1 # SPDX-License-Identifier: GPL-2.0
3 # General architecture dependent options
7 # Note: arch/$(SRCARCH)/Kconfig needs to be included first so that it can
8 # override the default values in this file.
10 source "arch/$(SRCARCH)/Kconfig"
12 config ARCH_CONFIGURES_CPU_MITIGATIONS
15 if !ARCH_CONFIGURES_CPU_MITIGATIONS
16 config CPU_MITIGATIONS
20 menu "General architecture-dependent options"
35 config ARCH_HAS_SUBPAGE_FAULTS
38 Select if the architecture can check permissions at sub-page
39 granularity (e.g. arm64 MTE). The probe_user_*() functions
45 config SMT_NUM_THREADS_DYNAMIC
54 depends on HAVE_KPROBES
56 select TASKS_RCU if PREEMPTION
58 Kprobes allows you to trap at almost any kernel address and
59 execute a callback function. register_kprobe() establishes
60 a probepoint and specifies the callback. Kprobes is useful
61 for kernel debugging, non-intrusive instrumentation and testing.
65 bool "Optimize very unlikely/likely branches"
66 depends on HAVE_ARCH_JUMP_LABEL
67 select OBJTOOL if HAVE_JUMP_LABEL_HACK
69 This option enables a transparent branch optimization that
70 makes certain almost-always-true or almost-always-false branch
71 conditions even cheaper to execute within the kernel.
73 Certain performance-sensitive kernel code, such as trace points,
74 scheduler functionality, networking code and KVM have such
75 branches and include support for this optimization technique.
77 If it is detected that the compiler has support for "asm goto",
78 the kernel will compile such branches with just a nop
79 instruction. When the condition flag is toggled to true, the
80 nop will be converted to a jump instruction to execute the
81 conditional block of instructions.
83 This technique lowers overhead and stress on the branch prediction
84 of the processor and generally makes the kernel faster. The update
85 of the condition is slower, but those are always very rare.
87 ( On 32-bit x86, the necessary options added to the compiler
88 flags may increase the size of the kernel slightly. )
90 config STATIC_KEYS_SELFTEST
91 bool "Static key selftest"
94 Boot time self-test of the branch patching code.
96 config STATIC_CALL_SELFTEST
97 bool "Static call selftest"
98 depends on HAVE_STATIC_CALL
100 Boot time self-test of the call patching code.
104 depends on KPROBES && HAVE_OPTPROBES
105 select TASKS_RCU if PREEMPTION
107 config KPROBES_ON_FTRACE
109 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
110 depends on DYNAMIC_FTRACE_WITH_REGS
112 If function tracer is enabled and the arch supports full
113 passing of pt_regs to function tracing, then kprobes can
114 optimize on top of function tracing.
118 depends on ARCH_SUPPORTS_UPROBES
120 Uprobes is the user-space counterpart to kprobes: they
121 enable instrumentation applications (such as 'perf probe')
122 to establish unintrusive probes in user-space binaries and
123 libraries, by executing handler functions when the probes
124 are hit by user-space applications.
126 ( These probes come in the form of single-byte breakpoints,
127 managed by the kernel and kept transparent to the probed
130 config HAVE_64BIT_ALIGNED_ACCESS
131 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
133 Some architectures require 64 bit accesses to be 64 bit
134 aligned, which also requires structs containing 64 bit values
135 to be 64 bit aligned too. This includes some 32 bit
136 architectures which can do 64 bit accesses, as well as 64 bit
137 architectures without unaligned access.
139 This symbol should be selected by an architecture if 64 bit
140 accesses are required to be 64 bit aligned in this way even
141 though it is not a 64 bit architecture.
143 See Documentation/core-api/unaligned-memory-access.rst for
144 more information on the topic of unaligned memory accesses.
146 config HAVE_EFFICIENT_UNALIGNED_ACCESS
149 Some architectures are unable to perform unaligned accesses
150 without the use of get_unaligned/put_unaligned. Others are
151 unable to perform such accesses efficiently (e.g. trap on
152 unaligned access and require fixing it up in the exception
155 This symbol should be selected by an architecture if it can
156 perform unaligned accesses efficiently to allow different
157 code paths to be selected for these cases. Some network
158 drivers, for example, could opt to not fix up alignment
159 problems with received packets if doing so would not help
162 See Documentation/core-api/unaligned-memory-access.rst for more
163 information on the topic of unaligned memory accesses.
165 config ARCH_USE_BUILTIN_BSWAP
168 Modern versions of GCC (since 4.4) have builtin functions
169 for handling byte-swapping. Using these, instead of the old
170 inline assembler that the architecture code provides in the
171 __arch_bswapXX() macros, allows the compiler to see what's
172 happening and offers more opportunity for optimisation. In
173 particular, the compiler will be able to combine the byteswap
174 with a nearby load or store and use load-and-swap or
175 store-and-swap instructions if the architecture has them. It
176 should almost *never* result in code which is worse than the
177 hand-coded assembler in <asm/swab.h>. But just in case it
178 does, the use of the builtins is optional.
180 Any architecture with load-and-swap or store-and-swap
181 instructions should set this. And it shouldn't hurt to set it
182 on architectures that don't have such instructions.
186 depends on KPROBES && (HAVE_KRETPROBES || HAVE_RETHOOK)
188 config KRETPROBE_ON_RETHOOK
190 depends on HAVE_RETHOOK
191 depends on KRETPROBES
194 config USER_RETURN_NOTIFIER
196 depends on HAVE_USER_RETURN_NOTIFIER
198 Provide a kernel-internal notification when a cpu is about to
201 config HAVE_IOREMAP_PROT
207 config HAVE_KRETPROBES
210 config HAVE_OPTPROBES
213 config HAVE_KPROBES_ON_FTRACE
216 config ARCH_CORRECT_STACKTRACE_ON_KRETPROBE
219 Since kretprobes modifies return address on the stack, the
220 stacktrace may see the kretprobe trampoline address instead
221 of correct one. If the architecture stacktrace code and
222 unwinder can adjust such entries, select this configuration.
224 config HAVE_FUNCTION_ERROR_INJECTION
230 config HAVE_FUNCTION_DESCRIPTORS
233 config TRACE_IRQFLAGS_SUPPORT
236 config TRACE_IRQFLAGS_NMI_SUPPORT
240 # An arch should select this if it provides all these things:
242 # task_pt_regs() in asm/processor.h or asm/ptrace.h
243 # arch_has_single_step() if there is hardware single-step support
244 # arch_has_block_step() if there is hardware block-step support
245 # asm/syscall.h supplying asm-generic/syscall.h interface
246 # linux/regset.h user_regset interfaces
247 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
248 # TIF_SYSCALL_TRACE calls ptrace_report_syscall_{entry,exit}
249 # TIF_NOTIFY_RESUME calls resume_user_mode_work()
251 config HAVE_ARCH_TRACEHOOK
254 config HAVE_DMA_CONTIGUOUS
257 config GENERIC_SMP_IDLE_THREAD
260 config GENERIC_IDLE_POLL_SETUP
263 config ARCH_HAS_FORTIFY_SOURCE
266 An architecture should select this when it can successfully
267 build and run with CONFIG_FORTIFY_SOURCE.
270 # Select if the arch provides a historic keepinit alias for the retain_initrd
271 # command line option
273 config ARCH_HAS_KEEPINITRD
276 # Select if arch has all set_memory_ro/rw/x/nx() functions in asm/cacheflush.h
277 config ARCH_HAS_SET_MEMORY
280 # Select if arch has all set_direct_map_invalid/default() functions
281 config ARCH_HAS_SET_DIRECT_MAP
285 # Select if the architecture provides the arch_dma_set_uncached symbol to
286 # either provide an uncached segment alias for a DMA allocation, or
287 # to remap the page tables in place.
289 config ARCH_HAS_DMA_SET_UNCACHED
293 # Select if the architectures provides the arch_dma_clear_uncached symbol
294 # to undo an in-place page table remap for uncached access.
296 config ARCH_HAS_DMA_CLEAR_UNCACHED
299 config ARCH_HAS_CPU_FINALIZE_INIT
302 # Select if arch init_task must go in the __init_task_data section
303 config ARCH_TASK_STRUCT_ON_STACK
306 # Select if arch has its private alloc_task_struct() function
307 config ARCH_TASK_STRUCT_ALLOCATOR
310 config HAVE_ARCH_THREAD_STRUCT_WHITELIST
312 depends on !ARCH_TASK_STRUCT_ALLOCATOR
314 An architecture should select this to provide hardened usercopy
315 knowledge about what region of the thread_struct should be
316 whitelisted for copying to userspace. Normally this is only the
317 FPU registers. Specifically, arch_thread_struct_whitelist()
318 should be implemented. Without this, the entire thread_struct
319 field in task_struct will be left whitelisted.
321 # Select if arch has its private alloc_thread_stack() function
322 config ARCH_THREAD_STACK_ALLOCATOR
325 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
326 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
329 config ARCH_WANTS_NO_INSTR
332 An architecture should select this if the noinstr macro is being used on
333 functions to denote that the toolchain should avoid instrumenting such
334 functions and is required for correctness.
336 config ARCH_32BIT_OFF_T
340 All new 32-bit architectures should have 64-bit off_t type on
341 userspace side which corresponds to the loff_t kernel type. This
342 is the requirement for modern ABIs. Some existing architectures
343 still support 32-bit off_t. This option is enabled for all such
344 architectures explicitly.
346 # Selected by 64 bit architectures which have a 32 bit f_tinode in struct ustat
347 config ARCH_32BIT_USTAT_F_TINODE
350 config HAVE_ASM_MODVERSIONS
353 This symbol should be selected by an architecture if it provides
354 <asm/asm-prototypes.h> to support the module versioning for symbols
355 exported from assembly code.
357 config HAVE_REGS_AND_STACK_ACCESS_API
360 This symbol should be selected by an architecture if it supports
361 the API needed to access registers and stack entries from pt_regs,
362 declared in asm/ptrace.h
363 For example the kprobes-based event tracer needs this API.
367 depends on HAVE_REGS_AND_STACK_ACCESS_API
369 This symbol should be selected by an architecture if it
370 supports an implementation of restartable sequences.
375 This symbol should be selected by an architecture if it
378 config HAVE_FUNCTION_ARG_ACCESS_API
381 This symbol should be selected by an architecture if it supports
382 the API needed to access function arguments from pt_regs,
383 declared in asm/ptrace.h
385 config HAVE_HW_BREAKPOINT
387 depends on PERF_EVENTS
389 config HAVE_MIXED_BREAKPOINTS_REGS
391 depends on HAVE_HW_BREAKPOINT
393 Depending on the arch implementation of hardware breakpoints,
394 some of them have separate registers for data and instruction
395 breakpoints addresses, others have mixed registers to store
396 them but define the access type in a control register.
397 Select this option if your arch implements breakpoints under the
400 config HAVE_USER_RETURN_NOTIFIER
403 config HAVE_PERF_EVENTS_NMI
406 System hardware can generate an NMI using the perf event
407 subsystem. Also has support for calculating CPU cycle events
408 to determine how many clock cycles in a given period.
410 config HAVE_HARDLOCKUP_DETECTOR_PERF
412 depends on HAVE_PERF_EVENTS_NMI
414 The arch chooses to use the generic perf-NMI-based hardlockup
415 detector. Must define HAVE_PERF_EVENTS_NMI.
417 config HAVE_NMI_WATCHDOG
421 The arch provides a low level NMI watchdog. It provides
422 asm/nmi.h, and defines its own arch_touch_nmi_watchdog().
424 config HAVE_HARDLOCKUP_DETECTOR_ARCH
426 select HAVE_NMI_WATCHDOG
428 The arch chooses to provide its own hardlockup detector, which is
429 a superset of the HAVE_NMI_WATCHDOG. It also conforms to config
430 interfaces and parameters provided by hardlockup detector subsystem.
432 config HAVE_PERF_REGS
435 Support selective register dumps for perf events. This includes
436 bit-mapping of each registers and a unique architecture id.
438 config HAVE_PERF_USER_STACK_DUMP
441 Support user stack dumps for perf event samples. This needs
442 access to the user stack pointer which is not unified across
445 config HAVE_ARCH_JUMP_LABEL
448 config HAVE_ARCH_JUMP_LABEL_RELATIVE
451 config MMU_GATHER_TABLE_FREE
454 config MMU_GATHER_RCU_TABLE_FREE
456 select MMU_GATHER_TABLE_FREE
458 config MMU_GATHER_PAGE_SIZE
461 config MMU_GATHER_NO_RANGE
463 select MMU_GATHER_MERGE_VMAS
465 config MMU_GATHER_NO_FLUSH_CACHE
468 config MMU_GATHER_MERGE_VMAS
471 config MMU_GATHER_NO_GATHER
473 depends on MMU_GATHER_TABLE_FREE
475 config ARCH_WANT_IRQS_OFF_ACTIVATE_MM
478 Temporary select until all architectures can be converted to have
479 irqs disabled over activate_mm. Architectures that do IPI based TLB
480 shootdowns should enable this.
482 config ARCH_HAVE_NMI_SAFE_CMPXCHG
485 config HAVE_ALIGNED_STRUCT_PAGE
488 This makes sure that struct pages are double word aligned and that
489 e.g. the SLUB allocator can perform double word atomic operations
490 on a struct page for better performance. However selecting this
491 might increase the size of a struct page by a word.
493 config HAVE_CMPXCHG_LOCAL
496 config HAVE_CMPXCHG_DOUBLE
499 config ARCH_WEAK_RELEASE_ACQUIRE
502 config ARCH_WANT_IPC_PARSE_VERSION
505 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
508 config ARCH_WANT_OLD_COMPAT_IPC
509 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
512 config HAVE_ARCH_SECCOMP
515 An arch should select this symbol to support seccomp mode 1 (the fixed
516 syscall policy), and must provide an overrides for __NR_seccomp_sigreturn,
517 and compat syscalls if the asm-generic/seccomp.h defaults need adjustment:
518 - __NR_seccomp_read_32
519 - __NR_seccomp_write_32
520 - __NR_seccomp_exit_32
521 - __NR_seccomp_sigreturn_32
523 config HAVE_ARCH_SECCOMP_FILTER
525 select HAVE_ARCH_SECCOMP
527 An arch should select this symbol if it provides all of these things:
528 - all the requirements for HAVE_ARCH_SECCOMP
530 - syscall_get_arguments()
532 - syscall_set_return_value()
533 - SIGSYS siginfo_t support
534 - secure_computing is called from a ptrace_event()-safe context
535 - secure_computing return value is checked and a return value of -1
536 results in the system call being skipped immediately.
537 - seccomp syscall wired up
538 - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE,
539 SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If
540 COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too.
543 prompt "Enable seccomp to safely execute untrusted bytecode"
545 depends on HAVE_ARCH_SECCOMP
547 This kernel feature is useful for number crunching applications
548 that may need to handle untrusted bytecode during their
549 execution. By using pipes or other transports made available
550 to the process as file descriptors supporting the read/write
551 syscalls, it's possible to isolate those applications in their
552 own address space using seccomp. Once seccomp is enabled via
553 prctl(PR_SET_SECCOMP) or the seccomp() syscall, it cannot be
554 disabled and the task is only allowed to execute a few safe
555 syscalls defined by each seccomp mode.
559 config SECCOMP_FILTER
561 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
563 Enable tasks to build secure computing environments defined
564 in terms of Berkeley Packet Filter programs which implement
565 task-defined system call filtering polices.
567 See Documentation/userspace-api/seccomp_filter.rst for details.
569 config SECCOMP_CACHE_DEBUG
570 bool "Show seccomp filter cache status in /proc/pid/seccomp_cache"
571 depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR
574 This enables the /proc/pid/seccomp_cache interface to monitor
575 seccomp cache data. The file format is subject to change. Reading
576 the file requires CAP_SYS_ADMIN.
578 This option is for debugging only. Enabling presents the risk that
579 an adversary may be able to infer the seccomp filter logic.
583 config HAVE_ARCH_STACKLEAK
586 An architecture should select this if it has the code which
587 fills the used part of the kernel stack with the STACKLEAK_POISON
588 value before returning from system calls.
590 config HAVE_STACKPROTECTOR
593 An arch should select this symbol if:
594 - it has implemented a stack canary (e.g. __stack_chk_guard)
596 config STACKPROTECTOR
597 bool "Stack Protector buffer overflow detection"
598 depends on HAVE_STACKPROTECTOR
599 depends on $(cc-option,-fstack-protector)
602 This option turns on the "stack-protector" GCC feature. This
603 feature puts, at the beginning of functions, a canary value on
604 the stack just before the return address, and validates
605 the value just before actually returning. Stack based buffer
606 overflows (that need to overwrite this return address) now also
607 overwrite the canary, which gets detected and the attack is then
608 neutralized via a kernel panic.
610 Functions will have the stack-protector canary logic added if they
611 have an 8-byte or larger character array on the stack.
613 This feature requires gcc version 4.2 or above, or a distribution
614 gcc with the feature backported ("-fstack-protector").
616 On an x86 "defconfig" build, this feature adds canary checks to
617 about 3% of all kernel functions, which increases kernel code size
620 config STACKPROTECTOR_STRONG
621 bool "Strong Stack Protector"
622 depends on STACKPROTECTOR
623 depends on $(cc-option,-fstack-protector-strong)
626 Functions will have the stack-protector canary logic added in any
627 of the following conditions:
629 - local variable's address used as part of the right hand side of an
630 assignment or function argument
631 - local variable is an array (or union containing an array),
632 regardless of array type or length
633 - uses register local variables
635 This feature requires gcc version 4.9 or above, or a distribution
636 gcc with the feature backported ("-fstack-protector-strong").
638 On an x86 "defconfig" build, this feature adds canary checks to
639 about 20% of all kernel functions, which increases the kernel code
642 config ARCH_SUPPORTS_SHADOW_CALL_STACK
645 An architecture should select this if it supports the compiler's
646 Shadow Call Stack and implements runtime support for shadow stack
649 config SHADOW_CALL_STACK
650 bool "Shadow Call Stack"
651 depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
652 depends on DYNAMIC_FTRACE_WITH_ARGS || DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
655 This option enables the compiler's Shadow Call Stack, which
656 uses a shadow stack to protect function return addresses from
657 being overwritten by an attacker. More information can be found
658 in the compiler's documentation:
660 - Clang: https://clang.llvm.org/docs/ShadowCallStack.html
661 - GCC: https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html#Instrumentation-Options
663 Note that security guarantees in the kernel differ from the
664 ones documented for user space. The kernel must store addresses
665 of shadow stacks in memory, which means an attacker capable of
666 reading and writing arbitrary memory may be able to locate them
667 and hijack control flow by modifying the stacks.
672 Selected if the kernel will be built using the compiler's LTO feature.
678 Selected if the kernel will be built using Clang's LTO feature.
680 config ARCH_SUPPORTS_LTO_CLANG
683 An architecture should select this option if it supports:
684 - compiling with Clang,
685 - compiling inline assembly with Clang's integrated assembler,
686 - and linking with LLD.
688 config ARCH_SUPPORTS_LTO_CLANG_THIN
691 An architecture should select this option if it can support Clang's
696 depends on CC_IS_CLANG && LD_IS_LLD && AS_IS_LLVM
697 depends on $(success,$(NM) --help | head -n 1 | grep -qi llvm)
698 depends on $(success,$(AR) --help | head -n 1 | grep -qi llvm)
699 depends on ARCH_SUPPORTS_LTO_CLANG
700 depends on !FTRACE_MCOUNT_USE_RECORDMCOUNT
701 depends on !KASAN || KASAN_HW_TAGS
702 depends on !GCOV_KERNEL
704 The compiler and Kconfig options support building with Clang's
708 prompt "Link Time Optimization (LTO)"
711 This option enables Link Time Optimization (LTO), which allows the
712 compiler to optimize binaries globally.
714 If unsure, select LTO_NONE. Note that LTO is very resource-intensive
715 so it's disabled by default.
720 Build the kernel normally, without Link Time Optimization (LTO).
722 config LTO_CLANG_FULL
723 bool "Clang Full LTO (EXPERIMENTAL)"
724 depends on HAS_LTO_CLANG
725 depends on !COMPILE_TEST
728 This option enables Clang's full Link Time Optimization (LTO), which
729 allows the compiler to optimize the kernel globally. If you enable
730 this option, the compiler generates LLVM bitcode instead of ELF
731 object files, and the actual compilation from bitcode happens at
732 the LTO link step, which may take several minutes depending on the
733 kernel configuration. More information can be found from LLVM's
736 https://llvm.org/docs/LinkTimeOptimization.html
738 During link time, this option can use a large amount of RAM, and
739 may take much longer than the ThinLTO option.
741 config LTO_CLANG_THIN
742 bool "Clang ThinLTO (EXPERIMENTAL)"
743 depends on HAS_LTO_CLANG && ARCH_SUPPORTS_LTO_CLANG_THIN
746 This option enables Clang's ThinLTO, which allows for parallel
747 optimization and faster incremental compiles compared to the
748 CONFIG_LTO_CLANG_FULL option. More information can be found
749 from Clang's documentation:
751 https://clang.llvm.org/docs/ThinLTO.html
756 config ARCH_SUPPORTS_CFI_CLANG
759 An architecture should select this option if it can support Clang's
760 Control-Flow Integrity (CFI) checking.
762 config ARCH_USES_CFI_TRAPS
766 bool "Use Clang's Control Flow Integrity (CFI)"
767 depends on ARCH_SUPPORTS_CFI_CLANG
768 depends on $(cc-option,-fsanitize=kcfi)
770 This option enables Clang’s forward-edge Control Flow Integrity
771 (CFI) checking, where the compiler injects a runtime check to each
772 indirect function call to ensure the target is a valid function with
773 the correct static type. This restricts possible call targets and
774 makes it more difficult for an attacker to exploit bugs that allow
775 the modification of stored function pointers. More information can be
776 found from Clang's documentation:
778 https://clang.llvm.org/docs/ControlFlowIntegrity.html
780 config CFI_PERMISSIVE
781 bool "Use CFI in permissive mode"
784 When selected, Control Flow Integrity (CFI) violations result in a
785 warning instead of a kernel panic. This option should only be used
786 for finding indirect call type mismatches during development.
790 config HAVE_ARCH_WITHIN_STACK_FRAMES
793 An architecture should select this if it can walk the kernel stack
794 frames to determine if an object is part of either the arguments
795 or local variables (i.e. that it excludes saved return addresses,
796 and similar) by implementing an inline arch_within_stack_frames(),
797 which is used by CONFIG_HARDENED_USERCOPY.
799 config HAVE_CONTEXT_TRACKING_USER
802 Provide kernel/user boundaries probes necessary for subsystems
803 that need it, such as userspace RCU extended quiescent state.
804 Syscalls need to be wrapped inside user_exit()-user_enter(), either
805 optimized behind static key or through the slow path using TIF_NOHZ
806 flag. Exceptions handlers must be wrapped as well. Irqs are already
807 protected inside ct_irq_enter/ct_irq_exit() but preemption or signal
808 handling on irq exit still need to be protected.
810 config HAVE_CONTEXT_TRACKING_USER_OFFSTACK
813 Architecture neither relies on exception_enter()/exception_exit()
814 nor on schedule_user(). Also preempt_schedule_notrace() and
815 preempt_schedule_irq() can't be called in a preemptible section
816 while context tracking is CONTEXT_USER. This feature reflects a sane
817 entry implementation where the following requirements are met on
818 critical entry code, ie: before user_exit() or after user_enter():
820 - Critical entry code isn't preemptible (or better yet:
822 - No use of RCU read side critical sections, unless ct_nmi_enter()
824 - No use of instrumentation, unless instrumentation_begin() got
830 Arch relies on TIF_NOHZ and syscall slow path to implement context
831 tracking calls to user_enter()/user_exit().
833 config HAVE_VIRT_CPU_ACCOUNTING
836 config HAVE_VIRT_CPU_ACCOUNTING_IDLE
839 Architecture has its own way to account idle CPU time and therefore
840 doesn't implement vtime_account_idle().
842 config ARCH_HAS_SCALED_CPUTIME
845 config HAVE_VIRT_CPU_ACCOUNTING_GEN
849 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
850 Before enabling this option, arch code must be audited
851 to ensure there are no races in concurrent read/write of
852 cputime_t. For example, reading/writing 64-bit cputime_t on
853 some 32-bit arches may require multiple accesses, so proper
854 locking is needed to protect against concurrent accesses.
856 config HAVE_IRQ_TIME_ACCOUNTING
859 Archs need to ensure they use a high enough resolution clock to
860 support irq time accounting and then call enable_sched_clock_irqtime().
865 Architectures that select this are able to move page tables at the
866 PUD level. If there are only 3 page table levels, the move effectively
867 happens at the PGD level.
872 Archs that select this are able to move page tables at the PMD level.
874 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
877 config HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
880 config HAVE_ARCH_HUGE_VMAP
884 # Archs that select this would be capable of PMD-sized vmaps (i.e.,
885 # arch_vmap_pmd_supported() returns true). The VM_ALLOW_HUGE_VMAP flag
886 # must be used to enable allocations to use hugepages.
888 config HAVE_ARCH_HUGE_VMALLOC
889 depends on HAVE_ARCH_HUGE_VMAP
892 config ARCH_WANT_HUGE_PMD_SHARE
895 config HAVE_ARCH_SOFT_DIRTY
898 config HAVE_MOD_ARCH_SPECIFIC
901 The arch uses struct mod_arch_specific to store data. Many arches
902 just need a simple module loader without arch specific data - those
903 should not enable this.
905 config MODULES_USE_ELF_RELA
908 Modules only use ELF RELA relocations. Modules with ELF REL
909 relocations will give an error.
911 config MODULES_USE_ELF_REL
914 Modules only use ELF REL relocations. Modules with ELF RELA
915 relocations will give an error.
917 config ARCH_WANTS_MODULES_DATA_IN_VMALLOC
920 For architectures like powerpc/32 which have constraints on module
921 allocation and need to allocate module data outside of module area.
923 config HAVE_IRQ_EXIT_ON_IRQ_STACK
926 Architecture doesn't only execute the irq handler on the irq stack
927 but also irq_exit(). This way we can process softirqs on this irq
928 stack instead of switching to a new one when we call __do_softirq()
929 in the end of an hardirq.
930 This spares a stack switch and improves cache usage on softirq
933 config HAVE_SOFTIRQ_ON_OWN_STACK
936 Architecture provides a function to run __do_softirq() on a
939 config SOFTIRQ_ON_OWN_STACK
940 def_bool HAVE_SOFTIRQ_ON_OWN_STACK && !PREEMPT_RT
942 config ALTERNATE_USER_ADDRESS_SPACE
945 Architectures set this when the CPU uses separate address
946 spaces for kernel and user space pointers. In this case, the
947 access_ok() check on a __user pointer is skipped.
949 config PGTABLE_LEVELS
953 config ARCH_HAS_ELF_RANDOMIZE
956 An architecture supports choosing randomized locations for
957 stack, mmap, brk, and ET_DYN. Defined functions:
959 - arch_randomize_brk()
961 config HAVE_ARCH_MMAP_RND_BITS
964 An arch should select this symbol if it supports setting a variable
965 number of bits for use in establishing the base address for mmap
966 allocations, has MMU enabled and provides values for both:
967 - ARCH_MMAP_RND_BITS_MIN
968 - ARCH_MMAP_RND_BITS_MAX
970 config HAVE_EXIT_THREAD
973 An architecture implements exit_thread.
975 config ARCH_MMAP_RND_BITS_MIN
978 config ARCH_MMAP_RND_BITS_MAX
981 config ARCH_MMAP_RND_BITS_DEFAULT
984 config ARCH_MMAP_RND_BITS
985 int "Number of bits to use for ASLR of mmap base address" if EXPERT
986 range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX
987 default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT
988 default ARCH_MMAP_RND_BITS_MIN
989 depends on HAVE_ARCH_MMAP_RND_BITS
991 This value can be used to select the number of bits to use to
992 determine the random offset to the base address of vma regions
993 resulting from mmap allocations. This value will be bounded
994 by the architecture's minimum and maximum supported values.
996 This value can be changed after boot using the
997 /proc/sys/vm/mmap_rnd_bits tunable
999 config HAVE_ARCH_MMAP_RND_COMPAT_BITS
1002 An arch should select this symbol if it supports running applications
1003 in compatibility mode, supports setting a variable number of bits for
1004 use in establishing the base address for mmap allocations, has MMU
1005 enabled and provides values for both:
1006 - ARCH_MMAP_RND_COMPAT_BITS_MIN
1007 - ARCH_MMAP_RND_COMPAT_BITS_MAX
1009 config ARCH_MMAP_RND_COMPAT_BITS_MIN
1012 config ARCH_MMAP_RND_COMPAT_BITS_MAX
1015 config ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1018 config ARCH_MMAP_RND_COMPAT_BITS
1019 int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT
1020 range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX
1021 default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1022 default ARCH_MMAP_RND_COMPAT_BITS_MIN
1023 depends on HAVE_ARCH_MMAP_RND_COMPAT_BITS
1025 This value can be used to select the number of bits to use to
1026 determine the random offset to the base address of vma regions
1027 resulting from mmap allocations for compatible applications This
1028 value will be bounded by the architecture's minimum and maximum
1031 This value can be changed after boot using the
1032 /proc/sys/vm/mmap_rnd_compat_bits tunable
1034 config HAVE_ARCH_COMPAT_MMAP_BASES
1037 This allows 64bit applications to invoke 32-bit mmap() syscall
1038 and vice-versa 32-bit applications to call 64-bit mmap().
1039 Required for applications doing different bitness syscalls.
1041 config PAGE_SIZE_LESS_THAN_64KB
1043 depends on !ARM64_64K_PAGES
1044 depends on !IA64_PAGE_SIZE_64KB
1045 depends on !PAGE_SIZE_64KB
1046 depends on !PARISC_PAGE_SIZE_64KB
1047 depends on PAGE_SIZE_LESS_THAN_256KB
1049 config PAGE_SIZE_LESS_THAN_256KB
1051 depends on !PAGE_SIZE_256KB
1053 # This allows to use a set of generic functions to determine mmap base
1054 # address by giving priority to top-down scheme only if the process
1055 # is not in legacy mode (compat task, unlimited stack size or
1056 # sysctl_legacy_va_layout).
1057 # Architecture that selects this option can provide its own version of:
1059 config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT
1062 select ARCH_HAS_ELF_RANDOMIZE
1067 config HAVE_JUMP_LABEL_HACK
1070 config HAVE_NOINSTR_HACK
1073 config HAVE_NOINSTR_VALIDATION
1076 config HAVE_UACCESS_VALIDATION
1080 config HAVE_STACK_VALIDATION
1083 Architecture supports objtool compile-time frame pointer rule
1086 config HAVE_RELIABLE_STACKTRACE
1089 Architecture has either save_stack_trace_tsk_reliable() or
1090 arch_stack_walk_reliable() function which only returns a stack trace
1091 if it can guarantee the trace is reliable.
1093 config HAVE_ARCH_HASH
1097 If this is set, the architecture provides an <asm/hash.h>
1098 file which provides platform-specific implementations of some
1099 functions in <linux/hash.h> or fs/namei.c.
1101 config HAVE_ARCH_NVRAM_OPS
1110 config CLONE_BACKWARDS
1113 Architecture has tls passed as the 4th argument of clone(2),
1116 config CLONE_BACKWARDS2
1119 Architecture has the first two arguments of clone(2) swapped.
1121 config CLONE_BACKWARDS3
1124 Architecture has tls passed as the 3rd argument of clone(2),
1127 config ODD_RT_SIGACTION
1130 Architecture has unusual rt_sigaction(2) arguments
1132 config OLD_SIGSUSPEND
1135 Architecture has old sigsuspend(2) syscall, of one-argument variety
1137 config OLD_SIGSUSPEND3
1140 Even weirder antique ABI - three-argument sigsuspend(2)
1142 config OLD_SIGACTION
1145 Architecture has old sigaction(2) syscall. Nope, not the same
1146 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
1147 but fairly different variant of sigaction(2), thanks to OSF/1
1150 config COMPAT_OLD_SIGACTION
1153 config COMPAT_32BIT_TIME
1154 bool "Provide system calls for 32-bit time_t"
1155 default !64BIT || COMPAT
1157 This enables 32 bit time_t support in addition to 64 bit time_t support.
1158 This is relevant on all 32-bit architectures, and 64-bit architectures
1159 as part of compat syscall handling.
1161 config ARCH_NO_PREEMPT
1164 config ARCH_EPHEMERAL_INODES
1167 An arch should select this symbol if it doesn't keep track of inode
1168 instances on its own, but instead relies on something else (e.g. the
1169 host kernel for an UML kernel).
1171 config ARCH_SUPPORTS_RT
1174 config CPU_NO_EFFICIENT_FFS
1177 config HAVE_ARCH_VMAP_STACK
1180 An arch should select this symbol if it can support kernel stacks
1181 in vmalloc space. This means:
1183 - vmalloc space must be large enough to hold many kernel stacks.
1184 This may rule out many 32-bit architectures.
1186 - Stacks in vmalloc space need to work reliably. For example, if
1187 vmap page tables are created on demand, either this mechanism
1188 needs to work while the stack points to a virtual address with
1189 unpopulated page tables or arch code (switch_to() and switch_mm(),
1190 most likely) needs to ensure that the stack's page table entries
1191 are populated before running on a possibly unpopulated stack.
1193 - If the stack overflows into a guard page, something reasonable
1194 should happen. The definition of "reasonable" is flexible, but
1195 instantly rebooting without logging anything would be unfriendly.
1199 bool "Use a virtually-mapped stack"
1200 depends on HAVE_ARCH_VMAP_STACK
1201 depends on !KASAN || KASAN_HW_TAGS || KASAN_VMALLOC
1203 Enable this if you want the use virtually-mapped kernel stacks
1204 with guard pages. This causes kernel stack overflows to be
1205 caught immediately rather than causing difficult-to-diagnose
1208 To use this with software KASAN modes, the architecture must support
1209 backing virtual mappings with real shadow memory, and KASAN_VMALLOC
1212 config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1215 An arch should select this symbol if it can support kernel stack
1216 offset randomization with calls to add_random_kstack_offset()
1217 during syscall entry and choose_random_kstack_offset() during
1218 syscall exit. Careful removal of -fstack-protector-strong and
1219 -fstack-protector should also be applied to the entry code and
1220 closely examined, as the artificial stack bump looks like an array
1221 to the compiler, so it will attempt to add canary checks regardless
1222 of the static branch state.
1224 config RANDOMIZE_KSTACK_OFFSET
1225 bool "Support for randomizing kernel stack offset on syscall entry" if EXPERT
1227 depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1228 depends on INIT_STACK_NONE || !CC_IS_CLANG || CLANG_VERSION >= 140000
1230 The kernel stack offset can be randomized (after pt_regs) by
1231 roughly 5 bits of entropy, frustrating memory corruption
1232 attacks that depend on stack address determinism or
1233 cross-syscall address exposures.
1235 The feature is controlled via the "randomize_kstack_offset=on/off"
1236 kernel boot param, and if turned off has zero overhead due to its use
1237 of static branches (see JUMP_LABEL).
1241 config RANDOMIZE_KSTACK_OFFSET_DEFAULT
1242 bool "Default state of kernel stack offset randomization"
1243 depends on RANDOMIZE_KSTACK_OFFSET
1245 Kernel stack offset randomization is controlled by kernel boot param
1246 "randomize_kstack_offset=on/off", and this config chooses the default
1249 config ARCH_OPTIONAL_KERNEL_RWX
1252 config ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1255 config ARCH_HAS_STRICT_KERNEL_RWX
1258 config STRICT_KERNEL_RWX
1259 bool "Make kernel text and rodata read-only" if ARCH_OPTIONAL_KERNEL_RWX
1260 depends on ARCH_HAS_STRICT_KERNEL_RWX
1261 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1263 If this is set, kernel text and rodata memory will be made read-only,
1264 and non-text memory will be made non-executable. This provides
1265 protection against certain security exploits (e.g. executing the heap
1268 These features are considered standard security practice these days.
1269 You should say Y here in almost all cases.
1271 config ARCH_HAS_STRICT_MODULE_RWX
1274 config STRICT_MODULE_RWX
1275 bool "Set loadable kernel module data as NX and text as RO" if ARCH_OPTIONAL_KERNEL_RWX
1276 depends on ARCH_HAS_STRICT_MODULE_RWX && MODULES
1277 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1279 If this is set, module text and rodata memory will be made read-only,
1280 and non-text memory will be made non-executable. This provides
1281 protection against certain security exploits (e.g. writing to text)
1283 # select if the architecture provides an asm/dma-direct.h header
1284 config ARCH_HAS_PHYS_TO_DMA
1287 config HAVE_ARCH_COMPILER_H
1290 An architecture can select this if it provides an
1291 asm/compiler.h header that should be included after
1292 linux/compiler-*.h in order to override macro definitions that those
1293 headers generally provide.
1295 config HAVE_ARCH_PREL32_RELOCATIONS
1298 May be selected by an architecture if it supports place-relative
1299 32-bit relocations, both in the toolchain and in the module loader,
1300 in which case relative references can be used in special sections
1301 for PCI fixup, initcalls etc which are only half the size on 64 bit
1302 architectures, and don't require runtime relocation on relocatable
1305 config ARCH_USE_MEMREMAP_PROT
1308 config LOCK_EVENT_COUNTS
1309 bool "Locking event counts collection"
1312 Enable light-weight counting of various locking related events
1313 in the system with minimal performance impact. This reduces
1314 the chance of application behavior change because of timing
1315 differences. The counts are reported via debugfs.
1317 # Select if the architecture has support for applying RELR relocations.
1318 config ARCH_HAS_RELR
1322 bool "Use RELR relocation packing"
1323 depends on ARCH_HAS_RELR && TOOLS_SUPPORT_RELR
1326 Store the kernel's dynamic relocations in the RELR relocation packing
1327 format. Requires a compatible linker (LLD supports this feature), as
1328 well as compatible NM and OBJCOPY utilities (llvm-nm and llvm-objcopy
1331 config ARCH_HAS_MEM_ENCRYPT
1334 config ARCH_HAS_CC_PLATFORM
1337 config HAVE_SPARSE_SYSCALL_NR
1340 An architecture should select this if its syscall numbering is sparse
1341 to save space. For example, MIPS architecture has a syscall array with
1342 entries at 4000, 5000 and 6000 locations. This option turns on syscall
1343 related optimizations for a given architecture.
1345 config ARCH_HAS_VDSO_DATA
1348 config HAVE_STATIC_CALL
1351 config HAVE_STATIC_CALL_INLINE
1353 depends on HAVE_STATIC_CALL
1356 config HAVE_PREEMPT_DYNAMIC
1359 config HAVE_PREEMPT_DYNAMIC_CALL
1361 depends on HAVE_STATIC_CALL
1362 select HAVE_PREEMPT_DYNAMIC
1364 An architecture should select this if it can handle the preemption
1365 model being selected at boot time using static calls.
1367 Where an architecture selects HAVE_STATIC_CALL_INLINE, any call to a
1368 preemption function will be patched directly.
1370 Where an architecture does not select HAVE_STATIC_CALL_INLINE, any
1371 call to a preemption function will go through a trampoline, and the
1372 trampoline will be patched.
1374 It is strongly advised to support inline static call to avoid any
1377 config HAVE_PREEMPT_DYNAMIC_KEY
1379 depends on HAVE_ARCH_JUMP_LABEL
1380 select HAVE_PREEMPT_DYNAMIC
1382 An architecture should select this if it can handle the preemption
1383 model being selected at boot time using static keys.
1385 Each preemption function will be given an early return based on a
1386 static key. This should have slightly lower overhead than non-inline
1387 static calls, as this effectively inlines each trampoline into the
1388 start of its callee. This may avoid redundant work, and may
1389 integrate better with CFI schemes.
1391 This will have greater overhead than using inline static calls as
1392 the call to the preemption function cannot be entirely elided.
1394 config ARCH_WANT_LD_ORPHAN_WARN
1397 An arch should select this symbol once all linker sections are explicitly
1398 included, size-asserted, or discarded in the linker scripts. This is
1399 important because we never want expected sections to be placed heuristically
1400 by the linker, since the locations of such sections can change between linker
1403 config HAVE_ARCH_PFN_VALID
1406 config ARCH_SUPPORTS_DEBUG_PAGEALLOC
1409 config ARCH_SUPPORTS_PAGE_TABLE_CHECK
1412 config ARCH_SPLIT_ARG64
1415 If a 32-bit architecture requires 64-bit arguments to be split into
1416 pairs of 32-bit arguments, select this option.
1418 config ARCH_HAS_ELFCORE_COMPAT
1421 config ARCH_HAS_PARANOID_L1D_FLUSH
1424 config ARCH_HAVE_TRACE_MMIO_ACCESS
1427 config DYNAMIC_SIGFRAME
1430 # Select, if arch has a named attribute group bound to NUMA device nodes.
1431 config HAVE_ARCH_NODE_DEV_GROUP
1434 config ARCH_HAS_NONLEAF_PMD_YOUNG
1437 Architectures that select this option are capable of setting the
1438 accessed bit in non-leaf PMD entries when using them as part of linear
1439 address translations. Page table walkers that clear the accessed bit
1440 may use this capability to reduce their search space.
1442 source "kernel/gcov/Kconfig"
1444 source "scripts/gcc-plugins/Kconfig"