This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Re: [RFC][PATCH][kprobe] enabling booster on the preemptible kernel, take 2
- From: "bibo,mao" <bibo dot mao at intel dot com>
- To: Masami Hiramatsu <masami dot hiramatsu dot pt at hitachi dot com>
- Cc: "Keshavamurthy, Anil S" <anil dot s dot keshavamurthy at intel dot com>, Ananth N Mavinakayanahalli <ananth at in dot ibm dot com>, Prasanna S Panchamukhi <prasanna at in dot ibm dot com>, Ingo Molnar <mingo at redhat dot com>, SystemTAP <systemtap at sources dot redhat dot com>, Satoshi Oshima <soshima at redhat dot com>, Hideo Aoki <haoki at redhat dot com>, Yumiko Sugita <yumiko dot sugita dot yf at hitachi dot com>
- Date: Mon, 30 Oct 2006 14:37:08 +0800
- Subject: Re: [RFC][PATCH][kprobe] enabling booster on the preemptible kernel, take 2
- References: <45338593.6090207@hitachi.com>
This patch will boost kprobe on preemptible kernel, I think
it is deserved to waster some memory for better performance
by deferring memory free after freeze_processes.
thanks
bibo,mao
Masami Hiramatsu wrote:
Hi,
Here is the patch which enables kprobe-booster on
the preemptive kernel.
When we are unregistering a kprobe-booster, we can't release
its buffer immediately on the preemptive kernel, because
some processes might be preempted on the buffer.
The freeze_processes() and thaw_processes() functions can
clean those processes up from the buffer. However, the
processing of those functions takes a long time.
So, this patch introduces the garbage collection mechanism
of insn_slot. It also introduces the "dirty" flag to
free_insn_slot because of efficiency.
The "clean" instruction slots (dirty flag is cleared) are
released immediately. But the "dirty" slots which are used
by boosted kprobes, are marked as garbages.
collect_garbage_slots() will be invoked to release "dirty"
slots if 1) there are more than INSNS_PER_PAGE garbage slots
or 2) there are no unused slots.
Thanks,
--
Masami HIRAMATSU
2nd Research Dept.
Hitachi, Ltd., Systems Development Laboratory
E-mail: masami.hiramatsu.pt@hitachi.com
---
arch/i386/kernel/kprobes.c | 4 -
arch/ia64/kernel/kprobes.c | 2
arch/powerpc/kernel/kprobes.c | 2
arch/s390/kernel/kprobes.c | 2
arch/x86_64/kernel/kprobes.c | 2
include/linux/kprobes.h | 2
kernel/kprobes.c | 101 +++++++++++++++++++++++++++++++++---------
7 files changed, 87 insertions(+), 28 deletions(-)
Index: linux-2.6.19-rc1-mm1/kernel/kprobes.c
===================================================================
--- linux-2.6.19-rc1-mm1.orig/kernel/kprobes.c 2006-10-16 10:40:02.000000000 +0900
+++ linux-2.6.19-rc1-mm1/kernel/kprobes.c 2006-10-16 21:50:44.000000000 +0900
@@ -38,6 +38,7 @@
#include <linux/module.h>
#include <linux/moduleloader.h>
#include <linux/kallsyms.h>
+#include <linux/sched.h>
#include <asm-generic/sections.h>
#include <asm/cacheflush.h>
#include <asm/errno.h>
@@ -83,9 +84,12 @@
kprobe_opcode_t *insns; /* Page of instruction slots */
char slot_used[INSNS_PER_PAGE];
int nused;
+ int ngarbage;
};
static struct hlist_head kprobe_insn_pages;
+static int kprobe_garbage_slots;
+static int collect_garbage_slots(void);
/**
* get_insn_slot() - Find a slot on an executable page for an instruction.
@@ -96,6 +100,7 @@
struct kprobe_insn_page *kip;
struct hlist_node *pos;
+ retry:
hlist_for_each(pos, &kprobe_insn_pages) {
kip = hlist_entry(pos, struct kprobe_insn_page, hlist);
if (kip->nused < INSNS_PER_PAGE) {
@@ -112,7 +117,11 @@
}
}
- /* All out of space. Need to allocate a new page. Use slot 0.*/
+ /* If there are any garbage slots, collect it and try again. */
+ if (kprobe_garbage_slots && collect_garbage_slots() == 0) {
I am not familiar with freeze_processes/thaw_process, but I only think that
it will bring performance downgrade greatly for the moment.I think kmalloc
method is better than collect_garbage_slots.
+ goto retry;
+ }
+ /* All out of space. Need to allocate a new page. Use slot 0. */
kip = kmalloc(sizeof(struct kprobe_insn_page), GFP_KERNEL);
if (!kip) {
return NULL;
@@ -133,10 +142,70 @@
memset(kip->slot_used, 0, INSNS_PER_PAGE);
kip->slot_used[0] = 1;
kip->nused = 1;
+ kip->ngarbage = 0;
return kip->insns;
}
-void __kprobes free_insn_slot(kprobe_opcode_t *slot)
+/* Return 1 if all garbages are collected, otherwise 0. */
+static int __kprobes collect_one_slot(struct kprobe_insn_page *kip, int idx)
+{
+ kip->slot_used[idx] = 0;
+ kip->nused--;
+ if (kip->nused == 0) {
+ /*
+ * Page is no longer in use. Free it unless
+ * it's the last one. We keep the last one
+ * so as not to have to set it up again the
+ * next time somebody inserts a probe.
+ */
+ hlist_del(&kip->hlist);
+ if (hlist_empty(&kprobe_insn_pages)) {
+ INIT_HLIST_NODE(&kip->hlist);
+ hlist_add_head(&kip->hlist,
+ &kprobe_insn_pages);
+ return 1;
+ } else {
+ module_free(NULL, kip->insns);
+ kfree(kip);
+ }
+ }
+ return 0;
+}
+
+static int __kprobes collect_garbage_slots(void)
+{
+ struct kprobe_insn_page *kip;
+ struct hlist_node *pos, *next;
+ int ret = -1;
+
+#if defined(CONFIG_PREEMPT) && defined(CONFIG_PM)
+ /* Ensure no-one is preepmted on the garbages */
+ if (freeze_processes() != 0)
I do not know whether there exists non-freezeable and preemptive
kernel thread, if there exist then this thread will not be frozen.
+ goto thaw_all;
+#endif
+ hlist_for_each_safe(pos, next, &kprobe_insn_pages) {
+ int i;
+ kip = hlist_entry(pos, struct kprobe_insn_page, hlist);
+ if (kip->ngarbage == 0)
+ continue;
+ kip->ngarbage = 0; /* we will collect all garbages */
+ for (i = 0; i < INSNS_PER_PAGE; i++) {
+ if (kip->slot_used[i] == -1 &&
+ collect_one_slot(kip, i))
if collect_one_slot executes kfree(kip) and return 0, then kernel will continue
execute the for () loop sentence and access freed kip point by kip->slot_used.
+ goto collected;
+ }
+ }
+ collected:
+ kprobe_garbage_slots = 0;
+ ret = 0;
+#if defined(CONFIG_PREEMPT) && defined(CONFIG_PM)
+ thaw_all:
+ thaw_processes();
+#endif
+ return ret;
+}
+
+void __kprobes free_insn_slot(kprobe_opcode_t * slot, int dirty)
{
struct kprobe_insn_page *kip;
struct hlist_node *pos;
@@ -146,28 +215,18 @@
if (kip->insns <= slot &&
slot < kip->insns + (INSNS_PER_PAGE * MAX_INSN_SIZE)) {
int i = (slot - kip->insns) / MAX_INSN_SIZE;
- kip->slot_used[i] = 0;
- kip->nused--;
- if (kip->nused == 0) {
- /*
- * Page is no longer in use. Free it unless
- * it's the last one. We keep the last one
- * so as not to have to set it up again the
- * next time somebody inserts a probe.
- */
- hlist_del(&kip->hlist);
- if (hlist_empty(&kprobe_insn_pages)) {
- INIT_HLIST_NODE(&kip->hlist);
- hlist_add_head(&kip->hlist,
- &kprobe_insn_pages);
- } else {
- module_free(NULL, kip->insns);
- kfree(kip);
- }
+ if (dirty) {
+ kip->slot_used[i] = -1;
+ kip->ngarbage++;
it seems that break sentence is missing.
+ } else {
+ collect_one_slot(kip, i);
+ break;
}
- return;
}
}
+ if (dirty && (++kprobe_garbage_slots > INSNS_PER_PAGE)) {
+ collect_garbage_slots();
+ }
}
#endif
Index: linux-2.6.19-rc1-mm1/arch/i386/kernel/kprobes.c
===================================================================
--- linux-2.6.19-rc1-mm1.orig/arch/i386/kernel/kprobes.c 2006-10-16
10:40:00.000000000 +0900
+++ linux-2.6.19-rc1-mm1/arch/i386/kernel/kprobes.c 2006-10-16
21:43:03.000000000 +0900
@@ -184,7 +184,7 @@
void __kprobes arch_remove_kprobe(struct kprobe *p)
{
mutex_lock(&kprobe_mutex);
- free_insn_slot(p->ainsn.insn);
+ free_insn_slot(p->ainsn.insn, (p->ainsn.boostable == 1));
mutex_unlock(&kprobe_mutex);
}
@@ -333,7 +333,7 @@
return 1;
ss_probe:
-#ifndef CONFIG_PREEMPT
+#if !defined(CONFIG_PREEMPT) || defined(CONFIG_PM)
if (p->ainsn.boostable == 1 && !p->post_handler){
/* Boost up -- we can execute copied instructions directly */
reset_current_kprobe();
Index: linux-2.6.19-rc1-mm1/arch/ia64/kernel/kprobes.c
===================================================================
--- linux-2.6.19-rc1-mm1.orig/arch/ia64/kernel/kprobes.c 2006-10-16
10:40:00.000000000 +0900
+++ linux-2.6.19-rc1-mm1/arch/ia64/kernel/kprobes.c 2006-10-16
10:54:09.000000000 +0900
@@ -481,7 +481,7 @@
void __kprobes arch_remove_kprobe(struct kprobe *p)
{
mutex_lock(&kprobe_mutex);
- free_insn_slot(p->ainsn.insn);
+ free_insn_slot(p->ainsn.insn, 0);
mutex_unlock(&kprobe_mutex);
}
/*
Index: linux-2.6.19-rc1-mm1/arch/powerpc/kernel/kprobes.c
===================================================================
--- linux-2.6.19-rc1-mm1.orig/arch/powerpc/kernel/kprobes.c 2006-10-16
10:40:00.000000000 +0900
+++ linux-2.6.19-rc1-mm1/arch/powerpc/kernel/kprobes.c 2006-10-16
10:54:09.000000000 +0900
@@ -85,7 +85,7 @@
void __kprobes arch_remove_kprobe(struct kprobe *p)
{
mutex_lock(&kprobe_mutex);
- free_insn_slot(p->ainsn.insn);
+ free_insn_slot(p->ainsn.insn, 0);
mutex_unlock(&kprobe_mutex);
}
Index: linux-2.6.19-rc1-mm1/arch/s390/kernel/kprobes.c
===================================================================
--- linux-2.6.19-rc1-mm1.orig/arch/s390/kernel/kprobes.c 2006-10-16
10:40:00.000000000 +0900
+++ linux-2.6.19-rc1-mm1/arch/s390/kernel/kprobes.c 2006-10-16
10:54:09.000000000 +0900
@@ -200,7 +200,7 @@
void __kprobes arch_remove_kprobe(struct kprobe *p)
{
mutex_lock(&kprobe_mutex);
- free_insn_slot(p->ainsn.insn);
+ free_insn_slot(p->ainsn.insn, 0);
mutex_unlock(&kprobe_mutex);
}
Index: linux-2.6.19-rc1-mm1/arch/x86_64/kernel/kprobes.c
===================================================================
--- linux-2.6.19-rc1-mm1.orig/arch/x86_64/kernel/kprobes.c 2006-10-16
10:40:00.000000000 +0900
+++ linux-2.6.19-rc1-mm1/arch/x86_64/kernel/kprobes.c 2006-10-16
10:54:09.000000000 +0900
@@ -224,7 +224,7 @@
void __kprobes arch_remove_kprobe(struct kprobe *p)
{
mutex_lock(&kprobe_mutex);
- free_insn_slot(p->ainsn.insn);
+ free_insn_slot(p->ainsn.insn, 0);
mutex_unlock(&kprobe_mutex);
}
Index: linux-2.6.19-rc1-mm1/include/linux/kprobes.h
===================================================================
--- linux-2.6.19-rc1-mm1.orig/include/linux/kprobes.h 2006-10-16
10:40:02.000000000 +0900
+++ linux-2.6.19-rc1-mm1/include/linux/kprobes.h 2006-10-16
21:43:07.000000000 +0900
@@ -165,7 +165,7 @@
extern int arch_init_kprobes(void);
extern void show_registers(struct pt_regs *regs);
extern kprobe_opcode_t *get_insn_slot(void);
-extern void free_insn_slot(kprobe_opcode_t *slot);
+extern void free_insn_slot(kprobe_opcode_t *slot, int dirty);
extern void kprobes_inc_nmissed_count(struct kprobe *p);
/* Get the kprobe at this addr (if any) - called with preemption disabled */