ftrace: create __mcount_loc section

This patch creates a section in the kernel called "__mcount_loc".
This will hold a list of pointers to the mcount relocation for
each call site of mcount.

For example:

objdump -dr init/main.o
[...]
Disassembly of section .text:

0000000000000000 <do_one_initcall>:
   0:   55                      push   %rbp
[...]
000000000000017b <init_post>:
 17b:   55                      push   %rbp
 17c:   48 89 e5                mov    %rsp,%rbp
 17f:   53                      push   %rbx
 180:   48 83 ec 08             sub    $0x8,%rsp
 184:   e8 00 00 00 00          callq  189 <init_post+0xe>
                        185: R_X86_64_PC32      mcount+0xfffffffffffffffc
[...]

We will add a section to point to each function call.

   .section __mcount_loc,"a",@progbits
[...]
   .quad .text + 0x185
[...]

The offset to of the mcount call site in init_post is an offset from
the start of the section, and not the start of the function init_post.
The mcount relocation is at the call site 0x185 from the start of the
.text section.

  .text + 0x185  == init_post + 0xa

We need a way to add this __mcount_loc section in a way that we do not
lose the relocations after final link.  The .text section here will
be attached to all other .text sections after final link and the
offsets will be meaningless.  We need to keep track of where these
.text sections are.

To do this, we use the start of the first function in the section.
do_one_initcall.  We can make a tmp.s file with this function as a reference
to the start of the .text section.

   .section __mcount_loc,"a",@progbits
[...]
   .quad do_one_initcall + 0x185
[...]

Then we can compile the tmp.s into a tmp.o

  gcc -c tmp.s -o tmp.o

And link it into back into main.o.

  ld -r main.o tmp.o -o tmp_main.o
  mv tmp_main.o main.o

But we have a problem.  What happens if the first function in a section
is not exported, and is a static function. The linker will not let
the tmp.o use it.  This case exists in main.o as well.

Disassembly of section .init.text:

0000000000000000 <set_reset_devices>:
   0:   55                      push   %rbp
   1:   48 89 e5                mov    %rsp,%rbp
   4:   e8 00 00 00 00          callq  9 <set_reset_devices+0x9>
                        5: R_X86_64_PC32        mcount+0xfffffffffffffffc

The first function in .init.text is a static function.

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 t set_reset_devices

The lowercase 't' means that set_reset_devices is local and is not exported.
If we simply try to link the tmp.o with the set_reset_devices we end
up with two symbols: one local and one global.

 .section __mcount_loc,"a",@progbits
 .quad set_reset_devices + 0x10

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 t set_reset_devices
                 U set_reset_devices

We still have an undefined reference to set_reset_devices, and if we try
to compile the kernel, we will end up with an undefined reference to
set_reset_devices, or even worst, it could be exported someplace else,
and then we will have a reference to the wrong location.

To handle this case, we make an intermediate step using objcopy.
We convert set_reset_devices into a global exported symbol before linking
it with tmp.o and set it back afterwards.

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 T set_reset_devices

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 T set_reset_devices

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 t set_reset_devices

Now we have a section in main.o called __mcount_loc that we can place
somewhere in the kernel using vmlinux.ld.S and access it to convert
all these locations that call mcount into nops before starting SMP
and thus, eliminating the need to do this with kstop_machine.

Note, A well documented perl script (scripts/recordmcount.pl) is used
to do all this in one location.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

4 files changed, 302 insertions, 0 deletions

diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index 3d8e472a09c8..838d9b2a0da1 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -37,6 +37,13 @@
 #define MEM_DISCARD(sec) *(.mem##sec)
 #endif
 
+#ifdef CONFIG_FTRACE_MCOUNT_RECORD
+#define MCOUNT_REC()    VMLINUX_SYMBOL(__start_mcount_loc) = .; \
+                        *(__mcount_loc)                         \
+                        VMLINUX_SYMBOL(__stop_mcount_loc) = .;
+#else
+#define MCOUNT_REC()
+#endif
 
 /* .data section */
 #define DATA_DATA                                                       \
@@ -192,6 +199,7 @@
         /* __*init sections */                                          \
         __init_rodata : AT(ADDR(__init_rodata) - LOAD_OFFSET) {         \
                 *(.ref.rodata)                                          \
+                MCOUNT_REC()                                            \
                 DEV_KEEP(init.rodata)                                   \
                 DEV_KEEP(exit.rodata)                                   \
                 CPU_KEEP(init.rodata)                                   \
diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
index cae2637d5e68..14d9505178ca 100644
--- a/kernel/trace/Kconfig
+++ b/kernel/trace/Kconfig
@@ -7,6 +7,9 @@ config HAVE_FTRACE
 config HAVE_DYNAMIC_FTRACE
         bool
 
+config HAVE_FTRACE_MCOUNT_RECORD
+        bool
+
 config TRACER_MAX_TRACE
         bool
 
@@ -122,6 +125,11 @@ config DYNAMIC_FTRACE
          were made. If so, it runs stop_machine (stops all CPUS)
          and modifies the code to jump over the call to ftrace.
 
+config FTRACE_MCOUNT_RECORD
+        def_bool y
+        depends on DYNAMIC_FTRACE
+        depends on HAVE_FTRACE_MCOUNT_RECORD
+
 config FTRACE_SELFTEST
         bool
 
diff --git a/scripts/Makefile.build b/scripts/Makefile.build
index 277cfe0b7100..463ddcc583ed 100644
--- a/scripts/Makefile.build
+++ b/scripts/Makefile.build
@@ -198,10 +198,16 @@ cmd_modversions =							\
         fi;
 endif
 
+ifdef CONFIG_FTRACE_MCOUNT_RECORD
+cmd_record_mcount = scripts/recordmcount.pl "$(ARCH)" \
+        "$(OBJDUMP)" "$(OBJCOPY)" "$(CC)" "$(LD)" "$(NM)" "$(RM)" "$(MV)" "$(@)";
+endif
+
 define rule_cc_o_c
         $(call echo-cmd,checksrc) $(cmd_checksrc)                         \
         $(call echo-cmd,cc_o_c) $(cmd_cc_o_c);                            \
         $(cmd_modversions)                                                \
+        $(cmd_record_mcount)                                              \
         scripts/basic/fixdep $(depfile) $@ '$(call make-cmd,cc_o_c)' >    \
                                                       $(dot-target).tmp;  \
         rm -f $(depfile);                                                 \
diff --git a/scripts/recordmcount.pl b/scripts/recordmcount.pl
new file mode 100755
index 000000000000..44b4b23e91b2
--- /dev/null
+++ b/scripts/recordmcount.pl
@@ -0,0 +1,280 @@
+#!/usr/bin/perl -w
+# (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.
+#
+#
+# What we want to end up with is a section in vmlinux called
+# __mcount_loc that contains a list of pointers to all the
+# call sites in the kernel that call mcount. 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.
+#
+# This is no easy feat. This script is called just after the original
+# object is compiled and before it is linked.
+#
+# The references to the call sites are offsets from the section of text
+# 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.
+#
+# The trick is to find a way to record the beginning of the section.
+# The way we do this is to look at the first function in the section
+# which will also be the location of that section after final link.
+# e.g.
+#
+#  .section ".text.sched"
+#  .globl my_func
+#  my_func:
+#        [...]
+#        call mcount  (offset: 0x5)
+#        [...]
+#        ret
+#  other_func:
+#        [...]
+#        call mcount (offset: 0x1b)
+#        [...]
+#
+# Both relocation offsets for the mcounts in the above example will be
+# offset from .text.sched. If we make another file called tmp.s with:
+#
+#  .section __mcount_loc
+#  .quad  my_func + 0x5
+#  .quad  my_func + 0x1b
+#
+# We can then compile this tmp.s into tmp.o, and link it to the original
+# object.
+#
+# But this gets hard if my_func is not globl (a static function).
+# In such a case we have:
+#
+#  .section ".text.sched"
+#  my_func:
+#        [...]
+#        call mcount  (offset: 0x5)
+#        [...]
+#        ret
+#  .globl my_func
+#  other_func:
+#        [...]
+#        call mcount (offset: 0x1b)
+#        [...]
+#
+# If we make the tmp.s the same as above, when we link together with
+# the original object, we will end up with two symbols for my_func:
+# one local, one global.  After final compile, we will end up with
+# an undefined reference to my_func.
+#
+# 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
+# file after it is linked together. To do this, we convert the my_func
+# into a global symbol before linking tmp.o. Then after we link tmp.o
+# we will only have a single symbol for my_func that is global.
+# We can convert my_func back into a local symbol and we are done.
+#
+# Here are the steps we take:
+#
+# 1) Record all the local symbols by using 'nm'
+# 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.
+#    End.
+# 8) Link the object with the list object.
+# 9) Move the result back to the original object.
+#    End.
+#
+
+use strict;
+
+my $P = $0;
+$P =~ s@.*/@@g;
+
+my $V = '0.1';
+
+if ($#ARGV < 6) {
+        print "usage: $P arch objdump objcopy cc ld nm rm mv inputfile\n";
+        print "version: $V\n";
+        exit(1);
+}
+
+my ($arch, $objdump, $objcopy, $cc, $ld, $nm, $rm, $mv, $inputfile) = @ARGV;
+
+$objdump = "objdump" if ((length $objdump) == 0);
+$objcopy = "objcopy" if ((length $objcopy) == 0);
+$cc = "gcc" if ((length $cc) == 0);
+$ld = "ld" if ((length $ld) == 0);
+$nm = "nm" if ((length $nm) == 0);
+$rm = "rm" if ((length $rm) == 0);
+$mv = "mv" if ((length $mv) == 0);
+
+#print STDERR "running: $P '$arch' '$objdump' '$objcopy' '$cc' '$ld' " .
+#    "'$nm' '$rm' '$mv' '$inputfile'\n";
+
+my %locals;
+my %convert;
+
+my $type;
+my $section_regex;      # Find the start of a section
+my $function_regex;     # Find the name of a function (return func name)
+my $mcount_regex;       # Find the call site to mcount (return offset)
+
+if ($arch eq "x86_64") {
+    $section_regex = "Disassembly of section";
+    $function_regex = "<(.*?)>:";
+    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\smcount([+-]0x[0-9a-zA-Z]+)?\$";
+    $type = ".quad";
+} elsif ($arch eq "i386") {
+    $section_regex = "Disassembly of section";
+    $function_regex = "<(.*?)>:";
+    $mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\smcount\$";
+    $type = ".long";
+} else {
+    die "Arch $arch is not supported with CONFIG_FTRACE_MCOUNT_RECORD";
+}
+
+my $text_found = 0;
+my $read_function = 0;
+my $opened = 0;
+my $text = "";
+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_" . $prefix . ".o";
+
+#
+# Step 1: find all the local symbols (static functions).
+#
+open (IN, "$nm $inputfile|") || die "error running $nm";
+while (<IN>) {
+    if (/^[0-9a-fA-F]+\s+t\s+(\S+)/) {
+        $locals{$1} = 1;
+    }
+}
+close(IN);
+
+#
+# Step 2: find the sections and mcount call sites
+#
+open(IN, "$objdump -dr $inputfile|") || die "error running $objdump";
+
+while (<IN>) {
+    # is it a section?
+    if (/$section_regex/) {
+        $read_function = 1;
+        $text_found = 0;
+    # section found, now is this a start of a function?
+    } elsif ($read_function && /$function_regex/) {
+        $read_function = 0;
+        $text_found = 1;
+        $text = $1;
+        # is this function static? If so, note this fact.
+        if (defined $locals{$text}) {
+            $convert{$text} = 1;
+        }
+    # is this a call site to mcount? If so, print the offset from the section
+    } elsif ($text_found && /$mcount_regex/) {
+        if (!$opened) {
+            open(FILE, ">$mcount_s") || die "can't create $mcount_s\n";
+            $opened = 1;
+            print FILE "\t.section $mcount_section,\"a\",\@progbits\n";
+        }
+        print FILE "\t$type $text + 0x$1\n";
+    }
+}
+
+# If we did not find any mcount callers, we are done (do nothing).
+if (!$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);