Loading...
1#!/usr/bin/perl
2
3use File::Basename;
4use Math::BigInt;
5use Getopt::Long;
6
7# Copyright 2008, Intel Corporation
8#
9# This file is part of the Linux kernel
10#
11# This program file is free software; you can redistribute it and/or modify it
12# under the terms of the GNU General Public License as published by the
13# Free Software Foundation; version 2 of the License.
14#
15# Authors:
16# Arjan van de Ven <arjan@linux.intel.com>
17
18
19my $cross_compile = "";
20my $vmlinux_name = "";
21my $modulefile = "";
22
23# Get options
24Getopt::Long::GetOptions(
25 'cross-compile|c=s' => \$cross_compile,
26 'module|m=s' => \$modulefile,
27 'help|h' => \&usage,
28) || usage ();
29my $vmlinux_name = $ARGV[0];
30if (!defined($vmlinux_name)) {
31 my $kerver = `uname -r`;
32 chomp($kerver);
33 $vmlinux_name = "/lib/modules/$kerver/build/vmlinux";
34 print "No vmlinux specified, assuming $vmlinux_name\n";
35}
36my $filename = $vmlinux_name;
37
38# Parse the oops to find the EIP value
39
40my $target = "0";
41my $function;
42my $module = "";
43my $func_offset = 0;
44my $vmaoffset = 0;
45
46my %regs;
47
48
49sub parse_x86_regs
50{
51 my ($line) = @_;
52 if ($line =~ /EAX: ([0-9a-f]+) EBX: ([0-9a-f]+) ECX: ([0-9a-f]+) EDX: ([0-9a-f]+)/) {
53 $regs{"%eax"} = $1;
54 $regs{"%ebx"} = $2;
55 $regs{"%ecx"} = $3;
56 $regs{"%edx"} = $4;
57 }
58 if ($line =~ /ESI: ([0-9a-f]+) EDI: ([0-9a-f]+) EBP: ([0-9a-f]+) ESP: ([0-9a-f]+)/) {
59 $regs{"%esi"} = $1;
60 $regs{"%edi"} = $2;
61 $regs{"%esp"} = $4;
62 }
63 if ($line =~ /RAX: ([0-9a-f]+) RBX: ([0-9a-f]+) RCX: ([0-9a-f]+)/) {
64 $regs{"%eax"} = $1;
65 $regs{"%ebx"} = $2;
66 $regs{"%ecx"} = $3;
67 }
68 if ($line =~ /RDX: ([0-9a-f]+) RSI: ([0-9a-f]+) RDI: ([0-9a-f]+)/) {
69 $regs{"%edx"} = $1;
70 $regs{"%esi"} = $2;
71 $regs{"%edi"} = $3;
72 }
73 if ($line =~ /RBP: ([0-9a-f]+) R08: ([0-9a-f]+) R09: ([0-9a-f]+)/) {
74 $regs{"%r08"} = $2;
75 $regs{"%r09"} = $3;
76 }
77 if ($line =~ /R10: ([0-9a-f]+) R11: ([0-9a-f]+) R12: ([0-9a-f]+)/) {
78 $regs{"%r10"} = $1;
79 $regs{"%r11"} = $2;
80 $regs{"%r12"} = $3;
81 }
82 if ($line =~ /R13: ([0-9a-f]+) R14: ([0-9a-f]+) R15: ([0-9a-f]+)/) {
83 $regs{"%r13"} = $1;
84 $regs{"%r14"} = $2;
85 $regs{"%r15"} = $3;
86 }
87}
88
89sub reg_name
90{
91 my ($reg) = @_;
92 $reg =~ s/r(.)x/e\1x/;
93 $reg =~ s/r(.)i/e\1i/;
94 $reg =~ s/r(.)p/e\1p/;
95 return $reg;
96}
97
98sub process_x86_regs
99{
100 my ($line, $cntr) = @_;
101 my $str = "";
102 if (length($line) < 40) {
103 return ""; # not an asm istruction
104 }
105
106 # find the arguments to the instruction
107 if ($line =~ /([0-9a-zA-Z\,\%\(\)\-\+]+)$/) {
108 $lastword = $1;
109 } else {
110 return "";
111 }
112
113 # we need to find the registers that get clobbered,
114 # since their value is no longer relevant for previous
115 # instructions in the stream.
116
117 $clobber = $lastword;
118 # first, remove all memory operands, they're read only
119 $clobber =~ s/\([a-z0-9\%\,]+\)//g;
120 # then, remove everything before the comma, thats the read part
121 $clobber =~ s/.*\,//g;
122
123 # if this is the instruction that faulted, we haven't actually done
124 # the write yet... nothing is clobbered.
125 if ($cntr == 0) {
126 $clobber = "";
127 }
128
129 foreach $reg (keys(%regs)) {
130 my $clobberprime = reg_name($clobber);
131 my $lastwordprime = reg_name($lastword);
132 my $val = $regs{$reg};
133 if ($val =~ /^[0]+$/) {
134 $val = "0";
135 } else {
136 $val =~ s/^0*//;
137 }
138
139 # first check if we're clobbering this register; if we do
140 # we print it with a =>, and then delete its value
141 if ($clobber =~ /$reg/ || $clobberprime =~ /$reg/) {
142 if (length($val) > 0) {
143 $str = $str . " $reg => $val ";
144 }
145 $regs{$reg} = "";
146 $val = "";
147 }
148 # now check if we're reading this register
149 if ($lastword =~ /$reg/ || $lastwordprime =~ /$reg/) {
150 if (length($val) > 0) {
151 $str = $str . " $reg = $val ";
152 }
153 }
154 }
155 return $str;
156}
157
158# parse the oops
159while (<STDIN>) {
160 my $line = $_;
161 if ($line =~ /EIP: 0060:\[\<([a-z0-9]+)\>\]/) {
162 $target = $1;
163 }
164 if ($line =~ /RIP: 0010:\[\<([a-z0-9]+)\>\]/) {
165 $target = $1;
166 }
167 if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+0x([0-9a-f]+)\/0x[a-f0-9]/) {
168 $function = $1;
169 $func_offset = $2;
170 }
171 if ($line =~ /RIP: 0010:\[\<[0-9a-f]+\>\] \[\<[0-9a-f]+\>\] ([a-zA-Z0-9\_]+)\+0x([0-9a-f]+)\/0x[a-f0-9]/) {
172 $function = $1;
173 $func_offset = $2;
174 }
175
176 # check if it's a module
177 if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
178 $module = $3;
179 }
180 if ($line =~ /RIP: 0010:\[\<[0-9a-f]+\>\] \[\<[0-9a-f]+\>\] ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
181 $module = $3;
182 }
183 parse_x86_regs($line);
184}
185
186my $decodestart = Math::BigInt->from_hex("0x$target") - Math::BigInt->from_hex("0x$func_offset");
187my $decodestop = Math::BigInt->from_hex("0x$target") + 8192;
188if ($target eq "0") {
189 print "No oops found!\n";
190 usage();
191}
192
193# if it's a module, we need to find the .ko file and calculate a load offset
194if ($module ne "") {
195 if ($modulefile eq "") {
196 $modulefile = `modinfo -F filename $module`;
197 chomp($modulefile);
198 }
199 $filename = $modulefile;
200 if ($filename eq "") {
201 print "Module .ko file for $module not found. Aborting\n";
202 exit;
203 }
204 # ok so we found the module, now we need to calculate the vma offset
205 open(FILE, $cross_compile."objdump -dS $filename |") || die "Cannot start objdump";
206 while (<FILE>) {
207 if ($_ =~ /^([0-9a-f]+) \<$function\>\:/) {
208 my $fu = $1;
209 $vmaoffset = Math::BigInt->from_hex("0x$target") - Math::BigInt->from_hex("0x$fu") - Math::BigInt->from_hex("0x$func_offset");
210 }
211 }
212 close(FILE);
213}
214
215my $counter = 0;
216my $state = 0;
217my $center = -1;
218my @lines;
219my @reglines;
220
221sub InRange {
222 my ($address, $target) = @_;
223 my $ad = "0x".$address;
224 my $ta = "0x".$target;
225 my $delta = Math::BigInt->from_hex($ad) - Math::BigInt->from_hex($ta);
226
227 if (($delta > -4096) && ($delta < 4096)) {
228 return 1;
229 }
230 return 0;
231}
232
233
234
235# first, parse the input into the lines array, but to keep size down,
236# we only do this for 4Kb around the sweet spot
237
238open(FILE, $cross_compile."objdump -dS --adjust-vma=$vmaoffset --start-address=$decodestart --stop-address=$decodestop $filename |") || die "Cannot start objdump";
239
240while (<FILE>) {
241 my $line = $_;
242 chomp($line);
243 if ($state == 0) {
244 if ($line =~ /^([a-f0-9]+)\:/) {
245 if (InRange($1, $target)) {
246 $state = 1;
247 }
248 }
249 }
250 if ($state == 1) {
251 if ($line =~ /^([a-f0-9][a-f0-9][a-f0-9][a-f0-9][a-f0-9][a-f0-9]+)\:/) {
252 my $val = $1;
253 if (!InRange($val, $target)) {
254 last;
255 }
256 if ($val eq $target) {
257 $center = $counter;
258 }
259 }
260 $lines[$counter] = $line;
261
262 $counter = $counter + 1;
263 }
264}
265
266close(FILE);
267
268if ($counter == 0) {
269 print "No matching code found \n";
270 exit;
271}
272
273if ($center == -1) {
274 print "No matching code found \n";
275 exit;
276}
277
278my $start;
279my $finish;
280my $codelines = 0;
281my $binarylines = 0;
282# now we go up and down in the array to find how much we want to print
283
284$start = $center;
285
286while ($start > 1) {
287 $start = $start - 1;
288 my $line = $lines[$start];
289 if ($line =~ /^([a-f0-9]+)\:/) {
290 $binarylines = $binarylines + 1;
291 } else {
292 $codelines = $codelines + 1;
293 }
294 if ($codelines > 10) {
295 last;
296 }
297 if ($binarylines > 20) {
298 last;
299 }
300}
301
302
303$finish = $center;
304$codelines = 0;
305$binarylines = 0;
306while ($finish < $counter) {
307 $finish = $finish + 1;
308 my $line = $lines[$finish];
309 if ($line =~ /^([a-f0-9]+)\:/) {
310 $binarylines = $binarylines + 1;
311 } else {
312 $codelines = $codelines + 1;
313 }
314 if ($codelines > 10) {
315 last;
316 }
317 if ($binarylines > 20) {
318 last;
319 }
320}
321
322
323my $i;
324
325
326# start annotating the registers in the asm.
327# this goes from the oopsing point back, so that the annotator
328# can track (opportunistically) which registers got written and
329# whos value no longer is relevant.
330
331$i = $center;
332while ($i >= $start) {
333 $reglines[$i] = process_x86_regs($lines[$i], $center - $i);
334 $i = $i - 1;
335}
336
337$i = $start;
338while ($i < $finish) {
339 my $line;
340 if ($i == $center) {
341 $line = "*$lines[$i] ";
342 } else {
343 $line = " $lines[$i] ";
344 }
345 print $line;
346 if (defined($reglines[$i]) && length($reglines[$i]) > 0) {
347 my $c = 60 - length($line);
348 while ($c > 0) { print " "; $c = $c - 1; };
349 print "| $reglines[$i]";
350 }
351 if ($i == $center) {
352 print "<--- faulting instruction";
353 }
354 print "\n";
355 $i = $i +1;
356}
357
358sub usage {
359 print <<EOT;
360Usage:
361 dmesg | perl $0 [OPTION] [VMLINUX]
362
363OPTION:
364 -c, --cross-compile CROSS_COMPILE Specify the prefix used for toolchain.
365 -m, --module MODULE_DIRNAME Specify the module filename.
366 -h, --help Help.
367EOT
368 exit;
369}
370
1#!/usr/bin/env perl
2# SPDX-License-Identifier: GPL-2.0-only
3
4use File::Basename;
5use Math::BigInt;
6use Getopt::Long;
7
8# Copyright 2008, Intel Corporation
9#
10# This file is part of the Linux kernel
11#
12# Authors:
13# Arjan van de Ven <arjan@linux.intel.com>
14
15
16my $cross_compile = "";
17my $vmlinux_name = "";
18my $modulefile = "";
19
20# Get options
21Getopt::Long::GetOptions(
22 'cross-compile|c=s' => \$cross_compile,
23 'module|m=s' => \$modulefile,
24 'help|h' => \&usage,
25) || usage ();
26my $vmlinux_name = $ARGV[0];
27if (!defined($vmlinux_name)) {
28 my $kerver = `uname -r`;
29 chomp($kerver);
30 $vmlinux_name = "/lib/modules/$kerver/build/vmlinux";
31 print "No vmlinux specified, assuming $vmlinux_name\n";
32}
33my $filename = $vmlinux_name;
34
35# Parse the oops to find the EIP value
36
37my $target = "0";
38my $function;
39my $module = "";
40my $func_offset = 0;
41my $vmaoffset = 0;
42
43my %regs;
44
45
46sub parse_x86_regs
47{
48 my ($line) = @_;
49 if ($line =~ /EAX: ([0-9a-f]+) EBX: ([0-9a-f]+) ECX: ([0-9a-f]+) EDX: ([0-9a-f]+)/) {
50 $regs{"%eax"} = $1;
51 $regs{"%ebx"} = $2;
52 $regs{"%ecx"} = $3;
53 $regs{"%edx"} = $4;
54 }
55 if ($line =~ /ESI: ([0-9a-f]+) EDI: ([0-9a-f]+) EBP: ([0-9a-f]+) ESP: ([0-9a-f]+)/) {
56 $regs{"%esi"} = $1;
57 $regs{"%edi"} = $2;
58 $regs{"%esp"} = $4;
59 }
60 if ($line =~ /RAX: ([0-9a-f]+) RBX: ([0-9a-f]+) RCX: ([0-9a-f]+)/) {
61 $regs{"%eax"} = $1;
62 $regs{"%ebx"} = $2;
63 $regs{"%ecx"} = $3;
64 }
65 if ($line =~ /RDX: ([0-9a-f]+) RSI: ([0-9a-f]+) RDI: ([0-9a-f]+)/) {
66 $regs{"%edx"} = $1;
67 $regs{"%esi"} = $2;
68 $regs{"%edi"} = $3;
69 }
70 if ($line =~ /RBP: ([0-9a-f]+) R08: ([0-9a-f]+) R09: ([0-9a-f]+)/) {
71 $regs{"%r08"} = $2;
72 $regs{"%r09"} = $3;
73 }
74 if ($line =~ /R10: ([0-9a-f]+) R11: ([0-9a-f]+) R12: ([0-9a-f]+)/) {
75 $regs{"%r10"} = $1;
76 $regs{"%r11"} = $2;
77 $regs{"%r12"} = $3;
78 }
79 if ($line =~ /R13: ([0-9a-f]+) R14: ([0-9a-f]+) R15: ([0-9a-f]+)/) {
80 $regs{"%r13"} = $1;
81 $regs{"%r14"} = $2;
82 $regs{"%r15"} = $3;
83 }
84}
85
86sub reg_name
87{
88 my ($reg) = @_;
89 $reg =~ s/r(.)x/e\1x/;
90 $reg =~ s/r(.)i/e\1i/;
91 $reg =~ s/r(.)p/e\1p/;
92 return $reg;
93}
94
95sub process_x86_regs
96{
97 my ($line, $cntr) = @_;
98 my $str = "";
99 if (length($line) < 40) {
100 return ""; # not an asm istruction
101 }
102
103 # find the arguments to the instruction
104 if ($line =~ /([0-9a-zA-Z\,\%\(\)\-\+]+)$/) {
105 $lastword = $1;
106 } else {
107 return "";
108 }
109
110 # we need to find the registers that get clobbered,
111 # since their value is no longer relevant for previous
112 # instructions in the stream.
113
114 $clobber = $lastword;
115 # first, remove all memory operands, they're read only
116 $clobber =~ s/\([a-z0-9\%\,]+\)//g;
117 # then, remove everything before the comma, thats the read part
118 $clobber =~ s/.*\,//g;
119
120 # if this is the instruction that faulted, we haven't actually done
121 # the write yet... nothing is clobbered.
122 if ($cntr == 0) {
123 $clobber = "";
124 }
125
126 foreach $reg (keys(%regs)) {
127 my $clobberprime = reg_name($clobber);
128 my $lastwordprime = reg_name($lastword);
129 my $val = $regs{$reg};
130 if ($val =~ /^[0]+$/) {
131 $val = "0";
132 } else {
133 $val =~ s/^0*//;
134 }
135
136 # first check if we're clobbering this register; if we do
137 # we print it with a =>, and then delete its value
138 if ($clobber =~ /$reg/ || $clobberprime =~ /$reg/) {
139 if (length($val) > 0) {
140 $str = $str . " $reg => $val ";
141 }
142 $regs{$reg} = "";
143 $val = "";
144 }
145 # now check if we're reading this register
146 if ($lastword =~ /$reg/ || $lastwordprime =~ /$reg/) {
147 if (length($val) > 0) {
148 $str = $str . " $reg = $val ";
149 }
150 }
151 }
152 return $str;
153}
154
155# parse the oops
156while (<STDIN>) {
157 my $line = $_;
158 if ($line =~ /EIP: 0060:\[\<([a-z0-9]+)\>\]/) {
159 $target = $1;
160 }
161 if ($line =~ /RIP: 0010:\[\<([a-z0-9]+)\>\]/) {
162 $target = $1;
163 }
164 if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+0x([0-9a-f]+)\/0x[a-f0-9]/) {
165 $function = $1;
166 $func_offset = $2;
167 }
168 if ($line =~ /RIP: 0010:\[\<[0-9a-f]+\>\] \[\<[0-9a-f]+\>\] ([a-zA-Z0-9\_]+)\+0x([0-9a-f]+)\/0x[a-f0-9]/) {
169 $function = $1;
170 $func_offset = $2;
171 }
172
173 # check if it's a module
174 if ($line =~ /EIP is at ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
175 $module = $3;
176 }
177 if ($line =~ /RIP: 0010:\[\<[0-9a-f]+\>\] \[\<[0-9a-f]+\>\] ([a-zA-Z0-9\_]+)\+(0x[0-9a-f]+)\/0x[a-f0-9]+\W\[([a-zA-Z0-9\_\-]+)\]/) {
178 $module = $3;
179 }
180 parse_x86_regs($line);
181}
182
183my $decodestart = Math::BigInt->from_hex("0x$target") - Math::BigInt->from_hex("0x$func_offset");
184my $decodestop = Math::BigInt->from_hex("0x$target") + 8192;
185if ($target eq "0") {
186 print "No oops found!\n";
187 usage();
188}
189
190# if it's a module, we need to find the .ko file and calculate a load offset
191if ($module ne "") {
192 if ($modulefile eq "") {
193 $modulefile = `modinfo -F filename $module`;
194 chomp($modulefile);
195 }
196 $filename = $modulefile;
197 if ($filename eq "") {
198 print "Module .ko file for $module not found. Aborting\n";
199 exit;
200 }
201 # ok so we found the module, now we need to calculate the vma offset
202 open(FILE, $cross_compile."objdump -dS $filename |") || die "Cannot start objdump";
203 while (<FILE>) {
204 if ($_ =~ /^([0-9a-f]+) \<$function\>\:/) {
205 my $fu = $1;
206 $vmaoffset = Math::BigInt->from_hex("0x$target") - Math::BigInt->from_hex("0x$fu") - Math::BigInt->from_hex("0x$func_offset");
207 }
208 }
209 close(FILE);
210}
211
212my $counter = 0;
213my $state = 0;
214my $center = -1;
215my @lines;
216my @reglines;
217
218sub InRange {
219 my ($address, $target) = @_;
220 my $ad = "0x".$address;
221 my $ta = "0x".$target;
222 my $delta = Math::BigInt->from_hex($ad) - Math::BigInt->from_hex($ta);
223
224 if (($delta > -4096) && ($delta < 4096)) {
225 return 1;
226 }
227 return 0;
228}
229
230
231
232# first, parse the input into the lines array, but to keep size down,
233# we only do this for 4Kb around the sweet spot
234
235open(FILE, $cross_compile."objdump -dS --adjust-vma=$vmaoffset --start-address=$decodestart --stop-address=$decodestop $filename |") || die "Cannot start objdump";
236
237while (<FILE>) {
238 my $line = $_;
239 chomp($line);
240 if ($state == 0) {
241 if ($line =~ /^([a-f0-9]+)\:/) {
242 if (InRange($1, $target)) {
243 $state = 1;
244 }
245 }
246 }
247 if ($state == 1) {
248 if ($line =~ /^([a-f0-9][a-f0-9][a-f0-9][a-f0-9][a-f0-9][a-f0-9]+)\:/) {
249 my $val = $1;
250 if (!InRange($val, $target)) {
251 last;
252 }
253 if ($val eq $target) {
254 $center = $counter;
255 }
256 }
257 $lines[$counter] = $line;
258
259 $counter = $counter + 1;
260 }
261}
262
263close(FILE);
264
265if ($counter == 0) {
266 print "No matching code found \n";
267 exit;
268}
269
270if ($center == -1) {
271 print "No matching code found \n";
272 exit;
273}
274
275my $start;
276my $finish;
277my $codelines = 0;
278my $binarylines = 0;
279# now we go up and down in the array to find how much we want to print
280
281$start = $center;
282
283while ($start > 1) {
284 $start = $start - 1;
285 my $line = $lines[$start];
286 if ($line =~ /^([a-f0-9]+)\:/) {
287 $binarylines = $binarylines + 1;
288 } else {
289 $codelines = $codelines + 1;
290 }
291 if ($codelines > 10) {
292 last;
293 }
294 if ($binarylines > 20) {
295 last;
296 }
297}
298
299
300$finish = $center;
301$codelines = 0;
302$binarylines = 0;
303while ($finish < $counter) {
304 $finish = $finish + 1;
305 my $line = $lines[$finish];
306 if ($line =~ /^([a-f0-9]+)\:/) {
307 $binarylines = $binarylines + 1;
308 } else {
309 $codelines = $codelines + 1;
310 }
311 if ($codelines > 10) {
312 last;
313 }
314 if ($binarylines > 20) {
315 last;
316 }
317}
318
319
320my $i;
321
322
323# start annotating the registers in the asm.
324# this goes from the oopsing point back, so that the annotator
325# can track (opportunistically) which registers got written and
326# whos value no longer is relevant.
327
328$i = $center;
329while ($i >= $start) {
330 $reglines[$i] = process_x86_regs($lines[$i], $center - $i);
331 $i = $i - 1;
332}
333
334$i = $start;
335while ($i < $finish) {
336 my $line;
337 if ($i == $center) {
338 $line = "*$lines[$i] ";
339 } else {
340 $line = " $lines[$i] ";
341 }
342 print $line;
343 if (defined($reglines[$i]) && length($reglines[$i]) > 0) {
344 my $c = 60 - length($line);
345 while ($c > 0) { print " "; $c = $c - 1; };
346 print "| $reglines[$i]";
347 }
348 if ($i == $center) {
349 print "<--- faulting instruction";
350 }
351 print "\n";
352 $i = $i +1;
353}
354
355sub usage {
356 print <<EOT;
357Usage:
358 dmesg | perl $0 [OPTION] [VMLINUX]
359
360OPTION:
361 -c, --cross-compile CROSS_COMPILE Specify the prefix used for toolchain.
362 -m, --module MODULE_DIRNAME Specify the module filename.
363 -h, --help Help.
364EOT
365 exit;
366}