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"
27 config ARCH_HAS_SUBPAGE_FAULTS
30 Select if the architecture can check permissions at sub-page
31 granularity (e.g. arm64 MTE). The probe_user_*() functions
43 depends on HAVE_KPROBES
45 select TASKS_RCU if PREEMPTION
47 Kprobes allows you to trap at almost any kernel address and
48 execute a callback function. register_kprobe() establishes
49 a probepoint and specifies the callback. Kprobes is useful
50 for kernel debugging, non-intrusive instrumentation and testing.
54 bool "Optimize very unlikely/likely branches"
55 depends on HAVE_ARCH_JUMP_LABEL
56 depends on CC_HAS_ASM_GOTO
57 select OBJTOOL if HAVE_JUMP_LABEL_HACK
59 This option enables a transparent branch optimization that
60 makes certain almost-always-true or almost-always-false branch
61 conditions even cheaper to execute within the kernel.
63 Certain performance-sensitive kernel code, such as trace points,
64 scheduler functionality, networking code and KVM have such
65 branches and include support for this optimization technique.
67 If it is detected that the compiler has support for "asm goto",
68 the kernel will compile such branches with just a nop
69 instruction. When the condition flag is toggled to true, the
70 nop will be converted to a jump instruction to execute the
71 conditional block of instructions.
73 This technique lowers overhead and stress on the branch prediction
74 of the processor and generally makes the kernel faster. The update
75 of the condition is slower, but those are always very rare.
77 ( On 32-bit x86, the necessary options added to the compiler
78 flags may increase the size of the kernel slightly. )
80 config STATIC_KEYS_SELFTEST
81 bool "Static key selftest"
84 Boot time self-test of the branch patching code.
86 config STATIC_CALL_SELFTEST
87 bool "Static call selftest"
88 depends on HAVE_STATIC_CALL
90 Boot time self-test of the call patching code.
94 depends on KPROBES && HAVE_OPTPROBES
95 select TASKS_RCU if PREEMPTION
97 config KPROBES_ON_FTRACE
99 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
100 depends on DYNAMIC_FTRACE_WITH_REGS
102 If function tracer is enabled and the arch supports full
103 passing of pt_regs to function tracing, then kprobes can
104 optimize on top of function tracing.
108 depends on ARCH_SUPPORTS_UPROBES
110 Uprobes is the user-space counterpart to kprobes: they
111 enable instrumentation applications (such as 'perf probe')
112 to establish unintrusive probes in user-space binaries and
113 libraries, by executing handler functions when the probes
114 are hit by user-space applications.
116 ( These probes come in the form of single-byte breakpoints,
117 managed by the kernel and kept transparent to the probed
120 config HAVE_64BIT_ALIGNED_ACCESS
121 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
123 Some architectures require 64 bit accesses to be 64 bit
124 aligned, which also requires structs containing 64 bit values
125 to be 64 bit aligned too. This includes some 32 bit
126 architectures which can do 64 bit accesses, as well as 64 bit
127 architectures without unaligned access.
129 This symbol should be selected by an architecture if 64 bit
130 accesses are required to be 64 bit aligned in this way even
131 though it is not a 64 bit architecture.
133 See Documentation/core-api/unaligned-memory-access.rst for
134 more information on the topic of unaligned memory accesses.
136 config HAVE_EFFICIENT_UNALIGNED_ACCESS
139 Some architectures are unable to perform unaligned accesses
140 without the use of get_unaligned/put_unaligned. Others are
141 unable to perform such accesses efficiently (e.g. trap on
142 unaligned access and require fixing it up in the exception
145 This symbol should be selected by an architecture if it can
146 perform unaligned accesses efficiently to allow different
147 code paths to be selected for these cases. Some network
148 drivers, for example, could opt to not fix up alignment
149 problems with received packets if doing so would not help
152 See Documentation/core-api/unaligned-memory-access.rst for more
153 information on the topic of unaligned memory accesses.
155 config ARCH_USE_BUILTIN_BSWAP
158 Modern versions of GCC (since 4.4) have builtin functions
159 for handling byte-swapping. Using these, instead of the old
160 inline assembler that the architecture code provides in the
161 __arch_bswapXX() macros, allows the compiler to see what's
162 happening and offers more opportunity for optimisation. In
163 particular, the compiler will be able to combine the byteswap
164 with a nearby load or store and use load-and-swap or
165 store-and-swap instructions if the architecture has them. It
166 should almost *never* result in code which is worse than the
167 hand-coded assembler in <asm/swab.h>. But just in case it
168 does, the use of the builtins is optional.
170 Any architecture with load-and-swap or store-and-swap
171 instructions should set this. And it shouldn't hurt to set it
172 on architectures that don't have such instructions.
176 depends on KPROBES && (HAVE_KRETPROBES || HAVE_RETHOOK)
178 config KRETPROBE_ON_RETHOOK
180 depends on HAVE_RETHOOK
181 depends on KRETPROBES
184 config USER_RETURN_NOTIFIER
186 depends on HAVE_USER_RETURN_NOTIFIER
188 Provide a kernel-internal notification when a cpu is about to
191 config HAVE_IOREMAP_PROT
197 config HAVE_KRETPROBES
200 config HAVE_OPTPROBES
203 config HAVE_KPROBES_ON_FTRACE
206 config ARCH_CORRECT_STACKTRACE_ON_KRETPROBE
209 Since kretprobes modifies return address on the stack, the
210 stacktrace may see the kretprobe trampoline address instead
211 of correct one. If the architecture stacktrace code and
212 unwinder can adjust such entries, select this configuration.
214 config HAVE_FUNCTION_ERROR_INJECTION
220 config HAVE_FUNCTION_DESCRIPTORS
223 config TRACE_IRQFLAGS_SUPPORT
227 # An arch should select this if it provides all these things:
229 # task_pt_regs() in asm/processor.h or asm/ptrace.h
230 # arch_has_single_step() if there is hardware single-step support
231 # arch_has_block_step() if there is hardware block-step support
232 # asm/syscall.h supplying asm-generic/syscall.h interface
233 # linux/regset.h user_regset interfaces
234 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
235 # TIF_SYSCALL_TRACE calls ptrace_report_syscall_{entry,exit}
236 # TIF_NOTIFY_RESUME calls resume_user_mode_work()
238 config HAVE_ARCH_TRACEHOOK
241 config HAVE_DMA_CONTIGUOUS
244 config GENERIC_SMP_IDLE_THREAD
247 config GENERIC_IDLE_POLL_SETUP
250 config ARCH_HAS_FORTIFY_SOURCE
253 An architecture should select this when it can successfully
254 build and run with CONFIG_FORTIFY_SOURCE.
257 # Select if the arch provides a historic keepinit alias for the retain_initrd
258 # command line option
260 config ARCH_HAS_KEEPINITRD
263 # Select if arch has all set_memory_ro/rw/x/nx() functions in asm/cacheflush.h
264 config ARCH_HAS_SET_MEMORY
267 # Select if arch has all set_direct_map_invalid/default() functions
268 config ARCH_HAS_SET_DIRECT_MAP
272 # Select if the architecture provides the arch_dma_set_uncached symbol to
273 # either provide an uncached segment alias for a DMA allocation, or
274 # to remap the page tables in place.
276 config ARCH_HAS_DMA_SET_UNCACHED
280 # Select if the architectures provides the arch_dma_clear_uncached symbol
281 # to undo an in-place page table remap for uncached access.
283 config ARCH_HAS_DMA_CLEAR_UNCACHED
286 # Select if arch init_task must go in the __init_task_data section
287 config ARCH_TASK_STRUCT_ON_STACK
290 # Select if arch has its private alloc_task_struct() function
291 config ARCH_TASK_STRUCT_ALLOCATOR
294 config HAVE_ARCH_THREAD_STRUCT_WHITELIST
296 depends on !ARCH_TASK_STRUCT_ALLOCATOR
298 An architecture should select this to provide hardened usercopy
299 knowledge about what region of the thread_struct should be
300 whitelisted for copying to userspace. Normally this is only the
301 FPU registers. Specifically, arch_thread_struct_whitelist()
302 should be implemented. Without this, the entire thread_struct
303 field in task_struct will be left whitelisted.
305 # Select if arch has its private alloc_thread_stack() function
306 config ARCH_THREAD_STACK_ALLOCATOR
309 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
310 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
313 config ARCH_WANTS_NO_INSTR
316 An architecture should select this if the noinstr macro is being used on
317 functions to denote that the toolchain should avoid instrumenting such
318 functions and is required for correctness.
320 config ARCH_32BIT_OFF_T
324 All new 32-bit architectures should have 64-bit off_t type on
325 userspace side which corresponds to the loff_t kernel type. This
326 is the requirement for modern ABIs. Some existing architectures
327 still support 32-bit off_t. This option is enabled for all such
328 architectures explicitly.
330 # Selected by 64 bit architectures which have a 32 bit f_tinode in struct ustat
331 config ARCH_32BIT_USTAT_F_TINODE
334 config HAVE_ASM_MODVERSIONS
337 This symbol should be selected by an architecture if it provides
338 <asm/asm-prototypes.h> to support the module versioning for symbols
339 exported from assembly code.
341 config HAVE_REGS_AND_STACK_ACCESS_API
344 This symbol should be selected by an architecture if it supports
345 the API needed to access registers and stack entries from pt_regs,
346 declared in asm/ptrace.h
347 For example the kprobes-based event tracer needs this API.
351 depends on HAVE_REGS_AND_STACK_ACCESS_API
353 This symbol should be selected by an architecture if it
354 supports an implementation of restartable sequences.
356 config HAVE_FUNCTION_ARG_ACCESS_API
359 This symbol should be selected by an architecture if it supports
360 the API needed to access function arguments from pt_regs,
361 declared in asm/ptrace.h
363 config HAVE_HW_BREAKPOINT
365 depends on PERF_EVENTS
367 config HAVE_MIXED_BREAKPOINTS_REGS
369 depends on HAVE_HW_BREAKPOINT
371 Depending on the arch implementation of hardware breakpoints,
372 some of them have separate registers for data and instruction
373 breakpoints addresses, others have mixed registers to store
374 them but define the access type in a control register.
375 Select this option if your arch implements breakpoints under the
378 config HAVE_USER_RETURN_NOTIFIER
381 config HAVE_PERF_EVENTS_NMI
384 System hardware can generate an NMI using the perf event
385 subsystem. Also has support for calculating CPU cycle events
386 to determine how many clock cycles in a given period.
388 config HAVE_HARDLOCKUP_DETECTOR_PERF
390 depends on HAVE_PERF_EVENTS_NMI
392 The arch chooses to use the generic perf-NMI-based hardlockup
393 detector. Must define HAVE_PERF_EVENTS_NMI.
395 config HAVE_NMI_WATCHDOG
399 The arch provides a low level NMI watchdog. It provides
400 asm/nmi.h, and defines its own arch_touch_nmi_watchdog().
402 config HAVE_HARDLOCKUP_DETECTOR_ARCH
404 select HAVE_NMI_WATCHDOG
406 The arch chooses to provide its own hardlockup detector, which is
407 a superset of the HAVE_NMI_WATCHDOG. It also conforms to config
408 interfaces and parameters provided by hardlockup detector subsystem.
410 config HAVE_PERF_REGS
413 Support selective register dumps for perf events. This includes
414 bit-mapping of each registers and a unique architecture id.
416 config HAVE_PERF_USER_STACK_DUMP
419 Support user stack dumps for perf event samples. This needs
420 access to the user stack pointer which is not unified across
423 config HAVE_ARCH_JUMP_LABEL
426 config HAVE_ARCH_JUMP_LABEL_RELATIVE
429 config MMU_GATHER_TABLE_FREE
432 config MMU_GATHER_RCU_TABLE_FREE
434 select MMU_GATHER_TABLE_FREE
436 config MMU_GATHER_PAGE_SIZE
439 config MMU_GATHER_NO_RANGE
442 config MMU_GATHER_NO_GATHER
444 depends on MMU_GATHER_TABLE_FREE
446 config ARCH_WANT_IRQS_OFF_ACTIVATE_MM
449 Temporary select until all architectures can be converted to have
450 irqs disabled over activate_mm. Architectures that do IPI based TLB
451 shootdowns should enable this.
453 config ARCH_HAVE_NMI_SAFE_CMPXCHG
456 config HAVE_ALIGNED_STRUCT_PAGE
459 This makes sure that struct pages are double word aligned and that
460 e.g. the SLUB allocator can perform double word atomic operations
461 on a struct page for better performance. However selecting this
462 might increase the size of a struct page by a word.
464 config HAVE_CMPXCHG_LOCAL
467 config HAVE_CMPXCHG_DOUBLE
470 config ARCH_WEAK_RELEASE_ACQUIRE
473 config ARCH_WANT_IPC_PARSE_VERSION
476 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
479 config ARCH_WANT_OLD_COMPAT_IPC
480 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
483 config HAVE_ARCH_SECCOMP
486 An arch should select this symbol to support seccomp mode 1 (the fixed
487 syscall policy), and must provide an overrides for __NR_seccomp_sigreturn,
488 and compat syscalls if the asm-generic/seccomp.h defaults need adjustment:
489 - __NR_seccomp_read_32
490 - __NR_seccomp_write_32
491 - __NR_seccomp_exit_32
492 - __NR_seccomp_sigreturn_32
494 config HAVE_ARCH_SECCOMP_FILTER
496 select HAVE_ARCH_SECCOMP
498 An arch should select this symbol if it provides all of these things:
499 - all the requirements for HAVE_ARCH_SECCOMP
501 - syscall_get_arguments()
503 - syscall_set_return_value()
504 - SIGSYS siginfo_t support
505 - secure_computing is called from a ptrace_event()-safe context
506 - secure_computing return value is checked and a return value of -1
507 results in the system call being skipped immediately.
508 - seccomp syscall wired up
509 - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE,
510 SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If
511 COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too.
514 prompt "Enable seccomp to safely execute untrusted bytecode"
516 depends on HAVE_ARCH_SECCOMP
518 This kernel feature is useful for number crunching applications
519 that may need to handle untrusted bytecode during their
520 execution. By using pipes or other transports made available
521 to the process as file descriptors supporting the read/write
522 syscalls, it's possible to isolate those applications in their
523 own address space using seccomp. Once seccomp is enabled via
524 prctl(PR_SET_SECCOMP) or the seccomp() syscall, it cannot be
525 disabled and the task is only allowed to execute a few safe
526 syscalls defined by each seccomp mode.
530 config SECCOMP_FILTER
532 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
534 Enable tasks to build secure computing environments defined
535 in terms of Berkeley Packet Filter programs which implement
536 task-defined system call filtering polices.
538 See Documentation/userspace-api/seccomp_filter.rst for details.
540 config SECCOMP_CACHE_DEBUG
541 bool "Show seccomp filter cache status in /proc/pid/seccomp_cache"
542 depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR
545 This enables the /proc/pid/seccomp_cache interface to monitor
546 seccomp cache data. The file format is subject to change. Reading
547 the file requires CAP_SYS_ADMIN.
549 This option is for debugging only. Enabling presents the risk that
550 an adversary may be able to infer the seccomp filter logic.
554 config HAVE_ARCH_STACKLEAK
557 An architecture should select this if it has the code which
558 fills the used part of the kernel stack with the STACKLEAK_POISON
559 value before returning from system calls.
561 config HAVE_STACKPROTECTOR
564 An arch should select this symbol if:
565 - it has implemented a stack canary (e.g. __stack_chk_guard)
567 config STACKPROTECTOR
568 bool "Stack Protector buffer overflow detection"
569 depends on HAVE_STACKPROTECTOR
570 depends on $(cc-option,-fstack-protector)
573 This option turns on the "stack-protector" GCC feature. This
574 feature puts, at the beginning of functions, a canary value on
575 the stack just before the return address, and validates
576 the value just before actually returning. Stack based buffer
577 overflows (that need to overwrite this return address) now also
578 overwrite the canary, which gets detected and the attack is then
579 neutralized via a kernel panic.
581 Functions will have the stack-protector canary logic added if they
582 have an 8-byte or larger character array on the stack.
584 This feature requires gcc version 4.2 or above, or a distribution
585 gcc with the feature backported ("-fstack-protector").
587 On an x86 "defconfig" build, this feature adds canary checks to
588 about 3% of all kernel functions, which increases kernel code size
591 config STACKPROTECTOR_STRONG
592 bool "Strong Stack Protector"
593 depends on STACKPROTECTOR
594 depends on $(cc-option,-fstack-protector-strong)
597 Functions will have the stack-protector canary logic added in any
598 of the following conditions:
600 - local variable's address used as part of the right hand side of an
601 assignment or function argument
602 - local variable is an array (or union containing an array),
603 regardless of array type or length
604 - uses register local variables
606 This feature requires gcc version 4.9 or above, or a distribution
607 gcc with the feature backported ("-fstack-protector-strong").
609 On an x86 "defconfig" build, this feature adds canary checks to
610 about 20% of all kernel functions, which increases the kernel code
613 config ARCH_SUPPORTS_SHADOW_CALL_STACK
616 An architecture should select this if it supports the compiler's
617 Shadow Call Stack and implements runtime support for shadow stack
620 config SHADOW_CALL_STACK
621 bool "Shadow Call Stack"
622 depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
623 depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
625 This option enables the compiler's Shadow Call Stack, which
626 uses a shadow stack to protect function return addresses from
627 being overwritten by an attacker. More information can be found
628 in the compiler's documentation:
630 - Clang: https://clang.llvm.org/docs/ShadowCallStack.html
631 - GCC: https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html#Instrumentation-Options
633 Note that security guarantees in the kernel differ from the
634 ones documented for user space. The kernel must store addresses
635 of shadow stacks in memory, which means an attacker capable of
636 reading and writing arbitrary memory may be able to locate them
637 and hijack control flow by modifying the stacks.
642 Selected if the kernel will be built using the compiler's LTO feature.
648 Selected if the kernel will be built using Clang's LTO feature.
650 config ARCH_SUPPORTS_LTO_CLANG
653 An architecture should select this option if it supports:
654 - compiling with Clang,
655 - compiling inline assembly with Clang's integrated assembler,
656 - and linking with LLD.
658 config ARCH_SUPPORTS_LTO_CLANG_THIN
661 An architecture should select this option if it can support Clang's
666 depends on CC_IS_CLANG && LD_IS_LLD && AS_IS_LLVM
667 depends on $(success,$(NM) --help | head -n 1 | grep -qi llvm)
668 depends on $(success,$(AR) --help | head -n 1 | grep -qi llvm)
669 depends on ARCH_SUPPORTS_LTO_CLANG
670 depends on !FTRACE_MCOUNT_USE_RECORDMCOUNT
671 depends on !KASAN || KASAN_HW_TAGS
672 depends on !GCOV_KERNEL
674 The compiler and Kconfig options support building with Clang's
678 prompt "Link Time Optimization (LTO)"
681 This option enables Link Time Optimization (LTO), which allows the
682 compiler to optimize binaries globally.
684 If unsure, select LTO_NONE. Note that LTO is very resource-intensive
685 so it's disabled by default.
690 Build the kernel normally, without Link Time Optimization (LTO).
692 config LTO_CLANG_FULL
693 bool "Clang Full LTO (EXPERIMENTAL)"
694 depends on HAS_LTO_CLANG
695 depends on !COMPILE_TEST
698 This option enables Clang's full Link Time Optimization (LTO), which
699 allows the compiler to optimize the kernel globally. If you enable
700 this option, the compiler generates LLVM bitcode instead of ELF
701 object files, and the actual compilation from bitcode happens at
702 the LTO link step, which may take several minutes depending on the
703 kernel configuration. More information can be found from LLVM's
706 https://llvm.org/docs/LinkTimeOptimization.html
708 During link time, this option can use a large amount of RAM, and
709 may take much longer than the ThinLTO option.
711 config LTO_CLANG_THIN
712 bool "Clang ThinLTO (EXPERIMENTAL)"
713 depends on HAS_LTO_CLANG && ARCH_SUPPORTS_LTO_CLANG_THIN
716 This option enables Clang's ThinLTO, which allows for parallel
717 optimization and faster incremental compiles compared to the
718 CONFIG_LTO_CLANG_FULL option. More information can be found
719 from Clang's documentation:
721 https://clang.llvm.org/docs/ThinLTO.html
726 config ARCH_SUPPORTS_CFI_CLANG
729 An architecture should select this option if it can support Clang's
730 Control-Flow Integrity (CFI) checking.
733 bool "Use Clang's Control Flow Integrity (CFI)"
734 depends on LTO_CLANG && ARCH_SUPPORTS_CFI_CLANG
735 depends on CLANG_VERSION >= 140000
738 This option enables Clang’s forward-edge Control Flow Integrity
739 (CFI) checking, where the compiler injects a runtime check to each
740 indirect function call to ensure the target is a valid function with
741 the correct static type. This restricts possible call targets and
742 makes it more difficult for an attacker to exploit bugs that allow
743 the modification of stored function pointers. More information can be
744 found from Clang's documentation:
746 https://clang.llvm.org/docs/ControlFlowIntegrity.html
748 config CFI_CLANG_SHADOW
749 bool "Use CFI shadow to speed up cross-module checks"
751 depends on CFI_CLANG && MODULES
753 If you select this option, the kernel builds a fast look-up table of
754 CFI check functions in loaded modules to reduce performance overhead.
758 config CFI_PERMISSIVE
759 bool "Use CFI in permissive mode"
762 When selected, Control Flow Integrity (CFI) violations result in a
763 warning instead of a kernel panic. This option should only be used
764 for finding indirect call type mismatches during development.
768 config HAVE_ARCH_WITHIN_STACK_FRAMES
771 An architecture should select this if it can walk the kernel stack
772 frames to determine if an object is part of either the arguments
773 or local variables (i.e. that it excludes saved return addresses,
774 and similar) by implementing an inline arch_within_stack_frames(),
775 which is used by CONFIG_HARDENED_USERCOPY.
777 config HAVE_CONTEXT_TRACKING
780 Provide kernel/user boundaries probes necessary for subsystems
781 that need it, such as userspace RCU extended quiescent state.
782 Syscalls need to be wrapped inside user_exit()-user_enter(), either
783 optimized behind static key or through the slow path using TIF_NOHZ
784 flag. Exceptions handlers must be wrapped as well. Irqs are already
785 protected inside rcu_irq_enter/rcu_irq_exit() but preemption or signal
786 handling on irq exit still need to be protected.
788 config HAVE_CONTEXT_TRACKING_OFFSTACK
791 Architecture neither relies on exception_enter()/exception_exit()
792 nor on schedule_user(). Also preempt_schedule_notrace() and
793 preempt_schedule_irq() can't be called in a preemptible section
794 while context tracking is CONTEXT_USER. This feature reflects a sane
795 entry implementation where the following requirements are met on
796 critical entry code, ie: before user_exit() or after user_enter():
798 - Critical entry code isn't preemptible (or better yet:
800 - No use of RCU read side critical sections, unless rcu_nmi_enter()
802 - No use of instrumentation, unless instrumentation_begin() got
808 Arch relies on TIF_NOHZ and syscall slow path to implement context
809 tracking calls to user_enter()/user_exit().
811 config HAVE_VIRT_CPU_ACCOUNTING
814 config HAVE_VIRT_CPU_ACCOUNTING_IDLE
817 Architecture has its own way to account idle CPU time and therefore
818 doesn't implement vtime_account_idle().
820 config ARCH_HAS_SCALED_CPUTIME
823 config HAVE_VIRT_CPU_ACCOUNTING_GEN
827 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
828 Before enabling this option, arch code must be audited
829 to ensure there are no races in concurrent read/write of
830 cputime_t. For example, reading/writing 64-bit cputime_t on
831 some 32-bit arches may require multiple accesses, so proper
832 locking is needed to protect against concurrent accesses.
834 config HAVE_IRQ_TIME_ACCOUNTING
837 Archs need to ensure they use a high enough resolution clock to
838 support irq time accounting and then call enable_sched_clock_irqtime().
843 Architectures that select this are able to move page tables at the
844 PUD level. If there are only 3 page table levels, the move effectively
845 happens at the PGD level.
850 Archs that select this are able to move page tables at the PMD level.
852 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
855 config HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
858 config HAVE_ARCH_HUGE_VMAP
862 # Archs that select this would be capable of PMD-sized vmaps (i.e.,
863 # arch_vmap_pmd_supported() returns true). The VM_ALLOW_HUGE_VMAP flag
864 # must be used to enable allocations to use hugepages.
866 config HAVE_ARCH_HUGE_VMALLOC
867 depends on HAVE_ARCH_HUGE_VMAP
870 config ARCH_WANT_HUGE_PMD_SHARE
873 config HAVE_ARCH_SOFT_DIRTY
876 config HAVE_MOD_ARCH_SPECIFIC
879 The arch uses struct mod_arch_specific to store data. Many arches
880 just need a simple module loader without arch specific data - those
881 should not enable this.
883 config MODULES_USE_ELF_RELA
886 Modules only use ELF RELA relocations. Modules with ELF REL
887 relocations will give an error.
889 config MODULES_USE_ELF_REL
892 Modules only use ELF REL relocations. Modules with ELF RELA
893 relocations will give an error.
895 config ARCH_WANTS_MODULES_DATA_IN_VMALLOC
898 For architectures like powerpc/32 which have constraints on module
899 allocation and need to allocate module data outside of module area.
901 config HAVE_IRQ_EXIT_ON_IRQ_STACK
904 Architecture doesn't only execute the irq handler on the irq stack
905 but also irq_exit(). This way we can process softirqs on this irq
906 stack instead of switching to a new one when we call __do_softirq()
907 in the end of an hardirq.
908 This spares a stack switch and improves cache usage on softirq
911 config HAVE_SOFTIRQ_ON_OWN_STACK
914 Architecture provides a function to run __do_softirq() on a
917 config ALTERNATE_USER_ADDRESS_SPACE
920 Architectures set this when the CPU uses separate address
921 spaces for kernel and user space pointers. In this case, the
922 access_ok() check on a __user pointer is skipped.
924 config PGTABLE_LEVELS
928 config ARCH_HAS_ELF_RANDOMIZE
931 An architecture supports choosing randomized locations for
932 stack, mmap, brk, and ET_DYN. Defined functions:
934 - arch_randomize_brk()
936 config HAVE_ARCH_MMAP_RND_BITS
939 An arch should select this symbol if it supports setting a variable
940 number of bits for use in establishing the base address for mmap
941 allocations, has MMU enabled and provides values for both:
942 - ARCH_MMAP_RND_BITS_MIN
943 - ARCH_MMAP_RND_BITS_MAX
945 config HAVE_EXIT_THREAD
948 An architecture implements exit_thread.
950 config ARCH_MMAP_RND_BITS_MIN
953 config ARCH_MMAP_RND_BITS_MAX
956 config ARCH_MMAP_RND_BITS_DEFAULT
959 config ARCH_MMAP_RND_BITS
960 int "Number of bits to use for ASLR of mmap base address" if EXPERT
961 range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX
962 default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT
963 default ARCH_MMAP_RND_BITS_MIN
964 depends on HAVE_ARCH_MMAP_RND_BITS
966 This value can be used to select the number of bits to use to
967 determine the random offset to the base address of vma regions
968 resulting from mmap allocations. This value will be bounded
969 by the architecture's minimum and maximum supported values.
971 This value can be changed after boot using the
972 /proc/sys/vm/mmap_rnd_bits tunable
974 config HAVE_ARCH_MMAP_RND_COMPAT_BITS
977 An arch should select this symbol if it supports running applications
978 in compatibility mode, supports setting a variable number of bits for
979 use in establishing the base address for mmap allocations, has MMU
980 enabled and provides values for both:
981 - ARCH_MMAP_RND_COMPAT_BITS_MIN
982 - ARCH_MMAP_RND_COMPAT_BITS_MAX
984 config ARCH_MMAP_RND_COMPAT_BITS_MIN
987 config ARCH_MMAP_RND_COMPAT_BITS_MAX
990 config ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
993 config ARCH_MMAP_RND_COMPAT_BITS
994 int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT
995 range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX
996 default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
997 default ARCH_MMAP_RND_COMPAT_BITS_MIN
998 depends on HAVE_ARCH_MMAP_RND_COMPAT_BITS
1000 This value can be used to select the number of bits to use to
1001 determine the random offset to the base address of vma regions
1002 resulting from mmap allocations for compatible applications This
1003 value will be bounded by the architecture's minimum and maximum
1006 This value can be changed after boot using the
1007 /proc/sys/vm/mmap_rnd_compat_bits tunable
1009 config HAVE_ARCH_COMPAT_MMAP_BASES
1012 This allows 64bit applications to invoke 32-bit mmap() syscall
1013 and vice-versa 32-bit applications to call 64-bit mmap().
1014 Required for applications doing different bitness syscalls.
1016 config PAGE_SIZE_LESS_THAN_64KB
1018 depends on !ARM64_64K_PAGES
1019 depends on !IA64_PAGE_SIZE_64KB
1020 depends on !PAGE_SIZE_64KB
1021 depends on !PARISC_PAGE_SIZE_64KB
1022 depends on PAGE_SIZE_LESS_THAN_256KB
1024 config PAGE_SIZE_LESS_THAN_256KB
1026 depends on !PAGE_SIZE_256KB
1028 # This allows to use a set of generic functions to determine mmap base
1029 # address by giving priority to top-down scheme only if the process
1030 # is not in legacy mode (compat task, unlimited stack size or
1031 # sysctl_legacy_va_layout).
1032 # Architecture that selects this option can provide its own version of:
1034 config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT
1037 select ARCH_HAS_ELF_RANDOMIZE
1042 config HAVE_JUMP_LABEL_HACK
1045 config HAVE_NOINSTR_HACK
1048 config HAVE_NOINSTR_VALIDATION
1051 config HAVE_UACCESS_VALIDATION
1055 config HAVE_STACK_VALIDATION
1058 Architecture supports objtool compile-time frame pointer rule
1061 config HAVE_RELIABLE_STACKTRACE
1064 Architecture has either save_stack_trace_tsk_reliable() or
1065 arch_stack_walk_reliable() function which only returns a stack trace
1066 if it can guarantee the trace is reliable.
1068 config HAVE_ARCH_HASH
1072 If this is set, the architecture provides an <asm/hash.h>
1073 file which provides platform-specific implementations of some
1074 functions in <linux/hash.h> or fs/namei.c.
1076 config HAVE_ARCH_NVRAM_OPS
1085 config CLONE_BACKWARDS
1088 Architecture has tls passed as the 4th argument of clone(2),
1091 config CLONE_BACKWARDS2
1094 Architecture has the first two arguments of clone(2) swapped.
1096 config CLONE_BACKWARDS3
1099 Architecture has tls passed as the 3rd argument of clone(2),
1102 config ODD_RT_SIGACTION
1105 Architecture has unusual rt_sigaction(2) arguments
1107 config OLD_SIGSUSPEND
1110 Architecture has old sigsuspend(2) syscall, of one-argument variety
1112 config OLD_SIGSUSPEND3
1115 Even weirder antique ABI - three-argument sigsuspend(2)
1117 config OLD_SIGACTION
1120 Architecture has old sigaction(2) syscall. Nope, not the same
1121 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
1122 but fairly different variant of sigaction(2), thanks to OSF/1
1125 config COMPAT_OLD_SIGACTION
1128 config COMPAT_32BIT_TIME
1129 bool "Provide system calls for 32-bit time_t"
1130 default !64BIT || COMPAT
1132 This enables 32 bit time_t support in addition to 64 bit time_t support.
1133 This is relevant on all 32-bit architectures, and 64-bit architectures
1134 as part of compat syscall handling.
1136 config ARCH_NO_PREEMPT
1139 config ARCH_EPHEMERAL_INODES
1142 An arch should select this symbol if it doesn't keep track of inode
1143 instances on its own, but instead relies on something else (e.g. the
1144 host kernel for an UML kernel).
1146 config ARCH_SUPPORTS_RT
1149 config CPU_NO_EFFICIENT_FFS
1152 config HAVE_ARCH_VMAP_STACK
1155 An arch should select this symbol if it can support kernel stacks
1156 in vmalloc space. This means:
1158 - vmalloc space must be large enough to hold many kernel stacks.
1159 This may rule out many 32-bit architectures.
1161 - Stacks in vmalloc space need to work reliably. For example, if
1162 vmap page tables are created on demand, either this mechanism
1163 needs to work while the stack points to a virtual address with
1164 unpopulated page tables or arch code (switch_to() and switch_mm(),
1165 most likely) needs to ensure that the stack's page table entries
1166 are populated before running on a possibly unpopulated stack.
1168 - If the stack overflows into a guard page, something reasonable
1169 should happen. The definition of "reasonable" is flexible, but
1170 instantly rebooting without logging anything would be unfriendly.
1174 bool "Use a virtually-mapped stack"
1175 depends on HAVE_ARCH_VMAP_STACK
1176 depends on !KASAN || KASAN_HW_TAGS || KASAN_VMALLOC
1178 Enable this if you want the use virtually-mapped kernel stacks
1179 with guard pages. This causes kernel stack overflows to be
1180 caught immediately rather than causing difficult-to-diagnose
1183 To use this with software KASAN modes, the architecture must support
1184 backing virtual mappings with real shadow memory, and KASAN_VMALLOC
1187 config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1190 An arch should select this symbol if it can support kernel stack
1191 offset randomization with calls to add_random_kstack_offset()
1192 during syscall entry and choose_random_kstack_offset() during
1193 syscall exit. Careful removal of -fstack-protector-strong and
1194 -fstack-protector should also be applied to the entry code and
1195 closely examined, as the artificial stack bump looks like an array
1196 to the compiler, so it will attempt to add canary checks regardless
1197 of the static branch state.
1199 config RANDOMIZE_KSTACK_OFFSET
1200 bool "Support for randomizing kernel stack offset on syscall entry" if EXPERT
1202 depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1203 depends on INIT_STACK_NONE || !CC_IS_CLANG || CLANG_VERSION >= 140000
1205 The kernel stack offset can be randomized (after pt_regs) by
1206 roughly 5 bits of entropy, frustrating memory corruption
1207 attacks that depend on stack address determinism or
1208 cross-syscall address exposures.
1210 The feature is controlled via the "randomize_kstack_offset=on/off"
1211 kernel boot param, and if turned off has zero overhead due to its use
1212 of static branches (see JUMP_LABEL).
1216 config RANDOMIZE_KSTACK_OFFSET_DEFAULT
1217 bool "Default state of kernel stack offset randomization"
1218 depends on RANDOMIZE_KSTACK_OFFSET
1220 Kernel stack offset randomization is controlled by kernel boot param
1221 "randomize_kstack_offset=on/off", and this config chooses the default
1224 config ARCH_OPTIONAL_KERNEL_RWX
1227 config ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1230 config ARCH_HAS_STRICT_KERNEL_RWX
1233 config STRICT_KERNEL_RWX
1234 bool "Make kernel text and rodata read-only" if ARCH_OPTIONAL_KERNEL_RWX
1235 depends on ARCH_HAS_STRICT_KERNEL_RWX
1236 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1238 If this is set, kernel text and rodata memory will be made read-only,
1239 and non-text memory will be made non-executable. This provides
1240 protection against certain security exploits (e.g. executing the heap
1243 These features are considered standard security practice these days.
1244 You should say Y here in almost all cases.
1246 config ARCH_HAS_STRICT_MODULE_RWX
1249 config STRICT_MODULE_RWX
1250 bool "Set loadable kernel module data as NX and text as RO" if ARCH_OPTIONAL_KERNEL_RWX
1251 depends on ARCH_HAS_STRICT_MODULE_RWX && MODULES
1252 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1254 If this is set, module text and rodata memory will be made read-only,
1255 and non-text memory will be made non-executable. This provides
1256 protection against certain security exploits (e.g. writing to text)
1258 # select if the architecture provides an asm/dma-direct.h header
1259 config ARCH_HAS_PHYS_TO_DMA
1262 config HAVE_ARCH_COMPILER_H
1265 An architecture can select this if it provides an
1266 asm/compiler.h header that should be included after
1267 linux/compiler-*.h in order to override macro definitions that those
1268 headers generally provide.
1270 config HAVE_ARCH_PREL32_RELOCATIONS
1273 May be selected by an architecture if it supports place-relative
1274 32-bit relocations, both in the toolchain and in the module loader,
1275 in which case relative references can be used in special sections
1276 for PCI fixup, initcalls etc which are only half the size on 64 bit
1277 architectures, and don't require runtime relocation on relocatable
1280 config ARCH_USE_MEMREMAP_PROT
1283 config LOCK_EVENT_COUNTS
1284 bool "Locking event counts collection"
1287 Enable light-weight counting of various locking related events
1288 in the system with minimal performance impact. This reduces
1289 the chance of application behavior change because of timing
1290 differences. The counts are reported via debugfs.
1292 # Select if the architecture has support for applying RELR relocations.
1293 config ARCH_HAS_RELR
1297 bool "Use RELR relocation packing"
1298 depends on ARCH_HAS_RELR && TOOLS_SUPPORT_RELR
1301 Store the kernel's dynamic relocations in the RELR relocation packing
1302 format. Requires a compatible linker (LLD supports this feature), as
1303 well as compatible NM and OBJCOPY utilities (llvm-nm and llvm-objcopy
1306 config ARCH_HAS_MEM_ENCRYPT
1309 config ARCH_HAS_CC_PLATFORM
1312 config HAVE_SPARSE_SYSCALL_NR
1315 An architecture should select this if its syscall numbering is sparse
1316 to save space. For example, MIPS architecture has a syscall array with
1317 entries at 4000, 5000 and 6000 locations. This option turns on syscall
1318 related optimizations for a given architecture.
1320 config ARCH_HAS_VDSO_DATA
1323 config HAVE_STATIC_CALL
1326 config HAVE_STATIC_CALL_INLINE
1328 depends on HAVE_STATIC_CALL
1331 config HAVE_PREEMPT_DYNAMIC
1334 config HAVE_PREEMPT_DYNAMIC_CALL
1336 depends on HAVE_STATIC_CALL
1337 select HAVE_PREEMPT_DYNAMIC
1339 An architecture should select this if it can handle the preemption
1340 model being selected at boot time using static calls.
1342 Where an architecture selects HAVE_STATIC_CALL_INLINE, any call to a
1343 preemption function will be patched directly.
1345 Where an architecture does not select HAVE_STATIC_CALL_INLINE, any
1346 call to a preemption function will go through a trampoline, and the
1347 trampoline will be patched.
1349 It is strongly advised to support inline static call to avoid any
1352 config HAVE_PREEMPT_DYNAMIC_KEY
1354 depends on HAVE_ARCH_JUMP_LABEL && CC_HAS_ASM_GOTO
1355 select HAVE_PREEMPT_DYNAMIC
1357 An architecture should select this if it can handle the preemption
1358 model being selected at boot time using static keys.
1360 Each preemption function will be given an early return based on a
1361 static key. This should have slightly lower overhead than non-inline
1362 static calls, as this effectively inlines each trampoline into the
1363 start of its callee. This may avoid redundant work, and may
1364 integrate better with CFI schemes.
1366 This will have greater overhead than using inline static calls as
1367 the call to the preemption function cannot be entirely elided.
1369 config ARCH_WANT_LD_ORPHAN_WARN
1372 An arch should select this symbol once all linker sections are explicitly
1373 included, size-asserted, or discarded in the linker scripts. This is
1374 important because we never want expected sections to be placed heuristically
1375 by the linker, since the locations of such sections can change between linker
1378 config HAVE_ARCH_PFN_VALID
1381 config ARCH_SUPPORTS_DEBUG_PAGEALLOC
1384 config ARCH_SUPPORTS_PAGE_TABLE_CHECK
1387 config ARCH_SPLIT_ARG64
1390 If a 32-bit architecture requires 64-bit arguments to be split into
1391 pairs of 32-bit arguments, select this option.
1393 config ARCH_HAS_ELFCORE_COMPAT
1396 config ARCH_HAS_PARANOID_L1D_FLUSH
1399 config DYNAMIC_SIGFRAME
1402 # Select, if arch has a named attribute group bound to NUMA device nodes.
1403 config HAVE_ARCH_NODE_DEV_GROUP
1406 source "kernel/gcov/Kconfig"
1408 source "scripts/gcc-plugins/Kconfig"