linux-next: ftrace tree build failure

linux-next: ftrace tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 744 bytes --]

Hi all,

Today's linux-next build (x86_64 allmodconfig) failed like this:

x86_64-unknown-linux-gnu-objcopy: unrecognized option `--globalize-symbol'

Introduced by commit 46483cdbb5dc06d84d9572bac1b90f6744c7fc95 ("ftrace:
create __mcount_loc section").

The currently documented minimum requirements for binutils is 2.12.

$ x86_64-unknown-linux-gnu-objcopy --version
GNU objcopy 2.16.1

For now I have reverted commit c2ffd56692cc4b762c175fad69b70a4ec0dfca3f
("ftrace: enable using mcount recording on x86") since that is the least
intrusive thing to do and the above change depends on
CONFIG_FTRACE_MCOUNT_RECORD.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: ftrace tree build failure

[Ingo Molnar]

* Stephen Rothwell <sfr@canb.auug.org.au> wrote:

> Hi all,
> 
> Today's linux-next build (x86_64 allmodconfig) failed like this:
> 
> x86_64-unknown-linux-gnu-objcopy: unrecognized option `--globalize-symbol'
> 
> Introduced by commit 46483cdbb5dc06d84d9572bac1b90f6744c7fc95 ("ftrace:
> create __mcount_loc section").
> 
> The currently documented minimum requirements for binutils is 2.12.
> 
> $ x86_64-unknown-linux-gnu-objcopy --version
> GNU objcopy 2.16.1
> 
> For now I have reverted commit c2ffd56692cc4b762c175fad69b70a4ec0dfca3f
> ("ftrace: enable using mcount recording on x86") since that is the least
> intrusive thing to do and the above change depends on
> CONFIG_FTRACE_MCOUNT_RECORD.

ok. The binutils i've used for testing is 2.18 and this is indeed a new 
option since 2.17. Steve?

	Ingo

Re: linux-next: ftrace tree build failure

[Steven Rostedt]

On Mon, 25 Aug 2008, Ingo Molnar wrote:

> 
> * Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> 
> > Hi all,
> > 
> > Today's linux-next build (x86_64 allmodconfig) failed like this:
> > 
> > x86_64-unknown-linux-gnu-objcopy: unrecognized option `--globalize-symbol'
> > 
> > Introduced by commit 46483cdbb5dc06d84d9572bac1b90f6744c7fc95 ("ftrace:
> > create __mcount_loc section").
> > 
> > The currently documented minimum requirements for binutils is 2.12.
> > 
> > $ x86_64-unknown-linux-gnu-objcopy --version
> > GNU objcopy 2.16.1
> > 
> > For now I have reverted commit c2ffd56692cc4b762c175fad69b70a4ec0dfca3f
> > ("ftrace: enable using mcount recording on x86") since that is the least
> > intrusive thing to do and the above change depends on
> > CONFIG_FTRACE_MCOUNT_RECORD.
> 
> ok. The binutils i've used for testing is 2.18 and this is indeed a new 
> option since 2.17. Steve?


Ouch! I didn't realize that.

OK, I can test to see if objdump is >= version 2.18 and if it is not, I 
can simply not mark those functions. The current mcount for MCOUNT_RECORD 
is simply a "ret". No saving of registers or anything.

-- Steve

Re: linux-next: ftrace tree build failure

[Steven Rostedt]

On Mon, 25 Aug 2008, Steven Rostedt wrote:
> > 
> > ok. The binutils i've used for testing is 2.18 and this is indeed a new 
> > option since 2.17. Steve?
> 
> 
> Ouch! I didn't realize that.
> 
> OK, I can test to see if objdump is >= version 2.18 and if it is not, I 
> can simply not mark those functions. The current mcount for MCOUNT_RECORD 
> is simply a "ret". No saving of registers or anything.

>= 2.17?

Hmm, I need to find a box with an old version of objcopy.

-- Steve

[PATCH] ftrace: objcopy version test for local symbols

[Steven Rostedt]

From: Steven Rostedt <srostedt@redhat.com>
Subject: ftrace: objcopy version test for local symbols

The --globalize-symbols option came out in objcopy version 2.17.
If the kernel is being compiled on a system with a lower version of
objcopy, then we can not use the globalize / localize trick to
link to symbols pointing to local functions.

This patch tests the version of objcopy and will only use the trick
if the version is greater than or equal to 2.17. Otherwise, if an
object has only local functions within a section, it will give a
nice warning and recommend the user to upgrade their objcopy.

Leaving the symbols unrecorded is not that big of a deal, since the
mcount record method changes the actual mcount code to be a simple
"ret" without recording registers or anything.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
---
 scripts/recordmcount.pl |   41 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 41 insertions(+)

Index: linux-tip.git/scripts/recordmcount.pl
===================================================================
--- linux-tip.git.orig/scripts/recordmcount.pl	2008-08-25 14:02:03.000000000 -0400
+++ linux-tip.git/scripts/recordmcount.pl	2008-08-25 14:38:36.000000000 -0400
@@ -187,6 +187,36 @@ my $mcount_s = $dirname . "/.tmp_mc_" . 
 my $mcount_o = $dirname . "/.tmp_mc_" . $prefix . ".o";
 
 #
+# --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.
+my $use_locals = 01;
+my $local_warn_once = 0;
+my $found_version = 0;
+
+open (IN, "$objcopy --version |") || die "error running $objcopy";
+while (<IN>) {
+    if (/objcopy.*\s(\d+)\.(\d+)/) {
+	my $major = $1;
+	my $minor = $2;
+
+	$found_version = 1;
+	if ($major < 2 ||
+	    ($major == 2 && $minor < 17)) {
+	    $use_locals = 0;
+	}
+	last;
+    }
+}
+close (IN);
+
+if (!$found_version) {
+    print STDERR "WARNING: could not find objcopy version.\n" .
+	"\tDisabling local function references.\n";
+}
+
+
+#
 # Step 1: find all the local (static functions) and weak symbols.
 #        't' is local, 'w/W' is weak (we never use a weak function)
 #
@@ -229,6 +259,17 @@ sub update_funcs
 
     # is this function static? If so, note this fact.
     if (defined $locals{$ref_func}) {
+
+	# only use locals if objcopy supports globalize-symbols
+	if (!$use_locals) {
+	    print STDERR
+		"$inputfile: WARNING: referencing local function " .
+		"$ref_func for mcount\n" .
+		"\tConsider upgrading objcopy to support the globalize-" .
+		"symbols option.\n"
+		if (!$local_warn_once++);
+	    return;
+	}
 	$convert{$ref_func} = 1;
     }
 

Re: [PATCH] ftrace: objcopy version test for local symbols

[Ingo Molnar]

* Steven Rostedt <rostedt@goodmis.org> wrote:

> From: Steven Rostedt <srostedt@redhat.com>
> Subject: ftrace: objcopy version test for local symbols
> 
> The --globalize-symbols option came out in objcopy version 2.17.
> If the kernel is being compiled on a system with a lower version of
> objcopy, then we can not use the globalize / localize trick to
> link to symbols pointing to local functions.
> 
> This patch tests the version of objcopy and will only use the trick
> if the version is greater than or equal to 2.17. Otherwise, if an
> object has only local functions within a section, it will give a
> nice warning and recommend the user to upgrade their objcopy.
> 
> Leaving the symbols unrecorded is not that big of a deal, since the
> mcount record method changes the actual mcount code to be a simple
> "ret" without recording registers or anything.
> 
> Signed-off-by: Steven Rostedt <srostedt@redhat.com>

applied to tip/tracing/ftrace, also have integrated it into 
auto-ftrace-next. (without much testing - i hope it goes well)

	Ingo

Re: [PATCH] ftrace: objcopy version test for local symbols

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 425 bytes --]

Hi Ingo, Steve,

On Mon, 25 Aug 2008 21:01:03 +0200 Ingo Molnar <mingo@elte.hu> wrote:
>
> applied to tip/tracing/ftrace, also have integrated it into 
> auto-ftrace-next. (without much testing - i hope it goes well)

Looks ok at a first glance, but it may spam my build logs a bit ...
Anyway, we will see.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: [PATCH] ftrace: objcopy version test for local symbols

[Stephen Rothwell]

Hi Ingo, Steve,

On Tue, 26 Aug 2008 09:46:27 +1000 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>
> On Mon, 25 Aug 2008 21:01:03 +0200 Ingo Molnar <mingo@elte.hu> wrote:
> >
> > applied to tip/tracing/ftrace, also have integrated it into 
> > auto-ftrace-next. (without much testing - i hope it goes well)
> 
> Looks ok at a first glance, but it may spam my build logs a bit ...
> Anyway, we will see.

I was right, it produces way to many messages :-(

I applied the following patch for now.  It would be nice if this check
could be done once per build and warned about once as well.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

>From 54bc3035c76a242d14df554966110e5964b3eac3 Mon Sep 17 00:00:00 2001
From: Stephen Rothwell <sfr@canb.auug.org.au>
Date: Tue, 26 Aug 2008 11:45:01 +1000
Subject: [PATCH] ftrace: silence overly verbose warning

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
---
 scripts/recordmcount.pl |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/scripts/recordmcount.pl b/scripts/recordmcount.pl
index ee9e126..ef0c6db 100755
--- a/scripts/recordmcount.pl
+++ b/scripts/recordmcount.pl
@@ -262,12 +262,12 @@ sub update_funcs
 
 	# only use locals if objcopy supports globalize-symbols
 	if (!$use_locals) {
-	    print STDERR
-		"$inputfile: WARNING: referencing local function " .
-		"$ref_func for mcount\n" .
-		"\tConsider upgrading objcopy to support the globalize-" .
-		"symbols option.\n"
-		if (!$local_warn_once++);
+#	    print STDERR
+#		"$inputfile: WARNING: referencing local function " .
+#		"$ref_func for mcount\n" .
+#		"\tConsider upgrading objcopy to support the globalize-" .
+#		"symbols option.\n"
+#		if (!$local_warn_once++);
 	    return;
 	}
 	$convert{$ref_func} = 1;
-- 
1.5.6.3

Re: [PATCH] ftrace: objcopy version test for local symbols

[Steven Rostedt]

On Tue, 26 Aug 2008, Stephen Rothwell wrote:

> Hi Ingo, Steve,
> 
> On Tue, 26 Aug 2008 09:46:27 +1000 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> >
> > On Mon, 25 Aug 2008 21:01:03 +0200 Ingo Molnar <mingo@elte.hu> wrote:
> > >
> > > applied to tip/tracing/ftrace, also have integrated it into 
> > > auto-ftrace-next. (without much testing - i hope it goes well)
> > 
> > Looks ok at a first glance, but it may spam my build logs a bit ...
> > Anyway, we will see.
> 
> I was right, it produces way to many messages :-(

I was afraid of that :-(

> 
> I applied the following patch for now.  It would be nice if this check
> could be done once per build and warned about once as well.

Actually, there is a way. I'll write one up and send it out.

-- Steve

[PATCH] ftrace: only warn once on old objcopy and local functions

[Steven Rostedt]

[
  Stephen, can you try to see if this works for you, with the
  issue of spamming warnings?

  Thanks,

  -- Steve
]

Older versions of objcopy do not support the --globalize-symbols
option, which prevent mcount calls in local functions within
their own sections from being reference by the mcount_loc table.

We want to warn the user about this out of date objcopy without spamming
them with warning messages. To do this, we create a .tmp_mcversion and
compare it with the .version to see if we have already warned the
user for that build.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
---
 Makefile                |    3 +-
 scripts/recordmcount.pl |   69 +++++++++++++++++++++++++++++++++++++++++++-----
 2 files changed, 65 insertions(+), 7 deletions(-)

Index: linux-tip.git/Makefile
===================================================================
--- linux-tip.git.orig/Makefile	2008-08-25 14:01:57.000000000 -0400
+++ linux-tip.git/Makefile	2008-08-25 23:17:57.000000000 -0400
@@ -1155,7 +1155,8 @@ endif # CONFIG_MODULES
 # Directories & files removed with 'make clean'
 CLEAN_DIRS  += $(MODVERDIR)
 CLEAN_FILES +=	vmlinux System.map \
-                .tmp_kallsyms* .tmp_version .tmp_vmlinux* .tmp_System.map
+                .tmp_kallsyms* .tmp_version .tmp_vmlinux* .tmp_System.map \
+		.tmp_mcversion
 
 # Directories & files removed with 'make mrproper'
 MRPROPER_DIRS  += include/config include2 usr/include
Index: linux-tip.git/scripts/recordmcount.pl
===================================================================
--- linux-tip.git.orig/scripts/recordmcount.pl	2008-08-25 16:23:29.000000000 -0400
+++ linux-tip.git/scripts/recordmcount.pl	2008-08-25 23:19:52.000000000 -0400
@@ -96,6 +96,8 @@
 use strict;
 
 my $P = $0;
+my $D = $P;
+$D =~ s@(.*)/.*@$1@;
 $P =~ s@.*/@@g;
 
 my $V = '0.1';
@@ -235,6 +237,66 @@ my $ref_func;		# reference function to u
 my $offset = 0;		# offset of ref_func to section beginning
 
 ##
+# warn_on_local - check to see if this make already had a warning
+#
+# We want to warn to the user if the objcopy is out of date
+# for referencing local functions, but we do not want to
+# spam the build with errors. We create a file ".tmp_mcversion"
+# and compare it to the .version file to see if we have already
+# warned the user in this compile or not.
+sub warn_on_local
+{
+    if ($local_warn_once) {
+	return;
+    }
+
+    $local_warn_once++;
+
+    # Check to see if we already warned on this build.
+    my $topdir = $D . "/..";
+    open(VER, "$topdir/.version") || return;
+    my $ver_num = -1;
+    while (<VER>) {
+	if (/(\d+)/) {
+	    $ver_num = $1;
+	    last;
+	}
+    }
+    close (VER);
+
+    my $last_ver = $topdir . "/.tmp_mcversion";
+    if (open(VER, $last_ver)) {
+	my $this_num = -2;
+	# see if the versions match
+	while (<VER>) {
+	    if (/(\d+)/) {
+		$this_num = $1;
+	    }
+	}
+	close (VER);
+
+	if ($ver_num == $this_num) {
+	    return;
+	}
+    }
+
+    # Either the file does not exist, or the versions do
+    # not match. Write the version we read, and warn about
+    # the message.
+
+    open (VER, ">$last_ver") || return;
+    print VER $ver_num . "\n";
+    close (VER);
+
+    print STDERR
+	"$inputfile: WARNING: referencing local function " .
+	"$ref_func for mcount\n" .
+	"\tConsider upgrading objcopy to support the globalize-" .
+	"symbols option.\n" .
+	"\tDisabling local function references.\n";
+}
+
+##
 # update_funcs - print out the current mcount callers
 #
 #  Go through the list of offsets to callers and write them to
@@ -262,12 +324,7 @@ sub update_funcs
 
 	# only use locals if objcopy supports globalize-symbols
 	if (!$use_locals) {
-	    print STDERR
-		"$inputfile: WARNING: referencing local function " .
-		"$ref_func for mcount\n" .
-		"\tConsider upgrading objcopy to support the globalize-" .
-		"symbols option.\n"
-		if (!$local_warn_once++);
+	    warn_on_local();
 	    return;
 	}
 	$convert{$ref_func} = 1;

Re: [PATCH] ftrace: only warn once on old objcopy and local functions

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 473 bytes --]

Hi Steve,

On Mon, 25 Aug 2008 23:25:25 -0400 (EDT) Steven Rostedt <rostedt@goodmis.org> wrote:
>
>   Stephen, can you try to see if this works for you, with the
>   issue of spamming warnings?

I get no warnings at all.  I think this is because I always build with
O=... so that the way you find $topdir (and therefore the .version file)
does not work.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: [PATCH] ftrace: only warn once on old objcopy and local functions

[Steven Rostedt]

On Tue, 26 Aug 2008, Stephen Rothwell wrote:

> Hi Steve,
> 
> On Mon, 25 Aug 2008 23:25:25 -0400 (EDT) Steven Rostedt <rostedt@goodmis.org> wrote:
> >
> >   Stephen, can you try to see if this works for you, with the
> >   issue of spamming warnings?
> 
> I get no warnings at all.  I think this is because I always build with
> O=... so that the way you find $topdir (and therefore the .version file)
> does not work.

Hmm, Well we don't need the warning, although it would be nice to give. 
There seems to be a lot of hacking to get a warning for this. Perhaps
we just ignore the local functions and not warn at all?

-- Steve

[PATCH] ftrace: remove warning of old objcopy and local functions

[Steven Rostedt]

The warning messages about old objcopy and local functions spam the
user quite drastically.  Remove the warning until we can find a nicer
way of tell the user to upgrade their objcopy.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
---
 scripts/recordmcount.pl |    6 ------
 1 file changed, 6 deletions(-)

Index: linux-tip.git/scripts/recordmcount.pl
===================================================================
--- linux-tip.git.orig/scripts/recordmcount.pl	2008-08-27 12:57:40.000000000 -0400
+++ linux-tip.git/scripts/recordmcount.pl	2008-08-27 12:58:03.000000000 -0400
@@ -262,12 +262,6 @@ sub update_funcs
 
 	# only use locals if objcopy supports globalize-symbols
 	if (!$use_locals) {
-	    print STDERR
-		"$inputfile: WARNING: referencing local function " .
-		"$ref_func for mcount\n" .
-		"\tConsider upgrading objcopy to support the globalize-" .
-		"symbols option.\n"
-		if (!$local_warn_once++);
 	    return;
 	}
 	$convert{$ref_func} = 1;

Re: [PATCH] ftrace: remove warning of old objcopy and local functions

[Ingo Molnar]

* Steven Rostedt <rostedt@goodmis.org> wrote:

> The warning messages about old objcopy and local functions spam the 
> user quite drastically.  Remove the warning until we can find a nicer 
> way of tell the user to upgrade their objcopy.

applied to tip/tracing/ftrace, thanks!

	Ingo