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 config SMT_NUM_THREADS_DYNAMIC
40 # Selected by HOTPLUG_CORE_SYNC_DEAD or HOTPLUG_CORE_SYNC_FULL
41 config HOTPLUG_CORE_SYNC
44 # Basic CPU dead synchronization selected by architecture
45 config HOTPLUG_CORE_SYNC_DEAD
47 select HOTPLUG_CORE_SYNC
49 # Full CPU synchronization with alive state selected by architecture
50 config HOTPLUG_CORE_SYNC_FULL
52 select HOTPLUG_CORE_SYNC_DEAD if HOTPLUG_CPU
53 select HOTPLUG_CORE_SYNC
55 config HOTPLUG_SPLIT_STARTUP
57 select HOTPLUG_CORE_SYNC_FULL
59 config HOTPLUG_PARALLEL
61 select HOTPLUG_SPLIT_STARTUP
69 depends on HAVE_KPROBES
71 select TASKS_RCU if PREEMPTION
73 Kprobes allows you to trap at almost any kernel address and
74 execute a callback function. register_kprobe() establishes
75 a probepoint and specifies the callback. Kprobes is useful
76 for kernel debugging, non-intrusive instrumentation and testing.
80 bool "Optimize very unlikely/likely branches"
81 depends on HAVE_ARCH_JUMP_LABEL
82 select OBJTOOL if HAVE_JUMP_LABEL_HACK
84 This option enables a transparent branch optimization that
85 makes certain almost-always-true or almost-always-false branch
86 conditions even cheaper to execute within the kernel.
88 Certain performance-sensitive kernel code, such as trace points,
89 scheduler functionality, networking code and KVM have such
90 branches and include support for this optimization technique.
92 If it is detected that the compiler has support for "asm goto",
93 the kernel will compile such branches with just a nop
94 instruction. When the condition flag is toggled to true, the
95 nop will be converted to a jump instruction to execute the
96 conditional block of instructions.
98 This technique lowers overhead and stress on the branch prediction
99 of the processor and generally makes the kernel faster. The update
100 of the condition is slower, but those are always very rare.
102 ( On 32-bit x86, the necessary options added to the compiler
103 flags may increase the size of the kernel slightly. )
105 config STATIC_KEYS_SELFTEST
106 bool "Static key selftest"
107 depends on JUMP_LABEL
109 Boot time self-test of the branch patching code.
111 config STATIC_CALL_SELFTEST
112 bool "Static call selftest"
113 depends on HAVE_STATIC_CALL
115 Boot time self-test of the call patching code.
119 depends on KPROBES && HAVE_OPTPROBES
120 select TASKS_RCU if PREEMPTION
122 config KPROBES_ON_FTRACE
124 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
125 depends on DYNAMIC_FTRACE_WITH_REGS
127 If function tracer is enabled and the arch supports full
128 passing of pt_regs to function tracing, then kprobes can
129 optimize on top of function tracing.
133 depends on ARCH_SUPPORTS_UPROBES
135 Uprobes is the user-space counterpart to kprobes: they
136 enable instrumentation applications (such as 'perf probe')
137 to establish unintrusive probes in user-space binaries and
138 libraries, by executing handler functions when the probes
139 are hit by user-space applications.
141 ( These probes come in the form of single-byte breakpoints,
142 managed by the kernel and kept transparent to the probed
145 config HAVE_64BIT_ALIGNED_ACCESS
146 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
148 Some architectures require 64 bit accesses to be 64 bit
149 aligned, which also requires structs containing 64 bit values
150 to be 64 bit aligned too. This includes some 32 bit
151 architectures which can do 64 bit accesses, as well as 64 bit
152 architectures without unaligned access.
154 This symbol should be selected by an architecture if 64 bit
155 accesses are required to be 64 bit aligned in this way even
156 though it is not a 64 bit architecture.
158 See Documentation/core-api/unaligned-memory-access.rst for
159 more information on the topic of unaligned memory accesses.
161 config HAVE_EFFICIENT_UNALIGNED_ACCESS
164 Some architectures are unable to perform unaligned accesses
165 without the use of get_unaligned/put_unaligned. Others are
166 unable to perform such accesses efficiently (e.g. trap on
167 unaligned access and require fixing it up in the exception
170 This symbol should be selected by an architecture if it can
171 perform unaligned accesses efficiently to allow different
172 code paths to be selected for these cases. Some network
173 drivers, for example, could opt to not fix up alignment
174 problems with received packets if doing so would not help
177 See Documentation/core-api/unaligned-memory-access.rst for more
178 information on the topic of unaligned memory accesses.
180 config ARCH_USE_BUILTIN_BSWAP
183 Modern versions of GCC (since 4.4) have builtin functions
184 for handling byte-swapping. Using these, instead of the old
185 inline assembler that the architecture code provides in the
186 __arch_bswapXX() macros, allows the compiler to see what's
187 happening and offers more opportunity for optimisation. In
188 particular, the compiler will be able to combine the byteswap
189 with a nearby load or store and use load-and-swap or
190 store-and-swap instructions if the architecture has them. It
191 should almost *never* result in code which is worse than the
192 hand-coded assembler in <asm/swab.h>. But just in case it
193 does, the use of the builtins is optional.
195 Any architecture with load-and-swap or store-and-swap
196 instructions should set this. And it shouldn't hurt to set it
197 on architectures that don't have such instructions.
201 depends on KPROBES && (HAVE_KRETPROBES || HAVE_RETHOOK)
203 config KRETPROBE_ON_RETHOOK
205 depends on HAVE_RETHOOK
206 depends on KRETPROBES
209 config USER_RETURN_NOTIFIER
211 depends on HAVE_USER_RETURN_NOTIFIER
213 Provide a kernel-internal notification when a cpu is about to
216 config HAVE_IOREMAP_PROT
222 config HAVE_KRETPROBES
225 config HAVE_OPTPROBES
228 config HAVE_KPROBES_ON_FTRACE
231 config ARCH_CORRECT_STACKTRACE_ON_KRETPROBE
234 Since kretprobes modifies return address on the stack, the
235 stacktrace may see the kretprobe trampoline address instead
236 of correct one. If the architecture stacktrace code and
237 unwinder can adjust such entries, select this configuration.
239 config HAVE_FUNCTION_ERROR_INJECTION
245 config HAVE_FUNCTION_DESCRIPTORS
248 config TRACE_IRQFLAGS_SUPPORT
251 config TRACE_IRQFLAGS_NMI_SUPPORT
255 # An arch should select this if it provides all these things:
257 # task_pt_regs() in asm/processor.h or asm/ptrace.h
258 # arch_has_single_step() if there is hardware single-step support
259 # arch_has_block_step() if there is hardware block-step support
260 # asm/syscall.h supplying asm-generic/syscall.h interface
261 # linux/regset.h user_regset interfaces
262 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
263 # TIF_SYSCALL_TRACE calls ptrace_report_syscall_{entry,exit}
264 # TIF_NOTIFY_RESUME calls resume_user_mode_work()
266 config HAVE_ARCH_TRACEHOOK
269 config HAVE_DMA_CONTIGUOUS
272 config GENERIC_SMP_IDLE_THREAD
275 config GENERIC_IDLE_POLL_SETUP
278 config ARCH_HAS_FORTIFY_SOURCE
281 An architecture should select this when it can successfully
282 build and run with CONFIG_FORTIFY_SOURCE.
285 # Select if the arch provides a historic keepinit alias for the retain_initrd
286 # command line option
288 config ARCH_HAS_KEEPINITRD
291 # Select if arch has all set_memory_ro/rw/x/nx() functions in asm/cacheflush.h
292 config ARCH_HAS_SET_MEMORY
295 # Select if arch has all set_direct_map_invalid/default() functions
296 config ARCH_HAS_SET_DIRECT_MAP
300 # Select if the architecture provides the arch_dma_set_uncached symbol to
301 # either provide an uncached segment alias for a DMA allocation, or
302 # to remap the page tables in place.
304 config ARCH_HAS_DMA_SET_UNCACHED
308 # Select if the architectures provides the arch_dma_clear_uncached symbol
309 # to undo an in-place page table remap for uncached access.
311 config ARCH_HAS_DMA_CLEAR_UNCACHED
314 config ARCH_HAS_CPU_FINALIZE_INIT
317 # Select if arch init_task must go in the __init_task_data section
318 config ARCH_TASK_STRUCT_ON_STACK
321 # Select if arch has its private alloc_task_struct() function
322 config ARCH_TASK_STRUCT_ALLOCATOR
325 config HAVE_ARCH_THREAD_STRUCT_WHITELIST
327 depends on !ARCH_TASK_STRUCT_ALLOCATOR
329 An architecture should select this to provide hardened usercopy
330 knowledge about what region of the thread_struct should be
331 whitelisted for copying to userspace. Normally this is only the
332 FPU registers. Specifically, arch_thread_struct_whitelist()
333 should be implemented. Without this, the entire thread_struct
334 field in task_struct will be left whitelisted.
336 # Select if arch has its private alloc_thread_stack() function
337 config ARCH_THREAD_STACK_ALLOCATOR
340 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
341 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
344 config ARCH_WANTS_NO_INSTR
347 An architecture should select this if the noinstr macro is being used on
348 functions to denote that the toolchain should avoid instrumenting such
349 functions and is required for correctness.
351 config ARCH_32BIT_OFF_T
355 All new 32-bit architectures should have 64-bit off_t type on
356 userspace side which corresponds to the loff_t kernel type. This
357 is the requirement for modern ABIs. Some existing architectures
358 still support 32-bit off_t. This option is enabled for all such
359 architectures explicitly.
361 # Selected by 64 bit architectures which have a 32 bit f_tinode in struct ustat
362 config ARCH_32BIT_USTAT_F_TINODE
365 config HAVE_ASM_MODVERSIONS
368 This symbol should be selected by an architecture if it provides
369 <asm/asm-prototypes.h> to support the module versioning for symbols
370 exported from assembly code.
372 config HAVE_REGS_AND_STACK_ACCESS_API
375 This symbol should be selected by an architecture if it supports
376 the API needed to access registers and stack entries from pt_regs,
377 declared in asm/ptrace.h
378 For example the kprobes-based event tracer needs this API.
382 depends on HAVE_REGS_AND_STACK_ACCESS_API
384 This symbol should be selected by an architecture if it
385 supports an implementation of restartable sequences.
390 This symbol should be selected by an architecture if it
393 config HAVE_FUNCTION_ARG_ACCESS_API
396 This symbol should be selected by an architecture if it supports
397 the API needed to access function arguments from pt_regs,
398 declared in asm/ptrace.h
400 config HAVE_HW_BREAKPOINT
402 depends on PERF_EVENTS
404 config HAVE_MIXED_BREAKPOINTS_REGS
406 depends on HAVE_HW_BREAKPOINT
408 Depending on the arch implementation of hardware breakpoints,
409 some of them have separate registers for data and instruction
410 breakpoints addresses, others have mixed registers to store
411 them but define the access type in a control register.
412 Select this option if your arch implements breakpoints under the
415 config HAVE_USER_RETURN_NOTIFIER
418 config HAVE_PERF_EVENTS_NMI
421 System hardware can generate an NMI using the perf event
422 subsystem. Also has support for calculating CPU cycle events
423 to determine how many clock cycles in a given period.
425 config HAVE_HARDLOCKUP_DETECTOR_PERF
427 depends on HAVE_PERF_EVENTS_NMI
429 The arch chooses to use the generic perf-NMI-based hardlockup
430 detector. Must define HAVE_PERF_EVENTS_NMI.
432 config HAVE_HARDLOCKUP_DETECTOR_ARCH
435 The arch provides its own hardlockup detector implementation instead
438 It uses the same command line parameters, and sysctl interface,
439 as the generic hardlockup detectors.
441 config HAVE_PERF_REGS
444 Support selective register dumps for perf events. This includes
445 bit-mapping of each registers and a unique architecture id.
447 config HAVE_PERF_USER_STACK_DUMP
450 Support user stack dumps for perf event samples. This needs
451 access to the user stack pointer which is not unified across
454 config HAVE_ARCH_JUMP_LABEL
457 config HAVE_ARCH_JUMP_LABEL_RELATIVE
460 config MMU_GATHER_TABLE_FREE
463 config MMU_GATHER_RCU_TABLE_FREE
465 select MMU_GATHER_TABLE_FREE
467 config MMU_GATHER_PAGE_SIZE
470 config MMU_GATHER_NO_RANGE
472 select MMU_GATHER_MERGE_VMAS
474 config MMU_GATHER_NO_FLUSH_CACHE
477 config MMU_GATHER_MERGE_VMAS
480 config MMU_GATHER_NO_GATHER
482 depends on MMU_GATHER_TABLE_FREE
484 config ARCH_WANT_IRQS_OFF_ACTIVATE_MM
487 Temporary select until all architectures can be converted to have
488 irqs disabled over activate_mm. Architectures that do IPI based TLB
489 shootdowns should enable this.
491 # Use normal mm refcounting for MMU_LAZY_TLB kernel thread references.
492 # MMU_LAZY_TLB_REFCOUNT=n can improve the scalability of context switching
493 # to/from kernel threads when the same mm is running on a lot of CPUs (a large
494 # multi-threaded application), by reducing contention on the mm refcount.
496 # This can be disabled if the architecture ensures no CPUs are using an mm as a
497 # "lazy tlb" beyond its final refcount (i.e., by the time __mmdrop frees the mm
498 # or its kernel page tables). This could be arranged by arch_exit_mmap(), or
499 # final exit(2) TLB flush, for example.
501 # To implement this, an arch *must*:
502 # Ensure the _lazy_tlb variants of mmgrab/mmdrop are used when manipulating
503 # the lazy tlb reference of a kthread's ->active_mm (non-arch code has been
504 # converted already).
505 config MMU_LAZY_TLB_REFCOUNT
507 depends on !MMU_LAZY_TLB_SHOOTDOWN
509 # This option allows MMU_LAZY_TLB_REFCOUNT=n. It ensures no CPUs are using an
510 # mm as a lazy tlb beyond its last reference count, by shooting down these
511 # users before the mm is deallocated. __mmdrop() first IPIs all CPUs that may
512 # be using the mm as a lazy tlb, so that they may switch themselves to using
513 # init_mm for their active mm. mm_cpumask(mm) is used to determine which CPUs
514 # may be using mm as a lazy tlb mm.
516 # To implement this, an arch *must*:
517 # - At the time of the final mmdrop of the mm, ensure mm_cpumask(mm) contains
518 # at least all possible CPUs in which the mm is lazy.
519 # - It must meet the requirements for MMU_LAZY_TLB_REFCOUNT=n (see above).
520 config MMU_LAZY_TLB_SHOOTDOWN
523 config ARCH_HAVE_NMI_SAFE_CMPXCHG
526 config ARCH_HAS_NMI_SAFE_THIS_CPU_OPS
529 config HAVE_ALIGNED_STRUCT_PAGE
532 This makes sure that struct pages are double word aligned and that
533 e.g. the SLUB allocator can perform double word atomic operations
534 on a struct page for better performance. However selecting this
535 might increase the size of a struct page by a word.
537 config HAVE_CMPXCHG_LOCAL
540 config HAVE_CMPXCHG_DOUBLE
543 config ARCH_WEAK_RELEASE_ACQUIRE
546 config ARCH_WANT_IPC_PARSE_VERSION
549 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
552 config ARCH_WANT_OLD_COMPAT_IPC
553 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
556 config HAVE_ARCH_SECCOMP
559 An arch should select this symbol to support seccomp mode 1 (the fixed
560 syscall policy), and must provide an overrides for __NR_seccomp_sigreturn,
561 and compat syscalls if the asm-generic/seccomp.h defaults need adjustment:
562 - __NR_seccomp_read_32
563 - __NR_seccomp_write_32
564 - __NR_seccomp_exit_32
565 - __NR_seccomp_sigreturn_32
567 config HAVE_ARCH_SECCOMP_FILTER
569 select HAVE_ARCH_SECCOMP
571 An arch should select this symbol if it provides all of these things:
572 - all the requirements for HAVE_ARCH_SECCOMP
574 - syscall_get_arguments()
576 - syscall_set_return_value()
577 - SIGSYS siginfo_t support
578 - secure_computing is called from a ptrace_event()-safe context
579 - secure_computing return value is checked and a return value of -1
580 results in the system call being skipped immediately.
581 - seccomp syscall wired up
582 - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE,
583 SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If
584 COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too.
587 prompt "Enable seccomp to safely execute untrusted bytecode"
589 depends on HAVE_ARCH_SECCOMP
591 This kernel feature is useful for number crunching applications
592 that may need to handle untrusted bytecode during their
593 execution. By using pipes or other transports made available
594 to the process as file descriptors supporting the read/write
595 syscalls, it's possible to isolate those applications in their
596 own address space using seccomp. Once seccomp is enabled via
597 prctl(PR_SET_SECCOMP) or the seccomp() syscall, it cannot be
598 disabled and the task is only allowed to execute a few safe
599 syscalls defined by each seccomp mode.
603 config SECCOMP_FILTER
605 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
607 Enable tasks to build secure computing environments defined
608 in terms of Berkeley Packet Filter programs which implement
609 task-defined system call filtering polices.
611 See Documentation/userspace-api/seccomp_filter.rst for details.
613 config SECCOMP_CACHE_DEBUG
614 bool "Show seccomp filter cache status in /proc/pid/seccomp_cache"
615 depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR
618 This enables the /proc/pid/seccomp_cache interface to monitor
619 seccomp cache data. The file format is subject to change. Reading
620 the file requires CAP_SYS_ADMIN.
622 This option is for debugging only. Enabling presents the risk that
623 an adversary may be able to infer the seccomp filter logic.
627 config HAVE_ARCH_STACKLEAK
630 An architecture should select this if it has the code which
631 fills the used part of the kernel stack with the STACKLEAK_POISON
632 value before returning from system calls.
634 config HAVE_STACKPROTECTOR
637 An arch should select this symbol if:
638 - it has implemented a stack canary (e.g. __stack_chk_guard)
640 config STACKPROTECTOR
641 bool "Stack Protector buffer overflow detection"
642 depends on HAVE_STACKPROTECTOR
643 depends on $(cc-option,-fstack-protector)
646 This option turns on the "stack-protector" GCC feature. This
647 feature puts, at the beginning of functions, a canary value on
648 the stack just before the return address, and validates
649 the value just before actually returning. Stack based buffer
650 overflows (that need to overwrite this return address) now also
651 overwrite the canary, which gets detected and the attack is then
652 neutralized via a kernel panic.
654 Functions will have the stack-protector canary logic added if they
655 have an 8-byte or larger character array on the stack.
657 This feature requires gcc version 4.2 or above, or a distribution
658 gcc with the feature backported ("-fstack-protector").
660 On an x86 "defconfig" build, this feature adds canary checks to
661 about 3% of all kernel functions, which increases kernel code size
664 config STACKPROTECTOR_STRONG
665 bool "Strong Stack Protector"
666 depends on STACKPROTECTOR
667 depends on $(cc-option,-fstack-protector-strong)
670 Functions will have the stack-protector canary logic added in any
671 of the following conditions:
673 - local variable's address used as part of the right hand side of an
674 assignment or function argument
675 - local variable is an array (or union containing an array),
676 regardless of array type or length
677 - uses register local variables
679 This feature requires gcc version 4.9 or above, or a distribution
680 gcc with the feature backported ("-fstack-protector-strong").
682 On an x86 "defconfig" build, this feature adds canary checks to
683 about 20% of all kernel functions, which increases the kernel code
686 config ARCH_SUPPORTS_SHADOW_CALL_STACK
689 An architecture should select this if it supports the compiler's
690 Shadow Call Stack and implements runtime support for shadow stack
693 config SHADOW_CALL_STACK
694 bool "Shadow Call Stack"
695 depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
696 depends on DYNAMIC_FTRACE_WITH_ARGS || DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
698 This option enables the compiler's Shadow Call Stack, which
699 uses a shadow stack to protect function return addresses from
700 being overwritten by an attacker. More information can be found
701 in the compiler's documentation:
703 - Clang: https://clang.llvm.org/docs/ShadowCallStack.html
704 - GCC: https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html#Instrumentation-Options
706 Note that security guarantees in the kernel differ from the
707 ones documented for user space. The kernel must store addresses
708 of shadow stacks in memory, which means an attacker capable of
709 reading and writing arbitrary memory may be able to locate them
710 and hijack control flow by modifying the stacks.
715 Set by the arch code if it relies on code patching to insert the
716 shadow call stack push and pop instructions rather than on the
722 Selected if the kernel will be built using the compiler's LTO feature.
728 Selected if the kernel will be built using Clang's LTO feature.
730 config ARCH_SUPPORTS_LTO_CLANG
733 An architecture should select this option if it supports:
734 - compiling with Clang,
735 - compiling inline assembly with Clang's integrated assembler,
736 - and linking with LLD.
738 config ARCH_SUPPORTS_LTO_CLANG_THIN
741 An architecture should select this option if it can support Clang's
746 depends on CC_IS_CLANG && LD_IS_LLD && AS_IS_LLVM
747 depends on $(success,$(NM) --help | head -n 1 | grep -qi llvm)
748 depends on $(success,$(AR) --help | head -n 1 | grep -qi llvm)
749 depends on ARCH_SUPPORTS_LTO_CLANG
750 depends on !FTRACE_MCOUNT_USE_RECORDMCOUNT
751 depends on !KASAN || KASAN_HW_TAGS
752 depends on !GCOV_KERNEL
754 The compiler and Kconfig options support building with Clang's
758 prompt "Link Time Optimization (LTO)"
761 This option enables Link Time Optimization (LTO), which allows the
762 compiler to optimize binaries globally.
764 If unsure, select LTO_NONE. Note that LTO is very resource-intensive
765 so it's disabled by default.
770 Build the kernel normally, without Link Time Optimization (LTO).
772 config LTO_CLANG_FULL
773 bool "Clang Full LTO (EXPERIMENTAL)"
774 depends on HAS_LTO_CLANG
775 depends on !COMPILE_TEST
778 This option enables Clang's full Link Time Optimization (LTO), which
779 allows the compiler to optimize the kernel globally. If you enable
780 this option, the compiler generates LLVM bitcode instead of ELF
781 object files, and the actual compilation from bitcode happens at
782 the LTO link step, which may take several minutes depending on the
783 kernel configuration. More information can be found from LLVM's
786 https://llvm.org/docs/LinkTimeOptimization.html
788 During link time, this option can use a large amount of RAM, and
789 may take much longer than the ThinLTO option.
791 config LTO_CLANG_THIN
792 bool "Clang ThinLTO (EXPERIMENTAL)"
793 depends on HAS_LTO_CLANG && ARCH_SUPPORTS_LTO_CLANG_THIN
796 This option enables Clang's ThinLTO, which allows for parallel
797 optimization and faster incremental compiles compared to the
798 CONFIG_LTO_CLANG_FULL option. More information can be found
799 from Clang's documentation:
801 https://clang.llvm.org/docs/ThinLTO.html
806 config ARCH_SUPPORTS_CFI_CLANG
809 An architecture should select this option if it can support Clang's
810 Control-Flow Integrity (CFI) checking.
812 config ARCH_USES_CFI_TRAPS
816 bool "Use Clang's Control Flow Integrity (CFI)"
817 depends on ARCH_SUPPORTS_CFI_CLANG
818 depends on $(cc-option,-fsanitize=kcfi)
820 This option enables Clang’s forward-edge Control Flow Integrity
821 (CFI) checking, where the compiler injects a runtime check to each
822 indirect function call to ensure the target is a valid function with
823 the correct static type. This restricts possible call targets and
824 makes it more difficult for an attacker to exploit bugs that allow
825 the modification of stored function pointers. More information can be
826 found from Clang's documentation:
828 https://clang.llvm.org/docs/ControlFlowIntegrity.html
830 config CFI_PERMISSIVE
831 bool "Use CFI in permissive mode"
834 When selected, Control Flow Integrity (CFI) violations result in a
835 warning instead of a kernel panic. This option should only be used
836 for finding indirect call type mismatches during development.
840 config HAVE_ARCH_WITHIN_STACK_FRAMES
843 An architecture should select this if it can walk the kernel stack
844 frames to determine if an object is part of either the arguments
845 or local variables (i.e. that it excludes saved return addresses,
846 and similar) by implementing an inline arch_within_stack_frames(),
847 which is used by CONFIG_HARDENED_USERCOPY.
849 config HAVE_CONTEXT_TRACKING_USER
852 Provide kernel/user boundaries probes necessary for subsystems
853 that need it, such as userspace RCU extended quiescent state.
854 Syscalls need to be wrapped inside user_exit()-user_enter(), either
855 optimized behind static key or through the slow path using TIF_NOHZ
856 flag. Exceptions handlers must be wrapped as well. Irqs are already
857 protected inside ct_irq_enter/ct_irq_exit() but preemption or signal
858 handling on irq exit still need to be protected.
860 config HAVE_CONTEXT_TRACKING_USER_OFFSTACK
863 Architecture neither relies on exception_enter()/exception_exit()
864 nor on schedule_user(). Also preempt_schedule_notrace() and
865 preempt_schedule_irq() can't be called in a preemptible section
866 while context tracking is CONTEXT_USER. This feature reflects a sane
867 entry implementation where the following requirements are met on
868 critical entry code, ie: before user_exit() or after user_enter():
870 - Critical entry code isn't preemptible (or better yet:
872 - No use of RCU read side critical sections, unless ct_nmi_enter()
874 - No use of instrumentation, unless instrumentation_begin() got
880 Arch relies on TIF_NOHZ and syscall slow path to implement context
881 tracking calls to user_enter()/user_exit().
883 config HAVE_VIRT_CPU_ACCOUNTING
886 config HAVE_VIRT_CPU_ACCOUNTING_IDLE
889 Architecture has its own way to account idle CPU time and therefore
890 doesn't implement vtime_account_idle().
892 config ARCH_HAS_SCALED_CPUTIME
895 config HAVE_VIRT_CPU_ACCOUNTING_GEN
899 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
900 Before enabling this option, arch code must be audited
901 to ensure there are no races in concurrent read/write of
902 cputime_t. For example, reading/writing 64-bit cputime_t on
903 some 32-bit arches may require multiple accesses, so proper
904 locking is needed to protect against concurrent accesses.
906 config HAVE_IRQ_TIME_ACCOUNTING
909 Archs need to ensure they use a high enough resolution clock to
910 support irq time accounting and then call enable_sched_clock_irqtime().
915 Architectures that select this are able to move page tables at the
916 PUD level. If there are only 3 page table levels, the move effectively
917 happens at the PGD level.
922 Archs that select this are able to move page tables at the PMD level.
924 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
927 config HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
930 config HAVE_ARCH_HUGE_VMAP
934 # Archs that select this would be capable of PMD-sized vmaps (i.e.,
935 # arch_vmap_pmd_supported() returns true). The VM_ALLOW_HUGE_VMAP flag
936 # must be used to enable allocations to use hugepages.
938 config HAVE_ARCH_HUGE_VMALLOC
939 depends on HAVE_ARCH_HUGE_VMAP
942 config ARCH_WANT_HUGE_PMD_SHARE
945 config HAVE_ARCH_SOFT_DIRTY
948 config HAVE_MOD_ARCH_SPECIFIC
951 The arch uses struct mod_arch_specific to store data. Many arches
952 just need a simple module loader without arch specific data - those
953 should not enable this.
955 config MODULES_USE_ELF_RELA
958 Modules only use ELF RELA relocations. Modules with ELF REL
959 relocations will give an error.
961 config MODULES_USE_ELF_REL
964 Modules only use ELF REL relocations. Modules with ELF RELA
965 relocations will give an error.
967 config ARCH_WANTS_MODULES_DATA_IN_VMALLOC
970 For architectures like powerpc/32 which have constraints on module
971 allocation and need to allocate module data outside of module area.
973 config HAVE_IRQ_EXIT_ON_IRQ_STACK
976 Architecture doesn't only execute the irq handler on the irq stack
977 but also irq_exit(). This way we can process softirqs on this irq
978 stack instead of switching to a new one when we call __do_softirq()
979 in the end of an hardirq.
980 This spares a stack switch and improves cache usage on softirq
983 config HAVE_SOFTIRQ_ON_OWN_STACK
986 Architecture provides a function to run __do_softirq() on a
989 config SOFTIRQ_ON_OWN_STACK
990 def_bool HAVE_SOFTIRQ_ON_OWN_STACK && !PREEMPT_RT
992 config ALTERNATE_USER_ADDRESS_SPACE
995 Architectures set this when the CPU uses separate address
996 spaces for kernel and user space pointers. In this case, the
997 access_ok() check on a __user pointer is skipped.
999 config PGTABLE_LEVELS
1003 config ARCH_HAS_ELF_RANDOMIZE
1006 An architecture supports choosing randomized locations for
1007 stack, mmap, brk, and ET_DYN. Defined functions:
1009 - arch_randomize_brk()
1011 config HAVE_ARCH_MMAP_RND_BITS
1014 An arch should select this symbol if it supports setting a variable
1015 number of bits for use in establishing the base address for mmap
1016 allocations, has MMU enabled and provides values for both:
1017 - ARCH_MMAP_RND_BITS_MIN
1018 - ARCH_MMAP_RND_BITS_MAX
1020 config HAVE_EXIT_THREAD
1023 An architecture implements exit_thread.
1025 config ARCH_MMAP_RND_BITS_MIN
1028 config ARCH_MMAP_RND_BITS_MAX
1031 config ARCH_MMAP_RND_BITS_DEFAULT
1034 config ARCH_MMAP_RND_BITS
1035 int "Number of bits to use for ASLR of mmap base address" if EXPERT
1036 range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX
1037 default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT
1038 default ARCH_MMAP_RND_BITS_MIN
1039 depends on HAVE_ARCH_MMAP_RND_BITS
1041 This value can be used to select the number of bits to use to
1042 determine the random offset to the base address of vma regions
1043 resulting from mmap allocations. This value will be bounded
1044 by the architecture's minimum and maximum supported values.
1046 This value can be changed after boot using the
1047 /proc/sys/vm/mmap_rnd_bits tunable
1049 config HAVE_ARCH_MMAP_RND_COMPAT_BITS
1052 An arch should select this symbol if it supports running applications
1053 in compatibility mode, supports setting a variable number of bits for
1054 use in establishing the base address for mmap allocations, has MMU
1055 enabled and provides values for both:
1056 - ARCH_MMAP_RND_COMPAT_BITS_MIN
1057 - ARCH_MMAP_RND_COMPAT_BITS_MAX
1059 config ARCH_MMAP_RND_COMPAT_BITS_MIN
1062 config ARCH_MMAP_RND_COMPAT_BITS_MAX
1065 config ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1068 config ARCH_MMAP_RND_COMPAT_BITS
1069 int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT
1070 range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX
1071 default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1072 default ARCH_MMAP_RND_COMPAT_BITS_MIN
1073 depends on HAVE_ARCH_MMAP_RND_COMPAT_BITS
1075 This value can be used to select the number of bits to use to
1076 determine the random offset to the base address of vma regions
1077 resulting from mmap allocations for compatible applications This
1078 value will be bounded by the architecture's minimum and maximum
1081 This value can be changed after boot using the
1082 /proc/sys/vm/mmap_rnd_compat_bits tunable
1084 config HAVE_ARCH_COMPAT_MMAP_BASES
1087 This allows 64bit applications to invoke 32-bit mmap() syscall
1088 and vice-versa 32-bit applications to call 64-bit mmap().
1089 Required for applications doing different bitness syscalls.
1091 config PAGE_SIZE_LESS_THAN_64KB
1093 depends on !ARM64_64K_PAGES
1094 depends on !IA64_PAGE_SIZE_64KB
1095 depends on !PAGE_SIZE_64KB
1096 depends on !PARISC_PAGE_SIZE_64KB
1097 depends on PAGE_SIZE_LESS_THAN_256KB
1099 config PAGE_SIZE_LESS_THAN_256KB
1101 depends on !PAGE_SIZE_256KB
1103 # This allows to use a set of generic functions to determine mmap base
1104 # address by giving priority to top-down scheme only if the process
1105 # is not in legacy mode (compat task, unlimited stack size or
1106 # sysctl_legacy_va_layout).
1107 # Architecture that selects this option can provide its own version of:
1109 config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT
1112 select ARCH_HAS_ELF_RANDOMIZE
1117 config HAVE_JUMP_LABEL_HACK
1120 config HAVE_NOINSTR_HACK
1123 config HAVE_NOINSTR_VALIDATION
1126 config HAVE_UACCESS_VALIDATION
1130 config HAVE_STACK_VALIDATION
1133 Architecture supports objtool compile-time frame pointer rule
1136 config HAVE_RELIABLE_STACKTRACE
1139 Architecture has either save_stack_trace_tsk_reliable() or
1140 arch_stack_walk_reliable() function which only returns a stack trace
1141 if it can guarantee the trace is reliable.
1143 config HAVE_ARCH_HASH
1147 If this is set, the architecture provides an <asm/hash.h>
1148 file which provides platform-specific implementations of some
1149 functions in <linux/hash.h> or fs/namei.c.
1151 config HAVE_ARCH_NVRAM_OPS
1160 config CLONE_BACKWARDS
1163 Architecture has tls passed as the 4th argument of clone(2),
1166 config CLONE_BACKWARDS2
1169 Architecture has the first two arguments of clone(2) swapped.
1171 config CLONE_BACKWARDS3
1174 Architecture has tls passed as the 3rd argument of clone(2),
1177 config ODD_RT_SIGACTION
1180 Architecture has unusual rt_sigaction(2) arguments
1182 config OLD_SIGSUSPEND
1185 Architecture has old sigsuspend(2) syscall, of one-argument variety
1187 config OLD_SIGSUSPEND3
1190 Even weirder antique ABI - three-argument sigsuspend(2)
1192 config OLD_SIGACTION
1195 Architecture has old sigaction(2) syscall. Nope, not the same
1196 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
1197 but fairly different variant of sigaction(2), thanks to OSF/1
1200 config COMPAT_OLD_SIGACTION
1203 config COMPAT_32BIT_TIME
1204 bool "Provide system calls for 32-bit time_t"
1205 default !64BIT || COMPAT
1207 This enables 32 bit time_t support in addition to 64 bit time_t support.
1208 This is relevant on all 32-bit architectures, and 64-bit architectures
1209 as part of compat syscall handling.
1211 config ARCH_NO_PREEMPT
1214 config ARCH_SUPPORTS_RT
1217 config CPU_NO_EFFICIENT_FFS
1220 config HAVE_ARCH_VMAP_STACK
1223 An arch should select this symbol if it can support kernel stacks
1224 in vmalloc space. This means:
1226 - vmalloc space must be large enough to hold many kernel stacks.
1227 This may rule out many 32-bit architectures.
1229 - Stacks in vmalloc space need to work reliably. For example, if
1230 vmap page tables are created on demand, either this mechanism
1231 needs to work while the stack points to a virtual address with
1232 unpopulated page tables or arch code (switch_to() and switch_mm(),
1233 most likely) needs to ensure that the stack's page table entries
1234 are populated before running on a possibly unpopulated stack.
1236 - If the stack overflows into a guard page, something reasonable
1237 should happen. The definition of "reasonable" is flexible, but
1238 instantly rebooting without logging anything would be unfriendly.
1242 bool "Use a virtually-mapped stack"
1243 depends on HAVE_ARCH_VMAP_STACK
1244 depends on !KASAN || KASAN_HW_TAGS || KASAN_VMALLOC
1246 Enable this if you want the use virtually-mapped kernel stacks
1247 with guard pages. This causes kernel stack overflows to be
1248 caught immediately rather than causing difficult-to-diagnose
1251 To use this with software KASAN modes, the architecture must support
1252 backing virtual mappings with real shadow memory, and KASAN_VMALLOC
1255 config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1258 An arch should select this symbol if it can support kernel stack
1259 offset randomization with calls to add_random_kstack_offset()
1260 during syscall entry and choose_random_kstack_offset() during
1261 syscall exit. Careful removal of -fstack-protector-strong and
1262 -fstack-protector should also be applied to the entry code and
1263 closely examined, as the artificial stack bump looks like an array
1264 to the compiler, so it will attempt to add canary checks regardless
1265 of the static branch state.
1267 config RANDOMIZE_KSTACK_OFFSET
1268 bool "Support for randomizing kernel stack offset on syscall entry" if EXPERT
1270 depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1271 depends on INIT_STACK_NONE || !CC_IS_CLANG || CLANG_VERSION >= 140000
1273 The kernel stack offset can be randomized (after pt_regs) by
1274 roughly 5 bits of entropy, frustrating memory corruption
1275 attacks that depend on stack address determinism or
1276 cross-syscall address exposures.
1278 The feature is controlled via the "randomize_kstack_offset=on/off"
1279 kernel boot param, and if turned off has zero overhead due to its use
1280 of static branches (see JUMP_LABEL).
1284 config RANDOMIZE_KSTACK_OFFSET_DEFAULT
1285 bool "Default state of kernel stack offset randomization"
1286 depends on RANDOMIZE_KSTACK_OFFSET
1288 Kernel stack offset randomization is controlled by kernel boot param
1289 "randomize_kstack_offset=on/off", and this config chooses the default
1292 config ARCH_OPTIONAL_KERNEL_RWX
1295 config ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1298 config ARCH_HAS_STRICT_KERNEL_RWX
1301 config STRICT_KERNEL_RWX
1302 bool "Make kernel text and rodata read-only" if ARCH_OPTIONAL_KERNEL_RWX
1303 depends on ARCH_HAS_STRICT_KERNEL_RWX
1304 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1306 If this is set, kernel text and rodata memory will be made read-only,
1307 and non-text memory will be made non-executable. This provides
1308 protection against certain security exploits (e.g. executing the heap
1311 These features are considered standard security practice these days.
1312 You should say Y here in almost all cases.
1314 config ARCH_HAS_STRICT_MODULE_RWX
1317 config STRICT_MODULE_RWX
1318 bool "Set loadable kernel module data as NX and text as RO" if ARCH_OPTIONAL_KERNEL_RWX
1319 depends on ARCH_HAS_STRICT_MODULE_RWX && MODULES
1320 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1322 If this is set, module text and rodata memory will be made read-only,
1323 and non-text memory will be made non-executable. This provides
1324 protection against certain security exploits (e.g. writing to text)
1326 # select if the architecture provides an asm/dma-direct.h header
1327 config ARCH_HAS_PHYS_TO_DMA
1330 config HAVE_ARCH_COMPILER_H
1333 An architecture can select this if it provides an
1334 asm/compiler.h header that should be included after
1335 linux/compiler-*.h in order to override macro definitions that those
1336 headers generally provide.
1338 config HAVE_ARCH_PREL32_RELOCATIONS
1341 May be selected by an architecture if it supports place-relative
1342 32-bit relocations, both in the toolchain and in the module loader,
1343 in which case relative references can be used in special sections
1344 for PCI fixup, initcalls etc which are only half the size on 64 bit
1345 architectures, and don't require runtime relocation on relocatable
1348 config ARCH_USE_MEMREMAP_PROT
1351 config LOCK_EVENT_COUNTS
1352 bool "Locking event counts collection"
1355 Enable light-weight counting of various locking related events
1356 in the system with minimal performance impact. This reduces
1357 the chance of application behavior change because of timing
1358 differences. The counts are reported via debugfs.
1360 # Select if the architecture has support for applying RELR relocations.
1361 config ARCH_HAS_RELR
1365 bool "Use RELR relocation packing"
1366 depends on ARCH_HAS_RELR && TOOLS_SUPPORT_RELR
1369 Store the kernel's dynamic relocations in the RELR relocation packing
1370 format. Requires a compatible linker (LLD supports this feature), as
1371 well as compatible NM and OBJCOPY utilities (llvm-nm and llvm-objcopy
1374 config ARCH_HAS_MEM_ENCRYPT
1377 config ARCH_HAS_CC_PLATFORM
1380 config HAVE_SPARSE_SYSCALL_NR
1383 An architecture should select this if its syscall numbering is sparse
1384 to save space. For example, MIPS architecture has a syscall array with
1385 entries at 4000, 5000 and 6000 locations. This option turns on syscall
1386 related optimizations for a given architecture.
1388 config ARCH_HAS_VDSO_DATA
1391 config HAVE_STATIC_CALL
1394 config HAVE_STATIC_CALL_INLINE
1396 depends on HAVE_STATIC_CALL
1399 config HAVE_PREEMPT_DYNAMIC
1402 config HAVE_PREEMPT_DYNAMIC_CALL
1404 depends on HAVE_STATIC_CALL
1405 select HAVE_PREEMPT_DYNAMIC
1407 An architecture should select this if it can handle the preemption
1408 model being selected at boot time using static calls.
1410 Where an architecture selects HAVE_STATIC_CALL_INLINE, any call to a
1411 preemption function will be patched directly.
1413 Where an architecture does not select HAVE_STATIC_CALL_INLINE, any
1414 call to a preemption function will go through a trampoline, and the
1415 trampoline will be patched.
1417 It is strongly advised to support inline static call to avoid any
1420 config HAVE_PREEMPT_DYNAMIC_KEY
1422 depends on HAVE_ARCH_JUMP_LABEL
1423 select HAVE_PREEMPT_DYNAMIC
1425 An architecture should select this if it can handle the preemption
1426 model being selected at boot time using static keys.
1428 Each preemption function will be given an early return based on a
1429 static key. This should have slightly lower overhead than non-inline
1430 static calls, as this effectively inlines each trampoline into the
1431 start of its callee. This may avoid redundant work, and may
1432 integrate better with CFI schemes.
1434 This will have greater overhead than using inline static calls as
1435 the call to the preemption function cannot be entirely elided.
1437 config ARCH_WANT_LD_ORPHAN_WARN
1440 An arch should select this symbol once all linker sections are explicitly
1441 included, size-asserted, or discarded in the linker scripts. This is
1442 important because we never want expected sections to be placed heuristically
1443 by the linker, since the locations of such sections can change between linker
1446 config HAVE_ARCH_PFN_VALID
1449 config ARCH_SUPPORTS_DEBUG_PAGEALLOC
1452 config ARCH_SUPPORTS_PAGE_TABLE_CHECK
1455 config ARCH_SPLIT_ARG64
1458 If a 32-bit architecture requires 64-bit arguments to be split into
1459 pairs of 32-bit arguments, select this option.
1461 config ARCH_HAS_ELFCORE_COMPAT
1464 config ARCH_HAS_PARANOID_L1D_FLUSH
1467 config ARCH_HAVE_TRACE_MMIO_ACCESS
1470 config DYNAMIC_SIGFRAME
1473 # Select, if arch has a named attribute group bound to NUMA device nodes.
1474 config HAVE_ARCH_NODE_DEV_GROUP
1477 config ARCH_HAS_NONLEAF_PMD_YOUNG
1480 Architectures that select this option are capable of setting the
1481 accessed bit in non-leaf PMD entries when using them as part of linear
1482 address translations. Page table walkers that clear the accessed bit
1483 may use this capability to reduce their search space.
1485 source "kernel/gcov/Kconfig"
1487 source "scripts/gcc-plugins/Kconfig"
1489 config FUNCTION_ALIGNMENT_4B
1492 config FUNCTION_ALIGNMENT_8B
1495 config FUNCTION_ALIGNMENT_16B
1498 config FUNCTION_ALIGNMENT_32B
1501 config FUNCTION_ALIGNMENT_64B
1504 config FUNCTION_ALIGNMENT
1506 default 64 if FUNCTION_ALIGNMENT_64B
1507 default 32 if FUNCTION_ALIGNMENT_32B
1508 default 16 if FUNCTION_ALIGNMENT_16B
1509 default 8 if FUNCTION_ALIGNMENT_8B
1510 default 4 if FUNCTION_ALIGNMENT_4B