scripts/markup_oops.pl

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