* [PATCH 1/4] [PATCH 1/4] function-graph/x86: replace unbalanced ret with jmp
2009-10-13 21:19 [PATCH 0/4] [GIT PUSH][v2.6.33] tracing: various updates (take 2) Steven Rostedt
@ 2009-10-13 21:19 ` Steven Rostedt
2009-10-13 21:19 ` [PATCH 2/4] [PATCH 2/4] tracing: support multiple pids in set_pid_ftrace file Steven Rostedt
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Steven Rostedt @ 2009-10-13 21:19 UTC (permalink / raw)
To: linux-kernel
Cc: Ingo Molnar, Andrew Morton, Frederic Weisbecker, Mathieu Desnoyers
[-- Attachment #1: 0001-function-graph-x86-replace-unbalanced-ret-with-jmp.patch --]
[-- Type: text/plain, Size: 2155 bytes --]
From: Steven Rostedt <srostedt@redhat.com>
The function graph tracer replaces the return address with a hook to
trace the exit of the function call. This hook will finish by returning
to the real location the function should return to.
But the current implementation uses a ret to jump to the real return
location. This causes a imbalance between calls and ret. That is
the original function does a call, the ret goes to the handler
and then the handler does a ret without a matching call.
Although the function graph tracer itself still breaks the branch
predictor by replacing the original ret, by using a second ret and
causing an imbalance, it breaks the predictor even more.
This patch replaces the ret with a jmp to keep the calls and ret
balanced. I tested this on one box and it showed a 1.7% increase in
performance. Another box only showed a small 0.3% increase. But no
box that I tested this on showed a decrease in performance by making this
change.
Cc: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
---
arch/x86/kernel/entry_32.S | 7 ++-----
arch/x86/kernel/entry_64.S | 6 +++---
2 files changed, 5 insertions(+), 8 deletions(-)
diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S
index c097e7d..7d52e9d 100644
--- a/arch/x86/kernel/entry_32.S
+++ b/arch/x86/kernel/entry_32.S
@@ -1185,17 +1185,14 @@ END(ftrace_graph_caller)
.globl return_to_handler
return_to_handler:
- pushl $0
pushl %eax
- pushl %ecx
pushl %edx
movl %ebp, %eax
call ftrace_return_to_handler
- movl %eax, 0xc(%esp)
+ movl %eax, %ecx
popl %edx
- popl %ecx
popl %eax
- ret
+ jmp *%ecx
#endif
.section .rodata,"a"
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index b5c061f..bd5bbdd 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -155,11 +155,11 @@ GLOBAL(return_to_handler)
call ftrace_return_to_handler
- movq %rax, 16(%rsp)
+ movq %rax, %rdi
movq 8(%rsp), %rdx
movq (%rsp), %rax
- addq $16, %rsp
- retq
+ addq $24, %rsp
+ jmp *%rdi
#endif
--
1.6.3.3
^ permalink raw reply [flat|nested] 5+ messages in thread* [PATCH 2/4] [PATCH 2/4] tracing: support multiple pids in set_pid_ftrace file
2009-10-13 21:19 [PATCH 0/4] [GIT PUSH][v2.6.33] tracing: various updates (take 2) Steven Rostedt
2009-10-13 21:19 ` [PATCH 1/4] [PATCH 1/4] function-graph/x86: replace unbalanced ret with jmp Steven Rostedt
@ 2009-10-13 21:19 ` Steven Rostedt
2009-10-13 21:19 ` [PATCH 3/4] [PATCH 3/4] tracing: enable records during the module load Steven Rostedt
2009-10-13 21:19 ` [PATCH 4/4] [PATCH 4/4] tracing: enabling "__cold" functions Steven Rostedt
3 siblings, 0 replies; 5+ messages in thread
From: Steven Rostedt @ 2009-10-13 21:19 UTC (permalink / raw)
To: linux-kernel; +Cc: Ingo Molnar, Andrew Morton, Frederic Weisbecker, Jiri Olsa
[-- Attachment #1: 0002-tracing-support-multiple-pids-in-set_pid_ftrace-file.patch --]
[-- Type: text/plain, Size: 8348 bytes --]
From: jolsa@redhat.com <jolsa@redhat.com>
Adding the possibility to set more than 1 pid in the set_pid_ftrace file,
thus allowing to trace more than 1 independent processes.
Usage:
sh-4.0# echo 284 > ./set_ftrace_pid
sh-4.0# cat ./set_ftrace_pid
284
sh-4.0# echo 1 >> ./set_ftrace_pid
sh-4.0# echo 0 >> ./set_ftrace_pid
sh-4.0# cat ./set_ftrace_pid
swapper tasks
1
284
sh-4.0# echo 4 > ./set_ftrace_pid
sh-4.0# cat ./set_ftrace_pid
4
sh-4.0# echo > ./set_ftrace_pid
sh-4.0# cat ./set_ftrace_pid
no pid
sh-4.0#
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
LKML-Reference: <1253088900-18506-1-git-send-email-jolsa@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
---
kernel/trace/ftrace.c | 234 ++++++++++++++++++++++++++++++++++---------------
kernel/trace/trace.h | 4 +-
2 files changed, 167 insertions(+), 71 deletions(-)
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index 45c9659..0c799d1 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -60,6 +60,13 @@ static int last_ftrace_enabled;
/* Quick disabling of function tracer. */
int function_trace_stop;
+/* List for set_ftrace_pid's pids. */
+LIST_HEAD(ftrace_pids);
+struct ftrace_pid {
+ struct list_head list;
+ struct pid *pid;
+};
+
/*
* ftrace_disabled is set when an anomaly is discovered.
* ftrace_disabled is much stronger than ftrace_enabled.
@@ -159,7 +166,7 @@ static int __register_ftrace_function(struct ftrace_ops *ops)
else
func = ftrace_list_func;
- if (ftrace_pid_trace) {
+ if (!list_empty(&ftrace_pids)) {
set_ftrace_pid_function(func);
func = ftrace_pid_func;
}
@@ -207,7 +214,7 @@ static int __unregister_ftrace_function(struct ftrace_ops *ops)
if (ftrace_list->next == &ftrace_list_end) {
ftrace_func_t func = ftrace_list->func;
- if (ftrace_pid_trace) {
+ if (!list_empty(&ftrace_pids)) {
set_ftrace_pid_function(func);
func = ftrace_pid_func;
}
@@ -235,7 +242,7 @@ static void ftrace_update_pid_func(void)
func = __ftrace_trace_function;
#endif
- if (ftrace_pid_trace) {
+ if (!list_empty(&ftrace_pids)) {
set_ftrace_pid_function(func);
func = ftrace_pid_func;
} else {
@@ -825,8 +832,6 @@ static __init void ftrace_profile_debugfs(struct dentry *d_tracer)
}
#endif /* CONFIG_FUNCTION_PROFILER */
-/* set when tracing only a pid */
-struct pid *ftrace_pid_trace;
static struct pid * const ftrace_swapper_pid = &init_struct_pid;
#ifdef CONFIG_DYNAMIC_FTRACE
@@ -2758,23 +2763,6 @@ static inline void ftrace_startup_enable(int command) { }
# define ftrace_shutdown_sysctl() do { } while (0)
#endif /* CONFIG_DYNAMIC_FTRACE */
-static ssize_t
-ftrace_pid_read(struct file *file, char __user *ubuf,
- size_t cnt, loff_t *ppos)
-{
- char buf[64];
- int r;
-
- if (ftrace_pid_trace == ftrace_swapper_pid)
- r = sprintf(buf, "swapper tasks\n");
- else if (ftrace_pid_trace)
- r = sprintf(buf, "%u\n", pid_vnr(ftrace_pid_trace));
- else
- r = sprintf(buf, "no pid\n");
-
- return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
-}
-
static void clear_ftrace_swapper(void)
{
struct task_struct *p;
@@ -2825,14 +2813,12 @@ static void set_ftrace_pid(struct pid *pid)
rcu_read_unlock();
}
-static void clear_ftrace_pid_task(struct pid **pid)
+static void clear_ftrace_pid_task(struct pid *pid)
{
- if (*pid == ftrace_swapper_pid)
+ if (pid == ftrace_swapper_pid)
clear_ftrace_swapper();
else
- clear_ftrace_pid(*pid);
-
- *pid = NULL;
+ clear_ftrace_pid(pid);
}
static void set_ftrace_pid_task(struct pid *pid)
@@ -2843,11 +2829,140 @@ static void set_ftrace_pid_task(struct pid *pid)
set_ftrace_pid(pid);
}
+static int ftrace_pid_add(int p)
+{
+ struct pid *pid;
+ struct ftrace_pid *fpid;
+ int ret = -EINVAL;
+
+ mutex_lock(&ftrace_lock);
+
+ if (!p)
+ pid = ftrace_swapper_pid;
+ else
+ pid = find_get_pid(p);
+
+ if (!pid)
+ goto out;
+
+ ret = 0;
+
+ list_for_each_entry(fpid, &ftrace_pids, list)
+ if (fpid->pid == pid)
+ goto out_put;
+
+ ret = -ENOMEM;
+
+ fpid = kmalloc(sizeof(*fpid), GFP_KERNEL);
+ if (!fpid)
+ goto out_put;
+
+ list_add(&fpid->list, &ftrace_pids);
+ fpid->pid = pid;
+
+ set_ftrace_pid_task(pid);
+
+ ftrace_update_pid_func();
+ ftrace_startup_enable(0);
+
+ mutex_unlock(&ftrace_lock);
+ return 0;
+
+out_put:
+ if (pid != ftrace_swapper_pid)
+ put_pid(pid);
+
+out:
+ mutex_unlock(&ftrace_lock);
+ return ret;
+}
+
+static void ftrace_pid_reset(void)
+{
+ struct ftrace_pid *fpid, *safe;
+
+ mutex_lock(&ftrace_lock);
+ list_for_each_entry_safe(fpid, safe, &ftrace_pids, list) {
+ struct pid *pid = fpid->pid;
+
+ clear_ftrace_pid_task(pid);
+
+ list_del(&fpid->list);
+ kfree(fpid);
+ }
+
+ ftrace_update_pid_func();
+ ftrace_startup_enable(0);
+
+ mutex_unlock(&ftrace_lock);
+}
+
+static void *fpid_start(struct seq_file *m, loff_t *pos)
+{
+ mutex_lock(&ftrace_lock);
+
+ if (list_empty(&ftrace_pids) && (!*pos))
+ return (void *) 1;
+
+ return seq_list_start(&ftrace_pids, *pos);
+}
+
+static void *fpid_next(struct seq_file *m, void *v, loff_t *pos)
+{
+ if (v == (void *)1)
+ return NULL;
+
+ return seq_list_next(v, &ftrace_pids, pos);
+}
+
+static void fpid_stop(struct seq_file *m, void *p)
+{
+ mutex_unlock(&ftrace_lock);
+}
+
+static int fpid_show(struct seq_file *m, void *v)
+{
+ const struct ftrace_pid *fpid = list_entry(v, struct ftrace_pid, list);
+
+ if (v == (void *)1) {
+ seq_printf(m, "no pid\n");
+ return 0;
+ }
+
+ if (fpid->pid == ftrace_swapper_pid)
+ seq_printf(m, "swapper tasks\n");
+ else
+ seq_printf(m, "%u\n", pid_vnr(fpid->pid));
+
+ return 0;
+}
+
+static const struct seq_operations ftrace_pid_sops = {
+ .start = fpid_start,
+ .next = fpid_next,
+ .stop = fpid_stop,
+ .show = fpid_show,
+};
+
+static int
+ftrace_pid_open(struct inode *inode, struct file *file)
+{
+ int ret = 0;
+
+ if ((file->f_mode & FMODE_WRITE) &&
+ (file->f_flags & O_TRUNC))
+ ftrace_pid_reset();
+
+ if (file->f_mode & FMODE_READ)
+ ret = seq_open(file, &ftrace_pid_sops);
+
+ return ret;
+}
+
static ssize_t
ftrace_pid_write(struct file *filp, const char __user *ubuf,
size_t cnt, loff_t *ppos)
{
- struct pid *pid;
char buf[64];
long val;
int ret;
@@ -2860,57 +2975,38 @@ ftrace_pid_write(struct file *filp, const char __user *ubuf,
buf[cnt] = 0;
+ /*
+ * Allow "echo > set_ftrace_pid" or "echo -n '' > set_ftrace_pid"
+ * to clean the filter quietly.
+ */
+ strstrip(buf);
+ if (strlen(buf) == 0)
+ return 1;
+
ret = strict_strtol(buf, 10, &val);
if (ret < 0)
return ret;
- mutex_lock(&ftrace_lock);
- if (val < 0) {
- /* disable pid tracing */
- if (!ftrace_pid_trace)
- goto out;
-
- clear_ftrace_pid_task(&ftrace_pid_trace);
-
- } else {
- /* swapper task is special */
- if (!val) {
- pid = ftrace_swapper_pid;
- if (pid == ftrace_pid_trace)
- goto out;
- } else {
- pid = find_get_pid(val);
-
- if (pid == ftrace_pid_trace) {
- put_pid(pid);
- goto out;
- }
- }
-
- if (ftrace_pid_trace)
- clear_ftrace_pid_task(&ftrace_pid_trace);
-
- if (!pid)
- goto out;
-
- ftrace_pid_trace = pid;
-
- set_ftrace_pid_task(ftrace_pid_trace);
- }
+ ret = ftrace_pid_add(val);
- /* update the function call */
- ftrace_update_pid_func();
- ftrace_startup_enable(0);
+ return ret ? ret : cnt;
+}
- out:
- mutex_unlock(&ftrace_lock);
+static int
+ftrace_pid_release(struct inode *inode, struct file *file)
+{
+ if (file->f_mode & FMODE_READ)
+ seq_release(inode, file);
- return cnt;
+ return 0;
}
static const struct file_operations ftrace_pid_fops = {
- .read = ftrace_pid_read,
- .write = ftrace_pid_write,
+ .open = ftrace_pid_open,
+ .write = ftrace_pid_write,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = ftrace_pid_release,
};
static __init int ftrace_init_debugfs(void)
diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h
index f22a7ac..acef8b4 100644
--- a/kernel/trace/trace.h
+++ b/kernel/trace/trace.h
@@ -496,12 +496,12 @@ print_graph_function(struct trace_iterator *iter)
}
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
-extern struct pid *ftrace_pid_trace;
+extern struct list_head ftrace_pids;
#ifdef CONFIG_FUNCTION_TRACER
static inline int ftrace_trace_task(struct task_struct *task)
{
- if (!ftrace_pid_trace)
+ if (list_empty(&ftrace_pids))
return 1;
return test_tsk_trace_trace(task);
--
1.6.3.3
^ permalink raw reply [flat|nested] 5+ messages in thread* [PATCH 3/4] [PATCH 3/4] tracing: enable records during the module load
2009-10-13 21:19 [PATCH 0/4] [GIT PUSH][v2.6.33] tracing: various updates (take 2) Steven Rostedt
2009-10-13 21:19 ` [PATCH 1/4] [PATCH 1/4] function-graph/x86: replace unbalanced ret with jmp Steven Rostedt
2009-10-13 21:19 ` [PATCH 2/4] [PATCH 2/4] tracing: support multiple pids in set_pid_ftrace file Steven Rostedt
@ 2009-10-13 21:19 ` Steven Rostedt
2009-10-13 21:19 ` [PATCH 4/4] [PATCH 4/4] tracing: enabling "__cold" functions Steven Rostedt
3 siblings, 0 replies; 5+ messages in thread
From: Steven Rostedt @ 2009-10-13 21:19 UTC (permalink / raw)
To: linux-kernel; +Cc: Ingo Molnar, Andrew Morton, Frederic Weisbecker, Jiri Olsa
[-- Attachment #1: 0003-tracing-enable-records-during-the-module-load.patch --]
[-- Type: text/plain, Size: 2697 bytes --]
From: Jiri Olsa <jolsa@redhat.com>
I was debuging some module using "function" and "function_graph" tracers
and noticed, that if you load module after you enabled tracing, the module's
hooks will convert only to NOP instructions.
The attached patch enables modules' hooks if there's function
trace allready on, thus allowing to trace module functions.
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
LKML-Reference: <1255183220-15640-1-git-send-email-jolsa@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
---
kernel/trace/ftrace.c | 38 ++++++++++++++++++++++++++++++--------
1 files changed, 30 insertions(+), 8 deletions(-)
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index 0c799d1..aaea9cd 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -1270,12 +1270,34 @@ static int ftrace_update_code(struct module *mod)
ftrace_new_addrs = p->newlist;
p->flags = 0L;
- /* convert record (i.e, patch mcount-call with NOP) */
- if (ftrace_code_disable(mod, p)) {
- p->flags |= FTRACE_FL_CONVERTED;
- ftrace_update_cnt++;
- } else
+ /*
+ * Do the initial record convertion from mcount jump
+ * to the NOP instructions.
+ */
+ if (!ftrace_code_disable(mod, p)) {
ftrace_free_rec(p);
+ continue;
+ }
+
+ p->flags |= FTRACE_FL_CONVERTED;
+ ftrace_update_cnt++;
+
+ /*
+ * If the tracing is enabled, go ahead and enable the record.
+ *
+ * The reason not to enable the record immediatelly is the
+ * inherent check of ftrace_make_nop/ftrace_make_call for
+ * correct previous instructions. Making first the NOP
+ * conversion puts the module to the correct state, thus
+ * passing the ftrace_make_call check.
+ */
+ if (ftrace_start_up) {
+ int failed = __ftrace_replace_code(p, 1);
+ if (failed) {
+ ftrace_bug(failed, p->ip);
+ ftrace_free_rec(p);
+ }
+ }
}
stop = ftrace_now(raw_smp_processor_id());
@@ -2609,7 +2631,7 @@ static __init int ftrace_init_dyn_debugfs(struct dentry *d_tracer)
return 0;
}
-static int ftrace_convert_nops(struct module *mod,
+static int ftrace_process_locs(struct module *mod,
unsigned long *start,
unsigned long *end)
{
@@ -2669,7 +2691,7 @@ static void ftrace_init_module(struct module *mod,
{
if (ftrace_disabled || start == end)
return;
- ftrace_convert_nops(mod, start, end);
+ ftrace_process_locs(mod, start, end);
}
static int ftrace_module_notify(struct notifier_block *self,
@@ -2730,7 +2752,7 @@ void __init ftrace_init(void)
last_ftrace_enabled = ftrace_enabled = 1;
- ret = ftrace_convert_nops(NULL,
+ ret = ftrace_process_locs(NULL,
__start_mcount_loc,
__stop_mcount_loc);
--
1.6.3.3
^ permalink raw reply [flat|nested] 5+ messages in thread