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
37 # Selected by HOTPLUG_CORE_SYNC_DEAD or HOTPLUG_CORE_SYNC_FULL
38 config HOTPLUG_CORE_SYNC
41 # Basic CPU dead synchronization selected by architecture
42 config HOTPLUG_CORE_SYNC_DEAD
44 select HOTPLUG_CORE_SYNC
46 # Full CPU synchronization with alive state selected by architecture
47 config HOTPLUG_CORE_SYNC_FULL
49 select HOTPLUG_CORE_SYNC_DEAD if HOTPLUG_CPU
50 select HOTPLUG_CORE_SYNC
52 config HOTPLUG_SPLIT_STARTUP
54 select HOTPLUG_CORE_SYNC_FULL
56 config HOTPLUG_PARALLEL
58 select HOTPLUG_SPLIT_STARTUP
66 depends on HAVE_KPROBES
68 select TASKS_RCU if PREEMPTION
70 Kprobes allows you to trap at almost any kernel address and
71 execute a callback function. register_kprobe() establishes
72 a probepoint and specifies the callback. Kprobes is useful
73 for kernel debugging, non-intrusive instrumentation and testing.
77 bool "Optimize very unlikely/likely branches"
78 depends on HAVE_ARCH_JUMP_LABEL
79 select OBJTOOL if HAVE_JUMP_LABEL_HACK
81 This option enables a transparent branch optimization that
82 makes certain almost-always-true or almost-always-false branch
83 conditions even cheaper to execute within the kernel.
85 Certain performance-sensitive kernel code, such as trace points,
86 scheduler functionality, networking code and KVM have such
87 branches and include support for this optimization technique.
89 If it is detected that the compiler has support for "asm goto",
90 the kernel will compile such branches with just a nop
91 instruction. When the condition flag is toggled to true, the
92 nop will be converted to a jump instruction to execute the
93 conditional block of instructions.
95 This technique lowers overhead and stress on the branch prediction
96 of the processor and generally makes the kernel faster. The update
97 of the condition is slower, but those are always very rare.
99 ( On 32-bit x86, the necessary options added to the compiler
100 flags may increase the size of the kernel slightly. )
102 config STATIC_KEYS_SELFTEST
103 bool "Static key selftest"
104 depends on JUMP_LABEL
106 Boot time self-test of the branch patching code.
108 config STATIC_CALL_SELFTEST
109 bool "Static call selftest"
110 depends on HAVE_STATIC_CALL
112 Boot time self-test of the call patching code.
116 depends on KPROBES && HAVE_OPTPROBES
117 select TASKS_RCU if PREEMPTION
119 config KPROBES_ON_FTRACE
121 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
122 depends on DYNAMIC_FTRACE_WITH_REGS
124 If function tracer is enabled and the arch supports full
125 passing of pt_regs to function tracing, then kprobes can
126 optimize on top of function tracing.
130 depends on ARCH_SUPPORTS_UPROBES
132 Uprobes is the user-space counterpart to kprobes: they
133 enable instrumentation applications (such as 'perf probe')
134 to establish unintrusive probes in user-space binaries and
135 libraries, by executing handler functions when the probes
136 are hit by user-space applications.
138 ( These probes come in the form of single-byte breakpoints,
139 managed by the kernel and kept transparent to the probed
142 config HAVE_64BIT_ALIGNED_ACCESS
143 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
145 Some architectures require 64 bit accesses to be 64 bit
146 aligned, which also requires structs containing 64 bit values
147 to be 64 bit aligned too. This includes some 32 bit
148 architectures which can do 64 bit accesses, as well as 64 bit
149 architectures without unaligned access.
151 This symbol should be selected by an architecture if 64 bit
152 accesses are required to be 64 bit aligned in this way even
153 though it is not a 64 bit architecture.
155 See Documentation/core-api/unaligned-memory-access.rst for
156 more information on the topic of unaligned memory accesses.
158 config HAVE_EFFICIENT_UNALIGNED_ACCESS
161 Some architectures are unable to perform unaligned accesses
162 without the use of get_unaligned/put_unaligned. Others are
163 unable to perform such accesses efficiently (e.g. trap on
164 unaligned access and require fixing it up in the exception
167 This symbol should be selected by an architecture if it can
168 perform unaligned accesses efficiently to allow different
169 code paths to be selected for these cases. Some network
170 drivers, for example, could opt to not fix up alignment
171 problems with received packets if doing so would not help
174 See Documentation/core-api/unaligned-memory-access.rst for more
175 information on the topic of unaligned memory accesses.
177 config ARCH_USE_BUILTIN_BSWAP
180 Modern versions of GCC (since 4.4) have builtin functions
181 for handling byte-swapping. Using these, instead of the old
182 inline assembler that the architecture code provides in the
183 __arch_bswapXX() macros, allows the compiler to see what's
184 happening and offers more opportunity for optimisation. In
185 particular, the compiler will be able to combine the byteswap
186 with a nearby load or store and use load-and-swap or
187 store-and-swap instructions if the architecture has them. It
188 should almost *never* result in code which is worse than the
189 hand-coded assembler in <asm/swab.h>. But just in case it
190 does, the use of the builtins is optional.
192 Any architecture with load-and-swap or store-and-swap
193 instructions should set this. And it shouldn't hurt to set it
194 on architectures that don't have such instructions.
198 depends on KPROBES && (HAVE_KRETPROBES || HAVE_RETHOOK)
200 config KRETPROBE_ON_RETHOOK
202 depends on HAVE_RETHOOK
203 depends on KRETPROBES
206 config USER_RETURN_NOTIFIER
208 depends on HAVE_USER_RETURN_NOTIFIER
210 Provide a kernel-internal notification when a cpu is about to
213 config HAVE_IOREMAP_PROT
219 config HAVE_KRETPROBES
222 config HAVE_OPTPROBES
225 config HAVE_KPROBES_ON_FTRACE
228 config ARCH_CORRECT_STACKTRACE_ON_KRETPROBE
231 Since kretprobes modifies return address on the stack, the
232 stacktrace may see the kretprobe trampoline address instead
233 of correct one. If the architecture stacktrace code and
234 unwinder can adjust such entries, select this configuration.
236 config HAVE_FUNCTION_ERROR_INJECTION
242 config HAVE_FUNCTION_DESCRIPTORS
245 config TRACE_IRQFLAGS_SUPPORT
248 config TRACE_IRQFLAGS_NMI_SUPPORT
252 # An arch should select this if it provides all these things:
254 # task_pt_regs() in asm/processor.h or asm/ptrace.h
255 # arch_has_single_step() if there is hardware single-step support
256 # arch_has_block_step() if there is hardware block-step support
257 # asm/syscall.h supplying asm-generic/syscall.h interface
258 # linux/regset.h user_regset interfaces
259 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
260 # TIF_SYSCALL_TRACE calls ptrace_report_syscall_{entry,exit}
261 # TIF_NOTIFY_RESUME calls resume_user_mode_work()
263 config HAVE_ARCH_TRACEHOOK
266 config HAVE_DMA_CONTIGUOUS
269 config GENERIC_SMP_IDLE_THREAD
272 config GENERIC_IDLE_POLL_SETUP
275 config ARCH_HAS_FORTIFY_SOURCE
278 An architecture should select this when it can successfully
279 build and run with CONFIG_FORTIFY_SOURCE.
282 # Select if the arch provides a historic keepinit alias for the retain_initrd
283 # command line option
285 config ARCH_HAS_KEEPINITRD
288 # Select if arch has all set_memory_ro/rw/x/nx() functions in asm/cacheflush.h
289 config ARCH_HAS_SET_MEMORY
292 # Select if arch has all set_direct_map_invalid/default() functions
293 config ARCH_HAS_SET_DIRECT_MAP
297 # Select if the architecture provides the arch_dma_set_uncached symbol to
298 # either provide an uncached segment alias for a DMA allocation, or
299 # to remap the page tables in place.
301 config ARCH_HAS_DMA_SET_UNCACHED
305 # Select if the architectures provides the arch_dma_clear_uncached symbol
306 # to undo an in-place page table remap for uncached access.
308 config ARCH_HAS_DMA_CLEAR_UNCACHED
311 config ARCH_HAS_CPU_FINALIZE_INIT
314 # Select if arch init_task must go in the __init_task_data section
315 config ARCH_TASK_STRUCT_ON_STACK
318 # Select if arch has its private alloc_task_struct() function
319 config ARCH_TASK_STRUCT_ALLOCATOR
322 config HAVE_ARCH_THREAD_STRUCT_WHITELIST
324 depends on !ARCH_TASK_STRUCT_ALLOCATOR
326 An architecture should select this to provide hardened usercopy
327 knowledge about what region of the thread_struct should be
328 whitelisted for copying to userspace. Normally this is only the
329 FPU registers. Specifically, arch_thread_struct_whitelist()
330 should be implemented. Without this, the entire thread_struct
331 field in task_struct will be left whitelisted.
333 # Select if arch has its private alloc_thread_stack() function
334 config ARCH_THREAD_STACK_ALLOCATOR
337 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
338 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
341 config ARCH_WANTS_NO_INSTR
344 An architecture should select this if the noinstr macro is being used on
345 functions to denote that the toolchain should avoid instrumenting such
346 functions and is required for correctness.
348 config ARCH_32BIT_OFF_T
352 All new 32-bit architectures should have 64-bit off_t type on
353 userspace side which corresponds to the loff_t kernel type. This
354 is the requirement for modern ABIs. Some existing architectures
355 still support 32-bit off_t. This option is enabled for all such
356 architectures explicitly.
358 # Selected by 64 bit architectures which have a 32 bit f_tinode in struct ustat
359 config ARCH_32BIT_USTAT_F_TINODE
362 config HAVE_ASM_MODVERSIONS
365 This symbol should be selected by an architecture if it provides
366 <asm/asm-prototypes.h> to support the module versioning for symbols
367 exported from assembly code.
369 config HAVE_REGS_AND_STACK_ACCESS_API
372 This symbol should be selected by an architecture if it supports
373 the API needed to access registers and stack entries from pt_regs,
374 declared in asm/ptrace.h
375 For example the kprobes-based event tracer needs this API.
379 depends on HAVE_REGS_AND_STACK_ACCESS_API
381 This symbol should be selected by an architecture if it
382 supports an implementation of restartable sequences.
387 This symbol should be selected by an architecture if it
390 config HAVE_FUNCTION_ARG_ACCESS_API
393 This symbol should be selected by an architecture if it supports
394 the API needed to access function arguments from pt_regs,
395 declared in asm/ptrace.h
397 config HAVE_HW_BREAKPOINT
399 depends on PERF_EVENTS
401 config HAVE_MIXED_BREAKPOINTS_REGS
403 depends on HAVE_HW_BREAKPOINT
405 Depending on the arch implementation of hardware breakpoints,
406 some of them have separate registers for data and instruction
407 breakpoints addresses, others have mixed registers to store
408 them but define the access type in a control register.
409 Select this option if your arch implements breakpoints under the
412 config HAVE_USER_RETURN_NOTIFIER
415 config HAVE_PERF_EVENTS_NMI
418 System hardware can generate an NMI using the perf event
419 subsystem. Also has support for calculating CPU cycle events
420 to determine how many clock cycles in a given period.
422 config HAVE_HARDLOCKUP_DETECTOR_PERF
424 depends on HAVE_PERF_EVENTS_NMI
426 The arch chooses to use the generic perf-NMI-based hardlockup
427 detector. Must define HAVE_PERF_EVENTS_NMI.
429 config HAVE_NMI_WATCHDOG
433 The arch provides a low level NMI watchdog. It provides
434 asm/nmi.h, and defines its own arch_touch_nmi_watchdog().
436 config HAVE_HARDLOCKUP_DETECTOR_ARCH
438 select HAVE_NMI_WATCHDOG
440 The arch chooses to provide its own hardlockup detector, which is
441 a superset of the HAVE_NMI_WATCHDOG. It also conforms to config
442 interfaces and parameters provided by hardlockup detector subsystem.
444 config HAVE_PERF_REGS
447 Support selective register dumps for perf events. This includes
448 bit-mapping of each registers and a unique architecture id.
450 config HAVE_PERF_USER_STACK_DUMP
453 Support user stack dumps for perf event samples. This needs
454 access to the user stack pointer which is not unified across
457 config HAVE_ARCH_JUMP_LABEL
460 config HAVE_ARCH_JUMP_LABEL_RELATIVE
463 config MMU_GATHER_TABLE_FREE
466 config MMU_GATHER_RCU_TABLE_FREE
468 select MMU_GATHER_TABLE_FREE
470 config MMU_GATHER_PAGE_SIZE
473 config MMU_GATHER_NO_RANGE
475 select MMU_GATHER_MERGE_VMAS
477 config MMU_GATHER_NO_FLUSH_CACHE
480 config MMU_GATHER_MERGE_VMAS
483 config MMU_GATHER_NO_GATHER
485 depends on MMU_GATHER_TABLE_FREE
487 config ARCH_WANT_IRQS_OFF_ACTIVATE_MM
490 Temporary select until all architectures can be converted to have
491 irqs disabled over activate_mm. Architectures that do IPI based TLB
492 shootdowns should enable this.
494 # Use normal mm refcounting for MMU_LAZY_TLB kernel thread references.
495 # MMU_LAZY_TLB_REFCOUNT=n can improve the scalability of context switching
496 # to/from kernel threads when the same mm is running on a lot of CPUs (a large
497 # multi-threaded application), by reducing contention on the mm refcount.
499 # This can be disabled if the architecture ensures no CPUs are using an mm as a
500 # "lazy tlb" beyond its final refcount (i.e., by the time __mmdrop frees the mm
501 # or its kernel page tables). This could be arranged by arch_exit_mmap(), or
502 # final exit(2) TLB flush, for example.
504 # To implement this, an arch *must*:
505 # Ensure the _lazy_tlb variants of mmgrab/mmdrop are used when manipulating
506 # the lazy tlb reference of a kthread's ->active_mm (non-arch code has been
507 # converted already).
508 config MMU_LAZY_TLB_REFCOUNT
510 depends on !MMU_LAZY_TLB_SHOOTDOWN
512 # This option allows MMU_LAZY_TLB_REFCOUNT=n. It ensures no CPUs are using an
513 # mm as a lazy tlb beyond its last reference count, by shooting down these
514 # users before the mm is deallocated. __mmdrop() first IPIs all CPUs that may
515 # be using the mm as a lazy tlb, so that they may switch themselves to using
516 # init_mm for their active mm. mm_cpumask(mm) is used to determine which CPUs
517 # may be using mm as a lazy tlb mm.
519 # To implement this, an arch *must*:
520 # - At the time of the final mmdrop of the mm, ensure mm_cpumask(mm) contains
521 # at least all possible CPUs in which the mm is lazy.
522 # - It must meet the requirements for MMU_LAZY_TLB_REFCOUNT=n (see above).
523 config MMU_LAZY_TLB_SHOOTDOWN
526 config ARCH_HAVE_NMI_SAFE_CMPXCHG
529 config ARCH_HAS_NMI_SAFE_THIS_CPU_OPS
532 config HAVE_ALIGNED_STRUCT_PAGE
535 This makes sure that struct pages are double word aligned and that
536 e.g. the SLUB allocator can perform double word atomic operations
537 on a struct page for better performance. However selecting this
538 might increase the size of a struct page by a word.
540 config HAVE_CMPXCHG_LOCAL
543 config HAVE_CMPXCHG_DOUBLE
546 config ARCH_WEAK_RELEASE_ACQUIRE
549 config ARCH_WANT_IPC_PARSE_VERSION
552 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
555 config ARCH_WANT_OLD_COMPAT_IPC
556 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
559 config HAVE_ARCH_SECCOMP
562 An arch should select this symbol to support seccomp mode 1 (the fixed
563 syscall policy), and must provide an overrides for __NR_seccomp_sigreturn,
564 and compat syscalls if the asm-generic/seccomp.h defaults need adjustment:
565 - __NR_seccomp_read_32
566 - __NR_seccomp_write_32
567 - __NR_seccomp_exit_32
568 - __NR_seccomp_sigreturn_32
570 config HAVE_ARCH_SECCOMP_FILTER
572 select HAVE_ARCH_SECCOMP
574 An arch should select this symbol if it provides all of these things:
575 - all the requirements for HAVE_ARCH_SECCOMP
577 - syscall_get_arguments()
579 - syscall_set_return_value()
580 - SIGSYS siginfo_t support
581 - secure_computing is called from a ptrace_event()-safe context
582 - secure_computing return value is checked and a return value of -1
583 results in the system call being skipped immediately.
584 - seccomp syscall wired up
585 - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE,
586 SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If
587 COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too.
590 prompt "Enable seccomp to safely execute untrusted bytecode"
592 depends on HAVE_ARCH_SECCOMP
594 This kernel feature is useful for number crunching applications
595 that may need to handle untrusted bytecode during their
596 execution. By using pipes or other transports made available
597 to the process as file descriptors supporting the read/write
598 syscalls, it's possible to isolate those applications in their
599 own address space using seccomp. Once seccomp is enabled via
600 prctl(PR_SET_SECCOMP) or the seccomp() syscall, it cannot be
601 disabled and the task is only allowed to execute a few safe
602 syscalls defined by each seccomp mode.
606 config SECCOMP_FILTER
608 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
610 Enable tasks to build secure computing environments defined
611 in terms of Berkeley Packet Filter programs which implement
612 task-defined system call filtering polices.
614 See Documentation/userspace-api/seccomp_filter.rst for details.
616 config SECCOMP_CACHE_DEBUG
617 bool "Show seccomp filter cache status in /proc/pid/seccomp_cache"
618 depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR
621 This enables the /proc/pid/seccomp_cache interface to monitor
622 seccomp cache data. The file format is subject to change. Reading
623 the file requires CAP_SYS_ADMIN.
625 This option is for debugging only. Enabling presents the risk that
626 an adversary may be able to infer the seccomp filter logic.
630 config HAVE_ARCH_STACKLEAK
633 An architecture should select this if it has the code which
634 fills the used part of the kernel stack with the STACKLEAK_POISON
635 value before returning from system calls.
637 config HAVE_STACKPROTECTOR
640 An arch should select this symbol if:
641 - it has implemented a stack canary (e.g. __stack_chk_guard)
643 config STACKPROTECTOR
644 bool "Stack Protector buffer overflow detection"
645 depends on HAVE_STACKPROTECTOR
646 depends on $(cc-option,-fstack-protector)
649 This option turns on the "stack-protector" GCC feature. This
650 feature puts, at the beginning of functions, a canary value on
651 the stack just before the return address, and validates
652 the value just before actually returning. Stack based buffer
653 overflows (that need to overwrite this return address) now also
654 overwrite the canary, which gets detected and the attack is then
655 neutralized via a kernel panic.
657 Functions will have the stack-protector canary logic added if they
658 have an 8-byte or larger character array on the stack.
660 This feature requires gcc version 4.2 or above, or a distribution
661 gcc with the feature backported ("-fstack-protector").
663 On an x86 "defconfig" build, this feature adds canary checks to
664 about 3% of all kernel functions, which increases kernel code size
667 config STACKPROTECTOR_STRONG
668 bool "Strong Stack Protector"
669 depends on STACKPROTECTOR
670 depends on $(cc-option,-fstack-protector-strong)
673 Functions will have the stack-protector canary logic added in any
674 of the following conditions:
676 - local variable's address used as part of the right hand side of an
677 assignment or function argument
678 - local variable is an array (or union containing an array),
679 regardless of array type or length
680 - uses register local variables
682 This feature requires gcc version 4.9 or above, or a distribution
683 gcc with the feature backported ("-fstack-protector-strong").
685 On an x86 "defconfig" build, this feature adds canary checks to
686 about 20% of all kernel functions, which increases the kernel code
689 config ARCH_SUPPORTS_SHADOW_CALL_STACK
692 An architecture should select this if it supports the compiler's
693 Shadow Call Stack and implements runtime support for shadow stack
696 config SHADOW_CALL_STACK
697 bool "Shadow Call Stack"
698 depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
699 depends on DYNAMIC_FTRACE_WITH_ARGS || DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
701 This option enables the compiler's Shadow Call Stack, which
702 uses a shadow stack to protect function return addresses from
703 being overwritten by an attacker. More information can be found
704 in the compiler's documentation:
706 - Clang: https://clang.llvm.org/docs/ShadowCallStack.html
707 - GCC: https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html#Instrumentation-Options
709 Note that security guarantees in the kernel differ from the
710 ones documented for user space. The kernel must store addresses
711 of shadow stacks in memory, which means an attacker capable of
712 reading and writing arbitrary memory may be able to locate them
713 and hijack control flow by modifying the stacks.
718 Set by the arch code if it relies on code patching to insert the
719 shadow call stack push and pop instructions rather than on the
725 Selected if the kernel will be built using the compiler's LTO feature.
731 Selected if the kernel will be built using Clang's LTO feature.
733 config ARCH_SUPPORTS_LTO_CLANG
736 An architecture should select this option if it supports:
737 - compiling with Clang,
738 - compiling inline assembly with Clang's integrated assembler,
739 - and linking with LLD.
741 config ARCH_SUPPORTS_LTO_CLANG_THIN
744 An architecture should select this option if it can support Clang's
749 depends on CC_IS_CLANG && LD_IS_LLD && AS_IS_LLVM
750 depends on $(success,$(NM) --help | head -n 1 | grep -qi llvm)
751 depends on $(success,$(AR) --help | head -n 1 | grep -qi llvm)
752 depends on ARCH_SUPPORTS_LTO_CLANG
753 depends on !FTRACE_MCOUNT_USE_RECORDMCOUNT
754 depends on !KASAN || KASAN_HW_TAGS
755 depends on !GCOV_KERNEL
757 The compiler and Kconfig options support building with Clang's
761 prompt "Link Time Optimization (LTO)"
764 This option enables Link Time Optimization (LTO), which allows the
765 compiler to optimize binaries globally.
767 If unsure, select LTO_NONE. Note that LTO is very resource-intensive
768 so it's disabled by default.
773 Build the kernel normally, without Link Time Optimization (LTO).
775 config LTO_CLANG_FULL
776 bool "Clang Full LTO (EXPERIMENTAL)"
777 depends on HAS_LTO_CLANG
778 depends on !COMPILE_TEST
781 This option enables Clang's full Link Time Optimization (LTO), which
782 allows the compiler to optimize the kernel globally. If you enable
783 this option, the compiler generates LLVM bitcode instead of ELF
784 object files, and the actual compilation from bitcode happens at
785 the LTO link step, which may take several minutes depending on the
786 kernel configuration. More information can be found from LLVM's
789 https://llvm.org/docs/LinkTimeOptimization.html
791 During link time, this option can use a large amount of RAM, and
792 may take much longer than the ThinLTO option.
794 config LTO_CLANG_THIN
795 bool "Clang ThinLTO (EXPERIMENTAL)"
796 depends on HAS_LTO_CLANG && ARCH_SUPPORTS_LTO_CLANG_THIN
799 This option enables Clang's ThinLTO, which allows for parallel
800 optimization and faster incremental compiles compared to the
801 CONFIG_LTO_CLANG_FULL option. More information can be found
802 from Clang's documentation:
804 https://clang.llvm.org/docs/ThinLTO.html
809 config ARCH_SUPPORTS_CFI_CLANG
812 An architecture should select this option if it can support Clang's
813 Control-Flow Integrity (CFI) checking.
815 config ARCH_USES_CFI_TRAPS
819 bool "Use Clang's Control Flow Integrity (CFI)"
820 depends on ARCH_SUPPORTS_CFI_CLANG
821 depends on $(cc-option,-fsanitize=kcfi)
823 This option enables Clang’s forward-edge Control Flow Integrity
824 (CFI) checking, where the compiler injects a runtime check to each
825 indirect function call to ensure the target is a valid function with
826 the correct static type. This restricts possible call targets and
827 makes it more difficult for an attacker to exploit bugs that allow
828 the modification of stored function pointers. More information can be
829 found from Clang's documentation:
831 https://clang.llvm.org/docs/ControlFlowIntegrity.html
833 config CFI_PERMISSIVE
834 bool "Use CFI in permissive mode"
837 When selected, Control Flow Integrity (CFI) violations result in a
838 warning instead of a kernel panic. This option should only be used
839 for finding indirect call type mismatches during development.
843 config HAVE_ARCH_WITHIN_STACK_FRAMES
846 An architecture should select this if it can walk the kernel stack
847 frames to determine if an object is part of either the arguments
848 or local variables (i.e. that it excludes saved return addresses,
849 and similar) by implementing an inline arch_within_stack_frames(),
850 which is used by CONFIG_HARDENED_USERCOPY.
852 config HAVE_CONTEXT_TRACKING_USER
855 Provide kernel/user boundaries probes necessary for subsystems
856 that need it, such as userspace RCU extended quiescent state.
857 Syscalls need to be wrapped inside user_exit()-user_enter(), either
858 optimized behind static key or through the slow path using TIF_NOHZ
859 flag. Exceptions handlers must be wrapped as well. Irqs are already
860 protected inside ct_irq_enter/ct_irq_exit() but preemption or signal
861 handling on irq exit still need to be protected.
863 config HAVE_CONTEXT_TRACKING_USER_OFFSTACK
866 Architecture neither relies on exception_enter()/exception_exit()
867 nor on schedule_user(). Also preempt_schedule_notrace() and
868 preempt_schedule_irq() can't be called in a preemptible section
869 while context tracking is CONTEXT_USER. This feature reflects a sane
870 entry implementation where the following requirements are met on
871 critical entry code, ie: before user_exit() or after user_enter():
873 - Critical entry code isn't preemptible (or better yet:
875 - No use of RCU read side critical sections, unless ct_nmi_enter()
877 - No use of instrumentation, unless instrumentation_begin() got
883 Arch relies on TIF_NOHZ and syscall slow path to implement context
884 tracking calls to user_enter()/user_exit().
886 config HAVE_VIRT_CPU_ACCOUNTING
889 config HAVE_VIRT_CPU_ACCOUNTING_IDLE
892 Architecture has its own way to account idle CPU time and therefore
893 doesn't implement vtime_account_idle().
895 config ARCH_HAS_SCALED_CPUTIME
898 config HAVE_VIRT_CPU_ACCOUNTING_GEN
902 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
903 Before enabling this option, arch code must be audited
904 to ensure there are no races in concurrent read/write of
905 cputime_t. For example, reading/writing 64-bit cputime_t on
906 some 32-bit arches may require multiple accesses, so proper
907 locking is needed to protect against concurrent accesses.
909 config HAVE_IRQ_TIME_ACCOUNTING
912 Archs need to ensure they use a high enough resolution clock to
913 support irq time accounting and then call enable_sched_clock_irqtime().
918 Architectures that select this are able to move page tables at the
919 PUD level. If there are only 3 page table levels, the move effectively
920 happens at the PGD level.
925 Archs that select this are able to move page tables at the PMD level.
927 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
930 config HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
933 config HAVE_ARCH_HUGE_VMAP
937 # Archs that select this would be capable of PMD-sized vmaps (i.e.,
938 # arch_vmap_pmd_supported() returns true). The VM_ALLOW_HUGE_VMAP flag
939 # must be used to enable allocations to use hugepages.
941 config HAVE_ARCH_HUGE_VMALLOC
942 depends on HAVE_ARCH_HUGE_VMAP
945 config ARCH_WANT_HUGE_PMD_SHARE
948 config HAVE_ARCH_SOFT_DIRTY
951 config HAVE_MOD_ARCH_SPECIFIC
954 The arch uses struct mod_arch_specific to store data. Many arches
955 just need a simple module loader without arch specific data - those
956 should not enable this.
958 config MODULES_USE_ELF_RELA
961 Modules only use ELF RELA relocations. Modules with ELF REL
962 relocations will give an error.
964 config MODULES_USE_ELF_REL
967 Modules only use ELF REL relocations. Modules with ELF RELA
968 relocations will give an error.
970 config ARCH_WANTS_MODULES_DATA_IN_VMALLOC
973 For architectures like powerpc/32 which have constraints on module
974 allocation and need to allocate module data outside of module area.
976 config HAVE_IRQ_EXIT_ON_IRQ_STACK
979 Architecture doesn't only execute the irq handler on the irq stack
980 but also irq_exit(). This way we can process softirqs on this irq
981 stack instead of switching to a new one when we call __do_softirq()
982 in the end of an hardirq.
983 This spares a stack switch and improves cache usage on softirq
986 config HAVE_SOFTIRQ_ON_OWN_STACK
989 Architecture provides a function to run __do_softirq() on a
992 config SOFTIRQ_ON_OWN_STACK
993 def_bool HAVE_SOFTIRQ_ON_OWN_STACK && !PREEMPT_RT
995 config ALTERNATE_USER_ADDRESS_SPACE
998 Architectures set this when the CPU uses separate address
999 spaces for kernel and user space pointers. In this case, the
1000 access_ok() check on a __user pointer is skipped.
1002 config PGTABLE_LEVELS
1006 config ARCH_HAS_ELF_RANDOMIZE
1009 An architecture supports choosing randomized locations for
1010 stack, mmap, brk, and ET_DYN. Defined functions:
1012 - arch_randomize_brk()
1014 config HAVE_ARCH_MMAP_RND_BITS
1017 An arch should select this symbol if it supports setting a variable
1018 number of bits for use in establishing the base address for mmap
1019 allocations, has MMU enabled and provides values for both:
1020 - ARCH_MMAP_RND_BITS_MIN
1021 - ARCH_MMAP_RND_BITS_MAX
1023 config HAVE_EXIT_THREAD
1026 An architecture implements exit_thread.
1028 config ARCH_MMAP_RND_BITS_MIN
1031 config ARCH_MMAP_RND_BITS_MAX
1034 config ARCH_MMAP_RND_BITS_DEFAULT
1037 config ARCH_MMAP_RND_BITS
1038 int "Number of bits to use for ASLR of mmap base address" if EXPERT
1039 range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX
1040 default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT
1041 default ARCH_MMAP_RND_BITS_MIN
1042 depends on HAVE_ARCH_MMAP_RND_BITS
1044 This value can be used to select the number of bits to use to
1045 determine the random offset to the base address of vma regions
1046 resulting from mmap allocations. This value will be bounded
1047 by the architecture's minimum and maximum supported values.
1049 This value can be changed after boot using the
1050 /proc/sys/vm/mmap_rnd_bits tunable
1052 config HAVE_ARCH_MMAP_RND_COMPAT_BITS
1055 An arch should select this symbol if it supports running applications
1056 in compatibility mode, supports setting a variable number of bits for
1057 use in establishing the base address for mmap allocations, has MMU
1058 enabled and provides values for both:
1059 - ARCH_MMAP_RND_COMPAT_BITS_MIN
1060 - ARCH_MMAP_RND_COMPAT_BITS_MAX
1062 config ARCH_MMAP_RND_COMPAT_BITS_MIN
1065 config ARCH_MMAP_RND_COMPAT_BITS_MAX
1068 config ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1071 config ARCH_MMAP_RND_COMPAT_BITS
1072 int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT
1073 range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX
1074 default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1075 default ARCH_MMAP_RND_COMPAT_BITS_MIN
1076 depends on HAVE_ARCH_MMAP_RND_COMPAT_BITS
1078 This value can be used to select the number of bits to use to
1079 determine the random offset to the base address of vma regions
1080 resulting from mmap allocations for compatible applications This
1081 value will be bounded by the architecture's minimum and maximum
1084 This value can be changed after boot using the
1085 /proc/sys/vm/mmap_rnd_compat_bits tunable
1087 config HAVE_ARCH_COMPAT_MMAP_BASES
1090 This allows 64bit applications to invoke 32-bit mmap() syscall
1091 and vice-versa 32-bit applications to call 64-bit mmap().
1092 Required for applications doing different bitness syscalls.
1094 config PAGE_SIZE_LESS_THAN_64KB
1096 depends on !ARM64_64K_PAGES
1097 depends on !IA64_PAGE_SIZE_64KB
1098 depends on !PAGE_SIZE_64KB
1099 depends on !PARISC_PAGE_SIZE_64KB
1100 depends on PAGE_SIZE_LESS_THAN_256KB
1102 config PAGE_SIZE_LESS_THAN_256KB
1104 depends on !PAGE_SIZE_256KB
1106 # This allows to use a set of generic functions to determine mmap base
1107 # address by giving priority to top-down scheme only if the process
1108 # is not in legacy mode (compat task, unlimited stack size or
1109 # sysctl_legacy_va_layout).
1110 # Architecture that selects this option can provide its own version of:
1112 config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT
1115 select ARCH_HAS_ELF_RANDOMIZE
1120 config HAVE_JUMP_LABEL_HACK
1123 config HAVE_NOINSTR_HACK
1126 config HAVE_NOINSTR_VALIDATION
1129 config HAVE_UACCESS_VALIDATION
1133 config HAVE_STACK_VALIDATION
1136 Architecture supports objtool compile-time frame pointer rule
1139 config HAVE_RELIABLE_STACKTRACE
1142 Architecture has either save_stack_trace_tsk_reliable() or
1143 arch_stack_walk_reliable() function which only returns a stack trace
1144 if it can guarantee the trace is reliable.
1146 config HAVE_ARCH_HASH
1150 If this is set, the architecture provides an <asm/hash.h>
1151 file which provides platform-specific implementations of some
1152 functions in <linux/hash.h> or fs/namei.c.
1154 config HAVE_ARCH_NVRAM_OPS
1163 config CLONE_BACKWARDS
1166 Architecture has tls passed as the 4th argument of clone(2),
1169 config CLONE_BACKWARDS2
1172 Architecture has the first two arguments of clone(2) swapped.
1174 config CLONE_BACKWARDS3
1177 Architecture has tls passed as the 3rd argument of clone(2),
1180 config ODD_RT_SIGACTION
1183 Architecture has unusual rt_sigaction(2) arguments
1185 config OLD_SIGSUSPEND
1188 Architecture has old sigsuspend(2) syscall, of one-argument variety
1190 config OLD_SIGSUSPEND3
1193 Even weirder antique ABI - three-argument sigsuspend(2)
1195 config OLD_SIGACTION
1198 Architecture has old sigaction(2) syscall. Nope, not the same
1199 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
1200 but fairly different variant of sigaction(2), thanks to OSF/1
1203 config COMPAT_OLD_SIGACTION
1206 config COMPAT_32BIT_TIME
1207 bool "Provide system calls for 32-bit time_t"
1208 default !64BIT || COMPAT
1210 This enables 32 bit time_t support in addition to 64 bit time_t support.
1211 This is relevant on all 32-bit architectures, and 64-bit architectures
1212 as part of compat syscall handling.
1214 config ARCH_NO_PREEMPT
1217 config ARCH_EPHEMERAL_INODES
1220 An arch should select this symbol if it doesn't keep track of inode
1221 instances on its own, but instead relies on something else (e.g. the
1222 host kernel for an UML kernel).
1224 config ARCH_SUPPORTS_RT
1227 config CPU_NO_EFFICIENT_FFS
1230 config HAVE_ARCH_VMAP_STACK
1233 An arch should select this symbol if it can support kernel stacks
1234 in vmalloc space. This means:
1236 - vmalloc space must be large enough to hold many kernel stacks.
1237 This may rule out many 32-bit architectures.
1239 - Stacks in vmalloc space need to work reliably. For example, if
1240 vmap page tables are created on demand, either this mechanism
1241 needs to work while the stack points to a virtual address with
1242 unpopulated page tables or arch code (switch_to() and switch_mm(),
1243 most likely) needs to ensure that the stack's page table entries
1244 are populated before running on a possibly unpopulated stack.
1246 - If the stack overflows into a guard page, something reasonable
1247 should happen. The definition of "reasonable" is flexible, but
1248 instantly rebooting without logging anything would be unfriendly.
1252 bool "Use a virtually-mapped stack"
1253 depends on HAVE_ARCH_VMAP_STACK
1254 depends on !KASAN || KASAN_HW_TAGS || KASAN_VMALLOC
1256 Enable this if you want the use virtually-mapped kernel stacks
1257 with guard pages. This causes kernel stack overflows to be
1258 caught immediately rather than causing difficult-to-diagnose
1261 To use this with software KASAN modes, the architecture must support
1262 backing virtual mappings with real shadow memory, and KASAN_VMALLOC
1265 config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1268 An arch should select this symbol if it can support kernel stack
1269 offset randomization with calls to add_random_kstack_offset()
1270 during syscall entry and choose_random_kstack_offset() during
1271 syscall exit. Careful removal of -fstack-protector-strong and
1272 -fstack-protector should also be applied to the entry code and
1273 closely examined, as the artificial stack bump looks like an array
1274 to the compiler, so it will attempt to add canary checks regardless
1275 of the static branch state.
1277 config RANDOMIZE_KSTACK_OFFSET
1278 bool "Support for randomizing kernel stack offset on syscall entry" if EXPERT
1280 depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1281 depends on INIT_STACK_NONE || !CC_IS_CLANG || CLANG_VERSION >= 140000
1283 The kernel stack offset can be randomized (after pt_regs) by
1284 roughly 5 bits of entropy, frustrating memory corruption
1285 attacks that depend on stack address determinism or
1286 cross-syscall address exposures.
1288 The feature is controlled via the "randomize_kstack_offset=on/off"
1289 kernel boot param, and if turned off has zero overhead due to its use
1290 of static branches (see JUMP_LABEL).
1294 config RANDOMIZE_KSTACK_OFFSET_DEFAULT
1295 bool "Default state of kernel stack offset randomization"
1296 depends on RANDOMIZE_KSTACK_OFFSET
1298 Kernel stack offset randomization is controlled by kernel boot param
1299 "randomize_kstack_offset=on/off", and this config chooses the default
1302 config ARCH_OPTIONAL_KERNEL_RWX
1305 config ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1308 config ARCH_HAS_STRICT_KERNEL_RWX
1311 config STRICT_KERNEL_RWX
1312 bool "Make kernel text and rodata read-only" if ARCH_OPTIONAL_KERNEL_RWX
1313 depends on ARCH_HAS_STRICT_KERNEL_RWX
1314 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1316 If this is set, kernel text and rodata memory will be made read-only,
1317 and non-text memory will be made non-executable. This provides
1318 protection against certain security exploits (e.g. executing the heap
1321 These features are considered standard security practice these days.
1322 You should say Y here in almost all cases.
1324 config ARCH_HAS_STRICT_MODULE_RWX
1327 config STRICT_MODULE_RWX
1328 bool "Set loadable kernel module data as NX and text as RO" if ARCH_OPTIONAL_KERNEL_RWX
1329 depends on ARCH_HAS_STRICT_MODULE_RWX && MODULES
1330 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1332 If this is set, module text and rodata memory will be made read-only,
1333 and non-text memory will be made non-executable. This provides
1334 protection against certain security exploits (e.g. writing to text)
1336 # select if the architecture provides an asm/dma-direct.h header
1337 config ARCH_HAS_PHYS_TO_DMA
1340 config HAVE_ARCH_COMPILER_H
1343 An architecture can select this if it provides an
1344 asm/compiler.h header that should be included after
1345 linux/compiler-*.h in order to override macro definitions that those
1346 headers generally provide.
1348 config HAVE_ARCH_PREL32_RELOCATIONS
1351 May be selected by an architecture if it supports place-relative
1352 32-bit relocations, both in the toolchain and in the module loader,
1353 in which case relative references can be used in special sections
1354 for PCI fixup, initcalls etc which are only half the size on 64 bit
1355 architectures, and don't require runtime relocation on relocatable
1358 config ARCH_USE_MEMREMAP_PROT
1361 config LOCK_EVENT_COUNTS
1362 bool "Locking event counts collection"
1365 Enable light-weight counting of various locking related events
1366 in the system with minimal performance impact. This reduces
1367 the chance of application behavior change because of timing
1368 differences. The counts are reported via debugfs.
1370 # Select if the architecture has support for applying RELR relocations.
1371 config ARCH_HAS_RELR
1375 bool "Use RELR relocation packing"
1376 depends on ARCH_HAS_RELR && TOOLS_SUPPORT_RELR
1379 Store the kernel's dynamic relocations in the RELR relocation packing
1380 format. Requires a compatible linker (LLD supports this feature), as
1381 well as compatible NM and OBJCOPY utilities (llvm-nm and llvm-objcopy
1384 config ARCH_HAS_MEM_ENCRYPT
1387 config ARCH_HAS_CC_PLATFORM
1390 config HAVE_SPARSE_SYSCALL_NR
1393 An architecture should select this if its syscall numbering is sparse
1394 to save space. For example, MIPS architecture has a syscall array with
1395 entries at 4000, 5000 and 6000 locations. This option turns on syscall
1396 related optimizations for a given architecture.
1398 config ARCH_HAS_VDSO_DATA
1401 config HAVE_STATIC_CALL
1404 config HAVE_STATIC_CALL_INLINE
1406 depends on HAVE_STATIC_CALL
1409 config HAVE_PREEMPT_DYNAMIC
1412 config HAVE_PREEMPT_DYNAMIC_CALL
1414 depends on HAVE_STATIC_CALL
1415 select HAVE_PREEMPT_DYNAMIC
1417 An architecture should select this if it can handle the preemption
1418 model being selected at boot time using static calls.
1420 Where an architecture selects HAVE_STATIC_CALL_INLINE, any call to a
1421 preemption function will be patched directly.
1423 Where an architecture does not select HAVE_STATIC_CALL_INLINE, any
1424 call to a preemption function will go through a trampoline, and the
1425 trampoline will be patched.
1427 It is strongly advised to support inline static call to avoid any
1430 config HAVE_PREEMPT_DYNAMIC_KEY
1432 depends on HAVE_ARCH_JUMP_LABEL
1433 select HAVE_PREEMPT_DYNAMIC
1435 An architecture should select this if it can handle the preemption
1436 model being selected at boot time using static keys.
1438 Each preemption function will be given an early return based on a
1439 static key. This should have slightly lower overhead than non-inline
1440 static calls, as this effectively inlines each trampoline into the
1441 start of its callee. This may avoid redundant work, and may
1442 integrate better with CFI schemes.
1444 This will have greater overhead than using inline static calls as
1445 the call to the preemption function cannot be entirely elided.
1447 config ARCH_WANT_LD_ORPHAN_WARN
1450 An arch should select this symbol once all linker sections are explicitly
1451 included, size-asserted, or discarded in the linker scripts. This is
1452 important because we never want expected sections to be placed heuristically
1453 by the linker, since the locations of such sections can change between linker
1456 config HAVE_ARCH_PFN_VALID
1459 config ARCH_SUPPORTS_DEBUG_PAGEALLOC
1462 config ARCH_SUPPORTS_PAGE_TABLE_CHECK
1465 config ARCH_SPLIT_ARG64
1468 If a 32-bit architecture requires 64-bit arguments to be split into
1469 pairs of 32-bit arguments, select this option.
1471 config ARCH_HAS_ELFCORE_COMPAT
1474 config ARCH_HAS_PARANOID_L1D_FLUSH
1477 config ARCH_HAVE_TRACE_MMIO_ACCESS
1480 config DYNAMIC_SIGFRAME
1483 # Select, if arch has a named attribute group bound to NUMA device nodes.
1484 config HAVE_ARCH_NODE_DEV_GROUP
1487 config ARCH_HAS_NONLEAF_PMD_YOUNG
1490 Architectures that select this option are capable of setting the
1491 accessed bit in non-leaf PMD entries when using them as part of linear
1492 address translations. Page table walkers that clear the accessed bit
1493 may use this capability to reduce their search space.
1495 source "kernel/gcov/Kconfig"
1497 source "scripts/gcc-plugins/Kconfig"
1499 config FUNCTION_ALIGNMENT_4B
1502 config FUNCTION_ALIGNMENT_8B
1505 config FUNCTION_ALIGNMENT_16B
1508 config FUNCTION_ALIGNMENT_32B
1511 config FUNCTION_ALIGNMENT_64B
1514 config FUNCTION_ALIGNMENT
1516 default 64 if FUNCTION_ALIGNMENT_64B
1517 default 32 if FUNCTION_ALIGNMENT_32B
1518 default 16 if FUNCTION_ALIGNMENT_16B
1519 default 8 if FUNCTION_ALIGNMENT_8B
1520 default 4 if FUNCTION_ALIGNMENT_4B