diff --git a/arch/x86/kernel/vsyscall_64.c b/arch/x86/kernel/vsyscall_64.c index 0b8b6690a86d184959703177b9cb6011bfd605ca..2f90202e59b33b4a014ac33297a5a5faeedcd14a 100644 --- a/arch/x86/kernel/vsyscall_64.c +++ b/arch/x86/kernel/vsyscall_64.c @@ -17,6 +17,14 @@ * want per guest time just set the kernel.vsyscall64 sysctl to 0. */ +/* Protect userspace from profiling */ +#ifdef CONFIG_TRACE_UNLIKELY_PROFILE +# undef likely +# undef unlikely +# define likely(x) likely_notrace(x) +# define unlikely(x) unlikely_notrace(x) +#endif + #include #include #include diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index 80744606bad172b57d2ab52dc01bb0bf33af293e..e10beb5335c93caeb3e2b8a46834eeb5b5cb72c0 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -45,6 +45,17 @@ #define MCOUNT_REC() #endif +#ifdef CONFIG_TRACE_UNLIKELY_PROFILE +#define LIKELY_PROFILE() VMLINUX_SYMBOL(__start_likely_profile) = .; \ + *(_ftrace_likely) \ + VMLINUX_SYMBOL(__stop_likely_profile) = .; \ + VMLINUX_SYMBOL(__start_unlikely_profile) = .; \ + *(_ftrace_unlikely) \ + VMLINUX_SYMBOL(__stop_unlikely_profile) = .; +#else +#define LIKELY_PROFILE() +#endif + /* .data section */ #define DATA_DATA \ *(.data) \ @@ -62,7 +73,8 @@ VMLINUX_SYMBOL(__stop___markers) = .; \ VMLINUX_SYMBOL(__start___tracepoints) = .; \ *(__tracepoints) \ - VMLINUX_SYMBOL(__stop___tracepoints) = .; + VMLINUX_SYMBOL(__stop___tracepoints) = .; \ + LIKELY_PROFILE() #define RO_DATA(align) \ . = ALIGN((align)); \ diff --git a/include/linux/compiler.h b/include/linux/compiler.h index 98115d9d04daa6c8008b528bee3014a8cee11078..935e30cfaf3c1635c67ae2f197a1ac4445038406 100644 --- a/include/linux/compiler.h +++ b/include/linux/compiler.h @@ -59,8 +59,65 @@ extern void __chk_io_ptr(const volatile void __iomem *); * specific implementations come from the above header files */ -#define likely(x) __builtin_expect(!!(x), 1) -#define unlikely(x) __builtin_expect(!!(x), 0) +#ifdef CONFIG_TRACE_UNLIKELY_PROFILE +struct ftrace_likely_data { + const char *func; + const char *file; + unsigned line; + unsigned long correct; + unsigned long incorrect; +}; +void ftrace_likely_update(struct ftrace_likely_data *f, int val, int expect); + +#define likely_notrace(x) __builtin_expect(!!(x), 1) +#define unlikely_notrace(x) __builtin_expect(!!(x), 0) + +#define likely_check(x) ({ \ + int ______r; \ + static struct ftrace_likely_data \ + __attribute__((__aligned__(4))) \ + __attribute__((section("_ftrace_likely"))) \ + ______f = { \ + .func = __func__, \ + .file = __FILE__, \ + .line = __LINE__, \ + }; \ + ______f.line = __LINE__; \ + ______r = likely_notrace(x); \ + ftrace_likely_update(&______f, ______r, 1); \ + ______r; \ + }) +#define unlikely_check(x) ({ \ + int ______r; \ + static struct ftrace_likely_data \ + __attribute__((__aligned__(4))) \ + __attribute__((section("_ftrace_unlikely"))) \ + ______f = { \ + .func = __func__, \ + .file = __FILE__, \ + .line = __LINE__, \ + }; \ + ______f.line = __LINE__; \ + ______r = unlikely_notrace(x); \ + ftrace_likely_update(&______f, ______r, 0); \ + ______r; \ + }) + +/* + * Using __builtin_constant_p(x) to ignore cases where the return + * value is always the same. This idea is taken from a similar patch + * written by Daniel Walker. + */ +# ifndef likely +# define likely(x) (__builtin_constant_p(x) ? !!(x) : likely_check(x)) +# endif +# ifndef unlikely +# define unlikely(x) (__builtin_constant_p(x) ? !!(x) : unlikely_check(x)) +# endif +#else +# define likely(x) __builtin_expect(!!(x), 1) +# define unlikely(x) __builtin_expect(!!(x), 0) +#endif /* Optimization barrier */ #ifndef barrier diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index d986216c8327a7003f5757f51d12e465410dc8c6..a604f24c755f2426768145f01aac7f03d92076d0 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig @@ -159,6 +159,22 @@ config BOOT_TRACER selected, because the self-tests are an initcall as well and that would invalidate the boot trace. ) +config TRACE_UNLIKELY_PROFILE + bool "Trace likely/unlikely profiler" + depends on DEBUG_KERNEL + select TRACING + help + This tracer profiles all the the likely and unlikely macros + in the kernel. It will display the results in: + + /debugfs/tracing/profile_likely + /debugfs/tracing/profile_unlikely + + Note: this will add a significant overhead, only turn this + on if you need to profile the system's use of these macros. + + Say N if unsure. + config STACK_TRACER bool "Trace max stack" depends on HAVE_FUNCTION_TRACER diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index 3e1f361bbc179d3c619267977057ba98ef708b79..98e70ee2798687c1e5f127ee19e8a802b0528481 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile @@ -25,5 +25,6 @@ obj-$(CONFIG_STACK_TRACER) += trace_stack.o obj-$(CONFIG_MMIOTRACE) += trace_mmiotrace.o obj-$(CONFIG_BOOT_TRACER) += trace_boot.o obj-$(CONFIG_FUNCTION_RET_TRACER) += trace_functions_return.o +obj-$(CONFIG_TRACE_UNLIKELY_PROFILE) += trace_unlikely.o libftrace-y := ftrace.o diff --git a/kernel/trace/trace_unlikely.c b/kernel/trace/trace_unlikely.c new file mode 100644 index 0000000000000000000000000000000000000000..94932696069ff7f95ded475856ca7b21d2026f9b --- /dev/null +++ b/kernel/trace/trace_unlikely.c @@ -0,0 +1,164 @@ +/* + * unlikely profiler + * + * Copyright (C) 2008 Steven Rostedt + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "trace.h" + +void ftrace_likely_update(struct ftrace_likely_data *f, int val, int expect) +{ + /* FIXME: Make this atomic! */ + if (val == expect) + f->correct++; + else + f->incorrect++; +} +EXPORT_SYMBOL(ftrace_likely_update); + +struct ftrace_pointer { + void *start; + void *stop; +}; + +static void * +t_next(struct seq_file *m, void *v, loff_t *pos) +{ + struct ftrace_pointer *f = m->private; + struct ftrace_likely_data *p = v; + + (*pos)++; + + if (v == (void *)1) + return f->start; + + ++p; + + if ((void *)p >= (void *)f->stop) + return NULL; + + return p; +} + +static void *t_start(struct seq_file *m, loff_t *pos) +{ + void *t = (void *)1; + loff_t l = 0; + + for (; t && l < *pos; t = t_next(m, t, &l)) + ; + + return t; +} + +static void t_stop(struct seq_file *m, void *p) +{ +} + +static int t_show(struct seq_file *m, void *v) +{ + struct ftrace_likely_data *p = v; + const char *f; + unsigned long percent; + + if (v == (void *)1) { + seq_printf(m, " correct incorrect %% " + " Function " + " File Line\n" + " ------- --------- - " + " -------- " + " ---- ----\n"); + return 0; + } + + /* Only print the file, not the path */ + f = p->file + strlen(p->file); + while (f >= p->file && *f != '/') + f--; + f++; + + if (p->correct) { + percent = p->incorrect * 100; + percent /= p->correct + p->incorrect; + } else + percent = p->incorrect ? 100 : 0; + + seq_printf(m, "%8lu %8lu %3lu ", p->correct, p->incorrect, percent); + seq_printf(m, "%-30.30s %-20.20s %d\n", p->func, f, p->line); + return 0; +} + +static struct seq_operations tracing_likely_seq_ops = { + .start = t_start, + .next = t_next, + .stop = t_stop, + .show = t_show, +}; + +static int tracing_likely_open(struct inode *inode, struct file *file) +{ + int ret; + + ret = seq_open(file, &tracing_likely_seq_ops); + if (!ret) { + struct seq_file *m = file->private_data; + m->private = (void *)inode->i_private; + } + + return ret; +} + +static struct file_operations tracing_likely_fops = { + .open = tracing_likely_open, + .read = seq_read, + .llseek = seq_lseek, +}; + +extern unsigned long __start_likely_profile[]; +extern unsigned long __stop_likely_profile[]; +extern unsigned long __start_unlikely_profile[]; +extern unsigned long __stop_unlikely_profile[]; + +static struct ftrace_pointer ftrace_likely_pos = { + .start = __start_likely_profile, + .stop = __stop_likely_profile, +}; + +static struct ftrace_pointer ftrace_unlikely_pos = { + .start = __start_unlikely_profile, + .stop = __stop_unlikely_profile, +}; + +static __init int ftrace_unlikely_init(void) +{ + struct dentry *d_tracer; + struct dentry *entry; + + d_tracer = tracing_init_dentry(); + + entry = debugfs_create_file("profile_likely", 0444, d_tracer, + &ftrace_likely_pos, + &tracing_likely_fops); + if (!entry) + pr_warning("Could not create debugfs 'profile_likely' entry\n"); + + entry = debugfs_create_file("profile_unlikely", 0444, d_tracer, + &ftrace_unlikely_pos, + &tracing_likely_fops); + if (!entry) + pr_warning("Could not create debugfs" + " 'profile_unlikely' entry\n"); + + return 0; +} + +device_initcall(ftrace_unlikely_init);