scripts/markup_oops.pl

   1 #!/usr/bin/perl
   2
   3 use File::Basename;
   4
   5 # Copyright 2008, Intel Corporation
   6 #
   7 # This file is part of the Linux kernel
   8 #
   9 # This program file is free software; you can redistribute it and/or modify it
  10 # under the terms of the GNU General Public License as published by the
  11 # Free Software Foundation; version 2 of the License.
  12 #
  13 # Authors:
  14 #       Arjan van de Ven <arjan@linux.intel.com>
  15
  16
  17 my $vmlinux_name = $ARGV[0];
  18 if (!defined($vmlinux_name)) {
  19         my $kerver = `uname -r`;
  20         chomp($kerver);
  21         $vmlinux_name = "/lib/modules/$kerver/build/vmlinux";
  22         print "No vmlinux specified, assuming $vmlinux_name\n";
  23 }
  24 my $filename = $vmlinux_name;
  25 #
  26 # Step 1: Parse the oops to find the EIP value
  27 #
  28
  29 my $target = "0";
  30 my $function;
  31 my $module = "";
  32 my $func_offset = 0;
  33 my $vmaoffset = 0;
  34
  35 my %regs;
  36
  37
  38 sub parse_x86_regs
  39 {
  40         my ($line) = @_;
  41         if ($line =~ /EAX: ([0-9a-f]+) EBX: ([0-9a-f]+) ECX: ([0-9a-f]+) EDX: ([0-9a-f]+)/) {
  42                 $regs{"%eax"} = $1;
  43                 $regs{"%ebx"} = $2;
  44                 $regs{"%ecx"} = $3;
  45                 $regs{"%edx"} = $4;
  46         }
  47         if ($line =~ /ESI: ([0-9a-f]+) EDI: ([0-9a-f]+) EBP: ([0-9a-f]+) ESP: ([0-9a-f]+)/) {
  48                 $regs{"%esi"} = $1;
  49                 $regs{"%edi"} = $2;
  50                 $regs{"%esp"} = $4;
  51         }
  52         if ($line =~ /RAX: ([0-9a-f]+) RBX: ([0-9a-f]+) RCX: ([0-9a-f]+)/) {
  53                 $regs{"%eax"} = $1;
  54                 $regs{"%ebx"} = $2;
  55                 $regs{"%ecx"} = $3;
  56         }
  57         if ($line =~ /RDX: ([0-9a-f]+) RSI: ([0-9a-f]+) RDI: ([0-9a-f]+)/) {
  58                 $regs{"%edx"} = $1;
  59                 $regs{"%esi"} = $2;
  60                 $regs{"%edi"} = $3;
  61         }
  62         if ($line =~ /RBP: ([0-9a-f]+) R08: ([0-9a-f]+) R09: ([0-9a-f]+)/) {
  63                 $regs{"%r08"} = $2;
  64                 $regs{"%r09"} = $3;
  65         }
  66         if ($line =~ /R10: ([0-9a-f]+) R11: ([0-9a-f]+) R12: ([0-9a-f]+)/) {
  67                 $regs{"%r10"} = $1;
  68                 $regs{"%r11"} = $2;
  69                 $regs{"%r12"} = $3;
  70         }
  71         if ($line =~ /R13: ([0-9a-f]+) R14: ([0-9a-f]+) R15: ([0-9a-f]+)/) {
  72                 $regs{"%r13"} = $1;
  73                 $regs{"%r14"} = $2;
  74                 $regs{"%r15"} = $3;
  75         }
  76 }
  77
  78 sub reg_name
  79 {
  80         my ($reg) = @_;
  81         $reg =~ s/r(.)x/e\1x/;
  82         $reg =~ s/r(.)i/e\1i/;
  83         $reg =~ s/r(.)p/e\1p/;
  84         return $reg;
  85 }
  86
  87 sub process_x86_regs
  88 {
  89         my ($line, $cntr) = @_;
  90         my $str = "";
  91         if (length($line) < 40) {
  92                 return ""; # not an asm istruction
  93         }
  94
  95         # find the arguments to the instruction
  96         if ($line =~ /([0-9a-zA-Z\,\%\(\)\-\+]+)$/) {
  97                 $lastword = $1;
  98         } else {
  99                 return "";
 100         }
 101
 102         # we need to find the registers that get clobbered,
 103         # since their value is no longer relevant for previous
 104         # instructions in the stream.
 105
 106         $clobber = $lastword;
 107         # first, remove all memory operands, they're read only
 108         $clobber =~ s/\([a-z0-9\%\,]+\)//g;
 109         # then, remove everything before the comma, thats the read part
 110         $clobber =~ s/.*\,//g;
 111
 112         # if this is the instruction that faulted, we haven't actually done
 113         # the write yet... nothing is clobbered.
 114         if ($cntr == 0) {
 115                 $clobber = "";
 116         }
 117
 118         foreach $reg (keys(%regs)) {
 119                 my $clobberprime = reg_name($clobber);
 120                 my $lastwordprime = reg_name($lastword);
 121                 my $val = $regs{$reg};
 122                 if ($val =~ /^[0]+$/) {
 123                         $val = "0";
 124                 } else {
 125                         $val =~ s/^0*//;
 126                 }
 127
 128                 # first check if we're clobbering this register; if we do
 129                 # we print it with a =>, and then delete its value
 130                 if ($clobber =~ /$reg/ || $clobberprime =~ /$reg/) {
 131                         if (length($val) > 0) {
 132                                 $str = $str . " $reg => $val ";
 133                         }
 134                         $regs{$reg} = "";
 135                         $val = "";
 136                 }
 137                 # now check if we're reading this register
 138                 if ($lastword =~ /$reg/ || $lastwordprime =~ /$reg/) {
 139                         if (length($val) > 0) {
 140                                 $str = $str . " $reg = $val ";
 141                         }
 142                 }
 143         }
 144         return $str;
 145 }
 146
 147 # parse the oops
 148 while (<STDIN>) {
 149         my $line = $_;
 150         if ($line =~ /EIP: 0060:\[\<([a-z0-9]+)\>\]/) {
 151                 $target = $1;
 152         }
 153         if ($line =~ /RIP: 0010:\[\<([a-z0-9]+)\>\]/) {
 154                 $target = $1;
 155         }
 156         if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]/) {
 157                 $function = $1;
 158                 $func_offset = $2;
 159         }
 160         if ($line =~ /RIP: 0010:\[\<[0-9a-f]+\>\]  \[\<[0-9a-f]+\>\] ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]/) {
 161                 $function = $1;
 162                 $func_offset = $2;
 163         }
 164
 165         # check if it's a module
 166         if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
 167                 $module = $3;
 168         }
 169         if ($line =~ /RIP: 0010:\[\<[0-9a-f]+\>\]  \[\<[0-9a-f]+\>\] ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
 170                 $module = $3;
 171         }
 172         parse_x86_regs($line);
 173 }
 174
 175 my $decodestart = hex($target) - hex($func_offset);
 176 my $decodestop = hex($target) + 8192;
 177 if ($target eq "0") {
 178         print "No oops found!\n";
 179         print "Usage: \n";
 180         print "    dmesg | perl scripts/markup_oops.pl vmlinux\n";
 181         exit;
 182 }
 183
 184 # if it's a module, we need to find the .ko file and calculate a load offset
 185 if ($module ne "") {
 186         my $dir = dirname($filename);
 187         $dir = $dir . "/";
 188         my $mod = $module . ".ko";
 189         my $modulefile = `find $dir -name $mod | head -1`;
 190         chomp($modulefile);
 191         $filename = $modulefile;
 192         if ($filename eq "") {
 193                 print "Module .ko file for $module not found. Aborting\n";
 194                 exit;
 195         }
 196         # ok so we found the module, now we need to calculate the vma offset
 197         open(FILE, "objdump -dS $filename |") || die "Cannot start objdump";
 198         while (<FILE>) {
 199                 if ($_ =~ /^([0-9a-f]+) \<$function\>\:/) {
 200                         my $fu = $1;
 201                         $vmaoffset = hex($target) - hex($fu) - hex($func_offset);
 202                 }
 203         }
 204         close(FILE);
 205 }
 206
 207 my $counter = 0;
 208 my $state   = 0;
 209 my $center  = 0;
 210 my @lines;
 211 my @reglines;
 212
 213 sub InRange {
 214         my ($address, $target) = @_;
 215         my $ad = "0x".$address;
 216         my $ta = "0x".$target;
 217         my $delta = hex($ad) - hex($ta);
 218
 219         if (($delta > -4096) && ($delta < 4096)) {
 220                 return 1;
 221         }
 222         return 0;
 223 }
 224
 225
 226
 227 # first, parse the input into the lines array, but to keep size down,
 228 # we only do this for 4Kb around the sweet spot
 229
 230 open(FILE, "objdump -dS --adjust-vma=$vmaoffset --start-address=$decodestart --stop-address=$decodestop $filename |") || die "Cannot start objdump";
 231
 232 while (<FILE>) {
 233         my $line = $_;
 234         chomp($line);
 235         if ($state == 0) {
 236                 if ($line =~ /^([a-f0-9]+)\:/) {
 237                         if (InRange($1, $target)) {
 238                                 $state = 1;
 239                         }
 240                 }
 241         } else {
 242                 if ($line =~ /^([a-f0-9][a-f0-9][a-f0-9][a-f0-9][a-f0-9][a-f0-9]+)\:/) {
 243                         my $val = $1;
 244                         if (!InRange($val, $target)) {
 245                                 last;
 246                         }
 247                         if ($val eq $target) {
 248                                 $center = $counter;
 249                         }
 250                 }
 251                 $lines[$counter] = $line;
 252
 253                 $counter = $counter + 1;
 254         }
 255 }
 256
 257 close(FILE);
 258
 259 if ($counter == 0) {
 260         print "No matching code found \n";
 261         exit;
 262 }
 263
 264 if ($center == 0) {
 265         print "No matching code found \n";
 266         exit;
 267 }
 268
 269 my $start;
 270 my $finish;
 271 my $codelines = 0;
 272 my $binarylines = 0;
 273 # now we go up and down in the array to find how much we want to print
 274
 275 $start = $center;
 276
 277 while ($start > 1) {
 278         $start = $start - 1;
 279         my $line = $lines[$start];
 280         if ($line =~ /^([a-f0-9]+)\:/) {
 281                 $binarylines = $binarylines + 1;
 282         } else {
 283                 $codelines = $codelines + 1;
 284         }
 285         if ($codelines > 10) {
 286                 last;
 287         }
 288         if ($binarylines > 20) {
 289                 last;
 290         }
 291 }
 292
 293
 294 $finish = $center;
 295 $codelines = 0;
 296 $binarylines = 0;
 297 while ($finish < $counter) {
 298         $finish = $finish + 1;
 299         my $line = $lines[$finish];
 300         if ($line =~ /^([a-f0-9]+)\:/) {
 301                 $binarylines = $binarylines + 1;
 302         } else {
 303                 $codelines = $codelines + 1;
 304         }
 305         if ($codelines > 10) {
 306                 last;
 307         }
 308         if ($binarylines > 20) {
 309                 last;
 310         }
 311 }
 312
 313
 314 my $i;
 315
 316
 317 # start annotating the registers in the asm.
 318 # this goes from the oopsing point back, so that the annotator
 319 # can track (opportunistically) which registers got written and
 320 # whos value no longer is relevant.
 321
 322 $i = $center;
 323 while ($i >= $start) {
 324         $reglines[$i] = process_x86_regs($lines[$i], $center - $i);
 325         $i = $i - 1;
 326 }
 327
 328 $i = $start;
 329 while ($i < $finish) {
 330         my $line;
 331         if ($i == $center) {
 332                 $line =  "*$lines[$i] ";
 333         } else {
 334                 $line =  " $lines[$i] ";
 335         }
 336         print $line;
 337         if (defined($reglines[$i]) && length($reglines[$i]) > 0) {
 338                 my $c = 60 - length($line);
 339                 while ($c > 0) { print " "; $c = $c - 1; };
 340                 print "| $reglines[$i]";
 341         }
 342         if ($i == $center) {
 343                 print "<--- faulting instruction";
 344         }
 345         print "\n";
 346         $i = $i +1;
 347 }
 348