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 menu "General architecture-dependent options"
36 depends on HAVE_KPROBES
39 Kprobes allows you to trap at almost any kernel address and
40 execute a callback function. register_kprobe() establishes
41 a probepoint and specifies the callback. Kprobes is useful
42 for kernel debugging, non-intrusive instrumentation and testing.
46 bool "Optimize very unlikely/likely branches"
47 depends on HAVE_ARCH_JUMP_LABEL
48 depends on CC_HAS_ASM_GOTO
50 This option enables a transparent branch optimization that
51 makes certain almost-always-true or almost-always-false branch
52 conditions even cheaper to execute within the kernel.
54 Certain performance-sensitive kernel code, such as trace points,
55 scheduler functionality, networking code and KVM have such
56 branches and include support for this optimization technique.
58 If it is detected that the compiler has support for "asm goto",
59 the kernel will compile such branches with just a nop
60 instruction. When the condition flag is toggled to true, the
61 nop will be converted to a jump instruction to execute the
62 conditional block of instructions.
64 This technique lowers overhead and stress on the branch prediction
65 of the processor and generally makes the kernel faster. The update
66 of the condition is slower, but those are always very rare.
68 ( On 32-bit x86, the necessary options added to the compiler
69 flags may increase the size of the kernel slightly. )
71 config STATIC_KEYS_SELFTEST
72 bool "Static key selftest"
75 Boot time self-test of the branch patching code.
77 config STATIC_CALL_SELFTEST
78 bool "Static call selftest"
79 depends on HAVE_STATIC_CALL
81 Boot time self-test of the call patching code.
85 depends on KPROBES && HAVE_OPTPROBES
86 select TASKS_RCU if PREEMPTION
88 config KPROBES_ON_FTRACE
90 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
91 depends on DYNAMIC_FTRACE_WITH_REGS
93 If function tracer is enabled and the arch supports full
94 passing of pt_regs to function tracing, then kprobes can
95 optimize on top of function tracing.
99 depends on ARCH_SUPPORTS_UPROBES
101 Uprobes is the user-space counterpart to kprobes: they
102 enable instrumentation applications (such as 'perf probe')
103 to establish unintrusive probes in user-space binaries and
104 libraries, by executing handler functions when the probes
105 are hit by user-space applications.
107 ( These probes come in the form of single-byte breakpoints,
108 managed by the kernel and kept transparent to the probed
111 config HAVE_64BIT_ALIGNED_ACCESS
112 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
114 Some architectures require 64 bit accesses to be 64 bit
115 aligned, which also requires structs containing 64 bit values
116 to be 64 bit aligned too. This includes some 32 bit
117 architectures which can do 64 bit accesses, as well as 64 bit
118 architectures without unaligned access.
120 This symbol should be selected by an architecture if 64 bit
121 accesses are required to be 64 bit aligned in this way even
122 though it is not a 64 bit architecture.
124 See Documentation/core-api/unaligned-memory-access.rst for
125 more information on the topic of unaligned memory accesses.
127 config HAVE_EFFICIENT_UNALIGNED_ACCESS
130 Some architectures are unable to perform unaligned accesses
131 without the use of get_unaligned/put_unaligned. Others are
132 unable to perform such accesses efficiently (e.g. trap on
133 unaligned access and require fixing it up in the exception
136 This symbol should be selected by an architecture if it can
137 perform unaligned accesses efficiently to allow different
138 code paths to be selected for these cases. Some network
139 drivers, for example, could opt to not fix up alignment
140 problems with received packets if doing so would not help
143 See Documentation/core-api/unaligned-memory-access.rst for more
144 information on the topic of unaligned memory accesses.
146 config ARCH_USE_BUILTIN_BSWAP
149 Modern versions of GCC (since 4.4) have builtin functions
150 for handling byte-swapping. Using these, instead of the old
151 inline assembler that the architecture code provides in the
152 __arch_bswapXX() macros, allows the compiler to see what's
153 happening and offers more opportunity for optimisation. In
154 particular, the compiler will be able to combine the byteswap
155 with a nearby load or store and use load-and-swap or
156 store-and-swap instructions if the architecture has them. It
157 should almost *never* result in code which is worse than the
158 hand-coded assembler in <asm/swab.h>. But just in case it
159 does, the use of the builtins is optional.
161 Any architecture with load-and-swap or store-and-swap
162 instructions should set this. And it shouldn't hurt to set it
163 on architectures that don't have such instructions.
167 depends on KPROBES && (HAVE_KRETPROBES || HAVE_RETHOOK)
169 config KRETPROBE_ON_RETHOOK
171 depends on HAVE_RETHOOK
172 depends on KRETPROBES
175 config USER_RETURN_NOTIFIER
177 depends on HAVE_USER_RETURN_NOTIFIER
179 Provide a kernel-internal notification when a cpu is about to
182 config HAVE_IOREMAP_PROT
188 config HAVE_KRETPROBES
191 config HAVE_OPTPROBES
194 config HAVE_KPROBES_ON_FTRACE
197 config ARCH_CORRECT_STACKTRACE_ON_KRETPROBE
200 Since kretprobes modifies return address on the stack, the
201 stacktrace may see the kretprobe trampoline address instead
202 of correct one. If the architecture stacktrace code and
203 unwinder can adjust such entries, select this configuration.
205 config HAVE_FUNCTION_ERROR_INJECTION
211 config HAVE_FUNCTION_DESCRIPTORS
214 config TRACE_IRQFLAGS_SUPPORT
218 # An arch should select this if it provides all these things:
220 # task_pt_regs() in asm/processor.h or asm/ptrace.h
221 # arch_has_single_step() if there is hardware single-step support
222 # arch_has_block_step() if there is hardware block-step support
223 # asm/syscall.h supplying asm-generic/syscall.h interface
224 # linux/regset.h user_regset interfaces
225 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
226 # TIF_SYSCALL_TRACE calls tracehook_report_syscall_{entry,exit}
227 # TIF_NOTIFY_RESUME calls tracehook_notify_resume()
228 # signal delivery calls tracehook_signal_handler()
230 config HAVE_ARCH_TRACEHOOK
233 config HAVE_DMA_CONTIGUOUS
236 config GENERIC_SMP_IDLE_THREAD
239 config GENERIC_IDLE_POLL_SETUP
242 config ARCH_HAS_FORTIFY_SOURCE
245 An architecture should select this when it can successfully
246 build and run with CONFIG_FORTIFY_SOURCE.
249 # Select if the arch provides a historic keepinit alias for the retain_initrd
250 # command line option
252 config ARCH_HAS_KEEPINITRD
255 # Select if arch has all set_memory_ro/rw/x/nx() functions in asm/cacheflush.h
256 config ARCH_HAS_SET_MEMORY
259 # Select if arch has all set_direct_map_invalid/default() functions
260 config ARCH_HAS_SET_DIRECT_MAP
264 # Select if the architecture provides the arch_dma_set_uncached symbol to
265 # either provide an uncached segment alias for a DMA allocation, or
266 # to remap the page tables in place.
268 config ARCH_HAS_DMA_SET_UNCACHED
272 # Select if the architectures provides the arch_dma_clear_uncached symbol
273 # to undo an in-place page table remap for uncached access.
275 config ARCH_HAS_DMA_CLEAR_UNCACHED
278 # Select if arch init_task must go in the __init_task_data section
279 config ARCH_TASK_STRUCT_ON_STACK
282 # Select if arch has its private alloc_task_struct() function
283 config ARCH_TASK_STRUCT_ALLOCATOR
286 config HAVE_ARCH_THREAD_STRUCT_WHITELIST
288 depends on !ARCH_TASK_STRUCT_ALLOCATOR
290 An architecture should select this to provide hardened usercopy
291 knowledge about what region of the thread_struct should be
292 whitelisted for copying to userspace. Normally this is only the
293 FPU registers. Specifically, arch_thread_struct_whitelist()
294 should be implemented. Without this, the entire thread_struct
295 field in task_struct will be left whitelisted.
297 # Select if arch has its private alloc_thread_stack() function
298 config ARCH_THREAD_STACK_ALLOCATOR
301 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
302 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
305 config ARCH_WANTS_NO_INSTR
308 An architecture should select this if the noinstr macro is being used on
309 functions to denote that the toolchain should avoid instrumenting such
310 functions and is required for correctness.
312 config ARCH_32BIT_OFF_T
316 All new 32-bit architectures should have 64-bit off_t type on
317 userspace side which corresponds to the loff_t kernel type. This
318 is the requirement for modern ABIs. Some existing architectures
319 still support 32-bit off_t. This option is enabled for all such
320 architectures explicitly.
322 # Selected by 64 bit architectures which have a 32 bit f_tinode in struct ustat
323 config ARCH_32BIT_USTAT_F_TINODE
326 config HAVE_ASM_MODVERSIONS
329 This symbol should be selected by an architecture if it provides
330 <asm/asm-prototypes.h> to support the module versioning for symbols
331 exported from assembly code.
333 config HAVE_REGS_AND_STACK_ACCESS_API
336 This symbol should be selected by an architecture if it supports
337 the API needed to access registers and stack entries from pt_regs,
338 declared in asm/ptrace.h
339 For example the kprobes-based event tracer needs this API.
343 depends on HAVE_REGS_AND_STACK_ACCESS_API
345 This symbol should be selected by an architecture if it
346 supports an implementation of restartable sequences.
348 config HAVE_FUNCTION_ARG_ACCESS_API
351 This symbol should be selected by an architecture if it supports
352 the API needed to access function arguments from pt_regs,
353 declared in asm/ptrace.h
355 config HAVE_HW_BREAKPOINT
357 depends on PERF_EVENTS
359 config HAVE_MIXED_BREAKPOINTS_REGS
361 depends on HAVE_HW_BREAKPOINT
363 Depending on the arch implementation of hardware breakpoints,
364 some of them have separate registers for data and instruction
365 breakpoints addresses, others have mixed registers to store
366 them but define the access type in a control register.
367 Select this option if your arch implements breakpoints under the
370 config HAVE_USER_RETURN_NOTIFIER
373 config HAVE_PERF_EVENTS_NMI
376 System hardware can generate an NMI using the perf event
377 subsystem. Also has support for calculating CPU cycle events
378 to determine how many clock cycles in a given period.
380 config HAVE_HARDLOCKUP_DETECTOR_PERF
382 depends on HAVE_PERF_EVENTS_NMI
384 The arch chooses to use the generic perf-NMI-based hardlockup
385 detector. Must define HAVE_PERF_EVENTS_NMI.
387 config HAVE_NMI_WATCHDOG
391 The arch provides a low level NMI watchdog. It provides
392 asm/nmi.h, and defines its own arch_touch_nmi_watchdog().
394 config HAVE_HARDLOCKUP_DETECTOR_ARCH
396 select HAVE_NMI_WATCHDOG
398 The arch chooses to provide its own hardlockup detector, which is
399 a superset of the HAVE_NMI_WATCHDOG. It also conforms to config
400 interfaces and parameters provided by hardlockup detector subsystem.
402 config HAVE_PERF_REGS
405 Support selective register dumps for perf events. This includes
406 bit-mapping of each registers and a unique architecture id.
408 config HAVE_PERF_USER_STACK_DUMP
411 Support user stack dumps for perf event samples. This needs
412 access to the user stack pointer which is not unified across
415 config HAVE_ARCH_JUMP_LABEL
418 config HAVE_ARCH_JUMP_LABEL_RELATIVE
421 config MMU_GATHER_TABLE_FREE
424 config MMU_GATHER_RCU_TABLE_FREE
426 select MMU_GATHER_TABLE_FREE
428 config MMU_GATHER_PAGE_SIZE
431 config MMU_GATHER_NO_RANGE
434 config MMU_GATHER_NO_GATHER
436 depends on MMU_GATHER_TABLE_FREE
438 config ARCH_WANT_IRQS_OFF_ACTIVATE_MM
441 Temporary select until all architectures can be converted to have
442 irqs disabled over activate_mm. Architectures that do IPI based TLB
443 shootdowns should enable this.
445 config ARCH_HAVE_NMI_SAFE_CMPXCHG
448 config HAVE_ALIGNED_STRUCT_PAGE
451 This makes sure that struct pages are double word aligned and that
452 e.g. the SLUB allocator can perform double word atomic operations
453 on a struct page for better performance. However selecting this
454 might increase the size of a struct page by a word.
456 config HAVE_CMPXCHG_LOCAL
459 config HAVE_CMPXCHG_DOUBLE
462 config ARCH_WEAK_RELEASE_ACQUIRE
465 config ARCH_WANT_IPC_PARSE_VERSION
468 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
471 config ARCH_WANT_OLD_COMPAT_IPC
472 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
475 config HAVE_ARCH_SECCOMP
478 An arch should select this symbol to support seccomp mode 1 (the fixed
479 syscall policy), and must provide an overrides for __NR_seccomp_sigreturn,
480 and compat syscalls if the asm-generic/seccomp.h defaults need adjustment:
481 - __NR_seccomp_read_32
482 - __NR_seccomp_write_32
483 - __NR_seccomp_exit_32
484 - __NR_seccomp_sigreturn_32
486 config HAVE_ARCH_SECCOMP_FILTER
488 select HAVE_ARCH_SECCOMP
490 An arch should select this symbol if it provides all of these things:
491 - all the requirements for HAVE_ARCH_SECCOMP
493 - syscall_get_arguments()
495 - syscall_set_return_value()
496 - SIGSYS siginfo_t support
497 - secure_computing is called from a ptrace_event()-safe context
498 - secure_computing return value is checked and a return value of -1
499 results in the system call being skipped immediately.
500 - seccomp syscall wired up
501 - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE,
502 SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If
503 COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too.
506 prompt "Enable seccomp to safely execute untrusted bytecode"
508 depends on HAVE_ARCH_SECCOMP
510 This kernel feature is useful for number crunching applications
511 that may need to handle untrusted bytecode during their
512 execution. By using pipes or other transports made available
513 to the process as file descriptors supporting the read/write
514 syscalls, it's possible to isolate those applications in their
515 own address space using seccomp. Once seccomp is enabled via
516 prctl(PR_SET_SECCOMP) or the seccomp() syscall, it cannot be
517 disabled and the task is only allowed to execute a few safe
518 syscalls defined by each seccomp mode.
522 config SECCOMP_FILTER
524 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
526 Enable tasks to build secure computing environments defined
527 in terms of Berkeley Packet Filter programs which implement
528 task-defined system call filtering polices.
530 See Documentation/userspace-api/seccomp_filter.rst for details.
532 config SECCOMP_CACHE_DEBUG
533 bool "Show seccomp filter cache status in /proc/pid/seccomp_cache"
534 depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR
537 This enables the /proc/pid/seccomp_cache interface to monitor
538 seccomp cache data. The file format is subject to change. Reading
539 the file requires CAP_SYS_ADMIN.
541 This option is for debugging only. Enabling presents the risk that
542 an adversary may be able to infer the seccomp filter logic.
546 config HAVE_ARCH_STACKLEAK
549 An architecture should select this if it has the code which
550 fills the used part of the kernel stack with the STACKLEAK_POISON
551 value before returning from system calls.
553 config HAVE_STACKPROTECTOR
556 An arch should select this symbol if:
557 - it has implemented a stack canary (e.g. __stack_chk_guard)
559 config STACKPROTECTOR
560 bool "Stack Protector buffer overflow detection"
561 depends on HAVE_STACKPROTECTOR
562 depends on $(cc-option,-fstack-protector)
565 This option turns on the "stack-protector" GCC feature. This
566 feature puts, at the beginning of functions, a canary value on
567 the stack just before the return address, and validates
568 the value just before actually returning. Stack based buffer
569 overflows (that need to overwrite this return address) now also
570 overwrite the canary, which gets detected and the attack is then
571 neutralized via a kernel panic.
573 Functions will have the stack-protector canary logic added if they
574 have an 8-byte or larger character array on the stack.
576 This feature requires gcc version 4.2 or above, or a distribution
577 gcc with the feature backported ("-fstack-protector").
579 On an x86 "defconfig" build, this feature adds canary checks to
580 about 3% of all kernel functions, which increases kernel code size
583 config STACKPROTECTOR_STRONG
584 bool "Strong Stack Protector"
585 depends on STACKPROTECTOR
586 depends on $(cc-option,-fstack-protector-strong)
589 Functions will have the stack-protector canary logic added in any
590 of the following conditions:
592 - local variable's address used as part of the right hand side of an
593 assignment or function argument
594 - local variable is an array (or union containing an array),
595 regardless of array type or length
596 - uses register local variables
598 This feature requires gcc version 4.9 or above, or a distribution
599 gcc with the feature backported ("-fstack-protector-strong").
601 On an x86 "defconfig" build, this feature adds canary checks to
602 about 20% of all kernel functions, which increases the kernel code
605 config ARCH_SUPPORTS_SHADOW_CALL_STACK
608 An architecture should select this if it supports the compiler's
609 Shadow Call Stack and implements runtime support for shadow stack
612 config SHADOW_CALL_STACK
613 bool "Shadow Call Stack"
614 depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
615 depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
617 This option enables the compiler's Shadow Call Stack, which
618 uses a shadow stack to protect function return addresses from
619 being overwritten by an attacker. More information can be found
620 in the compiler's documentation:
622 - Clang: https://clang.llvm.org/docs/ShadowCallStack.html
623 - GCC: https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html#Instrumentation-Options
625 Note that security guarantees in the kernel differ from the
626 ones documented for user space. The kernel must store addresses
627 of shadow stacks in memory, which means an attacker capable of
628 reading and writing arbitrary memory may be able to locate them
629 and hijack control flow by modifying the stacks.
634 Selected if the kernel will be built using the compiler's LTO feature.
640 Selected if the kernel will be built using Clang's LTO feature.
642 config ARCH_SUPPORTS_LTO_CLANG
645 An architecture should select this option if it supports:
646 - compiling with Clang,
647 - compiling inline assembly with Clang's integrated assembler,
648 - and linking with LLD.
650 config ARCH_SUPPORTS_LTO_CLANG_THIN
653 An architecture should select this option if it can support Clang's
658 depends on CC_IS_CLANG && LD_IS_LLD && AS_IS_LLVM
659 depends on $(success,$(NM) --help | head -n 1 | grep -qi llvm)
660 depends on $(success,$(AR) --help | head -n 1 | grep -qi llvm)
661 depends on ARCH_SUPPORTS_LTO_CLANG
662 depends on !FTRACE_MCOUNT_USE_RECORDMCOUNT
663 depends on !KASAN || KASAN_HW_TAGS
664 depends on !GCOV_KERNEL
666 The compiler and Kconfig options support building with Clang's
670 prompt "Link Time Optimization (LTO)"
673 This option enables Link Time Optimization (LTO), which allows the
674 compiler to optimize binaries globally.
676 If unsure, select LTO_NONE. Note that LTO is very resource-intensive
677 so it's disabled by default.
682 Build the kernel normally, without Link Time Optimization (LTO).
684 config LTO_CLANG_FULL
685 bool "Clang Full LTO (EXPERIMENTAL)"
686 depends on HAS_LTO_CLANG
687 depends on !COMPILE_TEST
690 This option enables Clang's full Link Time Optimization (LTO), which
691 allows the compiler to optimize the kernel globally. If you enable
692 this option, the compiler generates LLVM bitcode instead of ELF
693 object files, and the actual compilation from bitcode happens at
694 the LTO link step, which may take several minutes depending on the
695 kernel configuration. More information can be found from LLVM's
698 https://llvm.org/docs/LinkTimeOptimization.html
700 During link time, this option can use a large amount of RAM, and
701 may take much longer than the ThinLTO option.
703 config LTO_CLANG_THIN
704 bool "Clang ThinLTO (EXPERIMENTAL)"
705 depends on HAS_LTO_CLANG && ARCH_SUPPORTS_LTO_CLANG_THIN
708 This option enables Clang's ThinLTO, which allows for parallel
709 optimization and faster incremental compiles compared to the
710 CONFIG_LTO_CLANG_FULL option. More information can be found
711 from Clang's documentation:
713 https://clang.llvm.org/docs/ThinLTO.html
718 config ARCH_SUPPORTS_CFI_CLANG
721 An architecture should select this option if it can support Clang's
722 Control-Flow Integrity (CFI) checking.
725 bool "Use Clang's Control Flow Integrity (CFI)"
726 depends on LTO_CLANG && ARCH_SUPPORTS_CFI_CLANG
728 # - https://bugs.llvm.org/show_bug.cgi?id=46258
729 # - https://bugs.llvm.org/show_bug.cgi?id=47479
730 depends on CLANG_VERSION >= 120000
733 This option enables Clang’s forward-edge Control Flow Integrity
734 (CFI) checking, where the compiler injects a runtime check to each
735 indirect function call to ensure the target is a valid function with
736 the correct static type. This restricts possible call targets and
737 makes it more difficult for an attacker to exploit bugs that allow
738 the modification of stored function pointers. More information can be
739 found from Clang's documentation:
741 https://clang.llvm.org/docs/ControlFlowIntegrity.html
743 config CFI_CLANG_SHADOW
744 bool "Use CFI shadow to speed up cross-module checks"
746 depends on CFI_CLANG && MODULES
748 If you select this option, the kernel builds a fast look-up table of
749 CFI check functions in loaded modules to reduce performance overhead.
753 config CFI_PERMISSIVE
754 bool "Use CFI in permissive mode"
757 When selected, Control Flow Integrity (CFI) violations result in a
758 warning instead of a kernel panic. This option should only be used
759 for finding indirect call type mismatches during development.
763 config HAVE_ARCH_WITHIN_STACK_FRAMES
766 An architecture should select this if it can walk the kernel stack
767 frames to determine if an object is part of either the arguments
768 or local variables (i.e. that it excludes saved return addresses,
769 and similar) by implementing an inline arch_within_stack_frames(),
770 which is used by CONFIG_HARDENED_USERCOPY.
772 config HAVE_CONTEXT_TRACKING
775 Provide kernel/user boundaries probes necessary for subsystems
776 that need it, such as userspace RCU extended quiescent state.
777 Syscalls need to be wrapped inside user_exit()-user_enter(), either
778 optimized behind static key or through the slow path using TIF_NOHZ
779 flag. Exceptions handlers must be wrapped as well. Irqs are already
780 protected inside rcu_irq_enter/rcu_irq_exit() but preemption or signal
781 handling on irq exit still need to be protected.
783 config HAVE_CONTEXT_TRACKING_OFFSTACK
786 Architecture neither relies on exception_enter()/exception_exit()
787 nor on schedule_user(). Also preempt_schedule_notrace() and
788 preempt_schedule_irq() can't be called in a preemptible section
789 while context tracking is CONTEXT_USER. This feature reflects a sane
790 entry implementation where the following requirements are met on
791 critical entry code, ie: before user_exit() or after user_enter():
793 - Critical entry code isn't preemptible (or better yet:
795 - No use of RCU read side critical sections, unless rcu_nmi_enter()
797 - No use of instrumentation, unless instrumentation_begin() got
803 Arch relies on TIF_NOHZ and syscall slow path to implement context
804 tracking calls to user_enter()/user_exit().
806 config HAVE_VIRT_CPU_ACCOUNTING
809 config HAVE_VIRT_CPU_ACCOUNTING_IDLE
812 Architecture has its own way to account idle CPU time and therefore
813 doesn't implement vtime_account_idle().
815 config ARCH_HAS_SCALED_CPUTIME
818 config HAVE_VIRT_CPU_ACCOUNTING_GEN
822 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
823 Before enabling this option, arch code must be audited
824 to ensure there are no races in concurrent read/write of
825 cputime_t. For example, reading/writing 64-bit cputime_t on
826 some 32-bit arches may require multiple accesses, so proper
827 locking is needed to protect against concurrent accesses.
829 config HAVE_IRQ_TIME_ACCOUNTING
832 Archs need to ensure they use a high enough resolution clock to
833 support irq time accounting and then call enable_sched_clock_irqtime().
838 Architectures that select this are able to move page tables at the
839 PUD level. If there are only 3 page table levels, the move effectively
840 happens at the PGD level.
845 Archs that select this are able to move page tables at the PMD level.
847 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
850 config HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
853 config HAVE_ARCH_HUGE_VMAP
857 # Archs that select this would be capable of PMD-sized vmaps (i.e.,
858 # arch_vmap_pmd_supported() returns true), and they must make no assumptions
859 # that vmalloc memory is mapped with PAGE_SIZE ptes. The VM_NO_HUGE_VMAP flag
860 # can be used to prohibit arch-specific allocations from using hugepages to
861 # help with this (e.g., modules may require it).
863 config HAVE_ARCH_HUGE_VMALLOC
864 depends on HAVE_ARCH_HUGE_VMAP
867 config ARCH_WANT_HUGE_PMD_SHARE
870 config HAVE_ARCH_SOFT_DIRTY
873 config HAVE_MOD_ARCH_SPECIFIC
876 The arch uses struct mod_arch_specific to store data. Many arches
877 just need a simple module loader without arch specific data - those
878 should not enable this.
880 config MODULES_USE_ELF_RELA
883 Modules only use ELF RELA relocations. Modules with ELF REL
884 relocations will give an error.
886 config MODULES_USE_ELF_REL
889 Modules only use ELF REL relocations. Modules with ELF RELA
890 relocations will give an error.
892 config HAVE_IRQ_EXIT_ON_IRQ_STACK
895 Architecture doesn't only execute the irq handler on the irq stack
896 but also irq_exit(). This way we can process softirqs on this irq
897 stack instead of switching to a new one when we call __do_softirq()
898 in the end of an hardirq.
899 This spares a stack switch and improves cache usage on softirq
902 config HAVE_SOFTIRQ_ON_OWN_STACK
905 Architecture provides a function to run __do_softirq() on a
908 config ALTERNATE_USER_ADDRESS_SPACE
911 Architectures set this when the CPU uses separate address
912 spaces for kernel and user space pointers. In this case, the
913 access_ok() check on a __user pointer is skipped.
915 config PGTABLE_LEVELS
919 config ARCH_HAS_ELF_RANDOMIZE
922 An architecture supports choosing randomized locations for
923 stack, mmap, brk, and ET_DYN. Defined functions:
925 - arch_randomize_brk()
927 config HAVE_ARCH_MMAP_RND_BITS
930 An arch should select this symbol if it supports setting a variable
931 number of bits for use in establishing the base address for mmap
932 allocations, has MMU enabled and provides values for both:
933 - ARCH_MMAP_RND_BITS_MIN
934 - ARCH_MMAP_RND_BITS_MAX
936 config HAVE_EXIT_THREAD
939 An architecture implements exit_thread.
941 config ARCH_MMAP_RND_BITS_MIN
944 config ARCH_MMAP_RND_BITS_MAX
947 config ARCH_MMAP_RND_BITS_DEFAULT
950 config ARCH_MMAP_RND_BITS
951 int "Number of bits to use for ASLR of mmap base address" if EXPERT
952 range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX
953 default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT
954 default ARCH_MMAP_RND_BITS_MIN
955 depends on HAVE_ARCH_MMAP_RND_BITS
957 This value can be used to select the number of bits to use to
958 determine the random offset to the base address of vma regions
959 resulting from mmap allocations. This value will be bounded
960 by the architecture's minimum and maximum supported values.
962 This value can be changed after boot using the
963 /proc/sys/vm/mmap_rnd_bits tunable
965 config HAVE_ARCH_MMAP_RND_COMPAT_BITS
968 An arch should select this symbol if it supports running applications
969 in compatibility mode, supports setting a variable number of bits for
970 use in establishing the base address for mmap allocations, has MMU
971 enabled and provides values for both:
972 - ARCH_MMAP_RND_COMPAT_BITS_MIN
973 - ARCH_MMAP_RND_COMPAT_BITS_MAX
975 config ARCH_MMAP_RND_COMPAT_BITS_MIN
978 config ARCH_MMAP_RND_COMPAT_BITS_MAX
981 config ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
984 config ARCH_MMAP_RND_COMPAT_BITS
985 int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT
986 range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX
987 default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
988 default ARCH_MMAP_RND_COMPAT_BITS_MIN
989 depends on HAVE_ARCH_MMAP_RND_COMPAT_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 for compatible applications This
994 value will be bounded by the architecture's minimum and maximum
997 This value can be changed after boot using the
998 /proc/sys/vm/mmap_rnd_compat_bits tunable
1000 config HAVE_ARCH_COMPAT_MMAP_BASES
1003 This allows 64bit applications to invoke 32-bit mmap() syscall
1004 and vice-versa 32-bit applications to call 64-bit mmap().
1005 Required for applications doing different bitness syscalls.
1007 config PAGE_SIZE_LESS_THAN_64KB
1009 depends on !ARM64_64K_PAGES
1010 depends on !IA64_PAGE_SIZE_64KB
1011 depends on !PAGE_SIZE_64KB
1012 depends on !PARISC_PAGE_SIZE_64KB
1013 depends on !PPC_64K_PAGES
1014 depends on PAGE_SIZE_LESS_THAN_256KB
1016 config PAGE_SIZE_LESS_THAN_256KB
1018 depends on !PPC_256K_PAGES
1019 depends on !PAGE_SIZE_256KB
1021 # This allows to use a set of generic functions to determine mmap base
1022 # address by giving priority to top-down scheme only if the process
1023 # is not in legacy mode (compat task, unlimited stack size or
1024 # sysctl_legacy_va_layout).
1025 # Architecture that selects this option can provide its own version of:
1027 config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT
1030 select ARCH_HAS_ELF_RANDOMIZE
1032 config HAVE_STACK_VALIDATION
1035 Architecture supports the 'objtool check' host tool command, which
1036 performs compile-time stack metadata validation.
1038 config HAVE_RELIABLE_STACKTRACE
1041 Architecture has either save_stack_trace_tsk_reliable() or
1042 arch_stack_walk_reliable() function which only returns a stack trace
1043 if it can guarantee the trace is reliable.
1045 config HAVE_ARCH_HASH
1049 If this is set, the architecture provides an <asm/hash.h>
1050 file which provides platform-specific implementations of some
1051 functions in <linux/hash.h> or fs/namei.c.
1053 config HAVE_ARCH_NVRAM_OPS
1062 config CLONE_BACKWARDS
1065 Architecture has tls passed as the 4th argument of clone(2),
1068 config CLONE_BACKWARDS2
1071 Architecture has the first two arguments of clone(2) swapped.
1073 config CLONE_BACKWARDS3
1076 Architecture has tls passed as the 3rd argument of clone(2),
1079 config ODD_RT_SIGACTION
1082 Architecture has unusual rt_sigaction(2) arguments
1084 config OLD_SIGSUSPEND
1087 Architecture has old sigsuspend(2) syscall, of one-argument variety
1089 config OLD_SIGSUSPEND3
1092 Even weirder antique ABI - three-argument sigsuspend(2)
1094 config OLD_SIGACTION
1097 Architecture has old sigaction(2) syscall. Nope, not the same
1098 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
1099 but fairly different variant of sigaction(2), thanks to OSF/1
1102 config COMPAT_OLD_SIGACTION
1105 config COMPAT_32BIT_TIME
1106 bool "Provide system calls for 32-bit time_t"
1107 default !64BIT || COMPAT
1109 This enables 32 bit time_t support in addition to 64 bit time_t support.
1110 This is relevant on all 32-bit architectures, and 64-bit architectures
1111 as part of compat syscall handling.
1113 config ARCH_NO_PREEMPT
1116 config ARCH_EPHEMERAL_INODES
1119 An arch should select this symbol if it doesn't keep track of inode
1120 instances on its own, but instead relies on something else (e.g. the
1121 host kernel for an UML kernel).
1123 config ARCH_SUPPORTS_RT
1126 config CPU_NO_EFFICIENT_FFS
1129 config HAVE_ARCH_VMAP_STACK
1132 An arch should select this symbol if it can support kernel stacks
1133 in vmalloc space. This means:
1135 - vmalloc space must be large enough to hold many kernel stacks.
1136 This may rule out many 32-bit architectures.
1138 - Stacks in vmalloc space need to work reliably. For example, if
1139 vmap page tables are created on demand, either this mechanism
1140 needs to work while the stack points to a virtual address with
1141 unpopulated page tables or arch code (switch_to() and switch_mm(),
1142 most likely) needs to ensure that the stack's page table entries
1143 are populated before running on a possibly unpopulated stack.
1145 - If the stack overflows into a guard page, something reasonable
1146 should happen. The definition of "reasonable" is flexible, but
1147 instantly rebooting without logging anything would be unfriendly.
1151 bool "Use a virtually-mapped stack"
1152 depends on HAVE_ARCH_VMAP_STACK
1153 depends on !KASAN || KASAN_HW_TAGS || KASAN_VMALLOC
1155 Enable this if you want the use virtually-mapped kernel stacks
1156 with guard pages. This causes kernel stack overflows to be
1157 caught immediately rather than causing difficult-to-diagnose
1160 To use this with software KASAN modes, the architecture must support
1161 backing virtual mappings with real shadow memory, and KASAN_VMALLOC
1164 config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1167 An arch should select this symbol if it can support kernel stack
1168 offset randomization with calls to add_random_kstack_offset()
1169 during syscall entry and choose_random_kstack_offset() during
1170 syscall exit. Careful removal of -fstack-protector-strong and
1171 -fstack-protector should also be applied to the entry code and
1172 closely examined, as the artificial stack bump looks like an array
1173 to the compiler, so it will attempt to add canary checks regardless
1174 of the static branch state.
1176 config RANDOMIZE_KSTACK_OFFSET
1177 bool "Support for randomizing kernel stack offset on syscall entry" if EXPERT
1179 depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1180 depends on INIT_STACK_NONE || !CC_IS_CLANG || CLANG_VERSION >= 140000
1182 The kernel stack offset can be randomized (after pt_regs) by
1183 roughly 5 bits of entropy, frustrating memory corruption
1184 attacks that depend on stack address determinism or
1185 cross-syscall address exposures.
1187 The feature is controlled via the "randomize_kstack_offset=on/off"
1188 kernel boot param, and if turned off has zero overhead due to its use
1189 of static branches (see JUMP_LABEL).
1193 config RANDOMIZE_KSTACK_OFFSET_DEFAULT
1194 bool "Default state of kernel stack offset randomization"
1195 depends on RANDOMIZE_KSTACK_OFFSET
1197 Kernel stack offset randomization is controlled by kernel boot param
1198 "randomize_kstack_offset=on/off", and this config chooses the default
1201 config ARCH_OPTIONAL_KERNEL_RWX
1204 config ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1207 config ARCH_HAS_STRICT_KERNEL_RWX
1210 config STRICT_KERNEL_RWX
1211 bool "Make kernel text and rodata read-only" if ARCH_OPTIONAL_KERNEL_RWX
1212 depends on ARCH_HAS_STRICT_KERNEL_RWX
1213 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1215 If this is set, kernel text and rodata memory will be made read-only,
1216 and non-text memory will be made non-executable. This provides
1217 protection against certain security exploits (e.g. executing the heap
1220 These features are considered standard security practice these days.
1221 You should say Y here in almost all cases.
1223 config ARCH_HAS_STRICT_MODULE_RWX
1226 config STRICT_MODULE_RWX
1227 bool "Set loadable kernel module data as NX and text as RO" if ARCH_OPTIONAL_KERNEL_RWX
1228 depends on ARCH_HAS_STRICT_MODULE_RWX && MODULES
1229 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1231 If this is set, module text and rodata memory will be made read-only,
1232 and non-text memory will be made non-executable. This provides
1233 protection against certain security exploits (e.g. writing to text)
1235 # select if the architecture provides an asm/dma-direct.h header
1236 config ARCH_HAS_PHYS_TO_DMA
1239 config HAVE_ARCH_COMPILER_H
1242 An architecture can select this if it provides an
1243 asm/compiler.h header that should be included after
1244 linux/compiler-*.h in order to override macro definitions that those
1245 headers generally provide.
1247 config HAVE_ARCH_PREL32_RELOCATIONS
1250 May be selected by an architecture if it supports place-relative
1251 32-bit relocations, both in the toolchain and in the module loader,
1252 in which case relative references can be used in special sections
1253 for PCI fixup, initcalls etc which are only half the size on 64 bit
1254 architectures, and don't require runtime relocation on relocatable
1257 config ARCH_USE_MEMREMAP_PROT
1260 config LOCK_EVENT_COUNTS
1261 bool "Locking event counts collection"
1264 Enable light-weight counting of various locking related events
1265 in the system with minimal performance impact. This reduces
1266 the chance of application behavior change because of timing
1267 differences. The counts are reported via debugfs.
1269 # Select if the architecture has support for applying RELR relocations.
1270 config ARCH_HAS_RELR
1274 bool "Use RELR relocation packing"
1275 depends on ARCH_HAS_RELR && TOOLS_SUPPORT_RELR
1278 Store the kernel's dynamic relocations in the RELR relocation packing
1279 format. Requires a compatible linker (LLD supports this feature), as
1280 well as compatible NM and OBJCOPY utilities (llvm-nm and llvm-objcopy
1283 config ARCH_HAS_MEM_ENCRYPT
1286 config ARCH_HAS_CC_PLATFORM
1289 config HAVE_SPARSE_SYSCALL_NR
1292 An architecture should select this if its syscall numbering is sparse
1293 to save space. For example, MIPS architecture has a syscall array with
1294 entries at 4000, 5000 and 6000 locations. This option turns on syscall
1295 related optimizations for a given architecture.
1297 config ARCH_HAS_VDSO_DATA
1300 config HAVE_STATIC_CALL
1303 config HAVE_STATIC_CALL_INLINE
1305 depends on HAVE_STATIC_CALL
1307 config HAVE_PREEMPT_DYNAMIC
1310 config HAVE_PREEMPT_DYNAMIC_CALL
1312 depends on HAVE_STATIC_CALL
1313 select HAVE_PREEMPT_DYNAMIC
1315 An architecture should select this if it can handle the preemption
1316 model being selected at boot time using static calls.
1318 Where an architecture selects HAVE_STATIC_CALL_INLINE, any call to a
1319 preemption function will be patched directly.
1321 Where an architecture does not select HAVE_STATIC_CALL_INLINE, any
1322 call to a preemption function will go through a trampoline, and the
1323 trampoline will be patched.
1325 It is strongly advised to support inline static call to avoid any
1328 config HAVE_PREEMPT_DYNAMIC_KEY
1330 depends on HAVE_ARCH_JUMP_LABEL && CC_HAS_ASM_GOTO
1331 select HAVE_PREEMPT_DYNAMIC
1333 An architecture should select this if it can handle the preemption
1334 model being selected at boot time using static keys.
1336 Each preemption function will be given an early return based on a
1337 static key. This should have slightly lower overhead than non-inline
1338 static calls, as this effectively inlines each trampoline into the
1339 start of its callee. This may avoid redundant work, and may
1340 integrate better with CFI schemes.
1342 This will have greater overhead than using inline static calls as
1343 the call to the preemption function cannot be entirely elided.
1345 config ARCH_WANT_LD_ORPHAN_WARN
1348 An arch should select this symbol once all linker sections are explicitly
1349 included, size-asserted, or discarded in the linker scripts. This is
1350 important because we never want expected sections to be placed heuristically
1351 by the linker, since the locations of such sections can change between linker
1354 config HAVE_ARCH_PFN_VALID
1357 config ARCH_SUPPORTS_DEBUG_PAGEALLOC
1360 config ARCH_SUPPORTS_PAGE_TABLE_CHECK
1363 config ARCH_SPLIT_ARG64
1366 If a 32-bit architecture requires 64-bit arguments to be split into
1367 pairs of 32-bit arguments, select this option.
1369 config ARCH_HAS_ELFCORE_COMPAT
1372 config ARCH_HAS_PARANOID_L1D_FLUSH
1375 config DYNAMIC_SIGFRAME
1378 # Select, if arch has a named attribute group bound to NUMA device nodes.
1379 config HAVE_ARCH_NODE_DEV_GROUP
1382 source "kernel/gcov/Kconfig"
1384 source "scripts/gcc-plugins/Kconfig"