PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc.a while providing backward binary compatibility.

OK for trunk, 4.9 and 4.9 branches?

Thanks.


H.J.
---
libgcc/

	PR target/65612
	* config/i386/cpuinfo.c (__cpu_model): Initialize.
	(__cpu_indicator_init@GCC_4.8.0): New.
	(__cpu_model@GCC_4.8.0): Likewise.

gcc/testsuite/

	PR target/65612
	* g++.dg/ext/mv18.C: New test.
	* g++.dg/ext/mv19.C: Likewise.
	* g++.dg/ext/mv20.C: Likewise.
---
 gcc/testsuite/g++.dg/ext/mv18.C | 7 +++++++
 gcc/testsuite/g++.dg/ext/mv19.C | 7 +++++++
 gcc/testsuite/g++.dg/ext/mv20.C | 7 +++++++
 libgcc/config/i386/cpuinfo.c    | 7 ++++++-
 4 files changed, 27 insertions(+), 1 deletion(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/mv18.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv19.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv20.C

diff --git a/gcc/testsuite/g++.dg/ext/mv18.C b/gcc/testsuite/g++.dg/ext/mv18.C
new file mode 100644
index 0000000..1f024de
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv18.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv19.C b/gcc/testsuite/g++.dg/ext/mv19.C
new file mode 100644
index 0000000..d1ea788
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv19.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv20.C b/gcc/testsuite/g++.dg/ext/mv20.C
new file mode 100644
index 0000000..98f7408
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv20.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..9639c8d 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };
 
 
 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)
 
   return 0;
 }
+
+#if defined SHARED && !defined _WIN32
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
-- 
2.1.0

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Sun, Mar 29, 2015 at 7:25 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> We shouldn't call external function, __cpu_indicator_init, while an object
> is being relocated since its .got.plt section hasn't been updated.  It
> works for non-PIE since no update on .got.plt section is required.  This
> patch hides __cpu_indicator_init/__cpu_model from linker to force linker
> to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
> in libgcc.a while providing backward binary compatibility.
>
> OK for trunk, 4.9 and 4.9 branches?
>
> Thanks.
>
>
> H.J.
> ---
> libgcc/
>
>         PR target/65612
>         * config/i386/cpuinfo.c (__cpu_model): Initialize.
>         (__cpu_indicator_init@GCC_4.8.0): New.
>         (__cpu_model@GCC_4.8.0): Likewise.
>
> gcc/testsuite/
>
>         PR target/65612
>         * g++.dg/ext/mv18.C: New test.
>         * g++.dg/ext/mv19.C: Likewise.
>         * g++.dg/ext/mv20.C: Likewise.

It doesn' work for shared C++ library:

/export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -O2    -c
-o main.o main.cc
/export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -shared
-fPIC -O2  -o libmv20.so mv20.cc
/export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -O2  -o x
main.o libmv20.so -Wl,-R,.
/usr/local/bin/ld: x: hidden symbol `__cpu_model' in
/export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/../lib/gcc/x86_64-unknown-linux-gnu/5.0.0/libgcc.a(cpuinfo.o)
is referenced by DSO
/usr/local/bin/ld: final link failed: Bad value
collect2: error: ld returned 1 exit status
Makefile:12: recipe for target 'x' failed
make: *** [x] Error 1
[hjl@gnu-tools-1 pr65612]$

-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Sun, Mar 29, 2015 at 7:34 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Sun, Mar 29, 2015 at 7:25 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> We shouldn't call external function, __cpu_indicator_init, while an object
>> is being relocated since its .got.plt section hasn't been updated.  It
>> works for non-PIE since no update on .got.plt section is required.  This
>> patch hides __cpu_indicator_init/__cpu_model from linker to force linker
>> to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
>> in libgcc.a while providing backward binary compatibility.
>>
>> OK for trunk, 4.9 and 4.9 branches?
>>
>> Thanks.
>>
>>
>> H.J.
>> ---
>> libgcc/
>>
>>         PR target/65612
>>         * config/i386/cpuinfo.c (__cpu_model): Initialize.
>>         (__cpu_indicator_init@GCC_4.8.0): New.
>>         (__cpu_model@GCC_4.8.0): Likewise.
>>
>> gcc/testsuite/
>>
>>         PR target/65612
>>         * g++.dg/ext/mv18.C: New test.
>>         * g++.dg/ext/mv19.C: Likewise.
>>         * g++.dg/ext/mv20.C: Likewise.
>
> It doesn' work for shared C++ library:
>
> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -O2    -c
> -o main.o main.cc
> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -shared
> -fPIC -O2  -o libmv20.so mv20.cc
> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -O2  -o x
> main.o libmv20.so -Wl,-R,.
> /usr/local/bin/ld: x: hidden symbol `__cpu_model' in
> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/../lib/gcc/x86_64-unknown-linux-gnu/5.0.0/libgcc.a(cpuinfo.o)
> is referenced by DSO
> /usr/local/bin/ld: final link failed: Bad value
> collect2: error: ld returned 1 exit status
> Makefile:12: recipe for target 'x' failed
> make: *** [x] Error 1
> [hjl@gnu-tools-1 pr65612]$
>
> --
> H.J.

We need something like libgcc_nonshared.a, which contains cpuinfo.o, and
link together with -lgcc_s when creating executable or DSO.


-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

[-- Attachment #1: Type: text/plain, Size: 3420 bytes --]

On Sun, Mar 29, 2015 at 7:40 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Sun, Mar 29, 2015 at 7:34 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Sun, Mar 29, 2015 at 7:25 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> We shouldn't call external function, __cpu_indicator_init, while an object
>>> is being relocated since its .got.plt section hasn't been updated.  It
>>> works for non-PIE since no update on .got.plt section is required.  This
>>> patch hides __cpu_indicator_init/__cpu_model from linker to force linker
>>> to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
>>> in libgcc.a while providing backward binary compatibility.
>>>
>>> OK for trunk, 4.9 and 4.9 branches?
>>>
>>> Thanks.
>>>
>>>
>>> H.J.
>>> ---
>>> libgcc/
>>>
>>>         PR target/65612
>>>         * config/i386/cpuinfo.c (__cpu_model): Initialize.
>>>         (__cpu_indicator_init@GCC_4.8.0): New.
>>>         (__cpu_model@GCC_4.8.0): Likewise.
>>>
>>> gcc/testsuite/
>>>
>>>         PR target/65612
>>>         * g++.dg/ext/mv18.C: New test.
>>>         * g++.dg/ext/mv19.C: Likewise.
>>>         * g++.dg/ext/mv20.C: Likewise.
>>
>> It doesn' work for shared C++ library:
>>
>> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -O2    -c
>> -o main.o main.cc
>> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -shared
>> -fPIC -O2  -o libmv20.so mv20.cc
>> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -O2  -o x
>> main.o libmv20.so -Wl,-R,.
>> /usr/local/bin/ld: x: hidden symbol `__cpu_model' in
>> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/../lib/gcc/x86_64-unknown-linux-gnu/5.0.0/libgcc.a(cpuinfo.o)
>> is referenced by DSO
>> /usr/local/bin/ld: final link failed: Bad value
>> collect2: error: ld returned 1 exit status
>> Makefile:12: recipe for target 'x' failed
>> make: *** [x] Error 1
>> [hjl@gnu-tools-1 pr65612]$
>>
>> --
>> H.J.
>
> We need something like libgcc_nonshared.a, which contains cpuinfo.o, and
> link together with -lgcc_s when creating executable or DSO.
>

I am testing it on Linux/x86-64.  OK for master if regression test
passes?

-- 
H.J.
--
We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc_nonshared.a while providing backward binary compatibility.  The
new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.

gcc/

PR target/65612
* gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
to -lgcc_s/-lgcc/libgcc.a%s.

gcc/testsuite/

PR target/65612
* g++.dg/ext/mv18.C: New test.
* g++.dg/ext/mv19.C: Likewise.
* g++.dg/ext/mv20.C: Likewise.

libgcc/

PR target/65612
* Makefile.in (LIB2ADDSHARED): New.
(LIB2ADDNONSHARED): Likewise.
(libgcc-nonshared-objects): Likewise.
(libgcc_nonshared.a): Likewise.
(libgcc-s-objects): Add $(LIB2ADDSHARED).
(all): Depend on libgcc_nonshared.a.
($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
(install-leaf): Install libgcc_nonshared.a.
* config/i386/cpuinfo.c (__cpu_model): Initialize.
(__cpu_indicator_init@GCC_4.8.0): New.
(__cpu_model@GCC_4.8.0): Likewise.
* config/i386/t-cpuinfo (LIB2ADD): Renamed to ...
(LIB2ADDSHARED): This.

[-- Attachment #2: 0001-Hide-__cpu_indicator_init-__cpu_model-from-linker.patch --]
[-- Type: text/x-patch, Size: 7893 bytes --]

From 2dc6879c0c33069a02ae52d494f05603de9fe317 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sun, 29 Mar 2015 18:03:49 -0700
Subject: [PATCH] Hide __cpu_indicator_init/__cpu_model from linker

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc_nonshared.a while providing backward binary compatibility.  The
new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.

gcc/

	PR target/65612
	* gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
	to -lgcc_s/-lgcc/libgcc.a%s.

gcc/testsuite/

	PR target/65612
	* g++.dg/ext/mv18.C: New test.
	* g++.dg/ext/mv19.C: Likewise.
	* g++.dg/ext/mv20.C: Likewise.

libgcc/

	PR target/65612
	* Makefile.in (LIB2ADDSHARED): New.
	(LIB2ADDNONSHARED): Likewise.
	(libgcc-nonshared-objects): Likewise.
	(libgcc_nonshared.a): Likewise.
	(libgcc-s-objects): Add $(LIB2ADDSHARED).
	(all): Depend on libgcc_nonshared.a.
	($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
	(install-leaf): Install libgcc_nonshared.a.
	* config/i386/cpuinfo.c (__cpu_model): Initialize.
	(__cpu_indicator_init@GCC_4.8.0): New.
	(__cpu_model@GCC_4.8.0): Likewise.
	* config/i386/t-cpuinfo (LIB2ADD): Renamed to ...
	(LIB2ADDSHARED): This.
	(LIB2ADDNONSHARED): New.
---
 gcc/gcc.c                       |  4 ++++
 gcc/testsuite/g++.dg/ext/mv18.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv19.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv20.C |  7 +++++++
 libgcc/Makefile.in              | 20 +++++++++++++++++---
 libgcc/config/i386/cpuinfo.c    |  7 ++++++-
 libgcc/config/i386/t-cpuinfo    |  3 ++-
 7 files changed, 50 insertions(+), 5 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/mv18.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv19.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv20.C

diff --git a/gcc/gcc.c b/gcc/gcc.c
index d956c36..88a8dac 100644
--- a/gcc/gcc.c
+++ b/gcc/gcc.c
@@ -1566,11 +1566,13 @@ init_spec (void)
 	if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
 	  {
 	    init_gcc_specs (&obstack,
+			    "-lgcc_nonshared "
 			    "-lgcc_s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
 			    " -lunwind"
 #endif
 			    ,
+			    "-lgcc_nonshared "
 			    "-lgcc",
 			    "-lgcc_eh"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
@@ -1591,7 +1593,9 @@ init_spec (void)
 	    /* Ug.  We don't know shared library extensions.  Hope that
 	       systems that use this form don't do shared libraries.  */
 	    init_gcc_specs (&obstack,
+			    "libgcc_nonshared.a%s "
 			    "-lgcc_s",
+			    "libgcc_nonshared.a%s "
 			    "libgcc.a%s",
 			    "libgcc_eh.a%s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
diff --git a/gcc/testsuite/g++.dg/ext/mv18.C b/gcc/testsuite/g++.dg/ext/mv18.C
new file mode 100644
index 0000000..1f024de
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv18.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv19.C b/gcc/testsuite/g++.dg/ext/mv19.C
new file mode 100644
index 0000000..d1ea788
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv19.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv20.C b/gcc/testsuite/g++.dg/ext/mv20.C
new file mode 100644
index 0000000..98f7408
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv20.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/libgcc/Makefile.in b/libgcc/Makefile.in
index 88ddfea..cec4ae0 100644
--- a/libgcc/Makefile.in
+++ b/libgcc/Makefile.in
@@ -261,6 +261,14 @@ LIB2ADD =
 # Assembler files should have names ending in `.S'.
 LIB2ADD_ST =
 
+# List of extra C and assembler files to add to shared libgcc2.
+# Assembler files should have names ending in `.S'.
+LIB2ADDSHARED =
+
+# List of extra C and assembler files to add to libgcc_nonshared.a.
+# Assembler files should have names ending in `.S'.
+LIB2ADDNONSHARED =
+
 # Specify the directories to be searched for header files.
 # Both . and srcdir are used, in that order,
 # so that *config.h will be found in the compilation
@@ -806,13 +814,14 @@ endif
 
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD))))
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD_ST))))
+libgcc-nonshared-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADDNONSHARED))))
 
 c_flags :=
 iter-items := $(LIB2ADD) $(LIB2ADD_ST)
 include $(iterator)
 
 ifeq ($(enable_shared),yes)
-libgcc-s-objects += $(addsuffix _s$(objext),$(basename $(notdir $(LIB2ADD))))
+libgcc-s-objects += $(addsuffix _s$(objext),$(basename $(notdir $(LIB2ADD) $(LIB2ADDSHARED))))
 endif
 
 # Build LIB2ADDEH, LIB2ADDEHSTATIC, and LIB2ADDEHSHARED.  If we don't have
@@ -890,8 +899,9 @@ libgcc.a: $(libgcc-objects)
 libgcov.a: $(libgcov-objects)
 libunwind.a: $(libunwind-objects)
 libgcc_eh.a: $(libgcc-eh-objects)
+libgcc_nonshared.a: $(libgcc-nonshared-objects)
 
-libgcc.a libgcov.a libunwind.a libgcc_eh.a:
+libgcc.a libgcov.a libunwind.a libgcc_eh.a libgcc_nonshared.a:
 	-rm -f $@
 
 	objects="$(objects)";					\
@@ -905,7 +915,7 @@ libgcc.a libgcov.a libunwind.a libgcc_eh.a:
 
 	$(RANLIB) $@
 
-all: libgcc.a libgcov.a
+all: libgcc.a libgcc_nonshared.a libgcov.a
 
 ifneq ($(LIBUNWIND),)
 all: libunwind.a
@@ -1051,6 +1061,7 @@ libgcc-extra-parts: $(EXTRA_PARTS)
 all: $(extra-parts)
 
 $(libgcc-objects) $(libgcc-s-objects) $(libgcc-eh-objects) \
+	$(libgcc-nonshared-objects) \
 	$(libgcov-objects) \
 	$(libunwind-objects) $(libunwind-s-objects) \
 	$(EXTRA_PARTS): libgcc_tm.h
@@ -1118,6 +1129,9 @@ install-leaf: $(install-shared) $(install-libunwind)
 	$(INSTALL_DATA) libgcc.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc.a
+	$(INSTALL_DATA) libgcc_nonshared.a $(DESTDIR)$(inst_libdir)/
+	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
+	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
 	$(INSTALL_DATA) libgcov.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcov.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcov.a
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..9639c8d 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };
 
 
 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)
 
   return 0;
 }
+
+#if defined SHARED && !defined _WIN32
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
diff --git a/libgcc/config/i386/t-cpuinfo b/libgcc/config/i386/t-cpuinfo
index 31c8db9..8dbe9e6 100644
--- a/libgcc/config/i386/t-cpuinfo
+++ b/libgcc/config/i386/t-cpuinfo
@@ -1 +1,2 @@
-LIB2ADD += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDSHARED += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDNONSHARED += $(srcdir)/config/i386/cpuinfo.c
-- 
1.9.3

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jack Howarth]

HJ,
     This patch breaks the bootstrap on targets like darwin which
don't build libgcc_nonshared.a...

if test -z "$objects"; then \
  echo 'int __libgcc_eh_dummy;' > eh_dummy.c; \
  /sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xgcc
-B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/
-B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
-B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
/sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
/sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include    -g -O2 -m32 -O2
-g -O2 -DIN_GCC    -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual
-Wno-format -Wstrict-prototypes -Wmissing-prototypes
-Wold-style-definition  -isystem ./include   -pipe -fno-common -g
-DIN_LIBGCC2 -fbuilding-libgcc -fno-stack-protector   -pipe
-fno-common -I. -I. -I../../.././gcc
-I../../../../gcc-5-20150330/libgcc
-I../../../../gcc-5-20150330/libgcc/.
-I../../../../gcc-5-20150330/libgcc/../gcc
-I../../../../gcc-5-20150330/libgcc/../include  -DHAVE_CC_TLS
-DUSE_EMUTLS -fvisibility=hidden -DHIDE_EXPORTS -c eh_dummy.c \
     -o eh_dummy.o; \
  objects=eh_dummy.o; \
fi; \
ar  rc libgcc_nonshared.a $objects
ar: cpuinfo.o: No such file or directory
Makefile:905: recipe for target 'libgcc_nonshared.a' failed
make[5]: *** [libgcc_nonshared.a] Error 1
make[5]: Leaving directory
'/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/i386/libgcc'
Makefile:1168: recipe for target 'multi-do' failed
make[4]: *** [multi-do] Error 1
make[4]: Leaving directory
'/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
Makefile:117: recipe for target 'all-multi' failed
make[3]: *** [all-multi] Error 2
make[3]: Leaving directory
'/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
Makefile:14820: recipe for target 'all-stage1-target-libgcc' failed
make[2]: *** [all-stage1-target-libgcc] Error 2
make[2]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
Makefile:20760: recipe for target 'stage1-bubble' failed
make[1]: *** [stage1-bubble] Error 2
make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
Makefile:21064: recipe for target 'bootstrap' failed
make: *** [bootstrap] Error 2

          Jack

On Mon, Mar 30, 2015 at 6:25 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Sun, Mar 29, 2015 at 7:40 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Sun, Mar 29, 2015 at 7:34 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> On Sun, Mar 29, 2015 at 7:25 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>>> We shouldn't call external function, __cpu_indicator_init, while an object
>>>> is being relocated since its .got.plt section hasn't been updated.  It
>>>> works for non-PIE since no update on .got.plt section is required.  This
>>>> patch hides __cpu_indicator_init/__cpu_model from linker to force linker
>>>> to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
>>>> in libgcc.a while providing backward binary compatibility.
>>>>
>>>> OK for trunk, 4.9 and 4.9 branches?
>>>>
>>>> Thanks.
>>>>
>>>>
>>>> H.J.
>>>> ---
>>>> libgcc/
>>>>
>>>>         PR target/65612
>>>>         * config/i386/cpuinfo.c (__cpu_model): Initialize.
>>>>         (__cpu_indicator_init@GCC_4.8.0): New.
>>>>         (__cpu_model@GCC_4.8.0): Likewise.
>>>>
>>>> gcc/testsuite/
>>>>
>>>>         PR target/65612
>>>>         * g++.dg/ext/mv18.C: New test.
>>>>         * g++.dg/ext/mv19.C: Likewise.
>>>>         * g++.dg/ext/mv20.C: Likewise.
>>>
>>> It doesn' work for shared C++ library:
>>>
>>> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -O2    -c
>>> -o main.o main.cc
>>> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -shared
>>> -fPIC -O2  -o libmv20.so mv20.cc
>>> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/g++ -O2  -o x
>>> main.o libmv20.so -Wl,-R,.
>>> /usr/local/bin/ld: x: hidden symbol `__cpu_model' in
>>> /export/build/gnu/gcc-x32/release/usr/gcc-5.0.0-x32/bin/../lib/gcc/x86_64-unknown-linux-gnu/5.0.0/libgcc.a(cpuinfo.o)
>>> is referenced by DSO
>>> /usr/local/bin/ld: final link failed: Bad value
>>> collect2: error: ld returned 1 exit status
>>> Makefile:12: recipe for target 'x' failed
>>> make: *** [x] Error 1
>>> [hjl@gnu-tools-1 pr65612]$
>>>
>>> --
>>> H.J.
>>
>> We need something like libgcc_nonshared.a, which contains cpuinfo.o, and
>> link together with -lgcc_s when creating executable or DSO.
>>
>
> I am testing it on Linux/x86-64.  OK for master if regression test
> passes?
>
> --
> H.J.
> --
> We shouldn't call external function, __cpu_indicator_init, while an object
> is being relocated since its .got.plt section hasn't been updated.  It
> works for non-PIE since no update on .got.plt section is required.  This
> patch hides __cpu_indicator_init/__cpu_model from linker to force linker
> to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
> in libgcc_nonshared.a while providing backward binary compatibility.  The
> new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.
>
> gcc/
>
> PR target/65612
> * gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
> to -lgcc_s/-lgcc/libgcc.a%s.
>
> gcc/testsuite/
>
> PR target/65612
> * g++.dg/ext/mv18.C: New test.
> * g++.dg/ext/mv19.C: Likewise.
> * g++.dg/ext/mv20.C: Likewise.
>
> libgcc/
>
> PR target/65612
> * Makefile.in (LIB2ADDSHARED): New.
> (LIB2ADDNONSHARED): Likewise.
> (libgcc-nonshared-objects): Likewise.
> (libgcc_nonshared.a): Likewise.
> (libgcc-s-objects): Add $(LIB2ADDSHARED).
> (all): Depend on libgcc_nonshared.a.
> ($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
> (install-leaf): Install libgcc_nonshared.a.
> * config/i386/cpuinfo.c (__cpu_model): Initialize.
> (__cpu_indicator_init@GCC_4.8.0): New.
> (__cpu_model@GCC_4.8.0): Likewise.
> * config/i386/t-cpuinfo (LIB2ADD): Renamed to ...
> (LIB2ADDSHARED): This.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

[-- Attachment #1: Type: text/plain, Size: 2547 bytes --]

On Mon, Mar 30, 2015 at 5:53 PM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
> HJ,
>      This patch breaks the bootstrap on targets like darwin which
> don't build libgcc_nonshared.a...
>
> if test -z "$objects"; then \
>   echo 'int __libgcc_eh_dummy;' > eh_dummy.c; \
>   /sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xgcc
> -B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/
> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include    -g -O2 -m32 -O2
> -g -O2 -DIN_GCC    -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual
> -Wno-format -Wstrict-prototypes -Wmissing-prototypes
> -Wold-style-definition  -isystem ./include   -pipe -fno-common -g
> -DIN_LIBGCC2 -fbuilding-libgcc -fno-stack-protector   -pipe
> -fno-common -I. -I. -I../../.././gcc
> -I../../../../gcc-5-20150330/libgcc
> -I../../../../gcc-5-20150330/libgcc/.
> -I../../../../gcc-5-20150330/libgcc/../gcc
> -I../../../../gcc-5-20150330/libgcc/../include  -DHAVE_CC_TLS
> -DUSE_EMUTLS -fvisibility=hidden -DHIDE_EXPORTS -c eh_dummy.c \
>      -o eh_dummy.o; \
>   objects=eh_dummy.o; \
> fi; \
> ar  rc libgcc_nonshared.a $objects
> ar: cpuinfo.o: No such file or directory
> Makefile:905: recipe for target 'libgcc_nonshared.a' failed
> make[5]: *** [libgcc_nonshared.a] Error 1
> make[5]: Leaving directory
> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/i386/libgcc'
> Makefile:1168: recipe for target 'multi-do' failed
> make[4]: *** [multi-do] Error 1
> make[4]: Leaving directory
> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
> Makefile:117: recipe for target 'all-multi' failed
> make[3]: *** [all-multi] Error 2
> make[3]: Leaving directory
> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
> Makefile:14820: recipe for target 'all-stage1-target-libgcc' failed
> make[2]: *** [all-stage1-target-libgcc] Error 2
> make[2]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
> Makefile:20760: recipe for target 'stage1-bubble' failed
> make[1]: *** [stage1-bubble] Error 2
> make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
> Makefile:21064: recipe for target 'bootstrap' failed
> make: *** [bootstrap] Error 2
>
>           Jack
>

This one works.  I need to add $(LIB2ADDNONSHARED) $(LIB2ADDSHARED)
to iter-items and handle duplicated items in them.

-- 
H.J.

[-- Attachment #2: 0001-Hide-__cpu_indicator_init-__cpu_model-from-linker.patch --]
[-- Type: text/x-patch, Size: 8793 bytes --]

From ba83e7d541929204cb07a3f2d9de34b4f28264b7 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sun, 29 Mar 2015 18:03:49 -0700
Subject: [PATCH] Hide __cpu_indicator_init/__cpu_model from linker

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc_nonshared.a while providing backward binary compatibility.  The
new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.

gcc/

	PR target/65612
	* gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
	to -lgcc_s/-lgcc/libgcc.a%s.

gcc/testsuite/

	PR target/65612
	* g++.dg/ext/mv18.C: New test.
	* g++.dg/ext/mv19.C: Likewise.
	* g++.dg/ext/mv20.C: Likewise.

libgcc/

	PR target/65612
	* Makefile.in (LIB2ADDSHARED): New.
	(LIB2ADDNONSHARED): Likewise.
	(libgcc-nonshared-objects): Likewise.
	(libgcc_nonshared.a): Likewise.
	Check unsupported files in LIB2ADDNONSHARED or LIB2ADDSHARED.
	(iter-items): Add $(LIB2ADDNONSHARED) $(LIB2ADDSHARED).
	(libgcc-s-objects): Add $(LIB2ADDSHARED).
	(all): Depend on libgcc_nonshared.a.
	($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
	(install-leaf): Install libgcc_nonshared.a.
	* shared-object.mk: Check empty $o.
	* config/i386/cpuinfo.c (__cpu_model): Initialize.
	(__cpu_indicator_init@GCC_4.8.0): New.
	(__cpu_model@GCC_4.8.0): Likewise.
	* config/i386/t-cpuinfo (LIB2ADD): Renamed to ...
	(LIB2ADDSHARED): This.
	(LIB2ADDNONSHARED): New.
---
 gcc/gcc.c                       |  4 ++++
 gcc/testsuite/g++.dg/ext/mv18.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv19.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv20.C |  7 +++++++
 libgcc/Makefile.in              | 28 ++++++++++++++++++++++++----
 libgcc/config/i386/cpuinfo.c    |  7 ++++++-
 libgcc/config/i386/t-cpuinfo    |  3 ++-
 libgcc/shared-object.mk         |  2 ++
 8 files changed, 59 insertions(+), 6 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/mv18.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv19.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv20.C

diff --git a/gcc/gcc.c b/gcc/gcc.c
index d956c36..88a8dac 100644
--- a/gcc/gcc.c
+++ b/gcc/gcc.c
@@ -1566,11 +1566,13 @@ init_spec (void)
 	if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
 	  {
 	    init_gcc_specs (&obstack,
+			    "-lgcc_nonshared "
 			    "-lgcc_s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
 			    " -lunwind"
 #endif
 			    ,
+			    "-lgcc_nonshared "
 			    "-lgcc",
 			    "-lgcc_eh"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
@@ -1591,7 +1593,9 @@ init_spec (void)
 	    /* Ug.  We don't know shared library extensions.  Hope that
 	       systems that use this form don't do shared libraries.  */
 	    init_gcc_specs (&obstack,
+			    "libgcc_nonshared.a%s "
 			    "-lgcc_s",
+			    "libgcc_nonshared.a%s "
 			    "libgcc.a%s",
 			    "libgcc_eh.a%s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
diff --git a/gcc/testsuite/g++.dg/ext/mv18.C b/gcc/testsuite/g++.dg/ext/mv18.C
new file mode 100644
index 0000000..1f024de
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv18.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv19.C b/gcc/testsuite/g++.dg/ext/mv19.C
new file mode 100644
index 0000000..d1ea788
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv19.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv20.C b/gcc/testsuite/g++.dg/ext/mv20.C
new file mode 100644
index 0000000..98f7408
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv20.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/libgcc/Makefile.in b/libgcc/Makefile.in
index 88ddfea..f6eb047 100644
--- a/libgcc/Makefile.in
+++ b/libgcc/Makefile.in
@@ -261,6 +261,14 @@ LIB2ADD =
 # Assembler files should have names ending in `.S'.
 LIB2ADD_ST =
 
+# List of extra C and assembler files to add to shared libgcc2.
+# Assembler files should have names ending in `.S'.
+LIB2ADDSHARED =
+
+# List of extra C and assembler files to add to libgcc_nonshared.a.
+# Assembler files should have names ending in `.S'.
+LIB2ADDNONSHARED =
+
 # Specify the directories to be searched for header files.
 # Both . and srcdir are used, in that order,
 # so that *config.h will be found in the compilation
@@ -807,12 +815,19 @@ endif
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD))))
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD_ST))))
 
+# Build LIB2ADDNONSHARED and LIB2ADDSHARED.
+ifneq ($(filter-out %.c %.S %.asm,$(LIB2ADDNONSHARED) $(LIB2ADDSHARED)),)
+$(error Unsupported files in LIB2ADDNONSHARED or LIB2ADDSHARED.)
+endif
+
+libgcc-nonshared-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADDNONSHARED))))
+
 c_flags :=
-iter-items := $(LIB2ADD) $(LIB2ADD_ST)
+iter-items := $(LIB2ADD) $(LIB2ADD_ST) $(LIB2ADDNONSHARED) $(LIB2ADDSHARED)
 include $(iterator)
 
 ifeq ($(enable_shared),yes)
-libgcc-s-objects += $(addsuffix _s$(objext),$(basename $(notdir $(LIB2ADD))))
+libgcc-s-objects += $(addsuffix _s$(objext),$(basename $(notdir $(LIB2ADD) $(LIB2ADDSHARED))))
 endif
 
 # Build LIB2ADDEH, LIB2ADDEHSTATIC, and LIB2ADDEHSHARED.  If we don't have
@@ -890,8 +905,9 @@ libgcc.a: $(libgcc-objects)
 libgcov.a: $(libgcov-objects)
 libunwind.a: $(libunwind-objects)
 libgcc_eh.a: $(libgcc-eh-objects)
+libgcc_nonshared.a: $(libgcc-nonshared-objects)
 
-libgcc.a libgcov.a libunwind.a libgcc_eh.a:
+libgcc.a libgcov.a libunwind.a libgcc_eh.a libgcc_nonshared.a:
 	-rm -f $@
 
 	objects="$(objects)";					\
@@ -905,7 +921,7 @@ libgcc.a libgcov.a libunwind.a libgcc_eh.a:
 
 	$(RANLIB) $@
 
-all: libgcc.a libgcov.a
+all: libgcc.a libgcc_nonshared.a libgcov.a
 
 ifneq ($(LIBUNWIND),)
 all: libunwind.a
@@ -1051,6 +1067,7 @@ libgcc-extra-parts: $(EXTRA_PARTS)
 all: $(extra-parts)
 
 $(libgcc-objects) $(libgcc-s-objects) $(libgcc-eh-objects) \
+	$(libgcc-nonshared-objects) \
 	$(libgcov-objects) \
 	$(libunwind-objects) $(libunwind-s-objects) \
 	$(EXTRA_PARTS): libgcc_tm.h
@@ -1118,6 +1135,9 @@ install-leaf: $(install-shared) $(install-libunwind)
 	$(INSTALL_DATA) libgcc.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc.a
+	$(INSTALL_DATA) libgcc_nonshared.a $(DESTDIR)$(inst_libdir)/
+	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
+	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
 	$(INSTALL_DATA) libgcov.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcov.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcov.a
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..9639c8d 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };
 
 
 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)
 
   return 0;
 }
+
+#if defined SHARED && !defined _WIN32
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
diff --git a/libgcc/config/i386/t-cpuinfo b/libgcc/config/i386/t-cpuinfo
index 31c8db9..8dbe9e6 100644
--- a/libgcc/config/i386/t-cpuinfo
+++ b/libgcc/config/i386/t-cpuinfo
@@ -1 +1,2 @@
-LIB2ADD += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDSHARED += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDNONSHARED += $(srcdir)/config/i386/cpuinfo.c
diff --git a/libgcc/shared-object.mk b/libgcc/shared-object.mk
index efac797..306176d 100644
--- a/libgcc/shared-object.mk
+++ b/libgcc/shared-object.mk
@@ -4,6 +4,7 @@
 o := $(firstword $(iter-items))
 iter-items := $(filter-out $o,$(iter-items))
 
+ifneq ($o,)
 base := $(basename $(notdir $o))
 
 c_flags-$o := $(c_flags)
@@ -36,3 +37,4 @@ $(base)_s$(objext): $o
 	$(gcc_s_compile) -c $(as_flags-$<) $<
 
 endif
+endif
-- 
2.1.0

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Mon, Mar 30, 2015 at 7:08 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Mon, Mar 30, 2015 at 5:53 PM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>> HJ,
>>      This patch breaks the bootstrap on targets like darwin which
>> don't build libgcc_nonshared.a...
>>
>> if test -z "$objects"; then \
>>   echo 'int __libgcc_eh_dummy;' > eh_dummy.c; \
>>   /sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xgcc
>> -B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/
>> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
>> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
>> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
>> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include    -g -O2 -m32 -O2
>> -g -O2 -DIN_GCC    -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual
>> -Wno-format -Wstrict-prototypes -Wmissing-prototypes
>> -Wold-style-definition  -isystem ./include   -pipe -fno-common -g
>> -DIN_LIBGCC2 -fbuilding-libgcc -fno-stack-protector   -pipe
>> -fno-common -I. -I. -I../../.././gcc
>> -I../../../../gcc-5-20150330/libgcc
>> -I../../../../gcc-5-20150330/libgcc/.
>> -I../../../../gcc-5-20150330/libgcc/../gcc
>> -I../../../../gcc-5-20150330/libgcc/../include  -DHAVE_CC_TLS
>> -DUSE_EMUTLS -fvisibility=hidden -DHIDE_EXPORTS -c eh_dummy.c \
>>      -o eh_dummy.o; \
>>   objects=eh_dummy.o; \
>> fi; \
>> ar  rc libgcc_nonshared.a $objects
>> ar: cpuinfo.o: No such file or directory
>> Makefile:905: recipe for target 'libgcc_nonshared.a' failed
>> make[5]: *** [libgcc_nonshared.a] Error 1
>> make[5]: Leaving directory
>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/i386/libgcc'
>> Makefile:1168: recipe for target 'multi-do' failed
>> make[4]: *** [multi-do] Error 1
>> make[4]: Leaving directory
>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
>> Makefile:117: recipe for target 'all-multi' failed
>> make[3]: *** [all-multi] Error 2
>> make[3]: Leaving directory
>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
>> Makefile:14820: recipe for target 'all-stage1-target-libgcc' failed
>> make[2]: *** [all-stage1-target-libgcc] Error 2
>> make[2]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
>> Makefile:20760: recipe for target 'stage1-bubble' failed
>> make[1]: *** [stage1-bubble] Error 2
>> make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
>> Makefile:21064: recipe for target 'bootstrap' failed
>> make: *** [bootstrap] Error 2
>>
>>           Jack
>>
>
> This one works.  I need to add $(LIB2ADDNONSHARED) $(LIB2ADDSHARED)
> to iter-items and handle duplicated items in them.
>

There is no regression on Linux/x86-64.

-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jack Howarth]

On Mon, Mar 30, 2015 at 10:42 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Mon, Mar 30, 2015 at 7:08 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Mon, Mar 30, 2015 at 5:53 PM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>>> HJ,
>>>      This patch breaks the bootstrap on targets like darwin which
>>> don't build libgcc_nonshared.a...
>>>
>>> if test -z "$objects"; then \
>>>   echo 'int __libgcc_eh_dummy;' > eh_dummy.c; \
>>>   /sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xgcc
>>> -B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/
>>> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
>>> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
>>> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
>>> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include    -g -O2 -m32 -O2
>>> -g -O2 -DIN_GCC    -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual
>>> -Wno-format -Wstrict-prototypes -Wmissing-prototypes
>>> -Wold-style-definition  -isystem ./include   -pipe -fno-common -g
>>> -DIN_LIBGCC2 -fbuilding-libgcc -fno-stack-protector   -pipe
>>> -fno-common -I. -I. -I../../.././gcc
>>> -I../../../../gcc-5-20150330/libgcc
>>> -I../../../../gcc-5-20150330/libgcc/.
>>> -I../../../../gcc-5-20150330/libgcc/../gcc
>>> -I../../../../gcc-5-20150330/libgcc/../include  -DHAVE_CC_TLS
>>> -DUSE_EMUTLS -fvisibility=hidden -DHIDE_EXPORTS -c eh_dummy.c \
>>>      -o eh_dummy.o; \
>>>   objects=eh_dummy.o; \
>>> fi; \
>>> ar  rc libgcc_nonshared.a $objects
>>> ar: cpuinfo.o: No such file or directory
>>> Makefile:905: recipe for target 'libgcc_nonshared.a' failed
>>> make[5]: *** [libgcc_nonshared.a] Error 1
>>> make[5]: Leaving directory
>>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/i386/libgcc'
>>> Makefile:1168: recipe for target 'multi-do' failed
>>> make[4]: *** [multi-do] Error 1
>>> make[4]: Leaving directory
>>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
>>> Makefile:117: recipe for target 'all-multi' failed
>>> make[3]: *** [all-multi] Error 2
>>> make[3]: Leaving directory
>>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
>>> Makefile:14820: recipe for target 'all-stage1-target-libgcc' failed
>>> make[2]: *** [all-stage1-target-libgcc] Error 2
>>> make[2]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
>>> Makefile:20760: recipe for target 'stage1-bubble' failed
>>> make[1]: *** [stage1-bubble] Error 2
>>> make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
>>> Makefile:21064: recipe for target 'bootstrap' failed
>>> make: *** [bootstrap] Error 2
>>>
>>>           Jack
>>>
>>
>> This one works.  I need to add $(LIB2ADDNONSHARED) $(LIB2ADDSHARED)
>> to iter-items and handle duplicated items in them.
>>
>
> There is no regression on Linux/x86-64.
>
> --
> H.J.

Can this wait for 5.2? Creating new libgcc libraries seems really
invasive for stage4.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jack Howarth]

H.J.,
   This still breaks the darwin bootstrap but differently.

ar  rc libgcc_eh.a $objects
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib:
file: libgcc_eh.a(unwind-sjlj.o) has no symbols
ranlib libgcc_eh.a
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib:
file: libgcc_eh.a(unwind-sjlj.o) has no symbols
/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xgcc
-B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/
-B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
-B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
/sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
/sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include    -g -O2 -m32 -O2
-g -O2 -DIN_GCC    -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual
-Wno-format -Wstrict-prototypes -Wmissing-prototypes
-Wold-style-definition  -isystem ./include   -pipe -fno-common -g
-DIN_LIBGCC2 -fbuilding-libgcc -fno-stack-protector   -pipe
-fno-common -I. -I. -I../../.././gcc
-I../../../../gcc-5-20150330/libgcc
-I../../../../gcc-5-20150330/libgcc/.
-I../../../../gcc-5-20150330/libgcc/../gcc
-I../../../../gcc-5-20150330/libgcc/../include  -DHAVE_CC_TLS
-DUSE_EMUTLS -o cpuinfo_s.o -MT cpuinfo_s.o -MD -MP -MF cpuinfo_s.dep
-DSHARED  -c ../../../../gcc-5-20150330/libgcc/config/i386/cpuinfo.c
{standard input}:3:Unknown pseudo-op: .symver
{standard input}:3:Rest of line ignored. 1st junk character valued 95 (_).
{standard input}:4:Unknown pseudo-op: .symver
{standard input}:4:Rest of line ignored. 1st junk character valued 95 (_).
../../../../gcc-5-20150330/libgcc/shared-object.mk:18: recipe for
target 'cpuinfo_s.o' failed
make[5]: *** [cpuinfo_s.o] Error 1
make[5]: Leaving directory
'/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/i386/libgcc'
Makefile:1174: recipe for target 'multi-do' failed
make[4]: *** [multi-do] Error 1
make[4]: Leaving directory
'/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
Makefile:117: recipe for target 'all-multi' failed
make[3]: *** [all-multi] Error 2
make[3]: Leaving directory
'/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
Makefile:14820: recipe for target 'all-stage1-target-libgcc' failed
make[2]: *** [all-stage1-target-libgcc] Error 2
make[2]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
Makefile:20760: recipe for target 'stage1-bubble' failed
make[1]: *** [stage1-bubble] Error 2
make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
Makefile:21064: recipe for target 'bootstrap' failed
make: *** [bootstrap] Error 2

     Jack

On Mon, Mar 30, 2015 at 10:08 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Mon, Mar 30, 2015 at 5:53 PM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>> HJ,
>>      This patch breaks the bootstrap on targets like darwin which
>> don't build libgcc_nonshared.a...
>>
>> if test -z "$objects"; then \
>>   echo 'int __libgcc_eh_dummy;' > eh_dummy.c; \
>>   /sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xgcc
>> -B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/
>> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
>> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
>> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
>> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include    -g -O2 -m32 -O2
>> -g -O2 -DIN_GCC    -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual
>> -Wno-format -Wstrict-prototypes -Wmissing-prototypes
>> -Wold-style-definition  -isystem ./include   -pipe -fno-common -g
>> -DIN_LIBGCC2 -fbuilding-libgcc -fno-stack-protector   -pipe
>> -fno-common -I. -I. -I../../.././gcc
>> -I../../../../gcc-5-20150330/libgcc
>> -I../../../../gcc-5-20150330/libgcc/.
>> -I../../../../gcc-5-20150330/libgcc/../gcc
>> -I../../../../gcc-5-20150330/libgcc/../include  -DHAVE_CC_TLS
>> -DUSE_EMUTLS -fvisibility=hidden -DHIDE_EXPORTS -c eh_dummy.c \
>>      -o eh_dummy.o; \
>>   objects=eh_dummy.o; \
>> fi; \
>> ar  rc libgcc_nonshared.a $objects
>> ar: cpuinfo.o: No such file or directory
>> Makefile:905: recipe for target 'libgcc_nonshared.a' failed
>> make[5]: *** [libgcc_nonshared.a] Error 1
>> make[5]: Leaving directory
>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/i386/libgcc'
>> Makefile:1168: recipe for target 'multi-do' failed
>> make[4]: *** [multi-do] Error 1
>> make[4]: Leaving directory
>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
>> Makefile:117: recipe for target 'all-multi' failed
>> make[3]: *** [all-multi] Error 2
>> make[3]: Leaving directory
>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
>> Makefile:14820: recipe for target 'all-stage1-target-libgcc' failed
>> make[2]: *** [all-stage1-target-libgcc] Error 2
>> make[2]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
>> Makefile:20760: recipe for target 'stage1-bubble' failed
>> make[1]: *** [stage1-bubble] Error 2
>> make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
>> Makefile:21064: recipe for target 'bootstrap' failed
>> make: *** [bootstrap] Error 2
>>
>>           Jack
>>
>
> This one works.  I need to add $(LIB2ADDNONSHARED) $(LIB2ADDSHARED)
> to iter-items and handle duplicated items in them.
>
> --
> H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

[-- Attachment #1: Type: text/plain, Size: 2957 bytes --]

On Mon, Mar 30, 2015 at 8:09 PM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
> H.J.,
>    This still breaks the darwin bootstrap but differently.
>
> ar  rc libgcc_eh.a $objects
> /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib:
> file: libgcc_eh.a(unwind-sjlj.o) has no symbols
> ranlib libgcc_eh.a
> /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib:
> file: libgcc_eh.a(unwind-sjlj.o) has no symbols
> /sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xgcc
> -B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/
> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
> -B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
> /sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include    -g -O2 -m32 -O2
> -g -O2 -DIN_GCC    -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual
> -Wno-format -Wstrict-prototypes -Wmissing-prototypes
> -Wold-style-definition  -isystem ./include   -pipe -fno-common -g
> -DIN_LIBGCC2 -fbuilding-libgcc -fno-stack-protector   -pipe
> -fno-common -I. -I. -I../../.././gcc
> -I../../../../gcc-5-20150330/libgcc
> -I../../../../gcc-5-20150330/libgcc/.
> -I../../../../gcc-5-20150330/libgcc/../gcc
> -I../../../../gcc-5-20150330/libgcc/../include  -DHAVE_CC_TLS
> -DUSE_EMUTLS -o cpuinfo_s.o -MT cpuinfo_s.o -MD -MP -MF cpuinfo_s.dep
> -DSHARED  -c ../../../../gcc-5-20150330/libgcc/config/i386/cpuinfo.c
> {standard input}:3:Unknown pseudo-op: .symver
> {standard input}:3:Rest of line ignored. 1st junk character valued 95 (_).
> {standard input}:4:Unknown pseudo-op: .symver
> {standard input}:4:Rest of line ignored. 1st junk character valued 95 (_).
> ../../../../gcc-5-20150330/libgcc/shared-object.mk:18: recipe for
> target 'cpuinfo_s.o' failed
> make[5]: *** [cpuinfo_s.o] Error 1
> make[5]: Leaving directory
> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/i386/libgcc'
> Makefile:1174: recipe for target 'multi-do' failed
> make[4]: *** [multi-do] Error 1
> make[4]: Leaving directory
> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
> Makefile:117: recipe for target 'all-multi' failed
> make[3]: *** [all-multi] Error 2
> make[3]: Leaving directory
> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libgcc'
> Makefile:14820: recipe for target 'all-stage1-target-libgcc' failed
> make[2]: *** [all-stage1-target-libgcc] Error 2
> make[2]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
> Makefile:20760: recipe for target 'stage1-bubble' failed
> make[1]: *** [stage1-bubble] Error 2
> make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
> Makefile:21064: recipe for target 'bootstrap' failed
> make: *** [bootstrap] Error 2
>

Here is the updated patch.  I added

HOST_LIBGCC2_CFLAGS += -DUSE_ELF_SYMVER

when mkmap-symver.awk is used.


-- 
H.J.

[-- Attachment #2: 0001-Hide-__cpu_indicator_init-__cpu_model-from-linker.patch --]
[-- Type: text/x-patch, Size: 9237 bytes --]

From 00e6d397103e2acae8261e5e691704b4f4be3b2b Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sun, 29 Mar 2015 18:03:49 -0700
Subject: [PATCH] Hide __cpu_indicator_init/__cpu_model from linker

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc_nonshared.a while providing backward binary compatibility.  The
new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.

gcc/

	PR target/65612
	* gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
	to -lgcc_s/-lgcc/libgcc.a%s.

gcc/testsuite/

	PR target/65612
	* g++.dg/ext/mv18.C: New test.
	* g++.dg/ext/mv19.C: Likewise.
	* g++.dg/ext/mv20.C: Likewise.

libgcc/

	PR target/65612
	* Makefile.in (LIB2ADDSHARED): New.
	(LIB2ADDNONSHARED): Likewise.
	(libgcc-nonshared-objects): Likewise.
	(libgcc_nonshared.a): Likewise.
	Check unsupported files in LIB2ADDNONSHARED or LIB2ADDSHARED.
	(iter-items): Add $(LIB2ADDNONSHARED) $(LIB2ADDSHARED).
	(libgcc-s-objects): Add $(LIB2ADDSHARED).
	(all): Depend on libgcc_nonshared.a.
	($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
	(install-leaf): Install libgcc_nonshared.a.
	* shared-object.mk: Check empty $o.
	* config/i386/cpuinfo.c (__cpu_model): Initialize.
	(__cpu_indicator_init@GCC_4.8.0): New.
	(__cpu_model@GCC_4.8.0): Likewise.
	* config/t-slibgcc-elf-ver (HOST_LIBGCC2_CFLAGS ): Add
	-DUSE_ELF_SYMVER.
	* config/i386/t-cpuinfo (LIB2ADD): Renamed to ...
	(LIB2ADDSHARED): This.
	(LIB2ADDNONSHARED): New.
---
 gcc/gcc.c                       |  4 ++++
 gcc/testsuite/g++.dg/ext/mv18.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv19.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv20.C |  7 +++++++
 libgcc/Makefile.in              | 28 ++++++++++++++++++++++++----
 libgcc/config/i386/cpuinfo.c    |  7 ++++++-
 libgcc/config/i386/t-cpuinfo    |  3 ++-
 libgcc/config/t-slibgcc-elf-ver |  2 ++
 libgcc/shared-object.mk         |  2 ++
 9 files changed, 61 insertions(+), 6 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/mv18.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv19.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv20.C

diff --git a/gcc/gcc.c b/gcc/gcc.c
index d956c36..88a8dac 100644
--- a/gcc/gcc.c
+++ b/gcc/gcc.c
@@ -1566,11 +1566,13 @@ init_spec (void)
 	if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
 	  {
 	    init_gcc_specs (&obstack,
+			    "-lgcc_nonshared "
 			    "-lgcc_s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
 			    " -lunwind"
 #endif
 			    ,
+			    "-lgcc_nonshared "
 			    "-lgcc",
 			    "-lgcc_eh"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
@@ -1591,7 +1593,9 @@ init_spec (void)
 	    /* Ug.  We don't know shared library extensions.  Hope that
 	       systems that use this form don't do shared libraries.  */
 	    init_gcc_specs (&obstack,
+			    "libgcc_nonshared.a%s "
 			    "-lgcc_s",
+			    "libgcc_nonshared.a%s "
 			    "libgcc.a%s",
 			    "libgcc_eh.a%s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
diff --git a/gcc/testsuite/g++.dg/ext/mv18.C b/gcc/testsuite/g++.dg/ext/mv18.C
new file mode 100644
index 0000000..1f024de
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv18.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv19.C b/gcc/testsuite/g++.dg/ext/mv19.C
new file mode 100644
index 0000000..d1ea788
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv19.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv20.C b/gcc/testsuite/g++.dg/ext/mv20.C
new file mode 100644
index 0000000..98f7408
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv20.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/libgcc/Makefile.in b/libgcc/Makefile.in
index 88ddfea..f6eb047 100644
--- a/libgcc/Makefile.in
+++ b/libgcc/Makefile.in
@@ -261,6 +261,14 @@ LIB2ADD =
 # Assembler files should have names ending in `.S'.
 LIB2ADD_ST =
 
+# List of extra C and assembler files to add to shared libgcc2.
+# Assembler files should have names ending in `.S'.
+LIB2ADDSHARED =
+
+# List of extra C and assembler files to add to libgcc_nonshared.a.
+# Assembler files should have names ending in `.S'.
+LIB2ADDNONSHARED =
+
 # Specify the directories to be searched for header files.
 # Both . and srcdir are used, in that order,
 # so that *config.h will be found in the compilation
@@ -807,12 +815,19 @@ endif
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD))))
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD_ST))))
 
+# Build LIB2ADDNONSHARED and LIB2ADDSHARED.
+ifneq ($(filter-out %.c %.S %.asm,$(LIB2ADDNONSHARED) $(LIB2ADDSHARED)),)
+$(error Unsupported files in LIB2ADDNONSHARED or LIB2ADDSHARED.)
+endif
+
+libgcc-nonshared-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADDNONSHARED))))
+
 c_flags :=
-iter-items := $(LIB2ADD) $(LIB2ADD_ST)
+iter-items := $(LIB2ADD) $(LIB2ADD_ST) $(LIB2ADDNONSHARED) $(LIB2ADDSHARED)
 include $(iterator)
 
 ifeq ($(enable_shared),yes)
-libgcc-s-objects += $(addsuffix _s$(objext),$(basename $(notdir $(LIB2ADD))))
+libgcc-s-objects += $(addsuffix _s$(objext),$(basename $(notdir $(LIB2ADD) $(LIB2ADDSHARED))))
 endif
 
 # Build LIB2ADDEH, LIB2ADDEHSTATIC, and LIB2ADDEHSHARED.  If we don't have
@@ -890,8 +905,9 @@ libgcc.a: $(libgcc-objects)
 libgcov.a: $(libgcov-objects)
 libunwind.a: $(libunwind-objects)
 libgcc_eh.a: $(libgcc-eh-objects)
+libgcc_nonshared.a: $(libgcc-nonshared-objects)
 
-libgcc.a libgcov.a libunwind.a libgcc_eh.a:
+libgcc.a libgcov.a libunwind.a libgcc_eh.a libgcc_nonshared.a:
 	-rm -f $@
 
 	objects="$(objects)";					\
@@ -905,7 +921,7 @@ libgcc.a libgcov.a libunwind.a libgcc_eh.a:
 
 	$(RANLIB) $@
 
-all: libgcc.a libgcov.a
+all: libgcc.a libgcc_nonshared.a libgcov.a
 
 ifneq ($(LIBUNWIND),)
 all: libunwind.a
@@ -1051,6 +1067,7 @@ libgcc-extra-parts: $(EXTRA_PARTS)
 all: $(extra-parts)
 
 $(libgcc-objects) $(libgcc-s-objects) $(libgcc-eh-objects) \
+	$(libgcc-nonshared-objects) \
 	$(libgcov-objects) \
 	$(libunwind-objects) $(libunwind-s-objects) \
 	$(EXTRA_PARTS): libgcc_tm.h
@@ -1118,6 +1135,9 @@ install-leaf: $(install-shared) $(install-libunwind)
 	$(INSTALL_DATA) libgcc.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc.a
+	$(INSTALL_DATA) libgcc_nonshared.a $(DESTDIR)$(inst_libdir)/
+	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
+	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
 	$(INSTALL_DATA) libgcov.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcov.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcov.a
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..f6f91dd 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };
 
 
 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)
 
   return 0;
 }
+
+#if defined SHARED && defined USE_ELF_SYMVER
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
diff --git a/libgcc/config/i386/t-cpuinfo b/libgcc/config/i386/t-cpuinfo
index 31c8db9..8dbe9e6 100644
--- a/libgcc/config/i386/t-cpuinfo
+++ b/libgcc/config/i386/t-cpuinfo
@@ -1 +1,2 @@
-LIB2ADD += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDSHARED += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDNONSHARED += $(srcdir)/config/i386/cpuinfo.c
diff --git a/libgcc/config/t-slibgcc-elf-ver b/libgcc/config/t-slibgcc-elf-ver
index 4a00edc..a17659e 100644
--- a/libgcc/config/t-slibgcc-elf-ver
+++ b/libgcc/config/t-slibgcc-elf-ver
@@ -2,3 +2,5 @@
 
 SHLIB_MKMAP = $(srcdir)/mkmap-symver.awk
 SHLIB_MAPFILES = libgcc-std.ver
+
+HOST_LIBGCC2_CFLAGS += -DUSE_ELF_SYMVER
diff --git a/libgcc/shared-object.mk b/libgcc/shared-object.mk
index efac797..306176d 100644
--- a/libgcc/shared-object.mk
+++ b/libgcc/shared-object.mk
@@ -4,6 +4,7 @@
 o := $(firstword $(iter-items))
 iter-items := $(filter-out $o,$(iter-items))
 
+ifneq ($o,)
 base := $(basename $(notdir $o))
 
 c_flags-$o := $(c_flags)
@@ -36,3 +37,4 @@ $(base)_s$(objext): $o
 	$(gcc_s_compile) -c $(as_flags-$<) $<
 
 endif
+endif
-- 
2.1.0

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jakub Jelinek]

On Mon, Mar 30, 2015 at 07:08:00PM -0700, H.J. Lu wrote:
> --- a/gcc/gcc.c
> +++ b/gcc/gcc.c
> @@ -1566,11 +1566,13 @@ init_spec (void)
>  	if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
>  	  {
>  	    init_gcc_specs (&obstack,
> +			    "-lgcc_nonshared "
>  			    "-lgcc_s"
>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>  			    " -lunwind"
>  #endif
>  			    ,
> +			    "-lgcc_nonshared "
>  			    "-lgcc",
>  			    "-lgcc_eh"
>  #ifdef USE_LIBUNWIND_EXCEPTIONS
> @@ -1591,7 +1593,9 @@ init_spec (void)
>  	    /* Ug.  We don't know shared library extensions.  Hope that
>  	       systems that use this form don't do shared libraries.  */
>  	    init_gcc_specs (&obstack,
> +			    "libgcc_nonshared.a%s "
>  			    "-lgcc_s",
> +			    "libgcc_nonshared.a%s "
>  			    "libgcc.a%s",
>  			    "libgcc_eh.a%s"

Why do you need to link libgcc_nonshared.a twice here?  -lgcc_s surely won't
add any new __cpu* undefined references.

> @@ -424,3 +424,8 @@ __cpu_indicator_init (void)
>  
>    return 0;
>  }
> +
> +#if defined SHARED && !defined _WIN32
> +__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
> +__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
> +#endif

Will this work on Solaris?
I'd say you at least want to also guard with some configure check if
.symver is supported by assembler.

	Jakub

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Rainer Orth]

Jakub Jelinek <jakub@redhat.com> writes:

>> @@ -424,3 +424,8 @@ __cpu_indicator_init (void)
>>  
>>    return 0;
>>  }
>> +
>> +#if defined SHARED && !defined _WIN32
>> +__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
>> +__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
>> +#endif
>
> Will this work on Solaris?
> I'd say you at least want to also guard with some configure check if
> .symver is supported by assembler.

This will never work on Solaris: even if the assembler supports .symver,
ld.so.1 never will.

	Rainer

-- 
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

[-- Attachment #1: Type: text/plain, Size: 3496 bytes --]

On Mon, Mar 30, 2015 at 10:38 PM, Jakub Jelinek <jakub@redhat.com> wrote:
> On Mon, Mar 30, 2015 at 07:08:00PM -0700, H.J. Lu wrote:
>> --- a/gcc/gcc.c
>> +++ b/gcc/gcc.c
>> @@ -1566,11 +1566,13 @@ init_spec (void)
>>       if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
>>         {
>>           init_gcc_specs (&obstack,
>> +                         "-lgcc_nonshared "
>>                           "-lgcc_s"
>>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>>                           " -lunwind"
>>  #endif
>>                           ,
>> +                         "-lgcc_nonshared "
>>                           "-lgcc",
>>                           "-lgcc_eh"
>>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>> @@ -1591,7 +1593,9 @@ init_spec (void)
>>           /* Ug.  We don't know shared library extensions.  Hope that
>>              systems that use this form don't do shared libraries.  */
>>           init_gcc_specs (&obstack,
>> +                         "libgcc_nonshared.a%s "
>>                           "-lgcc_s",
>> +                         "libgcc_nonshared.a%s "
>>                           "libgcc.a%s",
>>                           "libgcc_eh.a%s"
>
> Why do you need to link libgcc_nonshared.a twice here?  -lgcc_s surely won't
> add any new __cpu* undefined references.

The one added for -lgcc_s is for building shared C++ library since -lgcc isn't
used and only -lgcc_s is used.  The one added for -lgcc is for static linking
since -lgcc_s isn't used.  I updated the patch to add some testcases
for -static.
I couldn't find a way to add tests for:

export/build/gnu/gcc/build-x86_64-linux/gcc/xg++
-B/export/build/gnu/gcc/build-x86_64-linux/gcc/ -O2    -c -o main.o
main.cc
/export/build/gnu/gcc/build-x86_64-linux/gcc/xg++
-B/export/build/gnu/gcc/build-x86_64-linux/gcc/ -shared -fPIC -O2  -o
libmv20.so mv20.cc
-L/export/build/gnu/gcc/build-x86_64-linux/x86_64-unknown-linux-gnu/libstdc++-v3/src/.libs
/export/build/gnu/gcc/build-x86_64-linux/gcc/xg++
-B/export/build/gnu/gcc/build-x86_64-linux/gcc/ -O2  -o x main.o
libmv20.so -Wl,-R,.
-L/export/build/gnu/gcc/build-x86_64-linux/x86_64-unknown-linux-gnu/libstdc++-v3/src/.libs
/usr/local/bin/ld: x: hidden symbol `__cpu_model' in
/export/build/gnu/gcc/build-x86_64-linux/gcc/libgcc_nonshared.a(cpuinfo.o)
is referenced by DSO
/usr/local/bin/ld: final link failed: Bad value
collect2: error: ld returned 1 exit status

>> @@ -424,3 +424,8 @@ __cpu_indicator_init (void)
>>
>>    return 0;
>>  }
>> +
>> +#if defined SHARED && !defined _WIN32
>> +__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
>> +__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
>> +#endif
>
> Will this work on Solaris?
> I'd say you at least want to also guard with some configure check if
> .symver is supported by assembler.
>

I updated my patch with:

@@ -424,3 +424,8 @@ __cpu_indicator_init (void)

   return 0;
 }
+
+#if defined SHARED && defined USE_ELF_SYMVER
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif

diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
index 4f47f7b..11bb46e 100644
--- a/libgcc/config/i386/t-linux
+++ b/libgcc/config/i386/t-linux
@@ -3,4 +3,4 @@
 # t-slibgcc-elf-ver and t-linux
 SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver

-HOST_LIBGCC2_CFLAGS += -mlong-double-80
+HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER

OK for master?

-- 
H.J.

[-- Attachment #2: 0001-Hide-__cpu_indicator_init-__cpu_model-from-linker.patch --]
[-- Type: text/x-patch, Size: 11008 bytes --]

From 259457afe0add56064ed49da1954d0770b1e1975 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sun, 29 Mar 2015 18:03:49 -0700
Subject: [PATCH] Hide __cpu_indicator_init/__cpu_model from linker

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc_nonshared.a while providing backward binary compatibility.  The
new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.

gcc/

	PR target/65612
	* gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
	to -lgcc_s/-lgcc/libgcc.a%s.

gcc/testsuite/

	PR target/65612
	* g++.dg/ext/mv18.C: New test.
	* g++.dg/ext/mv19.C: Likewise.
	* g++.dg/ext/mv20.C: Likewise.
	* g++.dg/ext/mv21.C: Likewise.
	* g++.dg/ext/mv22.C: Likewise.
	* g++.dg/ext/mv23.C: Likewise.

libgcc/

	PR target/65612
	* Makefile.in (LIB2ADDSHARED): New.
	(LIB2ADDNONSHARED): Likewise.
	(libgcc-nonshared-objects): Likewise.
	(libgcc_nonshared.a): Likewise.
	Check unsupported files in LIB2ADDNONSHARED or LIB2ADDSHARED.
	(iter-items): Add $(LIB2ADDNONSHARED) $(LIB2ADDSHARED).
	(libgcc-s-objects): Add $(LIB2ADDSHARED).
	(all): Depend on libgcc_nonshared.a.
	($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
	(install-leaf): Install libgcc_nonshared.a.
	* shared-object.mk: Check empty $o.
	* config/i386/cpuinfo.c (__cpu_model): Initialize.
	(__cpu_indicator_init@GCC_4.8.0): New.
	(__cpu_model@GCC_4.8.0): Likewise.
	* config/i386/t-cpuinfo (LIB2ADD): Renamed to ...
	(LIB2ADDSHARED): This.
	(LIB2ADDNONSHARED): New.
	* config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
	-DUSE_ELF_SYMVER.
---
 gcc/gcc.c                       |  4 ++++
 gcc/testsuite/g++.dg/ext/mv18.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv19.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv20.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv21.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv22.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv23.C |  7 +++++++
 libgcc/Makefile.in              | 28 ++++++++++++++++++++++++----
 libgcc/config/i386/cpuinfo.c    |  7 ++++++-
 libgcc/config/i386/t-cpuinfo    |  3 ++-
 libgcc/config/i386/t-linux      |  2 +-
 libgcc/shared-object.mk         |  2 ++
 12 files changed, 81 insertions(+), 7 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/mv18.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv19.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv20.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv21.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv22.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv23.C

diff --git a/gcc/gcc.c b/gcc/gcc.c
index d956c36..88a8dac 100644
--- a/gcc/gcc.c
+++ b/gcc/gcc.c
@@ -1566,11 +1566,13 @@ init_spec (void)
 	if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
 	  {
 	    init_gcc_specs (&obstack,
+			    "-lgcc_nonshared "
 			    "-lgcc_s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
 			    " -lunwind"
 #endif
 			    ,
+			    "-lgcc_nonshared "
 			    "-lgcc",
 			    "-lgcc_eh"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
@@ -1591,7 +1593,9 @@ init_spec (void)
 	    /* Ug.  We don't know shared library extensions.  Hope that
 	       systems that use this form don't do shared libraries.  */
 	    init_gcc_specs (&obstack,
+			    "libgcc_nonshared.a%s "
 			    "-lgcc_s",
+			    "libgcc_nonshared.a%s "
 			    "libgcc.a%s",
 			    "libgcc_eh.a%s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
diff --git a/gcc/testsuite/g++.dg/ext/mv18.C b/gcc/testsuite/g++.dg/ext/mv18.C
new file mode 100644
index 0000000..1f024de
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv18.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv19.C b/gcc/testsuite/g++.dg/ext/mv19.C
new file mode 100644
index 0000000..d1ea788
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv19.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv20.C b/gcc/testsuite/g++.dg/ext/mv20.C
new file mode 100644
index 0000000..98f7408
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv20.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv21.C b/gcc/testsuite/g++.dg/ext/mv21.C
new file mode 100644
index 0000000..9708ad9
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv21.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv22.C b/gcc/testsuite/g++.dg/ext/mv22.C
new file mode 100644
index 0000000..2550136
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv22.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv23.C b/gcc/testsuite/g++.dg/ext/mv23.C
new file mode 100644
index 0000000..f00afb0
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv23.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/libgcc/Makefile.in b/libgcc/Makefile.in
index 88ddfea..f6eb047 100644
--- a/libgcc/Makefile.in
+++ b/libgcc/Makefile.in
@@ -261,6 +261,14 @@ LIB2ADD =
 # Assembler files should have names ending in `.S'.
 LIB2ADD_ST =
 
+# List of extra C and assembler files to add to shared libgcc2.
+# Assembler files should have names ending in `.S'.
+LIB2ADDSHARED =
+
+# List of extra C and assembler files to add to libgcc_nonshared.a.
+# Assembler files should have names ending in `.S'.
+LIB2ADDNONSHARED =
+
 # Specify the directories to be searched for header files.
 # Both . and srcdir are used, in that order,
 # so that *config.h will be found in the compilation
@@ -807,12 +815,19 @@ endif
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD))))
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD_ST))))
 
+# Build LIB2ADDNONSHARED and LIB2ADDSHARED.
+ifneq ($(filter-out %.c %.S %.asm,$(LIB2ADDNONSHARED) $(LIB2ADDSHARED)),)
+$(error Unsupported files in LIB2ADDNONSHARED or LIB2ADDSHARED.)
+endif
+
+libgcc-nonshared-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADDNONSHARED))))
+
 c_flags :=
-iter-items := $(LIB2ADD) $(LIB2ADD_ST)
+iter-items := $(LIB2ADD) $(LIB2ADD_ST) $(LIB2ADDNONSHARED) $(LIB2ADDSHARED)
 include $(iterator)
 
 ifeq ($(enable_shared),yes)
-libgcc-s-objects += $(addsuffix _s$(objext),$(basename $(notdir $(LIB2ADD))))
+libgcc-s-objects += $(addsuffix _s$(objext),$(basename $(notdir $(LIB2ADD) $(LIB2ADDSHARED))))
 endif
 
 # Build LIB2ADDEH, LIB2ADDEHSTATIC, and LIB2ADDEHSHARED.  If we don't have
@@ -890,8 +905,9 @@ libgcc.a: $(libgcc-objects)
 libgcov.a: $(libgcov-objects)
 libunwind.a: $(libunwind-objects)
 libgcc_eh.a: $(libgcc-eh-objects)
+libgcc_nonshared.a: $(libgcc-nonshared-objects)
 
-libgcc.a libgcov.a libunwind.a libgcc_eh.a:
+libgcc.a libgcov.a libunwind.a libgcc_eh.a libgcc_nonshared.a:
 	-rm -f $@
 
 	objects="$(objects)";					\
@@ -905,7 +921,7 @@ libgcc.a libgcov.a libunwind.a libgcc_eh.a:
 
 	$(RANLIB) $@
 
-all: libgcc.a libgcov.a
+all: libgcc.a libgcc_nonshared.a libgcov.a
 
 ifneq ($(LIBUNWIND),)
 all: libunwind.a
@@ -1051,6 +1067,7 @@ libgcc-extra-parts: $(EXTRA_PARTS)
 all: $(extra-parts)
 
 $(libgcc-objects) $(libgcc-s-objects) $(libgcc-eh-objects) \
+	$(libgcc-nonshared-objects) \
 	$(libgcov-objects) \
 	$(libunwind-objects) $(libunwind-s-objects) \
 	$(EXTRA_PARTS): libgcc_tm.h
@@ -1118,6 +1135,9 @@ install-leaf: $(install-shared) $(install-libunwind)
 	$(INSTALL_DATA) libgcc.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc.a
+	$(INSTALL_DATA) libgcc_nonshared.a $(DESTDIR)$(inst_libdir)/
+	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
+	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
 	$(INSTALL_DATA) libgcov.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcov.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcov.a
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..f6f91dd 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };
 
 
 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)
 
   return 0;
 }
+
+#if defined SHARED && defined USE_ELF_SYMVER
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
diff --git a/libgcc/config/i386/t-cpuinfo b/libgcc/config/i386/t-cpuinfo
index 31c8db9..8dbe9e6 100644
--- a/libgcc/config/i386/t-cpuinfo
+++ b/libgcc/config/i386/t-cpuinfo
@@ -1 +1,2 @@
-LIB2ADD += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDSHARED += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDNONSHARED += $(srcdir)/config/i386/cpuinfo.c
diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
index 4f47f7b..11bb46e 100644
--- a/libgcc/config/i386/t-linux
+++ b/libgcc/config/i386/t-linux
@@ -3,4 +3,4 @@
 # t-slibgcc-elf-ver and t-linux
 SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver
 
-HOST_LIBGCC2_CFLAGS += -mlong-double-80
+HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER
diff --git a/libgcc/shared-object.mk b/libgcc/shared-object.mk
index efac797..306176d 100644
--- a/libgcc/shared-object.mk
+++ b/libgcc/shared-object.mk
@@ -4,6 +4,7 @@
 o := $(firstword $(iter-items))
 iter-items := $(filter-out $o,$(iter-items))
 
+ifneq ($o,)
 base := $(basename $(notdir $o))
 
 c_flags-$o := $(c_flags)
@@ -36,3 +37,4 @@ $(base)_s$(objext): $o
 	$(gcc_s_compile) -c $(as_flags-$<) $<
 
 endif
+endif
-- 
2.1.0

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jack Howarth]

H.J.,
     While the latest patch fails to bootstrap on x86_64-apple-darwin14...

make[2]: Entering directory
'/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libcilkrts'
/bin/sh ./libtool --tag=CXX   --mode=link
/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xg++
-B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/ -nostdinc++
-nostdinc++ -I/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/include/x86_64-apple-darwin14.3.0
-I/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/include
-I/sw/src/fink.build/gcc5-5.0.0-1/gcc-5-20150331/libstdc++-v3/libsupc++
-I/sw/src/fink.build/gcc5-5.0.0-1/gcc-5-20150331/libstdc++-v3/include/backward
-I/sw/src/fink.build/gcc5-5.0.0-1/gcc-5-20150331/libstdc++-v3/testsuite/util
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/src
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/src/.libs
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/libsupc++/.libs
-B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/src/.libs
-B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/libsupc++/.libs
-B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
-B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
/sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
/sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include     -g -O2
-version-info 5:0:0 -ldl -lpthread   -no-undefined  -o libcilkrts.la
-rpath /sw/lib/gcc5/lib cilk-abi-vla.lo os-unix-sysdep.lo bug.lo
cilk-abi.lo cilk-abi-cilk-for.lo cilk-abi-vla-internal.lo cilk_api.lo
cilk_fiber.lo cilk_fiber-unix.lo cilk_malloc.lo c_reducers.lo
except-gcc.lo frame_malloc.lo full_frame.lo global_state.lo jmpbuf.lo
local_state.lo metacall_impl.lo os_mutex-unix.lo os-unix.lo
pedigrees.lo record-replay.lo reducer_impl.lo scheduler.lo
signal_node.lo spin_mutex.lo stats.lo symbol_test.lo sysdep-unix.lo
worker_mutex.lo
libtool: link:
/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/xg++
-B/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/./gcc/ -nostdinc++
-nostdinc++ -I/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/include/x86_64-apple-darwin14.3.0
-I/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/include
-I/sw/src/fink.build/gcc5-5.0.0-1/gcc-5-20150331/libstdc++-v3/libsupc++
-I/sw/src/fink.build/gcc5-5.0.0-1/gcc-5-20150331/libstdc++-v3/include/backward
-I/sw/src/fink.build/gcc5-5.0.0-1/gcc-5-20150331/libstdc++-v3/testsuite/util
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/src
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/src/.libs
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/libsupc++/.libs
-B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/bin/
-B/sw/lib/gcc5/x86_64-apple-darwin14.3.0/lib/ -isystem
/sw/lib/gcc5/x86_64-apple-darwin14.3.0/include -isystem
/sw/lib/gcc5/x86_64-apple-darwin14.3.0/sys-include    -dynamiclib  -o
.libs/libcilkrts.5.dylib  .libs/cilk-abi-vla.o .libs/os-unix-sysdep.o
.libs/bug.o .libs/cilk-abi.o .libs/cilk-abi-cilk-for.o
.libs/cilk-abi-vla-internal.o .libs/cilk_api.o .libs/cilk_fiber.o
.libs/cilk_fiber-unix.o .libs/cilk_malloc.o .libs/c_reducers.o
.libs/except-gcc.o .libs/frame_malloc.o .libs/full_frame.o
.libs/global_state.o .libs/jmpbuf.o .libs/local_state.o
.libs/metacall_impl.o .libs/os_mutex-unix.o .libs/os-unix.o
.libs/pedigrees.o .libs/record-replay.o .libs/reducer_impl.o
.libs/scheduler.o .libs/signal_node.o .libs/spin_mutex.o .libs/stats.o
.libs/symbol_test.o .libs/sysdep-unix.o .libs/worker_mutex.o
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/src
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/src/.libs
-L/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libstdc++-v3/libsupc++/.libs
-ldl -lpthread    -install_name  /sw/lib/gcc5/lib/libcilkrts.5.dylib
-compatibility_version 6 -current_version 6.0 -Wl,-single_module
Undefined symbols for architecture x86_64:
  "___cpu_model", referenced from:
      _restore_x86_fp_state in os-unix-sysdep.o
      _sysdep_save_fp_ctrl_state in os-unix-sysdep.o
ld: symbol(s) not found for architecture x86_64
collect2: error: ld returned 1 exit status
Makefile:540: recipe for target 'libcilkrts.la' failed
make[2]: *** [libcilkrts.la] Error 1
make[2]: Leaving directory
'/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libcilkrts'
Makefile:13569: recipe for target 'all-target-libcilkrts' failed
make[1]: *** [all-target-libcilkrts] Error 2
make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
Makefile:21064: recipe for target 'bootstrap' failed
make: *** [bootstrap] Error 2

as darwin will require the new usage of libgcc_nonshared.a to be added
to the spec handling with...

Index: gcc/config/darwin.h
===================================================================
--- gcc/config/darwin.h (revision 221794)
+++ gcc/config/darwin.h (working copy)
@@ -325,7 +325,7 @@ extern GTY(()) int darwin_ms_struct;
    need symbols from -lgcc.  */
 #undef REAL_LIBGCC_SPEC
 #define REAL_LIBGCC_SPEC   \
-   "%{static-libgcc|static: -lgcc_eh -lgcc;   \
+   "%{static-libgcc|static: -lgcc_eh -lgcc_nonshared -lgcc;   \
       shared-libgcc|fexceptions|fgnu-runtime:   \
        %:version-compare(!> 10.5 mmacosx-version-min= -lgcc_s.10.4)   \
        %:version-compare(>< 10.5 10.6 mmacosx-version-min= -lgcc_s.10.5)   \
@@ -336,7 +336,7 @@ extern GTY(()) int darwin_ms_struct;
        %:version-compare(>< 10.5 10.6 mmacosx-version-min= -lgcc_s.10.5)   \
        %:version-compare(!> 10.5 mmacosx-version-min= -lgcc_ext.10.4)   \
        %:version-compare(>= 10.5 mmacosx-version-min= -lgcc_ext.10.5)   \
-       -lgcc }"
+       -lgcc_nonshared -lgcc }"

 /* We specify crt0.o as -lcrt0.o so that ld will search the library path.

             Jack
ps One minor nit...

Index: gcc/gcc.c
===================================================================
--- gcc/gcc.c (revision 221794)
+++ gcc/gcc.c (working copy)
@@ -1566,11 +1566,13 @@ init_spec (void)
  if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
   {
     init_gcc_specs (&obstack,
+    "-lgcc_nonshared "
     "-lgcc_s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
     " -lunwind"
 #endif
     ,
+    "-lgcc_nonshared "
     "-lgcc",
     "-lgcc_eh"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
@@ -1591,7 +1593,9 @@ init_spec (void)
     /* Ug.  We don't know shared library extensions.  Hope that
        systems that use this form don't do shared libraries.  */
     init_gcc_specs (&obstack,
+    "libgcc_nonshared.a%s "
     "-lgcc_s",
+    "libgcc_nonshared.a%s "
     "libgcc.a%s",
     "libgcc_eh.a%s"

You seem to have unnecessary trailing whitespace at the end of these flags.

On Tue, Mar 31, 2015 at 6:38 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Mon, Mar 30, 2015 at 10:38 PM, Jakub Jelinek <jakub@redhat.com> wrote:
>> On Mon, Mar 30, 2015 at 07:08:00PM -0700, H.J. Lu wrote:
>>> --- a/gcc/gcc.c
>>> +++ b/gcc/gcc.c
>>> @@ -1566,11 +1566,13 @@ init_spec (void)
>>>       if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
>>>         {
>>>           init_gcc_specs (&obstack,
>>> +                         "-lgcc_nonshared "
>>>                           "-lgcc_s"
>>>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>>>                           " -lunwind"
>>>  #endif
>>>                           ,
>>> +                         "-lgcc_nonshared "
>>>                           "-lgcc",
>>>                           "-lgcc_eh"
>>>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>>> @@ -1591,7 +1593,9 @@ init_spec (void)
>>>           /* Ug.  We don't know shared library extensions.  Hope that
>>>              systems that use this form don't do shared libraries.  */
>>>           init_gcc_specs (&obstack,
>>> +                         "libgcc_nonshared.a%s "
>>>                           "-lgcc_s",
>>> +                         "libgcc_nonshared.a%s "
>>>                           "libgcc.a%s",
>>>                           "libgcc_eh.a%s"
>>
>> Why do you need to link libgcc_nonshared.a twice here?  -lgcc_s surely won't
>> add any new __cpu* undefined references.
>
> The one added for -lgcc_s is for building shared C++ library since -lgcc isn't
> used and only -lgcc_s is used.  The one added for -lgcc is for static linking
> since -lgcc_s isn't used.  I updated the patch to add some testcases
> for -static.
> I couldn't find a way to add tests for:
>
> export/build/gnu/gcc/build-x86_64-linux/gcc/xg++
> -B/export/build/gnu/gcc/build-x86_64-linux/gcc/ -O2    -c -o main.o
> main.cc
> /export/build/gnu/gcc/build-x86_64-linux/gcc/xg++
> -B/export/build/gnu/gcc/build-x86_64-linux/gcc/ -shared -fPIC -O2  -o
> libmv20.so mv20.cc
> -L/export/build/gnu/gcc/build-x86_64-linux/x86_64-unknown-linux-gnu/libstdc++-v3/src/.libs
> /export/build/gnu/gcc/build-x86_64-linux/gcc/xg++
> -B/export/build/gnu/gcc/build-x86_64-linux/gcc/ -O2  -o x main.o
> libmv20.so -Wl,-R,.
> -L/export/build/gnu/gcc/build-x86_64-linux/x86_64-unknown-linux-gnu/libstdc++-v3/src/.libs
> /usr/local/bin/ld: x: hidden symbol `__cpu_model' in
> /export/build/gnu/gcc/build-x86_64-linux/gcc/libgcc_nonshared.a(cpuinfo.o)
> is referenced by DSO
> /usr/local/bin/ld: final link failed: Bad value
> collect2: error: ld returned 1 exit status
>
>>> @@ -424,3 +424,8 @@ __cpu_indicator_init (void)
>>>
>>>    return 0;
>>>  }
>>> +
>>> +#if defined SHARED && !defined _WIN32
>>> +__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
>>> +__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
>>> +#endif
>>
>> Will this work on Solaris?
>> I'd say you at least want to also guard with some configure check if
>> .symver is supported by assembler.
>>
>
> I updated my patch with:
>
> @@ -424,3 +424,8 @@ __cpu_indicator_init (void)
>
>    return 0;
>  }
> +
> +#if defined SHARED && defined USE_ELF_SYMVER
> +__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
> +__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
> +#endif
>
> diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
> index 4f47f7b..11bb46e 100644
> --- a/libgcc/config/i386/t-linux
> +++ b/libgcc/config/i386/t-linux
> @@ -3,4 +3,4 @@
>  # t-slibgcc-elf-ver and t-linux
>  SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver
>
> -HOST_LIBGCC2_CFLAGS += -mlong-double-80
> +HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER
>
> OK for master?
>
> --
> H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

[-- Attachment #1: Type: text/plain, Size: 4659 bytes --]

On Tue, Mar 31, 2015 at 7:25 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
> H.J.,
>      While the latest patch fails to bootstrap on x86_64-apple-darwin14...
>
>       _restore_x86_fp_state in os-unix-sysdep.o
>       _sysdep_save_fp_ctrl_state in os-unix-sysdep.o
> ld: symbol(s) not found for architecture x86_64
> collect2: error: ld returned 1 exit status
> Makefile:540: recipe for target 'libcilkrts.la' failed
> make[2]: *** [libcilkrts.la] Error 1
> make[2]: Leaving directory
> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libcilkrts'
> Makefile:13569: recipe for target 'all-target-libcilkrts' failed
> make[1]: *** [all-target-libcilkrts] Error 2
> make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
> Makefile:21064: recipe for target 'bootstrap' failed
> make: *** [bootstrap] Error 2
>
> as darwin will require the new usage of libgcc_nonshared.a to be added
> to the spec handling with...

Here is the updated patch to make libgcc_nonshared.a optional
so that it is only needed on Linux.

> Index: gcc/config/darwin.h
> ===================================================================
> --- gcc/config/darwin.h (revision 221794)
> +++ gcc/config/darwin.h (working copy)
> @@ -325,7 +325,7 @@ extern GTY(()) int darwin_ms_struct;
>     need symbols from -lgcc.  */
>  #undef REAL_LIBGCC_SPEC
>  #define REAL_LIBGCC_SPEC   \
> -   "%{static-libgcc|static: -lgcc_eh -lgcc;   \
> +   "%{static-libgcc|static: -lgcc_eh -lgcc_nonshared -lgcc;   \
>        shared-libgcc|fexceptions|fgnu-runtime:   \
>         %:version-compare(!> 10.5 mmacosx-version-min= -lgcc_s.10.4)   \
>         %:version-compare(>< 10.5 10.6 mmacosx-version-min= -lgcc_s.10.5)   \
> @@ -336,7 +336,7 @@ extern GTY(()) int darwin_ms_struct;
>         %:version-compare(>< 10.5 10.6 mmacosx-version-min= -lgcc_s.10.5)   \
>         %:version-compare(!> 10.5 mmacosx-version-min= -lgcc_ext.10.4)   \
>         %:version-compare(>= 10.5 mmacosx-version-min= -lgcc_ext.10.5)   \
> -       -lgcc }"
> +       -lgcc_nonshared -lgcc }"
>
>  /* We specify crt0.o as -lcrt0.o so that ld will search the library path.
>
>              Jack
> ps One minor nit...
>
> Index: gcc/gcc.c
> ===================================================================
> --- gcc/gcc.c (revision 221794)
> +++ gcc/gcc.c (working copy)
> @@ -1566,11 +1566,13 @@ init_spec (void)
>   if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
>    {
>      init_gcc_specs (&obstack,
> +    "-lgcc_nonshared "
>      "-lgcc_s"
>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>      " -lunwind"
>  #endif
>      ,
> +    "-lgcc_nonshared "
>      "-lgcc",
>      "-lgcc_eh"
>  #ifdef USE_LIBUNWIND_EXCEPTIONS
> @@ -1591,7 +1593,9 @@ init_spec (void)
>      /* Ug.  We don't know shared library extensions.  Hope that
>         systems that use this form don't do shared libraries.  */
>      init_gcc_specs (&obstack,
> +    "libgcc_nonshared.a%s "
>      "-lgcc_s",
> +    "libgcc_nonshared.a%s "
>      "libgcc.a%s",
>      "libgcc_eh.a%s"
>
> You seem to have unnecessary trailing whitespace at the end of these flags.
>

The white space is needed to avoid -lgcc_nonshared-lgcc_s.


-- 
H.J.
---

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc_nonshared.a while providing backward binary compatibility.  The
new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.

gcc/

PR target/65612
* gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
to -lgcc_s.

gcc/testsuite/

PR target/65612
* g++.dg/ext/mv18.C: New test.
* g++.dg/ext/mv19.C: Likewise.
* g++.dg/ext/mv20.C: Likewise.
* g++.dg/ext/mv21.C: Likewise.
* g++.dg/ext/mv22.C: Likewise.
* g++.dg/ext/mv23.C: Likewise.

libgcc/

PR target/65612
* Makefile.in (LIB2ADDNONSHARED): New.
(libgcc-nonshared-objects): Likewise.
(libgcc_nonshared.a): Likewise.
Check unsupported files in LIB2ADDNONSHARED.
(iter-items): Add $(LIB2ADDNONSHARED).
(all): Depend on libgcc_nonshared.a.
($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
(install-leaf): Install libgcc_nonshared.a.
* shared-object.mk: Check empty $o.
* config/i386/cpuinfo.c (__cpu_model): Initialize.
(__cpu_indicator_init@GCC_4.8.0): New.
(__cpu_model@GCC_4.8.0): Likewise.
* config/i386/t-cpuinfo (LIB2ADDNONSHARED): New.
* config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
-DUSE_ELF_SYMVER.
---

[-- Attachment #2: 0001-Hide-__cpu_indicator_init-__cpu_model-from-linker.patch --]
[-- Type: text/x-patch, Size: 10127 bytes --]

From 760b79124482860b4317f4d39fbe898cfbe8e47b Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sun, 29 Mar 2015 18:03:49 -0700
Subject: [PATCH] Hide __cpu_indicator_init/__cpu_model from linker

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc_nonshared.a while providing backward binary compatibility.  The
new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.

gcc/

	PR target/65612
	* gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
	to -lgcc_s.

gcc/testsuite/

	PR target/65612
	* g++.dg/ext/mv18.C: New test.
	* g++.dg/ext/mv19.C: Likewise.
	* g++.dg/ext/mv20.C: Likewise.
	* g++.dg/ext/mv21.C: Likewise.
	* g++.dg/ext/mv22.C: Likewise.
	* g++.dg/ext/mv23.C: Likewise.

libgcc/

	PR target/65612
	* Makefile.in (LIB2ADDNONSHARED): New.
	(libgcc-nonshared-objects): Likewise.
	(libgcc_nonshared.a): Likewise.
	Check unsupported files in LIB2ADDNONSHARED.
	(iter-items): Add $(LIB2ADDNONSHARED).
	(all): Depend on libgcc_nonshared.a.
	($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
	(install-leaf): Install libgcc_nonshared.a.
	* shared-object.mk: Check empty $o.
	* config/i386/cpuinfo.c (__cpu_model): Initialize.
	(__cpu_indicator_init@GCC_4.8.0): New.
	(__cpu_model@GCC_4.8.0): Likewise.
	* config/i386/t-cpuinfo (LIB2ADDNONSHARED): New.
	* config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
	-DUSE_ELF_SYMVER.
---
 gcc/gcc.c                       |  2 ++
 gcc/testsuite/g++.dg/ext/mv18.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv19.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv20.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv21.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv22.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv23.C |  7 +++++++
 libgcc/Makefile.in              | 22 +++++++++++++++++++---
 libgcc/config/i386/cpuinfo.c    |  7 ++++++-
 libgcc/config/i386/t-cpuinfo    |  1 +
 libgcc/config/i386/t-linux      |  2 +-
 libgcc/shared-object.mk         |  2 ++
 12 files changed, 73 insertions(+), 5 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/mv18.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv19.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv20.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv21.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv22.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv23.C

diff --git a/gcc/gcc.c b/gcc/gcc.c
index d956c36..3fbd549 100644
--- a/gcc/gcc.c
+++ b/gcc/gcc.c
@@ -1566,6 +1566,7 @@ init_spec (void)
 	if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
 	  {
 	    init_gcc_specs (&obstack,
+			    "-lgcc_nonshared "
 			    "-lgcc_s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
 			    " -lunwind"
@@ -1591,6 +1592,7 @@ init_spec (void)
 	    /* Ug.  We don't know shared library extensions.  Hope that
 	       systems that use this form don't do shared libraries.  */
 	    init_gcc_specs (&obstack,
+			    "libgcc_nonshared.a%s "
 			    "-lgcc_s",
 			    "libgcc.a%s",
 			    "libgcc_eh.a%s"
diff --git a/gcc/testsuite/g++.dg/ext/mv18.C b/gcc/testsuite/g++.dg/ext/mv18.C
new file mode 100644
index 0000000..1f024de
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv18.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv19.C b/gcc/testsuite/g++.dg/ext/mv19.C
new file mode 100644
index 0000000..d1ea788
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv19.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv20.C b/gcc/testsuite/g++.dg/ext/mv20.C
new file mode 100644
index 0000000..98f7408
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv20.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv21.C b/gcc/testsuite/g++.dg/ext/mv21.C
new file mode 100644
index 0000000..9708ad9
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv21.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv22.C b/gcc/testsuite/g++.dg/ext/mv22.C
new file mode 100644
index 0000000..2550136
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv22.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv23.C b/gcc/testsuite/g++.dg/ext/mv23.C
new file mode 100644
index 0000000..f00afb0
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv23.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/libgcc/Makefile.in b/libgcc/Makefile.in
index 88ddfea..f78b1ac 100644
--- a/libgcc/Makefile.in
+++ b/libgcc/Makefile.in
@@ -261,6 +261,10 @@ LIB2ADD =
 # Assembler files should have names ending in `.S'.
 LIB2ADD_ST =
 
+# List of extra C and assembler files to add to libgcc_nonshared.a.
+# Assembler files should have names ending in `.S'.
+LIB2ADDNONSHARED =
+
 # Specify the directories to be searched for header files.
 # Both . and srcdir are used, in that order,
 # so that *config.h will be found in the compilation
@@ -807,8 +811,15 @@ endif
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD))))
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD_ST))))
 
+# Build LIB2ADDNONSHARED.
+ifneq ($(filter-out %.c %.S %.asm,$(LIB2ADDNONSHARED)),)
+$(error Unsupported files in LIB2ADDNONSHARED.)
+endif
+
+libgcc-nonshared-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADDNONSHARED))))
+
 c_flags :=
-iter-items := $(LIB2ADD) $(LIB2ADD_ST)
+iter-items := $(LIB2ADD) $(LIB2ADD_ST) $(LIB2ADDNONSHARED)
 include $(iterator)
 
 ifeq ($(enable_shared),yes)
@@ -890,8 +901,9 @@ libgcc.a: $(libgcc-objects)
 libgcov.a: $(libgcov-objects)
 libunwind.a: $(libunwind-objects)
 libgcc_eh.a: $(libgcc-eh-objects)
+libgcc_nonshared.a: $(libgcc-nonshared-objects)
 
-libgcc.a libgcov.a libunwind.a libgcc_eh.a:
+libgcc.a libgcov.a libunwind.a libgcc_eh.a libgcc_nonshared.a:
 	-rm -f $@
 
 	objects="$(objects)";					\
@@ -905,7 +917,7 @@ libgcc.a libgcov.a libunwind.a libgcc_eh.a:
 
 	$(RANLIB) $@
 
-all: libgcc.a libgcov.a
+all: libgcc.a libgcc_nonshared.a libgcov.a
 
 ifneq ($(LIBUNWIND),)
 all: libunwind.a
@@ -1051,6 +1063,7 @@ libgcc-extra-parts: $(EXTRA_PARTS)
 all: $(extra-parts)
 
 $(libgcc-objects) $(libgcc-s-objects) $(libgcc-eh-objects) \
+	$(libgcc-nonshared-objects) \
 	$(libgcov-objects) \
 	$(libunwind-objects) $(libunwind-s-objects) \
 	$(EXTRA_PARTS): libgcc_tm.h
@@ -1118,6 +1131,9 @@ install-leaf: $(install-shared) $(install-libunwind)
 	$(INSTALL_DATA) libgcc.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc.a
+	$(INSTALL_DATA) libgcc_nonshared.a $(DESTDIR)$(inst_libdir)/
+	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
+	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
 	$(INSTALL_DATA) libgcov.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcov.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcov.a
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..f6f91dd 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };
 
 
 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)
 
   return 0;
 }
+
+#if defined SHARED && defined USE_ELF_SYMVER
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
diff --git a/libgcc/config/i386/t-cpuinfo b/libgcc/config/i386/t-cpuinfo
index 31c8db9..b20ef30 100644
--- a/libgcc/config/i386/t-cpuinfo
+++ b/libgcc/config/i386/t-cpuinfo
@@ -1 +1,2 @@
 LIB2ADD += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDNONSHARED += $(srcdir)/config/i386/cpuinfo.c
diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
index 4f47f7b..11bb46e 100644
--- a/libgcc/config/i386/t-linux
+++ b/libgcc/config/i386/t-linux
@@ -3,4 +3,4 @@
 # t-slibgcc-elf-ver and t-linux
 SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver
 
-HOST_LIBGCC2_CFLAGS += -mlong-double-80
+HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER
diff --git a/libgcc/shared-object.mk b/libgcc/shared-object.mk
index efac797..306176d 100644
--- a/libgcc/shared-object.mk
+++ b/libgcc/shared-object.mk
@@ -4,6 +4,7 @@
 o := $(firstword $(iter-items))
 iter-items := $(filter-out $o,$(iter-items))
 
+ifneq ($o,)
 base := $(basename $(notdir $o))
 
 c_flags-$o := $(c_flags)
@@ -36,3 +37,4 @@ $(base)_s$(objext): $o
 	$(gcc_s_compile) -c $(as_flags-$<) $<
 
 endif
+endif
-- 
1.9.3

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jack Howarth]

H.J.,
    Did you attach the correct version of the patch? I don't see
anything conditional on linux.
            Jack

On Tue, Mar 31, 2015 at 11:58 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Tue, Mar 31, 2015 at 7:25 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>> H.J.,
>>      While the latest patch fails to bootstrap on x86_64-apple-darwin14...
>>
>>       _restore_x86_fp_state in os-unix-sysdep.o
>>       _sysdep_save_fp_ctrl_state in os-unix-sysdep.o
>> ld: symbol(s) not found for architecture x86_64
>> collect2: error: ld returned 1 exit status
>> Makefile:540: recipe for target 'libcilkrts.la' failed
>> make[2]: *** [libcilkrts.la] Error 1
>> make[2]: Leaving directory
>> '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir/x86_64-apple-darwin14.3.0/libcilkrts'
>> Makefile:13569: recipe for target 'all-target-libcilkrts' failed
>> make[1]: *** [all-target-libcilkrts] Error 2
>> make[1]: Leaving directory '/sw/src/fink.build/gcc5-5.0.0-1/darwin_objdir'
>> Makefile:21064: recipe for target 'bootstrap' failed
>> make: *** [bootstrap] Error 2
>>
>> as darwin will require the new usage of libgcc_nonshared.a to be added
>> to the spec handling with...
>
> Here is the updated patch to make libgcc_nonshared.a optional
> so that it is only needed on Linux.
>
>> Index: gcc/config/darwin.h
>> ===================================================================
>> --- gcc/config/darwin.h (revision 221794)
>> +++ gcc/config/darwin.h (working copy)
>> @@ -325,7 +325,7 @@ extern GTY(()) int darwin_ms_struct;
>>     need symbols from -lgcc.  */
>>  #undef REAL_LIBGCC_SPEC
>>  #define REAL_LIBGCC_SPEC   \
>> -   "%{static-libgcc|static: -lgcc_eh -lgcc;   \
>> +   "%{static-libgcc|static: -lgcc_eh -lgcc_nonshared -lgcc;   \
>>        shared-libgcc|fexceptions|fgnu-runtime:   \
>>         %:version-compare(!> 10.5 mmacosx-version-min= -lgcc_s.10.4)   \
>>         %:version-compare(>< 10.5 10.6 mmacosx-version-min= -lgcc_s.10.5)   \
>> @@ -336,7 +336,7 @@ extern GTY(()) int darwin_ms_struct;
>>         %:version-compare(>< 10.5 10.6 mmacosx-version-min= -lgcc_s.10.5)   \
>>         %:version-compare(!> 10.5 mmacosx-version-min= -lgcc_ext.10.4)   \
>>         %:version-compare(>= 10.5 mmacosx-version-min= -lgcc_ext.10.5)   \
>> -       -lgcc }"
>> +       -lgcc_nonshared -lgcc }"
>>
>>  /* We specify crt0.o as -lcrt0.o so that ld will search the library path.
>>
>>              Jack
>> ps One minor nit...
>>
>> Index: gcc/gcc.c
>> ===================================================================
>> --- gcc/gcc.c (revision 221794)
>> +++ gcc/gcc.c (working copy)
>> @@ -1566,11 +1566,13 @@ init_spec (void)
>>   if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
>>    {
>>      init_gcc_specs (&obstack,
>> +    "-lgcc_nonshared "
>>      "-lgcc_s"
>>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>>      " -lunwind"
>>  #endif
>>      ,
>> +    "-lgcc_nonshared "
>>      "-lgcc",
>>      "-lgcc_eh"
>>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>> @@ -1591,7 +1593,9 @@ init_spec (void)
>>      /* Ug.  We don't know shared library extensions.  Hope that
>>         systems that use this form don't do shared libraries.  */
>>      init_gcc_specs (&obstack,
>> +    "libgcc_nonshared.a%s "
>>      "-lgcc_s",
>> +    "libgcc_nonshared.a%s "
>>      "libgcc.a%s",
>>      "libgcc_eh.a%s"
>>
>> You seem to have unnecessary trailing whitespace at the end of these flags.
>>
>
> The white space is needed to avoid -lgcc_nonshared-lgcc_s.
>
>
> --
> H.J.
> ---
>
> We shouldn't call external function, __cpu_indicator_init, while an object
> is being relocated since its .got.plt section hasn't been updated.  It
> works for non-PIE since no update on .got.plt section is required.  This
> patch hides __cpu_indicator_init/__cpu_model from linker to force linker
> to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
> in libgcc_nonshared.a while providing backward binary compatibility.  The
> new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.
>
> gcc/
>
> PR target/65612
> * gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
> to -lgcc_s.
>
> gcc/testsuite/
>
> PR target/65612
> * g++.dg/ext/mv18.C: New test.
> * g++.dg/ext/mv19.C: Likewise.
> * g++.dg/ext/mv20.C: Likewise.
> * g++.dg/ext/mv21.C: Likewise.
> * g++.dg/ext/mv22.C: Likewise.
> * g++.dg/ext/mv23.C: Likewise.
>
> libgcc/
>
> PR target/65612
> * Makefile.in (LIB2ADDNONSHARED): New.
> (libgcc-nonshared-objects): Likewise.
> (libgcc_nonshared.a): Likewise.
> Check unsupported files in LIB2ADDNONSHARED.
> (iter-items): Add $(LIB2ADDNONSHARED).
> (all): Depend on libgcc_nonshared.a.
> ($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
> (install-leaf): Install libgcc_nonshared.a.
> * shared-object.mk: Check empty $o.
> * config/i386/cpuinfo.c (__cpu_model): Initialize.
> (__cpu_indicator_init@GCC_4.8.0): New.
> (__cpu_model@GCC_4.8.0): Likewise.
> * config/i386/t-cpuinfo (LIB2ADDNONSHARED): New.
> * config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
> -DUSE_ELF_SYMVER.
> ---

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Tue, Mar 31, 2015 at 9:09 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
> H.J.,
>     Did you attach the correct version of the patch? I don't see
> anything conditional on linux.
>             Jack

My patch will build and install libgcc_nonshared.a for all targets.  If you
don't link against it, nothing is changed.  On Linux, it is used via the
init_spec change.

-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jack Howarth]

On Tue, Mar 31, 2015 at 12:14 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Tue, Mar 31, 2015 at 9:09 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>> H.J.,
>>     Did you attach the correct version of the patch? I don't see
>> anything conditional on linux.
>>             Jack
>
> My patch will build and install libgcc_nonshared.a for all targets.  If you
> don't link against it, nothing is changed.  On Linux, it is used via the
> init_spec change.

Isn't...

diff --git a/gcc/gcc.c b/gcc/gcc.c
index d956c36..3fbd549 100644
--- a/gcc/gcc.c
+++ b/gcc/gcc.c
@@ -1566,6 +1566,7 @@ init_spec (void)
        if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
          {
            init_gcc_specs (&obstack,
+                           "-lgcc_nonshared "
                            "-lgcc_s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
                            " -lunwind"
@@ -1591,6 +1592,7 @@ init_spec (void)
            /* Ug.  We don't know shared library extensions.  Hope that
               systems that use this form don't do shared libraries.  */
            init_gcc_specs (&obstack,
+                           "libgcc_nonshared.a%s "
                            "-lgcc_s",
                            "libgcc.a%s",
                            "libgcc_eh.a%s"

problematic for Solaris? I am unfamiliar with the Solaris spec
handling but sol2.h doesn't seem to have any instances of -lgcc which
might imply they use the stock compiler invocation which will now have
a non-existent libgcc_nonshared static library.
     Also, are you leaving the cpu symbols in libgcc.a on non-linux
targets? If not, the linkage failure reported in
https://gcc.gnu.org/ml/gcc-patches/2015-03/msg01668.html will occur,
no?
        Jack
>
> --
> H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Tue, Mar 31, 2015 at 9:39 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
> On Tue, Mar 31, 2015 at 12:14 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Tue, Mar 31, 2015 at 9:09 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>>> H.J.,
>>>     Did you attach the correct version of the patch? I don't see
>>> anything conditional on linux.
>>>             Jack
>>
>> My patch will build and install libgcc_nonshared.a for all targets.  If you
>> don't link against it, nothing is changed.  On Linux, it is used via the
>> init_spec change.
>
> Isn't...
>
> diff --git a/gcc/gcc.c b/gcc/gcc.c
> index d956c36..3fbd549 100644
> --- a/gcc/gcc.c
> +++ b/gcc/gcc.c
> @@ -1566,6 +1566,7 @@ init_spec (void)
>         if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
>           {
>             init_gcc_specs (&obstack,
> +                           "-lgcc_nonshared "
>                             "-lgcc_s"
>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>                             " -lunwind"
> @@ -1591,6 +1592,7 @@ init_spec (void)
>             /* Ug.  We don't know shared library extensions.  Hope that
>                systems that use this form don't do shared libraries.  */
>             init_gcc_specs (&obstack,
> +                           "libgcc_nonshared.a%s "
>                             "-lgcc_s",
>                             "libgcc.a%s",
>                             "libgcc_eh.a%s"
>
> problematic for Solaris? I am unfamiliar with the Solaris spec
> handling but sol2.h doesn't seem to have any instances of -lgcc which
> might imply they use the stock compiler invocation which will now have
> a non-existent libgcc_nonshared static library.

libgcc_nonshared.a is built and installed for all targets.

>      Also, are you leaving the cpu symbols in libgcc.a on non-linux
> targets? If not, the linkage failure reported in
> https://gcc.gnu.org/ml/gcc-patches/2015-03/msg01668.html will occur,
> no?

My current patch doesn't change what are in libgcc.a.  It
adds libgcc_nonshared.a for all targets, which contains
the same cpuinfo.o as in libgcc.a or a dummy .o if libgcc.a
doesn't have cpuinfo.o.

-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jack Howarth]

On Tue, Mar 31, 2015 at 1:00 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Tue, Mar 31, 2015 at 9:39 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>> On Tue, Mar 31, 2015 at 12:14 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> On Tue, Mar 31, 2015 at 9:09 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>>>> H.J.,
>>>>     Did you attach the correct version of the patch? I don't see
>>>> anything conditional on linux.
>>>>             Jack
>>>
>>> My patch will build and install libgcc_nonshared.a for all targets.  If you
>>> don't link against it, nothing is changed.  On Linux, it is used via the
>>> init_spec change.
>>
>> Isn't...
>>
>> diff --git a/gcc/gcc.c b/gcc/gcc.c
>> index d956c36..3fbd549 100644
>> --- a/gcc/gcc.c
>> +++ b/gcc/gcc.c
>> @@ -1566,6 +1566,7 @@ init_spec (void)
>>         if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
>>           {
>>             init_gcc_specs (&obstack,
>> +                           "-lgcc_nonshared "
>>                             "-lgcc_s"
>>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>>                             " -lunwind"
>> @@ -1591,6 +1592,7 @@ init_spec (void)
>>             /* Ug.  We don't know shared library extensions.  Hope that
>>                systems that use this form don't do shared libraries.  */
>>             init_gcc_specs (&obstack,
>> +                           "libgcc_nonshared.a%s "
>>                             "-lgcc_s",
>>                             "libgcc.a%s",
>>                             "libgcc_eh.a%s"
>>
>> problematic for Solaris? I am unfamiliar with the Solaris spec
>> handling but sol2.h doesn't seem to have any instances of -lgcc which
>> might imply they use the stock compiler invocation which will now have
>> a non-existent libgcc_nonshared static library.
>
> libgcc_nonshared.a is built and installed for all targets.
>
>>      Also, are you leaving the cpu symbols in libgcc.a on non-linux
>> targets? If not, the linkage failure reported in
>> https://gcc.gnu.org/ml/gcc-patches/2015-03/msg01668.html will occur,
>> no?
>
> My current patch doesn't change what are in libgcc.a.  It
> adds libgcc_nonshared.a for all targets, which contains
> the same cpuinfo.o as in libgcc.a or a dummy .o if libgcc.a
> doesn't have cpuinfo.o.


I can confirm that the most current patch bootstraps on
x86_64-apple-darwin14 and that all of the new tests show up as
unsupported in the test suite.
          Jack

>
> --
> H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

[-- Attachment #1: Type: text/plain, Size: 2634 bytes --]

On Tue, Mar 31, 2015 at 11:33 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
> On Tue, Mar 31, 2015 at 1:00 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Tue, Mar 31, 2015 at 9:39 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>>> On Tue, Mar 31, 2015 at 12:14 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>>> On Tue, Mar 31, 2015 at 9:09 AM, Jack Howarth <howarth.at.gcc@gmail.com> wrote:
>>>>> H.J.,
>>>>>     Did you attach the correct version of the patch? I don't see
>>>>> anything conditional on linux.
>>>>>             Jack
>>>>
>>>> My patch will build and install libgcc_nonshared.a for all targets.  If you
>>>> don't link against it, nothing is changed.  On Linux, it is used via the
>>>> init_spec change.
>>>
>>> Isn't...
>>>
>>> diff --git a/gcc/gcc.c b/gcc/gcc.c
>>> index d956c36..3fbd549 100644
>>> --- a/gcc/gcc.c
>>> +++ b/gcc/gcc.c
>>> @@ -1566,6 +1566,7 @@ init_spec (void)
>>>         if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
>>>           {
>>>             init_gcc_specs (&obstack,
>>> +                           "-lgcc_nonshared "
>>>                             "-lgcc_s"
>>>  #ifdef USE_LIBUNWIND_EXCEPTIONS
>>>                             " -lunwind"
>>> @@ -1591,6 +1592,7 @@ init_spec (void)
>>>             /* Ug.  We don't know shared library extensions.  Hope that
>>>                systems that use this form don't do shared libraries.  */
>>>             init_gcc_specs (&obstack,
>>> +                           "libgcc_nonshared.a%s "
>>>                             "-lgcc_s",
>>>                             "libgcc.a%s",
>>>                             "libgcc_eh.a%s"
>>>
>>> problematic for Solaris? I am unfamiliar with the Solaris spec
>>> handling but sol2.h doesn't seem to have any instances of -lgcc which
>>> might imply they use the stock compiler invocation which will now have
>>> a non-existent libgcc_nonshared static library.
>>
>> libgcc_nonshared.a is built and installed for all targets.
>>
>>>      Also, are you leaving the cpu symbols in libgcc.a on non-linux
>>> targets? If not, the linkage failure reported in
>>> https://gcc.gnu.org/ml/gcc-patches/2015-03/msg01668.html will occur,
>>> no?
>>
>> My current patch doesn't change what are in libgcc.a.  It
>> adds libgcc_nonshared.a for all targets, which contains
>> the same cpuinfo.o as in libgcc.a or a dummy .o if libgcc.a
>> doesn't have cpuinfo.o.
>
>
> I can confirm that the most current patch bootstraps on
> x86_64-apple-darwin14 and that all of the new tests show up as
> unsupported in the test suite.
>           Jack

I am re-posting this patch.  OK for trunk?

Thanks.

-- 
H.J.

[-- Attachment #2: 0001-Hide-__cpu_indicator_init-__cpu_model-from-linker.patch --]
[-- Type: text/x-patch, Size: 10127 bytes --]

From 9ddf859ba7112c3459f709fbb7aaaf9eabf485b6 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sun, 29 Mar 2015 18:03:49 -0700
Subject: [PATCH] Hide __cpu_indicator_init/__cpu_model from linker

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch hides __cpu_indicator_init/__cpu_model from linker to force linker
to resolve __cpu_indicator_init/__cpu_model to their hidden definitions
in libgcc_nonshared.a while providing backward binary compatibility.  The
new libgcc_nonshared.a is always linked togther with -lgcc_s and -lgcc.

gcc/

	PR target/65612
	* gcc.c (init_spec): Add -lgcc_nonshared/libgcc_nonshared.a%s
	to -lgcc_s.

gcc/testsuite/

	PR target/65612
	* g++.dg/ext/mv18.C: New test.
	* g++.dg/ext/mv19.C: Likewise.
	* g++.dg/ext/mv20.C: Likewise.
	* g++.dg/ext/mv21.C: Likewise.
	* g++.dg/ext/mv22.C: Likewise.
	* g++.dg/ext/mv23.C: Likewise.

libgcc/

	PR target/65612
	* Makefile.in (LIB2ADDNONSHARED): New.
	(libgcc-nonshared-objects): Likewise.
	(libgcc_nonshared.a): Likewise.
	Check unsupported files in LIB2ADDNONSHARED.
	(iter-items): Add $(LIB2ADDNONSHARED).
	(all): Depend on libgcc_nonshared.a.
	($(libgcc-nonshared-objects)): Depend on libgcc_tm.h.
	(install-leaf): Install libgcc_nonshared.a.
	* shared-object.mk: Check empty $o.
	* config/i386/cpuinfo.c (__cpu_model): Initialize.
	(__cpu_indicator_init@GCC_4.8.0): New.
	(__cpu_model@GCC_4.8.0): Likewise.
	* config/i386/t-cpuinfo (LIB2ADDNONSHARED): New.
	* config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
	-DUSE_ELF_SYMVER.
---
 gcc/gcc.c                       |  2 ++
 gcc/testsuite/g++.dg/ext/mv18.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv19.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv20.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv21.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv22.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/mv23.C |  7 +++++++
 libgcc/Makefile.in              | 22 +++++++++++++++++++---
 libgcc/config/i386/cpuinfo.c    |  7 ++++++-
 libgcc/config/i386/t-cpuinfo    |  1 +
 libgcc/config/i386/t-linux      |  2 +-
 libgcc/shared-object.mk         |  2 ++
 12 files changed, 73 insertions(+), 5 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/mv18.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv19.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv20.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv21.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv22.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv23.C

diff --git a/gcc/gcc.c b/gcc/gcc.c
index d956c36..3fbd549 100644
--- a/gcc/gcc.c
+++ b/gcc/gcc.c
@@ -1566,6 +1566,7 @@ init_spec (void)
 	if (in_sep && *p == '-' && strncmp (p, "-lgcc", 5) == 0)
 	  {
 	    init_gcc_specs (&obstack,
+			    "-lgcc_nonshared "
 			    "-lgcc_s"
 #ifdef USE_LIBUNWIND_EXCEPTIONS
 			    " -lunwind"
@@ -1591,6 +1592,7 @@ init_spec (void)
 	    /* Ug.  We don't know shared library extensions.  Hope that
 	       systems that use this form don't do shared libraries.  */
 	    init_gcc_specs (&obstack,
+			    "libgcc_nonshared.a%s "
 			    "-lgcc_s",
 			    "libgcc.a%s",
 			    "libgcc_eh.a%s"
diff --git a/gcc/testsuite/g++.dg/ext/mv18.C b/gcc/testsuite/g++.dg/ext/mv18.C
new file mode 100644
index 0000000..1f024de
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv18.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv19.C b/gcc/testsuite/g++.dg/ext/mv19.C
new file mode 100644
index 0000000..d1ea788
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv19.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv20.C b/gcc/testsuite/g++.dg/ext/mv20.C
new file mode 100644
index 0000000..98f7408
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv20.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv21.C b/gcc/testsuite/g++.dg/ext/mv21.C
new file mode 100644
index 0000000..9708ad9
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv21.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv22.C b/gcc/testsuite/g++.dg/ext/mv22.C
new file mode 100644
index 0000000..2550136
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv22.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv23.C b/gcc/testsuite/g++.dg/ext/mv23.C
new file mode 100644
index 0000000..f00afb0
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv23.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/libgcc/Makefile.in b/libgcc/Makefile.in
index 88ddfea..f78b1ac 100644
--- a/libgcc/Makefile.in
+++ b/libgcc/Makefile.in
@@ -261,6 +261,10 @@ LIB2ADD =
 # Assembler files should have names ending in `.S'.
 LIB2ADD_ST =
 
+# List of extra C and assembler files to add to libgcc_nonshared.a.
+# Assembler files should have names ending in `.S'.
+LIB2ADDNONSHARED =
+
 # Specify the directories to be searched for header files.
 # Both . and srcdir are used, in that order,
 # so that *config.h will be found in the compilation
@@ -807,8 +811,15 @@ endif
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD))))
 libgcc-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADD_ST))))
 
+# Build LIB2ADDNONSHARED.
+ifneq ($(filter-out %.c %.S %.asm,$(LIB2ADDNONSHARED)),)
+$(error Unsupported files in LIB2ADDNONSHARED.)
+endif
+
+libgcc-nonshared-objects += $(addsuffix $(objext),$(basename $(notdir $(LIB2ADDNONSHARED))))
+
 c_flags :=
-iter-items := $(LIB2ADD) $(LIB2ADD_ST)
+iter-items := $(LIB2ADD) $(LIB2ADD_ST) $(LIB2ADDNONSHARED)
 include $(iterator)
 
 ifeq ($(enable_shared),yes)
@@ -890,8 +901,9 @@ libgcc.a: $(libgcc-objects)
 libgcov.a: $(libgcov-objects)
 libunwind.a: $(libunwind-objects)
 libgcc_eh.a: $(libgcc-eh-objects)
+libgcc_nonshared.a: $(libgcc-nonshared-objects)
 
-libgcc.a libgcov.a libunwind.a libgcc_eh.a:
+libgcc.a libgcov.a libunwind.a libgcc_eh.a libgcc_nonshared.a:
 	-rm -f $@
 
 	objects="$(objects)";					\
@@ -905,7 +917,7 @@ libgcc.a libgcov.a libunwind.a libgcc_eh.a:
 
 	$(RANLIB) $@
 
-all: libgcc.a libgcov.a
+all: libgcc.a libgcc_nonshared.a libgcov.a
 
 ifneq ($(LIBUNWIND),)
 all: libunwind.a
@@ -1051,6 +1063,7 @@ libgcc-extra-parts: $(EXTRA_PARTS)
 all: $(extra-parts)
 
 $(libgcc-objects) $(libgcc-s-objects) $(libgcc-eh-objects) \
+	$(libgcc-nonshared-objects) \
 	$(libgcov-objects) \
 	$(libunwind-objects) $(libunwind-s-objects) \
 	$(EXTRA_PARTS): libgcc_tm.h
@@ -1118,6 +1131,9 @@ install-leaf: $(install-shared) $(install-libunwind)
 	$(INSTALL_DATA) libgcc.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc.a
+	$(INSTALL_DATA) libgcc_nonshared.a $(DESTDIR)$(inst_libdir)/
+	chmod 644 $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
+	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcc_nonshared.a
 	$(INSTALL_DATA) libgcov.a $(DESTDIR)$(inst_libdir)/
 	chmod 644 $(DESTDIR)$(inst_libdir)/libgcov.a
 	$(RANLIB) $(DESTDIR)$(inst_libdir)/libgcov.a
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..f6f91dd 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };
 
 
 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)
 
   return 0;
 }
+
+#if defined SHARED && defined USE_ELF_SYMVER
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
diff --git a/libgcc/config/i386/t-cpuinfo b/libgcc/config/i386/t-cpuinfo
index 31c8db9..b20ef30 100644
--- a/libgcc/config/i386/t-cpuinfo
+++ b/libgcc/config/i386/t-cpuinfo
@@ -1 +1,2 @@
 LIB2ADD += $(srcdir)/config/i386/cpuinfo.c
+LIB2ADDNONSHARED += $(srcdir)/config/i386/cpuinfo.c
diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
index 4f47f7b..11bb46e 100644
--- a/libgcc/config/i386/t-linux
+++ b/libgcc/config/i386/t-linux
@@ -3,4 +3,4 @@
 # t-slibgcc-elf-ver and t-linux
 SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver
 
-HOST_LIBGCC2_CFLAGS += -mlong-double-80
+HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER
diff --git a/libgcc/shared-object.mk b/libgcc/shared-object.mk
index efac797..306176d 100644
--- a/libgcc/shared-object.mk
+++ b/libgcc/shared-object.mk
@@ -4,6 +4,7 @@
 o := $(firstword $(iter-items))
 iter-items := $(filter-out $o,$(iter-items))
 
+ifneq ($o,)
 base := $(basename $(notdir $o))
 
 c_flags-$o := $(c_flags)
@@ -36,3 +37,4 @@ $(base)_s$(objext): $o
 	$(gcc_s_compile) -c $(as_flags-$<) $<
 
 endif
+endif
-- 
1.9.3

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Mike Stump]

On Apr 14, 2015, at 8:07 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> I can confirm that the most current patch bootstraps on
>> x86_64-apple-darwin14 and that all of the new tests show up as
>> unsupported in the test suite.
>>          Jack
> 
> I am re-posting this patch.  OK for trunk?

If Jack is happy, I’m happy.  :-)  That leaves the x86 people to comment on it.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Uros Bizjak]

On Thu, Apr 16, 2015 at 6:28 PM, Mike Stump <mikestump@comcast.net> wrote:
> On Apr 14, 2015, at 8:07 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> I can confirm that the most current patch bootstraps on
>>> x86_64-apple-darwin14 and that all of the new tests show up as
>>> unsupported in the test suite.
>>>          Jack
>>
>> I am re-posting this patch.  OK for trunk?
>
> If Jack is happy, I’m happy.  :-)  That leaves the x86 people to comment on it.

What about Solaris?

Uros.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Fri, Apr 17, 2015 at 1:05 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Thu, Apr 16, 2015 at 6:28 PM, Mike Stump <mikestump@comcast.net> wrote:
>> On Apr 14, 2015, at 8:07 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>>> I can confirm that the most current patch bootstraps on
>>>> x86_64-apple-darwin14 and that all of the new tests show up as
>>>> unsupported in the test suite.
>>>>          Jack
>>>
>>> I am re-posting this patch.  OK for trunk?
>>
>> If Jack is happy, I’m happy.  :-)  That leaves the x86 people to comment on it.
>
> What about Solaris?
>
> Uros.

There are

diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..f6f91dd 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };


 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)

   return 0;
 }
+
+#if defined SHARED && defined USE_ELF_SYMVER
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
index 4f47f7b..11bb46e 100644
--- a/libgcc/config/i386/t-linux
+++ b/libgcc/config/i386/t-linux
@@ -3,4 +3,4 @@
 # t-slibgcc-elf-ver and t-linux
 SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver

-HOST_LIBGCC2_CFLAGS += -mlong-double-80
+HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER

USE_ELF_SYMVER is only defined for Linux.  This patch won't
break Solaris.

-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Uros Bizjak]

On Fri, Apr 17, 2015 at 12:36 PM, H.J. Lu <hjl.tools@gmail.com> wrote:

>>>>> I can confirm that the most current patch bootstraps on
>>>>> x86_64-apple-darwin14 and that all of the new tests show up as
>>>>> unsupported in the test suite.
>>>>>          Jack
>>>>
>>>> I am re-posting this patch.  OK for trunk?
>>>
>>> If Jack is happy, I’m happy.  :-)  That leaves the x86 people to comment on it.
>>
>> What about Solaris?
>>
>> Uros.
>
> There are

[...]

Assuming Jakub is OK with the patch, let's go ahead with it.

OK for mainline.

Thanks,
Uros.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Uros Bizjak]

On Fri, Apr 17, 2015 at 1:04 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Fri, Apr 17, 2015 at 12:36 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>
>>>>>> I can confirm that the most current patch bootstraps on
>>>>>> x86_64-apple-darwin14 and that all of the new tests show up as
>>>>>> unsupported in the test suite.
>>>>>>          Jack
>>>>>
>>>>> I am re-posting this patch.  OK for trunk?
>>>>
>>>> If Jack is happy, I’m happy.  :-)  That leaves the x86 people to comment on it.
>>>
>>> What about Solaris?
>>>
>>> Uros.
>>
>> There are
>
> [...]
>
> Assuming Jakub is OK with the patch, let's go ahead with it.
>
> OK for mainline.

Ehm, the approval is for x86 part, you still need approval from libgcc
maintainer.

Uros.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jakub Jelinek]

On Fri, Apr 17, 2015 at 01:04:20PM +0200, Uros Bizjak wrote:
> On Fri, Apr 17, 2015 at 12:36 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> 
> >>>>> I can confirm that the most current patch bootstraps on
> >>>>> x86_64-apple-darwin14 and that all of the new tests show up as
> >>>>> unsupported in the test suite.
> >>>>>          Jack
> >>>>
> >>>> I am re-posting this patch.  OK for trunk?
> >>>
> >>> If Jack is happy, I’m happy.  :-)  That leaves the x86 people to comment on it.
> >>
> >> What about Solaris?
> >>
> >> Uros.
> >
> > There are
> 
> [...]
> 
> Assuming Jakub is OK with the patch, let's go ahead with it.
> 
> OK for mainline.

I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
creating yet another library for that.
So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
static_name is always linked in, in the switch combinations that it isn't
right now of course after shared_name rather than before that.
I thought we've fixed that years ago...

	Jakub

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Fri, Apr 17, 2015 at 4:37 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> On Fri, Apr 17, 2015 at 01:04:20PM +0200, Uros Bizjak wrote:
>> On Fri, Apr 17, 2015 at 12:36 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>
>> >>>>> I can confirm that the most current patch bootstraps on
>> >>>>> x86_64-apple-darwin14 and that all of the new tests show up as
>> >>>>> unsupported in the test suite.
>> >>>>>          Jack
>> >>>>
>> >>>> I am re-posting this patch.  OK for trunk?
>> >>>
>> >>> If Jack is happy, I’m happy.  :-)  That leaves the x86 people to comment on it.
>> >>
>> >> What about Solaris?
>> >>
>> >> Uros.
>> >
>> > There are
>>
>> [...]
>>
>> Assuming Jakub is OK with the patch, let's go ahead with it.
>>
>> OK for mainline.
>
> I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
> creating yet another library for that.
> So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
> of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
> static_name is always linked in, in the switch combinations that it isn't
> right now of course after shared_name rather than before that.
> I thought we've fixed that years ago...
>

We never pass -lgcc to linker when building C++ DSO:

 /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/collect2 -plugin
/usr/libexec/gcc/x86_64-redhat-linux/4.9.2/liblto_plugin.so
-plugin-opt=/usr/libexec/gcc/x86_64-redhat-linux/4.9.2/lto-wrapper
-plugin-opt=-fresolution=/tmp/ccZC7iqy.res
-plugin-opt=-pass-through=-lgcc_s -plugin-opt=-pass-through=-lc
-plugin-opt=-pass-through=-lgcc_s --build-id --no-add-needed
--eh-frame-hdr --hash-style=gnu -m elf_x86_64 -shared
/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crti.o
/usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtbeginS.o
-L/usr/lib/gcc/x86_64-redhat-linux/4.9.2
-L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64
-L/lib/../lib64 -L/usr/lib/../lib64
-L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../.. x.o -lstdc++ -lm
-lgcc_s -lc -lgcc_s /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtendS.o
/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crtn.o
[hjl@gnu-32 tmp]$

That is why libgcc_nonshared.a is needed.


-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jakub Jelinek]

On Fri, Apr 17, 2015 at 04:48:48AM -0700, H.J. Lu wrote:
> > I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
> > creating yet another library for that.
> > So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
> > of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
> > static_name is always linked in, in the switch combinations that it isn't
> > right now of course after shared_name rather than before that.
> > I thought we've fixed that years ago...
> >
> 
> We never pass -lgcc to linker when building C++ DSO:
> 
>  /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/collect2 -plugin
> /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/liblto_plugin.so
> -plugin-opt=/usr/libexec/gcc/x86_64-redhat-linux/4.9.2/lto-wrapper
> -plugin-opt=-fresolution=/tmp/ccZC7iqy.res
> -plugin-opt=-pass-through=-lgcc_s -plugin-opt=-pass-through=-lc
> -plugin-opt=-pass-through=-lgcc_s --build-id --no-add-needed
> --eh-frame-hdr --hash-style=gnu -m elf_x86_64 -shared
> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crti.o
> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtbeginS.o
> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2
> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64
> -L/lib/../lib64 -L/usr/lib/../lib64
> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../.. x.o -lstdc++ -lm
> -lgcc_s -lc -lgcc_s /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtendS.o
> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crtn.o
> [hjl@gnu-32 tmp]$
> 
> That is why libgcc_nonshared.a is needed.

See what I wrote.  I think it is a bug that we don't do that, in your case
we should pass -lgcc_s -lgcc -lc -lgcc_s -lgcc.
Or, if you don't want to change that, as the multi-versioning change is
i386/x86_64 only change, just ensure that those targets have
t-slibgcc-libgcc in libgcc/config.host and thus behave like most other linux
targets where -lgcc is linked in always after -lgcc_s.

	Jakub

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

[-- Attachment #1: Type: text/plain, Size: 2628 bytes --]

On Fri, Apr 17, 2015 at 4:59 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> On Fri, Apr 17, 2015 at 04:48:48AM -0700, H.J. Lu wrote:
>> > I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
>> > creating yet another library for that.
>> > So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
>> > of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
>> > static_name is always linked in, in the switch combinations that it isn't
>> > right now of course after shared_name rather than before that.
>> > I thought we've fixed that years ago...
>> >
>>
>> We never pass -lgcc to linker when building C++ DSO:
>>
>>  /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/collect2 -plugin
>> /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/liblto_plugin.so
>> -plugin-opt=/usr/libexec/gcc/x86_64-redhat-linux/4.9.2/lto-wrapper
>> -plugin-opt=-fresolution=/tmp/ccZC7iqy.res
>> -plugin-opt=-pass-through=-lgcc_s -plugin-opt=-pass-through=-lc
>> -plugin-opt=-pass-through=-lgcc_s --build-id --no-add-needed
>> --eh-frame-hdr --hash-style=gnu -m elf_x86_64 -shared
>> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crti.o
>> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtbeginS.o
>> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2
>> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64
>> -L/lib/../lib64 -L/usr/lib/../lib64
>> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../.. x.o -lstdc++ -lm
>> -lgcc_s -lc -lgcc_s /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtendS.o
>> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crtn.o
>> [hjl@gnu-32 tmp]$
>>
>> That is why libgcc_nonshared.a is needed.
>
> See what I wrote.  I think it is a bug that we don't do that, in your case
> we should pass -lgcc_s -lgcc -lc -lgcc_s -lgcc.
> Or, if you don't want to change that, as the multi-versioning change is
> i386/x86_64 only change, just ensure that those targets have
> t-slibgcc-libgcc in libgcc/config.host and thus behave like most other linux
> targets where -lgcc is linked in always after -lgcc_s.
>
>         Jakub

This patch works for me.  OK for trunk?

gcc/testsuite/

PR target/65612
* g++.dg/ext/mv18.C: New test.
* g++.dg/ext/mv19.C: Likewise.
* g++.dg/ext/mv20.C: Likewise.
* g++.dg/ext/mv21.C: Likewise.
* g++.dg/ext/mv22.C: Likewise.
* g++.dg/ext/mv23.C: Likewise.

libgcc/

PR target/65612
* config.host (tmake_file): Add t-slibgcc-libgcc for Linux/x86.
* config/i386/cpuinfo.c (__cpu_model): Initialize.
(__cpu_indicator_init@GCC_4.8.0): New.
(__cpu_model@GCC_4.8.0): Likewise.
* config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
-DUSE_ELF_SYMVER.

Thanks.

-- 
H.J.

[-- Attachment #2: 0001-Hide-__cpu_indicator_init-__cpu_model-from-linker.patch --]
[-- Type: application/octet-stream, Size: 6190 bytes --]

From ea6a6f31551b6cb96d27db4388499a259aaceeda Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sun, 29 Mar 2015 18:03:49 -0700
Subject: [PATCH] Hide __cpu_indicator_init/__cpu_model from linker

We shouldn't call external function, __cpu_indicator_init, while an object
is being relocated since its .got.plt section hasn't been updated.  It
works for non-PIE since no update on .got.plt section is required.  This
patch creates libgcc.so as a linker script, hides __cpu_indicator_init
and __cpu_model in libgcc.so.1 from linker, forces linker to resolve
__cpu_indicator_init and __cpu_model to their hidden definitions in
libgcc.a while providing backward binary compatibility.

gcc/testsuite/

	PR target/65612
	* g++.dg/ext/mv18.C: New test.
	* g++.dg/ext/mv19.C: Likewise.
	* g++.dg/ext/mv20.C: Likewise.
	* g++.dg/ext/mv21.C: Likewise.
	* g++.dg/ext/mv22.C: Likewise.
	* g++.dg/ext/mv23.C: Likewise.

libgcc/

	PR target/65612
	* config.host (tmake_file): Add t-slibgcc-libgcc for Linux/x86.
	* config/i386/cpuinfo.c (__cpu_model): Initialize.
	(__cpu_indicator_init@GCC_4.8.0): New.
	(__cpu_model@GCC_4.8.0): Likewise.
	* config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
	-DUSE_ELF_SYMVER.
---
 gcc/testsuite/g++.dg/ext/mv18.C | 7 +++++++
 gcc/testsuite/g++.dg/ext/mv19.C | 7 +++++++
 gcc/testsuite/g++.dg/ext/mv20.C | 7 +++++++
 gcc/testsuite/g++.dg/ext/mv21.C | 7 +++++++
 gcc/testsuite/g++.dg/ext/mv22.C | 7 +++++++
 gcc/testsuite/g++.dg/ext/mv23.C | 7 +++++++
 libgcc/config.host              | 2 +-
 libgcc/config/i386/cpuinfo.c    | 7 ++++++-
 libgcc/config/i386/t-linux      | 2 +-
 9 files changed, 50 insertions(+), 3 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/mv18.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv19.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv20.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv21.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv22.C
 create mode 100644 gcc/testsuite/g++.dg/ext/mv23.C

diff --git a/gcc/testsuite/g++.dg/ext/mv18.C b/gcc/testsuite/g++.dg/ext/mv18.C
new file mode 100644
index 0000000..1f024de
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv18.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv19.C b/gcc/testsuite/g++.dg/ext/mv19.C
new file mode 100644
index 0000000..d1ea788
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv19.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv20.C b/gcc/testsuite/g++.dg/ext/mv20.C
new file mode 100644
index 0000000..98f7408
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv20.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target pie } */
+/* { dg-options "-O2 -fPIE -pie -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv21.C b/gcc/testsuite/g++.dg/ext/mv21.C
new file mode 100644
index 0000000..9708ad9
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv21.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static" } */
+
+#include "mv1.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv22.C b/gcc/testsuite/g++.dg/ext/mv22.C
new file mode 100644
index 0000000..2550136
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv22.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static -march=x86-64" } */
+
+#include "mv14.C"
diff --git a/gcc/testsuite/g++.dg/ext/mv23.C b/gcc/testsuite/g++.dg/ext/mv23.C
new file mode 100644
index 0000000..f00afb0
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/mv23.C
@@ -0,0 +1,7 @@
+/* Test case to check if Multiversioning works.  */
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-ifunc "" }  */
+/* { dg-require-effective-target static } */
+/* { dg-options "-O2 -static -march=x86-64" } */
+
+#include "mv15.C"
diff --git a/libgcc/config.host b/libgcc/config.host
index 4b15895..ce14b9e 100644
--- a/libgcc/config.host
+++ b/libgcc/config.host
@@ -1306,7 +1306,7 @@ i[34567]86-*-linux* | x86_64-*-linux* | \
   i[34567]86-*-kfreebsd*-gnu | x86_64-*-kfreebsd*-gnu | \
   i[34567]86-*-knetbsd*-gnu | \
   i[34567]86-*-gnu*)
-	tmake_file="${tmake_file} t-tls i386/t-linux"
+	tmake_file="${tmake_file} t-tls i386/t-linux t-slibgcc-libgcc"
 	if test "$libgcc_cv_cfi" = "yes"; then
 		tmake_file="${tmake_file} t-stack i386/t-stack-i386"
 	fi
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index eaf2f10..f6f91dd 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -109,7 +109,7 @@ struct __processor_model
   unsigned int __cpu_type;
   unsigned int __cpu_subtype;
   unsigned int __cpu_features[1];
-} __cpu_model;
+} __cpu_model = { };
 
 
 /* Get the specific type of AMD CPU.  */
@@ -424,3 +424,8 @@ __cpu_indicator_init (void)
 
   return 0;
 }
+
+#if defined SHARED && defined USE_ELF_SYMVER
+__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
+__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#endif
diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
index 4f47f7b..11bb46e 100644
--- a/libgcc/config/i386/t-linux
+++ b/libgcc/config/i386/t-linux
@@ -3,4 +3,4 @@
 # t-slibgcc-elf-ver and t-linux
 SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver
 
-HOST_LIBGCC2_CFLAGS += -mlong-double-80
+HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER
-- 
2.1.0

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jakub Jelinek]

On Fri, Apr 17, 2015 at 05:36:30AM -0700, H.J. Lu wrote:
> This patch works for me.  OK for trunk?
> 
> gcc/testsuite/
> 
> PR target/65612
> * g++.dg/ext/mv18.C: New test.
> * g++.dg/ext/mv19.C: Likewise.
> * g++.dg/ext/mv20.C: Likewise.
> * g++.dg/ext/mv21.C: Likewise.
> * g++.dg/ext/mv22.C: Likewise.
> * g++.dg/ext/mv23.C: Likewise.
> 
> libgcc/
> 
> PR target/65612
> * config.host (tmake_file): Add t-slibgcc-libgcc for Linux/x86.
> * config/i386/cpuinfo.c (__cpu_model): Initialize.
> (__cpu_indicator_init@GCC_4.8.0): New.
> (__cpu_model@GCC_4.8.0): Likewise.
> * config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
> -DUSE_ELF_SYMVER.

LGTM.

	Jakub

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Rich Felker]

On Fri, Apr 17, 2015 at 05:36:30AM -0700, H.J. Lu wrote:
> On Fri, Apr 17, 2015@4:59 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> > On Fri, Apr 17, 2015 at 04:48:48AM -0700, H.J. Lu wrote:
> >> > I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
> >> > creating yet another library for that.
> >> > So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
> >> > of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
> >> > static_name is always linked in, in the switch combinations that it isn't
> >> > right now of course after shared_name rather than before that.
> >> > I thought we've fixed that years ago...
> >> >
> >>
> >> We never pass -lgcc to linker when building C++ DSO:
> >>
> >>  /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/collect2 -plugin
> >> /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/liblto_plugin.so
> >> -plugin-opt=/usr/libexec/gcc/x86_64-redhat-linux/4.9.2/lto-wrapper
> >> -plugin-opt=-fresolution=/tmp/ccZC7iqy.res
> >> -plugin-opt=-pass-through=-lgcc_s -plugin-opt=-pass-through=-lc
> >> -plugin-opt=-pass-through=-lgcc_s --build-id --no-add-needed
> >> --eh-frame-hdr --hash-style=gnu -m elf_x86_64 -shared
> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crti.o
> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtbeginS.o
> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2
> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64
> >> -L/lib/../lib64 -L/usr/lib/../lib64
> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../.. x.o -lstdc++ -lm
> >> -lgcc_s -lc -lgcc_s /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtendS.o
> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crtn.o
> >> [hjl@gnu-32 tmp]$
> >>
> >> That is why libgcc_nonshared.a is needed.
> >
> > See what I wrote.  I think it is a bug that we don't do that, in your case
> > we should pass -lgcc_s -lgcc -lc -lgcc_s -lgcc.
> > Or, if you don't want to change that, as the multi-versioning change is
> > i386/x86_64 only change, just ensure that those targets have
> > t-slibgcc-libgcc in libgcc/config.host and thus behave like most other linux
> > targets where -lgcc is linked in always after -lgcc_s.
> >
> >         Jakub
> 
> This patch works for me.  OK for trunk?
> 
> gcc/testsuite/
> 
> PR target/65612
> * g++.dg/ext/mv18.C: New test.
> * g++.dg/ext/mv19.C: Likewise.
> * g++.dg/ext/mv20.C: Likewise.
> * g++.dg/ext/mv21.C: Likewise.
> * g++.dg/ext/mv22.C: Likewise.
> * g++.dg/ext/mv23.C: Likewise.
> 
> libgcc/
> 
> PR target/65612
> * config.host (tmake_file): Add t-slibgcc-libgcc for Linux/x86.
> * config/i386/cpuinfo.c (__cpu_model): Initialize.
> (__cpu_indicator_init@GCC_4.8.0): New.
> (__cpu_model@GCC_4.8.0): Likewise.
> * config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
> -DUSE_ELF_SYMVER.
> 
> Thanks.

This patch seems to be making some trouble for musl's dynamic linker,
which I could go into details on if necessary, but I'm wondering if
there's any reason why simple visibility wasn't used if you want to
hide the symbols from the linker. Could you explain the motivation for
doing it this way? Is the intent for libgcc_s.so to refer to its own
internal copy of __cpu_model and __cpu_indicator_init? If so, it seems
wrong to have a relocation referring to them symbolically at all; this
should just be a relative relocation.

Rich

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Fri, May 8, 2015 at 4:00 PM, Rich Felker <dalias@libc.org> wrote:
> On Fri, Apr 17, 2015 at 05:36:30AM -0700, H.J. Lu wrote:
>> On Fri, Apr 17, 2015@4:59 AM, Jakub Jelinek <jakub@redhat.com> wrote:
>> > On Fri, Apr 17, 2015 at 04:48:48AM -0700, H.J. Lu wrote:
>> >> > I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
>> >> > creating yet another library for that.
>> >> > So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
>> >> > of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
>> >> > static_name is always linked in, in the switch combinations that it isn't
>> >> > right now of course after shared_name rather than before that.
>> >> > I thought we've fixed that years ago...
>> >> >
>> >>
>> >> We never pass -lgcc to linker when building C++ DSO:
>> >>
>> >>  /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/collect2 -plugin
>> >> /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/liblto_plugin.so
>> >> -plugin-opt=/usr/libexec/gcc/x86_64-redhat-linux/4.9.2/lto-wrapper
>> >> -plugin-opt=-fresolution=/tmp/ccZC7iqy.res
>> >> -plugin-opt=-pass-through=-lgcc_s -plugin-opt=-pass-through=-lc
>> >> -plugin-opt=-pass-through=-lgcc_s --build-id --no-add-needed
>> >> --eh-frame-hdr --hash-style=gnu -m elf_x86_64 -shared
>> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crti.o
>> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtbeginS.o
>> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2
>> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64
>> >> -L/lib/../lib64 -L/usr/lib/../lib64
>> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../.. x.o -lstdc++ -lm
>> >> -lgcc_s -lc -lgcc_s /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtendS.o
>> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crtn.o
>> >> [hjl@gnu-32 tmp]$
>> >>
>> >> That is why libgcc_nonshared.a is needed.
>> >
>> > See what I wrote.  I think it is a bug that we don't do that, in your case
>> > we should pass -lgcc_s -lgcc -lc -lgcc_s -lgcc.
>> > Or, if you don't want to change that, as the multi-versioning change is
>> > i386/x86_64 only change, just ensure that those targets have
>> > t-slibgcc-libgcc in libgcc/config.host and thus behave like most other linux
>> > targets where -lgcc is linked in always after -lgcc_s.
>> >
>> >         Jakub
>>
>> This patch works for me.  OK for trunk?
>>
>> gcc/testsuite/
>>
>> PR target/65612
>> * g++.dg/ext/mv18.C: New test.
>> * g++.dg/ext/mv19.C: Likewise.
>> * g++.dg/ext/mv20.C: Likewise.
>> * g++.dg/ext/mv21.C: Likewise.
>> * g++.dg/ext/mv22.C: Likewise.
>> * g++.dg/ext/mv23.C: Likewise.
>>
>> libgcc/
>>
>> PR target/65612
>> * config.host (tmake_file): Add t-slibgcc-libgcc for Linux/x86.
>> * config/i386/cpuinfo.c (__cpu_model): Initialize.
>> (__cpu_indicator_init@GCC_4.8.0): New.
>> (__cpu_model@GCC_4.8.0): Likewise.
>> * config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
>> -DUSE_ELF_SYMVER.
>>
>> Thanks.
>
> This patch seems to be making some trouble for musl's dynamic linker,
> which I could go into details on if necessary, but I'm wondering if

It sounds like a bug in musl's dynamic linker. foo@VERSION
is visible to ld.so, but invisible to ld.

> there's any reason why simple visibility wasn't used if you want to
> hide the symbols from the linker. Could you explain the motivation for
> doing it this way? Is the intent for libgcc_s.so to refer to its own
> internal copy of __cpu_model and __cpu_indicator_init? If so, it seems
> wrong to have a relocation referring to them symbolically at all; this
> should just be a relative relocation.

It is for backward binary compatibility so that the old binaries
which have references to libgcc_s.so.1 work with the new
libgcc_s.so.1.


-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Szabolcs Nagy]

* H.J. Lu <hjl.tools@gmail.com> [2015-04-17 05:36:30 -0700]:
> On Fri, Apr 17, 2015 at 4:59 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> > On Fri, Apr 17, 2015 at 04:48:48AM -0700, H.J. Lu wrote:
> >> > I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
> >> > creating yet another library for that.
> >> > So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
> >> > of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
> >> > static_name is always linked in, in the switch combinations that it isn't
> >> > right now of course after shared_name rather than before that.
> >> > I thought we've fixed that years ago...
> >> >

I think the patch committed for this is suboptimal.
(it breaks with musl libc on x86 if libgcc_s is linked into a binary)

the patch:
http://gcc.gnu.org/ml/gcc-patches/2015-04/msg00878.html
original thread:
http://gcc.gnu.org/ml/gcc-patches/2015-03/msg01520.html

The symbol versioning hack for __cpu_model and __cpu_indicator_init
makes them invisible to the musl dynamic linker so their relocation
fails with 'symbol not found' error.
(affects anything linked with -lgcc_s)

I don't yet understand all the constraints, what I gathered is

* x86 multi-versioning was broken with libgcc_s because ifunc
resolution tried to use a libgcc_s function before it was relocated.
(mv dispatch uses ifunc before relocation is finished, that calls
__builtin_cpu_init that turns into a call to __cpu_indicator_init
that is a constructor in libgcc_s.so to initialize __cpu_model).

* old binaries that happened to work still have to work with new
libgcc_s.so so __cpu_* must have a definition there.

* new binaries must get their own definition of __cpu_*
(ie statically linked from libgcc.a) to avoid the ifunc issue.

I would rename the __cpu_* symbols and put those only in libgcc.a
while keeping the old __cpu_* symbols still available in
libgcc_s.so: old binaries work with new gcc, new binaries get the
right definitions (so they work on old systems and new systems
too, and yes -lgcc should be passed to c++ dso linking).
Did i miss something?

(This is not in the gcc-5 branch so I assume It is OK to work
on a less intrusive fix.)


PS.: It would be nice if there was a specification for the ifunc
interface contract before any ifunc/multiversioned interface gets
into the compiler runtime (see libatomic.so) to avoid similar bugs
and incompatibilities between the dynamic linker/loader and compiler,
the dynamic linker cannot guarantee arbitrary ifunc calls to work.


> >>
> >> We never pass -lgcc to linker when building C++ DSO:
> >>
> >>  /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/collect2 -plugin
> >> /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/liblto_plugin.so
> >> -plugin-opt=/usr/libexec/gcc/x86_64-redhat-linux/4.9.2/lto-wrapper
> >> -plugin-opt=-fresolution=/tmp/ccZC7iqy.res
> >> -plugin-opt=-pass-through=-lgcc_s -plugin-opt=-pass-through=-lc
> >> -plugin-opt=-pass-through=-lgcc_s --build-id --no-add-needed
> >> --eh-frame-hdr --hash-style=gnu -m elf_x86_64 -shared
> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crti.o
> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtbeginS.o
> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2
> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64
> >> -L/lib/../lib64 -L/usr/lib/../lib64
> >> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../.. x.o -lstdc++ -lm
> >> -lgcc_s -lc -lgcc_s /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtendS.o
> >> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crtn.o
> >> [hjl@gnu-32 tmp]$
> >>
> >> That is why libgcc_nonshared.a is needed.
> >
> > See what I wrote.  I think it is a bug that we don't do that, in your case
> > we should pass -lgcc_s -lgcc -lc -lgcc_s -lgcc.
> > Or, if you don't want to change that, as the multi-versioning change is
> > i386/x86_64 only change, just ensure that those targets have
> > t-slibgcc-libgcc in libgcc/config.host and thus behave like most other linux
> > targets where -lgcc is linked in always after -lgcc_s.
> >
> >         Jakub
> 
> This patch works for me.  OK for trunk?
> 
> gcc/testsuite/
> 
> PR target/65612
> * g++.dg/ext/mv18.C: New test.
> * g++.dg/ext/mv19.C: Likewise.
> * g++.dg/ext/mv20.C: Likewise.
> * g++.dg/ext/mv21.C: Likewise.
> * g++.dg/ext/mv22.C: Likewise.
> * g++.dg/ext/mv23.C: Likewise.
> 
> libgcc/
> 
> PR target/65612
> * config.host (tmake_file): Add t-slibgcc-libgcc for Linux/x86.
> * config/i386/cpuinfo.c (__cpu_model): Initialize.
> (__cpu_indicator_init@GCC_4.8.0): New.
> (__cpu_model@GCC_4.8.0): Likewise.
> * config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
> -DUSE_ELF_SYMVER.
> 
> Thanks.
> 
> -- 
> H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[H.J. Lu]

On Sat, May 9, 2015 at 7:31 AM, Szabolcs Nagy <nsz@port70.net> wrote:
> * H.J. Lu <hjl.tools@gmail.com> [2015-04-17 05:36:30 -0700]:
>> On Fri, Apr 17, 2015 at 4:59 AM, Jakub Jelinek <jakub@redhat.com> wrote:
>> > On Fri, Apr 17, 2015 at 04:48:48AM -0700, H.J. Lu wrote:
>> >> > I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
>> >> > creating yet another library for that.
>> >> > So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
>> >> > of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
>> >> > static_name is always linked in, in the switch combinations that it isn't
>> >> > right now of course after shared_name rather than before that.
>> >> > I thought we've fixed that years ago...
>> >> >
>
> I think the patch committed for this is suboptimal.
> (it breaks with musl libc on x86 if libgcc_s is linked into a binary)
>
> the patch:
> http://gcc.gnu.org/ml/gcc-patches/2015-04/msg00878.html
> original thread:
> http://gcc.gnu.org/ml/gcc-patches/2015-03/msg01520.html
>
> The symbol versioning hack for __cpu_model and __cpu_indicator_init
> makes them invisible to the musl dynamic linker so their relocation
> fails with 'symbol not found' error.
> (affects anything linked with -lgcc_s)

There are

     4: 0000000000002b70   806 FUNC    GLOBAL DEFAULT   12
__cpu_indicator_init@GCC_4.8.0
    38: 00000000002153d0    16 OBJECT  GLOBAL DEFAULT   25 __cpu_model@GCC_4.8.0

and

000000000215000  0000000400000001 R_X86_64_64
0000000000002b70 __cpu_indicator_init@GCC_4.8.0 + 0
0000000000215220  0000002600000006 R_X86_64_GLOB_DAT
00000000002153d0 __cpu_model@GCC_4.8.0 + 0

in libgcc_s.so.1.  Musl ld.so must be fixed to handle it.



-- 
H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Szabolcs Nagy]

* H.J. Lu <hjl.tools@gmail.com> [2015-05-09 10:41:41 -0700]:
> On Sat, May 9, 2015 at 7:31 AM, Szabolcs Nagy <nsz@port70.net> wrote:
> >
> > The symbol versioning hack for __cpu_model and __cpu_indicator_init
> > makes them invisible to the musl dynamic linker so their relocation
> > fails with 'symbol not found' error.
> > (affects anything linked with -lgcc_s)
> 
> There are
> 
>      4: 0000000000002b70   806 FUNC    GLOBAL DEFAULT   12
> __cpu_indicator_init@GCC_4.8.0
>     38: 00000000002153d0    16 OBJECT  GLOBAL DEFAULT   25 __cpu_model@GCC_4.8.0
> 
> and
> 
> 000000000215000  0000000400000001 R_X86_64_64
> 0000000000002b70 __cpu_indicator_init@GCC_4.8.0 + 0
> 0000000000215220  0000002600000006 R_X86_64_GLOB_DAT
> 00000000002153d0 __cpu_model@GCC_4.8.0 + 0
> 
> in libgcc_s.so.1.  Musl ld.so must be fixed to handle it.
> 

(sorry, i missed the other discussion about the same issue)

i still think a solution that avoids depending on symbol
versioning in the dynamic linker would be preferable.
(versioning is a sun/gnu elf extension as far as i understand,
eg. not supported by posix dlsym).

i think it might be enough to add __cpu_indicator_init_local
as an alias to __cpu_indicator_init in libgcc.a and then use
the *_local symbol from the ifunc resolver, that way no new
dependency is added to libgcc_s.so handling.

(adding versioning support to musl is problematic with the
current code, a simpler workaround for musl is to provide its
own weak definition of __cpu_indicator_init, but that's ugly.)

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Szabolcs Nagy]

[-- Attachment #1: Type: text/plain, Size: 1681 bytes --]



On 09/05/15 19:57, Szabolcs Nagy wrote:
> * H.J. Lu <hjl.tools@gmail.com> [2015-05-09 10:41:41 -0700]:
>> There are
>>
>>      4: 0000000000002b70   806 FUNC    GLOBAL DEFAULT   12
>> __cpu_indicator_init@GCC_4.8.0
>>     38: 00000000002153d0    16 OBJECT  GLOBAL DEFAULT   25 __cpu_model@GCC_4.8.0
>>
>> and
>>
>> 000000000215000  0000000400000001 R_X86_64_64
>> 0000000000002b70 __cpu_indicator_init@GCC_4.8.0 + 0
>> 0000000000215220  0000002600000006 R_X86_64_GLOB_DAT
>> 00000000002153d0 __cpu_model@GCC_4.8.0 + 0
>>
>> in libgcc_s.so.1.  Musl ld.so must be fixed to handle it.
>>

Rich is looking at how to do this non-intrusively, but
it seems non-trivial (some users of musl prefer not to
resolve such versioned symbols).

> 
> i think it might be enough to add __cpu_indicator_init_local
> as an alias to __cpu_indicator_init in libgcc.a and then use
> the *_local symbol from the ifunc resolver, that way no new
> dependency is added to libgcc_s.so handling.

i tried this approach and it seems to work: passes all
multiversioning tests on x86_64.

i think it's no worse than the symver approach.

is it ok to change the current fix to this?


libgcc/Changelog:

2015-05-11  Szabolcs Nagy  <szabolcs.nagy@arm.com>

	* config/i386/cpuinfo.c (__cpu_indicator_init_local): Add.
	(__cpu_indicator_init@GCC_4.8.0, __cpu_model@GCC_4.8.0): Remove.

	* config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Remove -DUSE_ELF_SYMVER.

gcc/Changelog:

2015-05-11  Szabolcs Nagy  <szabolcs.nagy@arm.com>

	* config/i386/i386.c (ix86_expand_builtin): Make __builtin_cpu_init
	call __cpu_indicator_init_local instead of __cpu_indicator_init.

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: libgcc.diff --]
[-- Type: text/x-patch; name=libgcc.diff, Size: 1706 bytes --]

diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 7bd9ff3..7327cf3 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -38398,10 +38398,10 @@ ix86_expand_builtin (tree exp, rtx target, rtx subtarget,
     {
     case IX86_BUILTIN_CPU_INIT:
       {
-	/* Make it call __cpu_indicator_init in libgcc. */
+	/* Make it call __cpu_indicator_init_local in libgcc.a. */
 	tree call_expr, fndecl, type;
         type = build_function_type_list (integer_type_node, NULL_TREE); 
-	fndecl = build_fn_decl ("__cpu_indicator_init", type);
+	fndecl = build_fn_decl ("__cpu_indicator_init_local", type);
 	call_expr = build_call_expr (fndecl, 0); 
 	return expand_expr (call_expr, target, mode, EXPAND_NORMAL);
       }
diff --git a/libgcc/config/i386/cpuinfo.c b/libgcc/config/i386/cpuinfo.c
index f6f91dd..86b6d21 100644
--- a/libgcc/config/i386/cpuinfo.c
+++ b/libgcc/config/i386/cpuinfo.c
@@ -425,7 +425,7 @@ __cpu_indicator_init (void)
   return 0;
 }
 
-#if defined SHARED && defined USE_ELF_SYMVER
-__asm__ (".symver __cpu_indicator_init, __cpu_indicator_init@GCC_4.8.0");
-__asm__ (".symver __cpu_model, __cpu_model@GCC_4.8.0");
+#ifndef SHARED
+int __cpu_indicator_init_local (void)
+  __attribute__ ((weak, alias ("__cpu_indicator_init")));
 #endif
diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
index 11bb46e..4f47f7b 100644
--- a/libgcc/config/i386/t-linux
+++ b/libgcc/config/i386/t-linux
@@ -3,4 +3,4 @@
 # t-slibgcc-elf-ver and t-linux
 SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver
 
-HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER
+HOST_LIBGCC2_CFLAGS += -mlong-double-80

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jakub Jelinek]

On Mon, May 11, 2015 at 11:20:15AM +0100, Szabolcs Nagy wrote:
> 
> 
> On 09/05/15 19:57, Szabolcs Nagy wrote:
> > * H.J. Lu <hjl.tools@gmail.com> [2015-05-09 10:41:41 -0700]:
> >> There are
> >>
> >>      4: 0000000000002b70   806 FUNC    GLOBAL DEFAULT   12
> >> __cpu_indicator_init@GCC_4.8.0
> >>     38: 00000000002153d0    16 OBJECT  GLOBAL DEFAULT   25 __cpu_model@GCC_4.8.0
> >>
> >> and
> >>
> >> 000000000215000  0000000400000001 R_X86_64_64
> >> 0000000000002b70 __cpu_indicator_init@GCC_4.8.0 + 0
> >> 0000000000215220  0000002600000006 R_X86_64_GLOB_DAT
> >> 00000000002153d0 __cpu_model@GCC_4.8.0 + 0
> >>
> >> in libgcc_s.so.1.  Musl ld.so must be fixed to handle it.
> >>
> 
> Rich is looking at how to do this non-intrusively, but
> it seems non-trivial (some users of musl prefer not to
> resolve such versioned symbols).
> 
> > 
> > i think it might be enough to add __cpu_indicator_init_local
> > as an alias to __cpu_indicator_init in libgcc.a and then use
> > the *_local symbol from the ifunc resolver, that way no new
> > dependency is added to libgcc_s.so handling.
> 
> i tried this approach and it seems to work: passes all
> multiversioning tests on x86_64.
> 
> i think it's no worse than the symver approach.
> 
> is it ok to change the current fix to this?

No.  Instead of piling hacks like this just fix it in musl.
libgcc certainly isn't the only library that uses @ symbol versions,
e.g. libstdc++ does as well, as well as many other shared libraries.

	Jakub

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Szabolcs Nagy]

On 11/05/15 11:31, Jakub Jelinek wrote:
> On Mon, May 11, 2015 at 11:20:15AM +0100, Szabolcs Nagy wrote:
>>> i think it might be enough to add __cpu_indicator_init_local
>>> as an alias to __cpu_indicator_init in libgcc.a and then use
>>> the *_local symbol from the ifunc resolver, that way no new
>>> dependency is added to libgcc_s.so handling.
>>
>> i tried this approach and it seems to work: passes all
>> multiversioning tests on x86_64.
>>
>> i think it's no worse than the symver approach.
>>
>> is it ok to change the current fix to this?
> 
> No.  Instead of piling hacks like this just fix it in musl.
> libgcc certainly isn't the only library that uses @ symbol versions,
> e.g. libstdc++ does as well, as well as many other shared libraries.
> 

can you explain how using a standard elf feature is a hack,
but the current symver asm directive is not?

fyi, musl loader loads libstdc++ just fine because it has no
relocations for symbols which only has sym@version definition
(libgcc_s.so.1 has because __cpu_indicator_init is a ctor).

musl may end up supporting @version but that's an independent
quest.

(the one big pile of hacks here is multiversioning: each function
with mv has its own ifunc resolver repeating the same logic, then
the >1.5K cpuinfo.o is static linked into every single dso that
uses mv and the separate __cpu_model structs all have to be
initialized.. including the unused one in libgcc_s.so.1 because
of the ctor.. adding more startup overhead.

i'd gladly propose a patch to remove this feature if getting rid
of piling hacks has priority.. the current design seems problematic
to me for other archs that may need to call libc functions to do
the dispatch anyway.)

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jakub Jelinek]

On Mon, May 11, 2015 at 01:39:17PM +0100, Szabolcs Nagy wrote:
> fyi, musl loader loads libstdc++ just fine because it has no

But will it load any shared library that uses any of the 26 (if I count well
on x86_64) @ symbols from libstdc++.so.6?
readelf -Ws /lib64/libstdc++.so.6 | grep '@' | grep -v 'UND\|@@'
For libstdc++ that is primarily C++ apps and shared libraries
compiled/linked with GCC 4.0.0.

> musl may end up supporting @version but that's an independent
> quest.

It is not independent.  If musl claims to support symbol versioning, it
should support it properly, if not, then supposedly gcc configured for musl
can't be compatible with gcc configured for other linux C libraries, and
should force symbol versioning off.

	Jakub

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Szabolcs Nagy]

On 11/05/15 14:05, Jakub Jelinek wrote:
> On Mon, May 11, 2015 at 01:39:17PM +0100, Szabolcs Nagy wrote:
>> fyi, musl loader loads libstdc++ just fine because it has no
> 
> But will it load any shared library that uses any of the 26 (if I count well
> on x86_64) @ symbols from libstdc++.so.6?

i looked, but all of those symbols have @@ variant too
so at least the libraries would load with musl.

>> musl may end up supporting @version but that's an independent
>> quest.
> 
> It is not independent.  If musl claims to support symbol versioning, it
> should support it properly, if not, then supposedly gcc configured for musl
> can't be compatible with gcc configured for other linux C libraries, and
> should force symbol versioning off.

ok, but the current solution does not make that easy:
configuring gcc with

  --disable-gnu-indirect-function --disable-symvers

has no effect on libgcc_s.so.1 on linux.

(i can try to create a patch that removes the new
-DUSE_ELF_SYMVER from the libgcc cflags for musl, but that
seems a worse solution than the weak alias one).

(note that previously a simple spec file was enough to use an
existing gcc on linux to build things against musl...
this didnt work for c++ code that used libstdc++ headers, but
now it also fails for any build using -lgcc_s).

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Rich Felker]

On Mon, May 11, 2015 at 12:31:51PM +0200, Jakub Jelinek wrote:
> On Mon, May 11, 2015 at 11:20:15AM +0100, Szabolcs Nagy wrote:
> > 
> > 
> > On 09/05/15 19:57, Szabolcs Nagy wrote:
> > > * H.J. Lu <hjl.tools@gmail.com> [2015-05-09 10:41:41 -0700]:
> > >> There are
> > >>
> > >>      4: 0000000000002b70   806 FUNC    GLOBAL DEFAULT   12
> > >> __cpu_indicator_init@GCC_4.8.0
> > >>     38: 00000000002153d0    16 OBJECT  GLOBAL DEFAULT   25 __cpu_model@GCC_4.8.0
> > >>
> > >> and
> > >>
> > >> 000000000215000  0000000400000001 R_X86_64_64
> > >> 0000000000002b70 __cpu_indicator_init@GCC_4.8.0 + 0
> > >> 0000000000215220  0000002600000006 R_X86_64_GLOB_DAT
> > >> 00000000002153d0 __cpu_model@GCC_4.8.0 + 0
> > >>
> > >> in libgcc_s.so.1.  Musl ld.so must be fixed to handle it.
> > >>
> > 
> > Rich is looking at how to do this non-intrusively, but
> > it seems non-trivial (some users of musl prefer not to
> > resolve such versioned symbols).
> > 
> > > 
> > > i think it might be enough to add __cpu_indicator_init_local
> > > as an alias to __cpu_indicator_init in libgcc.a and then use
> > > the *_local symbol from the ifunc resolver, that way no new
> > > dependency is added to libgcc_s.so handling.
> > 
> > i tried this approach and it seems to work: passes all
> > multiversioning tests on x86_64.
> > 
> > i think it's no worse than the symver approach.
> > 
> > is it ok to change the current fix to this?
> 
> No.  Instead of piling hacks like this just fix it in musl.

I wouldn't call it piling hacks; it's an improvement as far as I can
tell since it remove symbolic relocations and replaces them with
relative ones.

> libgcc certainly isn't the only library that uses @ symbol versions,
> e.g. libstdc++ does as well, as well as many other shared libraries.

We haven't encountered such issues there.

Rich

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Rich Felker]

On Sat, May 09, 2015 at 10:41:41AM -0700, H.J. Lu wrote:
> On Sat, May 9, 2015 at 7:31 AM, Szabolcs Nagy <nsz@port70.net> wrote:
> > * H.J. Lu <hjl.tools@gmail.com> [2015-04-17 05:36:30 -0700]:
> >> On Fri, Apr 17, 2015 at 4:59 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> >> > On Fri, Apr 17, 2015 at 04:48:48AM -0700, H.J. Lu wrote:
> >> >> > I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
> >> >> > creating yet another library for that.
> >> >> > So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
> >> >> > of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
> >> >> > static_name is always linked in, in the switch combinations that it isn't
> >> >> > right now of course after shared_name rather than before that.
> >> >> > I thought we've fixed that years ago...
> >> >> >
> >
> > I think the patch committed for this is suboptimal.
> > (it breaks with musl libc on x86 if libgcc_s is linked into a binary)
> >
> > the patch:
> > http://gcc.gnu.org/ml/gcc-patches/2015-04/msg00878.html
> > original thread:
> > http://gcc.gnu.org/ml/gcc-patches/2015-03/msg01520.html
> >
> > The symbol versioning hack for __cpu_model and __cpu_indicator_init
> > makes them invisible to the musl dynamic linker so their relocation
> > fails with 'symbol not found' error.
> > (affects anything linked with -lgcc_s)
> 
> There are
> 
>      4: 0000000000002b70   806 FUNC    GLOBAL DEFAULT   12
> __cpu_indicator_init@GCC_4.8.0
>     38: 00000000002153d0    16 OBJECT  GLOBAL DEFAULT   25 __cpu_model@GCC_4.8.0
> 
> and
> 
> 000000000215000  0000000400000001 R_X86_64_64
> 0000000000002b70 __cpu_indicator_init@GCC_4.8.0 + 0
> 0000000000215220  0000002600000006 R_X86_64_GLOB_DAT
> 00000000002153d0 __cpu_model@GCC_4.8.0 + 0
> 
> in libgcc_s.so.1.  Musl ld.so must be fixed to handle it.

The intent in musl was always not to support symbol versioning. There
are various reasons for this, which I could go into, but I'd rather
not turn this thread into an argument about the merits of symbol
versioning.

Originally, musl ignored the version data completely, and would
happily have resolved the above symbol, but this also led to problems
where third-party libraries used symbol versioning and at runtime we
got the oldest-versioned symbol instead of the desired current one. So
I changed the lookup to use the same logic as ld, rejecting all
symbols with the hidden bit set, and getting us the newest one (which
is the only one we intended to support).

We can't just do the same version processing as glibc because we want
symbols in libc itself to resolve regardless of the version in the
reference (this is needed for the glibc ABI compat we provide). It
might work to honor symbol versions only when the library being
searched has versions, and otherwise allow symbols to match any
version requested, but I haven't thought this through entirely yet.

In any case I'd like the decision for what musl does about symbol
versions (which are not intended to be supported, anyway) to be
independent of libgcc's solution of this problem. We can solve it with
a hack on our side (just providing dummy symbols by those names) but
that's ugly too and I'd rather not do it.

Could you clarify the reasoning for why libgcc is using this hack with
a reference to an 'obsolete' symbol version rather than just
visibility?

Rich

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Jakub Jelinek]

On Sat, May 09, 2015 at 03:36:01PM -0400, Rich Felker wrote:
> Could you clarify the reasoning for why libgcc is using this hack with
> a reference to an 'obsolete' symbol version rather than just
> visibility?

Obviously for ABI compatibility reasons.  Older programs could be relying on
the symbols exported from libgcc_s.so.1 in the past.  We need to keep those
working, just don't want to allow linking new programs against this symbol
(because new programs should get their copy from libgcc.a now).

	Jakub

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Sriraman Tallam]

On Fri, Apr 17, 2015 at 5:36 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Fri, Apr 17, 2015 at 4:59 AM, Jakub Jelinek <jakub@redhat.com> wrote:
>> On Fri, Apr 17, 2015 at 04:48:48AM -0700, H.J. Lu wrote:
>>> > I don't like it.  Nonshared libgcc is libgcc.a, period.  No sense in
>>> > creating yet another library for that.
>>> > So, IMHO beyond making the __cpu* entrypoints compat symbols only (@ instead
>>> > of @@ symbol versions) the right fix is simply tweak init_gcc_spec, so that
>>> > static_name is always linked in, in the switch combinations that it isn't
>>> > right now of course after shared_name rather than before that.
>>> > I thought we've fixed that years ago...
>>> >
>>>
>>> We never pass -lgcc to linker when building C++ DSO:
>>>
>>>  /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/collect2 -plugin
>>> /usr/libexec/gcc/x86_64-redhat-linux/4.9.2/liblto_plugin.so
>>> -plugin-opt=/usr/libexec/gcc/x86_64-redhat-linux/4.9.2/lto-wrapper
>>> -plugin-opt=-fresolution=/tmp/ccZC7iqy.res
>>> -plugin-opt=-pass-through=-lgcc_s -plugin-opt=-pass-through=-lc
>>> -plugin-opt=-pass-through=-lgcc_s --build-id --no-add-needed
>>> --eh-frame-hdr --hash-style=gnu -m elf_x86_64 -shared
>>> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crti.o
>>> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtbeginS.o
>>> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2
>>> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64
>>> -L/lib/../lib64 -L/usr/lib/../lib64
>>> -L/usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../.. x.o -lstdc++ -lm
>>> -lgcc_s -lc -lgcc_s /usr/lib/gcc/x86_64-redhat-linux/4.9.2/crtendS.o
>>> /usr/lib/gcc/x86_64-redhat-linux/4.9.2/../../../../lib64/crtn.o
>>> [hjl@gnu-32 tmp]$
>>>
>>> That is why libgcc_nonshared.a is needed.
>>
>> See what I wrote.  I think it is a bug that we don't do that, in your case
>> we should pass -lgcc_s -lgcc -lc -lgcc_s -lgcc.
>> Or, if you don't want to change that, as the multi-versioning change is
>> i386/x86_64 only change, just ensure that those targets have
>> t-slibgcc-libgcc in libgcc/config.host and thus behave like most other linux
>> targets where -lgcc is linked in always after -lgcc_s.
>>
>>         Jakub
>
> This patch works for me.  OK for trunk?

H.J:  This patch assumes that libgcc.a is built with
-fvisibility=hidden and __cpu_indicator_init is LOCAL to libgcc.a.  We
have a config where libgcc.a is not built with hidden visibility.  Can
we consider this additional patch to make this explicit, mark
__cpu_indicator_init with hidden visiblity explicitly when not
building shared object.


--- config/i386/cpuinfo.c (revision 225800)
+++ config/i386/cpuinfo.c (working copy)
@@ -34,6 +34,9 @@
 #endif

 int __cpu_indicator_init (void)
+#if !defined(SHARED)
+__attribute__ ((visibility("hidden")))
+#endif
   __attribute__ ((constructor CONSTRUCTOR_PRIORITY));

 /* Get the specific type of AMD CPU.  */
@@ -321,6 +324,9 @@
    needs to be called explicitly there.  */

 int __attribute__ ((constructor CONSTRUCTOR_PRIORITY))
+#if !defined(SHARED)
+__attribute__ ((visibility("hidden")))
+#endif
 __cpu_indicator_init (void)


Also,  gold and ld have an incompatibility with symbol versioning as
discussed here:  https://sourceware.org/bugzilla/show_bug.cgi?id=18703

H.J. suggested this nice fix to solve this problem where BFD ld is
always used to build libgcc_s.so.1:

You can pass -fuse-ld=bfd to build libgcc_s.so.1 on Linux:

diff --git a/libgcc/config/i386/t-linux b/libgcc/config/i386/t-linux
index 11bb46e..12aab16 100644
--- a/libgcc/config/i386/t-linux
+++ b/libgcc/config/i386/t-linux
@@ -3,4 +3,8 @@
 # t-slibgcc-elf-ver and t-linux
 SHLIB_MAPFILES = libgcc-std.ver $(srcdir)/config/i386/libgcc-glibc.ver

+# Work around gold bug:
+# https://sourceware.org/bugzilla/show_bug.cgi?id=18703
+SHLIB_LDFLAGS += -fuse-ld=bfd
+
 HOST_LIBGCC2_CFLAGS += -mlong-double-80 -DUSE_ELF_SYMVER


Thanks
Sri

>
> gcc/testsuite/
>
> PR target/65612
> * g++.dg/ext/mv18.C: New test.
> * g++.dg/ext/mv19.C: Likewise.
> * g++.dg/ext/mv20.C: Likewise.
> * g++.dg/ext/mv21.C: Likewise.
> * g++.dg/ext/mv22.C: Likewise.
> * g++.dg/ext/mv23.C: Likewise.
>
> libgcc/
>
> PR target/65612
> * config.host (tmake_file): Add t-slibgcc-libgcc for Linux/x86.
> * config/i386/cpuinfo.c (__cpu_model): Initialize.
> (__cpu_indicator_init@GCC_4.8.0): New.
> (__cpu_model@GCC_4.8.0): Likewise.
> * config/i386/t-linux (HOST_LIBGCC2_CFLAGS): Add
> -DUSE_ELF_SYMVER.
>
> Thanks.
>
> --
> H.J.

Re: PATCH] PR target/65612: Multiversioning doesn't work with DSO nor PIE

[Mike Stump]

On Apr 17, 2015, at 1:05 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Thu, Apr 16, 2015 at 6:28 PM, Mike Stump <mikestump@comcast.net> wrote:
>> On Apr 14, 2015, at 8:07 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>>> I can confirm that the most current patch bootstraps on
>>>> x86_64-apple-darwin14 and that all of the new tests show up as
>>>> unsupported in the test suite.
>>>>         Jack
>>> 
>>> I am re-posting this patch.  OK for trunk?
>> 
>> If Jack is happy, I’m happy.  :-)  That leaves the x86 people to comment on it.
> 
> What about Solaris?

To be clear, I didn’t want to give the impression that my note was approval for the entire patch.  If I said, darwin parts ok, I think that would have been clearer...