Skip to contentSkip to conversation

Add BUG_XX() debugging options

PATCH Superseded Jeffrey Merkey 3 participants 8 messages Feb 1, 2016 – Feb 2, 2016

Subsystems (4)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/4 Add BUG_XX() debugging options

This patch series adds config options which can be set during compile to direct the compiler to output a breakpoint instruction anywhere a BUG() or WARN() macro has been placed in the kernel to trigger the system to enter a debugger if a bug is detected by the system. Use of this compile time option also allows conditional breakpoints to be set in the kernel with these currently used macros.

This addition is extremely useful for debugging hard and soft lockups real time and quickly from a console debugger, and other areas of the kernel.

---
 arch/x86/include/asm/bug.h | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)
arch/x86/include/asm/bug.h +16 -1
index ba38ebb..df26c2b 100644
@@ -11,6 +11,13 @@
 # define __BUG_C0	"2:\t.long 1b - 2b, %c0 - 2b\n"
 #endif
 
+#ifdef CONFIG_DEBUG_BUG
+#define BUG()							\
+do {								\
+	asm volatile("int3");					\
+	unreachable();						\
+} while (0)
+#else
 #define BUG()							\
 do {								\
 	asm volatile("1:\tud2\n"				\
@@ -23,7 +30,14 @@ do {								\
 		     "i" (sizeof(struct bug_entry)));		\
 	unreachable();						\
 } while (0)
-
+#endif
+#else
+#ifdef CONFIG_DEBUG_BUG
+#define BUG()							\
+do {								\
+	asm volatile("int3");					\
+	unreachable();						\
+} while (0)
 #else
 #define BUG()							\
 do {								\
@@ -31,6 +45,7 @@ do {								\
 	unreachable();						\
 } while (0)
 #endif
+#endif
 
 #include <asm-generic/bug.h>
 
Signature · 1 line
1.8.3.1
#1 lore raw

PATCH v4 2/4 Add WARN_XX() debugging options

This patch series adds config options which can be set during compile to direct the compiler to output a breakpoint instruction anywhere a BUG() or WARN() macro has been placed in the kernel to trigger the system to enter a debugger if a bug is detected by the system. Use of this compile time option also allows conditional breakpoints to be set in the kernel with these currently used macros.

This addition is extremely useful for debugging hard and soft lockups real time and quickly from a console debugger, and other areas of the kernel.

---
 include/asm-generic/bug.h | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)
include/asm-generic/bug.h +13 -1
index 630dd23..ed86cd2 100644
@@ -90,14 +90,26 @@ extern void warn_slowpath_null(const char *file, const int line);
 })
 #endif
 
+#ifdef CONFIG_DEBUG_WARN
 #ifndef WARN
-#define WARN(condition, format...) ({						\
+#define WARN(condition, format...) ({					\
 	int __ret_warn_on = !!(condition);				\
 	if (unlikely(__ret_warn_on))					\
 		__WARN_printf(format);					\
+	BUG();								\
 	unlikely(__ret_warn_on);					\
 })
 #endif
+#else
+#ifndef WARN
+#define WARN(condition, format...) ({					\
+	int __ret_warn_on = !!(condition);				\
+	if (unlikely(__ret_warn_on))					\
+		__WARN_printf(format);					\
+	unlikely(__ret_warn_on);					\
+})
+#endif
+#endif
 
 #define WARN_TAINT(condition, taint, format...) ({			\
 	int __ret_warn_on = !!(condition);				\
Signature · 1 line
1.8.3.1
#2 lore raw

PATCH v4 3/4 Add BUG_XX() and WARN_XX() debugging options

In reply to Jeffrey Merkey

This patch series adds config options which can be set during compile to direct the compiler to output a breakpoint instruction anywhere a BUG() or WARN() macro has been placed in the kernel to trigger the system to enter a debugger if a bug is detected by the system. Use of this compile time option also allows conditional breakpoints to be set in the kernel with these currently used macros.

This addition is extremely useful for debugging hard and soft lockups real time and quickly from a console debugger, and other areas of the kernel.

---
 lib/Kconfig.debug | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
lib/Kconfig.debug +22 -0
index 041f995..81b58cd 100644
@@ -326,6 +326,28 @@ config SECTION_MISMATCH_WARN_ONLY
 # is preferred to always offer frame pointers as a config
 # option on the architecture (regardless of KERNEL_DEBUG):
 #
+config DEBUG_BUG
+	bool "Set Breakpoint for Kernel BUG() macros"
+	depends on DEBUG_KERNEL && !S390
+	help
+	  Say Y here to enable the kernel to emit a breakpoint instruction
+	  anywhere a BUG_XX() macro has been placed in the system kernel code
+	  intended to catch bugs.
+
+	  This option is used for triggering the OS to try enter a debugger
+	  if one is presently installed on the system.
+
+config DEBUG_WARN
+	bool "Set Breakpoint for Kernel WARN() macros"
+	depends on DEBUG_KERNEL && !S390 && DEBUG_BUG
+	help
+	  Say Y here to enable the kernel to emit a breakpoint instruction
+	  anywhere a WARN_XX() macro has been placed in the system kernel code
+	  intended to catch bugs.
+
+	  This option is used for triggering the OS to try enter a debugger
+	  if one is presently installed on the system.
+
 config ARCH_WANT_FRAME_POINTERS
 	bool
 	help
Signature · 1 line
1.8.3.1
#3 lore raw

PATCH v4 4/4 pr_emerg add WARN_XX() debugger options

In reply to Jeffrey Merkey

This patch series adds config options which can be set during compile to direct the compiler to output a breakpoint instruction anywhere a BUG() or WARN() macro has been placed in the kernel to trigger the system to enter a debugger if a bug is detected by the system. Use of this compile time option also allows conditional breakpoints to be set in the kernel with these currently used macros.

This addition is extremely useful for debugging hard and soft lockups real time and quickly from a console debugger, and other areas of the kernel.

---
 include/linux/printk.h | 8 ++++++++
 1 file changed, 8 insertions(+)
include/linux/printk.h +8 -0
index 51dd6b8..dcfb270 100644
@@ -252,8 +252,16 @@ extern asmlinkage void dump_stack(void) __cold;
  * and other debug macros are compiled out unless either DEBUG is defined
  * or CONFIG_DYNAMIC_DEBUG is set.
  */
+#ifdef CONFIG_DEBUG_WARN
+#define pr_emerg(fmt, ...) \
+({								\
+	printk(KERN_EMERG pr_fmt(fmt), ##__VA_ARGS__);		\
+	BUG();							\
+})
+#else
 #define pr_emerg(fmt, ...) \
 	printk(KERN_EMERG pr_fmt(fmt), ##__VA_ARGS__)
+#endif
 #define pr_alert(fmt, ...) \
 	printk(KERN_ALERT pr_fmt(fmt), ##__VA_ARGS__)
 #define pr_crit(fmt, ...) \
Signature · 1 line
1.8.3.1
#4 lore raw

Re: [PATCH v4 4/4] pr_emerg add WARN_XX() debugger options

Quoted 9 lines · On Mon, 1 Feb 2016 16:35:52 -0700 Jeffrey Merkey <[email protected]> wrote:
---
 include/linux/printk.h | 8 ++++++++
 1 file changed, 8 insertions(+)
include/linux/printk.h +0 -0
index 51dd6b8..dcfb270 100644

include/linux/printk.h extern asmlinkage void dump_stack(void) __cold; · line 259

Quoted 8 lines
include/linux/printk.h +5 -0
@@ -252,8 +252,16 @@ extern asmlinkage void dump_stack(void) __cold;
  * and other debug macros are compiled out unless either DEBUG is defined
  * or CONFIG_DYNAMIC_DEBUG is set.
  */
+#ifdef CONFIG_DEBUG_WARN
+#define pr_emerg(fmt, ...) \
+({								\
+	printk(KERN_EMERG pr_fmt(fmt), ##__VA_ARGS__);		\
+	BUG();							\

This look rather heavy handed for a debug feature. This will crash any kernel on a pr-emerg(), similar to a panic on warning. Not only that, for cases that cal pr_emerg() more than once, it crashes on the first instance.

What about adding a condition that can be set set in /proc/sys/kernel/ A file called something like crash_on_print_emerg ?

	BUG_ON(crash_on_pr_emerg);

-- Steve

Quoted 8 lines
include/linux/printk.h +3 -0
+})
+#else
 #define pr_emerg(fmt, ...) \
 	printk(KERN_EMERG pr_fmt(fmt), ##__VA_ARGS__)
+#endif
 #define pr_alert(fmt, ...) \
 	printk(KERN_ALERT pr_fmt(fmt), ##__VA_ARGS__)
 #define pr_crit(fmt, ...) \
#5 lore raw

Re: [PATCH v4 4/4] pr_emerg add WARN_XX() debugger options

Quoted 26 lines · Steven Rostedt
Quoted 13 lines · On Mon, 1 Feb 2016 16:35:52 -0700 Jeffrey Merkey <[email protected]> wrote:
---
 include/linux/printk.h | 8 ++++++++
 1 file changed, 8 insertions(+)
include/linux/printk.h +0 -0
index 51dd6b8..dcfb270 100644
@@ -252,8 +252,16 @@ extern asmlinkage void dump_stack(void) __cold;
  * and other debug macros are compiled out unless either DEBUG is

defined
* or CONFIG_DYNAMIC_DEBUG is set.
*/

include/linux/printk.h extern asmlinkage void dump_stack(void) __cold; · line 259

Quoted 5 lines
include/linux/printk.h +5 -0
+#ifdef CONFIG_DEBUG_WARN
+#define pr_emerg(fmt, ...) \
+({								\
+	printk(KERN_EMERG pr_fmt(fmt), ##__VA_ARGS__);		\
+	BUG();							\

This look rather heavy handed for a debug feature. This will crash any kernel on a pr-emerg(), similar to a panic on warning. Not only that, for cases that cal pr_emerg() more than once, it crashes on the first instance.

Quoted 7 lines

What about adding a condition that can be set set in /proc/sys/kernel/ A file called something like crash_on_print_emerg ?

	BUG_ON(crash_on_pr_emerg);

-- Steve

If a debugger is loaded it will not crash, just enter the debugger. But yes, it will int3 if set and no debugger has been loaded to handle the int3 condition. Hmmm. Maybe its better just to skip calling pr_emerg and put this logic as a single call somewhere else.

Jeff

#6 lore raw

Re: [PATCH v4 4/4] pr_emerg add WARN_XX() debugger options

Quoted 4 lines · Jeffrey Merkey

If a debugger is loaded it will not crash, just enter the debugger. But yes, it will int3 if set and no debugger has been loaded to handle the int3 condition. Hmmm. Maybe its better just to skip calling pr_emerg and put this logic as a single call somewhere else.

What's the point? If you have a debugger loaded, then surely you can
just set a breakpoint anywhere you like using whatever user interface the debugger provides for setting breakpoints. You can actually set any number of software breakpoints you like wherever you like, depending on what you actually want to debug. I fail to see why it would have to be prearranged within the kernel -- do you have a configuration where you run the kernel from ROM by any chance?

Maciej

#7 lore raw

Re: [PATCH v4 4/4] pr_emerg add WARN_XX() debugger options

Quoted 15 lines · Maciej W. Rozycki
Quoted 4 lines · Jeffrey Merkey

If a debugger is loaded it will not crash, just enter the debugger. But yes, it will int3 if set and no debugger has been loaded to handle the int3 condition. Hmmm. Maybe its better just to skip calling pr_emerg and put this logic as a single call somewhere else.

What's the point? If you have a debugger loaded, then surely you can
just set a breakpoint anywhere you like using whatever user interface the debugger provides for setting breakpoints. You can actually set any number of software breakpoints you like wherever you like, depending on what you actually want to debug. I fail to see why it would have to be prearranged within the kernel -- do you have a configuration where you run the kernel from ROM by any chance?

Quoted 2 lines

Maciej

No ROM here. I resubmitted this as series 5 and dropped the change to pr_emerg for the very reasons you stated. I verified what you said was correct by reviewing several code paths and its not a good place for that. Folks expect this function to act like a printk on steroids, not as a terminal placement of a BUG().

Jeff