aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-02-28 13:17:55 -0500
committerLinus Torvalds <torvalds@linux-foundation.org>2010-02-28 13:17:55 -0500
commite0d272429a34ff143bfa04ee8e29dd4eed2964c7 (patch)
tree5a719135b245811b5d61ed084d7b8c1bc2e87031
parentd25e8dbdab203ed8b4fd0a174bb5259e35ecd87c (diff)
parent480917427b0b6ff39de55ffc81391055472e6c26 (diff)
Merge branch 'tracing-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'tracing-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (28 commits) ftrace: Add function names to dangling } in function graph tracer tracing: Simplify memory recycle of trace_define_field tracing: Remove unnecessary variable in print_graph_return tracing: Fix typo of info text in trace_kprobe.c tracing: Fix typo in prof_sysexit_enable() tracing: Remove CONFIG_TRACE_POWER from kernel config tracing: Fix ftrace_event_call alignment for use with gcc 4.5 ftrace: Remove memory barriers from NMI code when not needed tracing/kprobes: Add short documentation for HAVE_REGS_AND_STACK_ACCESS_API s390: Add pt_regs register and stack access API tracing/kprobes: Make Kconfig dependencies generic tracing: Unify arch_syscall_addr() implementations tracing: Add notrace to TRACE_EVENT implementation functions ftrace: Allow to remove a single function from function graph filter tracing: Add correct/incorrect to sort keys for branch annotation output tracing: Simplify test for function_graph tracing start point tracing: Drop the tr check from the graph tracing path tracing: Add stack dump to trace_printk if stacktrace option is set tracing: Use appropriate perl constructs in recordmcount.pl tracing: optimize recordmcount.pl for offsets-handling ...
Diffstat
-rw-r--r--Documentation/trace/ftrace-design.txt5
-rw-r--r--arch/Kconfig8
-rw-r--r--arch/s390/Kconfig1
-rw-r--r--arch/s390/include/asm/ptrace.h13
-rw-r--r--arch/s390/include/asm/syscall.h7
-rw-r--r--arch/s390/kernel/ftrace.c10
-rw-r--r--arch/s390/kernel/ptrace.c58
-rw-r--r--arch/sh/include/asm/syscall.h2
-rw-r--r--arch/sh/kernel/ftrace.c9
-rw-r--r--arch/sparc/include/asm/syscall.h7
-rw-r--r--arch/sparc/kernel/ftrace.c11
-rw-r--r--arch/x86/Kconfig1
-rw-r--r--arch/x86/include/asm/syscall.h2
-rw-r--r--arch/x86/kernel/ftrace.c36
-rw-r--r--include/linux/ftrace.h6
-rw-r--r--include/linux/ftrace_event.h3
-rw-r--r--include/linux/syscalls.h12
-rw-r--r--include/trace/ftrace.h195
-rw-r--r--include/trace/syscall.h4
-rw-r--r--kernel/trace/Kconfig11
-rw-r--r--kernel/trace/ftrace.c51
-rw-r--r--kernel/trace/trace.c144
-rw-r--r--kernel/trace/trace.h6
-rw-r--r--kernel/trace/trace_branch.c19
-rw-r--r--kernel/trace/trace_events.c81
-rw-r--r--kernel/trace/trace_export.c87
-rw-r--r--kernel/trace/trace_functions_graph.c78
-rw-r--r--kernel/trace/trace_kprobe.c108
-rw-r--r--kernel/trace/trace_syscalls.c113
-rwxr-xr-xscripts/recordmcount.pl39
30 files changed, 580 insertions, 547 deletions
diff --git a/Documentation/trace/ftrace-design.txt b/Documentation/trace/ftrace-design.txt
index 6a5a579126b0..f1f81afee8a0 100644
--- a/Documentation/trace/ftrace-design.txt
+++ b/Documentation/trace/ftrace-design.txt
@@ -238,11 +238,10 @@ HAVE_SYSCALL_TRACEPOINTS
238 238
239You need very few things to get the syscalls tracing in an arch. 239You need very few things to get the syscalls tracing in an arch.
240 240
241- Support HAVE_ARCH_TRACEHOOK (see arch/Kconfig).
241- Have a NR_syscalls variable in <asm/unistd.h> that provides the number 242- Have a NR_syscalls variable in <asm/unistd.h> that provides the number
242 of syscalls supported by the arch. 243 of syscalls supported by the arch.
243- Implement arch_syscall_addr() that resolves a syscall address from a 244- Support the TIF_SYSCALL_TRACEPOINT thread flags.
244 syscall number.
245- Support the TIF_SYSCALL_TRACEPOINT thread flags
246- Put the trace_sys_enter() and trace_sys_exit() tracepoints calls from ptrace 245- Put the trace_sys_enter() and trace_sys_exit() tracepoints calls from ptrace
247 in the ptrace syscalls tracing path. 246 in the ptrace syscalls tracing path.
248- Tag this arch as HAVE_SYSCALL_TRACEPOINTS. 247- Tag this arch as HAVE_SYSCALL_TRACEPOINTS.
diff --git a/arch/Kconfig b/arch/Kconfig
index 06a13729c8df..215e46073c45 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -105,6 +105,14 @@ config HAVE_DMA_ATTRS
105config USE_GENERIC_SMP_HELPERS 105config USE_GENERIC_SMP_HELPERS
106 bool 106 bool
107 107
108config HAVE_REGS_AND_STACK_ACCESS_API
109 bool
110 help
111 This symbol should be selected by an architecure if it supports
112 the API needed to access registers and stack entries from pt_regs,
113 declared in asm/ptrace.h
114 For example the kprobes-based event tracer needs this API.
115
108config HAVE_CLK 116config HAVE_CLK
109 bool 117 bool
110 help 118 help
diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index 19deda8d8875..0d8cd9bbe101 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -90,6 +90,7 @@ config S390
90 select HAVE_SYSCALL_TRACEPOINTS 90 select HAVE_SYSCALL_TRACEPOINTS
91 select HAVE_DYNAMIC_FTRACE 91 select HAVE_DYNAMIC_FTRACE
92 select HAVE_FUNCTION_GRAPH_TRACER 92 select HAVE_FUNCTION_GRAPH_TRACER
93 select HAVE_REGS_AND_STACK_ACCESS_API
93 select HAVE_DEFAULT_NO_SPIN_MUTEXES 94 select HAVE_DEFAULT_NO_SPIN_MUTEXES
94 select HAVE_OPROFILE 95 select HAVE_OPROFILE
95 select HAVE_KPROBES 96 select HAVE_KPROBES
diff --git a/arch/s390/include/asm/ptrace.h b/arch/s390/include/asm/ptrace.h
index 95dcf183a28d..dd2d913afcae 100644
--- a/arch/s390/include/asm/ptrace.h
+++ b/arch/s390/include/asm/ptrace.h
@@ -492,13 +492,24 @@ struct user_regs_struct
492struct task_struct; 492struct task_struct;
493extern void user_enable_single_step(struct task_struct *); 493extern void user_enable_single_step(struct task_struct *);
494extern void user_disable_single_step(struct task_struct *); 494extern void user_disable_single_step(struct task_struct *);
495extern void show_regs(struct pt_regs * regs);
495 496
496#define user_mode(regs) (((regs)->psw.mask & PSW_MASK_PSTATE) != 0) 497#define user_mode(regs) (((regs)->psw.mask & PSW_MASK_PSTATE) != 0)
497#define instruction_pointer(regs) ((regs)->psw.addr & PSW_ADDR_INSN) 498#define instruction_pointer(regs) ((regs)->psw.addr & PSW_ADDR_INSN)
498#define user_stack_pointer(regs)((regs)->gprs[15]) 499#define user_stack_pointer(regs)((regs)->gprs[15])
499#define regs_return_value(regs)((regs)->gprs[2]) 500#define regs_return_value(regs)((regs)->gprs[2])
500#define profile_pc(regs) instruction_pointer(regs) 501#define profile_pc(regs) instruction_pointer(regs)
501extern void show_regs(struct pt_regs * regs); 502
503int regs_query_register_offset(const char *name);
504const char *regs_query_register_name(unsigned int offset);
505unsigned long regs_get_register(struct pt_regs *regs, unsigned int offset);
506unsigned long regs_get_kernel_stack_nth(struct pt_regs *regs, unsigned int n);
507
508static inline unsigned long kernel_stack_pointer(struct pt_regs *regs)
509{
510 return regs->gprs[15] & PSW_ADDR_INSN;
511}
512
502#endif /* __KERNEL__ */ 513#endif /* __KERNEL__ */
503#endif /* __ASSEMBLY__ */ 514#endif /* __ASSEMBLY__ */
504 515
diff --git a/arch/s390/include/asm/syscall.h b/arch/s390/include/asm/syscall.h
index e0a73d3eb837..8429686951f9 100644
--- a/arch/s390/include/asm/syscall.h
+++ b/arch/s390/include/asm/syscall.h
@@ -15,6 +15,13 @@
15#include <linux/sched.h> 15#include <linux/sched.h>
16#include <asm/ptrace.h> 16#include <asm/ptrace.h>
17 17
18/*
19 * The syscall table always contains 32 bit pointers since we know that the
20 * address of the function to be called is (way) below 4GB. So the "int"
21 * type here is what we want [need] for both 32 bit and 64 bit systems.
22 */
23extern const unsigned int sys_call_table[];
24
18static inline long syscall_get_nr(struct task_struct *task, 25static inline long syscall_get_nr(struct task_struct *task,
19 struct pt_regs *regs) 26 struct pt_regs *regs)
20{ 27{
diff --git a/arch/s390/kernel/ftrace.c b/arch/s390/kernel/ftrace.c
index 314d8f09cf31..6a83d0581317 100644
--- a/arch/s390/kernel/ftrace.c
+++ b/arch/s390/kernel/ftrace.c
@@ -200,13 +200,3 @@ out:
200 return parent; 200 return parent;
201} 201}
202#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ 202#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
203
204#ifdef CONFIG_FTRACE_SYSCALLS
205
206extern unsigned int sys_call_table[];
207
208unsigned long __init arch_syscall_addr(int nr)
209{
210 return (unsigned long)sys_call_table[nr];
211}
212#endif
diff --git a/arch/s390/kernel/ptrace.c b/arch/s390/kernel/ptrace.c
index 7cf464234419..33fdc5a79764 100644
--- a/arch/s390/kernel/ptrace.c
+++ b/arch/s390/kernel/ptrace.c
@@ -992,3 +992,61 @@ const struct user_regset_view *task_user_regset_view(struct task_struct *task)
992#endif 992#endif
993 return &user_s390_view; 993 return &user_s390_view;
994} 994}
995
996static const char *gpr_names[NUM_GPRS] = {
997 "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
998 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15",
999};
1000
1001unsigned long regs_get_register(struct pt_regs *regs, unsigned int offset)
1002{
1003 if (offset >= NUM_GPRS)
1004 return 0;
1005 return regs->gprs[offset];
1006}
1007
1008int regs_query_register_offset(const char *name)
1009{
1010 unsigned long offset;
1011
1012 if (!name || *name != 'r')
1013 return -EINVAL;
1014 if (strict_strtoul(name + 1, 10, &offset))
1015 return -EINVAL;
1016 if (offset >= NUM_GPRS)
1017 return -EINVAL;
1018 return offset;
1019}
1020
1021const char *regs_query_register_name(unsigned int offset)
1022{
1023 if (offset >= NUM_GPRS)
1024 return NULL;
1025 return gpr_names[offset];
1026}
1027
1028static int regs_within_kernel_stack(struct pt_regs *regs, unsigned long addr)
1029{
1030 unsigned long ksp = kernel_stack_pointer(regs);
1031
1032 return (addr & ~(THREAD_SIZE - 1)) == (ksp & ~(THREAD_SIZE - 1));
1033}
1034
1035/**
1036 * regs_get_kernel_stack_nth() - get Nth entry of the stack
1037 * @regs:pt_regs which contains kernel stack pointer.
1038 * @n:stack entry number.
1039 *
1040 * regs_get_kernel_stack_nth() returns @n th entry of the kernel stack which
1041 * is specifined by @regs. If the @n th entry is NOT in the kernel stack,
1042 * this returns 0.
1043 */
1044unsigned long regs_get_kernel_stack_nth(struct pt_regs *regs, unsigned int n)
1045{
1046 unsigned long addr;
1047
1048 addr = kernel_stack_pointer(regs) + n * sizeof(long);
1049 if (!regs_within_kernel_stack(regs, addr))
1050 return 0;
1051 return *(unsigned long *)addr;
1052}
diff --git a/arch/sh/include/asm/syscall.h b/arch/sh/include/asm/syscall.h
index 6a381429ee9d..aa7777bdc370 100644
--- a/arch/sh/include/asm/syscall.h
+++ b/arch/sh/include/asm/syscall.h
@@ -1,6 +1,8 @@
1#ifndef __ASM_SH_SYSCALL_H 1#ifndef __ASM_SH_SYSCALL_H
2#define __ASM_SH_SYSCALL_H 2#define __ASM_SH_SYSCALL_H
3 3
4extern const unsigned long sys_call_table[];
5
4#ifdef CONFIG_SUPERH32 6#ifdef CONFIG_SUPERH32
5# include "syscall_32.h" 7# include "syscall_32.h"
6#else 8#else
diff --git a/arch/sh/kernel/ftrace.c b/arch/sh/kernel/ftrace.c
index a48cdedc73b5..30e13196d35b 100644
--- a/arch/sh/kernel/ftrace.c
+++ b/arch/sh/kernel/ftrace.c
@@ -399,12 +399,3 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
399 } 399 }
400} 400}
401#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ 401#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
402
403#ifdef CONFIG_FTRACE_SYSCALLS
404extern unsigned long *sys_call_table;
405
406unsigned long __init arch_syscall_addr(int nr)
407{
408 return (unsigned long)sys_call_table[nr];
409}
410#endif /* CONFIG_FTRACE_SYSCALLS */
diff --git a/arch/sparc/include/asm/syscall.h b/arch/sparc/include/asm/syscall.h
index 7486c605e23c..025a02ad2e31 100644
--- a/arch/sparc/include/asm/syscall.h
+++ b/arch/sparc/include/asm/syscall.h
@@ -5,6 +5,13 @@
5#include <linux/sched.h> 5#include <linux/sched.h>
6#include <asm/ptrace.h> 6#include <asm/ptrace.h>
7 7
8/*
9 * The syscall table always contains 32 bit pointers since we know that the
10 * address of the function to be called is (way) below 4GB. So the "int"
11 * type here is what we want [need] for both 32 bit and 64 bit systems.
12 */
13extern const unsigned int sys_call_table[];
14
8/* The system call number is given by the user in %g1 */ 15/* The system call number is given by the user in %g1 */
9static inline long syscall_get_nr(struct task_struct *task, 16static inline long syscall_get_nr(struct task_struct *task,
10 struct pt_regs *regs) 17 struct pt_regs *regs)
diff --git a/arch/sparc/kernel/ftrace.c b/arch/sparc/kernel/ftrace.c
index 29973daa9930..9103a56b39e8 100644
--- a/arch/sparc/kernel/ftrace.c
+++ b/arch/sparc/kernel/ftrace.c
@@ -91,14 +91,3 @@ int __init ftrace_dyn_arch_init(void *data)
91 return 0; 91 return 0;
92} 92}
93#endif 93#endif
94
95#ifdef CONFIG_FTRACE_SYSCALLS
96
97extern unsigned int sys_call_table[];
98
99unsigned long __init arch_syscall_addr(int nr)
100{
101 return (unsigned long)sys_call_table[nr];
102}
103
104#endif
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index eb4092568f9e..0896008f7509 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -45,6 +45,7 @@ config X86
45 select HAVE_GENERIC_DMA_COHERENT if X86_32 45 select HAVE_GENERIC_DMA_COHERENT if X86_32
46 select HAVE_EFFICIENT_UNALIGNED_ACCESS 46 select HAVE_EFFICIENT_UNALIGNED_ACCESS
47 select USER_STACKTRACE_SUPPORT 47 select USER_STACKTRACE_SUPPORT
48 select HAVE_REGS_AND_STACK_ACCESS_API
48 select HAVE_DMA_API_DEBUG 49 select HAVE_DMA_API_DEBUG
49 select HAVE_KERNEL_GZIP 50 select HAVE_KERNEL_GZIP
50 select HAVE_KERNEL_BZIP2 51 select HAVE_KERNEL_BZIP2
diff --git a/arch/x86/include/asm/syscall.h b/arch/x86/include/asm/syscall.h
index 8d33bc5462d1..c4a348f7bd43 100644
--- a/arch/x86/include/asm/syscall.h
+++ b/arch/x86/include/asm/syscall.h
@@ -16,6 +16,8 @@
16#include <linux/sched.h> 16#include <linux/sched.h>
17#include <linux/err.h> 17#include <linux/err.h>
18 18
19extern const unsigned long sys_call_table[];
20
19/* 21/*
20 * Only the low 32 bits of orig_ax are meaningful, so we return int. 22 * Only the low 32 bits of orig_ax are meaningful, so we return int.
21 * This importantly ignores the high bits on 64-bit, so comparisons 23 * This importantly ignores the high bits on 64-bit, so comparisons
diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
index 309689245431..cd37469b54ee 100644
--- a/arch/x86/kernel/ftrace.c
+++ b/arch/x86/kernel/ftrace.c
@@ -30,14 +30,32 @@
30 30
31#ifdef CONFIG_DYNAMIC_FTRACE 31#ifdef CONFIG_DYNAMIC_FTRACE
32 32
33/*
34 * modifying_code is set to notify NMIs that they need to use
35 * memory barriers when entering or exiting. But we don't want
36 * to burden NMIs with unnecessary memory barriers when code
37 * modification is not being done (which is most of the time).
38 *
39 * A mutex is already held when ftrace_arch_code_modify_prepare
40 * and post_process are called. No locks need to be taken here.
41 *
42 * Stop machine will make sure currently running NMIs are done
43 * and new NMIs will see the updated variable before we need
44 * to worry about NMIs doing memory barriers.
45 */
46static int modifying_code __read_mostly;
47static DEFINE_PER_CPU(int, save_modifying_code);
48
33int ftrace_arch_code_modify_prepare(void) 49int ftrace_arch_code_modify_prepare(void)
34{ 50{
35 set_kernel_text_rw(); 51 set_kernel_text_rw();
52 modifying_code = 1;
36 return 0; 53 return 0;
37} 54}
38 55
39int ftrace_arch_code_modify_post_process(void) 56int ftrace_arch_code_modify_post_process(void)
40{ 57{
58 modifying_code = 0;
41 set_kernel_text_ro(); 59 set_kernel_text_ro();
42 return 0; 60 return 0;
43} 61}
@@ -149,6 +167,11 @@ static void ftrace_mod_code(void)
149 167
150void ftrace_nmi_enter(void) 168void ftrace_nmi_enter(void)
151{ 169{
170 __get_cpu_var(save_modifying_code) = modifying_code;
171
172 if (!__get_cpu_var(save_modifying_code))
173 return;
174
152 if (atomic_inc_return(&nmi_running) & MOD_CODE_WRITE_FLAG) { 175 if (atomic_inc_return(&nmi_running) & MOD_CODE_WRITE_FLAG) {
153 smp_rmb(); 176 smp_rmb();
154 ftrace_mod_code(); 177 ftrace_mod_code();
@@ -160,6 +183,9 @@ void ftrace_nmi_enter(void)
160 183
161void ftrace_nmi_exit(void) 184void ftrace_nmi_exit(void)
162{ 185{
186 if (!__get_cpu_var(save_modifying_code))
187 return;
188
163 /* Finish all executions before clearing nmi_running */ 189 /* Finish all executions before clearing nmi_running */
164 smp_mb(); 190 smp_mb();
165 atomic_dec(&nmi_running); 191 atomic_dec(&nmi_running);
@@ -484,13 +510,3 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr,
484 } 510 }
485} 511}
486#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ 512#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
487
488#ifdef CONFIG_FTRACE_SYSCALLS
489
490extern unsigned long *sys_call_table;
491
492unsigned long __init arch_syscall_addr(int nr)
493{
494 return (unsigned long)(&sys_call_table)[nr];
495}
496#endif
diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
index 0b4f97d24d7f..1cbb36f2759c 100644
--- a/include/linux/ftrace.h
+++ b/include/linux/ftrace.h
@@ -511,4 +511,10 @@ static inline void trace_hw_branch_oops(void) {}
511 511
512#endif /* CONFIG_HW_BRANCH_TRACER */ 512#endif /* CONFIG_HW_BRANCH_TRACER */
513 513
514#ifdef CONFIG_FTRACE_SYSCALLS
515
516unsigned long arch_syscall_addr(int nr);
517
518#endif /* CONFIG_FTRACE_SYSCALLS */
519
514#endif /* _LINUX_FTRACE_H */ 520#endif /* _LINUX_FTRACE_H */
diff --git a/include/linux/ftrace_event.h b/include/linux/ftrace_event.h
index 2233c98d80df..84a5629adfd8 100644
--- a/include/linux/ftrace_event.h
+++ b/include/linux/ftrace_event.h
@@ -121,9 +121,8 @@ struct ftrace_event_call {
121 int (*regfunc)(struct ftrace_event_call *); 121 int (*regfunc)(struct ftrace_event_call *);
122 void (*unregfunc)(struct ftrace_event_call *); 122 void (*unregfunc)(struct ftrace_event_call *);
123 int id; 123 int id;
124 const char *print_fmt;
124 int (*raw_init)(struct ftrace_event_call *); 125 int (*raw_init)(struct ftrace_event_call *);
125 int (*show_format)(struct ftrace_event_call *,
126 struct trace_seq *);
127 int (*define_fields)(struct ftrace_event_call *); 126 int (*define_fields)(struct ftrace_event_call *);
128 struct list_head fields; 127 struct list_head fields;
129 int filter_active; 128 int filter_active;
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index 207466a49f3d..91bd7d78a07d 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -132,7 +132,8 @@ struct perf_event_attr;
132 132
133#define SYSCALL_TRACE_ENTER_EVENT(sname) \ 133#define SYSCALL_TRACE_ENTER_EVENT(sname) \
134 static const struct syscall_metadata __syscall_meta_##sname; \ 134 static const struct syscall_metadata __syscall_meta_##sname; \
135 static struct ftrace_event_call event_enter_##sname; \ 135 static struct ftrace_event_call \
136 __attribute__((__aligned__(4))) event_enter_##sname; \
136 static struct trace_event enter_syscall_print_##sname = { \ 137 static struct trace_event enter_syscall_print_##sname = { \
137 .trace = print_syscall_enter, \ 138 .trace = print_syscall_enter, \
138 }; \ 139 }; \
@@ -143,8 +144,7 @@ struct perf_event_attr;
143 .name = "sys_enter"#sname, \ 144 .name = "sys_enter"#sname, \
144 .system = "syscalls", \ 145 .system = "syscalls", \
145 .event = &enter_syscall_print_##sname, \ 146 .event = &enter_syscall_print_##sname, \
146 .raw_init = trace_event_raw_init, \ 147 .raw_init = init_syscall_trace, \
147 .show_format = syscall_enter_format, \
148 .define_fields = syscall_enter_define_fields, \ 148 .define_fields = syscall_enter_define_fields, \
149 .regfunc = reg_event_syscall_enter, \ 149 .regfunc = reg_event_syscall_enter, \
150 .unregfunc = unreg_event_syscall_enter, \ 150 .unregfunc = unreg_event_syscall_enter, \
@@ -154,7 +154,8 @@ struct perf_event_attr;
154 154
155#define SYSCALL_TRACE_EXIT_EVENT(sname) \ 155#define SYSCALL_TRACE_EXIT_EVENT(sname) \
156 static const struct syscall_metadata __syscall_meta_##sname; \ 156 static const struct syscall_metadata __syscall_meta_##sname; \
157 static struct ftrace_event_call event_exit_##sname; \ 157 static struct ftrace_event_call \
158 __attribute__((__aligned__(4))) event_exit_##sname; \
158 static struct trace_event exit_syscall_print_##sname = { \ 159 static struct trace_event exit_syscall_print_##sname = { \
159 .trace = print_syscall_exit, \ 160 .trace = print_syscall_exit, \
160 }; \ 161 }; \
@@ -165,8 +166,7 @@ struct perf_event_attr;
165 .name = "sys_exit"#sname, \ 166 .name = "sys_exit"#sname, \
166 .system = "syscalls", \ 167 .system = "syscalls", \
167 .event = &exit_syscall_print_##sname, \ 168 .event = &exit_syscall_print_##sname, \
168 .raw_init = trace_event_raw_init, \ 169 .raw_init = init_syscall_trace, \
169 .show_format = syscall_exit_format, \
170 .define_fields = syscall_exit_define_fields, \ 170 .define_fields = syscall_exit_define_fields, \
171 .regfunc = reg_event_syscall_exit, \ 171 .regfunc = reg_event_syscall_exit, \
172 .unregfunc = unreg_event_syscall_exit, \ 172 .unregfunc = unreg_event_syscall_exit, \
diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h
index c6fe03e902ca..f23a0ca6910a 100644
--- a/include/trace/ftrace.h
+++ b/include/trace/ftrace.h
@@ -65,7 +65,8 @@
65 }; 65 };
66#undef DEFINE_EVENT 66#undef DEFINE_EVENT
67#define DEFINE_EVENT(template, name, proto, args) \ 67#define DEFINE_EVENT(template, name, proto, args) \
68 static struct ftrace_event_call event_##name 68 static struct ftrace_event_call \
69 __attribute__((__aligned__(4))) event_##name
69 70
70#undef DEFINE_EVENT_PRINT 71#undef DEFINE_EVENT_PRINT
71#define DEFINE_EVENT_PRINT(template, name, proto, args, print) \ 72#define DEFINE_EVENT_PRINT(template, name, proto, args, print) \
@@ -131,130 +132,6 @@
131#include TRACE_INCLUDE(TRACE_INCLUDE_FILE) 132#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)
132 133
133/* 134/*
134 * Setup the showing format of trace point.
135 *
136 * int
137 * ftrace_format_##call(struct trace_seq *s)
138 * {
139 * struct ftrace_raw_##call field;
140 * int ret;
141 *
142 * ret = trace_seq_printf(s, #type " " #item ";"
143 * " offset:%u; size:%u;\n",
144 * offsetof(struct ftrace_raw_##call, item),
145 * sizeof(field.type));
146 *
147 * }
148 */
149
150#undef TP_STRUCT__entry
151#define TP_STRUCT__entry(args...) args
152
153#undef __field
154#define __field(type, item) \
155 ret = trace_seq_printf(s, "\tfield:" #type " " #item ";\t" \
156 "offset:%u;\tsize:%u;\tsigned:%u;\n", \
157 (unsigned int)offsetof(typeof(field), item), \
158 (unsigned int)sizeof(field.item), \
159 (unsigned int)is_signed_type(type)); \
160 if (!ret) \
161 return 0;
162
163#undef __field_ext
164#define __field_ext(type, item, filter_type) __field(type, item)
165
166#undef __array
167#define __array(type, item, len) \
168 ret = trace_seq_printf(s, "\tfield:" #type " " #item "[" #len "];\t" \
169 "offset:%u;\tsize:%u;\tsigned:%u;\n", \
170 (unsigned int)offsetof(typeof(field), item), \
171 (unsigned int)sizeof(field.item), \
172 (unsigned int)is_signed_type(type)); \
173 if (!ret) \
174 return 0;
175
176#undef __dynamic_array
177#define __dynamic_array(type, item, len) \
178 ret = trace_seq_printf(s, "\tfield:__data_loc " #type "[] " #item ";\t"\
179 "offset:%u;\tsize:%u;\tsigned:%u;\n", \
180 (unsigned int)offsetof(typeof(field), \
181 __data_loc_##item), \
182 (unsigned int)sizeof(field.__data_loc_##item), \
183 (unsigned int)is_signed_type(type)); \
184 if (!ret) \
185 return 0;
186
187#undef __string
188#define __string(item, src) __dynamic_array(char, item, -1)
189
190#undef __entry
191#define __entry REC
192
193#undef __print_symbolic
194#undef __get_dynamic_array
195#undef __get_str
196
197#undef TP_printk
198#define TP_printk(fmt, args...) "\"%s\", %s\n", fmt, __stringify(args)
199
200#undef TP_fast_assign
201#define TP_fast_assign(args...) args
202
203#undef TP_perf_assign
204#define TP_perf_assign(args...)
205
206#undef DECLARE_EVENT_CLASS
207#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, func, print) \
208static int \
209ftrace_format_setup_##call(struct ftrace_event_call *unused, \
210 struct trace_seq *s) \
211{ \
212 struct ftrace_raw_##call field __attribute__((unused)); \
213 int ret = 0; \
214 \
215 tstruct; \
216 \
217 return ret; \
218} \
219 \
220static int \
221ftrace_format_##call(struct ftrace_event_call *unused, \
222 struct trace_seq *s) \
223{ \
224 int ret = 0; \
225 \
226 ret = ftrace_format_setup_##call(unused, s); \
227 if (!ret) \
228 return ret; \
229 \
230 ret = trace_seq_printf(s, "\nprint fmt: " print); \
231 \
232 return ret; \
233}
234
235#undef DEFINE_EVENT
236#define DEFINE_EVENT(template, name, proto, args)
237
238#undef DEFINE_EVENT_PRINT
239#define DEFINE_EVENT_PRINT(template, name, proto, args, print) \
240static int \
241ftrace_format_##name(struct ftrace_event_call *unused, \
242 struct trace_seq *s) \
243{ \
244 int ret = 0; \
245 \
246 ret = ftrace_format_setup_##template(unused, s); \
247 if (!ret) \
248 return ret; \
249 \
250 trace_seq_printf(s, "\nprint fmt: " print); \
251 \
252 return ret; \
253}
254
255#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)
256
257/*
258 * Stage 3 of the trace events. 135 * Stage 3 of the trace events.
259 * 136 *
260 * Override the macros in <trace/trace_events.h> to include the following: 137 * Override the macros in <trace/trace_events.h> to include the following:
@@ -323,7 +200,7 @@ ftrace_format_##name(struct ftrace_event_call *unused, \
323 200
324#undef DECLARE_EVENT_CLASS 201#undef DECLARE_EVENT_CLASS
325#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \ 202#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \
326static enum print_line_t \ 203static notrace enum print_line_t \
327ftrace_raw_output_id_##call(int event_id, const char *name, \ 204ftrace_raw_output_id_##call(int event_id, const char *name, \
328 struct trace_iterator *iter, int flags) \ 205 struct trace_iterator *iter, int flags) \
329{ \ 206{ \
@@ -356,7 +233,7 @@ ftrace_raw_output_id_##call(int event_id, const char *name, \
356 233
357#undef DEFINE_EVENT 234#undef DEFINE_EVENT
358#define DEFINE_EVENT(template, name, proto, args) \ 235#define DEFINE_EVENT(template, name, proto, args) \
359static enum print_line_t \ 236static notrace enum print_line_t \
360ftrace_raw_output_##name(struct trace_iterator *iter, int flags) \ 237ftrace_raw_output_##name(struct trace_iterator *iter, int flags) \
361{ \ 238{ \
362 return ftrace_raw_output_id_##template(event_##name.id, \ 239 return ftrace_raw_output_id_##template(event_##name.id, \
@@ -365,7 +242,7 @@ ftrace_raw_output_##name(struct trace_iterator *iter, int flags) \
365 242
366#undef DEFINE_EVENT_PRINT 243#undef DEFINE_EVENT_PRINT
367#define DEFINE_EVENT_PRINT(template, call, proto, args, print) \ 244#define DEFINE_EVENT_PRINT(template, call, proto, args, print) \
368static enum print_line_t \ 245static notrace enum print_line_t \
369ftrace_raw_output_##call(struct trace_iterator *iter, int flags) \ 246ftrace_raw_output_##call(struct trace_iterator *iter, int flags) \
370{ \ 247{ \
371 struct trace_seq *s = &iter->seq; \ 248 struct trace_seq *s = &iter->seq; \
@@ -431,7 +308,7 @@ ftrace_raw_output_##call(struct trace_iterator *iter, int flags) \
431 308
432#undef DECLARE_EVENT_CLASS 309#undef DECLARE_EVENT_CLASS
433#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, func, print) \ 310#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, func, print) \
434static int \ 311static int notrace \
435ftrace_define_fields_##call(struct ftrace_event_call *event_call) \ 312ftrace_define_fields_##call(struct ftrace_event_call *event_call) \
436{ \ 313{ \
437 struct ftrace_raw_##call field; \ 314 struct ftrace_raw_##call field; \
@@ -479,7 +356,7 @@ ftrace_define_fields_##call(struct ftrace_event_call *event_call) \
479 356
480#undef DECLARE_EVENT_CLASS 357#undef DECLARE_EVENT_CLASS
481#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \ 358#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \
482static inline int ftrace_get_offsets_##call( \ 359static inline notrace int ftrace_get_offsets_##call( \
483 struct ftrace_data_offsets_##call *__data_offsets, proto) \ 360 struct ftrace_data_offsets_##call *__data_offsets, proto) \
484{ \ 361{ \
485 int __data_size = 0; \ 362 int __data_size = 0; \
@@ -526,12 +403,14 @@ static inline int ftrace_get_offsets_##call( \
526 \ 403 \
527static void ftrace_profile_##name(proto); \ 404static void ftrace_profile_##name(proto); \
528 \ 405 \
529static int ftrace_profile_enable_##name(struct ftrace_event_call *unused)\ 406static notrace int \
407ftrace_profile_enable_##name(struct ftrace_event_call *unused) \
530{ \ 408{ \
531 return register_trace_##name(ftrace_profile_##name); \ 409 return register_trace_##name(ftrace_profile_##name); \
532} \ 410} \
533 \ 411 \
534static void ftrace_profile_disable_##name(struct ftrace_event_call *unused)\ 412static notrace void \
413ftrace_profile_disable_##name(struct ftrace_event_call *unused) \
535{ \ 414{ \
536 unregister_trace_##name(ftrace_profile_##name); \ 415 unregister_trace_##name(ftrace_profile_##name); \
537} 416}
@@ -622,7 +501,6 @@ static void ftrace_profile_disable_##name(struct ftrace_event_call *unused)\
622 * .raw_init = trace_event_raw_init, 501 * .raw_init = trace_event_raw_init,
623 * .regfunc = ftrace_reg_event_<call>, 502 * .regfunc = ftrace_reg_event_<call>,
624 * .unregfunc = ftrace_unreg_event_<call>, 503 * .unregfunc = ftrace_unreg_event_<call>,
625 * .show_format = ftrace_format_<call>,
626 * } 504 * }
627 * 505 *
628 */ 506 */
@@ -657,10 +535,17 @@ static void ftrace_profile_disable_##name(struct ftrace_event_call *unused)\
657#define __assign_str(dst, src) \ 535#define __assign_str(dst, src) \
658 strcpy(__get_str(dst), src); 536 strcpy(__get_str(dst), src);
659 537
538#undef TP_fast_assign
539#define TP_fast_assign(args...) args
540
541#undef TP_perf_assign
542#define TP_perf_assign(args...)
543
660#undef DECLARE_EVENT_CLASS 544#undef DECLARE_EVENT_CLASS
661#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \ 545#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \
662 \ 546 \
663static void ftrace_raw_event_id_##call(struct ftrace_event_call *event_call, \ 547static notrace void \
548ftrace_raw_event_id_##call(struct ftrace_event_call *event_call, \
664 proto) \ 549 proto) \
665{ \ 550{ \
666 struct ftrace_data_offsets_##call __maybe_unused __data_offsets;\ 551 struct ftrace_data_offsets_##call __maybe_unused __data_offsets;\
@@ -697,17 +582,19 @@ static void ftrace_raw_event_id_##call(struct ftrace_event_call *event_call, \
697#undef DEFINE_EVENT 582#undef DEFINE_EVENT
698#define DEFINE_EVENT(template, call, proto, args) \ 583#define DEFINE_EVENT(template, call, proto, args) \
699 \ 584 \
700static void ftrace_raw_event_##call(proto) \ 585static notrace void ftrace_raw_event_##call(proto) \
701{ \ 586{ \
702 ftrace_raw_event_id_##template(&event_##call, args); \ 587 ftrace_raw_event_id_##template(&event_##call, args); \
703} \ 588} \
704 \ 589 \
705static int ftrace_raw_reg_event_##call(struct ftrace_event_call *unused)\ 590static notrace int \
591ftrace_raw_reg_event_##call(struct ftrace_event_call *unused) \
706{ \ 592{ \
707 return register_trace_##call(ftrace_raw_event_##call); \ 593 return register_trace_##call(ftrace_raw_event_##call); \
708} \ 594} \
709 \ 595 \
710static void ftrace_raw_unreg_event_##call(struct ftrace_event_call *unused)\ 596static notrace void \
597ftrace_raw_unreg_event_##call(struct ftrace_event_call *unused) \
711{ \ 598{ \
712 unregister_trace_##call(ftrace_raw_event_##call); \ 599 unregister_trace_##call(ftrace_raw_event_##call); \
713} \ 600} \
@@ -722,8 +609,20 @@ static struct trace_event ftrace_event_type_##call = { \
722 609
723#include TRACE_INCLUDE(TRACE_INCLUDE_FILE) 610#include TRACE_INCLUDE(TRACE_INCLUDE_FILE)
724 611
612#undef __entry
613#define __entry REC
614
615#undef __print_flags
616#undef __print_symbolic
617#undef __get_dynamic_array
618#undef __get_str
619
620#undef TP_printk
621#define TP_printk(fmt, args...) "\"" fmt "\", " __stringify(args)
622
725#undef DECLARE_EVENT_CLASS 623#undef DECLARE_EVENT_CLASS
726#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) 624#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \
625static const char print_fmt_##call[] = print;
727 626
728#undef DEFINE_EVENT 627#undef DEFINE_EVENT
729#define DEFINE_EVENT(template, call, proto, args) \ 628#define DEFINE_EVENT(template, call, proto, args) \
@@ -737,7 +636,7 @@ __attribute__((section("_ftrace_events"))) event_##call = { \
737 .raw_init = trace_event_raw_init, \ 636 .raw_init = trace_event_raw_init, \
738 .regfunc = ftrace_raw_reg_event_##call, \ 637 .regfunc = ftrace_raw_reg_event_##call, \
739 .unregfunc = ftrace_raw_unreg_event_##call, \ 638 .unregfunc = ftrace_raw_unreg_event_##call, \
740 .show_format = ftrace_format_##template, \ 639 .print_fmt = print_fmt_##template, \
741 .define_fields = ftrace_define_fields_##template, \ 640 .define_fields = ftrace_define_fields_##template, \
742 _TRACE_PROFILE_INIT(call) \ 641 _TRACE_PROFILE_INIT(call) \
743} 642}
@@ -745,6 +644,8 @@ __attribute__((section("_ftrace_events"))) event_##call = { \
745#undef DEFINE_EVENT_PRINT 644#undef DEFINE_EVENT_PRINT
746#define DEFINE_EVENT_PRINT(template, call, proto, args, print) \ 645#define DEFINE_EVENT_PRINT(template, call, proto, args, print) \
747 \ 646 \
647static const char print_fmt_##call[] = print; \
648 \
748static struct ftrace_event_call __used \ 649static struct ftrace_event_call __used \
749__attribute__((__aligned__(4))) \ 650__attribute__((__aligned__(4))) \
750__attribute__((section("_ftrace_events"))) event_##call = { \ 651__attribute__((section("_ftrace_events"))) event_##call = { \
@@ -754,7 +655,7 @@ __attribute__((section("_ftrace_events"))) event_##call = { \
754 .raw_init = trace_event_raw_init, \ 655 .raw_init = trace_event_raw_init, \
755 .regfunc = ftrace_raw_reg_event_##call, \ 656 .regfunc = ftrace_raw_reg_event_##call, \
756 .unregfunc = ftrace_raw_unreg_event_##call, \ 657 .unregfunc = ftrace_raw_unreg_event_##call, \
757 .show_format = ftrace_format_##call, \ 658 .print_fmt = print_fmt_##call, \
758 .define_fields = ftrace_define_fields_##template, \ 659 .define_fields = ftrace_define_fields_##template, \
759 _TRACE_PROFILE_INIT(call) \ 660 _TRACE_PROFILE_INIT(call) \
760} 661}
@@ -837,6 +738,16 @@ __attribute__((section("_ftrace_events"))) event_##call = { \
837 738
838#ifdef CONFIG_EVENT_PROFILE 739#ifdef CONFIG_EVENT_PROFILE
839 740
741#undef __entry
742#define __entry entry
743
744#undef __get_dynamic_array
745#define __get_dynamic_array(field) \
746 ((void *)__entry + (__entry->__data_loc_##field & 0xffff))
747
748#undef __get_str
749#define __get_str(field) (char *)__get_dynamic_array(field)
750
840#undef __perf_addr 751#undef __perf_addr
841#define __perf_addr(a) __addr = (a) 752#define __perf_addr(a) __addr = (a)
842 753
@@ -845,7 +756,7 @@ __attribute__((section("_ftrace_events"))) event_##call = { \
845 756
846#undef DECLARE_EVENT_CLASS 757#undef DECLARE_EVENT_CLASS
847#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \ 758#define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \
848static void \ 759static notrace void \
849ftrace_profile_templ_##call(struct ftrace_event_call *event_call, \ 760ftrace_profile_templ_##call(struct ftrace_event_call *event_call, \
850 proto) \ 761 proto) \
851{ \ 762{ \
@@ -915,7 +826,7 @@ end_recursion: \
915 826
916#undef DEFINE_EVENT 827#undef DEFINE_EVENT
917#define DEFINE_EVENT(template, call, proto, args) \ 828#define DEFINE_EVENT(template, call, proto, args) \
918static void ftrace_profile_##call(proto) \ 829static notrace void ftrace_profile_##call(proto) \
919{ \ 830{ \
920 struct ftrace_event_call *event_call = &event_##call; \ 831 struct ftrace_event_call *event_call = &event_##call; \
921 \ 832 \
diff --git a/include/trace/syscall.h b/include/trace/syscall.h
index 961fda3556bb..8cd410254456 100644
--- a/include/trace/syscall.h
+++ b/include/trace/syscall.h
@@ -34,10 +34,6 @@ struct syscall_metadata {
34extern unsigned long arch_syscall_addr(int nr); 34extern unsigned long arch_syscall_addr(int nr);
35extern int init_syscall_trace(struct ftrace_event_call *call); 35extern int init_syscall_trace(struct ftrace_event_call *call);
36 36
37extern int syscall_enter_format(struct ftrace_event_call *call,
38 struct trace_seq *s);
39extern int syscall_exit_format(struct ftrace_event_call *call,
40 struct trace_seq *s);
41extern int syscall_enter_define_fields(struct ftrace_event_call *call); 37extern int syscall_enter_define_fields(struct ftrace_event_call *call);
42extern int syscall_exit_define_fields(struct ftrace_event_call *call); 38extern int syscall_exit_define_fields(struct ftrace_event_call *call);
43extern int reg_event_syscall_enter(struct ftrace_event_call *call); 39extern int reg_event_syscall_enter(struct ftrace_event_call *call);
diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
index 60e2ce0181ee..13e13d428cd3 100644
--- a/kernel/trace/Kconfig
+++ b/kernel/trace/Kconfig
@@ -328,15 +328,6 @@ config BRANCH_TRACER
328 328
329 Say N if unsure. 329 Say N if unsure.
330 330
331config POWER_TRACER
332 bool "Trace power consumption behavior"
333 depends on X86
334 select GENERIC_TRACER
335 help
336 This tracer helps developers to analyze and optimize the kernel's
337 power management decisions, specifically the C-state and P-state
338 behavior.
339
340config KSYM_TRACER 331config KSYM_TRACER
341 bool "Trace read and write access on kernel memory locations" 332 bool "Trace read and write access on kernel memory locations"
342 depends on HAVE_HW_BREAKPOINT 333 depends on HAVE_HW_BREAKPOINT
@@ -449,7 +440,7 @@ config BLK_DEV_IO_TRACE
449 440
450config KPROBE_EVENT 441config KPROBE_EVENT
451 depends on KPROBES 442 depends on KPROBES
452 depends on X86 443 depends on HAVE_REGS_AND_STACK_ACCESS_API
453 bool "Enable kprobes-based dynamic events" 444 bool "Enable kprobes-based dynamic events"
454 select TRACING 445 select TRACING
455 default y 446 default y
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index 1e6640f80454..d996353473fd 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -2426,6 +2426,7 @@ static const struct file_operations ftrace_notrace_fops = {
2426static DEFINE_MUTEX(graph_lock); 2426static DEFINE_MUTEX(graph_lock);
2427 2427
2428int ftrace_graph_count; 2428int ftrace_graph_count;
2429int ftrace_graph_filter_enabled;
2429unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS] __read_mostly; 2430unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS] __read_mostly;
2430 2431
2431static void * 2432static void *
@@ -2448,7 +2449,7 @@ static void *g_start(struct seq_file *m, loff_t *pos)
2448 mutex_lock(&graph_lock); 2449 mutex_lock(&graph_lock);
2449 2450
2450 /* Nothing, tell g_show to print all functions are enabled */ 2451 /* Nothing, tell g_show to print all functions are enabled */
2451 if (!ftrace_graph_count && !*pos) 2452 if (!ftrace_graph_filter_enabled && !*pos)
2452 return (void *)1; 2453 return (void *)1;
2453 2454
2454 return __g_next(m, pos); 2455 return __g_next(m, pos);
@@ -2494,6 +2495,7 @@ ftrace_graph_open(struct inode *inode, struct file *file)
2494 mutex_lock(&graph_lock); 2495 mutex_lock(&graph_lock);
2495 if ((file->f_mode & FMODE_WRITE) && 2496 if ((file->f_mode & FMODE_WRITE) &&
2496 (file->f_flags & O_TRUNC)) { 2497 (file->f_flags & O_TRUNC)) {
2498 ftrace_graph_filter_enabled = 0;
2497 ftrace_graph_count = 0; 2499 ftrace_graph_count = 0;
2498 memset(ftrace_graph_funcs, 0, sizeof(ftrace_graph_funcs)); 2500 memset(ftrace_graph_funcs, 0, sizeof(ftrace_graph_funcs));
2499 } 2501 }
@@ -2519,7 +2521,7 @@ ftrace_set_func(unsigned long *array, int *idx, char *buffer)
2519 struct dyn_ftrace *rec; 2521 struct dyn_ftrace *rec;
2520 struct ftrace_page *pg; 2522 struct ftrace_page *pg;
2521 int search_len; 2523 int search_len;
2522 int found = 0; 2524 int fail = 1;
2523 int type, not; 2525 int type, not;
2524 char *search; 2526 char *search;
2525 bool exists; 2527 bool exists;
@@ -2530,37 +2532,51 @@ ftrace_set_func(unsigned long *array, int *idx, char *buffer)
2530 2532
2531 /* decode regex */ 2533 /* decode regex */
2532 type = filter_parse_regex(buffer, strlen(buffer), &search, &not); 2534 type = filter_parse_regex(buffer, strlen(buffer), &search, &not);
2533 if (not) 2535 if (!not && *idx >= FTRACE_GRAPH_MAX_FUNCS)
2534 return -EINVAL; 2536 return -EBUSY;
2535 2537
2536 search_len = strlen(search); 2538 search_len = strlen(search);
2537 2539
2538 mutex_lock(&ftrace_lock); 2540 mutex_lock(&ftrace_lock);
2539 do_for_each_ftrace_rec(pg, rec) { 2541 do_for_each_ftrace_rec(pg, rec) {
2540 2542
2541 if (*idx >= FTRACE_GRAPH_MAX_FUNCS)
2542 break;
2543
2544 if (rec->flags & (FTRACE_FL_FAILED | FTRACE_FL_FREE)) 2543 if (rec->flags & (FTRACE_FL_FAILED | FTRACE_FL_FREE))
2545 continue; 2544 continue;
2546 2545
2547 if (ftrace_match_record(rec, search, search_len, type)) { 2546 if (ftrace_match_record(rec, search, search_len, type)) {
2548 /* ensure it is not already in the array */ 2547 /* if it is in the array */
2549 exists = false; 2548 exists = false;
2550 for (i = 0; i < *idx; i++) 2549 for (i = 0; i < *idx; i++) {
2551 if (array[i] == rec->ip) { 2550 if (array[i] == rec->ip) {
2552 exists = true; 2551 exists = true;
2553 break; 2552 break;
2554 } 2553 }
2555 if (!exists) 2554 }
2556 array[(*idx)++] = rec->ip; 2555
2557 found = 1; 2556 if (!not) {
2557 fail = 0;
2558 if (!exists) {
2559 array[(*idx)++] = rec->ip;
2560 if (*idx >= FTRACE_GRAPH_MAX_FUNCS)
2561 goto out;
2562 }
2563 } else {
2564 if (exists) {
2565 array[i] = array[--(*idx)];
2566 array[*idx] = 0;
2567 fail = 0;
2568 }
2569 }
2558 } 2570 }
2559 } while_for_each_ftrace_rec(); 2571 } while_for_each_ftrace_rec();
2560 2572out:
2561 mutex_unlock(&ftrace_lock); 2573 mutex_unlock(&ftrace_lock);
2562 2574
2563 return found ? 0 : -EINVAL; 2575 if (fail)
2576 return -EINVAL;
2577
2578 ftrace_graph_filter_enabled = 1;
2579 return 0;
2564} 2580}
2565 2581
2566static ssize_t 2582static ssize_t
@@ -2570,16 +2586,11 @@ ftrace_graph_write(struct file *file, const char __user *ubuf,
2570 struct trace_parser parser; 2586 struct trace_parser parser;
2571 ssize_t read, ret; 2587 ssize_t read, ret;
2572 2588
2573 if (!cnt || cnt < 0) 2589 if (!cnt)
2574 return 0; 2590 return 0;
2575 2591
2576 mutex_lock(&graph_lock); 2592 mutex_lock(&graph_lock);
2577 2593
2578 if (ftrace_graph_count >= FTRACE_GRAPH_MAX_FUNCS) {
2579 ret = -EBUSY;
2580 goto out_unlock;
2581 }
2582
2583 if (trace_parser_get_init(&parser, FTRACE_BUFF_MAX)) { 2594 if (trace_parser_get_init(&parser, FTRACE_BUFF_MAX)) {
2584 ret = -ENOMEM; 2595 ret = -ENOMEM;
2585 goto out_unlock; 2596 goto out_unlock;
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index eac6875cb990..032c57ca6502 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -32,6 +32,7 @@
32#include <linux/splice.h> 32#include <linux/splice.h>
33#include <linux/kdebug.h> 33#include <linux/kdebug.h>
34#include <linux/string.h> 34#include <linux/string.h>
35#include <linux/rwsem.h>
35#include <linux/ctype.h> 36#include <linux/ctype.h>
36#include <linux/init.h> 37#include <linux/init.h>
37#include <linux/poll.h> 38#include <linux/poll.h>
@@ -102,9 +103,6 @@ static inline void ftrace_enable_cpu(void)
102 103
103static cpumask_var_t __read_mostly tracing_buffer_mask; 104static cpumask_var_t __read_mostly tracing_buffer_mask;
104 105
105/* Define which cpu buffers are currently read in trace_pipe */
106static cpumask_var_t tracing_reader_cpumask;
107
108#define for_each_tracing_cpu(cpu) \ 106#define for_each_tracing_cpu(cpu) \
109 for_each_cpu(cpu, tracing_buffer_mask) 107 for_each_cpu(cpu, tracing_buffer_mask)
110 108
@@ -243,12 +241,91 @@ static struct tracer *current_trace __read_mostly;
243 241
244/* 242/*
245 * trace_types_lock is used to protect the trace_types list. 243 * trace_types_lock is used to protect the trace_types list.
246 * This lock is also used to keep user access serialized.
247 * Accesses from userspace will grab this lock while userspace
248 * activities happen inside the kernel.
249 */ 244 */
250static DEFINE_MUTEX(trace_types_lock); 245static DEFINE_MUTEX(trace_types_lock);
251 246
247/*
248 * serialize the access of the ring buffer
249 *
250 * ring buffer serializes readers, but it is low level protection.
251 * The validity of the events (which returns by ring_buffer_peek() ..etc)
252 * are not protected by ring buffer.
253 *
254 * The content of events may become garbage if we allow other process consumes
255 * these events concurrently:
256 * A) the page of the consumed events may become a normal page
257 * (not reader page) in ring buffer, and this page will be rewrited
258 * by events producer.
259 * B) The page of the consumed events may become a page for splice_read,
260 * and this page will be returned to system.
261 *
262 * These primitives allow multi process access to different cpu ring buffer
263 * concurrently.
264 *
265 * These primitives don't distinguish read-only and read-consume access.
266 * Multi read-only access are also serialized.
267 */
268
269#ifdef CONFIG_SMP
270static DECLARE_RWSEM(all_cpu_access_lock);
271static DEFINE_PER_CPU(struct mutex, cpu_access_lock);
272
273static inline void trace_access_lock(int cpu)
274{
275 if (cpu == TRACE_PIPE_ALL_CPU) {
276 /* gain it for accessing the whole ring buffer. */
277 down_write(&all_cpu_access_lock);
278 } else {
279 /* gain it for accessing a cpu ring buffer. */
280
281 /* Firstly block other trace_access_lock(TRACE_PIPE_ALL_CPU). */
282 down_read(&all_cpu_access_lock);
283
284 /* Secondly block other access to this @cpu ring buffer. */
285 mutex_lock(&per_cpu(cpu_access_lock, cpu));
286 }
287}
288
289static inline void trace_access_unlock(int cpu)
290{
291 if (cpu == TRACE_PIPE_ALL_CPU) {
292 up_write(&all_cpu_access_lock);
293 } else {
294 mutex_unlock(&per_cpu(cpu_access_lock, cpu));
295 up_read(&all_cpu_access_lock);
296 }
297}
298
299static inline void trace_access_lock_init(void)
300{
301 int cpu;
302
303 for_each_possible_cpu(cpu)
304 mutex_init(&per_cpu(cpu_access_lock, cpu));
305}
306
307#else
308
309static DEFINE_MUTEX(access_lock);
310
311static inline void trace_access_lock(int cpu)
312{
313 (void)cpu;
314 mutex_lock(&access_lock);
315}
316
317static inline void trace_access_unlock(int cpu)
318{
319 (void)cpu;
320 mutex_unlock(&access_lock);
321}
322
323static inline void trace_access_lock_init(void)
324{
325}
326
327#endif
328
252/* trace_wait is a waitqueue for tasks blocked on trace_poll */ 329/* trace_wait is a waitqueue for tasks blocked on trace_poll */
253static DECLARE_WAIT_QUEUE_HEAD(trace_wait); 330static DECLARE_WAIT_QUEUE_HEAD(trace_wait);
254 331
@@ -1320,8 +1397,10 @@ int trace_vbprintk(unsigned long ip, const char *fmt, va_list args)
1320 entry->fmt = fmt; 1397 entry->fmt = fmt;
1321 1398
1322 memcpy(entry->buf, trace_buf, sizeof(u32) * len); 1399 memcpy(entry->buf, trace_buf, sizeof(u32) * len);
1323 if (!filter_check_discard(call, entry, buffer, event)) 1400 if (!filter_check_discard(call, entry, buffer, event)) {
1324 ring_buffer_unlock_commit(buffer, event); 1401 ring_buffer_unlock_commit(buffer, event);
1402 ftrace_trace_stack(buffer, flags, 6, pc);
1403 }
1325 1404
1326out_unlock: 1405out_unlock:
1327 arch_spin_unlock(&trace_buf_lock); 1406 arch_spin_unlock(&trace_buf_lock);
@@ -1394,8 +1473,10 @@ int trace_array_vprintk(struct trace_array *tr,
1394 1473
1395 memcpy(&entry->buf, trace_buf, len); 1474 memcpy(&entry->buf, trace_buf, len);
1396 entry->buf[len] = '\0'; 1475 entry->buf[len] = '\0';
1397 if (!filter_check_discard(call, entry, buffer, event)) 1476 if (!filter_check_discard(call, entry, buffer, event)) {
1398 ring_buffer_unlock_commit(buffer, event); 1477 ring_buffer_unlock_commit(buffer, event);
1478 ftrace_trace_stack(buffer, irq_flags, 6, pc);
1479 }
1399 1480
1400 out_unlock: 1481 out_unlock:
1401 arch_spin_unlock(&trace_buf_lock); 1482 arch_spin_unlock(&trace_buf_lock);
@@ -1585,12 +1666,6 @@ static void tracing_iter_reset(struct trace_iterator *iter, int cpu)
1585} 1666}
1586 1667
1587/* 1668/*
1588 * No necessary locking here. The worst thing which can
1589 * happen is loosing events consumed at the same time
1590 * by a trace_pipe reader.
1591 * Other than that, we don't risk to crash the ring buffer
1592 * because it serializes the readers.
1593 *
1594 * The current tracer is copied to avoid a global locking 1669 * The current tracer is copied to avoid a global locking
1595 * all around. 1670 * all around.
1596 */ 1671 */
@@ -1645,12 +1720,16 @@ static void *s_start(struct seq_file *m, loff_t *pos)
1645 } 1720 }
1646 1721
1647 trace_event_read_lock(); 1722 trace_event_read_lock();
1723 trace_access_lock(cpu_file);
1648 return p; 1724 return p;
1649} 1725}
1650 1726
1651static void s_stop(struct seq_file *m, void *p) 1727static void s_stop(struct seq_file *m, void *p)
1652{ 1728{
1729 struct trace_iterator *iter = m->private;
1730
1653 atomic_dec(&trace_record_cmdline_disabled); 1731 atomic_dec(&trace_record_cmdline_disabled);
1732 trace_access_unlock(iter->cpu_file);
1654 trace_event_read_unlock(); 1733 trace_event_read_unlock();
1655} 1734}
1656 1735
@@ -2841,22 +2920,6 @@ static int tracing_open_pipe(struct inode *inode, struct file *filp)
2841 2920
2842 mutex_lock(&trace_types_lock); 2921 mutex_lock(&trace_types_lock);
2843 2922
2844 /* We only allow one reader per cpu */
2845 if (cpu_file == TRACE_PIPE_ALL_CPU) {
2846 if (!cpumask_empty(tracing_reader_cpumask)) {
2847 ret = -EBUSY;
2848 goto out;
2849 }
2850 cpumask_setall(tracing_reader_cpumask);
2851 } else {
2852 if (!cpumask_test_cpu(cpu_file, tracing_reader_cpumask))
2853 cpumask_set_cpu(cpu_file, tracing_reader_cpumask);
2854 else {
2855 ret = -EBUSY;
2856 goto out;
2857 }
2858 }
2859
2860 /* create a buffer to store the information to pass to userspace */ 2923 /* create a buffer to store the information to pass to userspace */
2861 iter = kzalloc(sizeof(*iter), GFP_KERNEL); 2924 iter = kzalloc(sizeof(*iter), GFP_KERNEL);
2862 if (!iter) { 2925 if (!iter) {
@@ -2912,12 +2975,6 @@ static int tracing_release_pipe(struct inode *inode, struct file *file)
2912 2975
2913 mutex_lock(&trace_types_lock); 2976 mutex_lock(&trace_types_lock);
2914 2977
2915 if (iter->cpu_file == TRACE_PIPE_ALL_CPU)
2916 cpumask_clear(tracing_reader_cpumask);
2917 else
2918 cpumask_clear_cpu(iter->cpu_file, tracing_reader_cpumask);
2919
2920
2921 if (iter->trace->pipe_close) 2978 if (iter->trace->pipe_close)
2922 iter->trace->pipe_close(iter); 2979 iter->trace->pipe_close(iter);
2923 2980
@@ -3079,6 +3136,7 @@ waitagain:
3079 iter->pos = -1; 3136 iter->pos = -1;
3080 3137
3081 trace_event_read_lock(); 3138 trace_event_read_lock();
3139 trace_access_lock(iter->cpu_file);
3082 while (find_next_entry_inc(iter) != NULL) { 3140 while (find_next_entry_inc(iter) != NULL) {
3083 enum print_line_t ret; 3141 enum print_line_t ret;
3084 int len = iter->seq.len; 3142 int len = iter->seq.len;
@@ -3095,6 +3153,7 @@ waitagain:
3095 if (iter->seq.len >= cnt) 3153 if (iter->seq.len >= cnt)
3096 break; 3154 break;
3097 } 3155 }
3156 trace_access_unlock(iter->cpu_file);
3098 trace_event_read_unlock(); 3157 trace_event_read_unlock();
3099 3158
3100 /* Now copy what we have to the user */ 3159 /* Now copy what we have to the user */
@@ -3220,6 +3279,7 @@ static ssize_t tracing_splice_read_pipe(struct file *filp,
3220 } 3279 }
3221 3280
3222 trace_event_read_lock(); 3281 trace_event_read_lock();
3282 trace_access_lock(iter->cpu_file);
3223 3283
3224 /* Fill as many pages as possible. */ 3284 /* Fill as many pages as possible. */
3225 for (i = 0, rem = len; i < PIPE_BUFFERS && rem; i++) { 3285 for (i = 0, rem = len; i < PIPE_BUFFERS && rem; i++) {
@@ -3243,6 +3303,7 @@ static ssize_t tracing_splice_read_pipe(struct file *filp,
3243 trace_seq_init(&iter->seq); 3303 trace_seq_init(&iter->seq);
3244 } 3304 }
3245 3305
3306 trace_access_unlock(iter->cpu_file);
3246 trace_event_read_unlock(); 3307 trace_event_read_unlock();
3247 mutex_unlock(&iter->mutex); 3308 mutex_unlock(&iter->mutex);
3248 3309
@@ -3544,10 +3605,12 @@ tracing_buffers_read(struct file *filp, char __user *ubuf,
3544 3605
3545 info->read = 0; 3606 info->read = 0;
3546 3607
3608 trace_access_lock(info->cpu);
3547 ret = ring_buffer_read_page(info->tr->buffer, 3609 ret = ring_buffer_read_page(info->tr->buffer,
3548 &info->spare, 3610 &info->spare,
3549 count, 3611 count,
3550 info->cpu, 0); 3612 info->cpu, 0);
3613 trace_access_unlock(info->cpu);
3551 if (ret < 0) 3614 if (ret < 0)
3552 return 0; 3615 return 0;
3553 3616
@@ -3675,6 +3738,7 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos,
3675 len &= PAGE_MASK; 3738 len &= PAGE_MASK;
3676 } 3739 }
3677 3740
3741 trace_access_lock(info->cpu);
3678 entries = ring_buffer_entries_cpu(info->tr->buffer, info->cpu); 3742 entries = ring_buffer_entries_cpu(info->tr->buffer, info->cpu);
3679 3743
3680 for (i = 0; i < PIPE_BUFFERS && len && entries; i++, len -= PAGE_SIZE) { 3744 for (i = 0; i < PIPE_BUFFERS && len && entries; i++, len -= PAGE_SIZE) {
@@ -3722,6 +3786,7 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos,
3722 entries = ring_buffer_entries_cpu(info->tr->buffer, info->cpu); 3786 entries = ring_buffer_entries_cpu(info->tr->buffer, info->cpu);
3723 } 3787 }
3724 3788
3789 trace_access_unlock(info->cpu);
3725 spd.nr_pages = i; 3790 spd.nr_pages = i;
3726 3791
3727 /* did we read anything? */ 3792 /* did we read anything? */
@@ -4158,6 +4223,8 @@ static __init int tracer_init_debugfs(void)
4158 struct dentry *d_tracer; 4223 struct dentry *d_tracer;
4159 int cpu; 4224 int cpu;
4160 4225
4226 trace_access_lock_init();
4227
4161 d_tracer = tracing_init_dentry(); 4228 d_tracer = tracing_init_dentry();
4162 4229
4163 trace_create_file("tracing_enabled", 0644, d_tracer, 4230 trace_create_file("tracing_enabled", 0644, d_tracer,
@@ -4392,9 +4459,6 @@ __init static int tracer_alloc_buffers(void)
4392 if (!alloc_cpumask_var(&tracing_cpumask, GFP_KERNEL)) 4459 if (!alloc_cpumask_var(&tracing_cpumask, GFP_KERNEL))
4393 goto out_free_buffer_mask; 4460 goto out_free_buffer_mask;
4394 4461
4395 if (!zalloc_cpumask_var(&tracing_reader_cpumask, GFP_KERNEL))
4396 goto out_free_tracing_cpumask;
4397
4398 /* To save memory, keep the ring buffer size to its minimum */ 4462 /* To save memory, keep the ring buffer size to its minimum */
4399 if (ring_buffer_expanded) 4463 if (ring_buffer_expanded)
4400 ring_buf_size = trace_buf_size; 4464 ring_buf_size = trace_buf_size;
@@ -4452,8 +4516,6 @@ __init static int tracer_alloc_buffers(void)
4452 return 0; 4516 return 0;
4453 4517
4454out_free_cpumask: 4518out_free_cpumask:
4455 free_cpumask_var(tracing_reader_cpumask);
4456out_free_tracing_cpumask:
4457 free_cpumask_var(tracing_cpumask); 4519 free_cpumask_var(tracing_cpumask);
4458out_free_buffer_mask: 4520out_free_buffer_mask:
4459 free_cpumask_var(tracing_buffer_mask); 4521 free_cpumask_var(tracing_buffer_mask);
diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h
index 4df6a77eb196..fd05bcaf91b0 100644
--- a/kernel/trace/trace.h
+++ b/kernel/trace/trace.h
@@ -497,6 +497,7 @@ trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
497#ifdef CONFIG_DYNAMIC_FTRACE 497#ifdef CONFIG_DYNAMIC_FTRACE
498/* TODO: make this variable */ 498/* TODO: make this variable */
499#define FTRACE_GRAPH_MAX_FUNCS 32 499#define FTRACE_GRAPH_MAX_FUNCS 32
500extern int ftrace_graph_filter_enabled;
500extern int ftrace_graph_count; 501extern int ftrace_graph_count;
501extern unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS]; 502extern unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS];
502 503
@@ -504,7 +505,7 @@ static inline int ftrace_graph_addr(unsigned long addr)
504{ 505{
505 int i; 506 int i;
506 507
507 if (!ftrace_graph_count || test_tsk_trace_graph(current)) 508 if (!ftrace_graph_filter_enabled)
508 return 1; 509 return 1;
509 510
510 for (i = 0; i < ftrace_graph_count; i++) { 511 for (i = 0; i < ftrace_graph_count; i++) {
@@ -791,7 +792,8 @@ extern const char *__stop___trace_bprintk_fmt[];
791 792
792#undef FTRACE_ENTRY 793#undef FTRACE_ENTRY
793#define FTRACE_ENTRY(call, struct_name, id, tstruct, print) \ 794#define FTRACE_ENTRY(call, struct_name, id, tstruct, print) \
794 extern struct ftrace_event_call event_##call; 795 extern struct ftrace_event_call \
796 __attribute__((__aligned__(4))) event_##call;
795#undef FTRACE_ENTRY_DUP 797#undef FTRACE_ENTRY_DUP
796#define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print) \ 798#define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print) \
797 FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print)) 799 FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print))
diff --git a/kernel/trace/trace_branch.c b/kernel/trace/trace_branch.c
index 4a194f08f88c..b9bc4d470177 100644
--- a/kernel/trace/trace_branch.c
+++ b/kernel/trace/trace_branch.c
@@ -307,8 +307,23 @@ static int annotated_branch_stat_cmp(void *p1, void *p2)
307 return -1; 307 return -1;
308 if (percent_a > percent_b) 308 if (percent_a > percent_b)
309 return 1; 309 return 1;
310 else 310
311 return 0; 311 if (a->incorrect < b->incorrect)
312 return -1;
313 if (a->incorrect > b->incorrect)
314 return 1;
315
316 /*
317 * Since the above shows worse (incorrect) cases
318 * first, we continue that by showing best (correct)
319 * cases last.
320 */
321 if (a->correct > b->correct)
322 return -1;
323 if (a->correct < b->correct)
324 return 1;
325
326 return 0;
312} 327}
313 328
314static struct tracer_stat annotated_branch_stats = { 329static struct tracer_stat annotated_branch_stats = {
diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
index 189b09baf4fb..3f972ad98d04 100644
--- a/kernel/trace/trace_events.c
+++ b/kernel/trace/trace_events.c
@@ -60,10 +60,8 @@ int trace_define_field(struct ftrace_event_call *call, const char *type,
60 return 0; 60 return 0;
61 61
62err: 62err:
63 if (field) { 63 if (field)
64 kfree(field->name); 64 kfree(field->name);
65 kfree(field->type);
66 }
67 kfree(field); 65 kfree(field);
68 66
69 return -ENOMEM; 67 return -ENOMEM;
@@ -520,41 +518,16 @@ out:
520 return ret; 518 return ret;
521} 519}
522 520
523extern char *__bad_type_size(void);
524
525#undef FIELD
526#define FIELD(type, name) \
527 sizeof(type) != sizeof(field.name) ? __bad_type_size() : \
528 #type, "common_" #name, offsetof(typeof(field), name), \
529 sizeof(field.name), is_signed_type(type)
530
531static int trace_write_header(struct trace_seq *s)
532{
533 struct trace_entry field;
534
535 /* struct trace_entry */
536 return trace_seq_printf(s,
537 "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n"
538 "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n"
539 "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n"
540 "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n"
541 "\tfield:%s %s;\toffset:%zu;\tsize:%zu;\tsigned:%u;\n"
542 "\n",
543 FIELD(unsigned short, type),
544 FIELD(unsigned char, flags),
545 FIELD(unsigned char, preempt_count),
546 FIELD(int, pid),
547 FIELD(int, lock_depth));
548}
549
550static ssize_t 521static ssize_t
551event_format_read(struct file *filp, char __user *ubuf, size_t cnt, 522event_format_read(struct file *filp, char __user *ubuf, size_t cnt,
552 loff_t *ppos) 523 loff_t *ppos)
553{ 524{
554 struct ftrace_event_call *call = filp->private_data; 525 struct ftrace_event_call *call = filp->private_data;
526 struct ftrace_event_field *field;
555 struct trace_seq *s; 527 struct trace_seq *s;
528 int common_field_count = 5;
556 char *buf; 529 char *buf;
557 int r; 530 int r = 0;
558 531
559 if (*ppos) 532 if (*ppos)
560 return 0; 533 return 0;
@@ -565,14 +538,48 @@ event_format_read(struct file *filp, char __user *ubuf, size_t cnt,
565 538
566 trace_seq_init(s); 539 trace_seq_init(s);
567 540
568 /* If any of the first writes fail, so will the show_format. */
569
570 trace_seq_printf(s, "name: %s\n", call->name); 541 trace_seq_printf(s, "name: %s\n", call->name);
571 trace_seq_printf(s, "ID: %d\n", call->id); 542 trace_seq_printf(s, "ID: %d\n", call->id);
572 trace_seq_printf(s, "format:\n"); 543 trace_seq_printf(s, "format:\n");
573 trace_write_header(s);
574 544
575 r = call->show_format(call, s); 545 list_for_each_entry_reverse(field, &call->fields, link) {
546 /*
547 * Smartly shows the array type(except dynamic array).
548 * Normal:
549 * field:TYPE VAR
550 * If TYPE := TYPE[LEN], it is shown:
551 * field:TYPE VAR[LEN]
552 */
553 const char *array_descriptor = strchr(field->type, '[');
554
555 if (!strncmp(field->type, "__data_loc", 10))
556 array_descriptor = NULL;
557
558 if (!array_descriptor) {
559 r = trace_seq_printf(s, "\tfield:%s %s;\toffset:%u;"
560 "\tsize:%u;\tsigned:%d;\n",
561 field->type, field->name, field->offset,
562 field->size, !!field->is_signed);
563 } else {
564 r = trace_seq_printf(s, "\tfield:%.*s %s%s;\toffset:%u;"
565 "\tsize:%u;\tsigned:%d;\n",
566 (int)(array_descriptor - field->type),
567 field->type, field->name,
568 array_descriptor, field->offset,
569 field->size, !!field->is_signed);
570 }
571
572 if (--common_field_count == 0)
573 r = trace_seq_printf(s, "\n");
574
575 if (!r)
576 break;
577 }
578
579 if (r)
580 r = trace_seq_printf(s, "\nprint fmt: %s\n",
581 call->print_fmt);
582
576 if (!r) { 583 if (!r) {
577 /* 584 /*
578 * ug! The format output is bigger than a PAGE!! 585 * ug! The format output is bigger than a PAGE!!
@@ -948,10 +955,6 @@ event_create_dir(struct ftrace_event_call *call, struct dentry *d_events,
948 filter); 955 filter);
949 } 956 }
950 957
951 /* A trace may not want to export its format */
952 if (!call->show_format)
953 return 0;
954
955 trace_create_file("format", 0444, call->dir, call, 958 trace_create_file("format", 0444, call->dir, call,
956 format); 959 format);
957 960
diff --git a/kernel/trace/trace_export.c b/kernel/trace/trace_export.c
index d4fa5dc1ee4e..e091f64ba6ce 100644
--- a/kernel/trace/trace_export.c
+++ b/kernel/trace/trace_export.c
@@ -62,78 +62,6 @@ static void __always_unused ____ftrace_check_##name(void) \
62 62
63#include "trace_entries.h" 63#include "trace_entries.h"
64 64
65
66#undef __field
67#define __field(type, item) \
68 ret = trace_seq_printf(s, "\tfield:" #type " " #item ";\t" \
69 "offset:%zu;\tsize:%zu;\tsigned:%u;\n", \
70 offsetof(typeof(field), item), \
71 sizeof(field.item), is_signed_type(type)); \
72 if (!ret) \
73 return 0;
74
75#undef __field_desc
76#define __field_desc(type, container, item) \
77 ret = trace_seq_printf(s, "\tfield:" #type " " #item ";\t" \
78 "offset:%zu;\tsize:%zu;\tsigned:%u;\n", \
79 offsetof(typeof(field), container.item), \
80 sizeof(field.container.item), \
81 is_signed_type(type)); \
82 if (!ret) \
83 return 0;
84
85#undef __array
86#define __array(type, item, len) \
87 ret = trace_seq_printf(s, "\tfield:" #type " " #item "[" #len "];\t" \
88 "offset:%zu;\tsize:%zu;\tsigned:%u;\n", \
89 offsetof(typeof(field), item), \
90 sizeof(field.item), is_signed_type(type)); \
91 if (!ret) \
92 return 0;
93
94#undef __array_desc
95#define __array_desc(type, container, item, len) \
96 ret = trace_seq_printf(s, "\tfield:" #type " " #item "[" #len "];\t" \
97 "offset:%zu;\tsize:%zu;\tsigned:%u;\n", \
98 offsetof(typeof(field), container.item), \
99 sizeof(field.container.item), \
100 is_signed_type(type)); \
101 if (!ret) \
102 return 0;
103
104#undef __dynamic_array
105#define __dynamic_array(type, item) \
106 ret = trace_seq_printf(s, "\tfield:" #type " " #item ";\t" \
107 "offset:%zu;\tsize:0;\tsigned:%u;\n", \
108 offsetof(typeof(field), item), \
109 is_signed_type(type)); \
110 if (!ret) \
111 return 0;
112
113#undef F_printk
114#define F_printk(fmt, args...) "%s, %s\n", #fmt, __stringify(args)
115
116#undef __entry
117#define __entry REC
118
119#undef FTRACE_ENTRY
120#define FTRACE_ENTRY(name, struct_name, id, tstruct, print) \
121static int \
122ftrace_format_##name(struct ftrace_event_call *unused, \
123 struct trace_seq *s) \
124{ \
125 struct struct_name field __attribute__((unused)); \
126 int ret = 0; \
127 \
128 tstruct; \
129 \
130 trace_seq_printf(s, "\nprint fmt: " print); \
131 \
132 return ret; \
133}
134
135#include "trace_entries.h"
136
137#undef __field 65#undef __field
138#define __field(type, item) \ 66#define __field(type, item) \
139 ret = trace_define_field(event_call, #type, #item, \ 67 ret = trace_define_field(event_call, #type, #item, \
@@ -175,7 +103,12 @@ ftrace_format_##name(struct ftrace_event_call *unused, \
175 return ret; 103 return ret;
176 104
177#undef __dynamic_array 105#undef __dynamic_array
178#define __dynamic_array(type, item) 106#define __dynamic_array(type, item) \
107 ret = trace_define_field(event_call, #type, #item, \
108 offsetof(typeof(field), item), \
109 0, is_signed_type(type), FILTER_OTHER);\
110 if (ret) \
111 return ret;
179 112
180#undef FTRACE_ENTRY 113#undef FTRACE_ENTRY
181#define FTRACE_ENTRY(name, struct_name, id, tstruct, print) \ 114#define FTRACE_ENTRY(name, struct_name, id, tstruct, print) \
@@ -198,6 +131,9 @@ static int ftrace_raw_init_event(struct ftrace_event_call *call)
198 return 0; 131 return 0;
199} 132}
200 133
134#undef __entry
135#define __entry REC
136
201#undef __field 137#undef __field
202#define __field(type, item) 138#define __field(type, item)
203 139
@@ -213,6 +149,9 @@ static int ftrace_raw_init_event(struct ftrace_event_call *call)
213#undef __dynamic_array 149#undef __dynamic_array
214#define __dynamic_array(type, item) 150#define __dynamic_array(type, item)
215 151
152#undef F_printk
153#define F_printk(fmt, args...) #fmt ", " __stringify(args)
154
216#undef FTRACE_ENTRY 155#undef FTRACE_ENTRY
217#define FTRACE_ENTRY(call, struct_name, type, tstruct, print) \ 156#define FTRACE_ENTRY(call, struct_name, type, tstruct, print) \
218 \ 157 \
@@ -223,7 +162,7 @@ __attribute__((section("_ftrace_events"))) event_##call = { \
223 .id = type, \ 162 .id = type, \
224 .system = __stringify(TRACE_SYSTEM), \ 163 .system = __stringify(TRACE_SYSTEM), \
225 .raw_init = ftrace_raw_init_event, \ 164 .raw_init = ftrace_raw_init_event, \
226 .show_format = ftrace_format_##call, \ 165 .print_fmt = print, \
227 .define_fields = ftrace_define_fields_##call, \ 166 .define_fields = ftrace_define_fields_##call, \
228}; \ 167}; \
229 168
diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c
index b1342c5d37cf..e998a824e9db 100644
--- a/kernel/trace/trace_functions_graph.c
+++ b/kernel/trace/trace_functions_graph.c
@@ -18,6 +18,7 @@ struct fgraph_cpu_data {
18 pid_t last_pid; 18 pid_t last_pid;
19 int depth; 19 int depth;
20 int ignore; 20 int ignore;
21 unsigned long enter_funcs[FTRACE_RETFUNC_DEPTH];
21}; 22};
22 23
23struct fgraph_data { 24struct fgraph_data {
@@ -212,13 +213,11 @@ int trace_graph_entry(struct ftrace_graph_ent *trace)
212 int cpu; 213 int cpu;
213 int pc; 214 int pc;
214 215
215 if (unlikely(!tr))
216 return 0;
217
218 if (!ftrace_trace_task(current)) 216 if (!ftrace_trace_task(current))
219 return 0; 217 return 0;
220 218
221 if (!ftrace_graph_addr(trace->func)) 219 /* trace it when it is-nested-in or is a function enabled. */
220 if (!(trace->depth || ftrace_graph_addr(trace->func)))
222 return 0; 221 return 0;
223 222
224 local_irq_save(flags); 223 local_irq_save(flags);
@@ -231,9 +230,6 @@ int trace_graph_entry(struct ftrace_graph_ent *trace)
231 } else { 230 } else {
232 ret = 0; 231 ret = 0;
233 } 232 }
234 /* Only do the atomic if it is not already set */
235 if (!test_tsk_trace_graph(current))
236 set_tsk_trace_graph(current);
237 233
238 atomic_dec(&data->disabled); 234 atomic_dec(&data->disabled);
239 local_irq_restore(flags); 235 local_irq_restore(flags);
@@ -281,17 +277,24 @@ void trace_graph_return(struct ftrace_graph_ret *trace)
281 pc = preempt_count(); 277 pc = preempt_count();
282 __trace_graph_return(tr, trace, flags, pc); 278 __trace_graph_return(tr, trace, flags, pc);
283 } 279 }
284 if (!trace->depth)
285 clear_tsk_trace_graph(current);
286 atomic_dec(&data->disabled); 280 atomic_dec(&data->disabled);
287 local_irq_restore(flags); 281 local_irq_restore(flags);
288} 282}
289 283
284void set_graph_array(struct trace_array *tr)
285{
286 graph_array = tr;
287
288 /* Make graph_array visible before we start tracing */
289
290 smp_mb();
291}
292
290static int graph_trace_init(struct trace_array *tr) 293static int graph_trace_init(struct trace_array *tr)
291{ 294{
292 int ret; 295 int ret;
293 296
294 graph_array = tr; 297 set_graph_array(tr);
295 ret = register_ftrace_graph(&trace_graph_return, 298 ret = register_ftrace_graph(&trace_graph_return,
296 &trace_graph_entry); 299 &trace_graph_entry);
297 if (ret) 300 if (ret)
@@ -301,11 +304,6 @@ static int graph_trace_init(struct trace_array *tr)
301 return 0; 304 return 0;
302} 305}
303 306
304void set_graph_array(struct trace_array *tr)
305{
306 graph_array = tr;
307}
308
309static void graph_trace_reset(struct trace_array *tr) 307static void graph_trace_reset(struct trace_array *tr)
310{ 308{
311 tracing_stop_cmdline_record(); 309 tracing_stop_cmdline_record();
@@ -673,15 +671,21 @@ print_graph_entry_leaf(struct trace_iterator *iter,
673 duration = graph_ret->rettime - graph_ret->calltime; 671 duration = graph_ret->rettime - graph_ret->calltime;
674 672
675 if (data) { 673 if (data) {
674 struct fgraph_cpu_data *cpu_data;
676 int cpu = iter->cpu; 675 int cpu = iter->cpu;
677 int *depth = &(per_cpu_ptr(data->cpu_data, cpu)->depth); 676
677 cpu_data = per_cpu_ptr(data->cpu_data, cpu);
678 678
679 /* 679 /*
680 * Comments display at + 1 to depth. Since 680 * Comments display at + 1 to depth. Since
681 * this is a leaf function, keep the comments 681 * this is a leaf function, keep the comments
682 * equal to this depth. 682 * equal to this depth.
683 */ 683 */
684 *depth = call->depth - 1; 684 cpu_data->depth = call->depth - 1;
685
686 /* No need to keep this function around for this depth */
687 if (call->depth < FTRACE_RETFUNC_DEPTH)
688 cpu_data->enter_funcs[call->depth] = 0;
685 } 689 }
686 690
687 /* Overhead */ 691 /* Overhead */
@@ -721,10 +725,15 @@ print_graph_entry_nested(struct trace_iterator *iter,
721 int i; 725 int i;
722 726
723 if (data) { 727 if (data) {
728 struct fgraph_cpu_data *cpu_data;
724 int cpu = iter->cpu; 729 int cpu = iter->cpu;
725 int *depth = &(per_cpu_ptr(data->cpu_data, cpu)->depth);
726 730
727 *depth = call->depth; 731 cpu_data = per_cpu_ptr(data->cpu_data, cpu);
732 cpu_data->depth = call->depth;
733
734 /* Save this function pointer to see if the exit matches */
735 if (call->depth < FTRACE_RETFUNC_DEPTH)
736 cpu_data->enter_funcs[call->depth] = call->func;
728 } 737 }
729 738
730 /* No overhead */ 739 /* No overhead */
@@ -854,19 +863,28 @@ print_graph_return(struct ftrace_graph_ret *trace, struct trace_seq *s,
854 struct fgraph_data *data = iter->private; 863 struct fgraph_data *data = iter->private;
855 pid_t pid = ent->pid; 864 pid_t pid = ent->pid;
856 int cpu = iter->cpu; 865 int cpu = iter->cpu;
866 int func_match = 1;
857 int ret; 867 int ret;
858 int i; 868 int i;
859 869
860 if (data) { 870 if (data) {
871 struct fgraph_cpu_data *cpu_data;
861 int cpu = iter->cpu; 872 int cpu = iter->cpu;
862 int *depth = &(per_cpu_ptr(data->cpu_data, cpu)->depth); 873
874 cpu_data = per_cpu_ptr(data->cpu_data, cpu);
863 875
864 /* 876 /*
865 * Comments display at + 1 to depth. This is the 877 * Comments display at + 1 to depth. This is the
866 * return from a function, we now want the comments 878 * return from a function, we now want the comments
867 * to display at the same level of the bracket. 879 * to display at the same level of the bracket.
868 */ 880 */
869 *depth = trace->depth - 1; 881 cpu_data->depth = trace->depth - 1;
882
883 if (trace->depth < FTRACE_RETFUNC_DEPTH) {
884 if (cpu_data->enter_funcs[trace->depth] != trace->func)
885 func_match = 0;
886 cpu_data->enter_funcs[trace->depth] = 0;
887 }
870 } 888 }
871 889
872 if (print_graph_prologue(iter, s, 0, 0)) 890 if (print_graph_prologue(iter, s, 0, 0))
@@ -891,9 +909,21 @@ print_graph_return(struct ftrace_graph_ret *trace, struct trace_seq *s,
891 return TRACE_TYPE_PARTIAL_LINE; 909 return TRACE_TYPE_PARTIAL_LINE;
892 } 910 }
893 911
894 ret = trace_seq_printf(s, "}\n"); 912 /*
895 if (!ret) 913 * If the return function does not have a matching entry,
896 return TRACE_TYPE_PARTIAL_LINE; 914 * then the entry was lost. Instead of just printing
915 * the '}' and letting the user guess what function this
916 * belongs to, write out the function name.
917 */
918 if (func_match) {
919 ret = trace_seq_printf(s, "}\n");
920 if (!ret)
921 return TRACE_TYPE_PARTIAL_LINE;
922 } else {
923 ret = trace_seq_printf(s, "} (%ps)\n", (void *)trace->func);
924 if (!ret)
925 return TRACE_TYPE_PARTIAL_LINE;
926 }
897 927
898 /* Overrun */ 928 /* Overrun */
899 if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERRUN) { 929 if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERRUN) {
diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index 50b1b8239806..465b36bef4ca 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -651,12 +651,12 @@ static int create_trace_probe(int argc, char **argv)
651 event = strchr(group, '/') + 1; 651 event = strchr(group, '/') + 1;
652 event[-1] = '\0'; 652 event[-1] = '\0';
653 if (strlen(group) == 0) { 653 if (strlen(group) == 0) {
654 pr_info("Group name is not specifiled\n"); 654 pr_info("Group name is not specified\n");
655 return -EINVAL; 655 return -EINVAL;
656 } 656 }
657 } 657 }
658 if (strlen(event) == 0) { 658 if (strlen(event) == 0) {
659 pr_info("Event name is not specifiled\n"); 659 pr_info("Event name is not specified\n");
660 return -EINVAL; 660 return -EINVAL;
661 } 661 }
662 } 662 }
@@ -1174,80 +1174,60 @@ static int kretprobe_event_define_fields(struct ftrace_event_call *event_call)
1174 return 0; 1174 return 0;
1175} 1175}
1176 1176
1177static int __probe_event_show_format(struct trace_seq *s, 1177static int __set_print_fmt(struct trace_probe *tp, char *buf, int len)
1178 struct trace_probe *tp, const char *fmt,
1179 const char *arg)
1180{ 1178{
1181 int i; 1179 int i;
1180 int pos = 0;
1182 1181
1183 /* Show format */ 1182 const char *fmt, *arg;
1184 if (!trace_seq_printf(s, "\nprint fmt: \"%s", fmt))
1185 return 0;
1186 1183
1187 for (i = 0; i < tp->nr_args; i++) 1184 if (!probe_is_return(tp)) {
1188 if (!trace_seq_printf(s, " %s=%%lx", tp->args[i].name)) 1185 fmt = "(%lx)";
1189 return 0; 1186 arg = "REC->" FIELD_STRING_IP;
1187 } else {
1188 fmt = "(%lx <- %lx)";
1189 arg = "REC->" FIELD_STRING_FUNC ", REC->" FIELD_STRING_RETIP;
1190 }
1190 1191
1191 if (!trace_seq_printf(s, "\", %s", arg)) 1192 /* When len=0, we just calculate the needed length */
1192 return 0; 1193#define LEN_OR_ZERO (len ? len - pos : 0)
1193 1194
1194 for (i = 0; i < tp->nr_args; i++) 1195 pos += snprintf(buf + pos, LEN_OR_ZERO, "\"%s", fmt);
1195 if (!trace_seq_printf(s, ", REC->%s", tp->args[i].name))
1196 return 0;
1197 1196
1198 return trace_seq_puts(s, "\n"); 1197 for (i = 0; i < tp->nr_args; i++) {
1199} 1198 pos += snprintf(buf + pos, LEN_OR_ZERO, " %s=%%lx",
1199 tp->args[i].name);
1200 }
1200 1201
1201#undef SHOW_FIELD 1202 pos += snprintf(buf + pos, LEN_OR_ZERO, "\", %s", arg);
1202#define SHOW_FIELD(type, item, name) \
1203 do { \
1204 ret = trace_seq_printf(s, "\tfield:" #type " %s;\t" \
1205 "offset:%u;\tsize:%u;\tsigned:%d;\n", name,\
1206 (unsigned int)offsetof(typeof(field), item),\
1207 (unsigned int)sizeof(type), \
1208 is_signed_type(type)); \
1209 if (!ret) \
1210 return 0; \
1211 } while (0)
1212 1203
1213static int kprobe_event_show_format(struct ftrace_event_call *call, 1204 for (i = 0; i < tp->nr_args; i++) {
1214 struct trace_seq *s) 1205 pos += snprintf(buf + pos, LEN_OR_ZERO, ", REC->%s",
1215{ 1206 tp->args[i].name);
1216 struct kprobe_trace_entry field __attribute__((unused)); 1207 }
1217 int ret, i;
1218 struct trace_probe *tp = (struct trace_probe *)call->data;
1219
1220 SHOW_FIELD(unsigned long, ip, FIELD_STRING_IP);
1221 SHOW_FIELD(int, nargs, FIELD_STRING_NARGS);
1222 1208
1223 /* Show fields */ 1209#undef LEN_OR_ZERO
1224 for (i = 0; i < tp->nr_args; i++)
1225 SHOW_FIELD(unsigned long, args[i], tp->args[i].name);
1226 trace_seq_puts(s, "\n");
1227 1210
1228 return __probe_event_show_format(s, tp, "(%lx)", 1211 /* return the length of print_fmt */
1229 "REC->" FIELD_STRING_IP); 1212 return pos;
1230} 1213}
1231 1214
1232static int kretprobe_event_show_format(struct ftrace_event_call *call, 1215static int set_print_fmt(struct trace_probe *tp)
1233 struct trace_seq *s)
1234{ 1216{
1235 struct kretprobe_trace_entry field __attribute__((unused)); 1217 int len;
1236 int ret, i; 1218 char *print_fmt;
1237 struct trace_probe *tp = (struct trace_probe *)call->data;
1238 1219
1239 SHOW_FIELD(unsigned long, func, FIELD_STRING_FUNC); 1220 /* First: called with 0 length to calculate the needed length */
1240 SHOW_FIELD(unsigned long, ret_ip, FIELD_STRING_RETIP); 1221 len = __set_print_fmt(tp, NULL, 0);
1241 SHOW_FIELD(int, nargs, FIELD_STRING_NARGS); 1222 print_fmt = kmalloc(len + 1, GFP_KERNEL);
1223 if (!print_fmt)
1224 return -ENOMEM;
1242 1225
1243 /* Show fields */ 1226 /* Second: actually write the @print_fmt */
1244 for (i = 0; i < tp->nr_args; i++) 1227 __set_print_fmt(tp, print_fmt, len + 1);
1245 SHOW_FIELD(unsigned long, args[i], tp->args[i].name); 1228 tp->call.print_fmt = print_fmt;
1246 trace_seq_puts(s, "\n");
1247 1229
1248 return __probe_event_show_format(s, tp, "(%lx <- %lx)", 1230 return 0;
1249 "REC->" FIELD_STRING_FUNC
1250 ", REC->" FIELD_STRING_RETIP);
1251} 1231}
1252 1232
1253#ifdef CONFIG_EVENT_PROFILE 1233#ifdef CONFIG_EVENT_PROFILE
@@ -1448,18 +1428,20 @@ static int register_probe_event(struct trace_probe *tp)
1448 if (probe_is_return(tp)) { 1428 if (probe_is_return(tp)) {
1449 tp->event.trace = print_kretprobe_event; 1429 tp->event.trace = print_kretprobe_event;
1450 call->raw_init = probe_event_raw_init; 1430 call->raw_init = probe_event_raw_init;
1451 call->show_format = kretprobe_event_show_format;
1452 call->define_fields = kretprobe_event_define_fields; 1431 call->define_fields = kretprobe_event_define_fields;
1453 } else { 1432 } else {
1454 tp->event.trace = print_kprobe_event; 1433 tp->event.trace = print_kprobe_event;
1455 call->raw_init = probe_event_raw_init; 1434 call->raw_init = probe_event_raw_init;
1456 call->show_format = kprobe_event_show_format;
1457 call->define_fields = kprobe_event_define_fields; 1435 call->define_fields = kprobe_event_define_fields;
1458 } 1436 }
1437 if (set_print_fmt(tp) < 0)
1438 return -ENOMEM;
1459 call->event = &tp->event; 1439 call->event = &tp->event;
1460 call->id = register_ftrace_event(&tp->event); 1440 call->id = register_ftrace_event(&tp->event);
1461 if (!call->id) 1441 if (!call->id) {
1442 kfree(call->print_fmt);
1462 return -ENODEV; 1443 return -ENODEV;
1444 }
1463 call->enabled = 0; 1445 call->enabled = 0;
1464 call->regfunc = probe_event_enable; 1446 call->regfunc = probe_event_enable;
1465 call->unregfunc = probe_event_disable; 1447 call->unregfunc = probe_event_disable;
@@ -1472,6 +1454,7 @@ static int register_probe_event(struct trace_probe *tp)
1472 ret = trace_add_event_call(call); 1454 ret = trace_add_event_call(call);
1473 if (ret) { 1455 if (ret) {
1474 pr_info("Failed to register kprobe event: %s\n", call->name); 1456 pr_info("Failed to register kprobe event: %s\n", call->name);
1457 kfree(call->print_fmt);
1475 unregister_ftrace_event(&tp->event); 1458 unregister_ftrace_event(&tp->event);
1476 } 1459 }
1477 return ret; 1460 return ret;
@@ -1481,6 +1464,7 @@ static void unregister_probe_event(struct trace_probe *tp)
1481{ 1464{
1482 /* tp->event is unregistered in trace_remove_event_call() */ 1465 /* tp->event is unregistered in trace_remove_event_call() */
1483 trace_remove_event_call(&tp->call); 1466 trace_remove_event_call(&tp->call);
1467 kfree(tp->call.print_fmt);
1484} 1468}
1485 1469
1486/* Make a debugfs interface for controling probe points */ 1470/* Make a debugfs interface for controling probe points */
diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c
index 75289f372dd2..a1834dda85f4 100644
--- a/kernel/trace/trace_syscalls.c
+++ b/kernel/trace/trace_syscalls.c
@@ -143,70 +143,65 @@ extern char *__bad_type_size(void);
143 #type, #name, offsetof(typeof(trace), name), \ 143 #type, #name, offsetof(typeof(trace), name), \
144 sizeof(trace.name), is_signed_type(type) 144 sizeof(trace.name), is_signed_type(type)
145 145
146int syscall_enter_format(struct ftrace_event_call *call, struct trace_seq *s) 146static
147int __set_enter_print_fmt(struct syscall_metadata *entry, char *buf, int len)
147{ 148{
148 int i; 149 int i;
149 int ret; 150 int pos = 0;
150 struct syscall_metadata *entry = call->data;
151 struct syscall_trace_enter trace;
152 int offset = offsetof(struct syscall_trace_enter, args);
153 151
154 ret = trace_seq_printf(s, "\tfield:%s %s;\toffset:%zu;\tsize:%zu;" 152 /* When len=0, we just calculate the needed length */
155 "\tsigned:%u;\n", 153#define LEN_OR_ZERO (len ? len - pos : 0)
156 SYSCALL_FIELD(int, nr));
157 if (!ret)
158 return 0;
159 154
155 pos += snprintf(buf + pos, LEN_OR_ZERO, "\"");
160 for (i = 0; i < entry->nb_args; i++) { 156 for (i = 0; i < entry->nb_args; i++) {
161 ret = trace_seq_printf(s, "\tfield:%s %s;", entry->types[i], 157 pos += snprintf(buf + pos, LEN_OR_ZERO, "%s: 0x%%0%zulx%s",
162 entry->args[i]); 158 entry->args[i], sizeof(unsigned long),
163 if (!ret) 159 i == entry->nb_args - 1 ? "" : ", ");
164 return 0;
165 ret = trace_seq_printf(s, "\toffset:%d;\tsize:%zu;"
166 "\tsigned:%u;\n", offset,
167 sizeof(unsigned long),
168 is_signed_type(unsigned long));
169 if (!ret)
170 return 0;
171 offset += sizeof(unsigned long);
172 } 160 }
161 pos += snprintf(buf + pos, LEN_OR_ZERO, "\"");
173 162
174 trace_seq_puts(s, "\nprint fmt: \"");
175 for (i = 0; i < entry->nb_args; i++) { 163 for (i = 0; i < entry->nb_args; i++) {
176 ret = trace_seq_printf(s, "%s: 0x%%0%zulx%s", entry->args[i], 164 pos += snprintf(buf + pos, LEN_OR_ZERO,
177 sizeof(unsigned long), 165 ", ((unsigned long)(REC->%s))", entry->args[i]);
178 i == entry->nb_args - 1 ? "" : ", ");
179 if (!ret)
180 return 0;
181 } 166 }
182 trace_seq_putc(s, '"');
183 167
184 for (i = 0; i < entry->nb_args; i++) { 168#undef LEN_OR_ZERO
185 ret = trace_seq_printf(s, ", ((unsigned long)(REC->%s))",
186 entry->args[i]);
187 if (!ret)
188 return 0;
189 }
190 169
191 return trace_seq_putc(s, '\n'); 170 /* return the length of print_fmt */
171 return pos;
192} 172}
193 173
194int syscall_exit_format(struct ftrace_event_call *call, struct trace_seq *s) 174static int set_syscall_print_fmt(struct ftrace_event_call *call)
195{ 175{
196 int ret; 176 char *print_fmt;
197 struct syscall_trace_exit trace; 177 int len;
178 struct syscall_metadata *entry = call->data;
198 179
199 ret = trace_seq_printf(s, 180 if (entry->enter_event != call) {
200 "\tfield:%s %s;\toffset:%zu;\tsize:%zu;" 181 call->print_fmt = "\"0x%lx\", REC->ret";
201 "\tsigned:%u;\n"
202 "\tfield:%s %s;\toffset:%zu;\tsize:%zu;"
203 "\tsigned:%u;\n",
204 SYSCALL_FIELD(int, nr),
205 SYSCALL_FIELD(long, ret));
206 if (!ret)
207 return 0; 182 return 0;
183 }
208 184
209 return trace_seq_printf(s, "\nprint fmt: \"0x%%lx\", REC->ret\n"); 185 /* First: called with 0 length to calculate the needed length */
186 len = __set_enter_print_fmt(entry, NULL, 0);
187
188 print_fmt = kmalloc(len + 1, GFP_KERNEL);
189 if (!print_fmt)
190 return -ENOMEM;
191
192 /* Second: actually write the @print_fmt */
193 __set_enter_print_fmt(entry, print_fmt, len + 1);
194 call->print_fmt = print_fmt;
195
196 return 0;
197}
198
199static void free_syscall_print_fmt(struct ftrace_event_call *call)
200{
201 struct syscall_metadata *entry = call->data;
202
203 if (entry->enter_event == call)
204 kfree(call->print_fmt);
210} 205}
211 206
212int syscall_enter_define_fields(struct ftrace_event_call *call) 207int syscall_enter_define_fields(struct ftrace_event_call *call)
@@ -386,12 +381,22 @@ int init_syscall_trace(struct ftrace_event_call *call)
386{ 381{
387 int id; 382 int id;
388 383
389 id = register_ftrace_event(call->event); 384 if (set_syscall_print_fmt(call) < 0)
390 if (!id) 385 return -ENOMEM;
391 return -ENODEV; 386
392 call->id = id; 387 id = trace_event_raw_init(call);
393 INIT_LIST_HEAD(&call->fields); 388
394 return 0; 389 if (id < 0) {
390 free_syscall_print_fmt(call);
391 return id;
392 }
393
394 return id;
395}
396
397unsigned long __init arch_syscall_addr(int nr)
398{
399 return (unsigned long)sys_call_table[nr];
395} 400}
396 401
397int __init init_ftrace_syscalls(void) 402int __init init_ftrace_syscalls(void)
@@ -603,7 +608,7 @@ int prof_sysexit_enable(struct ftrace_event_call *call)
603 ret = register_trace_sys_exit(prof_syscall_exit); 608 ret = register_trace_sys_exit(prof_syscall_exit);
604 if (ret) { 609 if (ret) {
605 pr_info("event trace: Could not activate" 610 pr_info("event trace: Could not activate"
606 "syscall entry trace point"); 611 "syscall exit trace point");
607 } else { 612 } else {
608 set_bit(num, enabled_prof_exit_syscalls); 613 set_bit(num, enabled_prof_exit_syscalls);
609 sys_prof_refcount_exit++; 614 sys_prof_refcount_exit++;
diff --git a/scripts/recordmcount.pl b/scripts/recordmcount.pl
index ea6f6e3adaea..f3c9c0a90b98 100755
--- a/scripts/recordmcount.pl
+++ b/scripts/recordmcount.pl
@@ -136,13 +136,14 @@ my %text_sections = (
136 ".text.unlikely" => 1, 136 ".text.unlikely" => 1,
137); 137);
138 138
139$objdump = "objdump" if ((length $objdump) == 0); 139# Note: we are nice to C-programmers here, thus we skip the '||='-idiom.
140$objcopy = "objcopy" if ((length $objcopy) == 0); 140$objdump = 'objdump' if (!$objdump);
141$cc = "gcc" if ((length $cc) == 0); 141$objcopy = 'objcopy' if (!$objcopy);
142$ld = "ld" if ((length $ld) == 0); 142$cc = 'gcc' if (!$cc);
143$nm = "nm" if ((length $nm) == 0); 143$ld = 'ld' if (!$ld);
144$rm = "rm" if ((length $rm) == 0); 144$nm = 'nm' if (!$nm);
145$mv = "mv" if ((length $mv) == 0); 145$rm = 'rm' if (!$rm);
146$mv = 'mv' if (!$mv);
146 147
147#print STDERR "running: $P '$arch' '$objdump' '$objcopy' '$cc' '$ld' " . 148#print STDERR "running: $P '$arch' '$objdump' '$objcopy' '$cc' '$ld' " .
148# "'$nm' '$rm' '$mv' '$inputfile'\n"; 149# "'$nm' '$rm' '$mv' '$inputfile'\n";
@@ -432,14 +433,14 @@ sub update_funcs
432 433
433 # Loop through all the mcount caller offsets and print a reference 434 # Loop through all the mcount caller offsets and print a reference
434 # to the caller based from the ref_func. 435 # to the caller based from the ref_func.
435 for (my $i=0; $i <= $#offsets; $i++) { 436 if (!$opened) {
436 if (!$opened) { 437 open(FILE, ">$mcount_s") || die "can't create $mcount_s\n";
437 open(FILE, ">$mcount_s") || die "can't create $mcount_s\n"; 438 $opened = 1;
438 $opened = 1; 439 print FILE "\t.section $mcount_section,\"a\",$section_type\n";
439 print FILE "\t.section $mcount_section,\"a\",$section_type\n"; 440 print FILE "\t.align $alignment\n" if (defined($alignment));
440 print FILE "\t.align $alignment\n" if (defined($alignment)); 441 }
441 } 442 foreach my $cur_offset (@offsets) {
442 printf FILE "\t%s %s + %d\n", $type, $ref_func, $offsets[$i] - $offset; 443 printf FILE "\t%s %s + %d\n", $type, $ref_func, $cur_offset - $offset;
443 } 444 }
444} 445}
445 446
@@ -476,11 +477,7 @@ while (<IN>) {
476 $read_headers = 0; 477 $read_headers = 0;
477 478
478 # Only record text sections that we know are safe 479 # Only record text sections that we know are safe
479 if (defined($text_sections{$1})) { 480 $read_function = defined($text_sections{$1});
480 $read_function = 1;
481 } else {
482 $read_function = 0;
483 }
484 # print out any recorded offsets 481 # print out any recorded offsets
485 update_funcs(); 482 update_funcs();
486 483
@@ -514,7 +511,7 @@ while (<IN>) {
514 } 511 }
515 # is this a call site to mcount? If so, record it to print later 512 # is this a call site to mcount? If so, record it to print later
516 if ($text_found && /$mcount_regex/) { 513 if ($text_found && /$mcount_regex/) {
517 $offsets[$#offsets + 1] = hex $1; 514 push(@offsets, hex $1);
518 } 515 }
519} 516}
520 517
generated by cgit v1.2.2 (git 2.25.0) at 2024-09-25 01:29:08 -0400