Skip to contentSkip to conversation

backtracer stuck in apic_timer_interrupt

DISCUSSION Jan Beulich 2 participants 2 messages Aug 15, 2006 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw
Quoted 2 lines

Do you know why the backtracer gets stuck in apic_timer_interrupt? Somehow it didn't manage to go over to the process stack, but it

should

Quoted 1 line

have.

No, I don't, I just verified that the annotations are correct (in that unwinds can pass this function). If you see this again, I'd need the raw
stack dump to compare. However, while doing this I also saw a case where it wouldn't properly unwind (but one level up from apic_timer_interrupt), however in this case it was expectedly getting stuck because the interrupt had hit a procedure that already entered it
epilogue, which gcc fails to annotate.

Further, while doing this, I found myself ending up with an unwound stack trace followed by a legacy one, but without any separator indicating so. Therefore I looked at the fallback logic you recently changed and found that it was still not quite right. Hence I created below patch (for both i386 and x86-64).

Jan

The unwinder fallback logic still had potential for falling through to the legacy stack trace code without printing an indication (at once serving as a separator) of this.

Further, the stack pointer retrieval for the fallback should be as restrictive as possible (in order to avoid having the legacy stack tracer try to access invalid memory). The patch tightens that, but this could certainly be further improved.

Also making the call_trace command line option now conditional upon CONFIG_STACK_UNWIND (as it's meaningless otherwise).

--- linux-2.6.18-rc4/arch/i386/kernel/traps.c 2006-08-15
11:29:35.000000000 +0200
+++
2.6.18-rc4-unwind-x86-fallback/arch/i386/kernel/traps.c 2006-08-15
10:52:43.000000000 +0200
@@ -92,7 +92,11 @@ asmlinkage void spurious_interrupt_bug(v
asmlinkage void machine_check(void);

static int kstack_depth_to_print = 24;
+#ifdef CONFIG_STACK_UNWIND
static int call_trace = 1;
+#else
+#define call_trace (-1)
+#endif
ATOMIC_NOTIFIER_HEAD(i386die_chain);

int register_die_notifier(struct notifier_block *nb)
@@ -187,22 +191,21 @@ static void show_trace_log_lvl(struct ta
if (unwind_init_blocked(&info, task) == 0)
unw_ret = show_trace_unwind(&info,
log_lvl);
}
- if (unw_ret > 0 && !arch_unw_user_mode(&info)) {
-#ifdef CONFIG_STACK_UNWIND
- print_symbol("DWARF2 unwinder stuck at %s\n",
- UNW_PC(&info));
- if (call_trace == 1) {
- printk("Leftover inexact
backtrace:\n");
- if (UNW_SP(&info))
+ if (unw_ret > 0) {
+ if (call_trace == 1 &&
!arch_unw_user_mode(&info)) {
+ print_symbol("DWARF2 unwinder stuck at
%s\n",
+ UNW_PC(&info));
+ if (UNW_SP(&info) >= PAGE_OFFSET) {
+ printk("Leftover inexact
backtrace:\n");
stack = (void *)UNW_SP(&info);
- } else if (call_trace > 1)
+ } else
+ printk("Full inexact backtrace
again:\n");
+ } else if (call_trace >= 1)
return;
else
printk("Full inexact backtrace
again:\n");
-#else
+ } else
printk("Inexact backtrace:\n");
-#endif
- }
}

if (task == current) {
@@ -1241,6 +1244,7 @@ static int __init kstack_setup(char *s)
}
__setup("kstack=", kstack_setup);

+#ifdef CONFIG_STACK_UNWIND
static int __init call_trace_setup(char *s)
{
if (strcmp(s, "old") == 0)
@@ -1254,3 +1258,4 @@ static int __init call_trace_setup(char
return 1;
}
__setup("call_trace=", call_trace_setup);
+#endif
--- linux-2.6.18-rc4/arch/x86_64/kernel/traps.c 2006-08-15
11:29:41.000000000 +0200
+++
2.6.18-rc4-unwind-x86-fallback/arch/x86_64/kernel/traps.c 2006-08-15
10:49:34.000000000 +0200
@@ -107,7 +107,11 @@ static inline void preempt_conditional_c
}

static int kstack_depth_to_print = 12;
+#ifdef CONFIG_STACK_UNWIND
static int call_trace = 1;
+#else
+#define call_trace (-1)
+#endif

#ifdef CONFIG_KALLSYMS

 # include <linux/kallsyms.h>
@@ -274,21 +278,21 @@ void show_trace(struct task_struct *tsk,
 			if (unwind_init_blocked(&info, tsk) == 0)
 				unw_ret = show_trace_unwind(&info,
NULL);
 		}
-		if (unw_ret > 0 && !arch_unw_user_mode(&info)) {
-#ifdef CONFIG_STACK_UNWIND
-			unsigned long rip = info.regs.rip;
-			print_symbol("DWARF2 unwinder stuck at %s\n",
rip);
-			if (call_trace == 1) {
-				printk("Leftover inexact
backtrace:\n");
-				stack = (unsigned long *)info.regs.rsp;
-			} else if (call_trace > 1)
+		if (unw_ret > 0) {
+			if (call_trace == 1 &&
!arch_unw_user_mode(&info)) {
+				print_symbol("DWARF2 unwinder stuck at
%s\n",
+					     UNW_PC(&info));
+				if ((long)UNW_SP(&info) < 0) {
+					printk("Leftover inexact
backtrace:\n");
+					stack = (unsigned long
*)UNW_SP(&info);
+				} else
+					printk("Full inexact backtrace
again:\n");
+			} else if (call_trace >= 1)
 				return;
 			else
 				printk("Full inexact backtrace
again:\n");
-#else
+		} else
 			printk("Inexact backtrace:\n");
-#endif
-		}
 	}

/*
@@ -1120,6 +1124,7 @@ static int __init kstack_setup(char *s)
}
__setup("kstack=", kstack_setup);

+#ifdef CONFIG_STACK_UNWIND
static int __init call_trace_setup(char *s)
{
if (strcmp(s, "old") == 0)
@@ -1133,3 +1138,4 @@ static int __init call_trace_setup(char
return 1;
}
__setup("call_trace=", call_trace_setup);
+#endif
--- linux-2.6.18-rc4/include/asm-i386/unwind.h 2006-08-15
11:30:00.000000000 +0200
+++
2.6.18-rc4-unwind-x86-fallback/include/asm-i386/unwind.h 2006-08-15
09:36:19.000000000 +0200
@@ -87,6 +87,7 @@ static inline int arch_unw_user_mode(con
#else

#define UNW_PC(frame) ((void)(frame), 0)
+#define UNW_SP(frame) ((void)(frame), 0)

static inline int arch_unw_user_mode(const void *info) {
--- linux-2.6.18-rc4/include/asm-x86_64/unwind.h 2006-08-15
11:30:02.000000000 +0200
+++
2.6.18-rc4-unwind-x86-fallback/include/asm-x86_64/unwind.h 2006-08-15
09:36:19.000000000 +0200
@@ -95,6 +95,7 @@ static inline int arch_unw_user_mode(con
#else

#define UNW_PC(frame) ((void)(frame), 0)
+#define UNW_SP(frame) ((void)(frame), 0)

static inline int arch_unw_user_mode(const void *info) {

#2 lore raw
Quoted 3 lines

Also making the call_trace command line option now conditional upon CONFIG_STACK_UNWIND (as it's meaningless otherwise).

Thanks added (even though it was some effort because the patch was word wrapped and that code had changed in my tree)

-Andi