2 # General architecture dependent options
9 tristate "OProfile system profiling"
11 depends on HAVE_OPROFILE
13 select RING_BUFFER_ALLOW_SWAP
15 OProfile is a profiling system capable of profiling the
16 whole system, include the kernel, kernel modules, libraries,
21 config OPROFILE_EVENT_MULTIPLEX
22 bool "OProfile multiplexing support (EXPERIMENTAL)"
24 depends on OPROFILE && X86
26 The number of hardware counters is limited. The multiplexing
27 feature enables OProfile to gather more events than counters
28 are provided by the hardware. This is realized by switching
29 between events at an user specified time interval.
36 config OPROFILE_NMI_TIMER
38 depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI && !PPC64
43 depends on HAVE_KPROBES
46 Kprobes allows you to trap at almost any kernel address and
47 execute a callback function. register_kprobe() establishes
48 a probepoint and specifies the callback. Kprobes is useful
49 for kernel debugging, non-intrusive instrumentation and testing.
53 bool "Optimize very unlikely/likely branches"
54 depends on HAVE_ARCH_JUMP_LABEL
56 This option enables a transparent branch optimization that
57 makes certain almost-always-true or almost-always-false branch
58 conditions even cheaper to execute within the kernel.
60 Certain performance-sensitive kernel code, such as trace points,
61 scheduler functionality, networking code and KVM have such
62 branches and include support for this optimization technique.
64 If it is detected that the compiler has support for "asm goto",
65 the kernel will compile such branches with just a nop
66 instruction. When the condition flag is toggled to true, the
67 nop will be converted to a jump instruction to execute the
68 conditional block of instructions.
70 This technique lowers overhead and stress on the branch prediction
71 of the processor and generally makes the kernel faster. The update
72 of the condition is slower, but those are always very rare.
74 ( On 32-bit x86, the necessary options added to the compiler
75 flags may increase the size of the kernel slightly. )
77 config STATIC_KEYS_SELFTEST
78 bool "Static key selftest"
81 Boot time self-test of the branch patching code.
85 depends on KPROBES && HAVE_OPTPROBES
88 config KPROBES_ON_FTRACE
90 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
91 depends on DYNAMIC_FTRACE_WITH_REGS
93 If function tracer is enabled and the arch supports full
94 passing of pt_regs to function tracing, then kprobes can
95 optimize on top of function tracing.
100 Uprobes is the user-space counterpart to kprobes: they
101 enable instrumentation applications (such as 'perf probe')
102 to establish unintrusive probes in user-space binaries and
103 libraries, by executing handler functions when the probes
104 are hit by user-space applications.
106 ( These probes come in the form of single-byte breakpoints,
107 managed by the kernel and kept transparent to the probed
110 config HAVE_64BIT_ALIGNED_ACCESS
111 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
113 Some architectures require 64 bit accesses to be 64 bit
114 aligned, which also requires structs containing 64 bit values
115 to be 64 bit aligned too. This includes some 32 bit
116 architectures which can do 64 bit accesses, as well as 64 bit
117 architectures without unaligned access.
119 This symbol should be selected by an architecture if 64 bit
120 accesses are required to be 64 bit aligned in this way even
121 though it is not a 64 bit architecture.
123 See Documentation/unaligned-memory-access.txt for more
124 information on the topic of unaligned memory accesses.
126 config HAVE_EFFICIENT_UNALIGNED_ACCESS
129 Some architectures are unable to perform unaligned accesses
130 without the use of get_unaligned/put_unaligned. Others are
131 unable to perform such accesses efficiently (e.g. trap on
132 unaligned access and require fixing it up in the exception
135 This symbol should be selected by an architecture if it can
136 perform unaligned accesses efficiently to allow different
137 code paths to be selected for these cases. Some network
138 drivers, for example, could opt to not fix up alignment
139 problems with received packets if doing so would not help
142 See Documentation/unaligned-memory-access.txt for more
143 information on the topic of unaligned memory accesses.
145 config ARCH_USE_BUILTIN_BSWAP
148 Modern versions of GCC (since 4.4) have builtin functions
149 for handling byte-swapping. Using these, instead of the old
150 inline assembler that the architecture code provides in the
151 __arch_bswapXX() macros, allows the compiler to see what's
152 happening and offers more opportunity for optimisation. In
153 particular, the compiler will be able to combine the byteswap
154 with a nearby load or store and use load-and-swap or
155 store-and-swap instructions if the architecture has them. It
156 should almost *never* result in code which is worse than the
157 hand-coded assembler in <asm/swab.h>. But just in case it
158 does, the use of the builtins is optional.
160 Any architecture with load-and-swap or store-and-swap
161 instructions should set this. And it shouldn't hurt to set it
162 on architectures that don't have such instructions.
166 depends on KPROBES && HAVE_KRETPROBES
168 config USER_RETURN_NOTIFIER
170 depends on HAVE_USER_RETURN_NOTIFIER
172 Provide a kernel-internal notification when a cpu is about to
175 config HAVE_IOREMAP_PROT
181 config HAVE_KRETPROBES
184 config HAVE_OPTPROBES
187 config HAVE_KPROBES_ON_FTRACE
190 config HAVE_NMI_WATCHDOG
193 # An arch should select this if it provides all these things:
195 # task_pt_regs() in asm/processor.h or asm/ptrace.h
196 # arch_has_single_step() if there is hardware single-step support
197 # arch_has_block_step() if there is hardware block-step support
198 # asm/syscall.h supplying asm-generic/syscall.h interface
199 # linux/regset.h user_regset interfaces
200 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
201 # TIF_SYSCALL_TRACE calls tracehook_report_syscall_{entry,exit}
202 # TIF_NOTIFY_RESUME calls tracehook_notify_resume()
203 # signal delivery calls tracehook_signal_handler()
205 config HAVE_ARCH_TRACEHOOK
208 config HAVE_DMA_ATTRS
211 config HAVE_DMA_CONTIGUOUS
214 config GENERIC_SMP_IDLE_THREAD
217 config GENERIC_IDLE_POLL_SETUP
220 # Select if arch init_task initializer is different to init/init_task.c
221 config ARCH_INIT_TASK
224 # Select if arch has its private alloc_task_struct() function
225 config ARCH_TASK_STRUCT_ALLOCATOR
228 # Select if arch has its private alloc_thread_info() function
229 config ARCH_THREAD_INFO_ALLOCATOR
232 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
233 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
236 config HAVE_REGS_AND_STACK_ACCESS_API
239 This symbol should be selected by an architecure if it supports
240 the API needed to access registers and stack entries from pt_regs,
241 declared in asm/ptrace.h
242 For example the kprobes-based event tracer needs this API.
247 The <linux/clk.h> calls support software clock gating and
248 thus are a key power management tool on many systems.
250 config HAVE_DMA_API_DEBUG
253 config HAVE_HW_BREAKPOINT
255 depends on PERF_EVENTS
257 config HAVE_MIXED_BREAKPOINTS_REGS
259 depends on HAVE_HW_BREAKPOINT
261 Depending on the arch implementation of hardware breakpoints,
262 some of them have separate registers for data and instruction
263 breakpoints addresses, others have mixed registers to store
264 them but define the access type in a control register.
265 Select this option if your arch implements breakpoints under the
268 config HAVE_USER_RETURN_NOTIFIER
271 config HAVE_PERF_EVENTS_NMI
274 System hardware can generate an NMI using the perf event
275 subsystem. Also has support for calculating CPU cycle events
276 to determine how many clock cycles in a given period.
278 config HAVE_PERF_REGS
281 Support selective register dumps for perf events. This includes
282 bit-mapping of each registers and a unique architecture id.
284 config HAVE_PERF_USER_STACK_DUMP
287 Support user stack dumps for perf event samples. This needs
288 access to the user stack pointer which is not unified across
291 config HAVE_ARCH_JUMP_LABEL
294 config HAVE_RCU_TABLE_FREE
297 config ARCH_HAVE_NMI_SAFE_CMPXCHG
300 config HAVE_ALIGNED_STRUCT_PAGE
303 This makes sure that struct pages are double word aligned and that
304 e.g. the SLUB allocator can perform double word atomic operations
305 on a struct page for better performance. However selecting this
306 might increase the size of a struct page by a word.
308 config HAVE_CMPXCHG_LOCAL
311 config HAVE_CMPXCHG_DOUBLE
314 config ARCH_WANT_IPC_PARSE_VERSION
317 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
320 config ARCH_WANT_OLD_COMPAT_IPC
321 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
324 config HAVE_ARCH_SECCOMP_FILTER
327 An arch should select this symbol if it provides all of these things:
329 - syscall_get_arguments()
331 - syscall_set_return_value()
332 - SIGSYS siginfo_t support
333 - secure_computing is called from a ptrace_event()-safe context
334 - secure_computing return value is checked and a return value of -1
335 results in the system call being skipped immediately.
336 - seccomp syscall wired up
338 For best performance, an arch should use seccomp_phase1 and
339 seccomp_phase2 directly. It should call seccomp_phase1 for all
340 syscalls if TIF_SECCOMP is set, but seccomp_phase1 does not
341 need to be called from a ptrace-safe context. It must then
342 call seccomp_phase2 if seccomp_phase1 returns anything other
343 than SECCOMP_PHASE1_OK or SECCOMP_PHASE1_SKIP.
345 As an additional optimization, an arch may provide seccomp_data
346 directly to seccomp_phase1; this avoids multiple calls
347 to the syscall_xyz helpers for every syscall.
349 config SECCOMP_FILTER
351 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
353 Enable tasks to build secure computing environments defined
354 in terms of Berkeley Packet Filter programs which implement
355 task-defined system call filtering polices.
357 See Documentation/prctl/seccomp_filter.txt for details.
359 config HAVE_CC_STACKPROTECTOR
362 An arch should select this symbol if:
363 - its compiler supports the -fstack-protector option
364 - it has implemented a stack canary (e.g. __stack_chk_guard)
366 config CC_STACKPROTECTOR
369 Set when a stack-protector mode is enabled, so that the build
370 can enable kernel-side support for the GCC feature.
373 prompt "Stack Protector buffer overflow detection"
374 depends on HAVE_CC_STACKPROTECTOR
375 default CC_STACKPROTECTOR_NONE
377 This option turns on the "stack-protector" GCC feature. This
378 feature puts, at the beginning of functions, a canary value on
379 the stack just before the return address, and validates
380 the value just before actually returning. Stack based buffer
381 overflows (that need to overwrite this return address) now also
382 overwrite the canary, which gets detected and the attack is then
383 neutralized via a kernel panic.
385 config CC_STACKPROTECTOR_NONE
388 Disable "stack-protector" GCC feature.
390 config CC_STACKPROTECTOR_REGULAR
392 select CC_STACKPROTECTOR
394 Functions will have the stack-protector canary logic added if they
395 have an 8-byte or larger character array on the stack.
397 This feature requires gcc version 4.2 or above, or a distribution
398 gcc with the feature backported ("-fstack-protector").
400 On an x86 "defconfig" build, this feature adds canary checks to
401 about 3% of all kernel functions, which increases kernel code size
404 config CC_STACKPROTECTOR_STRONG
406 select CC_STACKPROTECTOR
408 Functions will have the stack-protector canary logic added in any
409 of the following conditions:
411 - local variable's address used as part of the right hand side of an
412 assignment or function argument
413 - local variable is an array (or union containing an array),
414 regardless of array type or length
415 - uses register local variables
417 This feature requires gcc version 4.9 or above, or a distribution
418 gcc with the feature backported ("-fstack-protector-strong").
420 On an x86 "defconfig" build, this feature adds canary checks to
421 about 20% of all kernel functions, which increases the kernel code
426 config HAVE_CONTEXT_TRACKING
429 Provide kernel/user boundaries probes necessary for subsystems
430 that need it, such as userspace RCU extended quiescent state.
431 Syscalls need to be wrapped inside user_exit()-user_enter() through
432 the slow path using TIF_NOHZ flag. Exceptions handlers must be
433 wrapped as well. Irqs are already protected inside
434 rcu_irq_enter/rcu_irq_exit() but preemption or signal handling on
435 irq exit still need to be protected.
437 config HAVE_VIRT_CPU_ACCOUNTING
440 config HAVE_VIRT_CPU_ACCOUNTING_GEN
444 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
445 Before enabling this option, arch code must be audited
446 to ensure there are no races in concurrent read/write of
447 cputime_t. For example, reading/writing 64-bit cputime_t on
448 some 32-bit arches may require multiple accesses, so proper
449 locking is needed to protect against concurrent accesses.
452 config HAVE_IRQ_TIME_ACCOUNTING
455 Archs need to ensure they use a high enough resolution clock to
456 support irq time accounting and then call enable_sched_clock_irqtime().
458 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
461 config HAVE_ARCH_HUGE_VMAP
464 config HAVE_ARCH_SOFT_DIRTY
467 config HAVE_MOD_ARCH_SPECIFIC
470 The arch uses struct mod_arch_specific to store data. Many arches
471 just need a simple module loader without arch specific data - those
472 should not enable this.
474 config MODULES_USE_ELF_RELA
477 Modules only use ELF RELA relocations. Modules with ELF REL
478 relocations will give an error.
480 config MODULES_USE_ELF_REL
483 Modules only use ELF REL relocations. Modules with ELF RELA
484 relocations will give an error.
486 config HAVE_UNDERSCORE_SYMBOL_PREFIX
489 Some architectures generate an _ in front of C symbols; things like
490 module loading and assembly files need to know about this.
492 config HAVE_IRQ_EXIT_ON_IRQ_STACK
495 Architecture doesn't only execute the irq handler on the irq stack
496 but also irq_exit(). This way we can process softirqs on this irq
497 stack instead of switching to a new one when we call __do_softirq()
498 in the end of an hardirq.
499 This spares a stack switch and improves cache usage on softirq
502 config PGTABLE_LEVELS
506 config ARCH_HAS_ELF_RANDOMIZE
509 An architecture supports choosing randomized locations for
510 stack, mmap, brk, and ET_DYN. Defined functions:
512 - arch_randomize_brk()
514 config HAVE_COPY_THREAD_TLS
517 Architecture provides copy_thread_tls to accept tls argument via
518 normal C parameter passing, rather than extracting the syscall
519 argument from pt_regs.
524 config CLONE_BACKWARDS
527 Architecture has tls passed as the 4th argument of clone(2),
530 config CLONE_BACKWARDS2
533 Architecture has the first two arguments of clone(2) swapped.
535 config CLONE_BACKWARDS3
538 Architecture has tls passed as the 3rd argument of clone(2),
541 config ODD_RT_SIGACTION
544 Architecture has unusual rt_sigaction(2) arguments
546 config OLD_SIGSUSPEND
549 Architecture has old sigsuspend(2) syscall, of one-argument variety
551 config OLD_SIGSUSPEND3
554 Even weirder antique ABI - three-argument sigsuspend(2)
559 Architecture has old sigaction(2) syscall. Nope, not the same
560 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
561 but fairly different variant of sigaction(2), thanks to OSF/1
564 config COMPAT_OLD_SIGACTION
567 source "kernel/gcov/Kconfig"