2 # General architecture dependent options
6 tristate "OProfile system profiling"
8 depends on HAVE_OPROFILE
10 select RING_BUFFER_ALLOW_SWAP
12 OProfile is a profiling system capable of profiling the
13 whole system, include the kernel, kernel modules, libraries,
18 config OPROFILE_EVENT_MULTIPLEX
19 bool "OProfile multiplexing support (EXPERIMENTAL)"
21 depends on OPROFILE && X86
23 The number of hardware counters is limited. The multiplexing
24 feature enables OProfile to gather more events than counters
25 are provided by the hardware. This is realized by switching
26 between events at an user specified time interval.
33 config OPROFILE_NMI_TIMER
35 depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI && !PPC64
40 depends on HAVE_KPROBES
43 Kprobes allows you to trap at almost any kernel address and
44 execute a callback function. register_kprobe() establishes
45 a probepoint and specifies the callback. Kprobes is useful
46 for kernel debugging, non-intrusive instrumentation and testing.
50 bool "Optimize very unlikely/likely branches"
51 depends on HAVE_ARCH_JUMP_LABEL
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.
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.
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.
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.
71 ( On 32-bit x86, the necessary options added to the compiler
72 flags may increase the size of the kernel slightly. )
76 depends on KPROBES && HAVE_OPTPROBES
79 config KPROBES_ON_FTRACE
81 depends on KPROBES && HAVE_KPROBES_ON_FTRACE
82 depends on DYNAMIC_FTRACE_WITH_REGS
84 If function tracer is enabled and the arch supports full
85 passing of pt_regs to function tracing, then kprobes can
86 optimize on top of function tracing.
91 Uprobes is the user-space counterpart to kprobes: they
92 enable instrumentation applications (such as 'perf probe')
93 to establish unintrusive probes in user-space binaries and
94 libraries, by executing handler functions when the probes
95 are hit by user-space applications.
97 ( These probes come in the form of single-byte breakpoints,
98 managed by the kernel and kept transparent to the probed
101 config HAVE_64BIT_ALIGNED_ACCESS
102 def_bool 64BIT && !HAVE_EFFICIENT_UNALIGNED_ACCESS
104 Some architectures require 64 bit accesses to be 64 bit
105 aligned, which also requires structs containing 64 bit values
106 to be 64 bit aligned too. This includes some 32 bit
107 architectures which can do 64 bit accesses, as well as 64 bit
108 architectures without unaligned access.
110 This symbol should be selected by an architecture if 64 bit
111 accesses are required to be 64 bit aligned in this way even
112 though it is not a 64 bit architecture.
114 See Documentation/unaligned-memory-access.txt for more
115 information on the topic of unaligned memory accesses.
117 config HAVE_EFFICIENT_UNALIGNED_ACCESS
120 Some architectures are unable to perform unaligned accesses
121 without the use of get_unaligned/put_unaligned. Others are
122 unable to perform such accesses efficiently (e.g. trap on
123 unaligned access and require fixing it up in the exception
126 This symbol should be selected by an architecture if it can
127 perform unaligned accesses efficiently to allow different
128 code paths to be selected for these cases. Some network
129 drivers, for example, could opt to not fix up alignment
130 problems with received packets if doing so would not help
133 See Documentation/unaligned-memory-access.txt for more
134 information on the topic of unaligned memory accesses.
136 config ARCH_USE_BUILTIN_BSWAP
139 Modern versions of GCC (since 4.4) have builtin functions
140 for handling byte-swapping. Using these, instead of the old
141 inline assembler that the architecture code provides in the
142 __arch_bswapXX() macros, allows the compiler to see what's
143 happening and offers more opportunity for optimisation. In
144 particular, the compiler will be able to combine the byteswap
145 with a nearby load or store and use load-and-swap or
146 store-and-swap instructions if the architecture has them. It
147 should almost *never* result in code which is worse than the
148 hand-coded assembler in <asm/swab.h>. But just in case it
149 does, the use of the builtins is optional.
151 Any architecture with load-and-swap or store-and-swap
152 instructions should set this. And it shouldn't hurt to set it
153 on architectures that don't have such instructions.
157 depends on KPROBES && HAVE_KRETPROBES
159 config USER_RETURN_NOTIFIER
161 depends on HAVE_USER_RETURN_NOTIFIER
163 Provide a kernel-internal notification when a cpu is about to
166 config HAVE_IOREMAP_PROT
172 config HAVE_KRETPROBES
175 config HAVE_OPTPROBES
178 config HAVE_KPROBES_ON_FTRACE
181 config HAVE_NMI_WATCHDOG
184 # An arch should select this if it provides all these things:
186 # task_pt_regs() in asm/processor.h or asm/ptrace.h
187 # arch_has_single_step() if there is hardware single-step support
188 # arch_has_block_step() if there is hardware block-step support
189 # asm/syscall.h supplying asm-generic/syscall.h interface
190 # linux/regset.h user_regset interfaces
191 # CORE_DUMP_USE_REGSET #define'd in linux/elf.h
192 # TIF_SYSCALL_TRACE calls tracehook_report_syscall_{entry,exit}
193 # TIF_NOTIFY_RESUME calls tracehook_notify_resume()
194 # signal delivery calls tracehook_signal_handler()
196 config HAVE_ARCH_TRACEHOOK
199 config HAVE_DMA_ATTRS
202 config HAVE_DMA_CONTIGUOUS
205 config GENERIC_SMP_IDLE_THREAD
208 config GENERIC_IDLE_POLL_SETUP
211 # Select if arch init_task initializer is different to init/init_task.c
212 config ARCH_INIT_TASK
215 # Select if arch has its private alloc_task_struct() function
216 config ARCH_TASK_STRUCT_ALLOCATOR
219 # Select if arch has its private alloc_thread_info() function
220 config ARCH_THREAD_INFO_ALLOCATOR
223 # Select if arch wants to size task_struct dynamically via arch_task_struct_size:
224 config ARCH_WANTS_DYNAMIC_TASK_STRUCT
227 config HAVE_REGS_AND_STACK_ACCESS_API
230 This symbol should be selected by an architecure if it supports
231 the API needed to access registers and stack entries from pt_regs,
232 declared in asm/ptrace.h
233 For example the kprobes-based event tracer needs this API.
238 The <linux/clk.h> calls support software clock gating and
239 thus are a key power management tool on many systems.
241 config HAVE_DMA_API_DEBUG
244 config HAVE_HW_BREAKPOINT
246 depends on PERF_EVENTS
248 config HAVE_MIXED_BREAKPOINTS_REGS
250 depends on HAVE_HW_BREAKPOINT
252 Depending on the arch implementation of hardware breakpoints,
253 some of them have separate registers for data and instruction
254 breakpoints addresses, others have mixed registers to store
255 them but define the access type in a control register.
256 Select this option if your arch implements breakpoints under the
259 config HAVE_USER_RETURN_NOTIFIER
262 config HAVE_PERF_EVENTS_NMI
265 System hardware can generate an NMI using the perf event
266 subsystem. Also has support for calculating CPU cycle events
267 to determine how many clock cycles in a given period.
269 config HAVE_PERF_REGS
272 Support selective register dumps for perf events. This includes
273 bit-mapping of each registers and a unique architecture id.
275 config HAVE_PERF_USER_STACK_DUMP
278 Support user stack dumps for perf event samples. This needs
279 access to the user stack pointer which is not unified across
282 config HAVE_ARCH_JUMP_LABEL
285 config HAVE_RCU_TABLE_FREE
288 config ARCH_HAVE_NMI_SAFE_CMPXCHG
291 config HAVE_ALIGNED_STRUCT_PAGE
294 This makes sure that struct pages are double word aligned and that
295 e.g. the SLUB allocator can perform double word atomic operations
296 on a struct page for better performance. However selecting this
297 might increase the size of a struct page by a word.
299 config HAVE_CMPXCHG_LOCAL
302 config HAVE_CMPXCHG_DOUBLE
305 config ARCH_WANT_IPC_PARSE_VERSION
308 config ARCH_WANT_COMPAT_IPC_PARSE_VERSION
311 config ARCH_WANT_OLD_COMPAT_IPC
312 select ARCH_WANT_COMPAT_IPC_PARSE_VERSION
315 config HAVE_ARCH_SECCOMP_FILTER
318 An arch should select this symbol if it provides all of these things:
320 - syscall_get_arguments()
322 - syscall_set_return_value()
323 - SIGSYS siginfo_t support
324 - secure_computing is called from a ptrace_event()-safe context
325 - secure_computing return value is checked and a return value of -1
326 results in the system call being skipped immediately.
327 - seccomp syscall wired up
329 For best performance, an arch should use seccomp_phase1 and
330 seccomp_phase2 directly. It should call seccomp_phase1 for all
331 syscalls if TIF_SECCOMP is set, but seccomp_phase1 does not
332 need to be called from a ptrace-safe context. It must then
333 call seccomp_phase2 if seccomp_phase1 returns anything other
334 than SECCOMP_PHASE1_OK or SECCOMP_PHASE1_SKIP.
336 As an additional optimization, an arch may provide seccomp_data
337 directly to seccomp_phase1; this avoids multiple calls
338 to the syscall_xyz helpers for every syscall.
340 config SECCOMP_FILTER
342 depends on HAVE_ARCH_SECCOMP_FILTER && SECCOMP && NET
344 Enable tasks to build secure computing environments defined
345 in terms of Berkeley Packet Filter programs which implement
346 task-defined system call filtering polices.
348 See Documentation/prctl/seccomp_filter.txt for details.
350 config HAVE_CC_STACKPROTECTOR
353 An arch should select this symbol if:
354 - its compiler supports the -fstack-protector option
355 - it has implemented a stack canary (e.g. __stack_chk_guard)
357 config CC_STACKPROTECTOR
360 Set when a stack-protector mode is enabled, so that the build
361 can enable kernel-side support for the GCC feature.
364 prompt "Stack Protector buffer overflow detection"
365 depends on HAVE_CC_STACKPROTECTOR
366 default CC_STACKPROTECTOR_NONE
368 This option turns on the "stack-protector" GCC feature. This
369 feature puts, at the beginning of functions, a canary value on
370 the stack just before the return address, and validates
371 the value just before actually returning. Stack based buffer
372 overflows (that need to overwrite this return address) now also
373 overwrite the canary, which gets detected and the attack is then
374 neutralized via a kernel panic.
376 config CC_STACKPROTECTOR_NONE
379 Disable "stack-protector" GCC feature.
381 config CC_STACKPROTECTOR_REGULAR
383 select CC_STACKPROTECTOR
385 Functions will have the stack-protector canary logic added if they
386 have an 8-byte or larger character array on the stack.
388 This feature requires gcc version 4.2 or above, or a distribution
389 gcc with the feature backported ("-fstack-protector").
391 On an x86 "defconfig" build, this feature adds canary checks to
392 about 3% of all kernel functions, which increases kernel code size
395 config CC_STACKPROTECTOR_STRONG
397 select CC_STACKPROTECTOR
399 Functions will have the stack-protector canary logic added in any
400 of the following conditions:
402 - local variable's address used as part of the right hand side of an
403 assignment or function argument
404 - local variable is an array (or union containing an array),
405 regardless of array type or length
406 - uses register local variables
408 This feature requires gcc version 4.9 or above, or a distribution
409 gcc with the feature backported ("-fstack-protector-strong").
411 On an x86 "defconfig" build, this feature adds canary checks to
412 about 20% of all kernel functions, which increases the kernel code
417 config HAVE_CONTEXT_TRACKING
420 Provide kernel/user boundaries probes necessary for subsystems
421 that need it, such as userspace RCU extended quiescent state.
422 Syscalls need to be wrapped inside user_exit()-user_enter() through
423 the slow path using TIF_NOHZ flag. Exceptions handlers must be
424 wrapped as well. Irqs are already protected inside
425 rcu_irq_enter/rcu_irq_exit() but preemption or signal handling on
426 irq exit still need to be protected.
428 config HAVE_VIRT_CPU_ACCOUNTING
431 config HAVE_VIRT_CPU_ACCOUNTING_GEN
435 With VIRT_CPU_ACCOUNTING_GEN, cputime_t becomes 64-bit.
436 Before enabling this option, arch code must be audited
437 to ensure there are no races in concurrent read/write of
438 cputime_t. For example, reading/writing 64-bit cputime_t on
439 some 32-bit arches may require multiple accesses, so proper
440 locking is needed to protect against concurrent accesses.
443 config HAVE_IRQ_TIME_ACCOUNTING
446 Archs need to ensure they use a high enough resolution clock to
447 support irq time accounting and then call enable_sched_clock_irqtime().
449 config HAVE_ARCH_TRANSPARENT_HUGEPAGE
452 config HAVE_ARCH_HUGE_VMAP
455 config HAVE_ARCH_SOFT_DIRTY
458 config HAVE_MOD_ARCH_SPECIFIC
461 The arch uses struct mod_arch_specific to store data. Many arches
462 just need a simple module loader without arch specific data - those
463 should not enable this.
465 config MODULES_USE_ELF_RELA
468 Modules only use ELF RELA relocations. Modules with ELF REL
469 relocations will give an error.
471 config MODULES_USE_ELF_REL
474 Modules only use ELF REL relocations. Modules with ELF RELA
475 relocations will give an error.
477 config HAVE_UNDERSCORE_SYMBOL_PREFIX
480 Some architectures generate an _ in front of C symbols; things like
481 module loading and assembly files need to know about this.
483 config HAVE_IRQ_EXIT_ON_IRQ_STACK
486 Architecture doesn't only execute the irq handler on the irq stack
487 but also irq_exit(). This way we can process softirqs on this irq
488 stack instead of switching to a new one when we call __do_softirq()
489 in the end of an hardirq.
490 This spares a stack switch and improves cache usage on softirq
493 config PGTABLE_LEVELS
497 config ARCH_HAS_ELF_RANDOMIZE
500 An architecture supports choosing randomized locations for
501 stack, mmap, brk, and ET_DYN. Defined functions:
503 - arch_randomize_brk()
505 config HAVE_COPY_THREAD_TLS
508 Architecture provides copy_thread_tls to accept tls argument via
509 normal C parameter passing, rather than extracting the syscall
510 argument from pt_regs.
515 config CLONE_BACKWARDS
518 Architecture has tls passed as the 4th argument of clone(2),
521 config CLONE_BACKWARDS2
524 Architecture has the first two arguments of clone(2) swapped.
526 config CLONE_BACKWARDS3
529 Architecture has tls passed as the 3rd argument of clone(2),
532 config ODD_RT_SIGACTION
535 Architecture has unusual rt_sigaction(2) arguments
537 config OLD_SIGSUSPEND
540 Architecture has old sigsuspend(2) syscall, of one-argument variety
542 config OLD_SIGSUSPEND3
545 Even weirder antique ABI - three-argument sigsuspend(2)
550 Architecture has old sigaction(2) syscall. Nope, not the same
551 as OLD_SIGSUSPEND | OLD_SIGSUSPEND3 - alpha has sigsuspend(2),
552 but fairly different variant of sigaction(2), thanks to OSF/1
555 config COMPAT_OLD_SIGACTION
558 source "kernel/gcov/Kconfig"