Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 23332c2e authored by Tong Li's avatar Tong Li Committed by Linus Torvalds
Browse files

[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: default avatarTong Li <tong.n.li@intel.com>
Cc: John Levon <levon@movementarian.org>
Cc: Philippe Elie <phil.el@wanadoo.fr>
Signed-off-by: default avatarAndrew Morton <akpm@osdl.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@osdl.org>
parent 81459169

arch/i386/oprofile/backtrace.c

+17 −2
Original line number Diff line number Diff line
@@ -49,7 +49,9 @@ dump_backtrace(struct frame_head * head) 
 * |    stack    |

 * --------------- saved regs->ebp value if valid (frame_head address)

 * .             .

 * --------------- struct pt_regs stored on stack (struct pt_regs *)

 * --------------- saved regs->rsp value if x86_64

 * |             |

 * --------------- struct pt_regs * stored on stack if 32-bit

 * |             |

 * .             .

 * |             |
@@ -57,13 +59,26 @@ dump_backtrace(struct frame_head * head) 
 * |             |

 * |             | \/ Lower addresses

 *

 * Thus, &pt_regs <-> stack base restricts the valid(ish) ebp values

 * 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;