer size
658 according to the number of CPUs. The value defines the contribution
659 of each CPU as a power of 2. The used space is typically only few
660 lines however it might be much more when problems are reported,
663 The increased size means that a new buffer has to be allocated and
664 the original static one is unused. It makes sense only on systems
665 with more CPUs. Therefore this value is used only when the sum of
666 contributions is greater than the half of the default kernel ring
667 buffer as defined by LOG_BUF_SHIFT. The default values are set
668 so that more than 64 CPUs are needed to trigger the allocation.
670 Also this option is ignored when "log_buf_len" kernel parameter is
671 used as it forces an exact (power of two) size of the ring buffer.
673 The number of possible CPUs is used for this computation ignoring
674 hotplugging making the computation optimal for the worst case
675 scenario while allowing a simple algorithm to be used from bootup.
677 Examples shift values and their meaning:
678 17 => 128 KB for each CPU
679 16 => 64 KB for each CPU
680 15 => 32 KB for each CPU
681 14 => 16 KB for each CPU
682 13 => 8 KB for each CPU
683 12 => 4 KB for each CPU
685 config PRINTK_SAFE_LOG_BUF_SHIFT
686 int "Temporary per-CPU printk log buffer size (12 => 4KB, 13 => 8KB)"
691 Select the size of an alternate printk per-CPU buffer where messages
692 printed from usafe contexts are temporary stored. One example would
693 be NMI messages, another one - printk recursion. The messages are
694 copied to the main log buffer in a safe context to avoid a deadlock.
695 The value defines the size as a power of 2.
697 Those messages are rare and limited. The largest one is when
698 a backtrace is printed. It usually fits into 4KB. Select
699 8KB if you want to be on the safe side.
702 17 => 128 KB for each CPU
703 16 => 64 KB for each CPU
704 15 => 32 KB for each CPU
705 14 => 16 KB for each CPU
706 13 => 8 KB for each CPU
707 12 => 4 KB for each CPU
710 # Architectures with an unreliable sched_clock() should select this:
712 config HAVE_UNSTABLE_SCHED_CLOCK
715 config GENERIC_SCHED_CLOCK
718 menu "Scheduler features"
721 bool "Enable utilization clamping for RT/FAIR tasks"
722 depends on CPU_FREQ_GOV_SCHEDUTIL
724 This feature enables the scheduler to track the clamped utilization
725 of each CPU based on RUNNABLE tasks scheduled on that CPU.
727 With this option, the user can specify the min and max CPU
728 utilization allowed for RUNNABLE tasks. The max utilization defines
729 the maximum frequency a task should use while the min utilization
730 defines the minimum frequency it should use.
732 Both min and max utilization clamp values are hints to the scheduler,
733 aiming at improving its frequency selection policy, but they do not
734 enforce or grant any specific bandwidth for tasks.
738 config UCLAMP_BUCKETS_COUNT
739 int "Number of supported utilization clamp buckets"
742 depends on UCLAMP_TASK
744 Defines the number of clamp buckets to use. The range of each bucket
745 will be SCHED_CAPACITY_SCALE/UCLAMP_BUCKETS_COUNT. The higher the
746 number of clamp buckets the finer their granularity and the higher
747 the precision of clamping aggregation and tracking at run-time.
749 For example, with the minimum configuration value we will have 5
750 clamp buckets tracking 20% utilization each. A 25% boosted tasks will
751 be refcounted in the [20..39]% bucket and will set the bucket clamp
752 effective value to 25%.
753 If a second 30% boosted task should be co-scheduled on the same CPU,
754 that task will be refcounted in the same bucket of the first task and
755 it will boost the bucket clamp effective value to 30%.
756 The clamp effective value of a bucket is reset to its nominal value
757 (20% in the example above) when there are no more tasks refcounted in
760 An additional boost/capping margin can be added to some tasks. In the
761 example above the 25% task will be boosted to 30% until it exits the
762 CPU. If that should be considered not acceptable on certain systems,
763 it's always possible to reduce the margin by increasing the number of
764 clamp buckets to trade off used memory for run-time tracking
767 If in doubt, use the default value.
772 # For architectures that want to enable the support for NUMA-affine scheduler
775 config ARCH_SUPPORTS_NUMA_BALANCING
779 # For architectures that prefer to flush all TLBs after a number of pages
780 # are unmapped instead of sending one IPI per page to flush. The architecture
781 # must provide guarantees on what happens if a clean TLB cache entry is
782 # written after the unmap. Details are in mm/rmap.c near the check for
783 # should_defer_flush. The architecture should also consider if the full flush
784 # and the refill costs are offset by the savings of sending fewer IPIs.
785 config ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
790 depends on !$(cc-option,-D__SIZEOF_INT128__=0)
793 # For architectures that know their GCC __int128 support is sound
795 config ARCH_SUPPORTS_INT128
798 # For architectures that (ab)use NUMA to represent different memory regions
799 # all cpu-local but of different latencies, such as SuperH.
801 config ARCH_WANT_NUMA_VARIABLE_LOCALITY
804 config NUMA_BALANCING
805 bool "Memory placement aware NUMA scheduler"
806 depends on ARCH_SUPPORTS_NUMA_BALANCING
807 depends on !ARCH_WANT_NUMA_VARIABLE_LOCALITY
808 depends on SMP && NUMA && MIGRATION
810 This option adds support for automatic NUMA aware memory/task placement.
811 The mechanism is quite primitive and is based on migrating memory when
812 it has references to the node the task is running on.
814 This system will be inactive on UMA systems.
816 config NUMA_BALANCING_DEFAULT_ENABLED
817 bool "Automatically enable NUMA aware memory/task placement"
819 depends on NUMA_BALANCING
821 If set, automatic NUMA balancing will be enabled if running on a NUMA
825 bool "Control Group support"
828 This option adds support for grouping sets of processes together, for
829 use with process control subsystems such as Cpusets, CFS, memory
830 controls or device isolation.
832 - Documentation/scheduler/sched-design-CFS.rst (CFS)
833 - Documentation/admin-guide/cgroup-v1/ (features for grouping, isolation
834 and resource control)
844 bool "Memory controller"
848 Provides control over the memory footprint of tasks in a cgroup.
851 bool "Swap controller"
852 depends on MEMCG && SWAP
854 Provides control over the swap space consumed by tasks in a cgroup.
856 config MEMCG_SWAP_ENABLED
857 bool "Swap controller enabled by default"
858 depends on MEMCG_SWAP
861 Memory Resource Controller Swap Extension comes with its price in
862 a bigger memory consumption. General purpose distribution kernels
863 which want to enable the feature but keep it disabled by default
864 and let the user enable it by swapaccount=1 boot command line
865 parameter should have this option unselected.
866 For those who want to have the feature enabled by default should
867 select this option (if, for some reason, they need to disable it
868 then swapaccount=0 does the trick).
872 depends on MEMCG && !SLOB
880 Generic block IO controller cgroup interface. This is the common
881 cgroup interface which should be used by various IO controlling
884 Currently, CFQ IO scheduler uses it to recognize task groups and
885 control disk bandwidth allocation (proportional time slice allocation)
886 to such task groups. It is also used by bio throttling logic in
887 block layer to implement upper limit in IO rates on a device.
889 This option only enables generic Block IO controller infrastructure.
890 One needs to also enable actual IO controlling logic/policy. For
891 enabling proportional weight division of disk bandwidth in CFQ, set
892 CONFIG_CFQ_GROUP_IOSCHED=y; for enabling throttling policy, set
893 CONFIG_BLK_DEV_THROTTLING=y.
895 See Documentation/admin-guide/cgroup-v1/blkio-controller.rst for more information.
897 config CGROUP_WRITEBACK
899 depends on MEMCG && BLK_CGROUP
902 menuconfig CGROUP_SCHED
903 bool "CPU controller"
906 This feature lets CPU scheduler recognize task groups and control CPU
907 bandwidth allocation to such task groups. It uses cgroups to group
911 config FAIR_GROUP_SCHED
912 bool "Group scheduling for SCHED_OTHER"
913 depends on CGROUP_SCHED
917 bool "CPU bandwidth provisioning for FAIR_GROUP_SCHED"
918 depends on FAIR_GROUP_SCHED
921 This option allows users to define CPU bandwidth rates (limits) for
922 tasks running within the fair group scheduler. Groups with no limit
923 set are considered to be unconstrained and will run with no
925 See Documentation/scheduler/sched-bwc.rst for more information.
927 config RT_GROUP_SCHED
928 bool "Group scheduling for SCHED_RR/FIFO"
929 depends on CGROUP_SCHED
932 This feature lets you explicitly allocate real CPU bandwidth
933 to task groups. If enabled, it will also make it impossible to
934 schedule realtime tasks for non-root users until you allocate
935 realtime bandwidth for them.
936 See Documentation/scheduler/sched-rt-group.rst for more information.
940 config UCLAMP_TASK_GROUP
941 bool "Utilization clamping per group of tasks"
942 depends on CGROUP_SCHED
943 depends on UCLAMP_TASK
946 This feature enables the scheduler to track the clamped utilization
947 of each CPU based on RUNNABLE tasks currently scheduled on that CPU.
949 When this option is enabled, the user can specify a min and max
950 CPU bandwidth which is allowed for each single task in a group.
951 The max bandwidth allows to clamp the maximum frequency a task
952 can use, while the min bandwidth allows to define a minimum
953 frequency a task will always use.
955 When task group based utilization clamping is enabled, an eventually
956 specified task-specific clamp value is constrained by the cgroup
957 specified clamp value. Both minimum and maximum task clamping cannot
958 be bigger than the corresponding clamping defined at task group level.
963 bool "PIDs controller"
965 Provides enforcement of process number limits in the scope of a
966 cgroup. Any attempt to fork more processes than is allowed in the
967 cgroup will fail. PIDs are fundamentally a global resource because it
968 is fairly trivial to reach PID exhaustion before you reach even a
969 conservative kmemcg limit. As a result, it is possible to grind a
970 system to halt without being limited by other cgroup policies. The
971 PIDs controller is designed to stop this from happening.
973 It should be noted that organisational operations (such as attaching
974 to a cgroup hierarchy) will *not* be blocked by the PIDs controller,
975 since the PIDs limit only affects a process's ability to fork, not to
979 bool "RDMA controller"
981 Provides enforcement of RDMA resources defined by IB stack.
982 It is fairly easy for consumers to exhaust RDMA resources, which
983 can result into resource unavailability to other consumers.
984 RDMA controller is designed to stop this from happening.
985 Attaching processes with active RDMA resources to the cgroup
986 hierarchy is allowed even if can cross the hierarchy's limit.
988 config CGROUP_FREEZER
989 bool "Freezer controller"
991 Provides a way to freeze and unfreeze all tasks in a
994 This option affects the ORIGINAL cgroup interface. The cgroup2 memory
995 controller includes important in-kernel memory consumers per default.
997 If you're using cgroup2, say N.
999 config CGROUP_HUGETLB
1000 bool "HugeTLB controller"
1001 depends on HUGETLB_PAGE
1005 Provides a cgroup controller for HugeTLB pages.
1006 When you enable this, you can put a per cgroup limit on HugeTLB usage.
1007 The limit is enforced during page fault. Since HugeTLB doesn't
1008 support page reclaim, enforcing the limit at page fault time implies
1009 that, the application will get SIGBUS signal if it tries to access
1010 HugeTLB pages beyond its limit. This requires the application to know
1011 beforehand how much HugeTLB pages it would require for its use. The
1012 control group is tracked in the third page lru pointer. This means
1013 that we cannot use the controller with huge page less than 3 pages.
1016 bool "Cpuset controller"
1019 This option will let you create and manage CPUSETs which
1020 allow dynamically partitioning a system into sets of CPUs and
1021 Memory Nodes and assigning tasks to run only within those sets.
1022 This is primarily useful on large SMP or NUMA systems.
1026 config PROC_PID_CPUSET
1027 bool "Include legacy /proc/<pid>/cpuset file"
1031 config CGROUP_DEVICE
1032 bool "Device controller"
1034 Provides a cgroup controller implementing whitelists for
1035 devices which a process in the cgroup can mknod or open.
1037 config CGROUP_CPUACCT
1038 bool "Simple CPU accounting controller"
1040 Provides a simple controller for monitoring the
1041 total CPU consumed by the tasks in a cgroup.
1044 bool "Perf controller"
1045 depends on PERF_EVENTS
1047 This option extends the perf per-cpu mode to restrict monitoring
1048 to threads which belong to the cgroup specified and run on the
1054 bool "Support for eBPF programs attached to cgroups"
1055 depends on BPF_SYSCALL
1056 select SOCK_CGROUP_DATA
1058 Allow attaching eBPF programs to a cgroup using the bpf(2)
1059 syscall command BPF_PROG_ATTACH.
1061 In which context these programs are accessed depends on the type
1062 of attachment. For instance, programs that are attached using
1063 BPF_CGROUP_INET_INGRESS will be executed on the ingress path of
1067 bool "Debug controller"
1069 depends on DEBUG_KERNEL
1071 This option enables a simple controller that exports
1072 debugging information about the cgroups framework. This
1073 controller is for control cgroup debugging only. Its
1074 interfaces are not stable.
1078 config SOCK_CGROUP_DATA
1084 menuconfig NAMESPACES
1085 bool "Namespaces support" if EXPERT
1086 depends on MULTIUSER
1089 Provides the way to make tasks work with different objects using
1090 the same id. For example same IPC id may refer to different objects
1091 or same user id or pid may refer to different tasks when used in
1092 different namespaces.
1097 bool "UTS namespace"
1100 In this namespace tasks see different info provided with the
1104 bool "IPC namespace"
1105 depends on (SYSVIPC || POSIX_MQUEUE)
1108 In this namespace tasks work with IPC ids which correspond to
1109 different IPC objects in different namespaces.
1112 bool "User namespace"
1115 This allows containers, i.e. vservers, to use user namespaces
1116 to provide different user info for different servers.
1118 When user namespaces are enabled in the kernel it is
1119 recommended that the MEMCG option also be enabled and that
1120 user-space use the memory control groups to limit the amount
1121 of memory a memory unprivileged users can use.
1126 bool "PID Namespaces"
1129 Support process id namespaces. This allows having multiple
1130 processes with the same pid as long as they are in different
1131 pid namespaces. This is a building block of containers.
1134 bool "Network namespace"
1138 Allow user space to create what appear to be multiple instances
1139 of the network stack.
1143 config CHECKPOINT_RESTORE
1144 bool "Checkpoint/restore support"
1145 select PROC_CHILDREN
1148 Enables additional kernel features in a sake of checkpoint/restore.
1149 In particular it adds auxiliary prctl codes to setup process text,
1150 data and heap segment sizes, and a few additional /proc filesystem
1153 If unsure, say N here.
1155 config SCHED_AUTOGROUP
1156 bool "Automatic process group scheduling"
1159 select FAIR_GROUP_SCHED
1161 This option optimizes the scheduler for common desktop workloads by
1162 automatically creating and populating task groups. This separation
1163 of workloads isolates aggressive CPU burners (like build jobs) from
1164 desktop applications. Task group autogeneration is currently based
1167 config SYSFS_DEPRECATED
1168 bool "Enable deprecated sysfs features to support old userspace tools"
1172 This option adds code that switches the layout of the "block" class
1173 devices, to not show up in /sys/class/block/, but only in
1176 This switch is only active when the sysfs.deprecated=1 boot option is
1177 passed or the SYSFS_DEPRECATED_V2 option is set.
1179 This option allows new kernels to run on old distributions and tools,
1180 which might get confused by /sys/class/block/. Since 2007/2008 all
1181 major distributions and tools handle this just fine.
1183 Recent distributions and userspace tools after 2009/2010 depend on
1184 the existence of /sys/class/block/, and will not work with this
1187 Only if you are using a new kernel on an old distribution, you might
1190 config SYSFS_DEPRECATED_V2
1191 bool "Enable deprecated sysfs features by default"
1194 depends on SYSFS_DEPRECATED
1196 Enable deprecated sysfs by default.
1198 See the CONFIG_SYSFS_DEPRECATED option for more details about this
1201 Only if you are using a new kernel on an old distribution, you might
1202 need to say Y here. Even then, odds are you would not need it
1203 enabled, you can always pass the boot option if absolutely necessary.
1206 bool "Kernel->user space relay support (formerly relayfs)"
1209 This option enables support for relay interface support in
1210 certain file systems (such as debugfs).
1211 It is designed to provide an efficient mechanism for tools and
1212 facilities to relay large amounts of data from kernel space to
1217 config BLK_DEV_INITRD
1218 bool "Initial RAM filesystem and RAM disk (initramfs/initrd) support"
1220 The initial RAM filesystem is a ramfs which is loaded by the
1221 boot loader (loadlin or lilo) and that is mounted as root
1222 before the normal boot procedure. It is typically used to
1223 load modules needed to mount the "real" root file system,
1224 etc. See <file:Documentation/admin-guide/initrd.rst> for details.
1226 If RAM disk support (BLK_DEV_RAM) is also included, this
1227 also enables initial RAM disk (initrd) support and adds
1228 15 Kbytes (more on some other architectures) to the kernel size.
1234 source "usr/Kconfig"
1239 prompt "Compiler optimization level"
1240 default CC_OPTIMIZE_FOR_PERFORMANCE
1242 config CC_OPTIMIZE_FOR_PERFORMANCE
1243 bool "Optimize for performance (-O2)"
1245 This is the default optimization level for the kernel, building
1246 with the "-O2" compiler flag for best performance and most
1247 helpful compile-time warnings.
1249 config CC_OPTIMIZE_FOR_PERFORMANCE_O3
1250 bool "Optimize more for performance (-O3)"
1252 imply CC_DISABLE_WARN_MAYBE_UNINITIALIZED # avoid false positives
1254 Choosing this option will pass "-O3" to your compiler to optimize
1255 the kernel yet more for performance.
1257 config CC_OPTIMIZE_FOR_SIZE
1258 bool "Optimize for size (-Os)"
1259 imply CC_DISABLE_WARN_MAYBE_UNINITIALIZED # avoid false positives
1261 Choosing this option will pass "-Os" to your compiler resulting
1262 in a smaller kernel.
1266 config HAVE_LD_DEAD_CODE_DATA_ELIMINATION
1269 This requires that the arch annotates or otherwise protects
1270 its external entry points from being discarded. Linker scripts
1271 must also merge .text.*, .data.*, and .bss.* correctly into
1272 output sections. Care must be taken not to pull in unrelated
1273 sections (e.g., '.text.init'). Typically '.' in section names
1274 is used to distinguish them from label names / C identifiers.
1276 config LD_DEAD_CODE_DATA_ELIMINATION
1277 bool "Dead code and data elimination (EXPERIMENTAL)"
1278 depends on HAVE_LD_DEAD_CODE_DATA_ELIMINATION
1280 depends on !(FUNCTION_TRACER && CC_IS_GCC && GCC_VERSION < 40800)
1281 depends on $(cc-option,-ffunction-sections -fdata-sections)
1282 depends on $(ld-option,--gc-sections)
1284 Enable this if you want to do dead code and data elimination with
1285 the linker by compiling with -ffunction-sections -fdata-sections,
1286 and linking with --gc-sections.
1288 This can reduce on disk and in-memory size of the kernel
1289 code and static data, particularly for small configs and
1290 on small systems. This has the possibility of introducing
1291 silently broken kernel if the required annotations are not
1292 present. This option is not well tested yet, so use at your
1301 config SYSCTL_EXCEPTION_TRACE
1304 Enable support for /proc/sys/debug/exception-trace.
1306 config SYSCTL_ARCH_UNALIGN_NO_WARN
1309 Enable support for /proc/sys/kernel/ignore-unaligned-usertrap
1310 Allows arch to define/use @no_unaligned_warning to possibly warn
1311 about unaligned access emulation going on under the hood.
1313 config SYSCTL_ARCH_UNALIGN_ALLOW
1316 Enable support for /proc/sys/kernel/unaligned-trap
1317 Allows arches to define/use @unaligned_enabled to runtime toggle
1318 the unaligned access emulation.
1319 see arch/parisc/kernel/unaligned.c for reference
1321 config HAVE_PCSPKR_PLATFORM
1324 # interpreter that classic socket filters depend on
1329 bool "Configure standard kernel features (expert users)"
1330 # Unhide debug options, to make the on-by-default options visible
1333 This option allows certain base kernel options and settings
1334 to be disabled or tweaked. This is for specialized
1335 environments which can tolerate a "non-standard" kernel.
1336 Only use this if you really know what you are doing.
1339 bool "Enable 16-bit UID system calls" if EXPERT
1340 depends on HAVE_UID16 && MULTIUSER
1343 This enables the legacy 16-bit UID syscall wrappers.
1346 bool "Multiple users, groups and capabilities support" if EXPERT
1349 This option enables support for non-root users, groups and
1352 If you say N here, all processes will run with UID 0, GID 0, and all
1353 possible capabilities. Saying N here also compiles out support for
1354 system calls related to UIDs, GIDs, and capabilities, such as setuid,
1357 If unsure, say Y here.
1359 config SGETMASK_SYSCALL
1360 bool "sgetmask/ssetmask syscalls support" if EXPERT
1361 def_bool PARISC || M68K || PPC || MIPS || X86 || SPARC || MICROBLAZE || SUPERH
1363 sys_sgetmask and sys_ssetmask are obsolete system calls
1364 no longer supported in libc but still enabled by default in some
1367 If unsure, leave the default option here.
1369 config SYSFS_SYSCALL
1370 bool "Sysfs syscall support" if EXPERT
1373 sys_sysfs is an obsolete system call no longer supported in libc.
1374 Note that disabling this option is more secure but might break
1375 compatibility with some systems.
1377 If unsure say Y here.
1379 config SYSCTL_SYSCALL
1380 bool "Sysctl syscall support" if EXPERT
1381 depends on PROC_SYSCTL
1385 sys_sysctl uses binary paths that have been found challenging
1386 to properly maintain and use. The interface in /proc/sys
1387 using paths with ascii names is now the primary path to this
1390 Almost nothing using the binary sysctl interface so if you are
1391 trying to save some space it is probably safe to disable this,
1392 making your kernel marginally smaller.
1394 If unsure say N here.
1397 bool "open by fhandle syscalls" if EXPERT
1401 If you say Y here, a user level program will be able to map
1402 file names to handle and then later use the handle for
1403 different file system operations. This is useful in implementing
1404 userspace file servers, which now track files using handles instead
1405 of names. The handle would remain the same even if file names
1406 get renamed. Enables open_by_handle_at(2) and name_to_handle_at(2)
1410 bool "Posix Clocks & timers" if EXPERT
1413 This includes native support for POSIX timers to the kernel.
1414 Some embedded systems have no use for them and therefore they
1415 can be configured out to reduce the size of the kernel image.
1417 When this option is disabled, the following syscalls won't be
1418 available: timer_create, timer_gettime: timer_getoverrun,
1419 timer_settime, timer_delete, clock_adjtime, getitimer,
1420 setitimer, alarm. Furthermore, the clock_settime, clock_gettime,
1421 clock_getres and clock_nanosleep syscalls will be limited to
1422 CLOCK_REALTIME, CLOCK_MONOTONIC and CLOCK_BOOTTIME only.
1428 bool "Enable support for printk" if EXPERT
1431 This option enables normal printk support. Removing it
1432 eliminates most of the message strings from the kernel image
1433 and makes the kernel more or less silent. As this makes it
1434 very difficult to diagnose system problems, saying N here is
1435 strongly discouraged.
1443 bool "BUG() support" if EXPERT
1446 Disabling this option eliminates support for BUG and WARN, reducing
1447 the size of your kernel image and potentially quietly ignoring
1448 numerous fatal conditions. You should only consider disabling this
1449 option for embedded systems with no facilities for reporting errors.
1455 bool "Enable ELF core dumps" if EXPERT
1457 Enable support for generating core dumps. Disabling saves about 4k.
1460 config PCSPKR_PLATFORM
1461 bool "Enable PC-Speaker support" if EXPERT
1462 depends on HAVE_PCSPKR_PLATFORM
1466 This option allows to disable the internal PC-Speaker
1467 support, saving some memory.
1471 bool "Enable full-sized data structures for core" if EXPERT
1473 Disabling this option reduces the size of miscellaneous core
1474 kernel data structures. This saves memory on small machines,
1475 but may reduce performance.
1478 bool "Enable futex support" if EXPERT
1482 Disabling this option will cause the kernel to be built without
1483 support for "fast userspace mutexes". The resulting kernel may not
1484 run glibc-based applications correctly.
1488 depends on FUTEX && RT_MUTEXES
1491 config HAVE_FUTEX_CMPXCHG
1495 Architectures should select this if futex_atomic_cmpxchg_inatomic()
1496 is implemented and always working. This removes a couple of runtime
1500 bool "Enable eventpoll support" if EXPERT
1503 Disabling this option will cause the kernel to be built without
1504 support for epoll family of system calls.
1507 bool "Enable signalfd() system call" if EXPERT
1510 Enable the signalfd() system call that allows to receive signals
1511 on a file descriptor.
1516 bool "Enable timerfd() system call" if EXPERT
1519 Enable the timerfd() system call that allows to receive timer
1520 events on a file descriptor.
1525 bool "Enable eventfd() system call" if EXPERT
1528 Enable the eventfd() system call that allows to receive both
1529 kernel notification (ie. KAIO) or userspace notifications.
1534 bool "Use full shmem filesystem" if EXPERT
1538 The shmem is an internal filesystem used to manage shared memory.
1539 It is backed by swap and manages resource limits. It is also exported
1540 to userspace as tmpfs if TMPFS is enabled. Disabling this
1541 option replaces shmem and tmpfs with the much simpler ramfs code,
1542 which may be appropriate on small systems without swap.
1545 bool "Enable AIO support" if EXPERT
1548 This option enables POSIX asynchronous I/O which may by used
1549 by some high performance threaded applications. Disabling
1550 this option saves about 7k.
1553 bool "Enable IO uring support" if EXPERT
1558 This option enables support for the io_uring interface, enabling
1559 applications to submit and complete IO through submission and
1560 completion rings that are shared between the kernel and application.
1562 config ADVISE_SYSCALLS
1563 bool "Enable madvise/fadvise syscalls" if EXPERT
1566 This option enables the madvise and fadvise syscalls, used by
1567 applications to advise the kernel about their future memory or file
1568 usage, improving performance. If building an embedded system where no
1569 applications use these syscalls, you can disable this option to save
1573 bool "Enable membarrier() system call" if EXPERT
1576 Enable the membarrier() system call that allows issuing memory
1577 barriers across all running threads, which can be used to distribute
1578 the cost of user-space memory barriers asymmetrically by transforming
1579 pairs of memory barriers into pairs consisting of membarrier() and a
1585 bool "Load all symbols for debugging/ksymoops" if EXPERT
1588 Say Y here to let the kernel print out symbolic crash information and
1589 symbolic stack backtraces. This increases the size of the kernel
1590 somewhat, as all symbols have to be loaded into the kernel image.
1593 bool "Include all symbols in kallsyms"
1594 depends on DEBUG_KERNEL && KALLSYMS
1596 Normally kallsyms only contains the symbols of functions for nicer
1597 OOPS messages and backtraces (i.e., symbols from the text and inittext
1598 sections). This is sufficient for most cases. And only in very rare
1599 cases (e.g., when a debugger is used) all symbols are required (e.g.,
1600 names of variables from the data sections, etc).
1602 This option makes sure that all symbols are loaded into the kernel
1603 image (i.e., symbols from all sections) in cost of increased kernel
1604 size (depending on the kernel configuration, it may be 300KiB or
1605 something like this).
1607 Say N unless you really need all symbols.
1609 config KALLSYMS_ABSOLUTE_PERCPU
1612 default X86_64 && SMP
1614 config KALLSYMS_BASE_RELATIVE
1619 Instead of emitting them as absolute values in the native word size,
1620 emit the symbol references in the kallsyms table as 32-bit entries,
1621 each containing a relative value in the range [base, base + U32_MAX]
1622 or, when KALLSYMS_ABSOLUTE_PERCPU is in effect, each containing either
1623 an absolute value in the range [0, S32_MAX] or a relative value in the
1624 range [base, base + S32_MAX], where base is the lowest relative symbol
1625 address encountered in the image.
1627 On 64-bit builds, this reduces the size of the address table by 50%,
1628 but more importantly, it results in entries whose values are build
1629 time constants, and no relocation pass is required at runtime to fix
1630 up the entries based on the runtime load address of the kernel.
1632 # end of the "standard kernel features (expert users)" menu
1634 # syscall, maps, verifier
1636 bool "Enable bpf() system call"
1641 Enable the bpf() system call that allows to manipulate eBPF
1642 programs and maps via file descriptors.
1644 config BPF_JIT_ALWAYS_ON
1645 bool "Permanently enable BPF JIT and remove BPF interpreter"
1646 depends on BPF_SYSCALL && HAVE_EBPF_JIT && BPF_JIT
1648 Enables BPF JIT and removes BPF interpreter to avoid
1649 speculative execution of BPF instructions by the interpreter
1652 bool "Enable userfaultfd() system call"
1655 Enable the userfaultfd() system call that allows to intercept and
1656 handle page faults in userland.
1658 config ARCH_HAS_MEMBARRIER_CALLBACKS
1661 config ARCH_HAS_MEMBARRIER_SYNC_CORE
1665 bool "Enable rseq() system call" if EXPERT
1667 depends on HAVE_RSEQ
1670 Enable the restartable sequences system call. It provides a
1671 user-space cache for the current CPU number value, which
1672 speeds up getting the current CPU number from user-space,
1673 as well as an ABI to speed up user-space operations on
1680 bool "Enabled debugging of rseq() system call" if EXPERT
1681 depends on RSEQ && DEBUG_KERNEL
1683 Enable extra debugging checks for the rseq system call.
1688 bool "Embedded system"
1689 option allnoconfig_y
1692 This option should be enabled if compiling the kernel for
1693 an embedded system so certain expert options are available
1696 config HAVE_PERF_EVENTS
1699 See tools/perf/design.txt for details.
1701 config PERF_USE_VMALLOC
1704 See tools/perf/design.txt for details
1707 bool "PC/104 support" if EXPERT
1709 Expose PC/104 form factor device drivers and options available for
1710 selection and configuration. Enable this option if your target
1711 machine has a PC/104 bus.
1713 menu "Kernel Performance Events And Counters"
1716 bool "Kernel performance events and counters"
1717 default y if PROFILING
1718 depends on HAVE_PERF_EVENTS
1722 Enable kernel support for various performance events provided
1723 by software and hardware.
1725 Software events are supported either built-in or via the
1726 use of generic tracepoints.
1728 Most modern CPUs support performance events via performance
1729 counter registers. These registers count the number of certain
1730 types of hw events: such as instructions executed, cachemisses
1731 suffered, or branches mis-predicted - without slowing down the
1732 kernel or applications. These registers can also trigger interrupts
1733 when a threshold number of events have passed - and can thus be
1734 used to profile the code that runs on that CPU.
1736 The Linux Performance Event subsystem provides an abstraction of
1737 these software and hardware event capabilities, available via a
1738 system call and used by the "perf" utility in tools/perf/. It
1739 provides per task and per CPU counters, and it provides event
1740 capabilities on top of those.
1744 config DEBUG_PERF_USE_VMALLOC
1746 bool "Debug: use vmalloc to back perf mmap() buffers"
1747 depends on PERF_EVENTS && DEBUG_KERNEL && !PPC
1748 select PERF_USE_VMALLOC
1750 Use vmalloc memory to back perf mmap() buffers.
1752 Mostly useful for debugging the vmalloc code on platforms
1753 that don't require it.
1759 config VM_EVENT_COUNTERS
1761 bool "Enable VM event counters for /proc/vmstat" if EXPERT
1763 VM event counters are needed for event counts to be shown.
1764 This option allows the disabling of the VM event counters
1765 on EXPERT systems. /proc/vmstat will only show page counts
1766 if VM event counters are disabled.
1770 bool "Enable SLUB debugging support" if EXPERT
1771 depends on SLUB && SYSFS
1773 SLUB has extensive debug support features. Disabling these can
1774 result in significant savings in code size. This also disables
1775 SLUB sysfs support. /sys/slab will not exist and there will be
1776 no support for cache validation etc.
1778 config SLUB_MEMCG_SYSFS_ON
1780 bool "Enable memcg SLUB sysfs support by default" if EXPERT
1781 depends on SLUB && SYSFS && MEMCG
1783 SLUB creates a directory under /sys/kernel/slab for each
1784 allocation cache to host info and debug files. If memory
1785 cgroup is enabled, each cache can have per memory cgroup
1786 caches. SLUB can create the same sysfs directories for these
1787 caches under /sys/kernel/slab/CACHE/cgroup but it can lead
1788 to a very high number of debug files being created. This is
1789 controlled by slub_memcg_sysfs boot parameter and this
1790 config option determines the parameter's default value.
1793 bool "Disable heap randomization"
1796 Randomizing heap placement makes heap exploits harder, but it
1797 also breaks ancient binaries (including anything libc5 based).
1798 This option changes the bootup default to heap randomization
1799 disabled, and can be overridden at runtime by setting
1800 /proc/sys/kernel/randomize_va_space to 2.
1802 On non-ancient distros (post-2000 ones) N is usually a safe choice.
1805 prompt "Choose SLAB allocator"
1808 This option allows to select a slab allocator.
1812 select HAVE_HARDENED_USERCOPY_ALLOCATOR
1814 The regular slab allocator that is established and known to work
1815 well in all environments. It organizes cache hot objects in
1816 per cpu and per node queues.
1819 bool "SLUB (Unqueued Allocator)"
1820 select HAVE_HARDENED_USERCOPY_ALLOCATOR
1822 SLUB is a slab allocator that minimizes cache line usage
1823 instead of managing queues of cached objects (SLAB approach).
1824 Per cpu caching is realized using slabs of objects instead
1825 of queues of objects. SLUB can use memory efficiently
1826 and has enhanced diagnostics. SLUB is the default choice for
1831 bool "SLOB (Simple Allocator)"
1833 SLOB replaces the stock allocator with a drastically simpler
1834 allocator. SLOB is generally more space efficient but
1835 does not perform as well on large systems.
1839 config SLAB_MERGE_DEFAULT
1840 bool "Allow slab caches to be merged"
1843 For reduced kernel memory fragmentation, slab caches can be
1844 merged when they share the same size and other characteristics.
1845 This carries a risk of kernel heap overflows being able to
1846 overwrite objects from merged caches (and more easily control
1847 cache layout), which makes such heap attacks easier to exploit
1848 by attackers. By keeping caches unmerged, these kinds of exploits
1849 can usually only damage objects in the same cache. To disable
1850 merging at runtime, "slab_nomerge" can be passed on the kernel
1853 config SLAB_FREELIST_RANDOM
1855 depends on SLAB || SLUB
1856 bool "SLAB freelist randomization"
1858 Randomizes the freelist order used on creating new pages. This
1859 security feature reduces the predictability of the kernel slab
1860 allocator against heap overflows.
1862 config SLAB_FREELIST_HARDENED
1863 bool "Harden slab freelist metadata"
1866 Many kernel heap attacks try to target slab cache metadata and
1867 other infrastructure. This options makes minor performance
1868 sacrifices to harden the kernel slab allocator against common
1869 freelist exploit methods.
1871 config SHUFFLE_PAGE_ALLOCATOR
1872 bool "Page allocator randomization"
1873 default SLAB_FREELIST_RANDOM && ACPI_NUMA
1875 Randomization of the page allocator improves the average
1876 utilization of a direct-mapped memory-side-cache. See section
1877 5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
1878 6.2a specification for an example of how a platform advertises
1879 the presence of a memory-side-cache. There are also incidental
1880 security benefits as it reduces the predictability of page
1881 allocations to compliment SLAB_FREELIST_RANDOM, but the
1882 default granularity of shuffling on the "MAX_ORDER - 1" i.e,
1883 10th order of pages is selected based on cache utilization
1886 While the randomization improves cache utilization it may
1887 negatively impact workloads on platforms without a cache. For
1888 this reason, by default, the randomization is enabled only
1889 after runtime detection of a direct-mapped memory-side-cache.
1890 Otherwise, the randomization may be force enabled with the
1891 'page_alloc.shuffle' kernel command line parameter.
1895 config SLUB_CPU_PARTIAL
1897 depends on SLUB && SMP
1898 bool "SLUB per cpu partial cache"
1900 Per cpu partial caches accelerate objects allocation and freeing
1901 that is local to a processor at the price of more indeterminism
1902 in the latency of the free. On overflow these caches will be cleared
1903 which requires the taking of locks that may cause latency spikes.
1904 Typically one would choose no for a realtime system.
1906 config MMAP_ALLOW_UNINITIALIZED
1907 bool "Allow mmapped anonymous memory to be uninitialized"
1908 depends on EXPERT && !MMU
1911 Normally, and according to the Linux spec, anonymous memory obtained
1912 from mmap() has its contents cleared before it is passed to
1913 userspace. Enabling this config option allows you to request that
1914 mmap() skip that if it is given an MAP_UNINITIALIZED flag, thus
1915 providing a huge performance boost. If this option is not enabled,
1916 then the flag will be ignored.
1918 This is taken advantage of by uClibc's malloc(), and also by
1919 ELF-FDPIC binfmt's brk and stack allocator.
1921 Because of the obvious security issues, this option should only be
1922 enabled on embedded devices where you control what is run in
1923 userspace. Since that isn't generally a problem on no-MMU systems,
1924 it is normally safe to say Y here.
1926 See Documentation/nommu-mmap.txt for more information.
1928 config SYSTEM_DATA_VERIFICATION
1930 select SYSTEM_TRUSTED_KEYRING
1934 select ASYMMETRIC_KEY_TYPE
1935 select ASYMMETRIC_PUBLIC_KEY_SUBTYPE
1938 select X509_CERTIFICATE_PARSER
1939 select PKCS7_MESSAGE_PARSER
1941 Provide PKCS#7 message verification using the contents of the system
1942 trusted keyring to provide public keys. This then can be used for
1943 module verification, kexec image verification and firmware blob
1947 bool "Profiling support"
1949 Say Y here to enable the extended profiling support mechanisms used
1950 by profilers such as OProfile.
1953 # Place an empty function call at each tracepoint site. Can be
1954 # dynamically changed for a probe function.
1959 endmenu # General setup
1961 source "arch/Kconfig"
1968 default 0 if BASE_FULL
1969 default 1 if !BASE_FULL
1971 config MODULE_SIG_FORMAT
1973 select SYSTEM_DATA_VERIFICATION
1976 bool "Enable loadable module support"
1979 Kernel modules are small pieces of compiled code which can
1980 be inserted in the running kernel, rather than being
1981 permanently built into the kernel. You use the "modprobe"
1982 tool to add (and sometimes remove) them. If you say Y here,
1983 many parts of the kernel can be built as modules (by
1984 answering M instead of Y where indicated): this is most
1985 useful for infrequently used options which are not required
1986 for booting. For more information, see the man pages for
1987 modprobe, lsmod, modinfo, insmod and rmmod.
1989 If you say Y here, you will need to run "make
1990 modules_install" to put the modules under /lib/modules/
1991 where modprobe can find them (you may need to be root to do
1998 config MODULE_FORCE_LOAD
1999 bool "Forced module loading"
2002 Allow loading of modules without version information (ie. modprobe
2003 --force). Forced module loading sets the 'F' (forced) taint flag and
2004 is usually a really bad idea.
2006 config MODULE_UNLOAD
2007 bool "Module unloading"
2009 Without this option you will not be able to unload any
2010 modules (note that some modules may not be unloadable
2011 anyway), which makes your kernel smaller, faster
2012 and simpler. If unsure, say Y.
2014 config MODULE_FORCE_UNLOAD
2015 bool "Forced module unloading"
2016 depends on MODULE_UNLOAD
2018 This option allows you to force a module to unload, even if the
2019 kernel believes it is unsafe: the kernel will remove the module
2020 without waiting for anyone to stop using it (using the -f option to
2021 rmmod). This is mainly for kernel developers and desperate users.
2025 bool "Module versioning support"
2027 Usually, you have to use modules compiled with your kernel.
2028 Saying Y here makes it sometimes possible to use modules
2029 compiled for different kernels, by adding enough information
2030 to the modules to (hopefully) spot any changes which would
2031 make them incompatible with the kernel you are running. If
2034 config ASM_MODVERSIONS
2036 default HAVE_ASM_MODVERSIONS && MODVERSIONS
2038 This enables module versioning for exported symbols also from
2039 assembly. This can be enabled only when the target architecture
2042 config MODULE_REL_CRCS
2044 depends on MODVERSIONS
2046 config MODULE_SRCVERSION_ALL
2047 bool "Source checksum for all modules"
2049 Modules which contain a MODULE_VERSION get an extra "srcversion"
2050 field inserted into their modinfo section, which contains a
2051 sum of the source files which made it. This helps maintainers
2052 see exactly which source was used to build a module (since
2053 others sometimes change the module source without updating
2054 the version). With this option, such a "srcversion" field
2055 will be created for all modules. If unsure, say N.
2058 bool "Module signature verification"
2059 select MODULE_SIG_FORMAT
2061 Check modules for valid signatures upon load: the signature
2062 is simply appended to the module. For more information see
2063 <file:Documentation/admin-guide/module-signing.rst>.
2065 Note that this option adds the OpenSSL development packages as a
2066 kernel build dependency so that the signing tool can use its crypto
2069 You should enable this option if you wish to use either
2070 CONFIG_SECURITY_LOCKDOWN_LSM or lockdown functionality imposed via
2071 another LSM - otherwise unsigned modules will be loadable regardless
2072 of the lockdown policy.
2074 !!!WARNING!!! If you enable this option, you MUST make sure that the
2075 module DOES NOT get stripped after being signed. This includes the
2076 debuginfo strip done by some packagers (such as rpmbuild) and
2077 inclusion into an initramfs that wants the module size reduced.
2079 config MODULE_SIG_FORCE
2080 bool "Require modules to be validly signed"
2081 depends on MODULE_SIG
2083 Reject unsigned modules or signed modules for which we don't have a
2084 key. Without this, such modules will simply taint the kernel.
2086 config MODULE_SIG_ALL
2087 bool "Automatically sign all modules"
2089 depends on MODULE_SIG
2091 Sign all modules during make modules_install. Without this option,
2092 modules must be signed manually, using the scripts/sign-file tool.
2094 comment "Do not forget to sign required modules with scripts/sign-file"
2095 depends on MODULE_SIG_FORCE && !MODULE_SIG_ALL
2098 prompt "Which hash algorithm should modules be signed with?"
2099 depends on MODULE_SIG
2101 This determines which sort of hashing algorithm will be used during
2102 signature generation. This algorithm _must_ be built into the kernel
2103 directly so that signature verification can take place. It is not
2104 possible to load a signed module containing the algorithm to check
2105 the signature on that module.
2107 config MODULE_SIG_SHA1
2108 bool "Sign modules with SHA-1"
2111 config MODULE_SIG_SHA224
2112 bool "Sign modules with SHA-224"
2113 select CRYPTO_SHA256
2115 config MODULE_SIG_SHA256
2116 bool "Sign modules with SHA-256"
2117 select CRYPTO_SHA256
2119 config MODULE_SIG_SHA384
2120 bool "Sign modules with SHA-384"
2121 select CRYPTO_SHA512
2123 config MODULE_SIG_SHA512
2124 bool "Sign modules with SHA-512"
2125 select CRYPTO_SHA512
2129 config MODULE_SIG_HASH
2131 depends on MODULE_SIG
2132 default "sha1" if MODULE_SIG_SHA1
2133 default "sha224" if MODULE_SIG_SHA224
2134 default "sha256" if MODULE_SIG_SHA256
2135 default "sha384" if MODULE_SIG_SHA384
2136 default "sha512" if MODULE_SIG_SHA512
2138 config MODULE_COMPRESS
2139 bool "Compress modules on installation"
2142 Compresses kernel modules when 'make modules_install' is run; gzip or
2143 xz depending on "Compression algorithm" below.
2145 module-init-tools MAY support gzip, and kmod MAY support gzip and xz.
2147 Out-of-tree kernel modules installed using Kbuild will also be
2148 compressed upon installation.
2150 Note: for modules inside an initrd or initramfs, it's more efficient
2151 to compress the whole initrd or initramfs instead.
2153 Note: This is fully compatible with signed modules.
2158 prompt "Compression algorithm"
2159 depends on MODULE_COMPRESS
2160 default MODULE_COMPRESS_GZIP
2162 This determines which sort of compression will be used during
2163 'make modules_install'.
2165 GZIP (default) and XZ are supported.
2167 config MODULE_COMPRESS_GZIP
2170 config MODULE_COMPRESS_XZ
2175 config MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS
2176 bool "Allow loading of modules with missing namespace imports"
2178 Symbols exported with EXPORT_SYMBOL_NS*() are considered exported in
2179 a namespace. A module that makes use of a symbol exported with such a
2180 namespace is required to import the namespace via MODULE_IMPORT_NS().
2181 There is no technical reason to enforce correct namespace imports,
2182 but it creates consistency between symbols defining namespaces and
2183 users importing namespaces they make use of. This option relaxes this
2184 requirement and lifts the enforcement when loading a module.
2188 config UNUSED_SYMBOLS
2189 bool "Enable unused/obsolete exported symbols"
2192 Unused but exported symbols make the kernel needlessly bigger. For
2193 that reason most of these unused exports will soon be removed. This
2194 option is provided temporarily to provide a transition period in case
2195 some external kernel module needs one of these symbols anyway. If you
2196 encounter such a case in your module, consider if you are actually
2197 using the right API. (rationale: since nobody in the kernel is using
2198 this in a module, there is a pretty good chance it's actually the
2199 wrong interface to use). If you really need the symbol, please send a
2200 mail to the linux kernel mailing list mentioning the symbol and why
2201 you really need it, and what the merge plan to the mainline kernel for
2204 config TRIM_UNUSED_KSYMS
2205 bool "Trim unused exported kernel symbols"
2206 depends on !UNUSED_SYMBOLS
2208 The kernel and some modules make many symbols available for
2209 other modules to use via EXPORT_SYMBOL() and variants. Depending
2210 on the set of modules being selected in your kernel configuration,
2211 many of those exported symbols might never be used.
2213 This option allows for unused exported symbols to be dropped from
2214 the build. In turn, this provides the compiler more opportunities
2215 (especially when using LTO) for optimizing the code and reducing
2216 binary size. This might have some security advantages as well.
2218 If unsure, or if you need to build out-of-tree modules, say N.
2222 config MODULES_TREE_LOOKUP
2224 depends on PERF_EVENTS || TRACING
2226 config INIT_ALL_POSSIBLE
2229 Back when each arch used to define their own cpu_online_mask and
2230 cpu_possible_mask, some of them chose to initialize cpu_possible_mask
2231 with all 1s, and others with all 0s. When they were centralised,
2232 it was better to provide this option than to break all the archs
2233 and have several arch maintainers pursuing me down dark alleys.
2235 source "block/Kconfig"
2237 config PREEMPT_NOTIFIERS
2247 Build a simple ASN.1 grammar compiler that produces a bytecode output
2248 that can be interpreted by the ASN.1 stream decoder and used to
2249 inform it as to what tags are to be expected in a stream and what
2250 functions to call on what tags.
2252 source "kernel/Kconfig.locks"
2254 config ARCH_HAS_SYNC_CORE_BEFORE_USERMODE
2257 # It may be useful for an architecture to override the definitions of the
2258 # SYSCALL_DEFINE() and __SYSCALL_DEFINEx() macros in <linux/syscalls.h>
2259 # and the COMPAT_ variants in <linux/compat.h>, in particular to use a
2260 # different calling convention for syscalls. They can also override the
2261 # macros for not-implemented syscalls in kernel/sys_ni.c and
2262 # kernel/time/posix-stubs.c. All these overrides need to be available in
2263 # <asm/syscall_wrapper.h>.
2264 config ARCH_HAS_SYSCALL_WRAPPER