1# SPDX-License-Identifier: GPL-2.0 2# 3# General architecture dependent options 4# 5 6# 7# Note: arch/$(SRCARCH)/Kconfig needs to be included first so that it can 8# override the default values in this file. 9# 10source "arch/$(SRCARCH)/Kconfig" 11 12menu "General architecture-dependent options" 13 14config CRASH_CORE 15 bool 16 17config KEXEC_CORE 18 select CRASH_CORE 19 bool 20 21config KEXEC_ELF 22 bool 23 24config HAVE_IMA_KEXEC 25 bool 26 27config SET_FS 28 bool 29 30config HOTPLUG_SMT 31 bool 32 33config GENERIC_ENTRY 34 bool 35 36config KPROBES 37 bool "Kprobes" 38 depends on MODULES 39 depends on HAVE_KPROBES 40 select KALLSYMS 41 help 42 Kprobes allows you to trap at almost any kernel address and 43 execute a callback function. register_kprobe() establishes 44 a probepoint and specifies the callback. Kprobes is useful 45 for kernel debugging, non-intrusive instrumentation and testing. 46 If in doubt, say "N". 47 48config JUMP_LABEL 49 bool "Optimize very unlikely/likely branches" 50 depends on HAVE_ARCH_JUMP_LABEL 51 depends on CC_HAS_ASM_GOTO 52 help 53 This option enables a transparent branch optimization that 54 makes certain almost-always-true or almost-always-false branch 55 conditions even cheaper to execute within the kernel. 56 57 Certain performance-sensitive kernel code, such as trace points, 58 scheduler functionality, networking code and KVM have such 59 branches and include support for this optimization technique. 60 61 If it is detected that the compiler has support for "asm goto", 62 the kernel will compile such branches with just a nop 63 instruction. When the condition flag is toggled to true, the 64 nop will be converted to a jump instruction to execute the 65 conditional block of instructions. 66 67 This technique lowers overhead and stress on the branch prediction 68 of the processor and generally makes the kernel faster. The update 69 of the condition is slower, but those are always very rare. 70 71 ( On 32-bit x86, the necessary options added to the compiler 72 flags may increase the size of the kernel slightly. ) 73 74config STATIC_KEYS_SELFTEST 75 bool "Static key selftest" 76 depends on JUMP_LABEL 77 help 78 Boot time self-test of the branch patching code. 79 80config STATIC_CALL_SELFTEST 81 bool "Static call selftest" 82 depends on HAVE_STATIC_CALL 83 help 84 Boot time self-test of the call patching code. 85 86config OPTPROBES 87 def_bool y 88 depends on KPROBES && HAVE_OPTPROBES 89 select TASKS_RCU if PREEMPTION 90 91config KPROBES_ON_FTRACE 92 def_bool y 93 depends on KPROBES && HAVE_KPROBES_ON_FTRACE 94 depends on DYNAMIC_FTRACE_WITH_REGS 95 help 96 If function tracer is enabled and the arch supports full 97 passing of pt_regs to function tracing, then kprobes can 98 optimize on top of function tracing. 99 100config UPROBES 101 def_bool n 102 depends on ARCH_SUPPORTS_UPROBES 103 help 104 Uprobes is the user-space counterpart to kprobes: they 105 enable instrumentation applications (such as 'perf probe') 106 to establish unintrusive probes in user-space binaries and 107 libraries, by executing handler functions when the probes 108 are hit by user-space applications. 109 110 ( These probes come in the form of single-byte breakpoints, 111 managed by the kernel and kept transparent to the probed 112 application. ) 113 114config HAVE_64BIT_ALIGNED_ACCESS 115 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS 116 help 117 Some architectures require 64 bit accesses to be 64 bit 118 aligned, which also requires structs containing 64 bit values 119 to be 64 bit aligned too. This includes some 32 bit 120 architectures which can do 64 bit accesses, as well as 64 bit 121 architectures without unaligned access. 122 123 This symbol should be selected by an architecture if 64 bit 124 accesses are required to be 64 bit aligned in this way even 125 though it is not a 64 bit architecture. 126 127 See Documentation/core-api/unaligned-memory-access.rst for 128 more information on the topic of unaligned memory accesses. 129 130config HAVE_EFFICIENT_UNALIGNED_ACCESS 131 bool 132 help 133 Some architectures are unable to perform unaligned accesses 134 without the use of get_unaligned/put_unaligned. Others are 135 unable to perform such accesses efficiently (e.g. trap on 136 unaligned access and require fixing it up in the exception 137 handler.) 138 139 This symbol should be selected by an architecture if it can 140 perform unaligned accesses efficiently to allow different 141 code paths to be selected for these cases. Some network 142 drivers, for example, could opt to not fix up alignment 143 problems with received packets if doing so would not help 144 much. 145 146 See Documentation/core-api/unaligned-memory-access.rst for more 147 information on the topic of unaligned memory accesses. 148 149config ARCH_USE_BUILTIN_BSWAP 150 bool 151 help 152 Modern versions of GCC (since 4.4) have builtin functions 153 for handling byte-swapping. Using these, instead of the old 154 inline assembler that the architecture code provides in the 155 __arch_bswapXX() macros, allows the compiler to see what's 156 happening and offers more opportunity for optimisation. In 157 particular, the compiler will be able to combine the byteswap 158 with a nearby load or store and use load-and-swap or 159 store-and-swap instructions if the architecture has them. It 160 should almost *never* result in code which is worse than the 161 hand-coded assembler in <asm/swab.h>. But just in case it 162 does, the use of the builtins is optional. 163 164 Any architecture with load-and-swap or store-and-swap 165 instructions should set this. And it shouldn't hurt to set it 166 on architectures that don't have such instructions. 167 168config KRETPROBES 169 def_bool y 170 depends on KPROBES && HAVE_KRETPROBES 171 172config USER_RETURN_NOTIFIER 173 bool 174 depends on HAVE_USER_RETURN_NOTIFIER 175 help 176 Provide a kernel-internal notification when a cpu is about to 177 switch to user mode. 178 179config HAVE_IOREMAP_PROT 180 bool 181 182config HAVE_KPROBES 183 bool 184 185config HAVE_KRETPROBES 186 bool 187 188config HAVE_OPTPROBES 189 bool 190 191config HAVE_KPROBES_ON_FTRACE 192 bool 193 194config HAVE_FUNCTION_ERROR_INJECTION 195 bool 196 197config HAVE_NMI 198 bool 199 200# 201# An arch should select this if it provides all these things: 202# 203# task_pt_regs() in asm/processor.h or asm/ptrace.h 204# arch_has_single_step() if there is hardware single-step support 205# arch_has_block_step() if there is hardware block-step support 206# asm/syscall.h supplying asm-generic/syscall.h interface 207# linux/regset.h user_regset interfaces 208# CORE_DUMP_USE_REGSET #define'd in linux/elf.h 209# TIF_SYSCALL_TRACE calls tracehook_report_syscall_{entry,exit} 210# TIF_NOTIFY_RESUME calls tracehook_notify_resume() 211# signal delivery calls tracehook_signal_handler() 212# 213config HAVE_ARCH_TRACEHOOK 214 bool 215 216config HAVE_DMA_CONTIGUOUS 217 bool 218 219config GENERIC_SMP_IDLE_THREAD 220 bool 221 222config GENERIC_IDLE_POLL_SETUP 223 bool 224 225config ARCH_HAS_FORTIFY_SOURCE 226 bool 227 help 228 An architecture should select this when it can successfully 229 build and run with CONFIG_FORTIFY_SOURCE. 230 231# 232# Select if the arch provides a historic keepinit alias for the retain_initrd 233# command line option 234# 235config ARCH_HAS_KEEPINITRD 236 bool 237 238# Select if arch has all set_memory_ro/rw/x/nx() functions in asm/cacheflush.h 239config ARCH_HAS_SET_MEMORY 240 bool 241 242# Select if arch has all set_direct_map_invalid/default() functions 243config ARCH_HAS_SET_DIRECT_MAP 244 bool 245 246# 247# Select if the architecture provides the arch_dma_set_uncached symbol to 248# either provide an uncached segment alias for a DMA allocation, or 249# to remap the page tables in place. 250# 251config ARCH_HAS_DMA_SET_UNCACHED 252 bool 253 254# 255# Select if the architectures provides the arch_dma_clear_uncached symbol 256# to undo an in-place page table remap for uncached access. 257# 258config ARCH_HAS_DMA_CLEAR_UNCACHED 259 bool 260 261# Select if arch init_task must go in the __init_task_data section 262config ARCH_TASK_STRUCT_ON_STACK 263 bool 264 265# Select if arch has its private alloc_task_struct() function 266config ARCH_TASK_STRUCT_ALLOCATOR 267 bool 268 269config HAVE_ARCH_THREAD_STRUCT_WHITELIST 270 bool 271 depends on !ARCH_TASK_STRUCT_ALLOCATOR 272 help 273 An architecture should select this to provide hardened usercopy 274 knowledge about what region of the thread_struct should be 275 whitelisted for copying to userspace. Normally this is only the 276 FPU registers. Specifically, arch_thread_struct_whitelist() 277 should be implemented. Without this, the entire thread_struct 278 field in task_struct will be left whitelisted. 279 280# Select if arch has its private alloc_thread_stack() function 281config ARCH_THREAD_STACK_ALLOCATOR 282 bool 283 284# Select if arch wants to size task_struct dynamically via arch_task_struct_size: 285config ARCH_WANTS_DYNAMIC_TASK_STRUCT 286 bool 287 288config ARCH_32BIT_OFF_T 289 bool 290 depends on !64BIT 291 help 292 All new 32-bit architectures should have 64-bit off_t type on 293 userspace side which corresponds to the loff_t kernel type. This 294 is the requirement for modern ABIs. Some existing architectures 295 still support 32-bit off_t. This option is enabled for all such 296 architectures explicitly. 297 298# Selected by 64 bit architectures which have a 32 bit f_tinode in struct ustat 299config ARCH_32BIT_USTAT_F_TINODE 300 bool 301 302config HAVE_ASM_MODVERSIONS 303 bool 304 help 305 This symbol should be selected by an architecture if it provides 306 <asm/asm-prototypes.h> to support the module versioning for symbols 307 exported from assembly code. 308 309config HAVE_REGS_AND_STACK_ACCESS_API 310 bool 311 help 312 This symbol should be selected by an architecture if it supports 313 the API needed to access registers and stack entries from pt_regs, 314 declared in asm/ptrace.h 315 For example the kprobes-based event tracer needs this API. 316 317config HAVE_RSEQ 318 bool 319 depends on HAVE_REGS_AND_STACK_ACCESS_API 320 help 321 This symbol should be selected by an architecture if it 322 supports an implementation of restartable sequences. 323 324config HAVE_FUNCTION_ARG_ACCESS_API 325 bool 326 help 327 This symbol should be selected by an architecture if it supports 328 the API needed to access function arguments from pt_regs, 329 declared in asm/ptrace.h 330 331config HAVE_HW_BREAKPOINT 332 bool 333 depends on PERF_EVENTS 334 335config HAVE_MIXED_BREAKPOINTS_REGS 336 bool 337 depends on HAVE_HW_BREAKPOINT 338 help 339 Depending on the arch implementation of hardware breakpoints, 340 some of them have separate registers for data and instruction 341 breakpoints addresses, others have mixed registers to store 342 them but define the access type in a control register. 343 Select this option if your arch implements breakpoints under the 344 latter fashion. 345 346config HAVE_USER_RETURN_NOTIFIER 347 bool 348 349config HAVE_PERF_EVENTS_NMI 350 bool 351 help 352 System hardware can generate an NMI using the perf event 353 subsystem. Also has support for calculating CPU cycle events 354 to determine how many clock cycles in a given period. 355 356config HAVE_HARDLOCKUP_DETECTOR_PERF 357 bool 358 depends on HAVE_PERF_EVENTS_NMI 359 help 360 The arch chooses to use the generic perf-NMI-based hardlockup 361 detector. Must define HAVE_PERF_EVENTS_NMI. 362 363config HAVE_NMI_WATCHDOG 364 depends on HAVE_NMI 365 bool 366 help 367 The arch provides a low level NMI watchdog. It provides 368 asm/nmi.h, and defines its own arch_touch_nmi_watchdog(). 369 370config HAVE_HARDLOCKUP_DETECTOR_ARCH 371 bool 372 select HAVE_NMI_WATCHDOG 373 help 374 The arch chooses to provide its own hardlockup detector, which is 375 a superset of the HAVE_NMI_WATCHDOG. It also conforms to config 376 interfaces and parameters provided by hardlockup detector subsystem. 377 378config HAVE_PERF_REGS 379 bool 380 help 381 Support selective register dumps for perf events. This includes 382 bit-mapping of each registers and a unique architecture id. 383 384config HAVE_PERF_USER_STACK_DUMP 385 bool 386 help 387 Support user stack dumps for perf event samples. This needs 388 access to the user stack pointer which is not unified across 389 architectures. 390 391config HAVE_ARCH_JUMP_LABEL 392 bool 393 394config HAVE_ARCH_JUMP_LABEL_RELATIVE 395 bool 396 397config MMU_GATHER_TABLE_FREE 398 bool 399 400config MMU_GATHER_RCU_TABLE_FREE 401 bool 402 select MMU_GATHER_TABLE_FREE 403 404config MMU_GATHER_PAGE_SIZE 405 bool 406 407config MMU_GATHER_NO_RANGE 408 bool 409 410config MMU_GATHER_NO_GATHER 411 bool 412 depends on MMU_GATHER_TABLE_FREE 413 414config ARCH_WANT_IRQS_OFF_ACTIVATE_MM 415 bool 416 help 417 Temporary select until all architectures can be converted to have 418 irqs disabled over activate_mm. Architectures that do IPI based TLB 419 shootdowns should enable this. 420 421config ARCH_HAVE_NMI_SAFE_CMPXCHG 422 bool 423 424config HAVE_ALIGNED_STRUCT_PAGE 425 bool 426 help 427 This makes sure that struct pages are double word aligned and that 428 e.g. the SLUB allocator can perform double word atomic operations 429 on a struct page for better performance. However selecting this 430 might increase the size of a struct page by a word. 431 432config HAVE_CMPXCHG_LOCAL 433 bool 434 435config HAVE_CMPXCHG_DOUBLE 436 bool 437 438config ARCH_WEAK_RELEASE_ACQUIRE 439 bool 440 441config ARCH_WANT_IPC_PARSE_VERSION 442 bool 443 444config ARCH_WANT_COMPAT_IPC_PARSE_VERSION 445 bool 446 447config ARCH_WANT_OLD_COMPAT_IPC 448 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION 449 bool 450 451config HAVE_ARCH_SECCOMP 452 bool 453 help 454 An arch should select this symbol to support seccomp mode 1 (the fixed 455 syscall policy), and must provide an overrides for __NR_seccomp_sigreturn, 456 and compat syscalls if the asm-generic/seccomp.h defaults need adjustment: 457 - __NR_seccomp_read_32 458 - __NR_seccomp_write_32 459 - __NR_seccomp_exit_32 460 - __NR_seccomp_sigreturn_32 461 462config HAVE_ARCH_SECCOMP_FILTER 463 bool 464 select HAVE_ARCH_SECCOMP 465 help 466 An arch should select this symbol if it provides all of these things: 467 - all the requirements for HAVE_ARCH_SECCOMP 468 - syscall_get_arch() 469 - syscall_get_arguments() 470 - syscall_rollback() 471 - syscall_set_return_value() 472 - SIGSYS siginfo_t support 473 - secure_computing is called from a ptrace_event()-safe context 474 - secure_computing return value is checked and a return value of -1 475 results in the system call being skipped immediately. 476 - seccomp syscall wired up 477 - if !HAVE_SPARSE_SYSCALL_NR, have SECCOMP_ARCH_NATIVE, 478 SECCOMP_ARCH_NATIVE_NR, SECCOMP_ARCH_NATIVE_NAME defined. If 479 COMPAT is supported, have the SECCOMP_ARCH_COMPAT* defines too. 480 481config SECCOMP 482 prompt "Enable seccomp to safely execute untrusted bytecode" 483 def_bool y 484 depends on HAVE_ARCH_SECCOMP 485 help 486 This kernel feature is useful for number crunching applications 487 that may need to handle untrusted bytecode during their 488 execution. By using pipes or other transports made available 489 to the process as file descriptors supporting the read/write 490 syscalls, it's possible to isolate those applications in their 491 own address space using seccomp. Once seccomp is enabled via 492 prctl(PR_SET_SECCOMP) or the seccomp() syscall, it cannot be 493 disabled and the task is only allowed to execute a few safe 494 syscalls defined by each seccomp mode. 495 496 If unsure, say Y. 497 498config SECCOMP_FILTER 499 def_bool y 500 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET 501 help 502 Enable tasks to build secure computing environments defined 503 in terms of Berkeley Packet Filter programs which implement 504 task-defined system call filtering polices. 505 506 See Documentation/userspace-api/seccomp_filter.rst for details. 507 508config SECCOMP_CACHE_DEBUG 509 bool "Show seccomp filter cache status in /proc/pid/seccomp_cache" 510 depends on SECCOMP_FILTER && !HAVE_SPARSE_SYSCALL_NR 511 depends on PROC_FS 512 help 513 This enables the /proc/pid/seccomp_cache interface to monitor 514 seccomp cache data. The file format is subject to change. Reading 515 the file requires CAP_SYS_ADMIN. 516 517 This option is for debugging only. Enabling presents the risk that 518 an adversary may be able to infer the seccomp filter logic. 519 520 If unsure, say N. 521 522config HAVE_ARCH_STACKLEAK 523 bool 524 help 525 An architecture should select this if it has the code which 526 fills the used part of the kernel stack with the STACKLEAK_POISON 527 value before returning from system calls. 528 529config HAVE_STACKPROTECTOR 530 bool 531 help 532 An arch should select this symbol if: 533 - it has implemented a stack canary (e.g. __stack_chk_guard) 534 535config STACKPROTECTOR 536 bool "Stack Protector buffer overflow detection" 537 depends on HAVE_STACKPROTECTOR 538 depends on $(cc-option,-fstack-protector) 539 default y 540 help 541 This option turns on the "stack-protector" GCC feature. This 542 feature puts, at the beginning of functions, a canary value on 543 the stack just before the return address, and validates 544 the value just before actually returning. Stack based buffer 545 overflows (that need to overwrite this return address) now also 546 overwrite the canary, which gets detected and the attack is then 547 neutralized via a kernel panic. 548 549 Functions will have the stack-protector canary logic added if they 550 have an 8-byte or larger character array on the stack. 551 552 This feature requires gcc version 4.2 or above, or a distribution 553 gcc with the feature backported ("-fstack-protector"). 554 555 On an x86 "defconfig" build, this feature adds canary checks to 556 about 3% of all kernel functions, which increases kernel code size 557 by about 0.3%. 558 559config STACKPROTECTOR_STRONG 560 bool "Strong Stack Protector" 561 depends on STACKPROTECTOR 562 depends on $(cc-option,-fstack-protector-strong) 563 default y 564 help 565 Functions will have the stack-protector canary logic added in any 566 of the following conditions: 567 568 - local variable's address used as part of the right hand side of an 569 assignment or function argument 570 - local variable is an array (or union containing an array), 571 regardless of array type or length 572 - uses register local variables 573 574 This feature requires gcc version 4.9 or above, or a distribution 575 gcc with the feature backported ("-fstack-protector-strong"). 576 577 On an x86 "defconfig" build, this feature adds canary checks to 578 about 20% of all kernel functions, which increases the kernel code 579 size by about 2%. 580 581config ARCH_SUPPORTS_SHADOW_CALL_STACK 582 bool 583 help 584 An architecture should select this if it supports Clang's Shadow 585 Call Stack and implements runtime support for shadow stack 586 switching. 587 588config SHADOW_CALL_STACK 589 bool "Clang Shadow Call Stack" 590 depends on CC_IS_CLANG && ARCH_SUPPORTS_SHADOW_CALL_STACK 591 depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER 592 help 593 This option enables Clang's Shadow Call Stack, which uses a 594 shadow stack to protect function return addresses from being 595 overwritten by an attacker. More information can be found in 596 Clang's documentation: 597 598 https://clang.llvm.org/docs/ShadowCallStack.html 599 600 Note that security guarantees in the kernel differ from the 601 ones documented for user space. The kernel must store addresses 602 of shadow stacks in memory, which means an attacker capable of 603 reading and writing arbitrary memory may be able to locate them 604 and hijack control flow by modifying the stacks. 605 606config LTO 607 bool 608 help 609 Selected if the kernel will be built using the compiler's LTO feature. 610 611config LTO_CLANG 612 bool 613 select LTO 614 help 615 Selected if the kernel will be built using Clang's LTO feature. 616 617config ARCH_SUPPORTS_LTO_CLANG 618 bool 619 help 620 An architecture should select this option if it supports: 621 - compiling with Clang, 622 - compiling inline assembly with Clang's integrated assembler, 623 - and linking with LLD. 624 625config ARCH_SUPPORTS_LTO_CLANG_THIN 626 bool 627 help 628 An architecture should select this option if it can support Clang's 629 ThinLTO mode. 630 631config HAS_LTO_CLANG 632 def_bool y 633 # Clang >= 11: https://github.com/ClangBuiltLinux/linux/issues/510 634 depends on CC_IS_CLANG && CLANG_VERSION >= 110000 && LD_IS_LLD && AS_IS_LLVM 635 depends on $(success,$(NM) --help | head -n 1 | grep -qi llvm) 636 depends on $(success,$(AR) --help | head -n 1 | grep -qi llvm) 637 depends on ARCH_SUPPORTS_LTO_CLANG 638 depends on !FTRACE_MCOUNT_USE_RECORDMCOUNT 639 depends on !KASAN || KASAN_HW_TAGS 640 depends on !GCOV_KERNEL 641 help 642 The compiler and Kconfig options support building with Clang's 643 LTO. 644 645choice 646 prompt "Link Time Optimization (LTO)" 647 default LTO_NONE 648 help 649 This option enables Link Time Optimization (LTO), which allows the 650 compiler to optimize binaries globally. 651 652 If unsure, select LTO_NONE. Note that LTO is very resource-intensive 653 so it's disabled by default. 654 655config LTO_NONE 656 bool "None" 657 help 658 Build the kernel normally, without Link Time Optimization (LTO). 659 660config LTO_CLANG_FULL 661 bool "Clang Full LTO (EXPERIMENTAL)" 662 depends on HAS_LTO_CLANG 663 depends on !COMPILE_TEST 664 select LTO_CLANG 665 help 666 This option enables Clang's full Link Time Optimization (LTO), which 667 allows the compiler to optimize the kernel globally. If you enable 668 this option, the compiler generates LLVM bitcode instead of ELF 669 object files, and the actual compilation from bitcode happens at 670 the LTO link step, which may take several minutes depending on the 671 kernel configuration. More information can be found from LLVM's 672 documentation: 673 674 https://llvm.org/docs/LinkTimeOptimization.html 675 676 During link time, this option can use a large amount of RAM, and 677 may take much longer than the ThinLTO option. 678 679config LTO_CLANG_THIN 680 bool "Clang ThinLTO (EXPERIMENTAL)" 681 depends on HAS_LTO_CLANG && ARCH_SUPPORTS_LTO_CLANG_THIN 682 select LTO_CLANG 683 help 684 This option enables Clang's ThinLTO, which allows for parallel 685 optimization and faster incremental compiles compared to the 686 CONFIG_LTO_CLANG_FULL option. More information can be found 687 from Clang's documentation: 688 689 https://clang.llvm.org/docs/ThinLTO.html 690 691 If unsure, say Y. 692endchoice 693 694config ARCH_SUPPORTS_CFI_CLANG 695 bool 696 help 697 An architecture should select this option if it can support Clang's 698 Control-Flow Integrity (CFI) checking. 699 700config CFI_CLANG 701 bool "Use Clang's Control Flow Integrity (CFI)" 702 depends on LTO_CLANG && ARCH_SUPPORTS_CFI_CLANG 703 # Clang >= 12: 704 # - https://bugs.llvm.org/show_bug.cgi?id=46258 705 # - https://bugs.llvm.org/show_bug.cgi?id=47479 706 depends on CLANG_VERSION >= 120000 707 select KALLSYMS 708 help 709 This option enables Clang’s forward-edge Control Flow Integrity 710 (CFI) checking, where the compiler injects a runtime check to each 711 indirect function call to ensure the target is a valid function with 712 the correct static type. This restricts possible call targets and 713 makes it more difficult for an attacker to exploit bugs that allow 714 the modification of stored function pointers. More information can be 715 found from Clang's documentation: 716 717 https://clang.llvm.org/docs/ControlFlowIntegrity.html 718 719config CFI_CLANG_SHADOW 720 bool "Use CFI shadow to speed up cross-module checks" 721 default y 722 depends on CFI_CLANG && MODULES 723 help 724 If you select this option, the kernel builds a fast look-up table of 725 CFI check functions in loaded modules to reduce performance overhead. 726 727 If unsure, say Y. 728 729config CFI_PERMISSIVE 730 bool "Use CFI in permissive mode" 731 depends on CFI_CLANG 732 help 733 When selected, Control Flow Integrity (CFI) violations result in a 734 warning instead of a kernel panic. This option should only be used 735 for finding indirect call type mismatches during development. 736 737 If unsure, say N. 738 739config HAVE_ARCH_WITHIN_STACK_FRAMES 740 bool 741 help 742 An architecture should select this if it can walk the kernel stack 743 frames to determine if an object is part of either the arguments 744 or local variables (i.e. that it excludes saved return addresses, 745 and similar) by implementing an inline arch_within_stack_frames(), 746 which is used by CONFIG_HARDENED_USERCOPY. 747 748config HAVE_CONTEXT_TRACKING 749 bool 750 help 751 Provide kernel/user boundaries probes necessary for subsystems 752 that need it, such as userspace RCU extended quiescent state. 753 Syscalls need to be wrapped inside user_exit()-user_enter(), either 754 optimized behind static key or through the slow path using TIF_NOHZ 755 flag. Exceptions handlers must be wrapped as well. Irqs are already 756 protected inside rcu_irq_enter/rcu_irq_exit() but preemption or signal 757 handling on irq exit still need to be protected. 758 759config HAVE_CONTEXT_TRACKING_OFFSTACK 760 bool 761 help 762 Architecture neither relies on exception_enter()/exception_exit() 763 nor on schedule_user(). Also preempt_schedule_notrace() and 764 preempt_schedule_irq() can't be called in a preemptible section 765 while context tracking is CONTEXT_USER. This feature reflects a sane 766 entry implementation where the following requirements are met on 767 critical entry code, ie: before user_exit() or after user_enter(): 768 769 - Critical entry code isn't preemptible (or better yet: 770 not interruptible). 771 - No use of RCU read side critical sections, unless rcu_nmi_enter() 772 got called. 773 - No use of instrumentation, unless instrumentation_begin() got 774 called. 775 776config HAVE_TIF_NOHZ 777 bool 778 help 779 Arch relies on TIF_NOHZ and syscall slow path to implement context 780 tracking calls to user_enter()/user_exit(). 781 782config HAVE_VIRT_CPU_ACCOUNTING 783 bool 784 785config HAVE_VIRT_CPU_ACCOUNTING_IDLE 786 bool 787 help 788 Architecture has its own way to account idle CPU time and therefore 789 doesn't implement vtime_account_idle(). 790 791config ARCH_HAS_SCALED_CPUTIME 792 bool 793 794config HAVE_VIRT_CPU_ACCOUNTING_GEN 795 bool 796 default y if 64BIT 797 help 798 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit. 799 Before enabling this option, arch code must be audited 800 to ensure there are no races in concurrent read/write of 801 cputime_t. For example, reading/writing 64-bit cputime_t on 802 some 32-bit arches may require multiple accesses, so proper 803 locking is needed to protect against concurrent accesses. 804 805config HAVE_IRQ_TIME_ACCOUNTING 806 bool 807 help 808 Archs need to ensure they use a high enough resolution clock to 809 support irq time accounting and then call enable_sched_clock_irqtime(). 810 811config HAVE_MOVE_PUD 812 bool 813 help 814 Architectures that select this are able to move page tables at the 815 PUD level. If there are only 3 page table levels, the move effectively 816 happens at the PGD level. 817 818config HAVE_MOVE_PMD 819 bool 820 help 821 Archs that select this are able to move page tables at the PMD level. 822 823config HAVE_ARCH_TRANSPARENT_HUGEPAGE 824 bool 825 826config HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD 827 bool 828 829config HAVE_ARCH_HUGE_VMAP 830 bool 831 832# 833# Archs that select this would be capable of PMD-sized vmaps (i.e., 834# arch_vmap_pmd_supported() returns true), and they must make no assumptions 835# that vmalloc memory is mapped with PAGE_SIZE ptes. The VM_NO_HUGE_VMAP flag 836# can be used to prohibit arch-specific allocations from using hugepages to 837# help with this (e.g., modules may require it). 838# 839config HAVE_ARCH_HUGE_VMALLOC 840 depends on HAVE_ARCH_HUGE_VMAP 841 bool 842 843config ARCH_WANT_HUGE_PMD_SHARE 844 bool 845 846config HAVE_ARCH_SOFT_DIRTY 847 bool 848 849config HAVE_MOD_ARCH_SPECIFIC 850 bool 851 help 852 The arch uses struct mod_arch_specific to store data. Many arches 853 just need a simple module loader without arch specific data - those 854 should not enable this. 855 856config MODULES_USE_ELF_RELA 857 bool 858 help 859 Modules only use ELF RELA relocations. Modules with ELF REL 860 relocations will give an error. 861 862config MODULES_USE_ELF_REL 863 bool 864 help 865 Modules only use ELF REL relocations. Modules with ELF RELA 866 relocations will give an error. 867 868config HAVE_IRQ_EXIT_ON_IRQ_STACK 869 bool 870 help 871 Architecture doesn't only execute the irq handler on the irq stack 872 but also irq_exit(). This way we can process softirqs on this irq 873 stack instead of switching to a new one when we call __do_softirq() 874 in the end of an hardirq. 875 This spares a stack switch and improves cache usage on softirq 876 processing. 877 878config HAVE_SOFTIRQ_ON_OWN_STACK 879 bool 880 help 881 Architecture provides a function to run __do_softirq() on a 882 seperate stack. 883 884config PGTABLE_LEVELS 885 int 886 default 2 887 888config ARCH_HAS_ELF_RANDOMIZE 889 bool 890 help 891 An architecture supports choosing randomized locations for 892 stack, mmap, brk, and ET_DYN. Defined functions: 893 - arch_mmap_rnd() 894 - arch_randomize_brk() 895 896config HAVE_ARCH_MMAP_RND_BITS 897 bool 898 help 899 An arch should select this symbol if it supports setting a variable 900 number of bits for use in establishing the base address for mmap 901 allocations, has MMU enabled and provides values for both: 902 - ARCH_MMAP_RND_BITS_MIN 903 - ARCH_MMAP_RND_BITS_MAX 904 905config HAVE_EXIT_THREAD 906 bool 907 help 908 An architecture implements exit_thread. 909 910config ARCH_MMAP_RND_BITS_MIN 911 int 912 913config ARCH_MMAP_RND_BITS_MAX 914 int 915 916config ARCH_MMAP_RND_BITS_DEFAULT 917 int 918 919config ARCH_MMAP_RND_BITS 920 int "Number of bits to use for ASLR of mmap base address" if EXPERT 921 range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX 922 default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT 923 default ARCH_MMAP_RND_BITS_MIN 924 depends on HAVE_ARCH_MMAP_RND_BITS 925 help 926 This value can be used to select the number of bits to use to 927 determine the random offset to the base address of vma regions 928 resulting from mmap allocations. This value will be bounded 929 by the architecture's minimum and maximum supported values. 930 931 This value can be changed after boot using the 932 /proc/sys/vm/mmap_rnd_bits tunable 933 934config HAVE_ARCH_MMAP_RND_COMPAT_BITS 935 bool 936 help 937 An arch should select this symbol if it supports running applications 938 in compatibility mode, supports setting a variable number of bits for 939 use in establishing the base address for mmap allocations, has MMU 940 enabled and provides values for both: 941 - ARCH_MMAP_RND_COMPAT_BITS_MIN 942 - ARCH_MMAP_RND_COMPAT_BITS_MAX 943 944config ARCH_MMAP_RND_COMPAT_BITS_MIN 945 int 946 947config ARCH_MMAP_RND_COMPAT_BITS_MAX 948 int 949 950config ARCH_MMAP_RND_COMPAT_BITS_DEFAULT 951 int 952 953config ARCH_MMAP_RND_COMPAT_BITS 954 int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT 955 range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX 956 default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT 957 default ARCH_MMAP_RND_COMPAT_BITS_MIN 958 depends on HAVE_ARCH_MMAP_RND_COMPAT_BITS 959 help 960 This value can be used to select the number of bits to use to 961 determine the random offset to the base address of vma regions 962 resulting from mmap allocations for compatible applications This 963 value will be bounded by the architecture's minimum and maximum 964 supported values. 965 966 This value can be changed after boot using the 967 /proc/sys/vm/mmap_rnd_compat_bits tunable 968 969config HAVE_ARCH_COMPAT_MMAP_BASES 970 bool 971 help 972 This allows 64bit applications to invoke 32-bit mmap() syscall 973 and vice-versa 32-bit applications to call 64-bit mmap(). 974 Required for applications doing different bitness syscalls. 975 976# This allows to use a set of generic functions to determine mmap base 977# address by giving priority to top-down scheme only if the process 978# is not in legacy mode (compat task, unlimited stack size or 979# sysctl_legacy_va_layout). 980# Architecture that selects this option can provide its own version of: 981# - STACK_RND_MASK 982config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT 983 bool 984 depends on MMU 985 select ARCH_HAS_ELF_RANDOMIZE 986 987config HAVE_STACK_VALIDATION 988 bool 989 help 990 Architecture supports the 'objtool check' host tool command, which 991 performs compile-time stack metadata validation. 992 993config HAVE_RELIABLE_STACKTRACE 994 bool 995 help 996 Architecture has either save_stack_trace_tsk_reliable() or 997 arch_stack_walk_reliable() function which only returns a stack trace 998 if it can guarantee the trace is reliable. 999 1000config HAVE_ARCH_HASH 1001 bool 1002 default n 1003 help 1004 If this is set, the architecture provides an <asm/hash.h> 1005 file which provides platform-specific implementations of some 1006 functions in <linux/hash.h> or fs/namei.c. 1007 1008config HAVE_ARCH_NVRAM_OPS 1009 bool 1010 1011config ISA_BUS_API 1012 def_bool ISA 1013 1014# 1015# ABI hall of shame 1016# 1017config CLONE_BACKWARDS 1018 bool 1019 help 1020 Architecture has tls passed as the 4th argument of clone(2), 1021 not the 5th one. 1022 1023config CLONE_BACKWARDS2 1024 bool 1025 help 1026 Architecture has the first two arguments of clone(2) swapped. 1027 1028config CLONE_BACKWARDS3 1029 bool 1030 help 1031 Architecture has tls passed as the 3rd argument of clone(2), 1032 not the 5th one. 1033 1034config ODD_RT_SIGACTION 1035 bool 1036 help 1037 Architecture has unusual rt_sigaction(2) arguments 1038 1039config OLD_SIGSUSPEND 1040 bool 1041 help 1042 Architecture has old sigsuspend(2) syscall, of one-argument variety 1043 1044config OLD_SIGSUSPEND3 1045 bool 1046 help 1047 Even weirder antique ABI - three-argument sigsuspend(2) 1048 1049config OLD_SIGACTION 1050 bool 1051 help 1052 Architecture has old sigaction(2) syscall. Nope, not the same 1053 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2), 1054 but fairly different variant of sigaction(2), thanks to OSF/1 1055 compatibility... 1056 1057config COMPAT_OLD_SIGACTION 1058 bool 1059 1060config COMPAT_32BIT_TIME 1061 bool "Provide system calls for 32-bit time_t" 1062 default !64BIT || COMPAT 1063 help 1064 This enables 32 bit time_t support in addition to 64 bit time_t support. 1065 This is relevant on all 32-bit architectures, and 64-bit architectures 1066 as part of compat syscall handling. 1067 1068config ARCH_NO_PREEMPT 1069 bool 1070 1071config ARCH_EPHEMERAL_INODES 1072 def_bool n 1073 help 1074 An arch should select this symbol if it doesn't keep track of inode 1075 instances on its own, but instead relies on something else (e.g. the 1076 host kernel for an UML kernel). 1077 1078config ARCH_SUPPORTS_RT 1079 bool 1080 1081config CPU_NO_EFFICIENT_FFS 1082 def_bool n 1083 1084config HAVE_ARCH_VMAP_STACK 1085 def_bool n 1086 help 1087 An arch should select this symbol if it can support kernel stacks 1088 in vmalloc space. This means: 1089 1090 - vmalloc space must be large enough to hold many kernel stacks. 1091 This may rule out many 32-bit architectures. 1092 1093 - Stacks in vmalloc space need to work reliably. For example, if 1094 vmap page tables are created on demand, either this mechanism 1095 needs to work while the stack points to a virtual address with 1096 unpopulated page tables or arch code (switch_to() and switch_mm(), 1097 most likely) needs to ensure that the stack's page table entries 1098 are populated before running on a possibly unpopulated stack. 1099 1100 - If the stack overflows into a guard page, something reasonable 1101 should happen. The definition of "reasonable" is flexible, but 1102 instantly rebooting without logging anything would be unfriendly. 1103 1104config VMAP_STACK 1105 default y 1106 bool "Use a virtually-mapped stack" 1107 depends on HAVE_ARCH_VMAP_STACK 1108 depends on !KASAN || KASAN_HW_TAGS || KASAN_VMALLOC 1109 help 1110 Enable this if you want the use virtually-mapped kernel stacks 1111 with guard pages. This causes kernel stack overflows to be 1112 caught immediately rather than causing difficult-to-diagnose 1113 corruption. 1114 1115 To use this with software KASAN modes, the architecture must support 1116 backing virtual mappings with real shadow memory, and KASAN_VMALLOC 1117 must be enabled. 1118 1119config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET 1120 def_bool n 1121 help 1122 An arch should select this symbol if it can support kernel stack 1123 offset randomization with calls to add_random_kstack_offset() 1124 during syscall entry and choose_random_kstack_offset() during 1125 syscall exit. Careful removal of -fstack-protector-strong and 1126 -fstack-protector should also be applied to the entry code and 1127 closely examined, as the artificial stack bump looks like an array 1128 to the compiler, so it will attempt to add canary checks regardless 1129 of the static branch state. 1130 1131config RANDOMIZE_KSTACK_OFFSET_DEFAULT 1132 bool "Randomize kernel stack offset on syscall entry" 1133 depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET 1134 help 1135 The kernel stack offset can be randomized (after pt_regs) by 1136 roughly 5 bits of entropy, frustrating memory corruption 1137 attacks that depend on stack address determinism or 1138 cross-syscall address exposures. This feature is controlled 1139 by kernel boot param "randomize_kstack_offset=on/off", and this 1140 config chooses the default boot state. 1141 1142config ARCH_OPTIONAL_KERNEL_RWX 1143 def_bool n 1144 1145config ARCH_OPTIONAL_KERNEL_RWX_DEFAULT 1146 def_bool n 1147 1148config ARCH_HAS_STRICT_KERNEL_RWX 1149 def_bool n 1150 1151config STRICT_KERNEL_RWX 1152 bool "Make kernel text and rodata read-only" if ARCH_OPTIONAL_KERNEL_RWX 1153 depends on ARCH_HAS_STRICT_KERNEL_RWX 1154 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT 1155 help 1156 If this is set, kernel text and rodata memory will be made read-only, 1157 and non-text memory will be made non-executable. This provides 1158 protection against certain security exploits (e.g. executing the heap 1159 or modifying text) 1160 1161 These features are considered standard security practice these days. 1162 You should say Y here in almost all cases. 1163 1164config ARCH_HAS_STRICT_MODULE_RWX 1165 def_bool n 1166 1167config STRICT_MODULE_RWX 1168 bool "Set loadable kernel module data as NX and text as RO" if ARCH_OPTIONAL_KERNEL_RWX 1169 depends on ARCH_HAS_STRICT_MODULE_RWX && MODULES 1170 default !ARCH_OPTIONAL_KERNEL_RWX || ARCH_OPTIONAL_KERNEL_RWX_DEFAULT 1171 help 1172 If this is set, module text and rodata memory will be made read-only, 1173 and non-text memory will be made non-executable. This provides 1174 protection against certain security exploits (e.g. writing to text) 1175 1176# select if the architecture provides an asm/dma-direct.h header 1177config ARCH_HAS_PHYS_TO_DMA 1178 bool 1179 1180config HAVE_ARCH_COMPILER_H 1181 bool 1182 help 1183 An architecture can select this if it provides an 1184 asm/compiler.h header that should be included after 1185 linux/compiler-*.h in order to override macro definitions that those 1186 headers generally provide. 1187 1188config HAVE_ARCH_PREL32_RELOCATIONS 1189 bool 1190 help 1191 May be selected by an architecture if it supports place-relative 1192 32-bit relocations, both in the toolchain and in the module loader, 1193 in which case relative references can be used in special sections 1194 for PCI fixup, initcalls etc which are only half the size on 64 bit 1195 architectures, and don't require runtime relocation on relocatable 1196 kernels. 1197 1198config ARCH_USE_MEMREMAP_PROT 1199 bool 1200 1201config LOCK_EVENT_COUNTS 1202 bool "Locking event counts collection" 1203 depends on DEBUG_FS 1204 help 1205 Enable light-weight counting of various locking related events 1206 in the system with minimal performance impact. This reduces 1207 the chance of application behavior change because of timing 1208 differences. The counts are reported via debugfs. 1209 1210# Select if the architecture has support for applying RELR relocations. 1211config ARCH_HAS_RELR 1212 bool 1213 1214config RELR 1215 bool "Use RELR relocation packing" 1216 depends on ARCH_HAS_RELR && TOOLS_SUPPORT_RELR 1217 default y 1218 help 1219 Store the kernel's dynamic relocations in the RELR relocation packing 1220 format. Requires a compatible linker (LLD supports this feature), as 1221 well as compatible NM and OBJCOPY utilities (llvm-nm and llvm-objcopy 1222 are compatible). 1223 1224config ARCH_HAS_MEM_ENCRYPT 1225 bool 1226 1227config HAVE_SPARSE_SYSCALL_NR 1228 bool 1229 help 1230 An architecture should select this if its syscall numbering is sparse 1231 to save space. For example, MIPS architecture has a syscall array with 1232 entries at 4000, 5000 and 6000 locations. This option turns on syscall 1233 related optimizations for a given architecture. 1234 1235config ARCH_HAS_VDSO_DATA 1236 bool 1237 1238config HAVE_STATIC_CALL 1239 bool 1240 1241config HAVE_STATIC_CALL_INLINE 1242 bool 1243 depends on HAVE_STATIC_CALL 1244 1245config HAVE_PREEMPT_DYNAMIC 1246 bool 1247 depends on HAVE_STATIC_CALL 1248 depends on GENERIC_ENTRY 1249 help 1250 Select this if the architecture support boot time preempt setting 1251 on top of static calls. It is strongly advised to support inline 1252 static call to avoid any overhead. 1253 1254config ARCH_WANT_LD_ORPHAN_WARN 1255 bool 1256 help 1257 An arch should select this symbol once all linker sections are explicitly 1258 included, size-asserted, or discarded in the linker scripts. This is 1259 important because we never want expected sections to be placed heuristically 1260 by the linker, since the locations of such sections can change between linker 1261 versions. 1262 1263config HAVE_ARCH_PFN_VALID 1264 bool 1265 1266config ARCH_SUPPORTS_DEBUG_PAGEALLOC 1267 bool 1268 1269config ARCH_SPLIT_ARG64 1270 bool 1271 help 1272 If a 32-bit architecture requires 64-bit arguments to be split into 1273 pairs of 32-bit arguments, select this option. 1274 1275config ARCH_HAS_ELFCORE_COMPAT 1276 bool 1277 1278source "kernel/gcov/Kconfig" 1279 1280source "scripts/gcc-plugins/Kconfig" 1281 1282endmenu 1283