1fb32e03fSMathieu Desnoyers# 2fb32e03fSMathieu Desnoyers# General architecture dependent options 3fb32e03fSMathieu Desnoyers# 4125e5645SMathieu Desnoyers 52965faa5SDave Youngconfig KEXEC_CORE 62965faa5SDave Young bool 72965faa5SDave Young 8125e5645SMathieu Desnoyersconfig OPROFILE 9b309a294SRobert Richter tristate "OProfile system profiling" 10125e5645SMathieu Desnoyers depends on PROFILING 11125e5645SMathieu Desnoyers depends on HAVE_OPROFILE 12d69d59f4SIngo Molnar select RING_BUFFER 139a5963ebSChristian Borntraeger select RING_BUFFER_ALLOW_SWAP 14125e5645SMathieu Desnoyers help 15125e5645SMathieu Desnoyers OProfile is a profiling system capable of profiling the 16125e5645SMathieu Desnoyers whole system, include the kernel, kernel modules, libraries, 17125e5645SMathieu Desnoyers and applications. 18125e5645SMathieu Desnoyers 19125e5645SMathieu Desnoyers If unsure, say N. 20125e5645SMathieu Desnoyers 214d4036e0SJason Yehconfig OPROFILE_EVENT_MULTIPLEX 224d4036e0SJason Yeh bool "OProfile multiplexing support (EXPERIMENTAL)" 234d4036e0SJason Yeh default n 244d4036e0SJason Yeh depends on OPROFILE && X86 254d4036e0SJason Yeh help 264d4036e0SJason Yeh The number of hardware counters is limited. The multiplexing 274d4036e0SJason Yeh feature enables OProfile to gather more events than counters 284d4036e0SJason Yeh are provided by the hardware. This is realized by switching 294d4036e0SJason Yeh between events at an user specified time interval. 304d4036e0SJason Yeh 314d4036e0SJason Yeh If unsure, say N. 324d4036e0SJason Yeh 33125e5645SMathieu Desnoyersconfig HAVE_OPROFILE 349ba16087SJan Beulich bool 35125e5645SMathieu Desnoyers 36dcfce4a0SRobert Richterconfig OPROFILE_NMI_TIMER 37dcfce4a0SRobert Richter def_bool y 38af9feebeSAnton Blanchard depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI && !PPC64 39dcfce4a0SRobert Richter 40125e5645SMathieu Desnoyersconfig KPROBES 41125e5645SMathieu Desnoyers bool "Kprobes" 4205ed160eSMasami Hiramatsu depends on MODULES 43125e5645SMathieu Desnoyers depends on HAVE_KPROBES 4405ed160eSMasami Hiramatsu select KALLSYMS 45125e5645SMathieu Desnoyers help 46125e5645SMathieu Desnoyers Kprobes allows you to trap at almost any kernel address and 47125e5645SMathieu Desnoyers execute a callback function. register_kprobe() establishes 48125e5645SMathieu Desnoyers a probepoint and specifies the callback. Kprobes is useful 49125e5645SMathieu Desnoyers for kernel debugging, non-intrusive instrumentation and testing. 50125e5645SMathieu Desnoyers If in doubt, say "N". 51125e5645SMathieu Desnoyers 5245f81b1cSSteven Rostedtconfig JUMP_LABEL 53c5905afbSIngo Molnar bool "Optimize very unlikely/likely branches" 5445f81b1cSSteven Rostedt depends on HAVE_ARCH_JUMP_LABEL 5545f81b1cSSteven Rostedt help 56c5905afbSIngo Molnar This option enables a transparent branch optimization that 57c5905afbSIngo Molnar makes certain almost-always-true or almost-always-false branch 58c5905afbSIngo Molnar conditions even cheaper to execute within the kernel. 5945f81b1cSSteven Rostedt 60c5905afbSIngo Molnar Certain performance-sensitive kernel code, such as trace points, 61c5905afbSIngo Molnar scheduler functionality, networking code and KVM have such 62c5905afbSIngo Molnar branches and include support for this optimization technique. 63c5905afbSIngo Molnar 64c5905afbSIngo Molnar If it is detected that the compiler has support for "asm goto", 65c5905afbSIngo Molnar the kernel will compile such branches with just a nop 66c5905afbSIngo Molnar instruction. When the condition flag is toggled to true, the 67c5905afbSIngo Molnar nop will be converted to a jump instruction to execute the 68c5905afbSIngo Molnar conditional block of instructions. 69c5905afbSIngo Molnar 70c5905afbSIngo Molnar This technique lowers overhead and stress on the branch prediction 71c5905afbSIngo Molnar of the processor and generally makes the kernel faster. The update 72c5905afbSIngo Molnar of the condition is slower, but those are always very rare. 73c5905afbSIngo Molnar 74c5905afbSIngo Molnar ( On 32-bit x86, the necessary options added to the compiler 75c5905afbSIngo Molnar flags may increase the size of the kernel slightly. ) 7645f81b1cSSteven Rostedt 771987c947SPeter Zijlstraconfig STATIC_KEYS_SELFTEST 781987c947SPeter Zijlstra bool "Static key selftest" 791987c947SPeter Zijlstra depends on JUMP_LABEL 801987c947SPeter Zijlstra help 811987c947SPeter Zijlstra Boot time self-test of the branch patching code. 821987c947SPeter Zijlstra 83afd66255SMasami Hiramatsuconfig OPTPROBES 845cc718b9SMasami Hiramatsu def_bool y 855cc718b9SMasami Hiramatsu depends on KPROBES && HAVE_OPTPROBES 86afd66255SMasami Hiramatsu depends on !PREEMPT 87afd66255SMasami Hiramatsu 88e7dbfe34SMasami Hiramatsuconfig KPROBES_ON_FTRACE 89e7dbfe34SMasami Hiramatsu def_bool y 90e7dbfe34SMasami Hiramatsu depends on KPROBES && HAVE_KPROBES_ON_FTRACE 91e7dbfe34SMasami Hiramatsu depends on DYNAMIC_FTRACE_WITH_REGS 92e7dbfe34SMasami Hiramatsu help 93e7dbfe34SMasami Hiramatsu If function tracer is enabled and the arch supports full 94e7dbfe34SMasami Hiramatsu passing of pt_regs to function tracing, then kprobes can 95e7dbfe34SMasami Hiramatsu optimize on top of function tracing. 96e7dbfe34SMasami Hiramatsu 972b144498SSrikar Dronamrajuconfig UPROBES 9809294e31SDavid A. Long def_bool n 992b144498SSrikar Dronamraju help 1007b2d81d4SIngo Molnar Uprobes is the user-space counterpart to kprobes: they 1017b2d81d4SIngo Molnar enable instrumentation applications (such as 'perf probe') 1027b2d81d4SIngo Molnar to establish unintrusive probes in user-space binaries and 1037b2d81d4SIngo Molnar libraries, by executing handler functions when the probes 1047b2d81d4SIngo Molnar are hit by user-space applications. 1057b2d81d4SIngo Molnar 1067b2d81d4SIngo Molnar ( These probes come in the form of single-byte breakpoints, 1077b2d81d4SIngo Molnar managed by the kernel and kept transparent to the probed 1087b2d81d4SIngo Molnar application. ) 1092b144498SSrikar Dronamraju 110c19fa94aSJames Hoganconfig HAVE_64BIT_ALIGNED_ACCESS 111c19fa94aSJames Hogan def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS 112c19fa94aSJames Hogan help 113c19fa94aSJames Hogan Some architectures require 64 bit accesses to be 64 bit 114c19fa94aSJames Hogan aligned, which also requires structs containing 64 bit values 115c19fa94aSJames Hogan to be 64 bit aligned too. This includes some 32 bit 116c19fa94aSJames Hogan architectures which can do 64 bit accesses, as well as 64 bit 117c19fa94aSJames Hogan architectures without unaligned access. 118c19fa94aSJames Hogan 119c19fa94aSJames Hogan This symbol should be selected by an architecture if 64 bit 120c19fa94aSJames Hogan accesses are required to be 64 bit aligned in this way even 121c19fa94aSJames Hogan though it is not a 64 bit architecture. 122c19fa94aSJames Hogan 123c19fa94aSJames Hogan See Documentation/unaligned-memory-access.txt for more 124c19fa94aSJames Hogan information on the topic of unaligned memory accesses. 125c19fa94aSJames Hogan 12658340a07SJohannes Bergconfig HAVE_EFFICIENT_UNALIGNED_ACCESS 1279ba16087SJan Beulich bool 12858340a07SJohannes Berg help 12958340a07SJohannes Berg Some architectures are unable to perform unaligned accesses 13058340a07SJohannes Berg without the use of get_unaligned/put_unaligned. Others are 13158340a07SJohannes Berg unable to perform such accesses efficiently (e.g. trap on 13258340a07SJohannes Berg unaligned access and require fixing it up in the exception 13358340a07SJohannes Berg handler.) 13458340a07SJohannes Berg 13558340a07SJohannes Berg This symbol should be selected by an architecture if it can 13658340a07SJohannes Berg perform unaligned accesses efficiently to allow different 13758340a07SJohannes Berg code paths to be selected for these cases. Some network 13858340a07SJohannes Berg drivers, for example, could opt to not fix up alignment 13958340a07SJohannes Berg problems with received packets if doing so would not help 14058340a07SJohannes Berg much. 14158340a07SJohannes Berg 14258340a07SJohannes Berg See Documentation/unaligned-memory-access.txt for more 14358340a07SJohannes Berg information on the topic of unaligned memory accesses. 14458340a07SJohannes Berg 145cf66bb93SDavid Woodhouseconfig ARCH_USE_BUILTIN_BSWAP 146cf66bb93SDavid Woodhouse bool 147cf66bb93SDavid Woodhouse help 148cf66bb93SDavid Woodhouse Modern versions of GCC (since 4.4) have builtin functions 149cf66bb93SDavid Woodhouse for handling byte-swapping. Using these, instead of the old 150cf66bb93SDavid Woodhouse inline assembler that the architecture code provides in the 151cf66bb93SDavid Woodhouse __arch_bswapXX() macros, allows the compiler to see what's 152cf66bb93SDavid Woodhouse happening and offers more opportunity for optimisation. In 153cf66bb93SDavid Woodhouse particular, the compiler will be able to combine the byteswap 154cf66bb93SDavid Woodhouse with a nearby load or store and use load-and-swap or 155cf66bb93SDavid Woodhouse store-and-swap instructions if the architecture has them. It 156cf66bb93SDavid Woodhouse should almost *never* result in code which is worse than the 157cf66bb93SDavid Woodhouse hand-coded assembler in <asm/swab.h>. But just in case it 158cf66bb93SDavid Woodhouse does, the use of the builtins is optional. 159cf66bb93SDavid Woodhouse 160cf66bb93SDavid Woodhouse Any architecture with load-and-swap or store-and-swap 161cf66bb93SDavid Woodhouse instructions should set this. And it shouldn't hurt to set it 162cf66bb93SDavid Woodhouse on architectures that don't have such instructions. 163cf66bb93SDavid Woodhouse 1649edddaa2SAnanth N Mavinakayanahalliconfig KRETPROBES 1659edddaa2SAnanth N Mavinakayanahalli def_bool y 1669edddaa2SAnanth N Mavinakayanahalli depends on KPROBES && HAVE_KRETPROBES 1679edddaa2SAnanth N Mavinakayanahalli 1687c68af6eSAvi Kivityconfig USER_RETURN_NOTIFIER 1697c68af6eSAvi Kivity bool 1707c68af6eSAvi Kivity depends on HAVE_USER_RETURN_NOTIFIER 1717c68af6eSAvi Kivity help 1727c68af6eSAvi Kivity Provide a kernel-internal notification when a cpu is about to 1737c68af6eSAvi Kivity switch to user mode. 1747c68af6eSAvi Kivity 17528b2ee20SRik van Rielconfig HAVE_IOREMAP_PROT 1769ba16087SJan Beulich bool 17728b2ee20SRik van Riel 178125e5645SMathieu Desnoyersconfig HAVE_KPROBES 1799ba16087SJan Beulich bool 1809edddaa2SAnanth N Mavinakayanahalli 1819edddaa2SAnanth N Mavinakayanahalliconfig HAVE_KRETPROBES 1829ba16087SJan Beulich bool 18374bc7ceeSArthur Kepner 184afd66255SMasami Hiramatsuconfig HAVE_OPTPROBES 185afd66255SMasami Hiramatsu bool 186d314d74cSCong Wang 187e7dbfe34SMasami Hiramatsuconfig HAVE_KPROBES_ON_FTRACE 188e7dbfe34SMasami Hiramatsu bool 189e7dbfe34SMasami Hiramatsu 19042a0bb3fSPetr Mladekconfig HAVE_NMI 19142a0bb3fSPetr Mladek bool 19242a0bb3fSPetr Mladek 193d314d74cSCong Wangconfig HAVE_NMI_WATCHDOG 19442a0bb3fSPetr Mladek depends on HAVE_NMI 195d314d74cSCong Wang bool 1961f5a4ad9SRoland McGrath# 1971f5a4ad9SRoland McGrath# An arch should select this if it provides all these things: 1981f5a4ad9SRoland McGrath# 1991f5a4ad9SRoland McGrath# task_pt_regs() in asm/processor.h or asm/ptrace.h 2001f5a4ad9SRoland McGrath# arch_has_single_step() if there is hardware single-step support 2011f5a4ad9SRoland McGrath# arch_has_block_step() if there is hardware block-step support 2021f5a4ad9SRoland McGrath# asm/syscall.h supplying asm-generic/syscall.h interface 2031f5a4ad9SRoland McGrath# linux/regset.h user_regset interfaces 2041f5a4ad9SRoland McGrath# CORE_DUMP_USE_REGSET #define'd in linux/elf.h 2051f5a4ad9SRoland McGrath# TIF_SYSCALL_TRACE calls tracehook_report_syscall_{entry,exit} 2061f5a4ad9SRoland McGrath# TIF_NOTIFY_RESUME calls tracehook_notify_resume() 2071f5a4ad9SRoland McGrath# signal delivery calls tracehook_signal_handler() 2081f5a4ad9SRoland McGrath# 2091f5a4ad9SRoland McGrathconfig HAVE_ARCH_TRACEHOOK 2109ba16087SJan Beulich bool 2111f5a4ad9SRoland McGrath 212c64be2bbSMarek Szyprowskiconfig HAVE_DMA_CONTIGUOUS 213c64be2bbSMarek Szyprowski bool 214c64be2bbSMarek Szyprowski 21529d5e047SThomas Gleixnerconfig GENERIC_SMP_IDLE_THREAD 21629d5e047SThomas Gleixner bool 21729d5e047SThomas Gleixner 218485cf5daSKevin Hilmanconfig GENERIC_IDLE_POLL_SETUP 219485cf5daSKevin Hilman bool 220485cf5daSKevin Hilman 221a6359d1eSThomas Gleixner# Select if arch init_task initializer is different to init/init_task.c 222a6359d1eSThomas Gleixnerconfig ARCH_INIT_TASK 223a4a2eb49SThomas Gleixner bool 224a4a2eb49SThomas Gleixner 225f5e10287SThomas Gleixner# Select if arch has its private alloc_task_struct() function 226f5e10287SThomas Gleixnerconfig ARCH_TASK_STRUCT_ALLOCATOR 227f5e10287SThomas Gleixner bool 228f5e10287SThomas Gleixner 229b235beeaSLinus Torvalds# Select if arch has its private alloc_thread_stack() function 230b235beeaSLinus Torvaldsconfig ARCH_THREAD_STACK_ALLOCATOR 231f5e10287SThomas Gleixner bool 232f5e10287SThomas Gleixner 2335aaeb5c0SIngo Molnar# Select if arch wants to size task_struct dynamically via arch_task_struct_size: 2345aaeb5c0SIngo Molnarconfig ARCH_WANTS_DYNAMIC_TASK_STRUCT 2355aaeb5c0SIngo Molnar bool 2365aaeb5c0SIngo Molnar 237f850c30cSHeiko Carstensconfig HAVE_REGS_AND_STACK_ACCESS_API 238f850c30cSHeiko Carstens bool 239e01292b1SHeiko Carstens help 240e01292b1SHeiko Carstens This symbol should be selected by an architecure if it supports 241e01292b1SHeiko Carstens the API needed to access registers and stack entries from pt_regs, 242e01292b1SHeiko Carstens declared in asm/ptrace.h 243e01292b1SHeiko Carstens For example the kprobes-based event tracer needs this API. 244f850c30cSHeiko Carstens 2459483a578SDavid Brownellconfig HAVE_CLK 2469ba16087SJan Beulich bool 2479483a578SDavid Brownell help 2489483a578SDavid Brownell The <linux/clk.h> calls support software clock gating and 2499483a578SDavid Brownell thus are a key power management tool on many systems. 2509483a578SDavid Brownell 2515ee00bd4SJoerg Roedelconfig HAVE_DMA_API_DEBUG 2525ee00bd4SJoerg Roedel bool 25336cd3c9fSHeiko Carstens 25462a038d3SK.Prasadconfig HAVE_HW_BREAKPOINT 25562a038d3SK.Prasad bool 25699e8c5a3SFrederic Weisbecker depends on PERF_EVENTS 25762a038d3SK.Prasad 2580102752eSFrederic Weisbeckerconfig HAVE_MIXED_BREAKPOINTS_REGS 2590102752eSFrederic Weisbecker bool 2600102752eSFrederic Weisbecker depends on HAVE_HW_BREAKPOINT 2610102752eSFrederic Weisbecker help 2620102752eSFrederic Weisbecker Depending on the arch implementation of hardware breakpoints, 2630102752eSFrederic Weisbecker some of them have separate registers for data and instruction 2640102752eSFrederic Weisbecker breakpoints addresses, others have mixed registers to store 2650102752eSFrederic Weisbecker them but define the access type in a control register. 2660102752eSFrederic Weisbecker Select this option if your arch implements breakpoints under the 2670102752eSFrederic Weisbecker latter fashion. 2680102752eSFrederic Weisbecker 2697c68af6eSAvi Kivityconfig HAVE_USER_RETURN_NOTIFIER 2707c68af6eSAvi Kivity bool 271a1922ed6SIngo Molnar 272c01d4323SFrederic Weisbeckerconfig HAVE_PERF_EVENTS_NMI 273c01d4323SFrederic Weisbecker bool 27423637d47SFrederic Weisbecker help 27523637d47SFrederic Weisbecker System hardware can generate an NMI using the perf event 27623637d47SFrederic Weisbecker subsystem. Also has support for calculating CPU cycle events 27723637d47SFrederic Weisbecker to determine how many clock cycles in a given period. 278c01d4323SFrederic Weisbecker 279c5e63197SJiri Olsaconfig HAVE_PERF_REGS 280c5e63197SJiri Olsa bool 281c5e63197SJiri Olsa help 282c5e63197SJiri Olsa Support selective register dumps for perf events. This includes 283c5e63197SJiri Olsa bit-mapping of each registers and a unique architecture id. 284c5e63197SJiri Olsa 285c5ebcedbSJiri Olsaconfig HAVE_PERF_USER_STACK_DUMP 286c5ebcedbSJiri Olsa bool 287c5ebcedbSJiri Olsa help 288c5ebcedbSJiri Olsa Support user stack dumps for perf event samples. This needs 289c5ebcedbSJiri Olsa access to the user stack pointer which is not unified across 290c5ebcedbSJiri Olsa architectures. 291c5ebcedbSJiri Olsa 292bf5438fcSJason Baronconfig HAVE_ARCH_JUMP_LABEL 293bf5438fcSJason Baron bool 294bf5438fcSJason Baron 29526723911SPeter Zijlstraconfig HAVE_RCU_TABLE_FREE 29626723911SPeter Zijlstra bool 29726723911SPeter Zijlstra 298df013ffbSHuang Yingconfig ARCH_HAVE_NMI_SAFE_CMPXCHG 299df013ffbSHuang Ying bool 300df013ffbSHuang Ying 30143570fd2SHeiko Carstensconfig HAVE_ALIGNED_STRUCT_PAGE 30243570fd2SHeiko Carstens bool 30343570fd2SHeiko Carstens help 30443570fd2SHeiko Carstens This makes sure that struct pages are double word aligned and that 30543570fd2SHeiko Carstens e.g. the SLUB allocator can perform double word atomic operations 30643570fd2SHeiko Carstens on a struct page for better performance. However selecting this 30743570fd2SHeiko Carstens might increase the size of a struct page by a word. 30843570fd2SHeiko Carstens 3094156153cSHeiko Carstensconfig HAVE_CMPXCHG_LOCAL 3104156153cSHeiko Carstens bool 3114156153cSHeiko Carstens 3122565409fSHeiko Carstensconfig HAVE_CMPXCHG_DOUBLE 3132565409fSHeiko Carstens bool 3142565409fSHeiko Carstens 315c1d7e01dSWill Deaconconfig ARCH_WANT_IPC_PARSE_VERSION 316c1d7e01dSWill Deacon bool 317c1d7e01dSWill Deacon 318c1d7e01dSWill Deaconconfig ARCH_WANT_COMPAT_IPC_PARSE_VERSION 319c1d7e01dSWill Deacon bool 320c1d7e01dSWill Deacon 32148b25c43SChris Metcalfconfig ARCH_WANT_OLD_COMPAT_IPC 322c1d7e01dSWill Deacon select ARCH_WANT_COMPAT_IPC_PARSE_VERSION 32348b25c43SChris Metcalf bool 32448b25c43SChris Metcalf 325e2cfabdfSWill Drewryconfig HAVE_ARCH_SECCOMP_FILTER 326e2cfabdfSWill Drewry bool 327e2cfabdfSWill Drewry help 328fb0fadf9SWill Drewry An arch should select this symbol if it provides all of these things: 329bb6ea430SWill Drewry - syscall_get_arch() 330bb6ea430SWill Drewry - syscall_get_arguments() 331bb6ea430SWill Drewry - syscall_rollback() 332bb6ea430SWill Drewry - syscall_set_return_value() 333fb0fadf9SWill Drewry - SIGSYS siginfo_t support 334fb0fadf9SWill Drewry - secure_computing is called from a ptrace_event()-safe context 335fb0fadf9SWill Drewry - secure_computing return value is checked and a return value of -1 336fb0fadf9SWill Drewry results in the system call being skipped immediately. 33748dc92b9SKees Cook - seccomp syscall wired up 338e2cfabdfSWill Drewry 339ff27f38eSAndy Lutomirski For best performance, an arch should use seccomp_phase1 and 340ff27f38eSAndy Lutomirski seccomp_phase2 directly. It should call seccomp_phase1 for all 341ff27f38eSAndy Lutomirski syscalls if TIF_SECCOMP is set, but seccomp_phase1 does not 342ff27f38eSAndy Lutomirski need to be called from a ptrace-safe context. It must then 343ff27f38eSAndy Lutomirski call seccomp_phase2 if seccomp_phase1 returns anything other 344ff27f38eSAndy Lutomirski than SECCOMP_PHASE1_OK or SECCOMP_PHASE1_SKIP. 345ff27f38eSAndy Lutomirski 346ff27f38eSAndy Lutomirski As an additional optimization, an arch may provide seccomp_data 347ff27f38eSAndy Lutomirski directly to seccomp_phase1; this avoids multiple calls 348ff27f38eSAndy Lutomirski to the syscall_xyz helpers for every syscall. 349ff27f38eSAndy Lutomirski 350e2cfabdfSWill Drewryconfig SECCOMP_FILTER 351e2cfabdfSWill Drewry def_bool y 352e2cfabdfSWill Drewry depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET 353e2cfabdfSWill Drewry help 354e2cfabdfSWill Drewry Enable tasks to build secure computing environments defined 355e2cfabdfSWill Drewry in terms of Berkeley Packet Filter programs which implement 356e2cfabdfSWill Drewry task-defined system call filtering polices. 357e2cfabdfSWill Drewry 358e2cfabdfSWill Drewry See Documentation/prctl/seccomp_filter.txt for details. 359e2cfabdfSWill Drewry 3606b90bd4bSEmese Revfyconfig HAVE_GCC_PLUGINS 3616b90bd4bSEmese Revfy bool 3626b90bd4bSEmese Revfy help 3636b90bd4bSEmese Revfy An arch should select this symbol if it supports building with 3646b90bd4bSEmese Revfy GCC plugins. 3656b90bd4bSEmese Revfy 3666b90bd4bSEmese Revfymenuconfig GCC_PLUGINS 3676b90bd4bSEmese Revfy bool "GCC plugins" 3686b90bd4bSEmese Revfy depends on HAVE_GCC_PLUGINS 369a519167eSKees Cook depends on !COMPILE_TEST 3706b90bd4bSEmese Revfy help 3716b90bd4bSEmese Revfy GCC plugins are loadable modules that provide extra features to the 3726b90bd4bSEmese Revfy compiler. They are useful for runtime instrumentation and static analysis. 3736b90bd4bSEmese Revfy 3746b90bd4bSEmese Revfy See Documentation/gcc-plugins.txt for details. 3756b90bd4bSEmese Revfy 3760dae776cSEmese Revfyconfig GCC_PLUGIN_CYC_COMPLEXITY 3770dae776cSEmese Revfy bool "Compute the cyclomatic complexity of a function" 3780dae776cSEmese Revfy depends on GCC_PLUGINS 3790dae776cSEmese Revfy help 3800dae776cSEmese Revfy The complexity M of a function's control flow graph is defined as: 3810dae776cSEmese Revfy M = E - N + 2P 3820dae776cSEmese Revfy where 3830dae776cSEmese Revfy 3840dae776cSEmese Revfy E = the number of edges 3850dae776cSEmese Revfy N = the number of nodes 3860dae776cSEmese Revfy P = the number of connected components (exit nodes). 3870dae776cSEmese Revfy 388543c37cbSEmese Revfyconfig GCC_PLUGIN_SANCOV 389543c37cbSEmese Revfy bool 390543c37cbSEmese Revfy depends on GCC_PLUGINS 391543c37cbSEmese Revfy help 392543c37cbSEmese Revfy This plugin inserts a __sanitizer_cov_trace_pc() call at the start of 393543c37cbSEmese Revfy basic blocks. It supports all gcc versions with plugin support (from 394543c37cbSEmese Revfy gcc-4.5 on). It is based on the commit "Add fuzzing coverage support" 395543c37cbSEmese Revfy by Dmitry Vyukov <dvyukov@google.com>. 396543c37cbSEmese Revfy 39719952a92SKees Cookconfig HAVE_CC_STACKPROTECTOR 39819952a92SKees Cook bool 39919952a92SKees Cook help 40019952a92SKees Cook An arch should select this symbol if: 40119952a92SKees Cook - its compiler supports the -fstack-protector option 40219952a92SKees Cook - it has implemented a stack canary (e.g. __stack_chk_guard) 40319952a92SKees Cook 40419952a92SKees Cookconfig CC_STACKPROTECTOR 4058779657dSKees Cook def_bool n 40619952a92SKees Cook help 4078779657dSKees Cook Set when a stack-protector mode is enabled, so that the build 4088779657dSKees Cook can enable kernel-side support for the GCC feature. 4098779657dSKees Cook 4108779657dSKees Cookchoice 4118779657dSKees Cook prompt "Stack Protector buffer overflow detection" 4128779657dSKees Cook depends on HAVE_CC_STACKPROTECTOR 4138779657dSKees Cook default CC_STACKPROTECTOR_NONE 4148779657dSKees Cook help 4158779657dSKees Cook This option turns on the "stack-protector" GCC feature. This 41619952a92SKees Cook feature puts, at the beginning of functions, a canary value on 41719952a92SKees Cook the stack just before the return address, and validates 41819952a92SKees Cook the value just before actually returning. Stack based buffer 41919952a92SKees Cook overflows (that need to overwrite this return address) now also 42019952a92SKees Cook overwrite the canary, which gets detected and the attack is then 42119952a92SKees Cook neutralized via a kernel panic. 42219952a92SKees Cook 4238779657dSKees Cookconfig CC_STACKPROTECTOR_NONE 4248779657dSKees Cook bool "None" 4258779657dSKees Cook help 4268779657dSKees Cook Disable "stack-protector" GCC feature. 4278779657dSKees Cook 4288779657dSKees Cookconfig CC_STACKPROTECTOR_REGULAR 4298779657dSKees Cook bool "Regular" 4308779657dSKees Cook select CC_STACKPROTECTOR 4318779657dSKees Cook help 4328779657dSKees Cook Functions will have the stack-protector canary logic added if they 4338779657dSKees Cook have an 8-byte or larger character array on the stack. 4348779657dSKees Cook 43519952a92SKees Cook This feature requires gcc version 4.2 or above, or a distribution 4368779657dSKees Cook gcc with the feature backported ("-fstack-protector"). 4378779657dSKees Cook 4388779657dSKees Cook On an x86 "defconfig" build, this feature adds canary checks to 4398779657dSKees Cook about 3% of all kernel functions, which increases kernel code size 4408779657dSKees Cook by about 0.3%. 4418779657dSKees Cook 4428779657dSKees Cookconfig CC_STACKPROTECTOR_STRONG 4438779657dSKees Cook bool "Strong" 4448779657dSKees Cook select CC_STACKPROTECTOR 4458779657dSKees Cook help 4468779657dSKees Cook Functions will have the stack-protector canary logic added in any 4478779657dSKees Cook of the following conditions: 4488779657dSKees Cook 4498779657dSKees Cook - local variable's address used as part of the right hand side of an 4508779657dSKees Cook assignment or function argument 4518779657dSKees Cook - local variable is an array (or union containing an array), 4528779657dSKees Cook regardless of array type or length 4538779657dSKees Cook - uses register local variables 4548779657dSKees Cook 4558779657dSKees Cook This feature requires gcc version 4.9 or above, or a distribution 4568779657dSKees Cook gcc with the feature backported ("-fstack-protector-strong"). 4578779657dSKees Cook 4588779657dSKees Cook On an x86 "defconfig" build, this feature adds canary checks to 4598779657dSKees Cook about 20% of all kernel functions, which increases the kernel code 4608779657dSKees Cook size by about 2%. 4618779657dSKees Cook 4628779657dSKees Cookendchoice 46319952a92SKees Cook 4640f60a8efSKees Cookconfig HAVE_ARCH_WITHIN_STACK_FRAMES 4650f60a8efSKees Cook bool 4660f60a8efSKees Cook help 4670f60a8efSKees Cook An architecture should select this if it can walk the kernel stack 4680f60a8efSKees Cook frames to determine if an object is part of either the arguments 4690f60a8efSKees Cook or local variables (i.e. that it excludes saved return addresses, 4700f60a8efSKees Cook and similar) by implementing an inline arch_within_stack_frames(), 4710f60a8efSKees Cook which is used by CONFIG_HARDENED_USERCOPY. 4720f60a8efSKees Cook 47391d1aa43SFrederic Weisbeckerconfig HAVE_CONTEXT_TRACKING 4742b1d5024SFrederic Weisbecker bool 4752b1d5024SFrederic Weisbecker help 47691d1aa43SFrederic Weisbecker Provide kernel/user boundaries probes necessary for subsystems 47791d1aa43SFrederic Weisbecker that need it, such as userspace RCU extended quiescent state. 47891d1aa43SFrederic Weisbecker Syscalls need to be wrapped inside user_exit()-user_enter() through 47991d1aa43SFrederic Weisbecker the slow path using TIF_NOHZ flag. Exceptions handlers must be 48091d1aa43SFrederic Weisbecker wrapped as well. Irqs are already protected inside 48191d1aa43SFrederic Weisbecker rcu_irq_enter/rcu_irq_exit() but preemption or signal handling on 48291d1aa43SFrederic Weisbecker irq exit still need to be protected. 4832b1d5024SFrederic Weisbecker 484b952741cSFrederic Weisbeckerconfig HAVE_VIRT_CPU_ACCOUNTING 485b952741cSFrederic Weisbecker bool 486b952741cSFrederic Weisbecker 487554b0004SKevin Hilmanconfig HAVE_VIRT_CPU_ACCOUNTING_GEN 488554b0004SKevin Hilman bool 489554b0004SKevin Hilman default y if 64BIT 490554b0004SKevin Hilman help 491554b0004SKevin Hilman With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit. 492554b0004SKevin Hilman Before enabling this option, arch code must be audited 493554b0004SKevin Hilman to ensure there are no races in concurrent read/write of 494554b0004SKevin Hilman cputime_t. For example, reading/writing 64-bit cputime_t on 495554b0004SKevin Hilman some 32-bit arches may require multiple accesses, so proper 496554b0004SKevin Hilman locking is needed to protect against concurrent accesses. 497554b0004SKevin Hilman 498554b0004SKevin Hilman 499fdf9c356SFrederic Weisbeckerconfig HAVE_IRQ_TIME_ACCOUNTING 500fdf9c356SFrederic Weisbecker bool 501fdf9c356SFrederic Weisbecker help 502fdf9c356SFrederic Weisbecker Archs need to ensure they use a high enough resolution clock to 503fdf9c356SFrederic Weisbecker support irq time accounting and then call enable_sched_clock_irqtime(). 504fdf9c356SFrederic Weisbecker 50515626062SGerald Schaeferconfig HAVE_ARCH_TRANSPARENT_HUGEPAGE 50615626062SGerald Schaefer bool 50715626062SGerald Schaefer 5080ddab1d2SToshi Kaniconfig HAVE_ARCH_HUGE_VMAP 5090ddab1d2SToshi Kani bool 5100ddab1d2SToshi Kani 5110f8975ecSPavel Emelyanovconfig HAVE_ARCH_SOFT_DIRTY 5120f8975ecSPavel Emelyanov bool 5130f8975ecSPavel Emelyanov 514786d35d4SDavid Howellsconfig HAVE_MOD_ARCH_SPECIFIC 515786d35d4SDavid Howells bool 516786d35d4SDavid Howells help 517786d35d4SDavid Howells The arch uses struct mod_arch_specific to store data. Many arches 518786d35d4SDavid Howells just need a simple module loader without arch specific data - those 519786d35d4SDavid Howells should not enable this. 520786d35d4SDavid Howells 521786d35d4SDavid Howellsconfig MODULES_USE_ELF_RELA 522786d35d4SDavid Howells bool 523786d35d4SDavid Howells help 524786d35d4SDavid Howells Modules only use ELF RELA relocations. Modules with ELF REL 525786d35d4SDavid Howells relocations will give an error. 526786d35d4SDavid Howells 527786d35d4SDavid Howellsconfig MODULES_USE_ELF_REL 528786d35d4SDavid Howells bool 529786d35d4SDavid Howells help 530786d35d4SDavid Howells Modules only use ELF REL relocations. Modules with ELF RELA 531786d35d4SDavid Howells relocations will give an error. 532786d35d4SDavid Howells 533b92021b0SRusty Russellconfig HAVE_UNDERSCORE_SYMBOL_PREFIX 534b92021b0SRusty Russell bool 535b92021b0SRusty Russell help 536b92021b0SRusty Russell Some architectures generate an _ in front of C symbols; things like 537b92021b0SRusty Russell module loading and assembly files need to know about this. 538b92021b0SRusty Russell 539cc1f0274SFrederic Weisbeckerconfig HAVE_IRQ_EXIT_ON_IRQ_STACK 540cc1f0274SFrederic Weisbecker bool 541cc1f0274SFrederic Weisbecker help 542cc1f0274SFrederic Weisbecker Architecture doesn't only execute the irq handler on the irq stack 543cc1f0274SFrederic Weisbecker but also irq_exit(). This way we can process softirqs on this irq 544cc1f0274SFrederic Weisbecker stack instead of switching to a new one when we call __do_softirq() 545cc1f0274SFrederic Weisbecker in the end of an hardirq. 546cc1f0274SFrederic Weisbecker This spares a stack switch and improves cache usage on softirq 547cc1f0274SFrederic Weisbecker processing. 548cc1f0274SFrederic Weisbecker 549235a8f02SKirill A. Shutemovconfig PGTABLE_LEVELS 550235a8f02SKirill A. Shutemov int 551235a8f02SKirill A. Shutemov default 2 552235a8f02SKirill A. Shutemov 5532b68f6caSKees Cookconfig ARCH_HAS_ELF_RANDOMIZE 5542b68f6caSKees Cook bool 5552b68f6caSKees Cook help 5562b68f6caSKees Cook An architecture supports choosing randomized locations for 5572b68f6caSKees Cook stack, mmap, brk, and ET_DYN. Defined functions: 5582b68f6caSKees Cook - arch_mmap_rnd() 559204db6edSKees Cook - arch_randomize_brk() 5602b68f6caSKees Cook 561d07e2259SDaniel Cashmanconfig HAVE_ARCH_MMAP_RND_BITS 562d07e2259SDaniel Cashman bool 563d07e2259SDaniel Cashman help 564d07e2259SDaniel Cashman An arch should select this symbol if it supports setting a variable 565d07e2259SDaniel Cashman number of bits for use in establishing the base address for mmap 566d07e2259SDaniel Cashman allocations, has MMU enabled and provides values for both: 567d07e2259SDaniel Cashman - ARCH_MMAP_RND_BITS_MIN 568d07e2259SDaniel Cashman - ARCH_MMAP_RND_BITS_MAX 569d07e2259SDaniel Cashman 5705f56a5dfSJiri Slabyconfig HAVE_EXIT_THREAD 5715f56a5dfSJiri Slaby bool 5725f56a5dfSJiri Slaby help 5735f56a5dfSJiri Slaby An architecture implements exit_thread. 5745f56a5dfSJiri Slaby 575d07e2259SDaniel Cashmanconfig ARCH_MMAP_RND_BITS_MIN 576d07e2259SDaniel Cashman int 577d07e2259SDaniel Cashman 578d07e2259SDaniel Cashmanconfig ARCH_MMAP_RND_BITS_MAX 579d07e2259SDaniel Cashman int 580d07e2259SDaniel Cashman 581d07e2259SDaniel Cashmanconfig ARCH_MMAP_RND_BITS_DEFAULT 582d07e2259SDaniel Cashman int 583d07e2259SDaniel Cashman 584d07e2259SDaniel Cashmanconfig ARCH_MMAP_RND_BITS 585d07e2259SDaniel Cashman int "Number of bits to use for ASLR of mmap base address" if EXPERT 586d07e2259SDaniel Cashman range ARCH_MMAP_RND_BITS_MIN ARCH_MMAP_RND_BITS_MAX 587d07e2259SDaniel Cashman default ARCH_MMAP_RND_BITS_DEFAULT if ARCH_MMAP_RND_BITS_DEFAULT 588d07e2259SDaniel Cashman default ARCH_MMAP_RND_BITS_MIN 589d07e2259SDaniel Cashman depends on HAVE_ARCH_MMAP_RND_BITS 590d07e2259SDaniel Cashman help 591d07e2259SDaniel Cashman This value can be used to select the number of bits to use to 592d07e2259SDaniel Cashman determine the random offset to the base address of vma regions 593d07e2259SDaniel Cashman resulting from mmap allocations. This value will be bounded 594d07e2259SDaniel Cashman by the architecture's minimum and maximum supported values. 595d07e2259SDaniel Cashman 596d07e2259SDaniel Cashman This value can be changed after boot using the 597d07e2259SDaniel Cashman /proc/sys/vm/mmap_rnd_bits tunable 598d07e2259SDaniel Cashman 599d07e2259SDaniel Cashmanconfig HAVE_ARCH_MMAP_RND_COMPAT_BITS 600d07e2259SDaniel Cashman bool 601d07e2259SDaniel Cashman help 602d07e2259SDaniel Cashman An arch should select this symbol if it supports running applications 603d07e2259SDaniel Cashman in compatibility mode, supports setting a variable number of bits for 604d07e2259SDaniel Cashman use in establishing the base address for mmap allocations, has MMU 605d07e2259SDaniel Cashman enabled and provides values for both: 606d07e2259SDaniel Cashman - ARCH_MMAP_RND_COMPAT_BITS_MIN 607d07e2259SDaniel Cashman - ARCH_MMAP_RND_COMPAT_BITS_MAX 608d07e2259SDaniel Cashman 609d07e2259SDaniel Cashmanconfig ARCH_MMAP_RND_COMPAT_BITS_MIN 610d07e2259SDaniel Cashman int 611d07e2259SDaniel Cashman 612d07e2259SDaniel Cashmanconfig ARCH_MMAP_RND_COMPAT_BITS_MAX 613d07e2259SDaniel Cashman int 614d07e2259SDaniel Cashman 615d07e2259SDaniel Cashmanconfig ARCH_MMAP_RND_COMPAT_BITS_DEFAULT 616d07e2259SDaniel Cashman int 617d07e2259SDaniel Cashman 618d07e2259SDaniel Cashmanconfig ARCH_MMAP_RND_COMPAT_BITS 619d07e2259SDaniel Cashman int "Number of bits to use for ASLR of mmap base address for compatible applications" if EXPERT 620d07e2259SDaniel Cashman range ARCH_MMAP_RND_COMPAT_BITS_MIN ARCH_MMAP_RND_COMPAT_BITS_MAX 621d07e2259SDaniel Cashman default ARCH_MMAP_RND_COMPAT_BITS_DEFAULT if ARCH_MMAP_RND_COMPAT_BITS_DEFAULT 622d07e2259SDaniel Cashman default ARCH_MMAP_RND_COMPAT_BITS_MIN 623d07e2259SDaniel Cashman depends on HAVE_ARCH_MMAP_RND_COMPAT_BITS 624d07e2259SDaniel Cashman help 625d07e2259SDaniel Cashman This value can be used to select the number of bits to use to 626d07e2259SDaniel Cashman determine the random offset to the base address of vma regions 627d07e2259SDaniel Cashman resulting from mmap allocations for compatible applications This 628d07e2259SDaniel Cashman value will be bounded by the architecture's minimum and maximum 629d07e2259SDaniel Cashman supported values. 630d07e2259SDaniel Cashman 631d07e2259SDaniel Cashman This value can be changed after boot using the 632d07e2259SDaniel Cashman /proc/sys/vm/mmap_rnd_compat_bits tunable 633d07e2259SDaniel Cashman 6343033f14aSJosh Triplettconfig HAVE_COPY_THREAD_TLS 6353033f14aSJosh Triplett bool 6363033f14aSJosh Triplett help 6373033f14aSJosh Triplett Architecture provides copy_thread_tls to accept tls argument via 6383033f14aSJosh Triplett normal C parameter passing, rather than extracting the syscall 6393033f14aSJosh Triplett argument from pt_regs. 6403033f14aSJosh Triplett 641b9ab5ebbSJosh Poimboeufconfig HAVE_STACK_VALIDATION 642b9ab5ebbSJosh Poimboeuf bool 643b9ab5ebbSJosh Poimboeuf help 644b9ab5ebbSJosh Poimboeuf Architecture supports the 'objtool check' host tool command, which 645b9ab5ebbSJosh Poimboeuf performs compile-time stack metadata validation. 646b9ab5ebbSJosh Poimboeuf 647468a9428SGeorge Spelvinconfig HAVE_ARCH_HASH 648468a9428SGeorge Spelvin bool 649468a9428SGeorge Spelvin default n 650468a9428SGeorge Spelvin help 651468a9428SGeorge Spelvin If this is set, the architecture provides an <asm/hash.h> 652468a9428SGeorge Spelvin file which provides platform-specific implementations of some 653468a9428SGeorge Spelvin functions in <linux/hash.h> or fs/namei.c. 654468a9428SGeorge Spelvin 6553a495511SWilliam Breathitt Grayconfig ISA_BUS_API 6563a495511SWilliam Breathitt Gray def_bool ISA 6573a495511SWilliam Breathitt Gray 658d2125043SAl Viro# 659d2125043SAl Viro# ABI hall of shame 660d2125043SAl Viro# 661d2125043SAl Viroconfig CLONE_BACKWARDS 662d2125043SAl Viro bool 663d2125043SAl Viro help 664d2125043SAl Viro Architecture has tls passed as the 4th argument of clone(2), 665d2125043SAl Viro not the 5th one. 666d2125043SAl Viro 667d2125043SAl Viroconfig CLONE_BACKWARDS2 668d2125043SAl Viro bool 669d2125043SAl Viro help 670d2125043SAl Viro Architecture has the first two arguments of clone(2) swapped. 671d2125043SAl Viro 672dfa9771aSMichal Simekconfig CLONE_BACKWARDS3 673dfa9771aSMichal Simek bool 674dfa9771aSMichal Simek help 675dfa9771aSMichal Simek Architecture has tls passed as the 3rd argument of clone(2), 676dfa9771aSMichal Simek not the 5th one. 677dfa9771aSMichal Simek 678eaca6eaeSAl Viroconfig ODD_RT_SIGACTION 679eaca6eaeSAl Viro bool 680eaca6eaeSAl Viro help 681eaca6eaeSAl Viro Architecture has unusual rt_sigaction(2) arguments 682eaca6eaeSAl Viro 6830a0e8cdfSAl Viroconfig OLD_SIGSUSPEND 6840a0e8cdfSAl Viro bool 6850a0e8cdfSAl Viro help 6860a0e8cdfSAl Viro Architecture has old sigsuspend(2) syscall, of one-argument variety 6870a0e8cdfSAl Viro 6880a0e8cdfSAl Viroconfig OLD_SIGSUSPEND3 6890a0e8cdfSAl Viro bool 6900a0e8cdfSAl Viro help 6910a0e8cdfSAl Viro Even weirder antique ABI - three-argument sigsuspend(2) 6920a0e8cdfSAl Viro 693495dfbf7SAl Viroconfig OLD_SIGACTION 694495dfbf7SAl Viro bool 695495dfbf7SAl Viro help 696495dfbf7SAl Viro Architecture has old sigaction(2) syscall. Nope, not the same 697495dfbf7SAl Viro as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2), 698495dfbf7SAl Viro but fairly different variant of sigaction(2), thanks to OSF/1 699495dfbf7SAl Viro compatibility... 700495dfbf7SAl Viro 701495dfbf7SAl Viroconfig COMPAT_OLD_SIGACTION 702495dfbf7SAl Viro bool 703495dfbf7SAl Viro 7040d4a619bSChristoph Hellwigconfig ARCH_NO_COHERENT_DMA_MMAP 7050d4a619bSChristoph Hellwig bool 7060d4a619bSChristoph Hellwig 707fff7fb0bSZhaoxiu Zengconfig CPU_NO_EFFICIENT_FFS 708fff7fb0bSZhaoxiu Zeng def_bool n 709fff7fb0bSZhaoxiu Zeng 710ba14a194SAndy Lutomirskiconfig HAVE_ARCH_VMAP_STACK 711ba14a194SAndy Lutomirski def_bool n 712ba14a194SAndy Lutomirski help 713ba14a194SAndy Lutomirski An arch should select this symbol if it can support kernel stacks 714ba14a194SAndy Lutomirski in vmalloc space. This means: 715ba14a194SAndy Lutomirski 716ba14a194SAndy Lutomirski - vmalloc space must be large enough to hold many kernel stacks. 717ba14a194SAndy Lutomirski This may rule out many 32-bit architectures. 718ba14a194SAndy Lutomirski 719ba14a194SAndy Lutomirski - Stacks in vmalloc space need to work reliably. For example, if 720ba14a194SAndy Lutomirski vmap page tables are created on demand, either this mechanism 721ba14a194SAndy Lutomirski needs to work while the stack points to a virtual address with 722ba14a194SAndy Lutomirski unpopulated page tables or arch code (switch_to() and switch_mm(), 723ba14a194SAndy Lutomirski most likely) needs to ensure that the stack's page table entries 724ba14a194SAndy Lutomirski are populated before running on a possibly unpopulated stack. 725ba14a194SAndy Lutomirski 726ba14a194SAndy Lutomirski - If the stack overflows into a guard page, something reasonable 727ba14a194SAndy Lutomirski should happen. The definition of "reasonable" is flexible, but 728ba14a194SAndy Lutomirski instantly rebooting without logging anything would be unfriendly. 729ba14a194SAndy Lutomirski 730ba14a194SAndy Lutomirskiconfig VMAP_STACK 731ba14a194SAndy Lutomirski default y 732ba14a194SAndy Lutomirski bool "Use a virtually-mapped stack" 733ba14a194SAndy Lutomirski depends on HAVE_ARCH_VMAP_STACK && !KASAN 734ba14a194SAndy Lutomirski ---help--- 735ba14a194SAndy Lutomirski Enable this if you want the use virtually-mapped kernel stacks 736ba14a194SAndy Lutomirski with guard pages. This causes kernel stack overflows to be 737ba14a194SAndy Lutomirski caught immediately rather than causing difficult-to-diagnose 738ba14a194SAndy Lutomirski corruption. 739ba14a194SAndy Lutomirski 740ba14a194SAndy Lutomirski This is presently incompatible with KASAN because KASAN expects 741ba14a194SAndy Lutomirski the stack to map directly to the KASAN shadow map using a formula 742ba14a194SAndy Lutomirski that is incorrect if the stack is in vmalloc space. 743ba14a194SAndy Lutomirski 7442521f2c2SPeter Oberparleitersource "kernel/gcov/Kconfig" 745