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 config ARCH_CONFIGURES_CPU_MITIGATIONS
15 if !ARCH_CONFIGURES_CPU_MITIGATIONS
16 config CPU_MITIGATIONS
20 menu "General architecture-dependent options"
22 config ARCH_HAS_SUBPAGE_FAULTS
25 Select if the architecture can check permissions at sub-page
26 granularity (e.g. arm64 MTE). The probe_user_*() functions
32 config SMT_NUM_THREADS_DYNAMIC
35 # Selected by HOTPLUG_CORE_SYNC_DEAD or HOTPLUG_CORE_SYNC_FULL
36 config HOTPLUG_CORE_SYNC
39 # Basic CPU dead synchronization selected by architecture
40 config HOTPLUG_CORE_SYNC_DEAD
42 select HOTPLUG_CORE_SYNC
44 # Full CPU synchronization with alive state selected by architecture
45 config HOTPLUG_CORE_SYNC_FULL
47 select HOTPLUG_CORE_SYNC_DEAD if HOTPLUG_CPU
48 select HOTPLUG_CORE_SYNC
50 config HOTPLUG_SPLIT_STARTUP
52 select HOTPLUG_CORE_SYNC_FULL
54 config HOTPLUG_PARALLEL
56 select HOTPLUG_SPLIT_STARTUP
64 depends on HAVE_KPROBES
66 select TASKS_RCU if PREEMPTION
68 Kprobes allows you to trap at almost any kernel address and
69 execute a callback function. register_kprobe() establishes
70 a probepoint and specifies the callback. Kprobes is useful
71 for kernel debugging, non-intrusive instrumentation and testing.
75 bool "Optimize very unlikely/likely branches"
76 depends on HAVE_ARCH_JUMP_LABEL
77 select OBJTOOL if HAVE_JUMP_LABEL_HACK
79 This option enables a transparent branch optimization that
80 makes certain almost-always-true or almost-always-false branch
81 conditions even cheaper to execute within the kernel.
83 Certain performance-sensitive kernel code, such as trace points,
84 scheduler functionality, networking code and KVM have such
85 branches and include support for this optimization technique.
87 If it is detected that the compiler has support for "asm goto",
88 the kernel will compile such branches with just a nop
89 instruction. When the condition flag is toggled to true, the
90 nop will be converted to a jump instruction to execute the
91 conditional block of instructions.
93 This technique lowers overhead and stress on the branch prediction
94 of the processor and generally makes the kernel faster. The update
95 of the condition is slower, but those are always very rare.
97 ( On 32-bit x86, the necessary options added to the compiler
98 flags may increase the size of the kernel slightly. )
100 config STATIC_KEYS_SELFTEST
101 bool "Static key selftest"
102 depends on JUMP_LABEL
104 Boot time self-test of the branch patching code.
106 config STATIC_CALL_SELFTEST
107 bool "Static call selftest"
108 depends on HAVE_STATIC_CALL
110 Boot time self-test of the call patching code.
114 depends on KPROBES && HAVE_OPTPROBES
115 select TASKS_RCU if PREEMPTION
117 config KPROBES_ON_FTRACE
119 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
120 depends on DYNAMIC_FTRACE_WITH_REGS
122 If function tracer is enabled and the arch supports full
123 passing of pt_regs to function tracing, then kprobes can
124 optimize on top of function tracing.
128 depends on ARCH_SUPPORTS_UPROBES
130 Uprobes is the user-space counterpart to kprobes: they
131 enable instrumentation applications (such as 'perf probe')
132 to establish unintrusive probes in user-space binaries and
133 libraries, by executing handler functions when the probes
134 are hit by user-space applications.
136 ( These probes come in the form of single-byte breakpoints,
137 managed by the kernel and kept transparent to the probed
140 config HAVE_64BIT_ALIGNED_ACCESS
141 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
143 Some architectures require 64 bit accesses to be 64 bit
144 aligned, which also requires structs containing 64 bit values
145 to be 64 bit aligned too. This includes some 32 bit
146 architectures which can do 64 bit accesses, as well as 64 bit
147 architectures without unaligned access.
149 This symbol should be selected by an architecture if 64 bit
150 accesses are required to be 64 bit aligned in this way even
151 though it is not a 64 bit architecture.
153 See Documentation/core-api/unaligned-memory-access.rst for
154 more information on the topic of unaligned memory accesses.
156 config HAVE_EFFICIENT_UNALIGNED_ACCESS
159 Some architectures are unable to perform unaligned accesses
160 without the use of get_unaligned/put_unaligned. Others are
161 unable to perform such accesses efficiently (e.g. trap on
162 unaligned access and require fixing it up in the exception
165 This symbol should be selected by an architecture if it can
166 perform unaligned accesses efficiently to allow different
167 code paths to be selected for these cases. Some network
168 drivers, for example, could opt to not fix up alignment
169 problems with received packets if doing so would not help
172 See Documentation/core-api/unaligned-memory-access.rst for more
173 information on the topic of unaligned memory accesses.
175 config ARCH_USE_BUILTIN_BSWAP
178 Modern versions of GCC (since 4.4) have builtin functions
179 for handling byte-swapping. Using these, instead of the old
180 inline assembler that the architecture code provides in the
181 __arch_bswapXX() macros, allows the compiler to see what's
182 happening and offers more opportunity for optimisation. In
183 particular, the compiler will be able to combine the byteswap
184 with a nearby load or store and use load-and-swap or
185 store-and-swap instructions if the architecture has them. It
186 should almost *never* result in code which is worse than the
187 hand-coded assembler in <asm/swab.h>. But just in case it
188 does, the use of the builtins is optional.
190 Any architecture with load-and-swap or store-and-swap
191 instructions should set this. And it shouldn't hurt to set it
192 on architectures that don't have such instructions.
196 depends on KPROBES && (HAVE_KRETPROBES || HAVE_RETHOOK)
198 config KRETPROBE_ON_RETHOOK
200 depends on HAVE_RETHOOK
201 depends on KRETPROBES
204 config USER_RETURN_NOTIFIER
206 depends on HAVE_USER_RETURN_NOTIFIER
208 Provide a kernel-internal notification when a cpu is about to
211 config HAVE_IOREMAP_PROT
217 config HAVE_KRETPROBES
220 config HAVE_OPTPROBES
223 config HAVE_KPROBES_ON_FTRACE
226 config ARCH_CORRECT_STACKTRACE_ON_KRETPROBE
229 Since kretprobes modifies return address on the stack, the
230 stacktrace may see the kretprobe trampoline address instead
231 of correct one. If the architecture stacktrace code and
232 unwinder can adjust such entries, select this configuration.
234 config HAVE_FUNCTION_ERROR_INJECTION
240 config HAVE_FUNCTION_DESCRIPTORS
243 config TRACE_IRQFLAGS_SUPPORT
246 config TRACE_IRQFLAGS_NMI_SUPPORT
250 # An arch should select this if it provides all these things:
252 # task_pt_regs() in asm/processor.h or asm/ptrace.h
253 # arch_has_single_step() if there is hardware single-step support
254 # arch_has_block_step() if there is hardware block-step support
255 # asm/syscall.h supplying asm-generic/syscall.h interface
256 # linux/regset.h user_regset interfaces
257 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
258 # TIF_SYSCALL_TRACE calls ptrace_report_syscall_{entry,exit}
259 # TIF_NOTIFY_RESUME calls resume_user_mode_work()
261 config HAVE_ARCH_TRACEHOOK
264 config HAVE_DMA_CONTIGUOUS
267 config GENERIC_SMP_IDLE_THREAD
270 config GENERIC_IDLE_POLL_SETUP
273 config ARCH_HAS_FORTIFY_SOURCE
276 An architecture should select this when it can successfully
277 build and run with CONFIG_FORTIFY_SOURCE.
280 # Select if the arch provides a historic keepinit alias for the retain_initrd
281 # command line option
283 config ARCH_HAS_KEEPINITRD
286 # Select if arch has all set_memory_ro/rw/x/nx() functions in asm/cacheflush.h
287 config ARCH_HAS_SET_MEMORY
290 # Select if arch has all set_direct_map_invalid/default() functions
291 config ARCH_HAS_SET_DIRECT_MAP
295 # Select if the architecture provides the arch_dma_set_uncached symbol to
296 # either provide an uncached segment alias for a DMA allocation, or
297 # to remap the page tables in place.
299 config ARCH_HAS_DMA_SET_UNCACHED
303 # Select if the architectures provides the arch_dma_clear_uncached symbol
304 # to undo an in-place page table remap for uncached access.
306 config ARCH_HAS_DMA_CLEAR_UNCACHED
309 config ARCH_HAS_CPU_FINALIZE_INIT
312 # The architecture has a per-task state that includes the mm's PASID
313 config ARCH_HAS_CPU_PASID
317 config HAVE_ARCH_THREAD_STRUCT_WHITELIST
320 An architecture should select this to provide hardened usercopy
321 knowledge about what region of the thread_struct should be
322 whitelisted for copying to userspace. Normally this is only the
323 FPU registers. Specifically, arch_thread_struct_whitelist()
324 should be implemented. Without this, the entire thread_struct
325 field in task_struct will be left whitelisted.
327 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
328 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
331 config ARCH_WANTS_NO_INSTR
334 An architecture should select this if the noinstr macro is being used on
335 functions to denote that the toolchain should avoid instrumenting such
336 functions and is required for correctness.
338 config ARCH_32BIT_OFF_T
342 All new 32-bit architectures should have 64-bit off_t type on
343 userspace side which corresponds to the loff_t kernel type. This
344 is the requirement for modern ABIs. Some existing architectures
345 still support 32-bit off_t. This option is enabled for all such
346 architectures explicitly.
348 # Selected by 64 bit architectures which have a 32 bit f_tinode in struct ustat
349 config ARCH_32BIT_USTAT_F_TINODE
352 config HAVE_ASM_MODVERSIONS
355 This symbol should be selected by an architecture if it provides
356 <asm/asm-prototypes.h> to support the module versioning for symbols
357 exported from assembly code.
359 config HAVE_REGS_AND_STACK_ACCESS_API
362 This symbol should be selected by an architecture if it supports
363 the API needed to access registers and stack entries from pt_regs,
364 declared in asm/ptrace.h
365 For example the kprobes-based event tracer needs this API.
369 depends on HAVE_REGS_AND_STACK_ACCESS_API
371 This symbol should be selected by an architecture if it
372 supports an implementation of restartable sequences.
377 This symbol should be selected by an architecture if it
380 config HAVE_FUNCTION_ARG_ACCESS_API
383 This symbol should be selected by an architecture if it supports
384 the API needed to access function arguments from pt_regs,
385 declared in asm/ptrace.h
387 config HAVE_HW_BREAKPOINT
389 depends on PERF_EVENTS
391 config HAVE_MIXED_BREAKPOINTS_REGS
393 depends on HAVE_HW_BREAKPOINT
395 Depending on the arch implementation of hardware breakpoints,
396 some of them have separate registers for data and instruction
397 breakpoints addresses, others have mixed registers to store
398 them but define the access type in a control register.
399 Select this option if your arch implements breakpoints under the
402 config HAVE_USER_RETURN_NOTIFIER
405 config HAVE_PERF_EVENTS_NMI
408 System hardware can generate an NMI using the perf event
409 subsystem. Also has support for calculating CPU cycle events
410 to determine how many clock cycles in a given period.
412 config HAVE_HARDLOCKUP_DETECTOR_PERF
414 depends on HAVE_PERF_EVENTS_NMI
416 The arch chooses to use the generic perf-NMI-based hardlockup
417 detector. Must define HAVE_PERF_EVENTS_NMI.
419 config HAVE_HARDLOCKUP_DETECTOR_ARCH
422 The arch provides its own hardlockup detector implementation instead
425 It uses the same command line parameters, and sysctl interface,
426 as the generic hardlockup detectors.
428 config HAVE_PERF_REGS
431 Support selective register dumps for perf events. This includes
432 bit-mapping of each registers and a unique architecture id.
434 config HAVE_PERF_USER_STACK_DUMP
437 Support user stack dumps for perf event samples. This needs
438 access to the user stack pointer which is not unified across
441 config HAVE_ARCH_JUMP_LABEL
444 config HAVE_ARCH_JUMP_LABEL_RELATIVE
447 config MMU_GATHER_TABLE_FREE
450 config MMU_GATHER_RCU_TABLE_FREE
452 select MMU_GATHER_TABLE_FREE
454 config MMU_GATHER_PAGE_SIZE
457 config MMU_GATHER_NO_RANGE
459 select MMU_GATHER_MERGE_VMAS
461 config MMU_GATHER_NO_FLUSH_CACHE
464 config MMU_GATHER_MERGE_VMAS
467 config MMU_GATHER_NO_GATHER
469 depends on MMU_GATHER_TABLE_FREE
471 config ARCH_WANT_IRQS_OFF_ACTIVATE_MM
474 Temporary select until all architectures can be converted to have
475 irqs disabled over activate_mm. Architectures that do IPI based TLB
476 shootdowns should enable this.
478 # Use normal mm refcounting for MMU_LAZY_TLB kernel thread references.
479 # MMU_LAZY_TLB_REFCOUNT=n can improve the scalability of context switching
480 # to/from kernel threads when the same mm is running on a lot of CPUs (a large
481 # multi-threaded application), by reducing contention on the mm refcount.
483 # This can be disabled if the architecture ensures no CPUs are using an mm as a
484 # "lazy tlb" beyond its final refcount (i.e., by the time __mmdrop frees the mm
485 # or its kernel page tables). This could be arranged by arch_exit_mmap(), or
486 # final exit(2) TLB flush, for example.
488 # To implement this, an arch *must*:
489 # Ensure the _lazy_tlb variants of mmgrab/mmdrop are used when manipulating
490 # the lazy tlb reference of a kthread's ->active_mm (non-arch code has been
491 # converted already).
492 config MMU_LAZY_TLB_REFCOUNT
494 depends on !MMU_LAZY_TLB_SHOOTDOWN
496 # This option allows MMU_LAZY_TLB_REFCOUNT=n. It ensures no CPUs are using an
497 # mm as a lazy tlb beyond its last reference count, by shooting down these
498 # users before the mm is deallocated. __mmdrop() first IPIs all CPUs that may
499 # be using the mm as a lazy tlb, so that they may switch themselves to using
500 # init_mm for their active mm. mm_cpumask(mm) is used to determine which CPUs
501 # may be using mm as a lazy tlb mm.
503 # To implement this, an arch *must*:
504 # - At the time of the final mmdrop of the mm, ensure mm_cpumask(mm) contains
505 # at least all possible CPUs in which the mm is lazy.
506 # - It must meet the requirements for MMU_LAZY_TLB_REFCOUNT=n (see above).
507 config MMU_LAZY_TLB_SHOOTDOWN
510 config ARCH_HAVE_NMI_SAFE_CMPXCHG
513 config ARCH_HAS_NMI_SAFE_THIS_CPU_OPS
516 config HAVE_ALIGNED_STRUCT_PAGE
519 This makes sure that struct pages are double word aligned and that
520 e.g. the SLUB allocator can perform double word atomic operations
521 on a struct page for better performance. However selecting this
522 might increase the size of a struct page by a word.
524 config HAVE_CMPXCHG_LOCAL
527 config HAVE_CMPXCHG_DOUBLE
530 config ARCH_WEAK_RELEASE_ACQUIRE
533 config ARCH_WANT_IPC_PARSE_VERSION
536 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
539 config ARCH_WANT_OLD_COMPAT_IPC
540 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
543 config HAVE_ARCH_SECCOMP
546 An arch should select this symbol to support seccomp mode 1 (the fixed
547 syscall policy), and must provide an overrides for __NR_seccomp_sigreturn,
548 and compat syscalls if the asm-generic/seccomp.h defaults need adjustment:
549 - __NR_seccomp_read_32
550 - __NR_seccomp_write_32
551 - __NR_seccomp_exit_32
552 - __NR_seccomp_sigreturn_32
554 config HAVE_ARCH_SECCOMP_FILTER
556 select HAVE_ARCH_SECCOMP
558 An arch should select this symbol if it provides all of these things:
559 - all the requirements for HAVE_ARCH_SECCOMP
561 - syscall_get_arguments()
563 - syscall_set_return_value()
564 - SIGSYS siginfo_t support
565 - secure_computing is called from a ptrace_event()-safe context
566 - secure_computing return value is checked and a return value of -1
567 results in the system call being skipped immediately.
568 - seccomp syscall wired up
569 - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE,
570 SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If
571 COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too.
574 prompt "Enable seccomp to safely execute untrusted bytecode"
576 depends on HAVE_ARCH_SECCOMP
578 This kernel feature is useful for number crunching applications
579 that may need to handle untrusted bytecode during their
580 execution. By using pipes or other transports made available
581 to the process as file descriptors supporting the read/write
582 syscalls, it's possible to isolate those applications in their
583 own address space using seccomp. Once seccomp is enabled via
584 prctl(PR_SET_SECCOMP) or the seccomp() syscall, it cannot be
585 disabled and the task is only allowed to execute a few safe
586 syscalls defined by each seccomp mode.
590 config SECCOMP_FILTER
592 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
594 Enable tasks to build secure computing environments defined
595 in terms of Berkeley Packet Filter programs which implement
596 task-defined system call filtering polices.
598 See Documentation/userspace-api/seccomp_filter.rst for details.
600 config SECCOMP_CACHE_DEBUG
601 bool "Show seccomp filter cache status in /proc/pid/seccomp_cache"
602 depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR
605 This enables the /proc/pid/seccomp_cache interface to monitor
606 seccomp cache data. The file format is subject to change. Reading
607 the file requires CAP_SYS_ADMIN.
609 This option is for debugging only. Enabling presents the risk that
610 an adversary may be able to infer the seccomp filter logic.
614 config HAVE_ARCH_STACKLEAK
617 An architecture should select this if it has the code which
618 fills the used part of the kernel stack with the STACKLEAK_POISON
619 value before returning from system calls.
621 config HAVE_STACKPROTECTOR
624 An arch should select this symbol if:
625 - it has implemented a stack canary (e.g. __stack_chk_guard)
627 config STACKPROTECTOR
628 bool "Stack Protector buffer overflow detection"
629 depends on HAVE_STACKPROTECTOR
630 depends on $(cc-option,-fstack-protector)
633 This option turns on the "stack-protector" GCC feature. This
634 feature puts, at the beginning of functions, a canary value on
635 the stack just before the return address, and validates
636 the value just before actually returning. Stack based buffer
637 overflows (that need to overwrite this return address) now also
638 overwrite the canary, which gets detected and the attack is then
639 neutralized via a kernel panic.
641 Functions will have the stack-protector canary logic added if they
642 have an 8-byte or larger character array on the stack.
644 This feature requires gcc version 4.2 or above, or a distribution
645 gcc with the feature backported ("-fstack-protector").
647 On an x86 "defconfig" build, this feature adds canary checks to
648 about 3% of all kernel functions, which increases kernel code size
651 config STACKPROTECTOR_STRONG
652 bool "Strong Stack Protector"
653 depends on STACKPROTECTOR
654 depends on $(cc-option,-fstack-protector-strong)
657 Functions will have the stack-protector canary logic added in any
658 of the following conditions:
660 - local variable's address used as part of the right hand side of an
661 assignment or function argument
662 - local variable is an array (or union containing an array),
663 regardless of array type or length
664 - uses register local variables
666 This feature requires gcc version 4.9 or above, or a distribution
667 gcc with the feature backported ("-fstack-protector-strong").
669 On an x86 "defconfig" build, this feature adds canary checks to
670 about 20% of all kernel functions, which increases the kernel code
673 config ARCH_SUPPORTS_SHADOW_CALL_STACK
676 An architecture should select this if it supports the compiler's
677 Shadow Call Stack and implements runtime support for shadow stack
680 config SHADOW_CALL_STACK
681 bool "Shadow Call Stack"
682 depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
683 depends on DYNAMIC_FTRACE_WITH_ARGS || DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
686 This option enables the compiler's Shadow Call Stack, which
687 uses a shadow stack to protect function return addresses from
688 being overwritten by an attacker. More information can be found
689 in the compiler's documentation:
691 - Clang: https://clang.llvm.org/docs/ShadowCallStack.html
692 - GCC: https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html#Instrumentation-Options
694 Note that security guarantees in the kernel differ from the
695 ones documented for user space. The kernel must store addresses
696 of shadow stacks in memory, which means an attacker capable of
697 reading and writing arbitrary memory may be able to locate them
698 and hijack control flow by modifying the stacks.
703 Set by the arch code if it relies on code patching to insert the
704 shadow call stack push and pop instructions rather than on the
710 Selected if the kernel will be built using the compiler's LTO feature.
716 Selected if the kernel will be built using Clang's LTO feature.
718 config ARCH_SUPPORTS_LTO_CLANG
721 An architecture should select this option if it supports:
722 - compiling with Clang,
723 - compiling inline assembly with Clang's integrated assembler,
724 - and linking with LLD.
726 config ARCH_SUPPORTS_LTO_CLANG_THIN
729 An architecture should select this option if it can support Clang's
734 depends on CC_IS_CLANG && LD_IS_LLD && AS_IS_LLVM
735 depends on $(success,$(NM) --help | head -n 1 | grep -qi llvm)
736 depends on $(success,$(AR) --help | head -n 1 | grep -qi llvm)
737 depends on ARCH_SUPPORTS_LTO_CLANG
738 depends on !FTRACE_MCOUNT_USE_RECORDMCOUNT
739 # https://github.com/ClangBuiltLinux/linux/issues/1721
740 depends on (!KASAN || KASAN_HW_TAGS || CLANG_VERSION >= 170000) || !DEBUG_INFO
741 depends on (!KCOV || CLANG_VERSION >= 170000) || !DEBUG_INFO
742 depends on !GCOV_KERNEL
744 The compiler and Kconfig options support building with Clang's
748 prompt "Link Time Optimization (LTO)"
751 This option enables Link Time Optimization (LTO), which allows the
752 compiler to optimize binaries globally.
754 If unsure, select LTO_NONE. Note that LTO is very resource-intensive
755 so it's disabled by default.
760 Build the kernel normally, without Link Time Optimization (LTO).
762 config LTO_CLANG_FULL
763 bool "Clang Full LTO (EXPERIMENTAL)"
764 depends on HAS_LTO_CLANG
765 depends on !COMPILE_TEST
768 This option enables Clang's full Link Time Optimization (LTO), which
769 allows the compiler to optimize the kernel globally. If you enable
770 this option, the compiler generates LLVM bitcode instead of ELF
771 object files, and the actual compilation from bitcode happens at
772 the LTO link step, which may take several minutes depending on the
773 kernel configuration. More information can be found from LLVM's
776 https://llvm.org/docs/LinkTimeOptimization.html
778 During link time, this option can use a large amount of RAM, and
779 may take much longer than the ThinLTO option.
781 config LTO_CLANG_THIN
782 bool "Clang ThinLTO (EXPERIMENTAL)"
783 depends on HAS_LTO_CLANG && ARCH_SUPPORTS_LTO_CLANG_THIN
786 This option enables Clang's ThinLTO, which allows for parallel
787 optimization and faster incremental compiles compared to the
788 CONFIG_LTO_CLANG_FULL option. More information can be found
789 from Clang's documentation:
791 https://clang.llvm.org/docs/ThinLTO.html
796 config ARCH_SUPPORTS_CFI_CLANG
799 An architecture should select this option if it can support Clang's
800 Control-Flow Integrity (CFI) checking.
802 config ARCH_USES_CFI_TRAPS
806 bool "Use Clang's Control Flow Integrity (CFI)"
807 depends on ARCH_SUPPORTS_CFI_CLANG
808 depends on $(cc-option,-fsanitize=kcfi)
810 This option enables Clang's forward-edge Control Flow Integrity
811 (CFI) checking, where the compiler injects a runtime check to each
812 indirect function call to ensure the target is a valid function with
813 the correct static type. This restricts possible call targets and
814 makes it more difficult for an attacker to exploit bugs that allow
815 the modification of stored function pointers. More information can be
816 found from Clang's documentation:
818 https://clang.llvm.org/docs/ControlFlowIntegrity.html
820 config CFI_PERMISSIVE
821 bool "Use CFI in permissive mode"
824 When selected, Control Flow Integrity (CFI) violations result in a
825 warning instead of a kernel panic. This option should only be used
826 for finding indirect call type mismatches during development.
830 config HAVE_ARCH_WITHIN_STACK_FRAMES
833 An architecture should select this if it can walk the kernel stack
834 frames to determine if an object is part of either the arguments
835 or local variables (i.e. that it excludes saved return addresses,
836 and similar) by implementing an inline arch_within_stack_frames(),
837 which is used by CONFIG_HARDENED_USERCOPY.
839 config HAVE_CONTEXT_TRACKING_USER
842 Provide kernel/user boundaries probes necessary for subsystems
843 that need it, such as userspace RCU extended quiescent state.
844 Syscalls need to be wrapped inside user_exit()-user_enter(), either
845 optimized behind static key or through the slow path using TIF_NOHZ
846 flag. Exceptions handlers must be wrapped as well. Irqs are already
847 protected inside ct_irq_enter/ct_irq_exit() but preemption or signal
848 handling on irq exit still need to be protected.
850 config HAVE_CONTEXT_TRACKING_USER_OFFSTACK
853 Architecture neither relies on exception_enter()/exception_exit()
854 nor on schedule_user(). Also preempt_schedule_notrace() and
855 preempt_schedule_irq() can't be called in a preemptible section
856 while context tracking is CONTEXT_USER. This feature reflects a sane
857 entry implementation where the following requirements are met on
858 critical entry code, ie: before user_exit() or after user_enter():
860 - Critical entry code isn't preemptible (or better yet:
862 - No use of RCU read side critical sections, unless ct_nmi_enter()
864 - No use of instrumentation, unless instrumentation_begin() got
870 Arch relies on TIF_NOHZ and syscall slow path to implement context
871 tracking calls to user_enter()/user_exit().
873 config HAVE_VIRT_CPU_ACCOUNTING
876 config HAVE_VIRT_CPU_ACCOUNTING_IDLE
879 Architecture has its own way to account idle CPU time and therefore
880 doesn't implement vtime_account_idle().
882 config ARCH_HAS_SCALED_CPUTIME
885 config HAVE_VIRT_CPU_ACCOUNTING_GEN
889 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
890 Before enabling this option, arch code must be audited
891 to ensure there are no races in concurrent read/write of
892 cputime_t. For example, reading/writing 64-bit cputime_t on
893 some 32-bit arches may require multiple accesses, so proper
894 locking is needed to protect against concurrent accesses.
896 config HAVE_IRQ_TIME_ACCOUNTING
899 Archs need to ensure they use a high enough resolution clock to
900 support irq time accounting and then call enable_sched_clock_irqtime().
905 Architectures that select this are able to move page tables at the
906 PUD level. If there are only 3 page table levels, the move effectively
907 happens at the PGD level.
912 Archs that select this are able to move page tables at the PMD level.
914 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
917 config HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
920 config HAVE_ARCH_HUGE_VMAP
924 # Archs that select this would be capable of PMD-sized vmaps (i.e.,
925 # arch_vmap_pmd_supported() returns true). The VM_ALLOW_HUGE_VMAP flag
926 # must be used to enable allocations to use hugepages.
928 config HAVE_ARCH_HUGE_VMALLOC
929 depends on HAVE_ARCH_HUGE_VMAP
932 config ARCH_WANT_HUGE_PMD_SHARE
935 # Archs that want to use pmd_mkwrite on kernel memory need it defined even
936 # if there are no userspace memory management features that use it
937 config ARCH_WANT_KERNEL_PMD_MKWRITE
940 config ARCH_WANT_PMD_MKWRITE
941 def_bool TRANSPARENT_HUGEPAGE || ARCH_WANT_KERNEL_PMD_MKWRITE
943 config HAVE_ARCH_SOFT_DIRTY
946 config HAVE_MOD_ARCH_SPECIFIC
949 The arch uses struct mod_arch_specific to store data. Many arches
950 just need a simple module loader without arch specific data - those
951 should not enable this.
953 config MODULES_USE_ELF_RELA
956 Modules only use ELF RELA relocations. Modules with ELF REL
957 relocations will give an error.
959 config MODULES_USE_ELF_REL
962 Modules only use ELF REL relocations. Modules with ELF RELA
963 relocations will give an error.
965 config ARCH_WANTS_MODULES_DATA_IN_VMALLOC
968 For architectures like powerpc/32 which have constraints on module
969 allocation and need to allocate module data outside of module area.
971 config HAVE_IRQ_EXIT_ON_IRQ_STACK
974 Architecture doesn't only execute the irq handler on the irq stack
975 but also irq_exit(). This way we can process softirqs on this irq
976 stack instead of switching to a new one when we call __do_softirq()
977 in the end of an hardirq.
978 This spares a stack switch and improves cache usage on softirq
981 config HAVE_SOFTIRQ_ON_OWN_STACK
984 Architecture provides a function to run __do_softirq() on a
987 config SOFTIRQ_ON_OWN_STACK
988 def_bool HAVE_SOFTIRQ_ON_OWN_STACK && !PREEMPT_RT
990 config ALTERNATE_USER_ADDRESS_SPACE
993 Architectures set this when the CPU uses separate address
994 spaces for kernel and user space pointers. In this case, the
995 access_ok() check on a __user pointer is skipped.
997 config PGTABLE_LEVELS
1001 config ARCH_HAS_ELF_RANDOMIZE
1004 An architecture supports choosing randomized locations for
1005 stack, mmap, brk, and ET_DYN. Defined functions:
1007 - arch_randomize_brk()
1009 config HAVE_ARCH_MMAP_RND_BITS
1012 An arch should select this symbol if it supports setting a variable
1013 number of bits for use in establishing the base address for mmap
1014 allocations, has MMU enabled and provides values for both:
1015 - ARCH_MMAP_RND_BITS_MIN
1016 - ARCH_MMAP_RND_BITS_MAX
1018 config HAVE_EXIT_THREAD
1021 An architecture implements exit_thread.
1023 config ARCH_MMAP_RND_BITS_MIN
1026 config ARCH_MMAP_RND_BITS_MAX
1029 config ARCH_MMAP_RND_BITS_DEFAULT
1032 config ARCH_MMAP_RND_BITS
1033 int "Number of bits to use for ASLR of mmap base address" if EXPERT
1034 range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX
1035 default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT
1036 default ARCH_MMAP_RND_BITS_MIN
1037 depends on HAVE_ARCH_MMAP_RND_BITS
1039 This value can be used to select the number of bits to use to
1040 determine the random offset to the base address of vma regions
1041 resulting from mmap allocations. This value will be bounded
1042 by the architecture's minimum and maximum supported values.
1044 This value can be changed after boot using the
1045 /proc/sys/vm/mmap_rnd_bits tunable
1047 config HAVE_ARCH_MMAP_RND_COMPAT_BITS
1050 An arch should select this symbol if it supports running applications
1051 in compatibility mode, supports setting a variable number of bits for
1052 use in establishing the base address for mmap allocations, has MMU
1053 enabled and provides values for both:
1054 - ARCH_MMAP_RND_COMPAT_BITS_MIN
1055 - ARCH_MMAP_RND_COMPAT_BITS_MAX
1057 config ARCH_MMAP_RND_COMPAT_BITS_MIN
1060 config ARCH_MMAP_RND_COMPAT_BITS_MAX
1063 config ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1066 config ARCH_MMAP_RND_COMPAT_BITS
1067 int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT
1068 range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX
1069 default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT
1070 default ARCH_MMAP_RND_COMPAT_BITS_MIN
1071 depends on HAVE_ARCH_MMAP_RND_COMPAT_BITS
1073 This value can be used to select the number of bits to use to
1074 determine the random offset to the base address of vma regions
1075 resulting from mmap allocations for compatible applications This
1076 value will be bounded by the architecture's minimum and maximum
1079 This value can be changed after boot using the
1080 /proc/sys/vm/mmap_rnd_compat_bits tunable
1082 config HAVE_ARCH_COMPAT_MMAP_BASES
1085 This allows 64bit applications to invoke 32-bit mmap() syscall
1086 and vice-versa 32-bit applications to call 64-bit mmap().
1087 Required for applications doing different bitness syscalls.
1089 config HAVE_PAGE_SIZE_4KB
1092 config HAVE_PAGE_SIZE_8KB
1095 config HAVE_PAGE_SIZE_16KB
1098 config HAVE_PAGE_SIZE_32KB
1101 config HAVE_PAGE_SIZE_64KB
1104 config HAVE_PAGE_SIZE_256KB
1108 prompt "MMU page size"
1110 config PAGE_SIZE_4KB
1112 depends on HAVE_PAGE_SIZE_4KB
1114 This option select the standard 4KiB Linux page size and the only
1115 available option on many architectures. Using 4KiB page size will
1116 minimize memory consumption and is therefore recommended for low
1118 Some software that is written for x86 systems makes incorrect
1119 assumptions about the page size and only runs on 4KiB pages.
1121 config PAGE_SIZE_8KB
1123 depends on HAVE_PAGE_SIZE_8KB
1125 This option is the only supported page size on a few older
1126 processors, and can be slightly faster than 4KiB pages.
1128 config PAGE_SIZE_16KB
1130 depends on HAVE_PAGE_SIZE_16KB
1132 This option is usually a good compromise between memory
1133 consumption and performance for typical desktop and server
1134 workloads, often saving a level of page table lookups compared
1135 to 4KB pages as well as reducing TLB pressure and overhead of
1136 per-page operations in the kernel at the expense of a larger
1139 config PAGE_SIZE_32KB
1141 depends on HAVE_PAGE_SIZE_32KB
1143 Using 32KiB page size will result in slightly higher performance
1144 kernel at the price of higher memory consumption compared to
1145 16KiB pages. This option is available only on cnMIPS cores.
1146 Note that you will need a suitable Linux distribution to
1149 config PAGE_SIZE_64KB
1151 depends on HAVE_PAGE_SIZE_64KB
1153 Using 64KiB page size will result in slightly higher performance
1154 kernel at the price of much higher memory consumption compared to
1155 4KiB or 16KiB pages.
1156 This is not suitable for general-purpose workloads but the
1157 better performance may be worth the cost for certain types of
1158 supercomputing or database applications that work mostly with
1159 large in-memory data rather than small files.
1161 config PAGE_SIZE_256KB
1163 depends on HAVE_PAGE_SIZE_256KB
1165 256KiB pages have little practical value due to their extreme
1166 memory usage. The kernel will only be able to run applications
1167 that have been compiled with '-zmax-page-size' set to 256KiB
1168 (the default is 64KiB or 4KiB on most architectures).
1172 config PAGE_SIZE_LESS_THAN_64KB
1174 depends on !PAGE_SIZE_64KB
1175 depends on PAGE_SIZE_LESS_THAN_256KB
1177 config PAGE_SIZE_LESS_THAN_256KB
1179 depends on !PAGE_SIZE_256KB
1183 default 12 if PAGE_SIZE_4KB
1184 default 13 if PAGE_SIZE_8KB
1185 default 14 if PAGE_SIZE_16KB
1186 default 15 if PAGE_SIZE_32KB
1187 default 16 if PAGE_SIZE_64KB
1188 default 18 if PAGE_SIZE_256KB
1190 # This allows to use a set of generic functions to determine mmap base
1191 # address by giving priority to top-down scheme only if the process
1192 # is not in legacy mode (compat task, unlimited stack size or
1193 # sysctl_legacy_va_layout).
1194 # Architecture that selects this option can provide its own version of:
1196 config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT
1199 select ARCH_HAS_ELF_RANDOMIZE
1204 config HAVE_JUMP_LABEL_HACK
1207 config HAVE_NOINSTR_HACK
1210 config HAVE_NOINSTR_VALIDATION
1213 config HAVE_UACCESS_VALIDATION
1217 config HAVE_STACK_VALIDATION
1220 Architecture supports objtool compile-time frame pointer rule
1223 config HAVE_RELIABLE_STACKTRACE
1226 Architecture has either save_stack_trace_tsk_reliable() or
1227 arch_stack_walk_reliable() function which only returns a stack trace
1228 if it can guarantee the trace is reliable.
1230 config HAVE_ARCH_HASH
1234 If this is set, the architecture provides an <asm/hash.h>
1235 file which provides platform-specific implementations of some
1236 functions in <linux/hash.h> or fs/namei.c.
1238 config HAVE_ARCH_NVRAM_OPS
1247 config CLONE_BACKWARDS
1250 Architecture has tls passed as the 4th argument of clone(2),
1253 config CLONE_BACKWARDS2
1256 Architecture has the first two arguments of clone(2) swapped.
1258 config CLONE_BACKWARDS3
1261 Architecture has tls passed as the 3rd argument of clone(2),
1264 config ODD_RT_SIGACTION
1267 Architecture has unusual rt_sigaction(2) arguments
1269 config OLD_SIGSUSPEND
1272 Architecture has old sigsuspend(2) syscall, of one-argument variety
1274 config OLD_SIGSUSPEND3
1277 Even weirder antique ABI - three-argument sigsuspend(2)
1279 config OLD_SIGACTION
1282 Architecture has old sigaction(2) syscall. Nope, not the same
1283 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
1284 but fairly different variant of sigaction(2), thanks to OSF/1
1287 config COMPAT_OLD_SIGACTION
1290 config COMPAT_32BIT_TIME
1291 bool "Provide system calls for 32-bit time_t"
1292 default !64BIT || COMPAT
1294 This enables 32 bit time_t support in addition to 64 bit time_t support.
1295 This is relevant on all 32-bit architectures, and 64-bit architectures
1296 as part of compat syscall handling.
1298 config ARCH_NO_PREEMPT
1301 config ARCH_SUPPORTS_RT
1304 config CPU_NO_EFFICIENT_FFS
1307 config HAVE_ARCH_VMAP_STACK
1310 An arch should select this symbol if it can support kernel stacks
1311 in vmalloc space. This means:
1313 - vmalloc space must be large enough to hold many kernel stacks.
1314 This may rule out many 32-bit architectures.
1316 - Stacks in vmalloc space need to work reliably. For example, if
1317 vmap page tables are created on demand, either this mechanism
1318 needs to work while the stack points to a virtual address with
1319 unpopulated page tables or arch code (switch_to() and switch_mm(),
1320 most likely) needs to ensure that the stack's page table entries
1321 are populated before running on a possibly unpopulated stack.
1323 - If the stack overflows into a guard page, something reasonable
1324 should happen. The definition of "reasonable" is flexible, but
1325 instantly rebooting without logging anything would be unfriendly.
1329 bool "Use a virtually-mapped stack"
1330 depends on HAVE_ARCH_VMAP_STACK
1331 depends on !KASAN || KASAN_HW_TAGS || KASAN_VMALLOC
1333 Enable this if you want the use virtually-mapped kernel stacks
1334 with guard pages. This causes kernel stack overflows to be
1335 caught immediately rather than causing difficult-to-diagnose
1338 To use this with software KASAN modes, the architecture must support
1339 backing virtual mappings with real shadow memory, and KASAN_VMALLOC
1342 config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1345 An arch should select this symbol if it can support kernel stack
1346 offset randomization with calls to add_random_kstack_offset()
1347 during syscall entry and choose_random_kstack_offset() during
1348 syscall exit. Careful removal of -fstack-protector-strong and
1349 -fstack-protector should also be applied to the entry code and
1350 closely examined, as the artificial stack bump looks like an array
1351 to the compiler, so it will attempt to add canary checks regardless
1352 of the static branch state.
1354 config RANDOMIZE_KSTACK_OFFSET
1355 bool "Support for randomizing kernel stack offset on syscall entry" if EXPERT
1357 depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
1358 depends on INIT_STACK_NONE || !CC_IS_CLANG || CLANG_VERSION >= 140000
1360 The kernel stack offset can be randomized (after pt_regs) by
1361 roughly 5 bits of entropy, frustrating memory corruption
1362 attacks that depend on stack address determinism or
1363 cross-syscall address exposures.
1365 The feature is controlled via the "randomize_kstack_offset=on/off"
1366 kernel boot param, and if turned off has zero overhead due to its use
1367 of static branches (see JUMP_LABEL).
1371 config RANDOMIZE_KSTACK_OFFSET_DEFAULT
1372 bool "Default state of kernel stack offset randomization"
1373 depends on RANDOMIZE_KSTACK_OFFSET
1375 Kernel stack offset randomization is controlled by kernel boot param
1376 "randomize_kstack_offset=on/off", and this config chooses the default
1379 config ARCH_OPTIONAL_KERNEL_RWX
1382 config ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1385 config ARCH_HAS_STRICT_KERNEL_RWX
1388 config STRICT_KERNEL_RWX
1389 bool "Make kernel text and rodata read-only" if ARCH_OPTIONAL_KERNEL_RWX
1390 depends on ARCH_HAS_STRICT_KERNEL_RWX
1391 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1393 If this is set, kernel text and rodata memory will be made read-only,
1394 and non-text memory will be made non-executable. This provides
1395 protection against certain security exploits (e.g. executing the heap
1398 These features are considered standard security practice these days.
1399 You should say Y here in almost all cases.
1401 config ARCH_HAS_STRICT_MODULE_RWX
1404 config STRICT_MODULE_RWX
1405 bool "Set loadable kernel module data as NX and text as RO" if ARCH_OPTIONAL_KERNEL_RWX
1406 depends on ARCH_HAS_STRICT_MODULE_RWX && MODULES
1407 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
1409 If this is set, module text and rodata memory will be made read-only,
1410 and non-text memory will be made non-executable. This provides
1411 protection against certain security exploits (e.g. writing to text)
1413 # select if the architecture provides an asm/dma-direct.h header
1414 config ARCH_HAS_PHYS_TO_DMA
1417 config HAVE_ARCH_COMPILER_H
1420 An architecture can select this if it provides an
1421 asm/compiler.h header that should be included after
1422 linux/compiler-*.h in order to override macro definitions that those
1423 headers generally provide.
1425 config HAVE_ARCH_PREL32_RELOCATIONS
1428 May be selected by an architecture if it supports place-relative
1429 32-bit relocations, both in the toolchain and in the module loader,
1430 in which case relative references can be used in special sections
1431 for PCI fixup, initcalls etc which are only half the size on 64 bit
1432 architectures, and don't require runtime relocation on relocatable
1435 config ARCH_USE_MEMREMAP_PROT
1438 config LOCK_EVENT_COUNTS
1439 bool "Locking event counts collection"
1442 Enable light-weight counting of various locking related events
1443 in the system with minimal performance impact. This reduces
1444 the chance of application behavior change because of timing
1445 differences. The counts are reported via debugfs.
1447 # Select if the architecture has support for applying RELR relocations.
1448 config ARCH_HAS_RELR
1452 bool "Use RELR relocation packing"
1453 depends on ARCH_HAS_RELR && TOOLS_SUPPORT_RELR
1456 Store the kernel's dynamic relocations in the RELR relocation packing
1457 format. Requires a compatible linker (LLD supports this feature), as
1458 well as compatible NM and OBJCOPY utilities (llvm-nm and llvm-objcopy
1461 config ARCH_HAS_MEM_ENCRYPT
1464 config ARCH_HAS_CC_PLATFORM
1467 config HAVE_SPARSE_SYSCALL_NR
1470 An architecture should select this if its syscall numbering is sparse
1471 to save space. For example, MIPS architecture has a syscall array with
1472 entries at 4000, 5000 and 6000 locations. This option turns on syscall
1473 related optimizations for a given architecture.
1475 config ARCH_HAS_VDSO_DATA
1478 config HAVE_STATIC_CALL
1481 config HAVE_STATIC_CALL_INLINE
1483 depends on HAVE_STATIC_CALL
1486 config HAVE_PREEMPT_DYNAMIC
1489 config HAVE_PREEMPT_DYNAMIC_CALL
1491 depends on HAVE_STATIC_CALL
1492 select HAVE_PREEMPT_DYNAMIC
1494 An architecture should select this if it can handle the preemption
1495 model being selected at boot time using static calls.
1497 Where an architecture selects HAVE_STATIC_CALL_INLINE, any call to a
1498 preemption function will be patched directly.
1500 Where an architecture does not select HAVE_STATIC_CALL_INLINE, any
1501 call to a preemption function will go through a trampoline, and the
1502 trampoline will be patched.
1504 It is strongly advised to support inline static call to avoid any
1507 config HAVE_PREEMPT_DYNAMIC_KEY
1509 depends on HAVE_ARCH_JUMP_LABEL
1510 select HAVE_PREEMPT_DYNAMIC
1512 An architecture should select this if it can handle the preemption
1513 model being selected at boot time using static keys.
1515 Each preemption function will be given an early return based on a
1516 static key. This should have slightly lower overhead than non-inline
1517 static calls, as this effectively inlines each trampoline into the
1518 start of its callee. This may avoid redundant work, and may
1519 integrate better with CFI schemes.
1521 This will have greater overhead than using inline static calls as
1522 the call to the preemption function cannot be entirely elided.
1524 config ARCH_WANT_LD_ORPHAN_WARN
1527 An arch should select this symbol once all linker sections are explicitly
1528 included, size-asserted, or discarded in the linker scripts. This is
1529 important because we never want expected sections to be placed heuristically
1530 by the linker, since the locations of such sections can change between linker
1533 config HAVE_ARCH_PFN_VALID
1536 config ARCH_SUPPORTS_DEBUG_PAGEALLOC
1539 config ARCH_SUPPORTS_PAGE_TABLE_CHECK
1542 config ARCH_SPLIT_ARG64
1545 If a 32-bit architecture requires 64-bit arguments to be split into
1546 pairs of 32-bit arguments, select this option.
1548 config ARCH_HAS_ELFCORE_COMPAT
1551 config ARCH_HAS_PARANOID_L1D_FLUSH
1554 config ARCH_HAVE_TRACE_MMIO_ACCESS
1557 config DYNAMIC_SIGFRAME
1560 # Select, if arch has a named attribute group bound to NUMA device nodes.
1561 config HAVE_ARCH_NODE_DEV_GROUP
1564 config ARCH_HAS_HW_PTE_YOUNG
1567 Architectures that select this option are capable of setting the
1568 accessed bit in PTE entries when using them as part of linear address
1569 translations. Architectures that require runtime check should select
1570 this option and override arch_has_hw_pte_young().
1572 config ARCH_HAS_NONLEAF_PMD_YOUNG
1575 Architectures that select this option are capable of setting the
1576 accessed bit in non-leaf PMD entries when using them as part of linear
1577 address translations. Page table walkers that clear the accessed bit
1578 may use this capability to reduce their search space.
1580 source "kernel/gcov/Kconfig"
1582 source "scripts/gcc-plugins/Kconfig"
1584 config FUNCTION_ALIGNMENT_4B
1587 config FUNCTION_ALIGNMENT_8B
1590 config FUNCTION_ALIGNMENT_16B
1593 config FUNCTION_ALIGNMENT_32B
1596 config FUNCTION_ALIGNMENT_64B
1599 config FUNCTION_ALIGNMENT
1601 default 64 if FUNCTION_ALIGNMENT_64B
1602 default 32 if FUNCTION_ALIGNMENT_32B
1603 default 16 if FUNCTION_ALIGNMENT_16B
1604 default 8 if FUNCTION_ALIGNMENT_8B
1605 default 4 if FUNCTION_ALIGNMENT_4B
1608 config CC_HAS_MIN_FUNCTION_ALIGNMENT
1609 # Detect availability of the GCC option -fmin-function-alignment which
1610 # guarantees minimal alignment for all functions, unlike
1611 # -falign-functions which the compiler ignores for cold functions.
1612 def_bool $(cc-option, -fmin-function-alignment=8)
1614 config CC_HAS_SANE_FUNCTION_ALIGNMENT
1615 # Set if the guaranteed alignment with -fmin-function-alignment is
1616 # available or extra care is required in the kernel. Clang provides
1617 # strict alignment always, even with -falign-functions.
1618 def_bool CC_HAS_MIN_FUNCTION_ALIGNMENT || CC_IS_CLANG