Skip to contentSkip to conversation

Improve the markup_oops.pl script

PATCH In review Arjan van de Ven 2 participants 5 messages Jan 13, 2009 – Feb 15, 2009

lore
Share & tools
lore
#0 lore raw

cover Improve the markup_oops.pl script

Hi,

the following 2 patches enhance the markup_oops.pl script in two ways Patch 1: Add support for putting register values in the asm dump Patch 2: Add support for 64 bit X86

The output (on a 32 bit OS with a "make me oops" example module) looks like the pasted dump; for the instruction where the oops happened, as well as a few instructions before that where the dumper can determine the register value. the value is appended on said line.

if (bar-1) {

 f8436012:	8b 45 fc             	mov    -0x4(%ebp),%eax 
 f8436015:	48                   	dec    %eax 
 f8436016:	74 09                	je     f8436021 <foo_bar+0x21> 
 		bar = bar * 5; 
 f8436018:	8b 45 fc             	mov    -0x4(%ebp),%eax 
 f843601b:	8d 04 80             	lea    (%eax,%eax,4),%eax  |  %eax => 0 
 f843601e:	89 45 fc             	mov    %eax,-0x4(%ebp)     |  %eax = 0 

}
*foo = 1;
*f8436021: c6 02 01 movb $0x1,(%edx) | %edx = 0 <--- faulting instruction
return bar;
f8436024: 8b 45 fc mov -0x4(%ebp),%eax
}
f8436027: c9 leave
f8436028: c3 ret

--
Arjan van de Ven Intel Open Source Technology Centre
For development, discussion and tips for power savings, visit http://www.lesswatts.org

#1 lore raw

PATCH 1/2 scripts: add x86 register parser to markup_oops.pl

Quoted 1 line

From ace27f35478ed449fd345d2381411362f6bcc9f9 Mon Sep 17 00:00:00 2001

From: Arjan van de Ven <[email protected]>
Date: Tue, 13 Jan 2009 05:50:35 +0000
Subject: [PATCH] scripts: add x86 register parser to markup_oops.pl

An oops dump also contains the register values.
This patch parses these for (32 bit) x86, and then annotates the disassembly with these values; this helps in analysis of the oops by the developer, for example, NULL pointer or other pointer bugs show up clearly this way.

---
 scripts/markup_oops.pl |  106 +++++++++++++++++++++++++++++++++++++++++++++---
 1 files changed, 100 insertions(+), 6 deletions(-)
scripts/markup_oops.pl +100 -6
index d40449c..bb20468 100644
@@ -32,6 +32,78 @@ my $module = "";
 my $func_offset;
 my $vmaoffset = 0;
 
+my %regs;
+
+
+sub parse_x86_regs
+{
+	my ($line) = @_;
+	if ($line =~ /EAX: ([0-9a-f]+) EBX: ([0-9a-f]+) ECX: ([0-9a-f]+) EDX: ([0-9a-f]+)/) {
+		$regs{"%eax"} = $1;
+		$regs{"%ebx"} = $2;
+		$regs{"%ecx"} = $3;
+		$regs{"%edx"} = $4;
+	}
+	if ($line =~ /ESI: ([0-9a-f]+) EDI: ([0-9a-f]+) EBP: ([0-9a-f]+) ESP: ([0-9a-f]+)/) {
+		$regs{"%esi"} = $1;
+		$regs{"%edi"} = $2;
+		$regs{"%esp"} = $4;
+	}
+}
+
+sub process_x86_regs
+{
+	my ($line, $cntr) = @_;
+	my $str = "";
+	if (length($line) < 40) {
+		return ""; # not an asm istruction
+	}
+
+	# find the arguments to the instruction
+	if ($line =~ /([0-9a-zA-Z\,\%\(\)\-\+]+)$/) {
+		$lastword = $1;
+	} else {
+		return "";
+	}
+
+	# we need to find the registers that get clobbered,
+	# since their value is no longer relevant for previous
+	# instructions in the stream.
+
+	$clobber = $lastword;
+	# first, remove all memory operands, they're read only
+	$clobber =~ s/\([a-z0-9\%\,]+\)//g;
+	# then, remove everything before the comma, thats the read part
+	$clobber =~ s/.*\,//g;
+
+	# if this is the instruction that faulted, we haven't actually done
+	# the write yet... nothing is clobbered.
+	if ($cntr == 0) {
+		$clobber = "";
+	}
+
+	foreach $reg (keys(%regs)) {
+		my $val = $regs{$reg};
+		# first check if we're clobbering this register; if we do
+		# we print it with a =>, and then delete its value
+		if ($clobber =~ /$reg/) {
+			if (length($val) > 0) {
+				$str = $str . " $reg => $val ";
+			}
+			$regs{$reg} = "";
+			$val = "";
+		}
+		# now check if we're reading this register
+		if ($lastword =~ /$reg/) {
+			if (length($val) > 0) {
+				$str = $str . " $reg = $val ";
+			}
+		}
+	}
+	return $str;
+}
+
+# parse the oops
 while (<STDIN>) {
 	my $line = $_;
 	if ($line =~ /EIP: 0060:\[\<([a-z0-9]+)\>\]/) {
@@ -46,10 +118,11 @@ while (<STDIN>) {
 	if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
 		$module = $3;
 	}
+	parse_x86_regs($line);
 }
 
 my $decodestart = hex($target) - hex($func_offset);
-my $decodestop = $decodestart + 8192;
+my $decodestop = hex($target) + 8192;
 if ($target eq "0") {
 	print "No oops found!\n";
 	print "Usage: \n";
@@ -84,6 +157,7 @@ my $counter = 0;
 my $state   = 0;
 my $center  = 0;
 my @lines;
+my @reglines;
 
 sub InRange {
 	my ($address, $target) = @_;
@@ -188,16 +262,36 @@ while ($finish < $counter) {
 
 my $i;
 
-my $fulltext = "";
+
+# start annotating the registers in the asm.
+# this goes from the oopsing point back, so that the annotator
+# can track (opportunistically) which registers got written and
+# whos value no longer is relevant.
+
+$i = $center;
+while ($i >= $start) {
+	$reglines[$i] = process_x86_regs($lines[$i], $center - $i);
+	$i = $i - 1;
+}
+
 $i = $start;
 while ($i < $finish) {
+	my $line;
 	if ($i == $center) {
-		$fulltext = $fulltext . "*$lines[$i]     <----- faulting instruction\n";
+		$line =  "*$lines[$i] ";
 	} else {
-		$fulltext = $fulltext .  " $lines[$i]\n";
+		$line =  " $lines[$i] ";
+	}
+	print $line;
+	if (defined($reglines[$i]) && length($reglines[$i]) > 0) {
+		my $c = 60 - length($line);
+		while ($c > 0) { print " "; $c = $c - 1; };
+		print "| $reglines[$i]";
 	}
+	if ($i == $center) {
+		print "<--- faulting instruction";
+	}
+	print "\n";
 	$i = $i +1;
 }
 
-print $fulltext;
-
Signature · 9 lines
1.6.0.6




-- 
Arjan van de Ven 	Intel Open Source Technology Centre
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org
#2 lore raw

Re: [PATCH 1/2] scripts: add x86 register parser to markup_oops.pl

Quoted 5 lines · Arjan van de Ven

From ace27f35478ed449fd345d2381411362f6bcc9f9 Mon Sep 17 00:00:00 2001
From: Arjan van de Ven <[email protected]>
Date: Tue, 13 Jan 2009 05:50:35 +0000
Subject: [PATCH] scripts: add x86 register parser to markup_oops.pl

Quoted 7 lines

An oops dump also contains the register values.
This patch parses these for (32 bit) x86, and then annotates the disassembly with these values; this helps in analysis of the oops by the developer, for example, NULL pointer or other pointer bugs show up clearly this way.

Added to kbuild-fixes.git

Sam

#3 lore raw

PATCH 2/2 scripts: add x86 64 bit support to the markup_oops.pl script

In reply to Arjan van de Ven
Quoted 1 line

From 231381ec24bf8f13fb87a06ae937032842a67225 Mon Sep 17 00:00:00 2001

From: Arjan van de Ven <[email protected]>
Date: Tue, 13 Jan 2009 09:45:53 +0000
Subject: [PATCH] scripts: add x86 64 bit support to the markup_oops.pl script

---
 scripts/markup_oops.pl |   59 ++++++++++++++++++++++++++++++++++++++++++++---
 1 files changed, 55 insertions(+), 4 deletions(-)
scripts/markup_oops.pl +55 -4
index bb20468..9b90b6c 100644
@@ -1,4 +1,4 @@
-#!/usr/bin/perl -w
+#!/usr/bin/perl 
 
 use File::Basename;
 
@@ -29,7 +29,7 @@ my $filename = $vmlinux_name;
 my $target = "0";
 my $function;
 my $module = "";
-my $func_offset;
+my $func_offset = 0;
 my $vmaoffset = 0;
 
 my %regs;
@@ -49,6 +49,39 @@ sub parse_x86_regs
 		$regs{"%edi"} = $2;
 		$regs{"%esp"} = $4;
 	}
+	if ($line =~ /RAX: ([0-9a-f]+) RBX: ([0-9a-f]+) RCX: ([0-9a-f]+)/) {
+		$regs{"%eax"} = $1;
+		$regs{"%ebx"} = $2;
+		$regs{"%ecx"} = $3;
+	}
+	if ($line =~ /RDX: ([0-9a-f]+) RSI: ([0-9a-f]+) RDI: ([0-9a-f]+)/) {
+		$regs{"%edx"} = $1;
+		$regs{"%esi"} = $2;
+		$regs{"%edi"} = $3;
+	}
+	if ($line =~ /RBP: ([0-9a-f]+) R08: ([0-9a-f]+) R09: ([0-9a-f]+)/) {
+		$regs{"%r08"} = $2;
+		$regs{"%r09"} = $3;
+	}
+	if ($line =~ /R10: ([0-9a-f]+) R11: ([0-9a-f]+) R12: ([0-9a-f]+)/) {
+		$regs{"%r10"} = $1;
+		$regs{"%r11"} = $2;
+		$regs{"%r12"} = $3;
+	}
+	if ($line =~ /R13: ([0-9a-f]+) R14: ([0-9a-f]+) R15: ([0-9a-f]+)/) {
+		$regs{"%r13"} = $1;
+		$regs{"%r14"} = $2;
+		$regs{"%r15"} = $3;
+	}
+}
+
+sub reg_name
+{
+	my ($reg) = @_;
+	$reg =~ s/r(.)x/e\1x/;
+	$reg =~ s/r(.)i/e\1i/;
+	$reg =~ s/r(.)p/e\1p/;
+	return $reg;
 }
 
 sub process_x86_regs
@@ -83,10 +116,18 @@ sub process_x86_regs
 	}
 
 	foreach $reg (keys(%regs)) {
+		my $clobberprime = reg_name($clobber);
+		my $lastwordprime = reg_name($lastword);
 		my $val = $regs{$reg};
+		if ($val =~ /^[0]+$/) {
+			$val = "0"; 
+		} else {
+			$val =~ s/^0*//;
+		}
+		
 		# first check if we're clobbering this register; if we do
 		# we print it with a =>, and then delete its value
-		if ($clobber =~ /$reg/) {
+		if ($clobber =~ /$reg/ || $clobberprime =~ /$reg/) {
 			if (length($val) > 0) {
 				$str = $str . " $reg => $val ";
 			}
@@ -94,7 +135,7 @@ sub process_x86_regs
 			$val = "";
 		}
 		# now check if we're reading this register
-		if ($lastword =~ /$reg/) {
+		if ($lastword =~ /$reg/ || $lastwordprime =~ /$reg/) {
 			if (length($val) > 0) {
 				$str = $str . " $reg = $val ";
 			}
@@ -109,15 +150,25 @@ while (<STDIN>) {
 	if ($line =~ /EIP: 0060:\[\<([a-z0-9]+)\>\]/) {
 		$target = $1;
 	}
+	if ($line =~ /RIP: 0010:\[\<([a-z0-9]+)\>\]/) {
+		$target = $1;
+	}
 	if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]/) {
 		$function = $1;
 		$func_offset = $2;
 	}
+	if ($line =~ /RIP: 0010:\[\<[0-9a-f]+\>\]  \[\<[0-9a-f]+\>\] ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]/) {
+		$function = $1;
+		$func_offset = $2;
+	}
 
 	# check if it's a module
 	if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
 		$module = $3;
 	}
+	if ($line =~ /RIP: 0010:\[\<[0-9a-f]+\>\]  \[\<[0-9a-f]+\>\] ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
+		$module = $3;
+	}
 	parse_x86_regs($line);
 }
 
Signature · 9 lines
1.6.0.6




-- 
Arjan van de Ven 	Intel Open Source Technology Centre
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org
#4 lore raw

Re: [PATCH 2/2] scripts: add x86 64 bit support to the markup_oops.pl script

Quoted 4 lines · Arjan van de Ven

From 231381ec24bf8f13fb87a06ae937032842a67225 Mon Sep 17 00:00:00 2001
From: Arjan van de Ven <[email protected]>
Date: Tue, 13 Jan 2009 09:45:53 +0000
Subject: [PATCH] scripts: add x86 64 bit support to the markup_oops.pl script

Added to kbuild-fixes.git

Sam