[PATCH 3/11] oprofile: i386 support for stack trace sampling

From: Greg Banks
Date: Tue Nov 09 2004 - 06:20:27 EST


--
Greg Banks, R&D Software Engineer, SGI Australian Software Group.
I don't speak for SGI.

oprofile i386 arch updates, including some internal
API changes and support for stack trace sampling.

Signed-off-by: John Levon <levon@xxxxxxxxxxxxxxxxx>
Signed-off-by: Greg Banks <gnb@xxxxxxxxxxxxxxxxx>
---

arch/i386/oprofile/Makefile | 2
arch/i386/oprofile/backtrace.c | 118 +++++++++++++++++++++++++++++++++++
arch/i386/oprofile/init.c | 12 ++-
arch/i386/oprofile/nmi_int.c | 47 ++++++-------
arch/i386/oprofile/nmi_timer_int.c | 16 +---
arch/i386/oprofile/op_model_athlon.c | 9 --
arch/i386/oprofile/op_model_p4.c | 9 --
arch/i386/oprofile/op_model_ppro.c | 9 --
arch/i386/oprofile/op_x86_model.h | 5 -
9 files changed, 164 insertions(+), 63 deletions(-)

Index: linux/arch/i386/oprofile/op_model_ppro.c
===================================================================
--- linux.orig/arch/i386/oprofile/op_model_ppro.c 2004-10-19 07:54:38.%N +1000
+++ linux/arch/i386/oprofile/op_model_ppro.c 2004-11-06 01:18:32.%N +1100
@@ -85,19 +85,16 @@ static void ppro_setup_ctrs(struct op_ms
}


-static int ppro_check_ctrs(unsigned int const cpu,
- struct op_msrs const * const msrs,
- struct pt_regs * const regs)
+static int ppro_check_ctrs(struct pt_regs * const regs,
+ struct op_msrs const * const msrs)
{
unsigned int low, high;
int i;
- unsigned long eip = profile_pc(regs);
- int is_kernel = !user_mode(regs);

for (i = 0 ; i < NUM_COUNTERS; ++i) {
CTR_READ(low, high, msrs, i);
if (CTR_OVERFLOWED(low)) {
- oprofile_add_sample(eip, is_kernel, i, cpu);
+ oprofile_add_sample(regs, i);
CTR_WRITE(reset_value[i], msrs, i);
}
}
Index: linux/arch/i386/oprofile/op_model_p4.c
===================================================================
--- linux.orig/arch/i386/oprofile/op_model_p4.c 2004-10-19 07:53:11.%N +1000
+++ linux/arch/i386/oprofile/op_model_p4.c 2004-11-06 01:18:32.%N +1100
@@ -619,14 +619,11 @@ static void p4_setup_ctrs(struct op_msrs
}


-static int p4_check_ctrs(unsigned int const cpu,
- struct op_msrs const * const msrs,
- struct pt_regs * const regs)
+static int p4_check_ctrs(struct pt_regs * const regs,
+ struct op_msrs const * const msrs)
{
unsigned long ctr, low, high, stag, real;
int i;
- unsigned long eip = profile_pc(regs);
- int is_kernel = !user_mode(regs);

stag = get_stagger();

@@ -657,7 +654,7 @@ static int p4_check_ctrs(unsigned int co
CCCR_READ(low, high, real);
CTR_READ(ctr, high, real);
if (CCCR_OVF_P(low) || CTR_OVERFLOW_P(ctr)) {
- oprofile_add_sample(eip, is_kernel, i, cpu);
+ oprofile_add_sample(regs, i);
CTR_WRITE(reset_value[i], real);
CCCR_CLEAR_OVF(low);
CCCR_WRITE(low, high, real);
Index: linux/arch/i386/oprofile/nmi_int.c
===================================================================
--- linux.orig/arch/i386/oprofile/nmi_int.c 2004-11-06 01:11:55.%N +1100
+++ linux/arch/i386/oprofile/nmi_int.c 2004-11-06 01:19:22.%N +1100
@@ -84,7 +84,7 @@ static void __exit exit_driverfs(void)

static int nmi_callback(struct pt_regs * regs, int cpu)
{
- return model->check_ctrs(cpu, &cpu_msrs[cpu], regs);
+ return model->check_ctrs(regs, &cpu_msrs[cpu]);
}


@@ -300,16 +300,7 @@ static int nmi_create_files(struct super
}


-struct oprofile_operations nmi_ops = {
- .create_files = nmi_create_files,
- .setup = nmi_setup,
- .shutdown = nmi_shutdown,
- .start = nmi_start,
- .stop = nmi_stop
-};
-
-
-static int __init p4_init(void)
+static int __init p4_init(char ** cpu_type)
{
__u8 cpu_model = boot_cpu_data.x86_model;

@@ -317,18 +308,18 @@ static int __init p4_init(void)
return 0;

#ifndef CONFIG_SMP
- nmi_ops.cpu_type = "i386/p4";
+ *cpu_type = "i386/p4";
model = &op_p4_spec;
return 1;
#else
switch (smp_num_siblings) {
case 1:
- nmi_ops.cpu_type = "i386/p4";
+ *cpu_type = "i386/p4";
model = &op_p4_spec;
return 1;

case 2:
- nmi_ops.cpu_type = "i386/p4-ht";
+ *cpu_type = "i386/p4-ht";
model = &op_p4_ht2_spec;
return 1;
}
@@ -340,7 +331,7 @@ static int __init p4_init(void)
}


-static int __init ppro_init(void)
+static int __init ppro_init(char ** cpu_type)
{
__u8 cpu_model = boot_cpu_data.x86_model;

@@ -348,13 +339,13 @@ static int __init ppro_init(void)
return 0;

if (cpu_model == 9) {
- nmi_ops.cpu_type = "i386/p6_mobile";
+ *cpu_type = "i386/p6_mobile";
} else if (cpu_model > 5) {
- nmi_ops.cpu_type = "i386/piii";
+ *cpu_type = "i386/piii";
} else if (cpu_model > 2) {
- nmi_ops.cpu_type = "i386/pii";
+ *cpu_type = "i386/pii";
} else {
- nmi_ops.cpu_type = "i386/ppro";
+ *cpu_type = "i386/ppro";
}

model = &op_ppro_spec;
@@ -364,10 +355,11 @@ static int __init ppro_init(void)
/* in order to get driverfs right */
static int using_nmi;

-int __init nmi_init(struct oprofile_operations ** ops)
+int __init nmi_init(struct oprofile_operations * ops)
{
__u8 vendor = boot_cpu_data.x86_vendor;
__u8 family = boot_cpu_data.x86;
+ char *cpu_type;

if (!cpu_has_apic)
return -ENODEV;
@@ -381,13 +373,13 @@ int __init nmi_init(struct oprofile_oper
return -ENODEV;
case 6:
model = &op_athlon_spec;
- nmi_ops.cpu_type = "i386/athlon";
+ cpu_type = "i386/athlon";
break;
case 0xf:
model = &op_athlon_spec;
/* Actually it could be i386/hammer too, but give
user space an consistent name. */
- nmi_ops.cpu_type = "x86-64/hammer";
+ cpu_type = "x86-64/hammer";
break;
}
break;
@@ -396,13 +388,13 @@ int __init nmi_init(struct oprofile_oper
switch (family) {
/* Pentium IV */
case 0xf:
- if (!p4_init())
+ if (!p4_init(&cpu_type))
return -ENODEV;
break;

/* A P6-class processor */
case 6:
- if (!ppro_init())
+ if (!ppro_init(&cpu_type))
return -ENODEV;
break;

@@ -417,7 +409,12 @@ int __init nmi_init(struct oprofile_oper

init_driverfs();
using_nmi = 1;
- *ops = &nmi_ops;
+ ops->create_files = nmi_create_files;
+ ops->setup = nmi_setup;
+ ops->shutdown = nmi_shutdown;
+ ops->start = nmi_start;
+ ops->stop = nmi_stop;
+ ops->cpu_type = cpu_type;
printk(KERN_INFO "oprofile: using NMI interrupt.\n");
return 0;
}
Index: linux/arch/i386/oprofile/nmi_timer_int.c
===================================================================
--- linux.orig/arch/i386/oprofile/nmi_timer_int.c 2004-10-19 07:53:45.%N +1000
+++ linux/arch/i386/oprofile/nmi_timer_int.c 2004-11-06 01:18:32.%N +1100
@@ -20,9 +20,7 @@

static int nmi_timer_callback(struct pt_regs * regs, int cpu)
{
- unsigned long eip = instruction_pointer(regs);
-
- oprofile_add_sample(eip, !user_mode(regs), 0, cpu);
+ oprofile_add_sample(regs, 0);
return 1;
}

@@ -42,20 +40,16 @@ static void timer_stop(void)
}


-static struct oprofile_operations nmi_timer_ops = {
- .start = timer_start,
- .stop = timer_stop,
- .cpu_type = "timer"
-};
-
-int __init nmi_timer_init(struct oprofile_operations ** ops)
+int __init nmi_timer_init(struct oprofile_operations * ops)
{
extern int nmi_active;

if (nmi_active <= 0)
return -ENODEV;

- *ops = &nmi_timer_ops;
+ ops->start = timer_start;
+ ops->stop = timer_stop;
+ ops->cpu_type = "timer";
printk(KERN_INFO "oprofile: using NMI timer interrupt.\n");
return 0;
}
Index: linux/arch/i386/oprofile/init.c
===================================================================
--- linux.orig/arch/i386/oprofile/init.c 2004-10-19 07:54:32.%N +1000
+++ linux/arch/i386/oprofile/init.c 2004-11-06 01:18:32.%N +1100
@@ -15,22 +15,24 @@
* with the NMI mode driver.
*/

-extern int nmi_init(struct oprofile_operations ** ops);
-extern int nmi_timer_init(struct oprofile_operations **ops);
+extern int nmi_init(struct oprofile_operations * ops);
+extern int nmi_timer_init(struct oprofile_operations * ops);
extern void nmi_exit(void);
+extern void x86_backtrace(struct pt_regs * const regs, unsigned int depth);

-int __init oprofile_arch_init(struct oprofile_operations ** ops)
+
+void __init oprofile_arch_init(struct oprofile_operations * ops)
{
int ret = -ENODEV;
+
#ifdef CONFIG_X86_LOCAL_APIC
ret = nmi_init(ops);
#endif
-
#ifdef CONFIG_X86_IO_APIC
if (ret < 0)
ret = nmi_timer_init(ops);
#endif
- return ret;
+ ops->backtrace = x86_backtrace;
}


Index: linux/arch/i386/oprofile/op_x86_model.h
===================================================================
--- linux.orig/arch/i386/oprofile/op_x86_model.h 2004-10-19 07:53:43.%N +1000
+++ linux/arch/i386/oprofile/op_x86_model.h 2004-11-06 01:18:32.%N +1100
@@ -36,9 +36,8 @@ struct op_x86_model_spec {
unsigned int const num_controls;
void (*fill_in_addresses)(struct op_msrs * const msrs);
void (*setup_ctrs)(struct op_msrs const * const msrs);
- int (*check_ctrs)(unsigned int const cpu,
- struct op_msrs const * const msrs,
- struct pt_regs * const regs);
+ int (*check_ctrs)(struct pt_regs * const regs,
+ struct op_msrs const * const msrs);
void (*start)(struct op_msrs const * const msrs);
void (*stop)(struct op_msrs const * const msrs);
};
Index: linux/arch/i386/oprofile/Makefile
===================================================================
--- linux.orig/arch/i386/oprofile/Makefile 2004-10-19 07:55:06.%N +1000
+++ linux/arch/i386/oprofile/Makefile 2004-11-06 01:18:32.%N +1100
@@ -6,7 +6,7 @@ DRIVER_OBJS = $(addprefix ../../../drive
oprofilefs.o oprofile_stats.o \
timer_int.o )

-oprofile-y := $(DRIVER_OBJS) init.o
+oprofile-y := $(DRIVER_OBJS) init.o backtrace.o
oprofile-$(CONFIG_X86_LOCAL_APIC) += nmi_int.o op_model_athlon.o \
op_model_ppro.o op_model_p4.o
oprofile-$(CONFIG_X86_IO_APIC) += nmi_timer_int.o
Index: linux/arch/i386/oprofile/op_model_athlon.c
===================================================================
--- linux.orig/arch/i386/oprofile/op_model_athlon.c 2004-10-19 07:55:36.%N +1000
+++ linux/arch/i386/oprofile/op_model_athlon.c 2004-11-06 01:18:32.%N +1100
@@ -90,19 +90,16 @@ static void athlon_setup_ctrs(struct op_
}


-static int athlon_check_ctrs(unsigned int const cpu,
- struct op_msrs const * const msrs,
- struct pt_regs * const regs)
+static int athlon_check_ctrs(struct pt_regs * const regs,
+ struct op_msrs const * const msrs)
{
unsigned int low, high;
int i;
- unsigned long eip = profile_pc(regs);
- int is_kernel = !user_mode(regs);

for (i = 0 ; i < NUM_COUNTERS; ++i) {
CTR_READ(low, high, msrs, i);
if (CTR_OVERFLOWED(low)) {
- oprofile_add_sample(eip, is_kernel, i, cpu);
+ oprofile_add_sample(regs, i);
CTR_WRITE(reset_value[i], msrs, i);
}
}
Index: linux/arch/i386/oprofile/backtrace.c
===================================================================
--- /dev/null 1970-01-01 00:00:00.000000000 +0000
+++ linux/arch/i386/oprofile/backtrace.c 2004-11-06 01:18:32.%N +1100
@@ -0,0 +1,118 @@
+/**
+ * @file backtrace.c
+ *
+ * @remark Copyright 2002 OProfile authors
+ * @remark Read the file COPYING
+ *
+ * @author John Levon
+ * @author David Smith
+ */
+
+#include <linux/oprofile.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <asm/ptrace.h>
+
+struct frame_head {
+ struct frame_head * ebp;
+ unsigned long ret;
+} __attribute__((packed));
+
+
+static struct frame_head *
+dump_backtrace(struct frame_head * head)
+{
+ oprofile_add_trace(head->ret);
+
+ /* frame pointers should strictly progress back up the stack
+ * (towards higher addresses) */
+ if (head >= head->ebp)
+ return 0;
+
+ return head->ebp;
+}
+
+
+#ifdef CONFIG_X86_4G
+/* With a 4G kernel/user split, user pages are not directly
+ * accessible from the kernel, so don't try
+ */
+static int pages_present(struct frame_head * head)
+{
+ return 0;
+}
+#else
+/* check that the page(s) containing the frame head are present */
+static int pages_present(struct frame_head * head)
+{
+ struct mm_struct * mm = current->mm;
+
+ /* FIXME: only necessary once per page */
+ if (!check_user_page_readable(mm, (unsigned long)head))
+ return 0;
+
+ return check_user_page_readable(mm, (unsigned long)(head + 1));
+}
+#endif /* CONFIG_X86_4G */
+
+
+/*
+ * | | /\ Higher addresses
+ * | |
+ * --------------- stack base (address of current_thread_info)
+ * | thread info |
+ * . .
+ * | stack |
+ * --------------- saved regs->ebp value if valid (frame_head address)
+ * . .
+ * --------------- struct pt_regs stored on stack (struct pt_regs *)
+ * | |
+ * . .
+ * | |
+ * --------------- %esp
+ * | |
+ * | | \/ Lower addresses
+ *
+ * Thus, &pt_regs <-> stack base restricts the valid(ish) ebp values
+ */
+#ifdef CONFIG_FRAME_POINTER
+static int valid_kernel_stack(struct frame_head * head, struct pt_regs * regs)
+{
+ unsigned long headaddr = (unsigned long)head;
+ unsigned long stack = (unsigned long)regs;
+ unsigned long stack_base = (stack & ~(THREAD_SIZE - 1)) + THREAD_SIZE;
+
+ return headaddr > stack && headaddr < stack_base;
+}
+#else
+/* without fp, it's just junk */
+static int valid_kernel_stack(struct frame_head * head, struct pt_regs * regs)
+{
+ return 0;
+}
+#endif
+
+
+void
+x86_backtrace(struct pt_regs * const regs, unsigned int depth)
+{
+ struct frame_head * head = (struct frame_head *)regs->ebp;
+
+ if (!user_mode(regs)) {
+ while (depth-- && valid_kernel_stack(head, regs))
+ head = dump_backtrace(head);
+ return;
+ }
+
+#ifdef CONFIG_SMP
+ if (!spin_trylock(&current->mm->page_table_lock))
+ return;
+#endif
+
+ while (depth-- && head && pages_present(head))
+ head = dump_backtrace(head);
+
+#ifdef CONFIG_SMP
+ spin_unlock(&current->mm->page_table_lock);
+#endif
+}