Steven Rostedt (VMware) 7aa1eaef9f function_graph: Allow multiple users to attach to function graph
Allow for multiple users to attach to function graph tracer at the same
time. Only 16 simultaneous users can attach to the tracer. This is because
there's an array that stores the pointers to the attached fgraph_ops. When
a function being traced is entered, each of the ftrace_ops entryfunc is
called and if it returns non zero, its index into the array will be added
to the shadow stack.

On exit of the function being traced, the shadow stack will contain the
indexes of the ftrace_ops on the array that want their retfunc to be
called.

Because a function may sleep for a long time (if a task sleeps itself),
the return of the function may be literally days later. If the ftrace_ops
is removed, its place on the array is replaced with a ftrace_ops that
contains the stub functions and that will be called when the function
finally returns.

If another ftrace_ops is added that happens to get the same index into the
array, its return function may be called. But that's actually the way
things current work with the old function graph tracer. If one tracer is
removed and another is added, the new one will get the return calls of the
function traced by the previous one, thus this is not a regression. This
can be fixed by adding a counter to each time the array item is updated and
save that on the shadow stack as well, such that it won't be called if the
index saved does not match the index on the array.

Note, being able to filter functions when both are called is not completely
handled yet, but that shouldn't be too hard to manage.

Co-developed with Masami Hiramatsu:
Link: https://lore.kernel.org/linux-trace-kernel/171509096221.162236.8806372072523195752.stgit@devnote2
Link: https://lore.kernel.org/linux-trace-kernel/20240603190821.555493396@goodmis.org

Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Florent Revest <revest@chromium.org>
Cc: Martin KaFai Lau <martin.lau@linux.dev>
Cc: bpf <bpf@vger.kernel.org>
Cc: Sven Schnelle <svens@linux.ibm.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alan Maguire <alan.maguire@oracle.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Guo Ren <guoren@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
2024-06-04 10:35:58 -04:00

984 lines
26 KiB
C

// SPDX-License-Identifier: GPL-2.0
/*
* Infrastructure to took into function calls and returns.
* Copyright (c) 2008-2009 Frederic Weisbecker <fweisbec@gmail.com>
* Mostly borrowed from function tracer which
* is Copyright (c) Steven Rostedt <srostedt@redhat.com>
*
* Highly modified by Steven Rostedt (VMware).
*/
#include <linux/bits.h>
#include <linux/jump_label.h>
#include <linux/suspend.h>
#include <linux/ftrace.h>
#include <linux/slab.h>
#include <trace/events/sched.h>
#include "ftrace_internal.h"
#include "trace.h"
#ifdef CONFIG_DYNAMIC_FTRACE
#define ASSIGN_OPS_HASH(opsname, val) \
.func_hash = val, \
.local_hash.regex_lock = __MUTEX_INITIALIZER(opsname.local_hash.regex_lock),
#else
#define ASSIGN_OPS_HASH(opsname, val)
#endif
/*
* FGRAPH_FRAME_SIZE: Size in bytes of the meta data on the shadow stack
* FGRAPH_FRAME_OFFSET: Size in long words of the meta data frame
*/
#define FGRAPH_FRAME_SIZE sizeof(struct ftrace_ret_stack)
#define FGRAPH_FRAME_OFFSET DIV_ROUND_UP(FGRAPH_FRAME_SIZE, sizeof(long))
/*
* On entry to a function (via function_graph_enter()), a new fgraph frame
* (ftrace_ret_stack) is pushed onto the stack as well as a word that
* holds a bitmask and a type (called "bitmap"). The bitmap is defined as:
*
* bits: 0 - 9 offset in words from the previous ftrace_ret_stack
* Currently, this will always be set to FGRAPH_FRAME_OFFSET
* to get to the fgraph frame.
*
* bits: 10 - 11 Type of storage
* 0 - reserved
* 1 - bitmap of fgraph_array index
*
* For type with "bitmap of fgraph_array index" (FGRAPH_TYPE_BITMAP):
* bits: 12 - 27 The bitmap of fgraph_ops fgraph_array index
* That is, it's a bitmask of 0-15 (16 bits)
* where if a corresponding ops in the fgraph_array[]
* expects a callback from the return of the function
* it's corresponding bit will be set.
*
*
* The top of the ret_stack (when not empty) will always have a reference
* word that points to the last fgraph frame that was saved.
*
* That is, at the end of function_graph_enter, if the first and forth
* fgraph_ops on the fgraph_array[] (index 0 and 3) needs their retfunc called
* on the return of the function being traced, this is what will be on the
* task's shadow ret_stack: (the stack grows upward)
*
* ret_stack[SHADOW_STACK_MAX_OFFSET]
* ...
* | | <- task->curr_ret_stack
* +--------------------------------------------+
* | (9 << 12) | (1 << 10) | FGRAPH_FRAME_OFFSET|
* | *or put another way* |
* | (BIT(3)|BIT(0)) << FGRAPH_INDEX_SHIFT | \ |
* | FGRAPH_TYPE_BITMAP << FGRAPH_TYPE_SHIFT| \ |
* | (offset:FGRAPH_FRAME_OFFSET) | <- the offset is from here
* +--------------------------------------------+
* | struct ftrace_ret_stack |
* | (stores the saved ret pointer) | <- the offset points here
* +--------------------------------------------+
* | (X) | (N) | ( N words away from
* | | previous ret_stack)
* ...
* ret_stack[0]
*
* If a backtrace is required, and the real return pointer needs to be
* fetched, then it looks at the task's curr_ret_stack offset, if it
* is greater than zero (reserved, or right before popped), it would mask
* the value by FGRAPH_FRAME_OFFSET_MASK to get the offset of the
* ftrace_ret_stack structure stored on the shadow stack.
*/
/*
* The following is for the top word on the stack:
*
* FGRAPH_FRAME_OFFSET (0-9) holds the offset delta to the fgraph frame
* FGRAPH_TYPE (10-11) holds the type of word this is.
* (RESERVED or BITMAP)
*/
#define FGRAPH_FRAME_OFFSET_BITS 10
#define FGRAPH_FRAME_OFFSET_MASK GENMASK(FGRAPH_FRAME_OFFSET_BITS - 1, 0)
#define FGRAPH_TYPE_BITS 2
#define FGRAPH_TYPE_MASK GENMASK(FGRAPH_TYPE_BITS - 1, 0)
#define FGRAPH_TYPE_SHIFT FGRAPH_FRAME_OFFSET_BITS
enum {
FGRAPH_TYPE_RESERVED = 0,
FGRAPH_TYPE_BITMAP = 1,
};
/*
* For BITMAP type:
* FGRAPH_INDEX (12-27) bits holding the gops index wanting return callback called
*/
#define FGRAPH_INDEX_BITS 16
#define FGRAPH_INDEX_MASK GENMASK(FGRAPH_INDEX_BITS - 1, 0)
#define FGRAPH_INDEX_SHIFT (FGRAPH_TYPE_SHIFT + FGRAPH_TYPE_BITS)
#define FGRAPH_ARRAY_SIZE FGRAPH_INDEX_BITS
/*
* SHADOW_STACK_SIZE: The size in bytes of the entire shadow stack
* SHADOW_STACK_OFFSET: The size in long words of the shadow stack
* SHADOW_STACK_MAX_OFFSET: The max offset of the stack for a new frame to be added
*/
#define SHADOW_STACK_SIZE (PAGE_SIZE)
#define SHADOW_STACK_OFFSET (SHADOW_STACK_SIZE / sizeof(long))
/* Leave on a buffer at the end */
#define SHADOW_STACK_MAX_OFFSET (SHADOW_STACK_OFFSET - (FGRAPH_FRAME_OFFSET + 1))
/* RET_STACK(): Return the frame from a given @offset from task @t */
#define RET_STACK(t, offset) ((struct ftrace_ret_stack *)(&(t)->ret_stack[offset]))
DEFINE_STATIC_KEY_FALSE(kill_ftrace_graph);
int ftrace_graph_active;
static int fgraph_array_cnt;
static struct fgraph_ops *fgraph_array[FGRAPH_ARRAY_SIZE];
/* Get the FRAME_OFFSET from the word from the @offset on ret_stack */
static inline int get_frame_offset(struct task_struct *t, int offset)
{
return t->ret_stack[offset] & FGRAPH_FRAME_OFFSET_MASK;
}
/* Get FGRAPH_TYPE from the word from the @offset at ret_stack */
static inline int get_fgraph_type(struct task_struct *t, int offset)
{
return (t->ret_stack[offset] >> FGRAPH_TYPE_SHIFT) & FGRAPH_TYPE_MASK;
}
/* For BITMAP type: get the bitmask from the @offset at ret_stack */
static inline unsigned long
get_bitmap_bits(struct task_struct *t, int offset)
{
return (t->ret_stack[offset] >> FGRAPH_INDEX_SHIFT) & FGRAPH_INDEX_MASK;
}
/* Write the bitmap to the ret_stack at @offset (does index, offset and bitmask) */
static inline void
set_bitmap(struct task_struct *t, int offset, unsigned long bitmap)
{
t->ret_stack[offset] = (bitmap << FGRAPH_INDEX_SHIFT) |
(FGRAPH_TYPE_BITMAP << FGRAPH_TYPE_SHIFT) | FGRAPH_FRAME_OFFSET;
}
/*
* @offset: The offset into @t->ret_stack to find the ret_stack entry
* @frame_offset: Where to place the offset into @t->ret_stack of that entry
*
* Returns a pointer to the previous ret_stack below @offset or NULL
* when it reaches the bottom of the stack.
*
* Calling this with:
*
* offset = task->curr_ret_stack;
* do {
* ret_stack = get_ret_stack(task, offset, &offset);
* } while (ret_stack);
*
* Will iterate through all the ret_stack entries from curr_ret_stack
* down to the first one.
*/
static inline struct ftrace_ret_stack *
get_ret_stack(struct task_struct *t, int offset, int *frame_offset)
{
int offs;
BUILD_BUG_ON(FGRAPH_FRAME_SIZE % sizeof(long));
if (unlikely(offset <= 0))
return NULL;
offs = get_frame_offset(t, --offset);
if (WARN_ON_ONCE(offs <= 0 || offs > offset))
return NULL;
offset -= offs;
*frame_offset = offset;
return RET_STACK(t, offset);
}
/* Both enabled by default (can be cleared by function_graph tracer flags */
static bool fgraph_sleep_time = true;
#ifdef CONFIG_DYNAMIC_FTRACE
/*
* archs can override this function if they must do something
* to enable hook for graph tracer.
*/
int __weak ftrace_enable_ftrace_graph_caller(void)
{
return 0;
}
/*
* archs can override this function if they must do something
* to disable hook for graph tracer.
*/
int __weak ftrace_disable_ftrace_graph_caller(void)
{
return 0;
}
#endif
int ftrace_graph_entry_stub(struct ftrace_graph_ent *trace)
{
return 0;
}
static void ftrace_graph_ret_stub(struct ftrace_graph_ret *trace)
{
}
static struct fgraph_ops fgraph_stub = {
.entryfunc = ftrace_graph_entry_stub,
.retfunc = ftrace_graph_ret_stub,
};
/**
* ftrace_graph_stop - set to permanently disable function graph tracing
*
* In case of an error int function graph tracing, this is called
* to try to keep function graph tracing from causing any more harm.
* Usually this is pretty severe and this is called to try to at least
* get a warning out to the user.
*/
void ftrace_graph_stop(void)
{
static_branch_enable(&kill_ftrace_graph);
}
/* Add a function return address to the trace stack on thread info.*/
static int
ftrace_push_return_trace(unsigned long ret, unsigned long func,
unsigned long frame_pointer, unsigned long *retp,
int fgraph_idx)
{
struct ftrace_ret_stack *ret_stack;
unsigned long long calltime;
unsigned long val;
int offset;
if (unlikely(ftrace_graph_is_dead()))
return -EBUSY;
if (!current->ret_stack)
return -EBUSY;
BUILD_BUG_ON(SHADOW_STACK_SIZE % sizeof(long));
/* Set val to "reserved" with the delta to the new fgraph frame */
val = (FGRAPH_TYPE_RESERVED << FGRAPH_TYPE_SHIFT) | FGRAPH_FRAME_OFFSET;
/*
* We must make sure the ret_stack is tested before we read
* anything else.
*/
smp_rmb();
/*
* Check if there's room on the shadow stack to fit a fraph frame
* and a bitmap word.
*/
if (current->curr_ret_stack + FGRAPH_FRAME_OFFSET + 1 >= SHADOW_STACK_MAX_OFFSET) {
atomic_inc(&current->trace_overrun);
return -EBUSY;
}
calltime = trace_clock_local();
offset = READ_ONCE(current->curr_ret_stack);
ret_stack = RET_STACK(current, offset);
offset += FGRAPH_FRAME_OFFSET;
/* ret offset = FGRAPH_FRAME_OFFSET ; type = reserved */
current->ret_stack[offset] = val;
ret_stack->ret = ret;
/*
* The unwinders expect curr_ret_stack to point to either zero
* or an offset where to find the next ret_stack. Even though the
* ret stack might be bogus, we want to write the ret and the
* offset to find the ret_stack before we increment the stack point.
* If an interrupt comes in now before we increment the curr_ret_stack
* it may blow away what we wrote. But that's fine, because the
* offset will still be correct (even though the 'ret' won't be).
* What we worry about is the offset being correct after we increment
* the curr_ret_stack and before we update that offset, as if an
* interrupt comes in and does an unwind stack dump, it will need
* at least a correct offset!
*/
barrier();
WRITE_ONCE(current->curr_ret_stack, offset + 1);
/*
* This next barrier is to ensure that an interrupt coming in
* will not corrupt what we are about to write.
*/
barrier();
/* Still keep it reserved even if an interrupt came in */
current->ret_stack[offset] = val;
ret_stack->ret = ret;
ret_stack->func = func;
ret_stack->calltime = calltime;
#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
ret_stack->fp = frame_pointer;
#endif
#ifdef HAVE_FUNCTION_GRAPH_RET_ADDR_PTR
ret_stack->retp = retp;
#endif
return offset;
}
/*
* Not all archs define MCOUNT_INSN_SIZE which is used to look for direct
* functions. But those archs currently don't support direct functions
* anyway, and ftrace_find_rec_direct() is just a stub for them.
* Define MCOUNT_INSN_SIZE to keep those archs compiling.
*/
#ifndef MCOUNT_INSN_SIZE
/* Make sure this only works without direct calls */
# ifdef CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS
# error MCOUNT_INSN_SIZE not defined with direct calls enabled
# endif
# define MCOUNT_INSN_SIZE 0
#endif
/* If the caller does not use ftrace, call this function. */
int function_graph_enter(unsigned long ret, unsigned long func,
unsigned long frame_pointer, unsigned long *retp)
{
struct ftrace_graph_ent trace;
unsigned long bitmap = 0;
int offset;
int i;
trace.func = func;
trace.depth = ++current->curr_ret_depth;
offset = ftrace_push_return_trace(ret, func, frame_pointer, retp, 0);
if (offset < 0)
goto out;
for (i = 0; i < fgraph_array_cnt; i++) {
struct fgraph_ops *gops = fgraph_array[i];
if (gops == &fgraph_stub)
continue;
if (gops->entryfunc(&trace))
bitmap |= BIT(i);
}
if (!bitmap)
goto out_ret;
/*
* Since this function uses fgraph_idx = 0 as a tail-call checking
* flag, set that bit always.
*/
set_bitmap(current, offset, bitmap | BIT(0));
return 0;
out_ret:
current->curr_ret_stack -= FGRAPH_FRAME_OFFSET + 1;
out:
current->curr_ret_depth--;
return -EBUSY;
}
/* Retrieve a function return address to the trace stack on thread info.*/
static struct ftrace_ret_stack *
ftrace_pop_return_trace(struct ftrace_graph_ret *trace, unsigned long *ret,
unsigned long frame_pointer, int *offset)
{
struct ftrace_ret_stack *ret_stack;
ret_stack = get_ret_stack(current, current->curr_ret_stack, offset);
if (unlikely(!ret_stack)) {
ftrace_graph_stop();
WARN(1, "Bad function graph ret_stack pointer: %d",
current->curr_ret_stack);
/* Might as well panic, otherwise we have no where to go */
*ret = (unsigned long)panic;
return NULL;
}
#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
/*
* The arch may choose to record the frame pointer used
* and check it here to make sure that it is what we expect it
* to be. If gcc does not set the place holder of the return
* address in the frame pointer, and does a copy instead, then
* the function graph trace will fail. This test detects this
* case.
*
* Currently, x86_32 with optimize for size (-Os) makes the latest
* gcc do the above.
*
* Note, -mfentry does not use frame pointers, and this test
* is not needed if CC_USING_FENTRY is set.
*/
if (unlikely(ret_stack->fp != frame_pointer)) {
ftrace_graph_stop();
WARN(1, "Bad frame pointer: expected %lx, received %lx\n"
" from func %ps return to %lx\n",
ret_stack->fp,
frame_pointer,
(void *)ret_stack->func,
ret_stack->ret);
*ret = (unsigned long)panic;
return NULL;
}
#endif
*offset += FGRAPH_FRAME_OFFSET;
*ret = ret_stack->ret;
trace->func = ret_stack->func;
trace->calltime = ret_stack->calltime;
trace->overrun = atomic_read(&current->trace_overrun);
trace->depth = current->curr_ret_depth;
/*
* We still want to trace interrupts coming in if
* max_depth is set to 1. Make sure the decrement is
* seen before ftrace_graph_return.
*/
barrier();
return ret_stack;
}
/*
* Hibernation protection.
* The state of the current task is too much unstable during
* suspend/restore to disk. We want to protect against that.
*/
static int
ftrace_suspend_notifier_call(struct notifier_block *bl, unsigned long state,
void *unused)
{
switch (state) {
case PM_HIBERNATION_PREPARE:
pause_graph_tracing();
break;
case PM_POST_HIBERNATION:
unpause_graph_tracing();
break;
}
return NOTIFY_DONE;
}
static struct notifier_block ftrace_suspend_notifier = {
.notifier_call = ftrace_suspend_notifier_call,
};
/* fgraph_ret_regs is not defined without CONFIG_FUNCTION_GRAPH_RETVAL */
struct fgraph_ret_regs;
/*
* Send the trace to the ring-buffer.
* @return the original return address.
*/
static unsigned long __ftrace_return_to_handler(struct fgraph_ret_regs *ret_regs,
unsigned long frame_pointer)
{
struct ftrace_ret_stack *ret_stack;
struct ftrace_graph_ret trace;
unsigned long bitmap;
unsigned long ret;
int offset;
int i;
ret_stack = ftrace_pop_return_trace(&trace, &ret, frame_pointer, &offset);
if (unlikely(!ret_stack)) {
ftrace_graph_stop();
WARN_ON(1);
/* Might as well panic. What else to do? */
return (unsigned long)panic;
}
trace.rettime = trace_clock_local();
#ifdef CONFIG_FUNCTION_GRAPH_RETVAL
trace.retval = fgraph_ret_regs_return_value(ret_regs);
#endif
bitmap = get_bitmap_bits(current, offset);
for (i = 0; i < FGRAPH_ARRAY_SIZE; i++) {
struct fgraph_ops *gops = fgraph_array[i];
if (!(bitmap & BIT(i)))
continue;
if (gops == &fgraph_stub)
continue;
gops->retfunc(&trace);
}
/*
* The ftrace_graph_return() may still access the current
* ret_stack structure, we need to make sure the update of
* curr_ret_stack is after that.
*/
barrier();
current->curr_ret_stack -= FGRAPH_FRAME_OFFSET + 1;
current->curr_ret_depth--;
return ret;
}
/*
* After all architecures have selected HAVE_FUNCTION_GRAPH_RETVAL, we can
* leave only ftrace_return_to_handler(ret_regs).
*/
#ifdef CONFIG_HAVE_FUNCTION_GRAPH_RETVAL
unsigned long ftrace_return_to_handler(struct fgraph_ret_regs *ret_regs)
{
return __ftrace_return_to_handler(ret_regs,
fgraph_ret_regs_frame_pointer(ret_regs));
}
#else
unsigned long ftrace_return_to_handler(unsigned long frame_pointer)
{
return __ftrace_return_to_handler(NULL, frame_pointer);
}
#endif
/**
* ftrace_graph_get_ret_stack - return the entry of the shadow stack
* @task: The task to read the shadow stack from.
* @idx: Index down the shadow stack
*
* Return the ret_struct on the shadow stack of the @task at the
* call graph at @idx starting with zero. If @idx is zero, it
* will return the last saved ret_stack entry. If it is greater than
* zero, it will return the corresponding ret_stack for the depth
* of saved return addresses.
*/
struct ftrace_ret_stack *
ftrace_graph_get_ret_stack(struct task_struct *task, int idx)
{
struct ftrace_ret_stack *ret_stack = NULL;
int offset = task->curr_ret_stack;
if (offset < 0)
return NULL;
do {
ret_stack = get_ret_stack(task, offset, &offset);
} while (ret_stack && --idx >= 0);
return ret_stack;
}
/**
* ftrace_graph_ret_addr - convert a potentially modified stack return address
* to its original value
*
* This function can be called by stack unwinding code to convert a found stack
* return address ('ret') to its original value, in case the function graph
* tracer has modified it to be 'return_to_handler'. If the address hasn't
* been modified, the unchanged value of 'ret' is returned.
*
* 'idx' is a state variable which should be initialized by the caller to zero
* before the first call.
*
* 'retp' is a pointer to the return address on the stack. It's ignored if
* the arch doesn't have HAVE_FUNCTION_GRAPH_RET_ADDR_PTR defined.
*/
#ifdef HAVE_FUNCTION_GRAPH_RET_ADDR_PTR
unsigned long ftrace_graph_ret_addr(struct task_struct *task, int *idx,
unsigned long ret, unsigned long *retp)
{
struct ftrace_ret_stack *ret_stack;
int i = task->curr_ret_stack;
if (ret != (unsigned long)dereference_kernel_function_descriptor(return_to_handler))
return ret;
while (i > 0) {
ret_stack = get_ret_stack(current, i, &i);
if (!ret_stack)
break;
if (ret_stack->retp == retp)
return ret_stack->ret;
}
return ret;
}
#else /* !HAVE_FUNCTION_GRAPH_RET_ADDR_PTR */
unsigned long ftrace_graph_ret_addr(struct task_struct *task, int *idx,
unsigned long ret, unsigned long *retp)
{
struct ftrace_ret_stack *ret_stack;
int offset = task->curr_ret_stack;
int i;
if (ret != (unsigned long)dereference_kernel_function_descriptor(return_to_handler))
return ret;
if (!idx)
return ret;
i = *idx;
do {
ret_stack = get_ret_stack(task, offset, &offset);
i--;
} while (i >= 0 && ret_stack);
if (ret_stack)
return ret_stack->ret;
return ret;
}
#endif /* HAVE_FUNCTION_GRAPH_RET_ADDR_PTR */
static struct ftrace_ops graph_ops = {
.func = ftrace_graph_func,
.flags = FTRACE_OPS_FL_INITIALIZED |
FTRACE_OPS_FL_PID |
FTRACE_OPS_GRAPH_STUB,
#ifdef FTRACE_GRAPH_TRAMP_ADDR
.trampoline = FTRACE_GRAPH_TRAMP_ADDR,
/* trampoline_size is only needed for dynamically allocated tramps */
#endif
ASSIGN_OPS_HASH(graph_ops, &global_ops.local_hash)
};
void ftrace_graph_sleep_time_control(bool enable)
{
fgraph_sleep_time = enable;
}
/*
* Simply points to ftrace_stub, but with the proper protocol.
* Defined by the linker script in linux/vmlinux.lds.h
*/
extern void ftrace_stub_graph(struct ftrace_graph_ret *);
/* The callbacks that hook a function */
trace_func_graph_ret_t ftrace_graph_return = ftrace_stub_graph;
trace_func_graph_ent_t ftrace_graph_entry = ftrace_graph_entry_stub;
static trace_func_graph_ent_t __ftrace_graph_entry = ftrace_graph_entry_stub;
/* Try to assign a return stack array on FTRACE_RETSTACK_ALLOC_SIZE tasks. */
static int alloc_retstack_tasklist(unsigned long **ret_stack_list)
{
int i;
int ret = 0;
int start = 0, end = FTRACE_RETSTACK_ALLOC_SIZE;
struct task_struct *g, *t;
for (i = 0; i < FTRACE_RETSTACK_ALLOC_SIZE; i++) {
ret_stack_list[i] = kmalloc(SHADOW_STACK_SIZE, GFP_KERNEL);
if (!ret_stack_list[i]) {
start = 0;
end = i;
ret = -ENOMEM;
goto free;
}
}
rcu_read_lock();
for_each_process_thread(g, t) {
if (start == end) {
ret = -EAGAIN;
goto unlock;
}
if (t->ret_stack == NULL) {
atomic_set(&t->trace_overrun, 0);
t->curr_ret_stack = 0;
t->curr_ret_depth = -1;
/* Make sure the tasks see the 0 first: */
smp_wmb();
t->ret_stack = ret_stack_list[start++];
}
}
unlock:
rcu_read_unlock();
free:
for (i = start; i < end; i++)
kfree(ret_stack_list[i]);
return ret;
}
static void
ftrace_graph_probe_sched_switch(void *ignore, bool preempt,
struct task_struct *prev,
struct task_struct *next,
unsigned int prev_state)
{
struct ftrace_ret_stack *ret_stack;
unsigned long long timestamp;
int offset;
/*
* Does the user want to count the time a function was asleep.
* If so, do not update the time stamps.
*/
if (fgraph_sleep_time)
return;
timestamp = trace_clock_local();
prev->ftrace_timestamp = timestamp;
/* only process tasks that we timestamped */
if (!next->ftrace_timestamp)
return;
/*
* Update all the counters in next to make up for the
* time next was sleeping.
*/
timestamp -= next->ftrace_timestamp;
for (offset = next->curr_ret_stack; offset > 0; ) {
ret_stack = get_ret_stack(next, offset, &offset);
if (ret_stack)
ret_stack->calltime += timestamp;
}
}
static int ftrace_graph_entry_test(struct ftrace_graph_ent *trace)
{
if (!ftrace_ops_test(&global_ops, trace->func, NULL))
return 0;
return __ftrace_graph_entry(trace);
}
/*
* The function graph tracer should only trace the functions defined
* by set_ftrace_filter and set_ftrace_notrace. If another function
* tracer ops is registered, the graph tracer requires testing the
* function against the global ops, and not just trace any function
* that any ftrace_ops registered.
*/
void update_function_graph_func(void)
{
struct ftrace_ops *op;
bool do_test = false;
/*
* The graph and global ops share the same set of functions
* to test. If any other ops is on the list, then
* the graph tracing needs to test if its the function
* it should call.
*/
do_for_each_ftrace_op(op, ftrace_ops_list) {
if (op != &global_ops && op != &graph_ops &&
op != &ftrace_list_end) {
do_test = true;
/* in double loop, break out with goto */
goto out;
}
} while_for_each_ftrace_op(op);
out:
if (do_test)
ftrace_graph_entry = ftrace_graph_entry_test;
else
ftrace_graph_entry = __ftrace_graph_entry;
}
static DEFINE_PER_CPU(unsigned long *, idle_ret_stack);
static void
graph_init_task(struct task_struct *t, unsigned long *ret_stack)
{
atomic_set(&t->trace_overrun, 0);
t->ftrace_timestamp = 0;
t->curr_ret_stack = 0;
t->curr_ret_depth = -1;
/* make curr_ret_stack visible before we add the ret_stack */
smp_wmb();
t->ret_stack = ret_stack;
}
/*
* Allocate a return stack for the idle task. May be the first
* time through, or it may be done by CPU hotplug online.
*/
void ftrace_graph_init_idle_task(struct task_struct *t, int cpu)
{
t->curr_ret_stack = 0;
t->curr_ret_depth = -1;
/*
* The idle task has no parent, it either has its own
* stack or no stack at all.
*/
if (t->ret_stack)
WARN_ON(t->ret_stack != per_cpu(idle_ret_stack, cpu));
if (ftrace_graph_active) {
unsigned long *ret_stack;
ret_stack = per_cpu(idle_ret_stack, cpu);
if (!ret_stack) {
ret_stack = kmalloc(SHADOW_STACK_SIZE, GFP_KERNEL);
if (!ret_stack)
return;
per_cpu(idle_ret_stack, cpu) = ret_stack;
}
graph_init_task(t, ret_stack);
}
}
/* Allocate a return stack for newly created task */
void ftrace_graph_init_task(struct task_struct *t)
{
/* Make sure we do not use the parent ret_stack */
t->ret_stack = NULL;
t->curr_ret_stack = 0;
t->curr_ret_depth = -1;
if (ftrace_graph_active) {
unsigned long *ret_stack;
ret_stack = kmalloc(SHADOW_STACK_SIZE, GFP_KERNEL);
if (!ret_stack)
return;
graph_init_task(t, ret_stack);
}
}
void ftrace_graph_exit_task(struct task_struct *t)
{
unsigned long *ret_stack = t->ret_stack;
t->ret_stack = NULL;
/* NULL must become visible to IRQs before we free it: */
barrier();
kfree(ret_stack);
}
/* Allocate a return stack for each task */
static int start_graph_tracing(void)
{
unsigned long **ret_stack_list;
int ret, cpu;
ret_stack_list = kmalloc(SHADOW_STACK_SIZE, GFP_KERNEL);
if (!ret_stack_list)
return -ENOMEM;
/* The cpu_boot init_task->ret_stack will never be freed */
for_each_online_cpu(cpu) {
if (!idle_task(cpu)->ret_stack)
ftrace_graph_init_idle_task(idle_task(cpu), cpu);
}
do {
ret = alloc_retstack_tasklist(ret_stack_list);
} while (ret == -EAGAIN);
if (!ret) {
ret = register_trace_sched_switch(ftrace_graph_probe_sched_switch, NULL);
if (ret)
pr_info("ftrace_graph: Couldn't activate tracepoint"
" probe to kernel_sched_switch\n");
}
kfree(ret_stack_list);
return ret;
}
int register_ftrace_graph(struct fgraph_ops *gops)
{
int ret = 0;
int i;
mutex_lock(&ftrace_lock);
if (!fgraph_array[0]) {
/* The array must always have real data on it */
for (i = 0; i < FGRAPH_ARRAY_SIZE; i++)
fgraph_array[i] = &fgraph_stub;
}
/* Look for an available spot */
for (i = 0; i < FGRAPH_ARRAY_SIZE; i++) {
if (fgraph_array[i] == &fgraph_stub)
break;
}
if (i >= FGRAPH_ARRAY_SIZE) {
ret = -EBUSY;
goto out;
}
fgraph_array[i] = gops;
if (i + 1 > fgraph_array_cnt)
fgraph_array_cnt = i + 1;
gops->idx = i;
ftrace_graph_active++;
if (ftrace_graph_active == 1) {
register_pm_notifier(&ftrace_suspend_notifier);
ret = start_graph_tracing();
if (ret) {
ftrace_graph_active--;
goto out;
}
ftrace_graph_return = gops->retfunc;
/*
* Update the indirect function to the entryfunc, and the
* function that gets called to the entry_test first. Then
* call the update fgraph entry function to determine if
* the entryfunc should be called directly or not.
*/
__ftrace_graph_entry = gops->entryfunc;
ftrace_graph_entry = ftrace_graph_entry_test;
update_function_graph_func();
ret = ftrace_startup(&graph_ops, FTRACE_START_FUNC_RET);
}
out:
mutex_unlock(&ftrace_lock);
return ret;
}
void unregister_ftrace_graph(struct fgraph_ops *gops)
{
int i;
mutex_lock(&ftrace_lock);
if (unlikely(!ftrace_graph_active))
goto out;
for (i = 0; i < fgraph_array_cnt; i++)
if (gops == fgraph_array[i])
break;
if (i >= fgraph_array_cnt)
goto out;
fgraph_array[i] = &fgraph_stub;
if (i + 1 == fgraph_array_cnt) {
for (; i >= 0; i--)
if (fgraph_array[i] != &fgraph_stub)
break;
fgraph_array_cnt = i + 1;
}
ftrace_graph_active--;
if (!ftrace_graph_active) {
ftrace_graph_return = ftrace_stub_graph;
ftrace_graph_entry = ftrace_graph_entry_stub;
__ftrace_graph_entry = ftrace_graph_entry_stub;
ftrace_shutdown(&graph_ops, FTRACE_STOP_FUNC_RET);
unregister_pm_notifier(&ftrace_suspend_notifier);
unregister_trace_sched_switch(ftrace_graph_probe_sched_switch, NULL);
}
out:
mutex_unlock(&ftrace_lock);
}