recordmcount.pl 18 KB
Newer Older
1
#!/usr/bin/env perl
2 3 4 5 6 7 8
# (c) 2008, Steven Rostedt <srostedt@redhat.com>
# Licensed under the terms of the GNU GPL License version 2
#
# recordmcount.pl - makes a section called __mcount_loc that holds
#                   all the offsets to the calls to mcount.
#
#
9 10 11 12 13 14 15
# What we want to end up with this is that each object file will have a
# section called __mcount_loc that will hold the list of pointers to mcount
# callers. After final linking, the vmlinux will have within .init.data the
# list of all callers to mcount between __start_mcount_loc and __stop_mcount_loc.
# Later on boot up, the kernel will read this list, save the locations and turn
# them into nops. When tracing or profiling is later enabled, these locations
# will then be converted back to pointers to some function.
16 17 18 19
#
# This is no easy feat. This script is called just after the original
# object is compiled and before it is linked.
#
20 21 22 23 24 25 26 27 28 29 30 31 32
# When parse this object file using 'objdump', the references to the call
# sites are offsets from the section that the call site is in. Hence, all
# functions in a section that has a call site to mcount, will have the
# offset from the beginning of the section and not the beginning of the
# function.
#
# But where this section will reside finally in vmlinx is undetermined at
# this point. So we can't use this kind of offsets to record the final
# address of this call site.
#
# The trick is to change the call offset referring the start of a section to
# referring a function symbol in this section. During the link step, 'ld' will
# compute the final address according to the information we record.
33 34 35
#
# e.g.
#
36
#  .section ".sched.text", "ax"
37
#        [...]
38 39 40 41 42 43 44 45
#  func1:
#        [...]
#        call mcount  (offset: 0x10)
#        [...]
#        ret
#  .globl fun2
#  func2:             (offset: 0x20)
#        [...]
46 47
#        [...]
#        ret
48
#  func3:
49
#        [...]
50
#        call mcount (offset: 0x30)
51 52 53
#        [...]
#
# Both relocation offsets for the mcounts in the above example will be
54 55
# offset from .sched.text. If we choose global symbol func2 as a reference and
# make another file called tmp.s with the new offsets:
56 57
#
#  .section __mcount_loc
58 59
#  .quad  func2 - 0x10
#  .quad  func2 + 0x10
60
#
61
# We can then compile this tmp.s into tmp.o, and link it back to the original
62 63
# object.
#
64 65 66
# In our algorithm, we will choose the first global function we meet in this
# section as the reference. But this gets hard if there is no global functions
# in this section. In such a case we have to select a local one. E.g. func1:
67
#
68
#  .section ".sched.text", "ax"
69
#  func1:
70
#        [...]
71
#        call mcount  (offset: 0x10)
72 73
#        [...]
#        ret
74
#  func2:
75
#        [...]
76
#        call mcount (offset: 0x20)
77
#        [...]
78
#  .section "other.section"
79 80
#
# If we make the tmp.s the same as above, when we link together with
81
# the original object, we will end up with two symbols for func1:
82
# one local, one global.  After final compile, we will end up with
83 84
# an undefined reference to func1 or a wrong reference to another global
# func1 in other files.
85 86 87
#
# Since local objects can reference local variables, we need to find
# a way to make tmp.o reference the local objects of the original object
88
# file after it is linked together. To do this, we convert func1
89
# into a global symbol before linking tmp.o. Then after we link tmp.o
90 91
# we will only have a single symbol for func1 that is global.
# We can convert func1 back into a local symbol and we are done.
92 93 94
#
# Here are the steps we take:
#
95
# 1) Record all the local and weak symbols by using 'nm'
96 97 98 99 100 101 102 103 104 105 106 107 108
# 2) Use objdump to find all the call site offsets and sections for
#    mcount.
# 3) Compile the list into its own object.
# 4) Do we have to deal with local functions? If not, go to step 8.
# 5) Make an object that converts these local functions to global symbols
#    with objcopy.
# 6) Link together this new object with the list object.
# 7) Convert the local functions back to local symbols and rename
#    the result as the original object.
# 8) Link the object with the list object.
# 9) Move the result back to the original object.
#

109
use warnings;
110 111 112 113 114 115 116
use strict;

my $P = $0;
$P =~ s@.*/@@g;

my $V = '0.1';

117 118
if ($#ARGV != 11) {
	print "usage: $P arch endian bits objdump objcopy cc ld nm rm mv is_module inputfile\n";
119 120 121 122
	print "version: $V\n";
	exit(1);
}

123
my ($arch, $endian, $bits, $objdump, $objcopy, $cc,
124
    $ld, $nm, $rm, $mv, $is_module, $inputfile) = @ARGV;
125

126
# This file refers to mcount and shouldn't be ftraced, so lets' ignore it
127
if ($inputfile =~ m,kernel/trace/ftrace\.o$,) {
128 129 130
    exit(0);
}

131 132 133
# Acceptable sections to record.
my %text_sections = (
     ".text" => 1,
134
     ".init.text" => 1,
135
     ".ref.text" => 1,
136 137
     ".sched.text" => 1,
     ".spinlock.text" => 1,
138
     ".irqentry.text" => 1,
139
     ".softirqentry.text" => 1,
140
     ".kprobes.text" => 1,
141
     ".cpuidle.text" => 1,
Jiri Olsa's avatar
Jiri Olsa committed
142
     ".text.unlikely" => 1,
143 144
);

145 146 147 148 149 150 151 152
# Note: we are nice to C-programmers here, thus we skip the '||='-idiom.
$objdump = 'objdump' if (!$objdump);
$objcopy = 'objcopy' if (!$objcopy);
$cc = 'gcc' if (!$cc);
$ld = 'ld' if (!$ld);
$nm = 'nm' if (!$nm);
$rm = 'rm' if (!$rm);
$mv = 'mv' if (!$mv);
153 154 155 156

#print STDERR "running: $P '$arch' '$objdump' '$objcopy' '$cc' '$ld' " .
#    "'$nm' '$rm' '$mv' '$inputfile'\n";

157 158 159
my %locals;		# List of local (static) functions
my %weak;		# List of weak functions
my %convert;		# List of local functions used that needs conversion
160 161

my $type;
162 163
my $local_regex;	# Match a local function (return function)
my $weak_regex; 	# Match a weak function (return function)
164
my $section_regex;	# Find the start of a section
165 166
my $function_regex;	# Find the name of a function
			#    (return offset and func name)
167
my $mcount_regex;	# Find the call site to mcount (return offset)
168
my $mcount_adjust;	# Address adjustment to mcount offset
169
my $alignment;		# The .align value to use for $mcount_section
170
my $section_type;	# Section header plus possible alignment command
171 172
my $can_use_local = 0; 	# If we can use local function references

173 174 175 176 177
# Shut up recordmcount if user has older objcopy
my $quiet_recordmcount = ".tmp_quiet_recordmcount";
my $print_warning = 1;
$print_warning = 0 if ( -f $quiet_recordmcount);

178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
##
# check_objcopy - whether objcopy supports --globalize-symbols
#
#  --globalize-symbols came out in 2.17, we must test the version
#  of objcopy, and if it is less than 2.17, then we can not
#  record local functions.
sub check_objcopy
{
    open (IN, "$objcopy --version |") or die "error running $objcopy";
    while (<IN>) {
	if (/objcopy.*\s(\d+)\.(\d+)/) {
	    $can_use_local = 1 if ($1 > 2 || ($1 == 2 && $2 >= 17));
	    last;
	}
    }
    close (IN);

195
    if (!$can_use_local && $print_warning) {
196 197
	print STDERR "WARNING: could not find objcopy version or version " .
	    "is less than 2.17.\n" .
198 199 200 201
	    "\tLocal function references are disabled.\n";
	open (QUIET, ">$quiet_recordmcount");
	printf QUIET "Disables the warning from recordmcount.pl\n";
	close QUIET;
202 203
    }
}
204

205
if ($arch =~ /(x86(_64)?)|(i386)/) {
206 207 208 209 210 211 212
    if ($bits == 64) {
	$arch = "x86_64";
    } else {
	$arch = "i386";
    }
}

213 214 215 216
#
# We base the defaults off of i386, the other archs may
# feel free to change them in the below if statements.
#
217 218
$local_regex = "^[0-9a-fA-F]+\\s+t\\s+(\\S+)";
$weak_regex = "^[0-9a-fA-F]+\\s+([wW])\\s+(\\S+)";
219 220
$section_regex = "Disassembly of section\\s+(\\S+):";
$function_regex = "^([0-9a-fA-F]+)\\s+<(.*?)>:";
221
$mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s(mcount|__fentry__)\$";
222
$section_type = '@progbits';
223
$mcount_adjust = 0;
224 225
$type = ".long";

226
if ($arch eq "x86_64") {
227
    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s(mcount|__fentry__)([+-]0x[0-9a-zA-Z]+)?\$";
228
    $type = ".quad";
229
    $alignment = 8;
230
    $mcount_adjust = -1;
231 232 233 234 235 236 237

    # force flags for this arch
    $ld .= " -m elf_x86_64";
    $objdump .= " -M x86-64";
    $objcopy .= " -O elf64-x86-64";
    $cc .= " -m64";

238
} elsif ($arch eq "i386") {
239
    $alignment = 4;
240
    $mcount_adjust = -1;
241 242 243 244 245 246 247

    # force flags for this arch
    $ld .= " -m elf_i386";
    $objdump .= " -M i386";
    $objcopy .= " -O elf32-i386";
    $cc .= " -m32";

248
} elsif ($arch eq "s390" && $bits == 64) {
249 250 251 252 253 254 255
    if ($cc =~ /-DCC_USING_HOTPATCH/) {
	$mcount_regex = "^\\s*([0-9a-fA-F]+):\\s*c0 04 00 00 00 00\\s*brcl\\s*0,[0-9a-f]+ <([^\+]*)>\$";
	$mcount_adjust = 0;
    } else {
	$mcount_regex = "^\\s*([0-9a-fA-F]+):\\s*R_390_(PC|PLT)32DBL\\s+_mcount\\+0x2\$";
	$mcount_adjust = -14;
    }
256 257 258 259 260
    $alignment = 8;
    $type = ".quad";
    $ld .= " -m elf64_s390";
    $cc .= " -m64";

Matt Fleming's avatar
Matt Fleming committed
261
} elsif ($arch eq "sh") {
262
    $alignment = 2;
Matt Fleming's avatar
Matt Fleming committed
263 264 265 266 267

    # force flags for this arch
    $ld .= " -m shlelf_linux";
    $objcopy .= " -O elf32-sh-linux";

268
} elsif ($arch eq "powerpc") {
269
    $local_regex = "^[0-9a-fA-F]+\\s+t\\s+(\\.?\\S+)";
270 271
    # See comment in the sparc64 section for why we use '\w'.
    $function_regex = "^([0-9a-fA-F]+)\\s+<(\\.?\\w*?)>:";
272
    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s\\.?_mcount\$";
273

274 275 276 277
    if ($bits == 64) {
	$type = ".quad";
    }

278 279 280
} elsif ($arch eq "arm") {
    $alignment = 2;
    $section_type = '%progbits';
281
    $mcount_regex = "^\\s*([0-9a-fA-F]+):\\s*R_ARM_(CALL|PC24|THM_CALL)" .
282
			"\\s+(__gnu_mcount_nc|mcount)\$";
283

284 285 286 287 288
} elsif ($arch eq "arm64") {
    $alignment = 3;
    $section_type = '%progbits';
    $mcount_regex = "^\\s*([0-9a-fA-F]+):\\s*R_AARCH64_CALL26\\s+_mcount\$";
    $type = ".quad";
289 290 291 292 293 294 295
} elsif ($arch eq "ia64") {
    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s_mcount\$";
    $type = "data8";

    if ($is_module eq "0") {
        $cc .= " -mconstant-gp";
    }
296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
} elsif ($arch eq "sparc64") {
    # In the objdump output there are giblets like:
    # 0000000000000000 <igmp_net_exit-0x18>:
    # As there's some data blobs that get emitted into the
    # text section before the first instructions and the first
    # real symbols.  We don't want to match that, so to combat
    # this we use '\w' so we'll match just plain symbol names,
    # and not those that also include hex offsets inside of the
    # '<>' brackets.  Actually the generic function_regex setting
    # could safely use this too.
    $function_regex = "^([0-9a-fA-F]+)\\s+<(\\w*?)>:";

    # Sparc64 calls '_mcount' instead of plain 'mcount'.
    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s_mcount\$";

    $alignment = 8;
    $type = ".xword";
    $ld .= " -m elf64_sparc";
    $cc .= " -m64";
    $objcopy .= " -O elf64-sparc";
316 317 318 319 320 321 322
} elsif ($arch eq "mips") {
    # To enable module support, we need to enable the -mlong-calls option
    # of gcc for module, after using this option, we can not get the real
    # offset of the calling to _mcount, but the offset of the lui
    # instruction or the addiu one. herein, we record the address of the
    # first one, and then we can replace this instruction by a branch
    # instruction to jump over the profiling function to filter the
323
    # indicated functions, or switch back to the lui instruction to trace
324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345
    # them, which means dynamic tracing.
    #
    #       c:	3c030000 	lui	v1,0x0
    #			c: R_MIPS_HI16	_mcount
    #			c: R_MIPS_NONE	*ABS*
    #			c: R_MIPS_NONE	*ABS*
    #      10:	64630000 	daddiu	v1,v1,0
    #			10: R_MIPS_LO16	_mcount
    #			10: R_MIPS_NONE	*ABS*
    #			10: R_MIPS_NONE	*ABS*
    #      14:	03e0082d 	move	at,ra
    #      18:	0060f809 	jalr	v1
    #
    # for the kernel:
    #
    #     10:   03e0082d        move    at,ra
    #	  14:   0c000000        jal     0 <loongson_halt>
    #                    14: R_MIPS_26   _mcount
    #                    14: R_MIPS_NONE *ABS*
    #                    14: R_MIPS_NONE *ABS*
    #	 18:   00020021        nop
    if ($is_module eq "0") {
346
	    $mcount_regex = "^\\s*([0-9a-fA-F]+): R_MIPS_26\\s+_mcount\$";
347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
    } else {
	    $mcount_regex = "^\\s*([0-9a-fA-F]+): R_MIPS_HI16\\s+_mcount\$";
    }
    $objdump .= " -Melf-trad".$endian."mips ";

    if ($endian eq "big") {
	    $endian = " -EB ";
	    $ld .= " -melf".$bits."btsmip";
    } else {
	    $endian = " -EL ";
	    $ld .= " -melf".$bits."ltsmip";
    }

    $cc .= " -mno-abicalls -fno-pic -mabi=" . $bits . $endian;
    $ld .= $endian;

    if ($bits == 64) {
	    $function_regex =
		"^([0-9a-fA-F]+)\\s+<(.|[^\$]L.*?|\$[^L].*?|[^\$][^L].*?)>:";
	    $type = ".dword";
    }
368 369 370
} elsif ($arch eq "microblaze") {
    # Microblaze calls '_mcount' instead of plain 'mcount'.
    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s_mcount\$";
371 372 373
} elsif ($arch eq "blackfin") {
    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s__mcount\$";
    $mcount_adjust = -4;
374 375
} elsif ($arch eq "tilegx" || $arch eq "tile") {
    # Default to the newer TILE-Gx architecture if only "tile" is given.
Tony Lu's avatar
Tony Lu committed
376 377 378
    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s__mcount\$";
    $type = ".quad";
    $alignment = 8;
379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411
} else {
    die "Arch $arch is not supported with CONFIG_FTRACE_MCOUNT_RECORD";
}

my $text_found = 0;
my $read_function = 0;
my $opened = 0;
my $mcount_section = "__mcount_loc";

my $dirname;
my $filename;
my $prefix;
my $ext;

if ($inputfile =~ m,^(.*)/([^/]*)$,) {
    $dirname = $1;
    $filename = $2;
} else {
    $dirname = ".";
    $filename = $inputfile;
}

if ($filename =~ m,^(.*)(\.\S),) {
    $prefix = $1;
    $ext = $2;
} else {
    $prefix = $filename;
    $ext = "";
}

my $mcount_s = $dirname . "/.tmp_mc_" . $prefix . ".s";
my $mcount_o = $dirname . "/.tmp_mc_" . $prefix . ".o";

412
check_objcopy();
413

414
#
415
# Step 1: find all the local (static functions) and weak symbols.
416
#         't' is local, 'w/W' is weak
417 418 419
#
open (IN, "$nm $inputfile|") || die "error running $nm";
while (<IN>) {
420
    if (/$local_regex/) {
421
	$locals{$1} = 1;
422
    } elsif (/$weak_regex/) {
423
	$weak{$2} = $1;
424 425 426 427
    }
}
close(IN);

428 429 430 431 432 433 434 435 436 437 438 439
my @offsets;		# Array of offsets of mcount callers
my $ref_func;		# reference function to use for offsets
my $offset = 0;		# offset of ref_func to section beginning

##
# update_funcs - print out the current mcount callers
#
#  Go through the list of offsets to callers and write them to
#  the output file in a format that can be read by an assembler.
#
sub update_funcs
{
440
    return unless ($ref_func and @offsets);
441

442 443
    # Sanity check on weak function. A weak function may be overwritten by
    # another function of the same name, making all these offsets incorrect.
444
    if (defined $weak{$ref_func}) {
445
	die "$inputfile: ERROR: referencing weak function" .
446 447 448 449 450
	    " $ref_func for mcount\n";
    }

    # is this function static? If so, note this fact.
    if (defined $locals{$ref_func}) {
451 452

	# only use locals if objcopy supports globalize-symbols
453
	if (!$can_use_local) {
454 455
	    return;
	}
456 457 458 459 460
	$convert{$ref_func} = 1;
    }

    # Loop through all the mcount caller offsets and print a reference
    # to the caller based from the ref_func.
461 462 463 464 465 466 467 468
    if (!$opened) {
	open(FILE, ">$mcount_s") || die "can't create $mcount_s\n";
	$opened = 1;
	print FILE "\t.section $mcount_section,\"a\",$section_type\n";
	print FILE "\t.align $alignment\n" if (defined($alignment));
    }
    foreach my $cur_offset (@offsets) {
	printf FILE "\t%s %s + %d\n", $type, $ref_func, $cur_offset - $offset;
469 470 471
    }
}

472 473 474
#
# Step 2: find the sections and mcount call sites
#
475
open(IN, "$objdump -hdr $inputfile|") || die "error running $objdump";
476

477 478
my $text;

479 480

# read headers first
481 482
my $read_headers = 1;

483
while (<IN>) {
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499

    if ($read_headers && /$mcount_section/) {
	#
	# Somehow the make process can execute this script on an
	# object twice. If it does, we would duplicate the mcount
	# section and it will cause the function tracer self test
	# to fail. Check if the mcount section exists, and if it does,
	# warn and exit.
	#
	print STDERR "ERROR: $mcount_section already in $inputfile\n" .
	    "\tThis may be an indication that your build is corrupted.\n" .
	    "\tDelete $inputfile and try again. If the same object file\n" .
	    "\tstill causes an issue, then disable CONFIG_DYNAMIC_FTRACE.\n";
	exit(-1);
    }

500 501
    # is it a section?
    if (/$section_regex/) {
502
	$read_headers = 0;
503 504

	# Only record text sections that we know are safe
505
	$read_function = defined($text_sections{$1});
506
	# print out any recorded offsets
507
	update_funcs();
508 509

	# reset all markers and arrays
510
	$text_found = 0;
511 512 513
	undef($ref_func);
	undef(@offsets);

514 515 516
    # section found, now is this a start of a function?
    } elsif ($read_function && /$function_regex/) {
	$text_found = 1;
517 518 519 520 521 522 523 524
	$text = $2;

	# if this is either a local function or a weak function
	# keep looking for functions that are global that
	# we can use safely.
	if (!defined($locals{$text}) && !defined($weak{$text})) {
	    $ref_func = $text;
	    $read_function = 0;
525
	    $offset = hex $1;
526 527
	} else {
	    # if we already have a function, and this is weak, skip it
528 529 530 531
	    if (!defined($ref_func) && !defined($weak{$text}) &&
		 # PPC64 can have symbols that start with .L and
		 # gcc considers these special. Don't use them!
		 $text !~ /^\.L/) {
532
		$ref_func = $text;
533
		$offset = hex $1;
534
	    }
535
	}
536 537 538
    }
    # is this a call site to mcount? If so, record it to print later
    if ($text_found && /$mcount_regex/) {
539
	push(@offsets, (hex $1) + $mcount_adjust);
540 541 542
    }
}

543
# dump out anymore offsets that may have been found
544
update_funcs();
545

546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611
# If we did not find any mcount callers, we are done (do nothing).
if (!$opened) {
    exit(0);
}

close(FILE);

#
# Step 3: Compile the file that holds the list of call sites to mcount.
#
`$cc -o $mcount_o -c $mcount_s`;

my @converts = keys %convert;

#
# Step 4: Do we have sections that started with local functions?
#
if ($#converts >= 0) {
    my $globallist = "";
    my $locallist = "";

    foreach my $con (@converts) {
	$globallist .= " --globalize-symbol $con";
	$locallist .= " --localize-symbol $con";
    }

    my $globalobj = $dirname . "/.tmp_gl_" . $filename;
    my $globalmix = $dirname . "/.tmp_mx_" . $filename;

    #
    # Step 5: set up each local function as a global
    #
    `$objcopy $globallist $inputfile $globalobj`;

    #
    # Step 6: Link the global version to our list.
    #
    `$ld -r $globalobj $mcount_o -o $globalmix`;

    #
    # Step 7: Convert the local functions back into local symbols
    #
    `$objcopy $locallist $globalmix $inputfile`;

    # Remove the temp files
    `$rm $globalobj $globalmix`;

} else {

    my $mix = $dirname . "/.tmp_mx_" . $filename;

    #
    # Step 8: Link the object with our list of call sites object.
    #
    `$ld -r $inputfile $mcount_o -o $mix`;

    #
    # Step 9: Move the result back to the original object.
    #
    `$mv $mix $inputfile`;
}

# Clean up the temp files
`$rm $mcount_o $mcount_s`;

exit(0);