linux/arch/i386/oprofile/backtrace.c
Tong Li 23332c2e9d [PATCH] OProfile: fixed x86_64 incorrect kernel call graphs
Fix the problem in kernel 2.6.15.1 (and early versions) that OProfile on
x86_64 does not correctly collect the stack traces for kernel functions.

The original code in valid_kernel_stack() in arch/i386/oprofile/backtrace.c
assumes that the frame pointer (headaddr) should be greater than stack
(i.e., regs).

This assumption is wrong for x86_64 because NMIs in x86_64 use a seperate
stack different from the kernel stack.  Therefore, the variable stack now
points to some location on the NMI stack, which turns out to be at a higher
address than the frame pointer (headaddr) on the kernel stack.  The correct
comparison here should be between headaddr and regs->rsp for x86_64.

Signed-off-by: Tong Li <tong.n.li@intel.com>
Cc: John Levon <levon@movementarian.org>
Cc: Philippe Elie <phil.el@wanadoo.fr>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-03 08:32:04 -08:00

115 lines
3.0 KiB
C

/**
* @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>
#include <asm/uaccess.h>
struct frame_head {
struct frame_head * ebp;
unsigned long ret;
} __attribute__((packed));
static struct frame_head *
dump_backtrace(struct frame_head * head)
{
struct frame_head bufhead[2];
/* Also check accessibility of one struct frame_head beyond */
if (!access_ok(VERIFY_READ, head, sizeof(bufhead)))
return NULL;
if (__copy_from_user_inatomic(bufhead, head, sizeof(bufhead)))
return NULL;
oprofile_add_trace(bufhead[0].ret);
/* frame pointers should strictly progress back up the stack
* (towards higher addresses) */
if (head >= bufhead[0].ebp)
return NULL;
return bufhead[0].ebp;
}
/*
* | | /\ Higher addresses
* | |
* --------------- stack base (address of current_thread_info)
* | thread info |
* . .
* | stack |
* --------------- saved regs->ebp value if valid (frame_head address)
* . .
* --------------- saved regs->rsp value if x86_64
* | |
* --------------- struct pt_regs * stored on stack if 32-bit
* | |
* . .
* | |
* --------------- %esp
* | |
* | | \/ Lower addresses
*
* Thus, regs (or regs->rsp for x86_64) <-> stack base restricts the
* valid(ish) ebp values. Note: (1) for x86_64, NMI and several other
* exceptions use special stacks, maintained by the interrupt stack table
* (IST). These stacks are set up in trap_init() in
* arch/x86_64/kernel/traps.c. Thus, for x86_64, regs now does not point
* to the kernel stack; instead, it points to some location on the NMI
* stack. On the other hand, regs->rsp is the stack pointer saved when the
* NMI occurred. (2) For 32-bit, regs->esp is not valid because the
* processor does not save %esp on the kernel stack when interrupts occur
* in the kernel mode.
*/
#ifdef CONFIG_FRAME_POINTER
static int valid_kernel_stack(struct frame_head * head, struct pt_regs * regs)
{
unsigned long headaddr = (unsigned long)head;
#ifdef CONFIG_X86_64
unsigned long stack = (unsigned long)regs->rsp;
#else
unsigned long stack = (unsigned long)regs;
#endif
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;
#ifdef CONFIG_X86_64
head = (struct frame_head *)regs->rbp;
#else
head = (struct frame_head *)regs->ebp;
#endif
if (!user_mode_vm(regs)) {
while (depth-- && valid_kernel_stack(head, regs))
head = dump_backtrace(head);
return;
}
while (depth-- && head)
head = dump_backtrace(head);
}