Skip to contentSkip to conversation

changes about Call Trace:

PATCH In review Akinobu Mita 6 participants 21 messages Jan 16, 2006 – Jan 17, 2006

lore
Share & tools
lore
#0 lore raw

cover changes about Call Trace:

Hello,

I realized two things when I was porting my small script oops2line to x86-64. (read call trace, find correspondance modules, calculate addr and addr2line)

If I'm missing something, please let me know.

a) On x86-64 we get different Call Trace format than other architectures when we get oops or press SysRq-t:

<ffffffffa008ef6c>{:jbd:kjournald+1030}

There is a architecture independent function print_symbol(). How about using it on x86-64? But it changes to:

[<ffffffffa008ef6c>] kjournald+0x406/0x578 [jbd]

b) I can't find useful usage for the symbol size in print_symbol(). And symbolsize seems to be fixed when vmlinux or modules are compiled. So we can calculate it from vmlinux or modules.
How about removing the field of symbolsize in print_symbol()?

[<ffffffffa008ef6c>] kjournald+0x406 [jbd]

#1 lore raw

PATCH 1/3 makes print_symbol() return int

This patch makes print_symbol() return the number of characters printed.

----

 include/linux/kallsyms.h |   17 ++++++++++-------
 kernel/kallsyms.c        |    4 ++--
 2 files changed, 12 insertions(+), 9 deletions(-)
include/linux/kallsyms.h +10 -7
@@ -20,7 +20,7 @@ const char *kallsyms_lookup(unsigned lon
 			    char **modname, char *namebuf);
 
 /* Replace "%s" in format with address, if found */
-extern void __print_symbol(const char *fmt, unsigned long address);
+extern int __print_symbol(const char *fmt, unsigned long address);
 
 #else /* !CONFIG_KALLSYMS */
 
@@ -38,7 +38,10 @@ static inline const char *kallsyms_looku
 }
 
 /* Stupid that this does nothing, but I didn't create this mess. */
-#define __print_symbol(fmt, addr)
+static inline int __print_symbol(const char *fmt, unsigned long addr)
+{
+	return 0;
+}
 #endif /*CONFIG_KALLSYMS*/
 
 /* This macro allows us to keep printk typechecking */
@@ -58,10 +61,10 @@ do {						\
 #define print_fn_descriptor_symbol(fmt, addr) print_symbol(fmt, addr)
 #endif
 
-#define print_symbol(fmt, addr)			\
-do {						\
-	__check_printsym_format(fmt, "");	\
-	__print_symbol(fmt, addr);		\
-} while(0)
+static inline int print_symbol(const char *fmt, unsigned long addr)
+{
+	__check_printsym_format(fmt, "");
+	return __print_symbol(fmt, addr);
+}
 
 #endif /*_LINUX_KALLSYMS_H*/
kernel/kallsyms.c +2 -2
@@ -231,7 +231,7 @@ const char *kallsyms_lookup(unsigned lon
 }
 
 /* Replace "%s" in format with address, or returns -errno. */
-void __print_symbol(const char *fmt, unsigned long address)
+int __print_symbol(const char *fmt, unsigned long address)
 {
 	char *modname;
 	const char *name;
@@ -251,7 +251,7 @@ void __print_symbol(const char *fmt, uns
 		else
 			sprintf(buffer, "%s+%#lx/%#lx", name, offset, size);
 	}
-	printk(fmt, buffer);
+	return printk(fmt, buffer);
 }
 
 /* To avoid using get_symbol_offset for every symbol, we carry prefix along. */
#5 lore raw

PATCH 2/3 use usual call trace format on x86-64

In reply to Akinobu Mita

Use print_symbol() to dump call trace.

----

 traps.c |   29 +++++++----------------------
 1 files changed, 7 insertions(+), 22 deletions(-)
arch/x86_64/kernel/traps.c renamed from arch/x86_64/kernel/traps.c.orig +7 -22
@@ -30,6 +30,7 @@
 #include <linux/moduleparam.h>
 #include <linux/nmi.h>
 #include <linux/kprobes.h>
+#include <linux/kallsyms.h> 
 
 #include <asm/system.h>
 #include <asm/uaccess.h>
@@ -93,30 +94,14 @@ static inline void conditional_sti(struc
 
 static int kstack_depth_to_print = 10;
 
-#ifdef CONFIG_KALLSYMS
-#include <linux/kallsyms.h> 
-int printk_address(unsigned long address)
-{ 
-	unsigned long offset = 0, symsize;
-	const char *symname;
-	char *modname;
-	char *delim = ":"; 
-	char namebuf[128];
-
-	symname = kallsyms_lookup(address, &symsize, &offset, &modname, namebuf); 
-	if (!symname) 
-		return printk("[<%016lx>]", address);
-	if (!modname) 
-		modname = delim = ""; 		
-        return printk("<%016lx>{%s%s%s%s%+ld}",
-		      address,delim,modname,delim,symname,offset); 
-} 
-#else
 int printk_address(unsigned long address)
 { 
-	return printk("[<%016lx>]", address);
-} 
-#endif
+	int len;
+
+	len = printk("[<%016lx>]", address);
+	len += print_symbol(" %s", address);
+	return len;
+}
 
 static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
 					unsigned *usedp, const char **idp)
#9 lore raw

PATCH 3/3 omit symbol size field in print_symbol()

In reply to Akinobu Mita

I can't find useful usage for the symbol size in print_symbol(). And symbolsize seems to be fixed when vmlinux or modules are compiled. So we can calculate it from vmlinux or modules.

----

 kallsyms.c |    5 ++---
 1 files changed, 2 insertions(+), 3 deletions(-)
kernel/kallsyms.c renamed from kernel/kallsyms.c.orig +2 -3
@@ -246,10 +246,9 @@ int __print_symbol(const char *fmt, unsi
 		sprintf(buffer, "0x%lx", address);
 	else {
 		if (modname)
-			sprintf(buffer, "%s+%#lx/%#lx [%s]", name, offset,
-				size, modname);
+			sprintf(buffer, "%s+%#lx [%s]", name, offset, modname);
 		else
-			sprintf(buffer, "%s+%#lx/%#lx", name, offset, size);
+			sprintf(buffer, "%s+%#lx", name, offset);
 	}
 	return printk(fmt, buffer);
 }
#14 lore raw
In reply to Akinobu Mita
Quoted 11 lines · Akinobu Mita

If I'm missing something, please let me know.

a) On x86-64 we get different Call Trace format than other architectures when we get oops or press SysRq-t:

<ffffffffa008ef6c>{:jbd:kjournald+1030}

There is a architecture independent function print_symbol(). How about using it on x86-64? But it changes to:

   [<ffffffffa008ef6c>] kjournald+0x406/0x578 [jbd]

The x86-64 format is more compact.

Quoted 6 lines

b) I can't find useful usage for the symbol size in print_symbol(). And symbolsize seems to be fixed when vmlinux or modules are compiled. So we can calculate it from vmlinux or modules.
How about removing the field of symbolsize in print_symbol()?

   [<ffffffffa008ef6c>] kjournald+0x406 [jbd]

It's a double check that the oops is matching the vmlinux you're looking at.

-Andi

#10 lore raw

Re: [PATCH 3/3] omit symbol size field in print_symbol()

In reply to Akinobu Mita
Quoted 3 lines · Akinobu Mita

I can't find useful usage for the symbol size in print_symbol(). And symbolsize seems to be fixed when vmlinux or modules are compiled. So we can calculate it from vmlinux or modules.

Also not applied.

-Andi

#7 lore raw

Re: [PATCH 2/3] use usual call trace format on x86-64

In reply to Andi Kleen
Quoted 4 lines · Andi Kleen
Quoted 1 line · Akinobu Mita

Use print_symbol() to dump call trace.

Rejected.

Why? This is a lot more readable than what's there, and similar to other architectures.

#8 lore raw

Re: [PATCH 2/3] use usual call trace format on x86-64

Quoted 8 lines · Christoph Hellwig
Quoted 4 lines · Andi Kleen
Quoted 1 line · Akinobu Mita

Use print_symbol() to dump call trace.

Rejected.

Why? This is a lot more readable than what's there, and similar to other architectures.

See my reply to 0/0

-Andi

#15 lore raw
In reply to Andi Kleen
Quoted 6 lines · Andi Kleen
Quoted 5 lines · Akinobu Mita

If I'm missing something, please let me know.

a) On x86-64 we get different Call Trace format than other architectures when we get oops or press SysRq-t:

Quoted 8 lines
Quoted 6 lines

<ffffffffa008ef6c>{:jbd:kjournald+1030}

There is a architecture independent function print_symbol(). How about using it on x86-64? But it changes to:

   [<ffffffffa008ef6c>] kjournald+0x406/0x578 [jbd]

The x86-64 format is more compact.

How about this update?

1/3: change from "[<...>]" to "<...>".
2/3: change the format of offset from hexadecimal to decimal in.

#16 lore raw

Use print_symbol() to dump call trace.

arch/x86_64/kernel/traps.c renamed from arch/x86_64/kernel/traps.c.orig +7 -22
@@ -30,6 +30,7 @@
 #include <linux/moduleparam.h>
 #include <linux/nmi.h>
 #include <linux/kprobes.h>
+#include <linux/kallsyms.h> 
 
 #include <asm/system.h>
 #include <asm/uaccess.h>
@@ -92,30 +93,14 @@ static inline void conditional_sti(struc
 
 static int kstack_depth_to_print = 10;
 
-#ifdef CONFIG_KALLSYMS
-#include <linux/kallsyms.h> 
-int printk_address(unsigned long address)
-{ 
-	unsigned long offset = 0, symsize;
-	const char *symname;
-	char *modname;
-	char *delim = ":"; 
-	char namebuf[128];
-
-	symname = kallsyms_lookup(address, &symsize, &offset, &modname, namebuf); 
-	if (!symname) 
-		return printk("[<%016lx>]", address);
-	if (!modname) 
-		modname = delim = ""; 		
-        return printk("<%016lx>{%s%s%s%s%+ld}",
-		      address,delim,modname,delim,symname,offset); 
-} 
-#else
 int printk_address(unsigned long address)
 { 
-	return printk("[<%016lx>]", address);
-} 
-#endif
+	int len;
+
+	len = printk("<%016lx>", address);
+	len += print_symbol(" %s", address);
+	return len;
+}
 
 static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
 					unsigned *usedp, const char **idp)
#17 lore raw
In reply to Akinobu Mita

remove symbolsize, and change offset format from hexadecimal to decimal in print_symbol()

kernel/kallsyms.c renamed from kernel/kallsyms.c.orig +3 -4
@@ -237,7 +237,7 @@ int __print_symbol(const char *fmt, unsi
 	const char *name;
 	unsigned long offset, size;
 	char namebuf[KSYM_NAME_LEN+1];
-	char buffer[sizeof("%s+%#lx/%#lx [%s]") + KSYM_NAME_LEN +
+	char buffer[sizeof("%s+%ld [%s]") + KSYM_NAME_LEN +
 		    2*(BITS_PER_LONG*3/10) + MODULE_NAME_LEN + 1];
 
 	name = kallsyms_lookup(address, &size, &offset, &modname, namebuf);
@@ -246,10 +246,9 @@ int __print_symbol(const char *fmt, unsi
 		sprintf(buffer, "0x%lx", address);
 	else {
 		if (modname)
-			sprintf(buffer, "%s+%#lx/%#lx [%s]", name, offset,
-				size, modname);
+			sprintf(buffer, "%s+%ld [%s]", name, offset, modname);
 		else
-			sprintf(buffer, "%s+%#lx/%#lx", name, offset, size);
+			sprintf(buffer, "%s+%ld", name, offset);
 	}
 	return printk(fmt, buffer);
 }
#2 lore raw

Re: [PATCH 1/3] makes print_symbol() return int

In reply to Akinobu Mita
Quoted 2 lines · Akinobu Mita

This patch makes print_symbol() return the number of characters printed.

Why?
Who are the users of this?
If there are users who can bennefit, then where's the patch to make them use this new return value?

--
Jesper Juhl <[email protected]>
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please http://www.expita.com/nomime.html

#3 lore raw

Re: [PATCH 1/3] makes print_symbol() return int

Quoted 7 lines · Jesper Juhl
Quoted 2 lines · Akinobu Mita

This patch makes print_symbol() return the number of characters printed.

Why?
Who are the users of this?
If there are users who can bennefit, then where's the patch to make them use this new return value?

Please see 2/3

#11 lore raw

Re: [PATCH 3/3] omit symbol size field in print_symbol()

In reply to Akinobu Mita
Quoted 3 lines · Akinobu Mita

I can't find useful usage for the symbol size in print_symbol(). And symbolsize seems to be fixed when vmlinux or modules are compiled. So we can calculate it from vmlinux or modules.

the use is that you can see if the EIP actually is inside the function, or if the decoder is going bonkers. Quite useful feature that...

#4 lore raw

Re: [PATCH 1/3] makes print_symbol() return int

In reply to Akinobu Mita
Quoted 11 lines · Akinobu Mita
Quoted 7 lines · Jesper Juhl
Quoted 2 lines · Akinobu Mita

This patch makes print_symbol() return the number of characters printed.

Why?
Who are the users of this?
If there are users who can bennefit, then where's the patch to make them use this new return value?

Please see 2/3

Ahh, for some reason I have not recieved that patch via LKML, but I found it in the archives.
Thanks.

--
Jesper Juhl <[email protected]>
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please http://www.expita.com/nomime.html

#12 lore raw

Re: [PATCH 3/3] omit symbol size field in print_symbol()

In reply to Arjan van de Ven
Quoted 9 lines · Arjan van de Ven
Quoted 3 lines · Akinobu Mita

I can't find useful usage for the symbol size in print_symbol(). And symbolsize seems to be fixed when vmlinux or modules are compiled. So we can calculate it from vmlinux or modules.

the use is that you can see if the EIP actually is inside the function, or if the decoder is going bonkers. Quite useful feature that...

If it is really useful, should we have it on x86-64 too?
The patches 1/3 and 2/3 will do that.
Andi Kleen doesn't want because The call trace on x86-64 could have more than one symbol per line, but my patches will print more characters than now.

#18 lore raw
In reply to Akinobu Mita
Quoted 12 lines · Akinobu Mita
Quoted 11 lines · Andi Kleen
Quoted 10 lines · Akinobu Mita

If I'm missing something, please let me know.

a) On x86-64 we get different Call Trace format than other architectures when we get oops or press SysRq-t:

<ffffffffa008ef6c>{:jbd:kjournald+1030}

There is a architecture independent function print_symbol(). How about using it on x86-64? But it changes to:

Quoted 8 lines
Quoted 3 lines
Quoted 1 line
   [<ffffffffa008ef6c>] kjournald+0x406/0x578 [jbd]

The x86-64 format is more compact.

How about this update?

1/3: change from "[<...>]" to "<...>".
2/3: change the format of offset from hexadecimal to decimal in.

Can you please repost it in a fresh thread? I lost track of what was the latest patch.

In general if you can make the call trace more compact without losing information it's ok for me. Better wrapping sounds like a promising approach.

-Andi

#20 lore raw
In reply to Andi Kleen
Quoted 9 lines · Andi Kleen
Quoted 6 lines

b) I can't find useful usage for the symbol size in print_symbol(). And symbolsize seems to be fixed when vmlinux or modules are compiled. So we can calculate it from vmlinux or modules.
How about removing the field of symbolsize in print_symbol()?

   [<ffffffffa008ef6c>] kjournald+0x406 [jbd]

It's a double check that the oops is matching the vmlinux you're looking at.

If we add system_utsname.version in oops so that we can compare with linux_banner[] in vmlinux, Will it be more precise and easier way to double check than checking symbolsize?

arch/i386/kernel/traps.c renamed from arch/i386/kernel/traps.c.orig +3 -2
@@ -239,9 +239,10 @@ void show_registers(struct pt_regs *regs
 	}
 	print_modules();
 	printk(KERN_EMERG "CPU:    %d\nEIP:    %04x:[<%08lx>]    %s VLI\n"
-			"EFLAGS: %08lx   (%s) \n",
+			"EFLAGS: %08lx   (%s %s) \n",
 		smp_processor_id(), 0xffff & regs->xcs, regs->eip,
-		print_tainted(), regs->eflags, system_utsname.release);
+		print_tainted(), regs->eflags, system_utsname.release,
+		system_utsname.version);
 	print_symbol(KERN_EMERG "EIP is at %s\n", regs->eip);
 	printk(KERN_EMERG "eax: %08lx   ebx: %08lx   ecx: %08lx   edx: %08lx\n",
 		regs->eax, regs->ebx, regs->ecx, regs->edx);
#13 lore raw

Re: [PATCH 3/3] omit symbol size field in print_symbol()

In reply to Arjan van de Ven
Quoted 9 lines · Arjan van de Ven
Quoted 3 lines · Akinobu Mita

I can't find useful usage for the symbol size in print_symbol(). And symbolsize seems to be fixed when vmlinux or modules are compiled. So we can calculate it from vmlinux or modules.

the use is that you can see if the EIP actually is inside the function, or if the decoder is going bonkers. Quite useful feature that...

I still cannot understand the importance of symbolsize in print_symbol() very well. When is the decoder going bonkers for example? Isn't it the task for kallsyms itself to detect that bonkers?

#19 lore raw
In reply to Andi Kleen
Quoted 6 lines
Quoted 6 lines
Quoted 1 line

The x86-64 format is more compact.

How about this update?

1/3: change from "[<...>]" to "<...>".
2/3: change the format of offset from hexadecimal to decimal in.

i386 should also get these two things.

Jan Engelhardt