* Fwd: [PATCH] kretprobe instance recycled by parent process
[not found] ` <67029b170602141936v69b85832q@mail.gmail.com>
@ 2006-02-15 3:39 ` Zhou Yingchao
2006-02-15 12:55 ` bibo mao
0 siblings, 1 reply; 5+ messages in thread
From: Zhou Yingchao @ 2006-02-15 3:39 UTC (permalink / raw)
To: linux-kernel
2006/2/15, bibo mao <bibo_mao@linux.intel.com>:
> When kretprobe probe schedule() function, if probed process exit then
> schedule() function will never return, so some kretprobe instance will
> never be recycled. By this patch the parent process will recycle
> retprobe instance of probed function, there will be no memory leak of
> kretprobe instance. This patch is based on 2.6.16-rc3.
Is there any process which can exit without go through the do_exit() path?
--
Yingchao Zhou
***********************************************
Institute Of Computing Technology
Chinese Academy of Sciences
Tel(O) : 010-62613792-28
***********************************************
--
Yingchao Zhou
***********************************************
Institute Of Computing Technology
Chinese Academy of Sciences
Tel(O) : 010-62613792-28
***********************************************
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: Fwd: [PATCH] kretprobe instance recycled by parent process
2006-02-15 12:55 ` bibo mao
@ 2006-02-15 5:59 ` Zhou Yingchao
2006-02-15 6:21 ` bibo mao
0 siblings, 1 reply; 5+ messages in thread
From: Zhou Yingchao @ 2006-02-15 5:59 UTC (permalink / raw)
To: bibo mao; +Cc: linux-kernel
> >> When kretprobe probe schedule() function, if probed process exit then
> >> schedule() function will never return, so some kretprobe instance will
> >> never be recycled. By this patch the parent process will recycle
> >> retprobe instance of probed function, there will be no memory leak of
> >> kretprobe instance. This patch is based on 2.6.16-rc3.
> >
> > Is there any process which can exit without go through the do_exit() path?
> > --
> When process exits through do_exit() function, it will call schedule()
> function. But if schedule() function is probed by kretprobe, this time
> schedule() function will not return never because process has exited.
>
> bibo,mao
>
In the original path, doesn't the call path of
do_exit()->exit_thread()->kprobe_flush_task(current) recycle the
kretprobe instance? Is there anything misundstood?
--
Yingchao Zhou
***********************************************
Institute Of Computing Technology
Chinese Academy of Sciences
Tel(O) : 010-62613792-28
***********************************************
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: Fwd: [PATCH] kretprobe instance recycled by parent process
2006-02-15 5:59 ` Zhou Yingchao
@ 2006-02-15 6:21 ` bibo mao
0 siblings, 0 replies; 5+ messages in thread
From: bibo mao @ 2006-02-15 6:21 UTC (permalink / raw)
To: Zhou Yingchao; +Cc: linux-kernel
Zhou Yingchao wrote:
>>>> When kretprobe probe schedule() function, if probed process exit then
>>>> schedule() function will never return, so some kretprobe instance will
>>>> never be recycled. By this patch the parent process will recycle
>>>> retprobe instance of probed function, there will be no memory leak of
>>>> kretprobe instance. This patch is based on 2.6.16-rc3.
>>> Is there any process which can exit without go through the do_exit() path?
>>> --
>> When process exits through do_exit() function, it will call schedule()
>> function. But if schedule() function is probed by kretprobe, this time
>> schedule() function will not return never because process has exited.
>>
>> bibo,mao
>>
>
> In the original path, doesn't the call path of
> do_exit()->exit_thread()->kprobe_flush_task(current) recycle the
> kretprobe instance? Is there anything misundstood?
> --
yes, it is right. The old recycle method is
do_exit()->exit_thread()->kprobe_flush_task(current)
->schedule()
At last line of do_exit() it will call schedule() function, and this
time it will never return. But if schedule function is probed, who is
responsible for recycling it?
bibo,mao
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH] kretprobe instance recycled by parent process
@ 2006-02-15 10:42 bibo mao
[not found] ` <67029b170602141936v69b85832q@mail.gmail.com>
0 siblings, 1 reply; 5+ messages in thread
From: bibo mao @ 2006-02-15 10:42 UTC (permalink / raw)
To: akpm; +Cc: anil.s.keshavamurthy, yanmin.zhang, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 6335 bytes --]
When kretprobe probe schedule() function, if probed process exit then
schedule() function will never return, so some kretprobe instance will
never be recycled. By this patch the parent process will recycle
retprobe instance of probed function, there will be no memory leak of
kretprobe instance. This patch is based on 2.6.16-rc3.
Signed-off-by: bibo mao <bibo.mao@intel.com>
arch/i386/kernel/process.c | 8 --------
arch/ia64/kernel/process.c | 8 --------
arch/powerpc/kernel/process.c | 2 --
arch/x86_64/kernel/process.c | 8 --------
kernel/kprobes.c | 10 +++++-----
kernel/sched.c | 9 ++++++++-
6 files changed, 13 insertions(+), 32 deletions(-)
diff -Nruap linux-2.6.16-rc3.org/arch/i386/kernel/process.c
linux-2.6.16-rc3/arch/i386/kernel/process.c
--- linux-2.6.16-rc3.org/arch/i386/kernel/process.c 2006-02-14
05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/arch/i386/kernel/process.c 2006-02-14
05:48:06.000000000 +0800
@@ -38,7 +38,6 @@
#include <linux/kallsyms.h>
#include <linux/ptrace.h>
#include <linux/random.h>
-#include <linux/kprobes.h>
#include <asm/uaccess.h>
#include <asm/pgtable.h>
@@ -364,13 +363,6 @@ void exit_thread(void)
struct task_struct *tsk = current;
struct thread_struct *t = &tsk->thread;
- /*
- * Remove function-return probe instances associated with this task
- * and put them back on the free list. Do not insert an exit probe for
- * this function, it will be disabled by kprobe_flush_task if you do.
- */
- kprobe_flush_task(tsk);
-
/* The process may have allocated an io port bitmap... nuke it. */
if (unlikely(NULL != t->io_bitmap_ptr)) {
int cpu = get_cpu();
diff -Nruap linux-2.6.16-rc3.org/arch/ia64/kernel/process.c
linux-2.6.16-rc3/arch/ia64/kernel/process.c
--- linux-2.6.16-rc3.org/arch/ia64/kernel/process.c 2006-02-14
05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/arch/ia64/kernel/process.c 2006-02-14
05:48:40.000000000 +0800
@@ -30,7 +30,6 @@
#include <linux/efi.h>
#include <linux/interrupt.h>
#include <linux/delay.h>
-#include <linux/kprobes.h>
#include <asm/cpu.h>
#include <asm/delay.h>
@@ -738,13 +737,6 @@ void
exit_thread (void)
{
- /*
- * Remove function-return probe instances associated with this task
- * and put them back on the free list. Do not insert an exit probe for
- * this function, it will be disabled by kprobe_flush_task if you do.
- */
- kprobe_flush_task(current);
-
ia64_drop_fpu(current);
#ifdef CONFIG_PERFMON
/* if needed, stop monitoring and flush state to perfmon context */
diff -Nruap linux-2.6.16-rc3.org/arch/powerpc/kernel/process.c
linux-2.6.16-rc3/arch/powerpc/kernel/process.c
--- linux-2.6.16-rc3.org/arch/powerpc/kernel/process.c 2006-02-14
05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/arch/powerpc/kernel/process.c 2006-02-14
05:49:33.000000000 +0800
@@ -37,7 +37,6 @@
#include <linux/mqueue.h>
#include <linux/hardirq.h>
#include <linux/utsname.h>
-#include <linux/kprobes.h>
#include <asm/pgtable.h>
#include <asm/uaccess.h>
@@ -457,7 +456,6 @@ void show_regs(struct pt_regs * regs)
void exit_thread(void)
{
- kprobe_flush_task(current);
discard_lazy_cpu_state();
}
diff -Nruap linux-2.6.16-rc3.org/arch/x86_64/kernel/process.c
linux-2.6.16-rc3/arch/x86_64/kernel/process.c
--- linux-2.6.16-rc3.org/arch/x86_64/kernel/process.c 2006-02-14
05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/arch/x86_64/kernel/process.c 2006-02-14
05:50:02.000000000 +0800
@@ -35,7 +35,6 @@
#include <linux/ptrace.h>
#include <linux/utsname.h>
#include <linux/random.h>
-#include <linux/kprobes.h>
#include <linux/notifier.h>
#include <asm/uaccess.h>
@@ -353,13 +352,6 @@ void exit_thread(void)
struct task_struct *me = current;
struct thread_struct *t = &me->thread;
- /*
- * Remove function-return probe instances associated with this task
- * and put them back on the free list. Do not insert an exit probe for
- * this function, it will be disabled by kprobe_flush_task if you do.
- */
- kprobe_flush_task(me);
-
if (me->thread.io_bitmap_ptr) {
struct tss_struct *tss = &per_cpu(init_tss, get_cpu());
diff -Nruap linux-2.6.16-rc3.org/kernel/kprobes.c
linux-2.6.16-rc3/kernel/kprobes.c
--- linux-2.6.16-rc3.org/kernel/kprobes.c 2006-02-14 05:46:38.000000000
+0800
+++ linux-2.6.16-rc3/kernel/kprobes.c 2006-02-14 08:16:55.000000000 +0800
@@ -323,10 +323,10 @@ struct hlist_head __kprobes *kretprobe_i
}
/*
- * This function is called from exit_thread or flush_thread when task tk's
- * stack is being recycled so that we can recycle any function-return probe
- * instances associated with this task. These left over instances represent
- * probed functions that have been called but will never return.
+ * This function is called from finish_task_switch when task tk becomes
dead,
+ * so that we can recycle any function-return probe instances associated
+ * with this task. These left over instances represent probed functions
+ * that have been called but will never return.
*/
void __kprobes kprobe_flush_task(struct task_struct *tk)
{
@@ -336,7 +336,7 @@ void __kprobes kprobe_flush_task(struct
unsigned long flags = 0;
spin_lock_irqsave(&kretprobe_lock, flags);
- head = kretprobe_inst_table_head(current);
+ head = kretprobe_inst_table_head(tk);
hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
if (ri->task == tk)
recycle_rp_inst(ri);
diff -Nruap linux-2.6.16-rc3.org/kernel/sched.c
linux-2.6.16-rc3/kernel/sched.c
--- linux-2.6.16-rc3.org/kernel/sched.c 2006-02-14 05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/kernel/sched.c 2006-02-14 06:11:57.000000000 +0800
@@ -49,6 +49,7 @@
#include <linux/syscalls.h>
#include <linux/times.h>
#include <linux/acct.h>
+#include <linux/kprobes.h>
#include <asm/tlb.h>
#include <asm/unistd.h>
@@ -1566,8 +1567,14 @@ static inline void finish_task_switch(ru
finish_lock_switch(rq, prev);
if (mm)
mmdrop(mm);
- if (unlikely(prev_task_flags & PF_DEAD))
+ if (unlikely(prev_task_flags & PF_DEAD)){
+ /*
+ * Remove function-return probe instances associated with this task
+ * and put them back on the free list.
+ */
+ kprobe_flush_task(prev);
put_task_struct(prev);
+ }
}
/**
[-- Attachment #2: kretprobe_instance_patch_2.6.16-rc3 --]
[-- Type: text/plain, Size: 5889 bytes --]
arch/i386/kernel/process.c | 8 --------
arch/ia64/kernel/process.c | 8 --------
arch/powerpc/kernel/process.c | 2 --
arch/x86_64/kernel/process.c | 8 --------
kernel/kprobes.c | 10 +++++-----
kernel/sched.c | 9 ++++++++-
6 files changed, 13 insertions(+), 32 deletions(-)
diff -Nruap linux-2.6.16-rc3.org/arch/i386/kernel/process.c linux-2.6.16-rc3/arch/i386/kernel/process.c
--- linux-2.6.16-rc3.org/arch/i386/kernel/process.c 2006-02-14 05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/arch/i386/kernel/process.c 2006-02-14 05:48:06.000000000 +0800
@@ -38,7 +38,6 @@
#include <linux/kallsyms.h>
#include <linux/ptrace.h>
#include <linux/random.h>
-#include <linux/kprobes.h>
#include <asm/uaccess.h>
#include <asm/pgtable.h>
@@ -364,13 +363,6 @@ void exit_thread(void)
struct task_struct *tsk = current;
struct thread_struct *t = &tsk->thread;
- /*
- * Remove function-return probe instances associated with this task
- * and put them back on the free list. Do not insert an exit probe for
- * this function, it will be disabled by kprobe_flush_task if you do.
- */
- kprobe_flush_task(tsk);
-
/* The process may have allocated an io port bitmap... nuke it. */
if (unlikely(NULL != t->io_bitmap_ptr)) {
int cpu = get_cpu();
diff -Nruap linux-2.6.16-rc3.org/arch/ia64/kernel/process.c linux-2.6.16-rc3/arch/ia64/kernel/process.c
--- linux-2.6.16-rc3.org/arch/ia64/kernel/process.c 2006-02-14 05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/arch/ia64/kernel/process.c 2006-02-14 05:48:40.000000000 +0800
@@ -30,7 +30,6 @@
#include <linux/efi.h>
#include <linux/interrupt.h>
#include <linux/delay.h>
-#include <linux/kprobes.h>
#include <asm/cpu.h>
#include <asm/delay.h>
@@ -738,13 +737,6 @@ void
exit_thread (void)
{
- /*
- * Remove function-return probe instances associated with this task
- * and put them back on the free list. Do not insert an exit probe for
- * this function, it will be disabled by kprobe_flush_task if you do.
- */
- kprobe_flush_task(current);
-
ia64_drop_fpu(current);
#ifdef CONFIG_PERFMON
/* if needed, stop monitoring and flush state to perfmon context */
diff -Nruap linux-2.6.16-rc3.org/arch/powerpc/kernel/process.c linux-2.6.16-rc3/arch/powerpc/kernel/process.c
--- linux-2.6.16-rc3.org/arch/powerpc/kernel/process.c 2006-02-14 05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/arch/powerpc/kernel/process.c 2006-02-14 05:49:33.000000000 +0800
@@ -37,7 +37,6 @@
#include <linux/mqueue.h>
#include <linux/hardirq.h>
#include <linux/utsname.h>
-#include <linux/kprobes.h>
#include <asm/pgtable.h>
#include <asm/uaccess.h>
@@ -457,7 +456,6 @@ void show_regs(struct pt_regs * regs)
void exit_thread(void)
{
- kprobe_flush_task(current);
discard_lazy_cpu_state();
}
diff -Nruap linux-2.6.16-rc3.org/arch/x86_64/kernel/process.c linux-2.6.16-rc3/arch/x86_64/kernel/process.c
--- linux-2.6.16-rc3.org/arch/x86_64/kernel/process.c 2006-02-14 05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/arch/x86_64/kernel/process.c 2006-02-14 05:50:02.000000000 +0800
@@ -35,7 +35,6 @@
#include <linux/ptrace.h>
#include <linux/utsname.h>
#include <linux/random.h>
-#include <linux/kprobes.h>
#include <linux/notifier.h>
#include <asm/uaccess.h>
@@ -353,13 +352,6 @@ void exit_thread(void)
struct task_struct *me = current;
struct thread_struct *t = &me->thread;
- /*
- * Remove function-return probe instances associated with this task
- * and put them back on the free list. Do not insert an exit probe for
- * this function, it will be disabled by kprobe_flush_task if you do.
- */
- kprobe_flush_task(me);
-
if (me->thread.io_bitmap_ptr) {
struct tss_struct *tss = &per_cpu(init_tss, get_cpu());
diff -Nruap linux-2.6.16-rc3.org/kernel/kprobes.c linux-2.6.16-rc3/kernel/kprobes.c
--- linux-2.6.16-rc3.org/kernel/kprobes.c 2006-02-14 05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/kernel/kprobes.c 2006-02-14 08:16:55.000000000 +0800
@@ -323,10 +323,10 @@ struct hlist_head __kprobes *kretprobe_i
}
/*
- * This function is called from exit_thread or flush_thread when task tk's
- * stack is being recycled so that we can recycle any function-return probe
- * instances associated with this task. These left over instances represent
- * probed functions that have been called but will never return.
+ * This function is called from finish_task_switch when task tk becomes dead,
+ * so that we can recycle any function-return probe instances associated
+ * with this task. These left over instances represent probed functions
+ * that have been called but will never return.
*/
void __kprobes kprobe_flush_task(struct task_struct *tk)
{
@@ -336,7 +336,7 @@ void __kprobes kprobe_flush_task(struct
unsigned long flags = 0;
spin_lock_irqsave(&kretprobe_lock, flags);
- head = kretprobe_inst_table_head(current);
+ head = kretprobe_inst_table_head(tk);
hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
if (ri->task == tk)
recycle_rp_inst(ri);
diff -Nruap linux-2.6.16-rc3.org/kernel/sched.c linux-2.6.16-rc3/kernel/sched.c
--- linux-2.6.16-rc3.org/kernel/sched.c 2006-02-14 05:46:38.000000000 +0800
+++ linux-2.6.16-rc3/kernel/sched.c 2006-02-14 06:11:57.000000000 +0800
@@ -49,6 +49,7 @@
#include <linux/syscalls.h>
#include <linux/times.h>
#include <linux/acct.h>
+#include <linux/kprobes.h>
#include <asm/tlb.h>
#include <asm/unistd.h>
@@ -1566,8 +1567,14 @@ static inline void finish_task_switch(ru
finish_lock_switch(rq, prev);
if (mm)
mmdrop(mm);
- if (unlikely(prev_task_flags & PF_DEAD))
+ if (unlikely(prev_task_flags & PF_DEAD)){
+ /*
+ * Remove function-return probe instances associated with this task
+ * and put them back on the free list.
+ */
+ kprobe_flush_task(prev);
put_task_struct(prev);
+ }
}
/**
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: Fwd: [PATCH] kretprobe instance recycled by parent process
2006-02-15 3:39 ` Fwd: " Zhou Yingchao
@ 2006-02-15 12:55 ` bibo mao
2006-02-15 5:59 ` Zhou Yingchao
0 siblings, 1 reply; 5+ messages in thread
From: bibo mao @ 2006-02-15 12:55 UTC (permalink / raw)
To: Zhou Yingchao; +Cc: linux-kernel
Zhou Yingchao wrote:
> 2006/2/15, bibo mao <bibo_mao@linux.intel.com>:
>> When kretprobe probe schedule() function, if probed process exit then
>> schedule() function will never return, so some kretprobe instance will
>> never be recycled. By this patch the parent process will recycle
>> retprobe instance of probed function, there will be no memory leak of
>> kretprobe instance. This patch is based on 2.6.16-rc3.
>
> Is there any process which can exit without go through the do_exit() path?
> --
When process exits through do_exit() function, it will call schedule()
function. But if schedule() function is probed by kretprobe, this time
schedule() function will not return never because process has exited.
bibo,mao
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2006-02-15 6:22 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-02-15 10:42 [PATCH] kretprobe instance recycled by parent process bibo mao
[not found] ` <67029b170602141936v69b85832q@mail.gmail.com>
2006-02-15 3:39 ` Fwd: " Zhou Yingchao
2006-02-15 12:55 ` bibo mao
2006-02-15 5:59 ` Zhou Yingchao
2006-02-15 6:21 ` bibo mao
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox