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"
14 config ARCH_HAS_SUBPAGE_FAULTS
17 Select if the architecture can check permissions at sub-page
18 granularity (e.g. arm64 MTE). The probe_user_*() functions
24 config SMT_NUM_THREADS_DYNAMIC
27 # Selected by HOTPLUG_CORE_SYNC_DEAD or HOTPLUG_CORE_SYNC_FULL
28 config HOTPLUG_CORE_SYNC
31 # Basic CPU dead synchronization selected by architecture
32 config HOTPLUG_CORE_SYNC_DEAD
34 select HOTPLUG_CORE_SYNC
36 # Full CPU synchronization with alive state selected by architecture
37 config HOTPLUG_CORE_SYNC_FULL
39 select HOTPLUG_CORE_SYNC_DEAD if HOTPLUG_CPU
40 select HOTPLUG_CORE_SYNC
42 config HOTPLUG_SPLIT_STARTUP
44 select HOTPLUG_CORE_SYNC_FULL
46 config HOTPLUG_PARALLEL
48 select HOTPLUG_SPLIT_STARTUP
56 depends on HAVE_KPROBES
58 select TASKS_RCU if PREEMPTION
60 Kprobes allows you to trap at almost any kernel address and
61 execute a callback function. register_kprobe() establishes
62 a probepoint and specifies the callback. Kprobes is useful
63 for kernel debugging, non-intrusive instrumentation and testing.
67 bool "Optimize very unlikely/likely branches"
68 depends on HAVE_ARCH_JUMP_LABEL
69 select OBJTOOL if HAVE_JUMP_LABEL_HACK
71 This option enables a transparent branch optimization that
72 makes certain almost-always-true or almost-always-false branch
73 conditions even cheaper to execute within the kernel.
75 Certain performance-sensitive kernel code, such as trace points,
76 scheduler functionality, networking code and KVM have such
77 branches and include support for this optimization technique.
79 If it is detected that the compiler has support for "asm goto",
80 the kernel will compile such branches with just a nop
81 instruction. When the condition flag is toggled to true, the
82 nop will be converted to a jump instruction to execute the
83 conditional block of instructions.
85 This technique lowers overhead and stress on the branch prediction
86 of the processor and generally makes the kernel faster. The update
87 of the condition is slower, but those are always very rare.
89 ( On 32-bit x86, the necessary options added to the compiler
90 flags may increase the size of the kernel slightly. )
92 config STATIC_KEYS_SELFTEST
93 bool "Static key selftest"
96 Boot time self-test of the branch patching code.
98 config STATIC_CALL_SELFTEST
99 bool "Static call selftest"
100 depends on HAVE_STATIC_CALL
102 Boot time self-test of the call patching code.
106 depends on KPROBES && HAVE_OPTPROBES
107 select TASKS_RCU if PREEMPTION
109 config KPROBES_ON_FTRACE
111 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
112 depends on DYNAMIC_FTRACE_WITH_REGS
114 If function tracer is enabled and the arch supports full
115 passing of pt_regs to function tracing, then kprobes can
116 optimize on top of function tracing.
120 depends on ARCH_SUPPORTS_UPROBES
122 Uprobes is the user-space counterpart to kprobes: they
123 enable instrumentation applications (such as 'perf probe')
124 to establish unintrusive probes in user-space binaries and
125 libraries, by executing handler functions when the probes
126 are hit by user-space applications.
128 ( These probes come in the form of single-byte breakpoints,
129 managed by the kernel and kept transparent to the probed
132 config HAVE_64BIT_ALIGNED_ACCESS
133 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
135 Some architectures require 64 bit accesses to be 64 bit
136 aligned, which also requires structs containing 64 bit values
137 to be 64 bit aligned too. This includes some 32 bit
138 architectures which can do 64 bit accesses, as well as 64 bit
139 architectures without unaligned access.
141 This symbol should be selected by an architecture if 64 bit
142 accesses are required to be 64 bit aligned in this way even
143 though it is not a 64 bit architecture.
145 See Documentation/core-api/unaligned-memory-access.rst for
146 more information on the topic of unaligned memory accesses.
148 config HAVE_EFFICIENT_UNALIGNED_ACCESS
151 Some architectures are unable to perform unaligned accesses
152 without the use of get_unaligned/put_unaligned. Others are
153 unable to perform such accesses efficiently (e.g. trap on
154 unaligned access and require fixing it up in the exception
157 This symbol should be selected by an architecture if it can
158 perform unaligned accesses efficiently to allow different
159 code paths to be selected for these cases. Some network
160 drivers, for example, could opt to not fix up alignment
161 problems with received packets if doing so would not help
164 See Documentation/core-api/unaligned-memory-access.rst for more
165 information on the topic of unaligned memory accesses.
167 config ARCH_USE_BUILTIN_BSWAP
170 Modern versions of GCC (since 4.4) have builtin functions
171 for handling byte-swapping. Using these, instead of the old
172 inline assembler that the architecture code provides in the
173 __arch_bswapXX() macros, allows the compiler to see what's
174 happening and offers more opportunity for optimisation. In
175 particular, the compiler will be able to combine the byteswap
176 with a nearby load or store and use load-and-swap or
177 store-and-swap instructions if the architecture has them. It
178 should almost *never* result in code which is worse than the
179 hand-coded assembler in <asm/swab.h>. But just in case it
180 does, the use of the builtins is optional.
182 Any architecture with load-and-swap or store-and-swap
183 instructions should set this. And it shouldn't hurt to set it
184 on architectures that don't have such instructions.
188 depends on KPROBES && (HAVE_KRETPROBES || HAVE_RETHOOK)
190 config KRETPROBE_ON_RETHOOK
192 depends on HAVE_RETHOOK
193 depends on KRETPROBES
196 config USER_RETURN_NOTIFIER
198 depends on HAVE_USER_RETURN_NOTIFIER
200 Provide a kernel-internal notification when a cpu is about to
203 config HAVE_IOREMAP_PROT
209 config HAVE_KRETPROBES
212 config HAVE_OPTPROBES
215 config HAVE_KPROBES_ON_FTRACE
218 config ARCH_CORRECT_STACKTRACE_ON_KRETPROBE
221 Since kretprobes modifies return address on the stack, the
222 stacktrace may see the kretprobe trampoline address instead
223 of correct one. If the architecture stacktrace code and
224 unwinder can adjust such entries, select this configuration.
226 config HAVE_FUNCTION_ERROR_INJECTION
232 config HAVE_FUNCTION_DESCRIPTORS
235 config TRACE_IRQFLAGS_SUPPORT
238 config TRACE_IRQFLAGS_NMI_SUPPORT
242 # An arch should select this if it provides all these things:
244 # task_pt_regs() in asm/processor.h or asm/ptrace.h
245 # arch_has_single_step() if there is hardware single-step support
246 # arch_has_block_step() if there is hardware block-step support
247 # asm/syscall.h supplying asm-generic/syscall.h interface
248 # linux/regset.h user_regset interfaces
249 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
250 # TIF_SYSCALL_TRACE calls ptrace_report_syscall_{entry,exit}
251 # TIF_NOTIFY_RESUME calls resume_user_mode_work()
253 config HAVE_ARCH_TRACEHOOK
256 config HAVE_DMA_CONTIGUOUS
259 config GENERIC_SMP_IDLE_THREAD
262 config GENERIC_IDLE_POLL_SETUP
265 config ARCH_HAS_FORTIFY_SOURCE
268 An architecture should select this when it can successfully
269 build and run with CONFIG_FORTIFY_SOURCE.
272 # Select if the arch provides a historic keepinit alias for the retain_initrd
273 # command line option
275 config ARCH_HAS_KEEPINITRD
278 # Select if arch has all set_memory_ro/rw/x/nx() functions in asm/cacheflush.h
279 config ARCH_HAS_SET_MEMORY
282 # Select if arch has all set_direct_map_invalid/default() functions
283 config ARCH_HAS_SET_DIRECT_MAP
287 # Select if the architecture provides the arch_dma_set_uncached symbol to
288 # either provide an uncached segment alias for a DMA allocation, or
289 # to remap the page tables in place.
291 config ARCH_HAS_DMA_SET_UNCACHED
295 # Select if the architectures provides the arch_dma_clear_uncached symbol
296 # to undo an in-place page table remap for uncached access.
298 config ARCH_HAS_DMA_CLEAR_UNCACHED
301 config ARCH_HAS_CPU_FINALIZE_INIT
304 config HAVE_ARCH_THREAD_STRUCT_WHITELIST
307 An architecture should select this to provide hardened usercopy
308 knowledge about what region of the thread_struct should be
309 whitelisted for copying to userspace. Normally this is only the
310 FPU registers. Specifically, arch_thread_struct_whitelist()
311 should be implemented. Without this, the entire thread_struct
312 field in task_struct will be left whitelisted.
314 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
315 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
318 config ARCH_WANTS_NO_INSTR
321 An architecture should select this if the noinstr macro is being used on
322 functions to denote that the toolchain should avoid instrumenting such
323 functions and is required for correctness.
325 config ARCH_32BIT_OFF_T
329 All new 32-bit architectures should have 64-bit off_t type on
330 userspace side which corresponds to the loff_t kernel type. This
331 is the requirement for modern ABIs. Some existing architectures
332 still support 32-bit off_t. This option is enabled for all such
333 architectures explicitly.
335 # Selected by 64 bit architectures which have a 32 bit f_tinode in struct ustat
336 config ARCH_32BIT_USTAT_F_TINODE
339 config HAVE_ASM_MODVERSIONS
342 This symbol should be selected by an architecture if it provides
343 <asm/asm-prototypes.h> to support the module versioning for symbols
344 exported from assembly code.
346 config HAVE_REGS_AND_STACK_ACCESS_API
349 This symbol should be selected by an architecture if it supports
350 the API needed to access registers and stack entries from pt_regs,
351 declared in asm/ptrace.h
352 For example the kprobes-based event tracer needs this API.
356 depends on HAVE_REGS_AND_STACK_ACCESS_API
358 This symbol should be selected by an architecture if it
359 supports an implementation of restartable sequences.
364 This symbol should be selected by an architecture if it
367 config HAVE_FUNCTION_ARG_ACCESS_API
370 This symbol should be selected by an architecture if it supports
371 the API needed to access function arguments from pt_regs,
372 declared in asm/ptrace.h
374 config HAVE_HW_BREAKPOINT
376 depends on PERF_EVENTS
378 config HAVE_MIXED_BREAKPOINTS_REGS
380 depends on HAVE_HW_BREAKPOINT
382 Depending on the arch implementation of hardware breakpoints,
383 some of them have separate registers for data and instruction
384 breakpoints addresses, others have mixed registers to store
385 them but define the access type in a control register.
386 Select this option if your arch implements breakpoints under the
389 config HAVE_USER_RETURN_NOTIFIER
392 config HAVE_PERF_EVENTS_NMI
395 System hardware can generate an NMI using the perf event
396 subsystem. Also has support for calculating CPU cycle events
397 to determine how many clock cycles in a given period.
399 config HAVE_HARDLOCKUP_DETECTOR_PERF
401 depends on HAVE_PERF_EVENTS_NMI
403 The arch chooses to use the generic perf-NMI-based hardlockup
404 detector. Must define HAVE_PERF_EVENTS_NMI.
406 config HAVE_HARDLOCKUP_DETECTOR_ARCH
409 The arch provides its own hardlockup detector implementation instead
412 It uses the same command line parameters, and sysctl interface,
413 as the generic hardlockup detectors.
415 config HAVE_PERF_REGS
418 Support selective register dumps for perf events. This includes
419 bit-mapping of each registers and a unique architecture id.
421 config HAVE_PERF_USER_STACK_DUMP
424 Support user stack dumps for perf event samples. This needs
425 access to the user stack pointer which is not unified across
428 config HAVE_ARCH_JUMP_LABEL
431 config HAVE_ARCH_JUMP_LABEL_RELATIVE
434 config MMU_GATHER_TABLE_FREE
437 config MMU_GATHER_RCU_TABLE_FREE
439 select MMU_GATHER_TABLE_FREE
441 config MMU_GATHER_PAGE_SIZE
444 config MMU_GATHER_NO_RANGE
446 select MMU_GATHER_MERGE_VMAS
448 config MMU_GATHER_NO_FLUSH_CACHE
451 config MMU_GATHER_MERGE_VMAS
454 config MMU_GATHER_NO_GATHER
456 depends on MMU_GATHER_TABLE_FREE
458 config ARCH_WANT_IRQS_OFF_ACTIVATE_MM
461 Temporary select until all architectures can be converted to have
462 irqs disabled over activate_mm. Architectures that do IPI based TLB
463 shootdowns should enable this.
465 # Use normal mm refcounting for MMU_LAZY_TLB kernel thread references.
466 # MMU_LAZY_TLB_REFCOUNT=n can improve the scalability of context switching
467 # to/from kernel threads when the same mm is running on a lot of CPUs (a large
468 # multi-threaded application), by reducing contention on the mm refcount.
470 # This can be disabled if the architecture ensures no CPUs are using an mm as a
471 # "lazy tlb" beyond its final refcount (i.e., by the time __mmdrop frees the mm
472 # or its kernel page tables). This could be arranged by arch_exit_mmap(), or
473 # final exit(2) TLB flush, for example.
475 # To implement this, an arch *must*:
476 # Ensure the _lazy_tlb variants of mmgrab/mmdrop are used when manipulating
477 # the lazy tlb reference of a kthread's ->active_mm (non-arch code has been
478 # converted already).
479 config MMU_LAZY_TLB_REFCOUNT
481 depends on !MMU_LAZY_TLB_SHOOTDOWN
483 # This option allows MMU_LAZY_TLB_REFCOUNT=n. It ensures no CPUs are using an
484 # mm as a lazy tlb beyond its last reference count, by shooting down these
485 # users before the mm is deallocated. __mmdrop() first IPIs all CPUs that may
486 # be using the mm as a lazy tlb, so that they may switch themselves to using
487 # init_mm for their active mm. mm_cpumask(mm) is used to determine which CPUs
488 # may be using mm as a lazy tlb mm.
490 # To implement this, an arch *must*:
491 # - At the time of the final mmdrop of the mm, ensure mm_cpumask(mm) contains
492 # at least all possible CPUs in which the mm is lazy.
493 # - It must meet the requirements for MMU_LAZY_TLB_REFCOUNT=n (see above).
494 config MMU_LAZY_TLB_SHOOTDOWN
497 config ARCH_HAVE_NMI_SAFE_CMPXCHG
500 config ARCH_HAS_NMI_SAFE_THIS_CPU_OPS
503 config HAVE_ALIGNED_STRUCT_PAGE
506 This makes sure that struct pages are double word aligned and that
507 e.g. the SLUB allocator can perform double word atomic operations
508 on a struct page for better performance. However selecting this
509 might increase the size of a struct page by a word.
511 config HAVE_CMPXCHG_LOCAL
514 config HAVE_CMPXCHG_DOUBLE
517 config ARCH_WEAK_RELEASE_ACQUIRE
520 config ARCH_WANT_IPC_PARSE_VERSION
523 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
526 config ARCH_WANT_OLD_COMPAT_IPC
527 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
530 config HAVE_ARCH_SECCOMP
533 An arch should select this symbol to support seccomp mode 1 (the fixed
534 syscall policy), and must provide an overrides for __NR_seccomp_sigreturn,
535 and compat syscalls if the asm-generic/seccomp.h defaults need adjustment:
536 - __NR_seccomp_read_32
537 - __NR_seccomp_write_32
538 - __NR_seccomp_exit_32
539 - __NR_seccomp_sigreturn_32
541 config HAVE_ARCH_SECCOMP_FILTER
543 select HAVE_ARCH_SECCOMP
545 An arch should select this symbol if it provides all of these things:
546 - all the requirements for HAVE_ARCH_SECCOMP
548 - syscall_get_arguments()
550 - syscall_set_return_value()
551 - SIGSYS siginfo_t support
552 - secure_computing is called from a ptrace_event()-safe context
553 - secure_computing return value is checked and a return value of -1
554 results in the system call being skipped immediately.
555 - seccomp syscall wired up
556 - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE,
557 SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If
558 COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too.
561 prompt "Enable seccomp to safely execute untrusted bytecode"
563 depends on HAVE_ARCH_SECCOMP
565 This kernel feature is useful for number crunching applications
566 that may need to handle untrusted bytecode during their
567 execution. By using pipes or other transports made available
568 to the process as file descriptors supporting the read/write
569 syscalls, it's possible to isolate those applications in their
570 own address space using seccomp. Once seccomp is enabled via
571 prctl(PR_SET_SECCOMP) or the seccomp() syscall, it cannot be
572 disabled and the task is only allowed to execute a few safe
573 syscalls defined by each seccomp mode.
577 config SECCOMP_FILTER
579 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
581 Enable tasks to build secure computing environments defined
582 in terms of Berkeley Packet Filter programs which implement
583 task-defined system call filtering polices.
585 See Documentation/userspace-api/seccomp_filter.rst for details.
587 config SECCOMP_CACHE_DEBUG
588 bool "Show seccomp filter cache status in /proc/pid/seccomp_cache"
589 depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR
592 This enables the /proc/pid/seccomp_cache interface to monitor
593 seccomp cache data. The file format is subject to change. Reading
594 the file requires CAP_SYS_ADMIN.
596 This option is for debugging only. Enabling presents the risk that
597 an adversary may be able to infer the seccomp filter logic.
601 config HAVE_ARCH_STACKLEAK
604 An architecture should select this if it has the code which
605 fills the used part of the kernel stack with the STACKLEAK_POISON
606 value before returning from system calls.
608 config HAVE_STACKPROTECTOR
611 An arch should select this symbol if:
612 - it has implemented a stack canary (e.g. __stack_chk_guard)
614 config STACKPROTECTOR
615 bool "Stack Protector buffer overflow detection"
616 depends on HAVE_STACKPROTECTOR
617 depends on $(cc-option,-fstack-protector)
620 This option turns on the "stack-protector" GCC feature. This
621 feature puts, at the beginning of functions, a canary value on
622 the stack just before the return address, and validates
623 the value just before actually returning. Stack based buffer
624 overflows (that need to overwrite this return address) now also
625 overwrite the canary, which gets detected and the attack is then
626 neutralized via a kernel panic.
628 Functions will have the stack-protector canary logic added if they
629 have an 8-byte or larger character array on the stack.
631 This feature requires gcc version 4.2 or above, or a distribution
632 gcc with the feature backported ("-fstack-protector").
634 On an x86 "defconfig" build, this feature adds canary checks to
635 about 3% of all kernel functions, which increases kernel code size
638 config STACKPROTECTOR_STRONG
639 bool "Strong Stack Protector"
640 depends on STACKPROTECTOR
641 depends on $(cc-option,-fstack-protector-strong)
644 Functions will have the stack-protector canary logic added in any
645 of the following conditions:
647 - local variable's address used as part of the right hand side of an
648 assignment or function argument
649 - local variable is an array (or union containing an array),
650 regardless of array type or length
651 - uses register local variables
653 This feature requires gcc version 4.9 or above, or a distribution
654 gcc with the feature backported ("-fstack-protector-strong").
656 On an x86 "defconfig" build, this feature adds canary checks to
657 about 20% of all kernel functions, which increases the kernel code
660 config ARCH_SUPPORTS_SHADOW_CALL_STACK
663 An architecture should select this if it supports the compiler's
664 Shadow Call Stack and implements runtime support for shadow stack
667 config SHADOW_CALL_STACK
668 bool "Shadow Call Stack"
669 depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
670 depends on DYNAMIC_FTRACE_WITH_ARGS || DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
672 This option enables the compiler's Shadow Call Stack, which
673 uses a shadow stack to protect function return addresses from
674 being overwritten by an attacker. More information can be found
675 in the compiler's documentation:
677 - Clang: https://clang.llvm.org/docs/ShadowCallStack.html
678 - GCC: https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html#Instrumentation-Options
680 Note that security guarantees in the kernel differ from the
681 ones documented for user space. The kernel must store addresses
682 of shadow stacks in memory, which means an attacker capable of
683 reading and writing arbitrary memory may be able to locate them
684 and hijack control flow by modifying the stacks.
689 Set by the arch code if it relies on code patching to insert the
690 shadow call stack push and pop instructions rather than on the
696 Selected if the kernel will be built using the compiler's LTO feature.
702 Selected if the kernel will be built using Clang's LTO feature.
704 config ARCH_SUPPORTS_LTO_CLANG
707 An architecture should select this option if it supports:
708 - compiling with Clang,
709 - compiling inline assembly with Clang's integrated assembler,
710 - and linking with LLD.
712 config ARCH_SUPPORTS_LTO_CLANG_THIN
715 An architecture should select this option if it can support Clang's
720 depends on CC_IS_CLANG && LD_IS_LLD && AS_IS_LLVM
721 depends on $(success,$(NM) --help | head -n 1 | grep -qi llvm)
722 depends on $(success,$(AR) --help | head -n 1 | grep -qi llvm)
723 depends on ARCH_SUPPORTS_LTO_CLANG
724 depends on !FTRACE_MCOUNT_USE_RECORDMCOUNT
725 # https://github.com/ClangBuiltLinux/linux/issues/1721
726 depends on (!KASAN || KASAN_HW_TAGS || CLANG_VERSION >= 170000) || !DEBUG_INFO
727 depends on (!KCOV || CLANG_VERSION >= 170000) || !DEBUG_INFO
728 depends on !GCOV_KERNEL
730 The compiler and Kconfig options support building with Clang's
734 prompt "Link Time Optimization (LTO)"
737 This option enables Link Time Optimization (LTO), which allows the
738 compiler to optimize binaries globally.
740 If unsure, select LTO_NONE. Note that LTO is very resource-intensive
741 so it's disabled by default.
746 Build the kernel normally, without Link Time Optimization (LTO).
748 config LTO_CLANG_FULL
749 bool "Clang Full LTO (EXPERIMENTAL)"
750 depends on HAS_LTO_CLANG
751 depends on !COMPILE_TEST
754 This option enables Clang's full Link Time Optimization (LTO), which
755 allows the compiler to optimize the kernel globally. If you enable
756 this option, the compiler generates LLVM bitcode instead of ELF
757 object files, and the actual compilation from bitcode happens at
758 the LTO link step, which may take several minutes depending on the
759 kernel configuration. More information can be found from LLVM's
762 https://llvm.org/docs/LinkTimeOptimization.html
764 During link time, this option can use a large amount of RAM, and
765 may take much longer than the ThinLTO option.
767 config LTO_CLANG_THIN
768 bool "Clang ThinLTO (EXPERIMENTAL)"
769 depends on HAS_LTO_CLANG && ARCH_SUPPORTS_LTO_CLANG_THIN
772 This option enables Clang's ThinLTO, which allows for parallel
773 optimization and faster incremental compiles compared to the
774 CONFIG_LTO_CLANG_FULL option. More information can be found
775 from Clang's documentation:
777 https://clang.llvm.org/docs/ThinLTO.html
782 config ARCH_SUPPORTS_CFI_CLANG
785 An architecture should select this option if it can support Clang's
786 Control-Flow Integrity (CFI) checking.
788 config ARCH_USES_CFI_TRAPS
792 bool "Use Clang's Control Flow Integrity (CFI)"
793 depends on ARCH_SUPPORTS_CFI_CLANG
794 depends on $(cc-option,-fsanitize=kcfi)
796 This option enables Clang’s forward-edge Control Flow Integrity
797 (CFI) checking, where the compiler injects a runtime check to each
798 indirect function call to ensure the target is a valid function with
799 the correct static type. This restricts possible call targets and
800 makes it more difficult for an attacker to exploit bugs that allow
801 the modification of stored function pointers. More information can be
802 found from Clang's documentation:
804 https://clang.llvm.org/docs/ControlFlowIntegrity.html
806 config CFI_PERMISSIVE
807 bool "Use CFI in permissive mode"
810 When selected, Control Flow Integrity (CFI) violations result in a
811 warning instead of a kernel panic. This option should only be used
812 for finding indirect call type mismatches during development.
816 config HAVE_ARCH_WITHIN_STACK_FRAMES
819 An architecture should select this if it can walk the kernel stack
820 frames to determine if an object is part of either the arguments
821 or local variables (i.e. that it excludes saved return addresses,
822 and similar) by implementing an inline arch_within_stack_frames(),
823 which is used by CONFIG_HARDENED_USERCOPY.
825 config HAVE_CONTEXT_TRACKING_USER
828 Provide kernel/user boundaries probes necessary for subsystems
829 that need it, such as userspace RCU extended quiescent state.
830 Syscalls need to be wrapped inside user_exit()-user_enter(), either
831 optimized behind static key or through the slow path using TIF_NOHZ
832 flag. Exceptions handlers must be wrapped as well. Irqs are already
833 protected inside ct_irq_enter/ct_irq_exit() but preemption or signal
834 handling on irq exit still need to be protected.
836 config HAVE_CONTEXT_TRACKING_USER_OFFSTACK
839 Architecture neither relies on exception_enter()/exception_exit()
840 nor on schedule_user(). Also preempt_schedule_notrace() and
841 preempt_schedule_irq() can't be called in a preemptible section
842 while context tracking is CONTEXT_USER. This feature reflects a sane
843 entry implementation where the following requirements are met on
844 critical entry code, ie: before user_exit() or after user_enter():
846 - Critical entry code isn't preemptible (or better yet:
848 - No use of RCU read side critical sections, unless ct_nmi_enter()
850 - No use of instrumentation, unless instrumentation_begin() got
856 Arch relies on TIF_NOHZ and syscall slow path to implement context
857 tracking calls to user_enter()/user_exit().
859 config HAVE_VIRT_CPU_ACCOUNTING
862 config HAVE_VIRT_CPU_ACCOUNTING_IDLE
865 Architecture has its own way to account idle CPU time and therefore
866 doesn't implement vtime_account_idle().
868 config ARCH_HAS_SCALED_CPUTIME
871 config HAVE_VIRT_CPU_ACCOUNTING_GEN
875 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
876 Before enabling this option, arch code must be audited
877 to ensure there are no races in concurrent read/write of
878 cputime_t. For example, reading/writing 64-bit cputime_t on
879 some 32-bit arches may require multiple accesses, so proper
880 locking is needed to protect against concurrent accesses.
882 config HAVE_IRQ_TIME_ACCOUNTING
885 Archs need to ensure they use a high enough resolution clock to
886 support irq time accounting and then call enable_sched_clock_irqtime().
891 Architectures that select this are able to move page tables at the
892 PUD level. If there are only 3 page table levels, the move effectively
893 happens at the PGD level.
898 Archs that select this are able to move page tables at the PMD level.
900 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
903 config HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
906 config HAVE_ARCH_HUGE_VMAP
910 # Archs that select this would be capable of PMD-sized vmaps (i.e.,
911 # arch_vmap_pmd_supported() returns true). The VM_ALLOW_HUGE_VMAP flag
912 # must be used to enable allocations to use hugepages.
914 config HAVE_ARCH_HUGE_VMALLOC
915 depends on HAVE_ARCH_HUGE_VMAP
918 config ARCH_WANT_HUGE_PMD_SHARE
921 # Archs that want to use pmd_mkwrite on kernel memory need it defined even
922 # if there are no userspace memory management features that use it
923 config ARCH_WANT_KERNEL_PMD_MKWRITE
926 config ARCH_WANT_PMD_MKWRITE
927 def_bool TRANSPARENT_HUGEPAGE || ARCH_WANT_KERNEL_PMD_MKWRITE
929 config HAVE_ARCH_SOFT_DIRTY
932 config HAVE_MOD_ARCH_SPECIFIC
935 The arch uses struct mod_arch_specific to store data. Many arches
936 just need a simple module loader without arch specific data - those
937 should not enable this.
939 config MODULES_USE_ELF_RELA
942 Modules only use ELF RELA relocations. Modules with ELF REL
943 relocations will give an error.
945 config MODULES_USE_ELF_REL
948 Modules only use ELF REL relocations. Modules with ELF RELA
949 relocations will give an error.
951 config ARCH_WANTS_MODULES_DATA_IN_VMALLOC
954 For architectures like powerpc/32 which have constraints on module
955 allocation and need to allocate module data outside of module area.
957 config HAVE_IRQ_EXIT_ON_IRQ_STACK
960 Architecture doesn't only execute the irq handler on the irq stack
961 but also irq_exit(). This way we can process softirqs on this irq
962 stack instead of switching to a new one when we call __do_softirq()
963 in the end of an hardirq.
964 This spares a stack switch and improves cache usage on softirq
967 config HAVE_SOFTIRQ_ON_OWN_STACK
970 Architecture provides a function to run __do_softirq() on a
973 config SOFTIRQ_ON_OWN_STACK
974 def_bool HAVE_SOFTIRQ_ON_OWN_STACK && !PREEMPT_RT
976 config ALTERNATE_USER_ADDRESS_SPACE
979 Architectures set this when the CPU uses separate address
980 spaces for kernel and user space pointers. In this case, the
981 access_ok() check on a __user pointer is skipped.
983 config PGTABLE_LEVELS
987 config ARCH_HAS_ELF_RANDOMIZE
990 An architecture supports choosing randomized locations for
991 stack, mmap, brk, and ET_DYN. Defined functions:
993 - arch_randomize_brk()
995 config HAVE_ARCH_MMAP_RND_BITS
998 An arch should select this symbol if it supports setting a variable
999 number of bits for use in establishing the base address for mmap
1000 allocations, has MMU enabled and provides values for both:
1001 - ARCH_MMAP_RND_BITS_MIN
1002 - ARCH_MMAP_RND_BITS_MAX
1004 config HAVE_EXIT_THREAD
1007 An architecture implements exit_thread.
1009 config ARCH_MMAP_RND_BITS_MIN
1012 config ARCH_MMAP_RND_BITS_MAX
1015 config ARCH_MMAP_RND_BITS_DEFAULT
1018 config ARCH_MMAP_RND_BITS
1019 int "Number of bits to use for ASLR of mmap base address" if EXPERT
1020 range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX
1021 default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT
1022 default ARCH_MMAP_RND_BITS_MIN
1023 depends on HAVE_ARCH_MMAP_RND_BITS
1025 This value can be used to select the number of bits to use to
1026 determine the random offset to the base address of vma regions
1027 resulting from mmap allocations. This value will be bounded
1028 by the architecture's minimum and maximum supported values.
1030 This value can be changed after boot using the
1031 /proc/sys/vm/mmap_rnd_bits tunable
1033 config HAVE_ARCH_MMAP_RND_COMPAT_BITS
1036 An arch should select this symbol if it supports running applications
1037 in compatibility mode, supports setting a variable number of bits for
1038 use in establishing the base address for mmap allocations, has MMU
1039 enabled and provides values for both:
1040 - ARCH_MMAP_RND_COMPAT_BITS_MIN
1041 - ARCH_MMAP_RND_COMPAT_BITS_MAX
1043 config ARCH_MMAP_RND_COMPAT_BITS_MIN
1046 config ARCH_MMAP_RND_COMPAT_BITS_MAX
1049 config ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1052 config ARCH_MMAP_RND_COMPAT_BITS
1053 int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT
1054 range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX
1055 default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1056 default ARCH_MMAP_RND_COMPAT_BITS_MIN
1057 depends on HAVE_ARCH_MMAP_RND_COMPAT_BITS
1059 This value can be used to select the number of bits to use to
1060 determine the random offset to the base address of vma regions
1061 resulting from mmap allocations for compatible applications This
1062 value will be bounded by the architecture's minimum and maximum
1065 This value can be changed after boot using the
1066 /proc/sys/vm/mmap_rnd_compat_bits tunable
1068 config HAVE_ARCH_COMPAT_MMAP_BASES
1071 This allows 64bit applications to invoke 32-bit mmap() syscall
1072 and vice-versa 32-bit applications to call 64-bit mmap().
1073 Required for applications doing different bitness syscalls.
1075 config PAGE_SIZE_LESS_THAN_64KB
1077 depends on !ARM64_64K_PAGES
1078 depends on !PAGE_SIZE_64KB
1079 depends on !PARISC_PAGE_SIZE_64KB
1080 depends on PAGE_SIZE_LESS_THAN_256KB
1082 config PAGE_SIZE_LESS_THAN_256KB
1084 depends on !PAGE_SIZE_256KB
1086 # This allows to use a set of generic functions to determine mmap base
1087 # address by giving priority to top-down scheme only if the process
1088 # is not in legacy mode (compat task, unlimited stack size or
1089 # sysctl_legacy_va_layout).
1090 # Architecture that selects this option can provide its own version of:
1092 config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT
1095 select ARCH_HAS_ELF_RANDOMIZE
1100 config HAVE_JUMP_LABEL_HACK
1103 config HAVE_NOINSTR_HACK
1106 config HAVE_NOINSTR_VALIDATION
1109 config HAVE_UACCESS_VALIDATION
1113 config HAVE_STACK_VALIDATION
1116 Architecture supports objtool compile-time frame pointer rule
1119 config HAVE_RELIABLE_STACKTRACE
1122 Architecture has either save_stack_trace_tsk_reliable() or
1123 arch_stack_walk_reliable() function which only returns a stack trace
1124 if it can guarantee the trace is reliable.
1126 config HAVE_ARCH_HASH
1130 If this is set, the architecture provides an <asm/hash.h>
1131 file which provides platform-specific implementations of some
1132 functions in <linux/hash.h> or fs/namei.c.
1134 config HAVE_ARCH_NVRAM_OPS
1143 config CLONE_BACKWARDS
1146 Architecture has tls passed as the 4th argument of clone(2),
1149 config CLONE_BACKWARDS2
1152 Architecture has the first two arguments of clone(2) swapped.
1154 config CLONE_BACKWARDS3
1157 Architecture has tls passed as the 3rd argument of clone(2),
1160 config ODD_RT_SIGACTION
1163 Architecture has unusual rt_sigaction(2) arguments
1165 config OLD_SIGSUSPEND
1168 Architecture has old sigsuspend(2) syscall, of one-argument variety
1170 config OLD_SIGSUSPEND3
1173 Even weirder antique ABI - three-argument sigsuspend(2)
1175 config OLD_SIGACTION
1178 Architecture has old sigaction(2) syscall. Nope, not the same
1179 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
1180 but fairly different variant of sigaction(2), thanks to OSF/1
1183 config COMPAT_OLD_SIGACTION
1186 config COMPAT_32BIT_TIME
1187 bool "Provide system calls for 32-bit time_t"
1188 default !64BIT || COMPAT
1190 This enables 32 bit time_t support in addition to 64 bit time_t support.
1191 This is relevant on all 32-bit architectures, and 64-bit architectures
1192 as part of compat syscall handling.
1194 config ARCH_NO_PREEMPT
1197 config ARCH_SUPPORTS_RT
1200 config CPU_NO_EFFICIENT_FFS
1203 config HAVE_ARCH_VMAP_STACK
1206 An arch should select this symbol if it can support kernel stacks
1207 in vmalloc space. This means:
1209 - vmalloc space must be large enough to hold many kernel stacks.
1210 This may rule out many 32-bit architectures.
1212 - Stacks in vmalloc space need to work reliably. For example, if
1213 vmap page tables are created on demand, either this mechanism
1214 needs to work while the stack points to a virtual address with
1215 unpopulated page tables or arch code (switch_to() and switch_mm(),
1216 most likely) needs to ensure that the stack's page table entries
1217 are populated before running on a possibly unpopulated stack.
1219 - If the stack overflows into a guard page, something reasonable
1220 should happen. The definition of "reasonable" is flexible, but
1221 instantly rebooting without logging anything would be unfriendly.
1225 bool "Use a virtually-mapped stack"
1226 depends on HAVE_ARCH_VMAP_STACK
1227 depends on !KASAN || KASAN_HW_TAGS || KASAN_VMALLOC
1229 Enable this if you want the use virtually-mapped kernel stacks
1230 with guard pages. This causes kernel stack overflows to be
1231 caught immediately rather than causing difficult-to-diagnose
1234 To use this with software KASAN modes, the architecture must support
1235 backing virtual mappings with real shadow memory, and KASAN_VMALLOC
1238 config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1241 An arch should select this symbol if it can support kernel stack
1242 offset randomization with calls to add_random_kstack_offset()
1243 during syscall entry and choose_random_kstack_offset() during
1244 syscall exit. Careful removal of -fstack-protector-strong and
1245 -fstack-protector should also be applied to the entry code and
1246 closely examined, as the artificial stack bump looks like an array
1247 to the compiler, so it will attempt to add canary checks regardless
1248 of the static branch state.
1250 config RANDOMIZE_KSTACK_OFFSET
1251 bool "Support for randomizing kernel stack offset on syscall entry" if EXPERT
1253 depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1254 depends on INIT_STACK_NONE || !CC_IS_CLANG || CLANG_VERSION >= 140000
1256 The kernel stack offset can be randomized (after pt_regs) by
1257 roughly 5 bits of entropy, frustrating memory corruption
1258 attacks that depend on stack address determinism or
1259 cross-syscall address exposures.
1261 The feature is controlled via the "randomize_kstack_offset=on/off"
1262 kernel boot param, and if turned off has zero overhead due to its use
1263 of static branches (see JUMP_LABEL).
1267 config RANDOMIZE_KSTACK_OFFSET_DEFAULT
1268 bool "Default state of kernel stack offset randomization"
1269 depends on RANDOMIZE_KSTACK_OFFSET
1271 Kernel stack offset randomization is controlled by kernel boot param
1272 "randomize_kstack_offset=on/off", and this config chooses the default
1275 config ARCH_OPTIONAL_KERNEL_RWX
1278 config ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1281 config ARCH_HAS_STRICT_KERNEL_RWX
1284 config STRICT_KERNEL_RWX
1285 bool "Make kernel text and rodata read-only" if ARCH_OPTIONAL_KERNEL_RWX
1286 depends on ARCH_HAS_STRICT_KERNEL_RWX
1287 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1289 If this is set, kernel text and rodata memory will be made read-only,
1290 and non-text memory will be made non-executable. This provides
1291 protection against certain security exploits (e.g. executing the heap
1294 These features are considered standard security practice these days.
1295 You should say Y here in almost all cases.
1297 config ARCH_HAS_STRICT_MODULE_RWX
1300 config STRICT_MODULE_RWX
1301 bool "Set loadable kernel module data as NX and text as RO" if ARCH_OPTIONAL_KERNEL_RWX
1302 depends on ARCH_HAS_STRICT_MODULE_RWX && MODULES
1303 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1305 If this is set, module text and rodata memory will be made read-only,
1306 and non-text memory will be made non-executable. This provides
1307 protection against certain security exploits (e.g. writing to text)
1309 # select if the architecture provides an asm/dma-direct.h header
1310 config ARCH_HAS_PHYS_TO_DMA
1313 config HAVE_ARCH_COMPILER_H
1316 An architecture can select this if it provides an
1317 asm/compiler.h header that should be included after
1318 linux/compiler-*.h in order to override macro definitions that those
1319 headers generally provide.
1321 config HAVE_ARCH_PREL32_RELOCATIONS
1324 May be selected by an architecture if it supports place-relative
1325 32-bit relocations, both in the toolchain and in the module loader,
1326 in which case relative references can be used in special sections
1327 for PCI fixup, initcalls etc which are only half the size on 64 bit
1328 architectures, and don't require runtime relocation on relocatable
1331 config ARCH_USE_MEMREMAP_PROT
1334 config LOCK_EVENT_COUNTS
1335 bool "Locking event counts collection"
1338 Enable light-weight counting of various locking related events
1339 in the system with minimal performance impact. This reduces
1340 the chance of application behavior change because of timing
1341 differences. The counts are reported via debugfs.
1343 # Select if the architecture has support for applying RELR relocations.
1344 config ARCH_HAS_RELR
1348 bool "Use RELR relocation packing"
1349 depends on ARCH_HAS_RELR && TOOLS_SUPPORT_RELR
1352 Store the kernel's dynamic relocations in the RELR relocation packing
1353 format. Requires a compatible linker (LLD supports this feature), as
1354 well as compatible NM and OBJCOPY utilities (llvm-nm and llvm-objcopy
1357 config ARCH_HAS_MEM_ENCRYPT
1360 config ARCH_HAS_CC_PLATFORM
1363 config HAVE_SPARSE_SYSCALL_NR
1366 An architecture should select this if its syscall numbering is sparse
1367 to save space. For example, MIPS architecture has a syscall array with
1368 entries at 4000, 5000 and 6000 locations. This option turns on syscall
1369 related optimizations for a given architecture.
1371 config ARCH_HAS_VDSO_DATA
1374 config HAVE_STATIC_CALL
1377 config HAVE_STATIC_CALL_INLINE
1379 depends on HAVE_STATIC_CALL
1382 config HAVE_PREEMPT_DYNAMIC
1385 config HAVE_PREEMPT_DYNAMIC_CALL
1387 depends on HAVE_STATIC_CALL
1388 select HAVE_PREEMPT_DYNAMIC
1390 An architecture should select this if it can handle the preemption
1391 model being selected at boot time using static calls.
1393 Where an architecture selects HAVE_STATIC_CALL_INLINE, any call to a
1394 preemption function will be patched directly.
1396 Where an architecture does not select HAVE_STATIC_CALL_INLINE, any
1397 call to a preemption function will go through a trampoline, and the
1398 trampoline will be patched.
1400 It is strongly advised to support inline static call to avoid any
1403 config HAVE_PREEMPT_DYNAMIC_KEY
1405 depends on HAVE_ARCH_JUMP_LABEL
1406 select HAVE_PREEMPT_DYNAMIC
1408 An architecture should select this if it can handle the preemption
1409 model being selected at boot time using static keys.
1411 Each preemption function will be given an early return based on a
1412 static key. This should have slightly lower overhead than non-inline
1413 static calls, as this effectively inlines each trampoline into the
1414 start of its callee. This may avoid redundant work, and may
1415 integrate better with CFI schemes.
1417 This will have greater overhead than using inline static calls as
1418 the call to the preemption function cannot be entirely elided.
1420 config ARCH_WANT_LD_ORPHAN_WARN
1423 An arch should select this symbol once all linker sections are explicitly
1424 included, size-asserted, or discarded in the linker scripts. This is
1425 important because we never want expected sections to be placed heuristically
1426 by the linker, since the locations of such sections can change between linker
1429 config HAVE_ARCH_PFN_VALID
1432 config ARCH_SUPPORTS_DEBUG_PAGEALLOC
1435 config ARCH_SUPPORTS_PAGE_TABLE_CHECK
1438 config ARCH_SPLIT_ARG64
1441 If a 32-bit architecture requires 64-bit arguments to be split into
1442 pairs of 32-bit arguments, select this option.
1444 config ARCH_HAS_ELFCORE_COMPAT
1447 config ARCH_HAS_PARANOID_L1D_FLUSH
1450 config ARCH_HAVE_TRACE_MMIO_ACCESS
1453 config DYNAMIC_SIGFRAME
1456 # Select, if arch has a named attribute group bound to NUMA device nodes.
1457 config HAVE_ARCH_NODE_DEV_GROUP
1460 config ARCH_HAS_HW_PTE_YOUNG
1463 Architectures that select this option are capable of setting the
1464 accessed bit in PTE entries when using them as part of linear address
1465 translations. Architectures that require runtime check should select
1466 this option and override arch_has_hw_pte_young().
1468 config ARCH_HAS_NONLEAF_PMD_YOUNG
1471 Architectures that select this option are capable of setting the
1472 accessed bit in non-leaf PMD entries when using them as part of linear
1473 address translations. Page table walkers that clear the accessed bit
1474 may use this capability to reduce their search space.
1476 source "kernel/gcov/Kconfig"
1478 source "scripts/gcc-plugins/Kconfig"
1480 config FUNCTION_ALIGNMENT_4B
1483 config FUNCTION_ALIGNMENT_8B
1486 config FUNCTION_ALIGNMENT_16B
1489 config FUNCTION_ALIGNMENT_32B
1492 config FUNCTION_ALIGNMENT_64B
1495 config FUNCTION_ALIGNMENT
1497 default 64 if FUNCTION_ALIGNMENT_64B
1498 default 32 if FUNCTION_ALIGNMENT_32B
1499 default 16 if FUNCTION_ALIGNMENT_16B
1500 default 8 if FUNCTION_ALIGNMENT_8B
1501 default 4 if FUNCTION_ALIGNMENT_4B