Nabble - Sourceware - glibc

Re: Is DT_TEXTREL supposed to be supported on s390 31-bit or not?

2009-12-17T02:18:03Z

Hi Jakub,

On Wed, 16 Dec 2009 18:10:23 +0100
Jakub Jelinek <jakub@...> wrote:

> On Wed, Dec 16, 2009 at 05:43:54PM +0100, Martin Schwidefsky wrote:
> > On Wed, 16 Dec 2009 16:56:35 +0100
> > Jakub Jelinek <jakub@...> wrote:
> >
> > > While DT_TEXTREL libraries are clearly unsupported on many architectures and
> > > on some even ld fails to link them, s390 31-bit (unlike 64-bit) DT_TEXTREL
> > > libraries used to work. But apparently only when targetting old CPUs, with
> > > say -march=z900 or newer relocations like R_390_PC32DBL are created and
> > > ld.so (nor prelink) handles them.
> > > So my question is, do we just want to say that DT_TEXTREL on s390 31-bit is
> > > not supported at all, or do we want to handle the relocations that can
> > > appear in -fno-pic 31-bit libraries?
> >
> > DT_TEXTREL on 64-bit is not supported because of weak global symbols and
> > the limitation of the LARL instruction to a 4GB range.
> >
> > DT_TEXTREL on 31-bit used to work fine though and I can't see a reason
> > why it should not work on newer CPUs. With -march=z900 the compiler
> > will use LARL to get the address of global symbols, but the library will
> > be loaded below 4GB. That should work.
>
> If so, can you please figure out what relocations might be needed
> by -m31 -march=z900 (or later) code and add support for those relocs into
> sysdeps/s390/s390-32/dl-machine.h ? R_390_PC32DBL is the only one I've
> noticed during prelink checking, but that doesn't mean it isn't the only
> one.

Ok, I've reproduced the problem. The load of a non-pic library compiled
with -march=z9-109 gives the following error:

error while loading shared libraries: <lib>: unexpected reloc type 0x13

Reloc type 0x13 is indeed R_390_PC32DBL. The 32-bit variant
of dl-machine.h does not handle this relocation which is used to access
global variables for -m31 -march=z9-109. The following patches fixes
the problem for me:

2009-12-17 Martin Schwidefsky <schwidefsky@...>

* sysdeps/s390/s390-32/dl-machine.h (elf_machine_rela): Handle
R_390_PC32DBL. Remove unneeded R_390_PLT16DBL.
* sysdeps/s390/s390-32/dl-machine.h (elf_machine_rela): Remove
unneeded R_390_PLT16DBL and R_390_PLT32DBL.

--
diff -urpN glibc/sysdeps/s390/s390-32/dl-machine.h glibc-s390/sysdeps/s390/s390-32/dl-machine.h
--- glibc/sysdeps/s390/s390-32/dl-machine.h 2009-12-17 10:48:49.000000000 +0100
+++ glibc-s390/sysdeps/s390/s390-32/dl-machine.h 2009-12-17 10:47:30.000000000 +0100
@@ -389,10 +389,13 @@ elf_machine_rela (struct link_map *map,
*reloc_addr = value + reloc->r_addend - (Elf32_Addr) reloc_addr;
break;
case R_390_PC16DBL:
- case R_390_PLT16DBL:
*(unsigned short *) reloc_addr = (unsigned short)
((short) (value + reloc->r_addend - (Elf32_Addr) reloc_addr) >> 1);
break;
+ case R_390_PC32DBL:
+ *(unsigned int *) reloc_addr = (unsigned int)
+ ((int) (value + reloc->r_addend - (Elf32_Addr) reloc_addr) >> 1);
+ break;
case R_390_PC16:
*(unsigned short *) reloc_addr =
value + reloc->r_addend - (Elf32_Addr) reloc_addr;
diff -urpN glibc/sysdeps/s390/s390-64/dl-machine.h glibc-s390/sysdeps/s390/s390-64/dl-machine.h
--- glibc/sysdeps/s390/s390-64/dl-machine.h 2009-12-17 10:48:49.000000000 +0100
+++ glibc-s390/sysdeps/s390/s390-64/dl-machine.h 2009-12-17 10:47:20.000000000 +0100
@@ -371,7 +371,6 @@ elf_machine_rela (struct link_map *map,
*reloc_addr = value +reloc->r_addend - (Elf64_Addr) reloc_addr;
break;
case R_390_PC32DBL:
- case R_390_PLT32DBL:
*(unsigned int *) reloc_addr = (unsigned int)
((int) (value + reloc->r_addend - (Elf64_Addr) reloc_addr) >> 1);
break;
@@ -380,7 +379,6 @@ elf_machine_rela (struct link_map *map,
value + reloc->r_addend - (Elf64_Addr) reloc_addr;
break;
case R_390_PC16DBL:
- case R_390_PLT16DBL:
*(unsigned short *) reloc_addr = (unsigned short)
((short) (value + reloc->r_addend - (Elf64_Addr) reloc_addr) >> 1);
break;
--
blue skies,
Martin.

"Reality continues to ruin my life." - Calvin.

From forum: Sourceware - libc-alpha

[Bug libc/11099] INT_FIELD_MAYBE_NULL changed behaviour on x86_64

2009-12-17T01:20:55Z

------- Additional Comments From schwab at linux-m68k dot org 2009-12-17 09:20 -------
The CVS->git conversion is broken, use the CVS log.

[BZ #9706]
* nss/nss_files/files-parse.c (strtou32): New function.
(INT_FIELD): Use strotu32 instead of strtoul to unify behavior
across 32-bit and 64-bit platforms.
(INT_FIELD_MAYBE_NULL): Likewise.

--

http://sourceware.org/bugzilla/show_bug.cgi?id=11099

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

[Bug libc/11100] New: sysdeps/unix/sysv/linux/ptrace.c PTRACE_SETOPTIONS inappropriate pointer check

2009-12-16T14:11:17Z

In sysdeps/unix/sysv/linux/ptrace.c, on line 102, if __BOUNDED__POINTERS__ is
true, checks are done that pointers point at appropriate locations. Such a
check is done on the "data" argument to ptrace() if the "request" argument is
PTRACE_SETOPTIONS. However, according to ptrace(2) (and, indeed, according to
the gdb sources and the kernel sources), the PTRACE_SETOPTIONS request accepts
a bitmask of PTRACE_O_* values as its "data" argument, *not* a pointer.

--
Summary: sysdeps/unix/sysv/linux/ptrace.c PTRACE_SETOPTIONS
inappropriate pointer check
Product: glibc
Version: unspecified
Status: NEW
Severity: minor
Priority: P2
Component: libc
AssignedTo: drepper at redhat dot com
ReportedBy: chris2k01 at hotmail dot com
CC: glibc-bugs at sources dot redhat dot com

http://sourceware.org/bugzilla/show_bug.cgi?id=11100

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

Re: [MIPS,M68K] Update bits/poll.h

2009-12-16T10:53:56Z

"Joseph S. Myers" <joseph@...> writes:

> 2009-12-16 Joseph Myers <joseph@...>
>
> + [BZ #11093]
> + * sysdeps/unix/sysv/linux/m68k/bits/poll.h: Define POLLRDNORM,
> + POLLRDBAND, POLLWRNORM, and POLLWRBAND also for POSIX 2008.
> +

Ok.

Andreas.

--
Andreas Schwab, schwab@...
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5
"And now for something completely different."

From forum: Sourceware - libc-ports

[Bug libc/11099] New: INT_FIELD_MAYBE_NULL changed behaviour on x86_64

2009-12-16T10:41:19Z

We had password aging bugs on Mandriva and I found that when a field of
/etc/shadow contains -1, it gets 32 bits value of -1 on x86_64 even if the field
is long in the struct.

According to the doc it should indeed be empty to disable instead of containing
-1, but well the field is signed and nothing prevents from having a negative value

It used to work and the behavior change broke tests like spw->sp_* != -1 in pam
modules

The change was introduced by
http://repo.or.cz/w/glibc.git/commitdiff/1e545d0185072ac0d4646223d509c3692cbe550a with
a commit message unrelated to this part of the commit

--
Summary: INT_FIELD_MAYBE_NULL changed behaviour on x86_64
Product: glibc
Version: 2.11
Status: NEW
Severity: normal
Priority: P2
Component: libc
AssignedTo: drepper at redhat dot com
ReportedBy: pterjan at mandriva dot com
CC: glibc-bugs at sources dot redhat dot com

http://sourceware.org/bugzilla/show_bug.cgi?id=11099

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

Re: Breaking ARM's mcount implementations into a separate file

2009-12-16T10:04:41Z

On Tue, 15 Dec 2009, Thomas Schwinge wrote:

> > You'll need to send the patch in a form appropriate for committing, tested
> > as both ARM and Thumb-2.
>
> Thusly done with the following patch.

Thank you, I have committed this patch.

--
Joseph S. Myers
joseph@...

From forum: Sourceware - libc-ports

[MIPS,M68K] Update bits/poll.h

2009-12-16T09:50:02Z

This patch updates ports bits/poll.h headers in line with a recent libc
change. I've committed the MIPS patch; OK to commit the M68K one?

diff --git a/ChangeLog.m68k b/ChangeLog.m68k
index 8411570..978cd1a 100644
--- a/ChangeLog.m68k
+++ b/ChangeLog.m68k
@@ -1,5 +1,11 @@
2009-12-16 Joseph Myers <joseph@...>

+ [BZ #11093]
+ * sysdeps/unix/sysv/linux/m68k/bits/poll.h: Define POLLRDNORM,
+ POLLRDBAND, POLLWRNORM, and POLLWRBAND also for POSIX 2008.
+
+2009-12-16 Joseph Myers <joseph@...>
+
* sysdeps/unix/sysv/linux/m68k/bits/fcntl.h: Redefine O_SYNC and
O_DSYNC to match 2.6.33+ kernels.

diff --git a/ChangeLog.mips b/ChangeLog.mips
index 0f8f7c8..68ba90d 100644
--- a/ChangeLog.mips
+++ b/ChangeLog.mips
@@ -1,5 +1,11 @@
2009-12-16 Joseph Myers <joseph@...>

+ [BZ #11093]
+ * sysdeps/unix/sysv/linux/mips/bits/poll.h: Define POLLRDNORM,
+ POLLRDBAND, POLLWRNORM, and POLLWRBAND also for POSIX 2008.
+
+2009-12-16 Joseph Myers <joseph@...>
+
* sysdeps/mips/dl-lookup.c: Update from generic version.

2009-12-15 Joseph Myers <joseph@...>
diff --git a/sysdeps/unix/sysv/linux/m68k/bits/poll.h b/sysdeps/unix/sysv/linux/m68k/bits/poll.h
index bc28579..84219ba 100644
--- a/sysdeps/unix/sysv/linux/m68k/bits/poll.h
+++ b/sysdeps/unix/sysv/linux/m68k/bits/poll.h
@@ -1,4 +1,4 @@
-/* Copyright (C) 1997, 2001, 2008 Free Software Foundation, Inc.
+/* Copyright (C) 1997, 2001, 2008, 2009 Free Software Foundation, Inc.
This file is part of the GNU C Library.

The GNU C Library is free software; you can redistribute it and/or
@@ -27,7 +27,7 @@
#define POLLPRI 0x002 /* There is urgent data to read. */
#define POLLOUT 0x004 /* Writing now will not block. */

-#ifdef __USE_XOPEN
+#if defined __USE_XOPEN || defined __USE_XOPEN2K8
/* These values are defined in XPG4.2. */
# define POLLRDNORM 0x040 /* Normal data may be read. */
# define POLLRDBAND 0x080 /* Priority data may be read. */
diff --git a/sysdeps/unix/sysv/linux/mips/bits/poll.h b/sysdeps/unix/sysv/linux/mips/bits/poll.h
index eee4ea2..8231239 100644
--- a/sysdeps/unix/sysv/linux/mips/bits/poll.h
+++ b/sysdeps/unix/sysv/linux/mips/bits/poll.h
@@ -1,4 +1,4 @@
-/* Copyright (C) 1997, 2001, 2006 Free Software Foundation, Inc.
+/* Copyright (C) 1997, 2001, 2006, 2009 Free Software Foundation, Inc.
This file is part of the GNU C Library.

The GNU C Library is free software; you can redistribute it and/or
@@ -27,7 +27,7 @@
#define POLLPRI 0x002 /* There is urgent data to read. */
#define POLLOUT 0x004 /* Writing now will not block. */

-#ifdef __USE_XOPEN
+#if defined __USE_XOPEN || defined __USE_XOPEN2K8
/* These values are defined in XPG4.2. */
# define POLLRDNORM 0x040 /* Normal data may be read. */
# define POLLRDBAND 0x080 /* Priority data may be read. */

--
Joseph S. Myers
joseph@...

From forum: Sourceware - libc-ports

Another MIPS dl-lookup.c update

2009-12-16T09:44:54Z

Another change has gone into the libc dl-lookup.c, so I've applied the
patch also to the MIPS version.

diff --git a/ChangeLog.mips b/ChangeLog.mips
index 9dfe2a5..0f8f7c8 100644
--- a/ChangeLog.mips
+++ b/ChangeLog.mips
@@ -1,3 +1,7 @@
+2009-12-16 Joseph Myers <joseph@...>
+
+ * sysdeps/mips/dl-lookup.c: Update from generic version.
+
2009-12-15 Joseph Myers <joseph@...>

* sysdeps/unix/sysv/linux/mips/bits/fcntl.h: Redefine O_SYNC and
diff --git a/sysdeps/mips/dl-lookup.c b/sysdeps/mips/dl-lookup.c
index c7deb4d..a9fff59 100644
--- a/sysdeps/mips/dl-lookup.c
+++ b/sysdeps/mips/dl-lookup.c
@@ -794,7 +794,7 @@ _dl_lookup_symbol_x (const char *undef_name, struct link_map *undef_map,
if (__builtin_expect (protected != 0, 0))
{
/* It is very tricky. We need to figure out what value to
- return for the protected symbol. */
+ return for the protected symbol. */
if (type_class == ELF_RTYPE_CLASS_PLT)
{
if (current_value.s != NULL && current_value.m != undef_map)
@@ -839,7 +839,8 @@ _dl_lookup_symbol_x (const char *undef_name, struct link_map *undef_map,
version, type_class, flags, skip_map);

/* The object is used. */
- current_value.m->l_used = 1;
+ if (__builtin_expect (current_value.m->l_used == 0, 0))
+ current_value.m->l_used = 1;

if (__builtin_expect (GLRO(dl_debug_mask)
& (DL_DEBUG_BINDINGS|DL_DEBUG_PRELINK), 0))
@@ -861,7 +862,7 @@ _dl_setup_hash (struct link_map *map)
Elf_Symndx nchain;

if (__builtin_expect (map->l_info[DT_ADDRTAGIDX (DT_GNU_HASH) + DT_NUM
- + DT_THISPROCNUM + DT_VERSIONTAGNUM
+ + DT_THISPROCNUM + DT_VERSIONTAGNUM
+ DT_EXTRANUM + DT_VALNUM] != NULL, 1))
{
Elf32_Word *hash32

--
Joseph S. Myers
joseph@...

From forum: Sourceware - libc-ports

Re: Is DT_TEXTREL supposed to be supported on s390 31-bit or not?

2009-12-16T09:10:23Z

On Wed, Dec 16, 2009 at 05:43:54PM +0100, Martin Schwidefsky wrote:

> On Wed, 16 Dec 2009 16:56:35 +0100
> Jakub Jelinek <jakub@...> wrote:
>
> > While DT_TEXTREL libraries are clearly unsupported on many architectures and
> > on some even ld fails to link them, s390 31-bit (unlike 64-bit) DT_TEXTREL
> > libraries used to work. But apparently only when targetting old CPUs, with
> > say -march=z900 or newer relocations like R_390_PC32DBL are created and
> > ld.so (nor prelink) handles them.
> > So my question is, do we just want to say that DT_TEXTREL on s390 31-bit is
> > not supported at all, or do we want to handle the relocations that can
> > appear in -fno-pic 31-bit libraries?
>
> DT_TEXTREL on 64-bit is not supported because of weak global symbols and
> the limitation of the LARL instruction to a 4GB range.
>
> DT_TEXTREL on 31-bit used to work fine though and I can't see a reason
> why it should not work on newer CPUs. With -march=z900 the compiler
> will use LARL to get the address of global symbols, but the library will
> be loaded below 4GB. That should work.

If so, can you please figure out what relocations might be needed
by -m31 -march=z900 (or later) code and add support for those relocs into
sysdeps/s390/s390-32/dl-machine.h ? R_390_PC32DBL is the only one I've
noticed during prelink checking, but that doesn't mean it isn't the only
one.

Jakub

From forum: Sourceware - libc-alpha

Re: SH, S390: (not) calling __gmon_start__ from pt-initfini.c

2009-12-16T08:46:19Z

On 12/16/2009 08:34 AM, Martin Schwidefsky wrote:
> Ok, from looking at the code it seems that the call to __gmon_start__
> from libpthread is redundant. Andreas checked and tested the change,
> that is good enough for me. Ok from my side.

Somebody will have to submit a new patch. I'm not trying to second
guess from the various messages what has to be done.

--
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖

From forum: Sourceware - libc-alpha

Re: Is DT_TEXTREL supposed to be supported on s390 31-bit or not?

2009-12-16T08:43:54Z

Hi Jakub,

On Wed, 16 Dec 2009 16:56:35 +0100
Jakub Jelinek <jakub@...> wrote:

> While DT_TEXTREL libraries are clearly unsupported on many architectures and
> on some even ld fails to link them, s390 31-bit (unlike 64-bit) DT_TEXTREL
> libraries used to work. But apparently only when targetting old CPUs, with
> say -march=z900 or newer relocations like R_390_PC32DBL are created and
> ld.so (nor prelink) handles them.
> So my question is, do we just want to say that DT_TEXTREL on s390 31-bit is
> not supported at all, or do we want to handle the relocations that can
> appear in -fno-pic 31-bit libraries?

DT_TEXTREL on 64-bit is not supported because of weak global symbols and
the limitation of the LARL instruction to a 4GB range.

DT_TEXTREL on 31-bit used to work fine though and I can't see a reason
why it should not work on newer CPUs. With -march=z900 the compiler
will use LARL to get the address of global symbols, but the library will
be loaded below 4GB. That should work.

--
blue skies,
Martin.

"Reality continues to ruin my life." - Calvin.

From forum: Sourceware - libc-alpha

Re: SH, S390: (not) calling __gmon_start__ from pt-initfini.c

2009-12-16T08:34:31Z

On Wed, 09 Dec 2009 15:26:43 +0100
Andreas Krebbel <krebbel@...> wrote:

> Hi Thomas,
>
> the S/390 part looks ok to me but I think the ".L22" label should go away as well.
>
> I've run the libc testsuite on s390 and s390x with your patch without seeing
> regressions.
>
> However I cannot approve your patch. Let's wait for comments from Martin.

Ok, from looking at the code it seems that the call to __gmon_start__
from libpthread is redundant. Andreas checked and tested the change,
that is good enough for me. Ok from my side.

--
blue skies,
Martin.

"Reality continues to ruin my life." - Calvin.

From forum: Sourceware - libc-alpha

Is DT_TEXTREL supposed to be supported on s390 31-bit or not?

2009-12-16T07:56:35Z

Hi!

While DT_TEXTREL libraries are clearly unsupported on many architectures and
on some even ld fails to link them, s390 31-bit (unlike 64-bit) DT_TEXTREL
libraries used to work. But apparently only when targetting old CPUs, with
say -march=z900 or newer relocations like R_390_PC32DBL are created and
ld.so (nor prelink) handles them.
So my question is, do we just want to say that DT_TEXTREL on s390 31-bit is
not supported at all, or do we want to handle the relocations that can
appear in -fno-pic 31-bit libraries?

Jakub

From forum: Sourceware - libc-alpha

Re: [ARM,MIPS,M68K] O_SYNC fcntl.h updates

2009-12-15T15:47:58Z

"Joseph S. Myers" <joseph@...> writes:

> +2009-12-15 Joseph Myers <joseph@...>
> +
> + * sysdeps/unix/sysv/linux/m68k/bits/fcntl.h: Redefine O_SYNC and
> + O_DSYNC to match 2.6.33+ kernels.
> +

Ok.

Andreas.

--
Andreas Schwab, schwab@...
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5
"And now for something completely different."

From forum: Sourceware - libc-ports

Re: Breaking ARM's mcount implementations into a separate file

2009-12-15T13:54:43Z

Hello!

On 2009-12-15 13:55, Joseph S. Myers wrote:

> On Tue, 15 Dec 2009, Thomas Schwinge wrote:
>> On 2009-12-15 02:05, Joseph S. Myers wrote:
>> > I'm not sure of the significance of having separate ARM/Thumb versions,
>>
>> Well, if I understand this correctly, it was you, together with Paul
>> Brook, who have added this code on 2009-02-05.
>
> You need to look more carefully at the history before I merged Thumb-2
> support from EGLIBC to FSF ports. The new mcount entry point dates back
> at least to <http://www.eglibc.org/archives/patches/msg00449.html> and
> other discussions around that time, while Thumb-2 mcount comes from
> <http://www.eglibc.org/archives/patches/msg00143.html>. I.e., on
> 2009-02-05 I merged existing code without knowing the significance of
> every bit of it.

Thanks for the clarification. I had simply been looking at ChangeLog.arm
which contains this:

2009-02-05 Paul Brook <paul@...>
Joseph Myers <joseph@...>

[...]
* sysdeps/arm/machine-gmon.h (MCOUNT): Add Thumb-2 implementation.
[...]

..., so I supposed that you had been directly involved with that.

> You'll need to send the patch in a form appropriate for committing, tested
> as both ARM and Thumb-2.

Thusly done with the following patch.

I folded the two _mcount definitions into one -- but that one now
consists of nearly an equal amount of CPP conditionals compared to
assembly instructions, so if you'd rather have a bit of assembly source
code duplication, but better readability (?), then please tell and I'll
modify accordingly.

2009-12-15 Thomas Schwinge <thomas@...>

* sysdeps/arm/eabi/Makefile [gmon] (sysdep_routines): Add arm-mcount.
* sysdeps/arm/eabi/machine-gmon.h (mcount_internal): Define to
`__mcount_internal'.
(MCOUNT (__gnu_mcount_nc), OLD_MCOUNT (_mcount), mcount): Move to...
* sysdeps/arm/eabi/arm-mcount.S: ... this new file as __gnu_mcount_nc,
_mcount, and mcount, respectively.

diff --git a/glibc-ports-mainline/sysdeps/arm/eabi/Makefile b/glibc-ports-mainline/sysdeps/arm/eabi/Makefile
index 36ddb8e..73185cb 100644
--- a/glibc-ports-mainline/sysdeps/arm/eabi/Makefile
+++ b/glibc-ports-mainline/sysdeps/arm/eabi/Makefile
@@ -15,6 +15,10 @@ static-only-routines += $(aeabi_constants)
gen-as-const-headers += rtld-global-offsets.sym
endif

+ifeq ($(subdir),gmon)
+sysdep_routines += arm-mcount
+endif
+
ifeq ($(subdir),debug)
CFLAGS-backtrace.c += -funwind-tables
CFLAGS-tst-backtrace2.c += -funwind-tables
diff --git a/glibc-ports-mainline/sysdeps/arm/eabi/arm-mcount.S b/glibc-ports-mainline/sysdeps/arm/eabi/arm-mcount.S
new file mode 100644
index 0000000..2aa50b7
--- /dev/null
+++ b/glibc-ports-mainline/sysdeps/arm/eabi/arm-mcount.S
@@ -0,0 +1,83 @@
+/* Implementation of profiling support. ARM EABI version.
+ Copyright (C) 2008, 2009 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, write to the Free
+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+ 02111-1307 USA. */
+
+/* Don't call mcount when calling mcount... */
+#undef PROF
+
+#include <sysdep.h>
+
+
+#ifdef __thumb2__
+ .thumb
+ .syntax unified
+#endif
+
+
+/* Use an assembly stub with a special ABI. The calling lr has been
+ pushed to the stack (which will be misaligned). We should preserve
+ all registers except ip and pop a word off the stack.
+
+ NOTE: This assumes mcount_internal does not clobber any non-core
+ (coprocessor) registers. Currently this is true, but may require
+ additional attention in the future.
+
+ The calling sequence looks something like:
+func:
+ push {lr}
+ bl __gnu_mount_nc
+ <function body>
+*/
+
+ENTRY(__gnu_mcount_nc)
+ push {r0, r1, r2, r3, lr}
+ bic r1, lr, #1
+ ldr r0, [sp, #20]
+ bl __mcount_internal
+ pop {r0, r1, r2, r3, ip, lr}
+ bx ip
+END(__gnu_mcount_nc)
+
+
+/* Provide old mcount for backwards compatibility. This requires
+ code be compiled with APCS frame pointers. */
+
+ENTRY(_mcount)
+ stmdb sp!, {r0, r1, r2, r3, fp, lr}
+#ifdef __thumb2__
+ movs r0, fp
+ ittt ne
+ ldrne r0, [r0, #-4]
+#else
+ movs fp, fp
+ ldrne r0, [fp, #-4]
+#endif
+ movnes r1, lr
+ blne __mcount_internal
+#ifdef __thumb2__
+ ldmia sp!, {r0, r1, r2, r3, fp, pc}
+#else
+ ldmia sp!, {r0, r1, r2, r3, fp, lr}
+ bx lr
+#endif
+END(_mcount)
+
+/* The canonical name for the function is `_mcount' in both C and asm,
+ but some old asm code might assume it's `mcount'. */
+#undef mcount
+weak_alias (_mcount, mcount)
diff --git a/glibc-ports-mainline/sysdeps/arm/eabi/machine-gmon.h b/glibc-ports-mainline/sysdeps/arm/eabi/machine-gmon.h
index 189a9a3..aca23c9 100644
--- a/glibc-ports-mainline/sysdeps/arm/eabi/machine-gmon.h
+++ b/glibc-ports-mainline/sysdeps/arm/eabi/machine-gmon.h
@@ -1,5 +1,5 @@
/* Machine-dependent definitions for profiling support. ARM EABI version.
- Copyright (C) 2008 Free Software Foundation, Inc.
+ Copyright (C) 2008, 2009 Free Software Foundation, Inc.
This file is part of the GNU C Library.

The GNU C Library is free software; you can redistribute it and/or
@@ -17,83 +17,18 @@
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */

+#include <sysdep.h>
+
/* GCC for the ARM cannot compile __builtin_return_address(N) for N != 0,
so we must use an assembly stub. */

-#include <sysdep.h>
-static void mcount_internal (u_long frompc, u_long selfpc) __attribute_used__;
+/* We must not pollute the global namespace. */
+#define mcount_internal __mcount_internal

+extern void mcount_internal (u_long frompc, u_long selfpc) internal_function;
#define _MCOUNT_DECL(frompc, selfpc) \
-static void mcount_internal (u_long frompc, u_long selfpc)
-
-/* Use an assembly stub with a special ABI. The calling lr has been
- pushed to the stack (which will be misaligned). We should preserve
- all registers except ip and pop a word off the stack.
-
- NOTE: This assumes mcount_internal does not clobber any non-core
- (coprocessor) registers. Currently this is true, but may require
- additional attention in the future.
-
- The calling sequence looks something like:
-func:
- push {lr}
- bl __gnu_mount_nc
- <function body>
- */
-
-
-#define MCOUNT \
-void __attribute__((__naked__)) __gnu_mcount_nc(void) \
-{ \
- asm ("push {r0, r1, r2, r3, lr}\n\t" \
- "bic r1, lr, #1\n\t" \
- "ldr r0, [sp, #20]\n\t" \
- "bl mcount_internal\n\t" \
- "pop {r0, r1, r2, r3, ip, lr}\n\t" \
- "bx ip"); \
-} \
-OLD_MCOUNT
-
-/* Provide old mcount for backwards compatibility. This requires
- code be compiled with APCS frame pointers. */
-
-#ifndef NO_UNDERSCORES
-/* The asm symbols for C functions are `_function'.
- The canonical name for the counter function is `mcount', no _. */
-void _mcount (void) asm ("mcount");
-#else
-/* The canonical name for the function is `_mcount' in both C and asm,
- but some old asm code might assume it's `mcount'. */
-void _mcount (void);
-weak_alias (_mcount, mcount)
-#endif
-
-#ifdef __thumb2__
-
-#define OLD_MCOUNT \
-void __attribute__((__naked__)) _mcount (void) \
-{ \
- __asm__("push {r0, r1, r2, r3, fp, lr};" \
- "movs r0, fp;" \
- "ittt ne;" \
- "ldrne r0, [r0, #-4];" \
- "movsne r1, lr;" \
- "blne mcount_internal;" \
- "pop {r0, r1, r2, r3, fp, pc}"); \
-}
-
-#else
+ void internal_function mcount_internal (u_long frompc, u_long selfpc)

-#define OLD_MCOUNT \
-void __attribute__((__naked__)) _mcount (void) \
-{ \
- __asm__("stmdb sp!, {r0, r1, r2, r3, fp, lr};" \
- "movs fp, fp;" \
- "ldrne r0, [fp, #-4];" \
- "movnes r1, lr;" \
- "blne mcount_internal;" \
- "ldmia sp!, {r0, r1, r2, r3, fp, lr};" \
- "bx lr"); \
-}

-#endif
+/* Define MCOUNT as empty since we have the implementation in another file. */
+#define MCOUNT

Regards,
Thomas

attachment0 (202 bytes) Download Attachment

From forum: Sourceware - libc-ports

[Bug libc/11093] does not define POLL[RD|WR]* macros with _POSIX_C_SOURCE="200809L"

2009-12-15T13:05:41Z

------- Additional Comments From drepper at redhat dot com 2009-12-15 21:05 -------
Fixed in git.

--
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution| |FIXED

http://sourceware.org/bugzilla/show_bug.cgi?id=11093

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

[Bug libc/11093] New: does not define POLL[RD|WR]* macros with _POSIX_C_SOURCE="200809L"

2009-12-15T12:42:36Z

As of POSIX.1-2008, the <poll.h> header was moved completely to the base
specification [1]. Therefore, a definition of _POSIX_C_SOURCE to 200809L alone
should be enough to get all definitions in <poll.h> specified by POSIX.1-2008.
Currently, however, the POLLRDNORM, POLLRDBAND, POLLWRNORM and POLLWRBAND depend
on _XOPEN_SOURCE.

Test program:

#define _POSIX_C_SOURCE 200809L

#include <poll.h>

int main (void)
{
struct pollfd fds[] = {
{
.fd = 0,
.events = POLLIN | POLLRDNORM | POLLRDBAND
},
{
.fd = 1,
.events = POLLOUT | POLLWRNORM | POLLWRBAND
}
};

(void)fds;
}

[1] http://www.opengroup.org/onlinepubs/9699919799/basedefs/poll.h.html

--
Summary: <poll.h> does not define POLL[RD|WR]* macros with
_POSIX_C_SOURCE="200809L"
Product: glibc
Version: 2.9
Status: NEW
Severity: normal
Priority: P2
Component: libc
AssignedTo: drepper at redhat dot com
ReportedBy: andrey dot vihrov at gmail dot com
CC: glibc-bugs at sources dot redhat dot com
GCC build triplet: x86_64-pc-linux-gnu
GCC host triplet: x86_64-pc-linux-gnu
GCC target triplet: x86_64-pc-linux-gnu

http://sourceware.org/bugzilla/show_bug.cgi?id=11093

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

[ARM,MIPS,M68K] O_SYNC fcntl.h updates

2009-12-15T10:16:32Z

This patch updates O_SYNC in bits/fcntl.h headers in line with recent libc
and kernel changes. (The previously used value had only O_DSYNC semantics
and keeps those for binary compatibility, so is kept as the value of
O_DSYNC, but O_SYNC gets a new value which has full O_SYNC semantics in
new kernels.)

I've committed the ARM and MIPS changes; OK to commit the M68K ones?
Other ports maintainers will also need to update their ports; HPPA will
need special care in view of the "Note that parisc really screwed up their
headers" in the commit message for the relevant kernel patch.

diff --git a/ChangeLog.arm b/ChangeLog.arm
index c20b095..03bfd34 100644
--- a/ChangeLog.arm
+++ b/ChangeLog.arm
@@ -1,3 +1,8 @@
+2009-12-15 Joseph Myers <joseph@...>
+
+ * sysdeps/unix/sysv/linux/arm/bits/fcntl.h: Redefine O_SYNC and
+ O_DSYNC to match 2.6.33+ kernels.
+
2009-11-19 Joseph Myers <joseph@...>

* sysdeps/unix/sysv/linux/arm/bits/fcntl.h: Introduce new symbol
diff --git a/ChangeLog.m68k b/ChangeLog.m68k
index 42d6a91..b1eb009 100644
--- a/ChangeLog.m68k
+++ b/ChangeLog.m68k
@@ -1,3 +1,8 @@
+2009-12-15 Joseph Myers <joseph@...>
+
+ * sysdeps/unix/sysv/linux/m68k/bits/fcntl.h: Redefine O_SYNC and
+ O_DSYNC to match 2.6.33+ kernels.
+
2009-11-23 Andreas Schwab <schwab@...>

* sysdeps/unix/sysv/linux/m68k/bits/stat.h: Use struct timespec
diff --git a/ChangeLog.mips b/ChangeLog.mips
index b3125f4..9dfe2a5 100644
--- a/ChangeLog.mips
+++ b/ChangeLog.mips
@@ -1,3 +1,8 @@
+2009-12-15 Joseph Myers <joseph@...>
+
+ * sysdeps/unix/sysv/linux/mips/bits/fcntl.h: Redefine O_SYNC and
+ O_DSYNC to match 2.6.33+ kernels.
+
2009-12-09 Joseph Myers <joseph@...>

* sysdeps/unix/sysv/linux/mips/bits/socket.h: Define mmsghdr and
diff --git a/sysdeps/unix/sysv/linux/arm/bits/fcntl.h b/sysdeps/unix/sysv/linux/arm/bits/fcntl.h
index d597c10..4971473 100644
--- a/sysdeps/unix/sysv/linux/arm/bits/fcntl.h
+++ b/sysdeps/unix/sysv/linux/arm/bits/fcntl.h
@@ -41,7 +41,7 @@
#define O_APPEND 02000
#define O_NONBLOCK 04000
#define O_NDELAY O_NONBLOCK
-#define O_SYNC 010000
+#define O_SYNC 04010000
#define O_FSYNC O_SYNC
#define O_ASYNC 020000

@@ -57,7 +57,7 @@
We define the symbols here but let them do the same as O_SYNC since
this is a superset. */
#if defined __USE_POSIX199309 || defined __USE_UNIX98
-# define O_DSYNC O_SYNC /* Synchronize data. */
+# define O_DSYNC 010000 /* Synchronize data. */
# define O_RSYNC O_SYNC /* Synchronize read operations. */
#endif

diff --git a/sysdeps/unix/sysv/linux/m68k/bits/fcntl.h b/sysdeps/unix/sysv/linux/m68k/bits/fcntl.h
index 8ed8eb6..1f98cf5 100644
--- a/sysdeps/unix/sysv/linux/m68k/bits/fcntl.h
+++ b/sysdeps/unix/sysv/linux/m68k/bits/fcntl.h
@@ -40,7 +40,7 @@
#define O_APPEND 02000
#define O_NONBLOCK 04000
#define O_NDELAY O_NONBLOCK
-#define O_SYNC 010000
+#define O_SYNC 04010000
#define O_FSYNC O_SYNC
#define O_ASYNC 020000

@@ -56,7 +56,7 @@
We define the symbols here but let them do the same as O_SYNC since
this is a superset. */
#if defined __USE_POSIX199309 || defined __USE_UNIX98
-# define O_DSYNC O_SYNC /* Synchronize data. */
+# define O_DSYNC 010000 /* Synchronize data. */
# define O_RSYNC O_SYNC /* Synchronize read operations. */
#endif

diff --git a/sysdeps/unix/sysv/linux/mips/bits/fcntl.h b/sysdeps/unix/sysv/linux/mips/bits/fcntl.h
index 5496912..24fb974 100644
--- a/sysdeps/unix/sysv/linux/mips/bits/fcntl.h
+++ b/sysdeps/unix/sysv/linux/mips/bits/fcntl.h
@@ -36,7 +36,7 @@
#define O_WRONLY 01
#define O_RDWR 02
#define O_APPEND 0x0008
-#define O_SYNC 0x0010
+#define O_SYNC 0x4010
#define O_NONBLOCK 0x0080
#define O_NDELAY O_NONBLOCK
#define O_CREAT 0x0100 /* not fcntl */
@@ -58,7 +58,7 @@
We define the symbols here but let them do the same as O_SYNC since
this is a superset. */
#if defined __USE_POSIX199309 || defined __USE_UNIX98
-# define O_DSYNC O_SYNC /* Synchronize data. */
+# define O_DSYNC 0x0010 /* Synchronize data. */
# define O_RSYNC O_SYNC /* Synchronize read operations. */
#endif

--
Joseph S. Myers
joseph@...

From forum: Sourceware - libc-ports

[PATCH] Additional setcontext(), etc. POSIX conformance tests

2009-12-15T09:57:32Z

Hello,

While developing the implementation of the setcontext(), etc. API for the
MIPS platform earlier this year, I used these additional tests to verify
POSIX conformance (cf.
http://www.opengroup.org/onlinepubs/009695399/functions/setcontext.html).
Their purpose is as follows:

- tst-makecontext3.c -- verify that for ABIs that use a mixed
register/stack argument passing convention arguments are placed in their
designated registers and stack slots correctly -- the existing tests
would not hit the argument register limit for some ABIs, such as MIPS
n64 and n32 ABIs that define eight argument registers,

- tst-sigcontext.c -- verify that a user context obtained from a signal
handler can be used with setcontext(), etc.

I've thought it might be worth sharing these tests with the community so
that developers can readily verify the respective implementations for
platforms they are interested in and do not have to reinvent the wheel.

2009-12-15 Maciej W. Rozycki <macro@...>

* stdlib/tst-makecontext3.c: New test.
* stdlib/tst-sigcontext.c: Likewise.
* stdlib/Makefile (tests): Add them.

* stdlib/tst-setcontext.c (check_called): Fix a typo.

Please apply.

Maciej

glibc-2.11.90-20091215-tst-ucontext-3.patch
diff --git a/stdlib/Makefile b/stdlib/Makefile
index d361dd8..882302f 100644
--- a/stdlib/Makefile
+++ b/stdlib/Makefile
@@ -70,7 +70,8 @@ tests := tst-strtol tst-strtod testmb testrand testsort testdiv \
test-a64l tst-qsort tst-system testmb2 bug-strtod2 \
tst-atof1 tst-atof2 tst-strtod2 tst-strtod3 tst-rand48-2 \
tst-makecontext tst-strtod4 tst-strtod5 tst-qsort2 \
- tst-makecontext2 tst-strtod6 tst-unsetenv1
+ tst-makecontext2 tst-strtod6 tst-unsetenv1 \
+ tst-makecontext3 tst-sigcontext

include ../Makeconfig

diff --git a/stdlib/tst-makecontext3.c b/stdlib/tst-makecontext3.c
new file mode 100644
index 0000000..f127c6a
--- /dev/null
+++ b/stdlib/tst-makecontext3.c
@@ -0,0 +1,217 @@
+/* Copyright (C) 2001,2002,2004,2006,2009 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, write to the Free
+ Software Foundation, 51 Franklin Street - Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <ucontext.h>
+#include <unistd.h>
+
+static ucontext_t ctx[3];
+
+static int was_in_f1;
+static int was_in_f2;
+
+static char st2[32768];
+
+static volatile int flag;
+
+static void
+f1 (int a00, int a01, int a02, int a03, int a04, int a05, int a06, int a07,
+ int a08, int a09, int a10, int a11, int a12, int a13, int a14, int a15,
+ int a16, int a17, int a18, int a19, int a20, int a21, int a22, int a23,
+ int a24, int a25, int a26, int a27, int a28, int a29, int a30, int a31,
+ int a32)
+{
+ printf ("start f1(a00=%08x,a01=%08x,a02=%08x,a03=%08x,\n"
+ " a04=%08x,a05=%08x,a06=%08x,a07=%08x,\n"
+ " a08=%08x,a09=%08x,a10=%08x,a11=%08x,\n"
+ " a12=%08x,a13=%08x,a14=%08x,a15=%08x,\n"
+ " a16=%08x,a17=%08x,a18=%08x,a19=%08x,\n"
+ " a20=%08x,a21=%08x,a22=%08x,a23=%08x,\n"
+ " a24=%08x,a25=%08x,a26=%08x,a27=%08x,\n"
+ " a28=%08x,a29=%08x,a30=%08x,a31=%08x,\n"
+ " a32=%08x) [%d]\n",
+ a00, a01, a02, a03, a04, a05, a06, a07,
+ a08, a09, a10, a11, a12, a13, a14, a15,
+ a16, a17, a18, a19, a20, a21, a22, a23,
+ a24, a25, a26, a27, a28, a29, a30, a31,
+ a32, flag);
+
+ if (a00 != (0x00000001 << flag) || a01 != (0x00000004 << flag)
+ || a02 != (0x00000012 << flag) || a03 != (0x00000048 << flag)
+ || a04 != (0x00000123 << flag) || a05 != (0x0000048d << flag)
+ || a06 != (0x00001234 << flag) || a07 != (0x000048d1 << flag)
+ || a08 != (0x00012345 << flag) || a09 != (0x00048d15 << flag)
+ || a10 != (0x00123456 << flag) || a11 != (0x0048d159 << flag)
+ || a12 != (0x01234567 << flag) || a13 != (0x048d159e << flag)
+ || a14 != (0x12345678 << flag) || a15 != (0x48d159e2 << flag)
+ || a16 != (0x23456789 << flag) || a17 != (0x8d159e26 << flag)
+ || a18 != (0x3456789a << flag) || a19 != (0xd159e26a << flag)
+ || a20 != (0x456789ab << flag) || a21 != (0x159e26af << flag)
+ || a22 != (0x56789abc << flag) || a23 != (0x59e26af3 << flag)
+ || a24 != (0x6789abcd << flag) || a25 != (0x9e26af37 << flag)
+ || a26 != (0x789abcde << flag) || a27 != (0xe26af37b << flag)
+ || a28 != (0x89abcdef << flag) || a29 != (0x26af37bc << flag)
+ || a30 != (0x9abcdef0 << flag) || a31 != (0x6af37bc3 << flag)
+ || a32 != (0xabcdef0f << flag))
+ {
+ puts ("arg mismatch");
+ exit (-1);
+ }
+
+ if (flag && swapcontext (&ctx[1], &ctx[2]) != 0)
+ {
+ printf ("%s: swapcontext: %m\n", __FUNCTION__);
+ exit (1);
+ }
+ printf ("finish f1 [%d]\n", flag);
+ flag++;
+ was_in_f1++;
+}
+
+static void
+f2 (void)
+{
+ puts ("start f2");
+ if (swapcontext (&ctx[2], &ctx[1]) != 0)
+ {
+ printf ("%s: swapcontext: %m\n", __FUNCTION__);
+ exit (1);
+ }
+ puts ("finish f2");
+ was_in_f2 = 1;
+}
+
+volatile int global;
+
+
+static int back_in_main;
+
+
+static void
+check_called (void)
+{
+ if (back_in_main == 0)
+ {
+ puts ("program did not reach main again");
+ _exit (1);
+ }
+}
+
+
+int
+main (void)
+{
+ atexit (check_called);
+
+ char st1[32768];
+
+ puts ("making contexts");
+ if (getcontext (&ctx[0]) != 0)
+ {
+ if (errno == ENOSYS)
+ {
+ back_in_main = 1;
+ exit (0);
+ }
+
+ printf ("%s: getcontext: %m\n", __FUNCTION__);
+ exit (1);
+ }
+
+ ctx[1] = ctx[0];
+ ctx[1].uc_stack.ss_sp = st1;
+ ctx[1].uc_stack.ss_size = sizeof st1;
+ ctx[1].uc_link = &ctx[0];
+ {
+ ucontext_t tempctx = ctx[1];
+ makecontext (&ctx[1], (void (*) (void)) f1, 33,
+ 0x00000001 << flag, 0x00000004 << flag,
+ 0x00000012 << flag, 0x00000048 << flag,
+ 0x00000123 << flag, 0x0000048d << flag,
+ 0x00001234 << flag, 0x000048d1 << flag,
+ 0x00012345 << flag, 0x00048d15 << flag,
+ 0x00123456 << flag, 0x0048d159 << flag,
+ 0x01234567 << flag, 0x048d159e << flag,
+ 0x12345678 << flag, 0x48d159e2 << flag,
+ 0x23456789 << flag, 0x8d159e26 << flag,
+ 0x3456789a << flag, 0xd159e26a << flag,
+ 0x456789ab << flag, 0x159e26af << flag,
+ 0x56789abc << flag, 0x59e26af3 << flag,
+ 0x6789abcd << flag, 0x9e26af37 << flag,
+ 0x789abcde << flag, 0xe26af37b << flag,
+ 0x89abcdef << flag, 0x26af37bc << flag,
+ 0x9abcdef0 << flag, 0x6af37bc3 << flag,
+ 0xabcdef0f << flag);
+
+ /* Without this check, a stub makecontext can make us spin forever. */
+ if (memcmp (&tempctx, &ctx[1], sizeof ctx[1]) == 0)
+ {
+ puts ("makecontext was a no-op, presuming not implemented");
+ return 0;
+ }
+ }
+
+ /* Play some tricks with this context. */
+ if (++global == 1)
+ if (setcontext (&ctx[1]) != 0)
+ {
+ printf ("%s: setcontext: %m\n", __FUNCTION__);
+ exit (1);
+ }
+ if (global != 2)
+ {
+ printf ("%s: 'global' not incremented twice\n", __FUNCTION__);
+ exit (1);
+ }
+
+ if (getcontext (&ctx[2]) != 0)
+ {
+ printf ("%s: second getcontext: %m\n", __FUNCTION__);
+ exit (1);
+ }
+ ctx[2].uc_stack.ss_sp = st2;
+ ctx[2].uc_stack.ss_size = sizeof st2;
+ ctx[2].uc_link = &ctx[1];
+ makecontext (&ctx[2], f2, 0);
+
+ puts ("swapping contexts");
+ if (swapcontext (&ctx[0], &ctx[2]) != 0)
+ {
+ printf ("%s: swapcontext: %m\n", __FUNCTION__);
+ exit (1);
+ }
+ puts ("back at main program");
+ back_in_main = 1;
+
+ if (was_in_f1 < 2)
+ {
+ puts ("didn't reach f1 twice");
+ exit (1);
+ }
+ if (was_in_f2 == 0)
+ {
+ puts ("didn't reach f2");
+ exit (1);
+ }
+
+ puts ("test succeeded");
+ return 0;
+}
diff --git a/stdlib/tst-setcontext.c b/stdlib/tst-setcontext.c
index 51296f7..9a6d79b 100644
--- a/stdlib/tst-setcontext.c
+++ b/stdlib/tst-setcontext.c
@@ -133,7 +133,7 @@ check_called (void)
{
if (back_in_main == 0)
{
- puts ("program did no reach main again");
+ puts ("program did not reach main again");
_exit (1);
}
}
diff --git a/stdlib/tst-sigcontext.c b/stdlib/tst-sigcontext.c
new file mode 100644
index 0000000..45b1a06
--- /dev/null
+++ b/stdlib/tst-sigcontext.c
@@ -0,0 +1,312 @@
+/* Copyright (C) 2009 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Maciej W. Rozycki <macro@...>.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, write to the Free
+ Software Foundation, 51 Franklin Street - Fifth Floor, Boston, MA
+ 02110-1301, USA. */
+
+#include <errno.h>
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+#include <ucontext.h>
+#include <unistd.h>
+
+static volatile ucontext_t ctx[3];
+static volatile int flags[3];
+static volatile int oflag;
+static volatile int flag;
+
+static volatile int sig_errno_set;
+static volatile int sig_errno;
+static volatile int sig_ready;
+
+static int was_in_f1;
+static int was_in_f2;
+
+static char st2[32768];
+
+static long long int max;
+static volatile long long int count;
+
+/* These have to be run with the caller's frame. */
+static inline int sig_getcontext (int) __attribute__ ((always_inline));
+static inline int sig_swapcontext (int, int) __attribute__ ((always_inline));
+
+static void
+sig_gethandler (int signo, siginfo_t *info, void *context)
+{
+ ucontext_t *ucp = context;
+
+ if (! sig_ready)
+ {
+ alarm (1);
+ return;
+ }
+
+ ctx[flag] = *ucp;
+ flags[flag] = 1;
+
+ sig_errno_set = 1;
+ sig_errno = 0;
+}
+
+static void
+sig_swaphandler (int signo, siginfo_t *info, void *context)
+{
+ ucontext_t *oucp = context;
+ ucontext_t *ucp = (ucontext_t *) (ctx + flag);
+
+ if (! sig_ready)
+ {
+ alarm (1);
+ return;
+ }
+
+ ctx[oflag] = *oucp;
+ flags[oflag] = 1;
+
+ sig_errno_set = 1;
+ sig_errno = setcontext (ucp);
+}
+
+static int
+sig_getcontext (int uci)
+{
+ struct sigaction sa;
+
+ sigemptyset (&sa.sa_mask);
+ sa.sa_flags = SA_SIGINFO;
+ sa.sa_sigaction = sig_gethandler;
+ if (sigaction (SIGALRM, &sa, NULL) < 0)
+ {
+ printf ("%s: sigaction ALRM failed: %m\n", __FUNCTION__);
+ exit (1);
+ }
+
+ flag = uci;
+ sig_errno_set = 0;
+ sig_ready = 0;
+ alarm (1);
+
+ /* Need to get the signal in this frame. */
+ sig_ready = 1;
+ for (count = max; count > 0; count--)
+ if (sig_errno_set)
+ break;
+
+ if (! sig_errno_set)
+ {
+ printf ("%s: ALRM signal not received\n", __FUNCTION__);
+ exit (1);
+ }
+
+ return 0;
+}
+
+static int
+sig_swapcontext (int ouci, int uci)
+{
+ struct sigaction sa;
+
+ sigemptyset (&sa.sa_mask);
+ sa.sa_flags = SA_SIGINFO;
+ sa.sa_sigaction = sig_swaphandler;
+ if (sigaction (SIGALRM, &sa, NULL) < 0)
+ {
+ printf ("%s: sigaction ALRM failed: %m\n", __FUNCTION__);
+ exit (1);
+ }
+
+ flag = uci;
+ oflag = ouci;
+ sig_errno_set = 0;
+ sig_ready = 0;
+ alarm (1);
+
+ /* Need to get the signal in this frame. */
+ sig_ready = 1;
+ for (count = max; count > 0; count--)
+ if (sig_errno_set)
+ break;
+
+ if (! sig_errno_set)
+ {
+ printf ("%s: ALRM signal not received\n", __FUNCTION__);
+ exit (1);
+ }
+
+ return sig_errno;
+}
+
+
+static void
+f1 (int a0, int a1, int a2, int a3)
+{
+ int status;
+
+ printf ("start f1(a0=%x,a1=%x,a2=%x,a3=%x)\n", a0, a1, a2, a3);
+
+ if (a0 != 1 || a1 != 2 || a2 != 3 || a3 != -4)
+ {
+ puts ("arg mismatch");
+ exit (-1);
+ }
+
+ status = sig_swapcontext (1, 2);
+ if (status != 0)
+ {
+ printf ("%s: swapping contexts failed: %s\n",
+ __FUNCTION__, strerror (status));
+ exit (1);
+ }
+ puts ("finish f1");
+ was_in_f1 = 1;
+}
+
+static void
+f2 (void)
+{
+ int status;
+
+ puts ("start f2");
+
+ status = sig_swapcontext (2, 1);
+ if (status != 0)
+ {
+ printf ("%s: swapping contexts failed: %s\n",
+ __FUNCTION__, strerror (status));
+ exit (1);
+ }
+ puts ("finish f2");
+ was_in_f2 = 1;
+}
+
+
+static int back_in_main;
+
+static void
+check_called (void)
+{
+ if (back_in_main == 0)
+ {
+ printf ("%s: program did not reach main again", __FUNCTION__);
+ _exit (1);
+ }
+}
+
+
+volatile int global;
+
+int
+main (void)
+{
+ char st1[32768];
+ time_t t0, t1;
+ int status;
+
+ atexit (check_called);
+
+ puts ("calibrating delay loop");
+ for (max = 0x10000; max < LLONG_MAX / 2; max *= 2)
+ {
+ t0 = time (NULL);
+ for (count = max; count > 0; count--);
+ t1 = time (NULL);
+ if (difftime (t1, t0) > 10.0)
+ break;
+ }
+
+ puts ("making contexts");
+
+ sig_getcontext (1);
+ if (! flags[1])
+ {
+ printf ("%s: context not retrieved\n", __FUNCTION__);
+ exit (1);
+ }
+
+ /* Play some tricks with this context. */
+ if (++global == 1)
+ if (setcontext ((ucontext_t *) &ctx[1]) != 0)
+ {
+ if (errno == ENOSYS)
+ {
+ back_in_main = 1;
+ exit (0);
+ }
+ printf ("%s: setcontext: %m\n", __FUNCTION__);
+ exit (1);
+ }
+ if (global != 2)
+ {
+ printf ("%s: 'global' not incremented twice\n", __FUNCTION__);
+ exit (1);
+ }
+
+ ctx[1].uc_stack.ss_sp = st1;
+ ctx[1].uc_stack.ss_size = sizeof st1;
+ ctx[1].uc_link = (ucontext_t *) &ctx[0];
+ {
+ ucontext_t tempctx = ctx[1];
+ makecontext ((ucontext_t *) &ctx[1], (void (*) (void)) f1, 4, 1, 2, 3, -4);
+
+ /* Without this check, a stub makecontext can make us spin forever. */
+ if (memcmp (&tempctx, (void *) &ctx[1], sizeof ctx[1]) == 0)
+ {
+ puts ("makecontext was a no-op, presuming not implemented");
+ return 0;
+ }
+ }
+
+ sig_getcontext (2);
+ if (! flags[2])
+ {
+ printf ("%s: second context not retrieved\n", __FUNCTION__);
+ exit (1);
+ }
+
+ ctx[2].uc_stack.ss_sp = st2;
+ ctx[2].uc_stack.ss_size = sizeof st2;
+ ctx[2].uc_link = (ucontext_t *) &ctx[1];
+ makecontext ((ucontext_t *) &ctx[2], f2, 0);
+
+ puts ("swapping contexts");
+ status = sig_swapcontext (0, 2);
+ if (status != 0)
+ {
+ printf ("%s: swapping contexts failed: %s\n",
+ __FUNCTION__, strerror (status));
+ exit (1);
+ }
+ puts ("back at main program");
+ back_in_main = 1;
+
+ if (was_in_f1 == 0)
+ {
+ puts ("didn't reach f1");
+ exit (1);
+ }
+ if (was_in_f2 == 0)
+ {
+ puts ("didn't reach f2");
+ exit (1);
+ }
+
+ puts ("test succeeded");
+ return 0;
+}

From forum: Sourceware - libc-alpha

Re: [ARM] {OABI, EABI} x {LinuxThreads, NPTL} (was: Breaking ARM's mcount implementations into a separate file)

2009-12-15T06:35:55Z

On Tue, Dec 15, 2009 at 10:41:51AM +0100, Thomas Schwinge wrote:
> Hello!
>
> While we're at that -- another question: what's the general policy on
> keeping ARM's OABI and LinuxThreads code up to date that is still present
> in the repository? Obviously it needs time to maintain / enhace these
> code paths when doing changes to EABI / NPTL code -- but are these
> changes at all used by anyone, or is it in fact wasted time? Let alone
> the difficulty with testing such changes. Should maybe even the OABI and
> LinuxThread code blocks / files be removed from the repository?

We've left it there in case anyone wants to use it - but still don't
bother updating it. I expect a few people still use the OABI NPTL
port, and no one uses new glibc sources with OABI LinuxThreads.

--
Daniel Jacobowitz
CodeSourcery

From forum: Sourceware - libc-ports

Re: Breaking ARM's mcount implementations into a separate file

2009-12-15T05:55:16Z

On Tue, 15 Dec 2009, Thomas Schwinge wrote:

> Hello!
>
> On 2009-12-15 02:05, Joseph S. Myers wrote:
> > On Tue, 15 Dec 2009, Thomas Schwinge wrote:
> >> I inspected the pre / post libraries (symbols, disassembly), and did test
> >> runs to confirm that profiling still works in the ``normal'' and -mthumb
> >> cases.
> >
> > I'm not sure of the significance of having separate ARM/Thumb versions,
>
> Well, if I understand this correctly, it was you, together with Paul
> Brook, who have added this code on 2009-02-05.

You need to look more carefully at the history before I merged Thumb-2
support from EGLIBC to FSF ports. The new mcount entry point dates back
at least to <http://www.eglibc.org/archives/patches/msg00449.html> and
other discussions around that time, while Thumb-2 mcount comes from
<http://www.eglibc.org/archives/patches/msg00143.html>. I.e., on
2009-02-05 I merged existing code without knowing the significance of
every bit of it.

> > #if defined(__thumb2__)
> > .thumb
> > .syntax unified
> > #endif
>
> This is correct. I transformed the code without thinking about that. Do
> I need to resend the patch, or can the one who is comitting it easily add
> the two assembler directives to the [__thumb2__] code?

You'll need to send the patch in a form appropriate for committing, tested
as both ARM and Thumb-2.

--
Joseph S. Myers
joseph@...

From forum: Sourceware - libc-ports

[ARM] {OABI, EABI} x {LinuxThreads, NPTL} (was: Breaking ARM's mcount implementations into a separate file)

2009-12-15T01:41:51Z

Hello!

While we're at that -- another question: what's the general policy on
keeping ARM's OABI and LinuxThreads code up to date that is still present
in the repository? Obviously it needs time to maintain / enhace these
code paths when doing changes to EABI / NPTL code -- but are these
changes at all used by anyone, or is it in fact wasted time? Let alone
the difficulty with testing such changes. Should maybe even the OABI and
LinuxThread code blocks / files be removed from the repository?

Regards,
Thomas

attachment0 (202 bytes) Download Attachment

From forum: Sourceware - libc-ports

Re: Breaking ARM's mcount implementations into a separate file

2009-12-15T00:54:41Z

Hello!

On 2009-12-15 02:05, Joseph S. Myers wrote:
> On Tue, 15 Dec 2009, Thomas Schwinge wrote:
>> I inspected the pre / post libraries (symbols, disassembly), and did test
>> runs to confirm that profiling still works in the ``normal'' and -mthumb
>> cases.
>
> I'm not sure of the significance of having separate ARM/Thumb versions,

Well, if I understand this correctly, it was you, together with Paul
Brook, who have added this code on 2009-02-05.

> but it does not make much sense to me to have different __thumb2__
> versions that still get compiled in ARM mode when glibc is built as
> Thumb-2. Remember that .S files get built in ARM mode unless they contain
> assembler directives to use Thumb, but __thumb2__ is defined if the
> compiler would generate Thumb-2. If you want particular .S sources built
> as Thumb-2 code, you need something like sysdeps/arm/elf/start.S:
>
> #if defined(__thumb2__)
> .thumb
> .syntax unified
> #endif

This is correct. I transformed the code without thinking about that. Do
I need to resend the patch, or can the one who is comitting it easily add
the two assembler directives to the [__thumb2__] code?

Regards,
Thomas

attachment0 (202 bytes) Download Attachment

From forum: Sourceware - libc-ports

Re: Breaking ARM's mcount implementations into a separate file

2009-12-14T18:05:23Z

On Tue, 15 Dec 2009, Thomas Schwinge wrote:

> I inspected the pre / post libraries (symbols, disassembly), and did test
> runs to confirm that profiling still works in the ``normal'' and -mthumb
> cases.

I'm not sure of the significance of having separate ARM/Thumb versions,
but it does not make much sense to me to have different __thumb2__
versions that still get compiled in ARM mode when glibc is built as
Thumb-2. Remember that .S files get built in ARM mode unless they contain
assembler directives to use Thumb, but __thumb2__ is defined if the
compiler would generate Thumb-2. If you want particular .S sources built
as Thumb-2 code, you need something like sysdeps/arm/elf/start.S:

#if defined(__thumb2__)
.thumb
.syntax unified
#endif

--
Joseph S. Myers
joseph@...

From forum: Sourceware - libc-ports

Breaking ARM's mcount implementations into a separate file

2009-12-14T16:08:24Z

Hello!

ARM currently implements the mcount functions as inline assembly in
__attribute__ (__naked__) functions. This interferes with my ongoing
work to add CFI statements to ARM's assembly code, as I can't be sure
whether .cfi_startproc has already been issued (-g with recent GCC) or
not (and I'd have to do it manually in the inline assembly code then, but
only in this latter case). Also, this patch simplifies the code by
undoing the wrapping of macro definitions containing C functions
containing (only) inline assembly code, and aligns to what the mainline
code is using.

2009-12-15 Thomas Schwinge <thomas@...>

* sysdeps/arm/eabi/Makefile [gmon] (sysdep_routines): Add arm-mcount.
* sysdeps/arm/eabi/machine-gmon.h (mcount_internal): Define to
`__mcount_internal'.
(MCOUNT (__gnu_mcount_nc), OLD_MCOUNT (_mcount), mcount): Move to...
* sysdeps/arm/eabi/arm-mcount.S: ... this new file as __gnu_mcount_nc,
_mcount, and mcount, respectively.

diff --git a/glibc-ports-mainline/sysdeps/arm/eabi/Makefile b/glibc-ports-mainline/sysdeps/arm/eabi/Makefile
index 36ddb8e..73185cb 100644
--- a/glibc-ports-mainline/sysdeps/arm/eabi/Makefile
+++ b/glibc-ports-mainline/sysdeps/arm/eabi/Makefile
@@ -15,6 +15,10 @@ static-only-routines += $(aeabi_constants)
gen-as-const-headers += rtld-global-offsets.sym
endif

+ifeq ($(subdir),gmon)
+sysdep_routines += arm-mcount
+endif
+
ifeq ($(subdir),debug)
CFLAGS-backtrace.c += -funwind-tables
CFLAGS-tst-backtrace2.c += -funwind-tables
diff --git a/glibc-ports-mainline/sysdeps/arm/eabi/arm-mcount.S b/glibc-ports-mainline/sysdeps/arm/eabi/arm-mcount.S
new file mode 100644
index 0000000..82ef02b
--- /dev/null
+++ b/glibc-ports-mainline/sysdeps/arm/eabi/arm-mcount.S
@@ -0,0 +1,78 @@
+/* Implementation of profiling support. ARM EABI version.
+ Copyright (C) 2008, 2009 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, write to the Free
+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+ 02111-1307 USA. */
+
+#include <sysdep.h>
+
+#undef CALL_MCOUNT
+#define CALL_MCOUNT /* Don't call mcount when calling mcount... */
+
+/* Use an assembly stub with a special ABI. The calling lr has been
+ pushed to the stack (which will be misaligned). We should preserve
+ all registers except ip and pop a word off the stack.
+
+ NOTE: This assumes mcount_internal does not clobber any non-core
+ (coprocessor) registers. Currently this is true, but may require
+ additional attention in the future.
+
+ The calling sequence looks something like:
+func:
+ push {lr}
+ bl __gnu_mount_nc
+ <function body>
+*/
+
+ENTRY(__gnu_mcount_nc)
+ push {r0, r1, r2, r3, lr}
+ bic r1, lr, #1
+ ldr r0, [sp, #20]
+ bl __mcount_internal
+ pop {r0, r1, r2, r3, ip, lr}
+ bx ip
+END(__gnu_mcount_nc)
+
+
+/* Provide old mcount for backwards compatibility. This requires
+ code be compiled with APCS frame pointers. */
+
+#ifdef __thumb2__
+ENTRY(_mcount)
+ push {r0, r1, r2, r3, fp, lr}
+ movs r0, fp
+ ittt ne
+ ldrne r0, [r0, #-4]
+ movsne r1, lr
+ blne __mcount_internal
+ pop {r0, r1, r2, r3, fp, pc}
+END(_mcount)
+#else
+ENTRY(_mcount)
+ stmdb sp!, {r0, r1, r2, r3, fp, lr}
+ movs fp, fp
+ ldrne r0, [fp, #-4]
+ movnes r1, lr
+ blne __mcount_internal
+ ldmia sp!, {r0, r1, r2, r3, fp, lr}
+ bx lr
+END(_mcount)
+#endif
+
+/* The canonical name for the function is `_mcount' in both C and asm,
+ but some old asm code might assume it's `mcount'. */
+#undef mcount
+weak_alias (_mcount, mcount)
diff --git a/glibc-ports-mainline/sysdeps/arm/eabi/machine-gmon.h b/glibc-ports-mainline/sysdeps/arm/eabi/machine-gmon.h
index 189a9a3..aca23c9 100644
--- a/glibc-ports-mainline/sysdeps/arm/eabi/machine-gmon.h
+++ b/glibc-ports-mainline/sysdeps/arm/eabi/machine-gmon.h
@@ -1,5 +1,5 @@
/* Machine-dependent definitions for profiling support. ARM EABI version.
- Copyright (C) 2008 Free Software Foundation, Inc.
+ Copyright (C) 2008, 2009 Free Software Foundation, Inc.
This file is part of the GNU C Library.

The GNU C Library is free software; you can redistribute it and/or
@@ -17,83 +17,18 @@
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */

+#include <sysdep.h>
+
/* GCC for the ARM cannot compile __builtin_return_address(N) for N != 0,
so we must use an assembly stub. */

-#include <sysdep.h>
-static void mcount_internal (u_long frompc, u_long selfpc) __attribute_used__;
+/* We must not pollute the global namespace. */
+#define mcount_internal __mcount_internal

+extern void mcount_internal (u_long frompc, u_long selfpc) internal_function;
#define _MCOUNT_DECL(frompc, selfpc) \
-static void mcount_internal (u_long frompc, u_long selfpc)
-
-/* Use an assembly stub with a special ABI. The calling lr has been
- pushed to the stack (which will be misaligned). We should preserve
- all registers except ip and pop a word off the stack.
-
- NOTE: This assumes mcount_internal does not clobber any non-core
- (coprocessor) registers. Currently this is true, but may require
- additional attention in the future.
-
- The calling sequence looks something like:
-func:
- push {lr}
- bl __gnu_mount_nc
- <function body>
- */
-
-
-#define MCOUNT \
-void __attribute__((__naked__)) __gnu_mcount_nc(void) \
-{ \
- asm ("push {r0, r1, r2, r3, lr}\n\t" \
- "bic r1, lr, #1\n\t" \
- "ldr r0, [sp, #20]\n\t" \
- "bl mcount_internal\n\t" \
- "pop {r0, r1, r2, r3, ip, lr}\n\t" \
- "bx ip"); \
-} \
-OLD_MCOUNT
-
-/* Provide old mcount for backwards compatibility. This requires
- code be compiled with APCS frame pointers. */
-
-#ifndef NO_UNDERSCORES
-/* The asm symbols for C functions are `_function'.
- The canonical name for the counter function is `mcount', no _. */
-void _mcount (void) asm ("mcount");
-#else
-/* The canonical name for the function is `_mcount' in both C and asm,
- but some old asm code might assume it's `mcount'. */
-void _mcount (void);
-weak_alias (_mcount, mcount)
-#endif
-
-#ifdef __thumb2__
-
-#define OLD_MCOUNT \
-void __attribute__((__naked__)) _mcount (void) \
-{ \
- __asm__("push {r0, r1, r2, r3, fp, lr};" \
- "movs r0, fp;" \
- "ittt ne;" \
- "ldrne r0, [r0, #-4];" \
- "movsne r1, lr;" \
- "blne mcount_internal;" \
- "pop {r0, r1, r2, r3, fp, pc}"); \
-}
-
-#else
+ void internal_function mcount_internal (u_long frompc, u_long selfpc)

-#define OLD_MCOUNT \
-void __attribute__((__naked__)) _mcount (void) \
-{ \
- __asm__("stmdb sp!, {r0, r1, r2, r3, fp, lr};" \
- "movs fp, fp;" \
- "ldrne r0, [fp, #-4];" \
- "movnes r1, lr;" \
- "blne mcount_internal;" \
- "ldmia sp!, {r0, r1, r2, r3, fp, lr};" \
- "bx lr"); \
-}

-#endif
+/* Define MCOUNT as empty since we have the implementation in another file. */
+#define MCOUNT

I inspected the pre / post libraries (symbols, disassembly), and did test
runs to confirm that profiling still works in the ``normal'' and -mthumb
cases.

Regards,
Thomas

attachment0 (202 bytes) Download Attachment

From forum: Sourceware - libc-ports

The Linux binutils 2.20.51.0.4 is released

2009-12-14T09:37:24Z

This is the beta release of binutils 2.20.51.0.4 for Linux, which is
based on binutils 2009 1214 in CVS on sourceware.org plus various
changes. It is purely for Linux.

All relevant patches in patches have been applied to the source tree.
You can take a look at patches/README to see what have been applied and
in what order they have been applied.

Starting from the 2.20.51.0.4 release, no diffs against the previous
release will be provided.

You can enable both gold and bfd ld with --enable-gold=both. Gold will
be installed as ld.gold and bfd ld will be installed as ld.bfd. By
default, ld.bfd will be installed as ld. You can use the configure
option, --enable-linker=[bfd,gold] to choose the default linker, ld.
IA-32 binary and X64_64 binary tar balls are configured with
--enable-gold=both --enable-plugins --enable-threads.

Starting from the 2.18.50.0.4 release, the x86 assembler no longer
accepts

fnstsw %eax

fnstsw stores 16bit into %ax and the upper 16bit of %eax is unchanged.
Please use

fnstsw %ax

Starting from the 2.17.50.0.4 release, the default output section LMA
(load memory address) has changed for allocatable sections from being
equal to VMA (virtual memory address), to keeping the difference between
LMA and VMA the same as the previous output section in the same region.

For

.data.init_task : { *(.data.init_task) }

LMA of .data.init_task section is equal to its VMA with the old linker.
With the new linker, it depends on the previous output section. You
can use

.data.init_task : AT (ADDR(.data.init_task)) { *(.data.init_task) }

to ensure that LMA of .data.init_task section is always equal to its
VMA. The linker script in the older 2.6 x86-64 kernel depends on the
old behavior. You can add AT (ADDR(section)) to force LMA of
.data.init_task section equal to its VMA. It will work with both old
and new linkers. The x86-64 kernel linker script in kernel 2.6.13 and
above is OK.

The new x86_64 assembler no longer accepts

monitor %eax,%ecx,%edx

You should use

monitor %rax,%ecx,%edx

or
monitor

which works with both old and new x86_64 assemblers. They should
generate the same opcode.

The new i386/x86_64 assemblers no longer accept instructions for moving
between a segment register and a 32bit memory location, i.e.,

movl (%eax),%ds
movl %ds,(%eax)

To generate instructions for moving between a segment register and a
16bit memory location without the 16bit operand size prefix, 0x66,

mov (%eax),%ds
mov %ds,(%eax)

should be used. It will work with both new and old assemblers. The
assembler starting from 2.16.90.0.1 will also support

movw (%eax),%ds
movw %ds,(%eax)

without the 0x66 prefix. Patches for 2.4 and 2.6 Linux kernels are
available at

http://www.kernel.org/pub/linux/devel/binutils/linux-2.4-seg-4.patch
http://www.kernel.org/pub/linux/devel/binutils/linux-2.6-seg-5.patch

The ia64 assembler is now defaulted to tune for Itanium 2 processors.
To build a kernel for Itanium 1 processors, you will need to add

ifeq ($(CONFIG_ITANIUM),y)
CFLAGS += -Wa,-mtune=itanium1
AFLAGS += -Wa,-mtune=itanium1
endif

to arch/ia64/Makefile in your kernel source tree.

Please report any bugs related to binutils 2.20.51.0.3 to
hjl.tools@...

and

http://www.sourceware.org/bugzilla/

Changes from binutils 2.20.51.0.3:

1. Update from binutils 2009 1214.
2. Update x86 assembler to check lockable instructions for lock prefix.
3. Update x86 disassembler to display all prefixes.
4. Support AMD XOP new instructions.
5. Fix an x86 assembler regression on Intel syntax. PR 11037.
6. Improve ia64 linker relaxation. PR 10955.
7. Add --no-relax linker option.
8. Update readelf to dump .debug_pubtype sections.
9. Improve gold:
a. Support linking against STT_GNU_IFUNC symbols defined in
shared libraries.
b. Support linking with STB_GNU_UNIQUE symbols.
10. Improve arm support.
11. Improve m68k support.
12. Improve mips support.
13. Improve ppc support.

Changes from binutils 2.20.51.0.2:

1. Update from binutils 2009 1109.
2. Fix "ld -s -static" with STT_GNU_IFUNC symbols. PR 10911.
3. Fix file permission on PIE with objcopy. PR 10802.
4. Fix x86 Intel syntax assembler with relocation. PR 10856.
5. Fix x86 Intel syntax assembler with far jump. PR 10740.
6. Add AMD LWP support.
7. Renamed linker option --add-needed to --copy-dt-needed-entries.
8. Support enabling both ld and gold with --enable-gold=both and
--enable-linker=[bfd,gold].
9. Improve gold.
10. Improve arm support.
11. Improve cris support.
12. Improve hppa support.
13. Improve m68k support.
14. Improve RX support.
15. Improve spu support.
16. Improve vax support.
17. Improve MacOS support.
18. Improve Windows support.

Changes from binutils 2.20.51.0.1:

1. Update from binutils 2009 1009.
2. Add .cfi_sections to assembler.
3. Fix a linker bug with local dynamic symbols. PR 10630.
4. Add DWARF-3/DWARF-4 support.
5. Fix the x86 assembler PIC bug. PR 10677.
6. Fix the x86-64 displacement assembler bug. PR 10636.
7. Fix the x86 assembler bug with Intel memory syntax. PR 10637.
8. Fix the x86 PIC assembler bug with Intel syntax. PR 10704.
9. Add RX support.
10. Improve gold.
11. Improve arm support.
12. Improve bfin support.
13. Improve cr16 support.
14. Improve m68k support.
15. Improve mips support.
16. Improve ppc support.

Changes from binutils 2.19.51.0.14:

1. Update from binutils 2009 0905.
2. Add Intel L1OM support.
3. Add MicroBlaze support.
4. Fix assembler for DWARF info without .file/.loc directives. PR 10531.
5. Improve -pie with TLS relocations on ia32 and x86-64. PRs 6443/10434.
6. Fix linker page size support. PR 10569.
7. Fix wildcard in linker version script. PR 10518.
8. Fix strip with STB_GNU_UNIQUE. PR 10492.
9. Fix strip on unwriteable files. PR 10364.
10. Fix crash with "ld --build-id /usr/lib/libc.a". PR 10555.
11. Fix linker for Linux kernel build. PR 10429.
12. Support string merge on .comment section.
13. Improve build with C++ compiler.
14. Improve gold.
15. Improve arm support.
16. Improve bfin support.
11. Improve m32c support.
17. Improve m68k support.
18. Improve mep support.
19. Improve mips support.
20. Improve ppc support.
21. Improve spu support.
22. Improve xtensa support.

Changes from binutils 2.19.51.0.13:

1. Update from binutils 2009 0722.
2. Fix linker for STT_GNU_IFUNC symbols in static executables. PR 10433.
3. Fix linker bug for Linux kernel build. PR 10429.

Changes from binutils 2.19.51.0.12:

1. Update from binutils 2009 0721.
2. Fix linker for undefined STT_GNU_IFUNC symbols. PR 10426.
3. Fix x86 assembler for nops in 64bit. PR 10420.
4. Add a new option, --insn-width, to objdump.
5. Improve arm support.
6. Improve mips support.
7. Improve gold support.

Changes from binutils 2.19.51.0.11:

1. Update from binutils 2009 0716.
2. Fix x86 assembler for jumping to local STT_GNU_IFUNC symbols.
3. Fix x86 linker for relocatable link with local STT_GNU_IFUNC symbols.
4. Implement ppc STT_GNU_IFUNC support.
5. Support x86 FMA4.
6. Fix linker regression with Linux kernel build.
7. Support unordered references in DWARF reader.
8. Improve PE/COFF support.
8. Improve arm support.
9. Improve m10300 support.
10. Improve ppc support.
11. Improve spu support.
12. Improve gold support.

Changes from binutils 2.19.51.0.10:

1. Update from binutils 2009 0627.
2. Fix strip on static executable with STT_GNU_IFUNC symbol. PR 10337.
3. Add STB_GNU_UNIQUE support.
4. Fix objcopy on empty file. PR 10321.
5. Fix debug section for PE-COFF.
6. Suport build with gcc 4.5.0.
7. Improve arm support.
8. Improve ppc support.
9. Improve m10300 support.
10. Improve mep support.
11. Improve MacOS support.
12. Improve gold support.

Changes from binutils 2.19.51.0.9:

1. Update from binutils 2009 0618.
2. Update STT_GNU_IFUNC symbol support. PR 10269/10270.
3. Fix an assembler CFI bug. PR 10255.
4. Improve objdump. PR 10263/10288
5. Improve readelf.
6. Improve arm support.
7. Improve moxie support.
8. Improve spu support.
9. Improve vax support.
10. Improve COFF/PE support.
11. Improve MacOS support.

Changes from binutils 2.19.51.0.8:

1. Update from binutils 2009 0606.
2. Update STT_GNU_IFUNC symbol support.

Changes from binutils 2.19.51.0.7:

1. Update from binutils 2009 0603.
2. Fix STT_GNU_IFUNC symbol with pointer equality.

Changes from binutils 2.19.51.0.6:

1. Update from binutils 2009 0601.
2. Update STT_GNU_IFUNC support. PR 10205.
3. Fix x86 asssembler Intel syntax regression with '$'. PR 10198.

Changes from binutils 2.19.51.0.5:

1. Update from binutils 2009 0529.
2. Rewrite STT_GNU_IFUNC, R_386_IRELATIVE and R_X86_64_IRELATIVE linker
support for STT_GNU_IFUNC symbols in shared library, dynamic executable
and static executable.
3. Add plugin support.
4. Improve spu support.

Changes from binutils 2.19.51.0.4:

1. Update from binutils 2009 0525.
2. Add STT_GNU_IFUNC, R_386_IRELATIVE and R_X86_64_IRELATIVE support to
assembler and linker.
3. Add LD_AS_NEEDED support to linker.
4. Remove AMD SSE5 support.
5. A new Intel syntax parser in x86 assembler.
6. Add DWARF discriminator support.
7. Add --64 support for x86 PE/COFF assembler.
8. Support common symbol with alignment for PE/COFF.
9. Improve gold support.
10. Improve arm support.
11. Improve mep support.
12. Improve mips support.
13. Improve ppc support.
14. Improve spu support.

Changes from binutils 2.19.51.0.3:

1. Update from binutils 2009 0418.
2. Remove EFI targets and use PEI targets for EFI. Add --file-alignment,
--heap, --image-base, --section-alignment, --stack and --subsystem command
line options for objcopy. PR 10074.
3. Update linker to warn alternate ELF machine code.
4. Fix x86 linker TLS transition. PR 9938.
5. Improve DWARF dumper to check relocations against STT_SECTION
symbol.
6. Guard DWARF dumper on bad DWARF input.
7. Add EM_ETPU and EM_SLE9X. Reserve 3 ELF machine types for Intel.
8. Adding a linker missing entry symbol warning for -pie. PR 9970.
9. Make the -e option for linker to imply -u. PR 6766.
10. Properly handle paging for PEI targets.
11. Fix assembler listing with input from stdin.
12. Update objcopy/string to generate symbol table if there is any
relocation in output. PR 9945.
13. Require texinfo 4.7 for build. PR 10039.
14. Add moxie support.
15. Improve gold support.
16. Improve AIX support.
17. Improve arm support.
18. Improve cris support.
19. Improve crx support.
20. Improve mips support.
21. Improve ppc support.
22. Improve s390 support.
23. Improve spu support.
24. Improve vax support.

Changes from binutils 2.19.51.0.2:

1. Update from binutils 2009 0310.
2. Fix strip on common symbols in relocatable file. PR 9933.
3. Fix --enable-targets=all build.
4. Fix ia64 build with -Wformat-security. PR 9874.
5. Add REGION_ALIAS support in linker script.
6. Add think archive support to readelf.
7. Improve DWARF support in objdump.
8. Improve alpha support.
9. Improve arm support.
10. Improve hppa support.
11. Improve m68k support.
12. Improve mips support.
13. Improve ppc support.
14. Improve xtensa support.
15. Add score 7 support.

Changes from binutils 2.19.51.0.1:

1. Update from binutils 2009 0204.
2. Support AVX Programming Reference (January, 2009)
3. Improve .s suffix support in x86 disassembler.
4. Add --prefix/--prefix-strip for objdump -S. PR 9784.
5. Change "ld --as-needed" to resolve undefined references in DSO.
6. Add -Ttext-segment to ld to set address of text segment.
7. Fix "ld -r --gc-sections --entry" crash with COMDAT group. PR 9727.
8. Improve linker compatibility for g++ 3.4 `.gnu.linkonce.r.*.
9. Add VMS/ia64 support.
10. Improve arm support.
11. Improve cris support.
12. Improve m68k support.
13. Improve mips support.
14. Improve spu support.

Changes from binutils 2.19.50.0.1:

1. Update from binutils 2009 0106.
2. Support AVX Programming Reference (December, 2008)
2. Encode AVX insns with 2byte VEX prefix if possible.
4. Add .s suffix support to swap register operands to x86 assembler.
5. Properly select NOP insns for code alignment in x86 assembler.
6. Fix 2 symbol visibility linker bugs. PRs 9676/9679.
7. Fix an ia64 linker relaxation bug. PR 7036.
8. Fix a symbol versioning bug. PR 7047.
9. Fix unitialized data in linker. PR 7028.
10. Avoid a linker crash on bad input. PR 7023.
11. Fix a linker memory leak. PR 7012.
12. Fix strip/objcopy crash on PT_GNU_RELRO. PR 7011.
13. Improve MacOS support.
14. Fix a COFF linker bug. PR 6945.
15. Add LM32 support.
16. Fix various arm bugs.
17. Fix various avr bugs.
18. Fix various CR16 bugs.
19. Fix various cris bugs.
20. Fix various m32c bugs.
21. Fix various m68k bugs.
22. Fix various mips bugs.
23. Fix various ppc bugs.
24. Fix various s390 bugs.
25. Fix various sparc bugs.
26. Fix various spu bugs.
27. Fix various xtensa bugs.

The file list:

1. binutils-2.20.51.0.3.tar.bz2. Source code.
2. binutils-2.20.51.0.3.i686.tar.bz2. IA-32 binary tar ball for RedHat
EL 4.
3. binutils-2.20.51.0.3.ia64.tar.bz2. IA-64 binary tar ball for RedHat
EL 4.
4. binutils-2.20.51.0.3.x86_64.tar.bz2. X64_64 binary tar ball for RedHat
EL 4.

The primary sites for the beta Linux binutils are:

1. http://www.kernel.org/pub/linux/devel/binutils/

Thanks.

H.J. Lu
hjl.tools@...
12/14/2009

From forum: Sourceware - libc-alpha

[Bug libc/11087] New: mallinfo miscounting hblks because of missing mutex

2009-12-14T00:56:34Z

I have an issue with non-atomic increment/
decrement
of hblks, (i.e. n_mmaps) in a multi-threaded mission critical process
that monitors
heap usage as part of supervision. This value can drift astray of the
true count and
rarerly triggers an assertion in the total heap calculatio. There is a
similar disabled
assert in glibc:

#ifdef NO_THREADS
assert(total <= (unsigned long)(mp_.max_total_mem));
assert(mp_.n_mmaps >= 0);
#endif

Seem to be a problem of wanting to pay the cost of mutex contention
instead of
the (application specific) cost of providing wrong values. In the
context of a malloc call
that asks for 2 Mb, I have not figured out that part yet, but I can't
imagine that keeping
a correct total count in a multi-threaded is beyond state of the art
(but am sure it may
be beyond desire of hackers or motivation of maintainers). There is a
disabled
THREAD_STATS switch but that doesn't seem to handle this case. The
problem might be
resolved, but at home I can still reproduce it with latest glibc:

with Text_Io;
with Unchecked_Deallocation;
procedure T is

type Mallinfo_Field is (Arena, Ordblks, Smblks, Hblks, Hblkhd,
Usmblks, Fsmblks, Uordblks, Fordblks, Keepcost);

type Mallinfo_Stats is array (Mallinfo_Field) of Integer;
-- on 32 bit.

function mallinfo return Mallinfo_Stats;
pragma Interface (C, mallinfo);

type Large is array (1..2**22) of Character; -- 4M
type Large_Ptr is access Large;
procedure Free is new Unchecked_Deallocation (Large, Large_Ptr);

task type Alloc is
pragma Priority (0);
end Alloc;

task body Alloc is
L : array (1..10) of Large_Ptr;
begin
for Count in Natural loop
for I in L'Range loop
L(I) := new Large;
end loop;
for I in L'Range loop
Free (L(I));
end loop;
-- To check for sufficient parallelism
-- Test on multi-core/multi-processor
if Count mod 10000 = 0 then
Text_Io.Put_Line (Count'Img);
end if;
end loop;
end;

Loads : array (1..4) of Alloc;

procedure Show_Stats is
Result : constant Mallinfo_Stats := mallinfo;
begin
for Field in Result'Range loop
Text_IO.Put (Field'Img & "=" & Result(Field)'Img & " ");
end loop;
Text_Io.New_Line;
if Result(Hblks) not in 0..40 then
for T in Loads'Range loop
abort Loads(T);
end loop;
raise Program_Error;
end if;
end Show_Stats;

begin
loop
delay 1.0;
Show_Stats;
end loop;
end;

output :

ARENA= 135984 ORDBLKS= 1 SMBLKS= 0 HBLKS= 0 HBLKHD=-20992000 USMBLKS=
0 FSMBLKS= 0 UORDBLKS= 62424 FORDBLKS= 73560 KEEPCOST= 73560
22700000
22620000
ARENA= 135984 ORDBLKS= 1 SMBLKS= 0 HBLKS= 17 HBLKHD= 50380800 USMBLKS=
0 FSMBLKS= 0 UORDBLKS= 62424 FORDBLKS= 73560 KEEPCOST= 73560
22630000
22570000
ARENA= 135984 ORDBLKS= 1 SMBLKS= 0 HBLKS= 7 HBLKHD= 8396800 USMBLKS= 0
FSMBLKS= 0 UORDBLKS= 62424 FORDBLKS= 73560 KEEPCOST= 73560
22710000
22630000
22640000
22580000
ARENA= 135984 ORDBLKS= 1 SMBLKS= 0 HBLKS= 2 HBLKHD=-12595200 USMBLKS=
0 FSMBLKS= 0 UORDBLKS= 62424 FORDBLKS= 73560 KEEPCOST= 73560
22720000
22640000
ARENA= 135984 ORDBLKS= 1 SMBLKS= 0 HBLKS= 0 HBLKHD=-20992000 USMBLKS=
0 FSMBLKS= 0 UORDBLKS= 62424 FORDBLKS= 73560 KEEPCOST= 73560
22650000
22590000
ARENA= 135984 ORDBLKS= 1 SMBLKS= 0 HBLKS= 26 HBLKHD= 88166400 USMBLKS=
0 FSMBLKS= 0 UORDBLKS= 62424 FORDBLKS= 73560 KEEPCOST= 73560
22730000
22650000
22660000
22600000
ARENA= 135984 ORDBLKS= 1 SMBLKS= 0 HBLKS=-1 HBLKHD=-25190400 USMBLKS=
0 FSMBLKS= 0 UORDBLKS= 62424 FORDBLKS= 73560 KEEPCOST= 73560

raised PROGRAM_ERROR : t.adb:52 explicit raise

Posted on :
http://groups.google.com/group/comp.os.linux.development.apps/browse_thread/thread/80b0c703fe7004c2/f512a2856b0eb3e3?lnk=raot&pli=1

Found a comment stating mutexing was expensive :
http://groups.google.com/group/comp.os.linux.development.system/browse_thread/thread/c25c962a87d81895/674575250fd87044?q=mallinfo+hblk#674575250fd87044

I've tried adding a mutex, and I don't see what is expensive. When I compare the
latest version of glibc malloc after adding the mutex locking against the system
default glibc behaviour I don't measure a difference, results are within 3%
accurate (some fluctuations on the measure, just used time command), I guess
mutex cost is much smaller than mmap cost. Anyway, that is only performance of
iterative malloc/free of a 4 MB block, adding nothing more but a zero-fill of
this memory makes the operation about 1000 times slower, so I don't see a
significant issue with this mutex overhead.

In case overhead would be significant, at least a #define THREAD_SAFE 1 should
activate a fix, or a dynamic option, but surely correct total heap count should
be achievable.

My experiment :

$ diff -U3 arena.c.orig arena.c
--- arena.c.orig 2009-12-13 18:00:55.000000000 +0100
+++ arena.c 2009-12-13 18:03:12.000000000 +0100
@@ -93,6 +93,13 @@
/* Mapped memory in non-main arenas (reliable only for NO_THREADS). */
static unsigned long arena_mem;

+static mutex_t mp_mmap_mutex;
+/*Serialize access to (malloc.c) mp_.
+ int n_mmaps;
+ int n_mmaps_max;
+ int max_n_mmaps;
+*/
+
/* Already initialized? */
int __malloc_initialized = -1;

@@ -354,6 +361,8 @@
if(ar_ptr == &main_arena) break;
}
mutex_init(&list_lock);
+ mutex_init(&mp_mmap_mutex);
+
atfork_recursive_cntr = 0;
}

$ diff -U3 malloc.c.org malloc.c
--- malloc.c.org 2009-12-13 13:26:22.000000000 +0100
+++ malloc.c 2009-12-13 18:13:13.000000000 +0100
@@ -2359,6 +2359,8 @@
----------- Internal state representation and initialization -----------
*/

+#define THREAD_STATS 1
+
struct malloc_state {
/* Serialize access. */
mutex_t mutex;
@@ -2443,6 +2445,12 @@
/* There is only one instance of the malloc parameters. */

static struct malloc_par mp_;
+static mutex_t mp_mmap_mutex;
+/*Serialize access to (malloc.c) mp_.
+ int n_mmaps;
+ int n_mmaps_max;
+ int max_n_mmaps;
+*/

#ifdef PER_THREAD
@@ -3055,7 +3063,8 @@
set_head(p, size|IS_MMAPPED);
}

- /* update statistics */
+ /* begin update statistics */
+ (void)mutex_lock(&mp_mmap_mutex);

if (++mp_.n_mmaps > mp_.max_n_mmaps)
mp_.max_n_mmaps = mp_.n_mmaps;
@@ -3071,6 +3080,10 @@

check_chunk(av, p);

+ (void)mutex_unlock(&mp_mmap_mutex);
+ /* end update statistics */
+
+
return chunk2mem(p);
}
}
@@ -3521,6 +3534,7 @@
#endif
{
INTERNAL_SIZE_T size = chunksize(p);
+ (void)mutex_lock(&mp_mmap_mutex);

assert (chunk_is_mmapped(p));
#if 0
@@ -3544,6 +3558,7 @@

mp_.n_mmaps--;
mp_.mmapped_mem -= total_size;
+ (void)mutex_unlock(&mp_mmap_mutex);

int ret __attribute__ ((unused)) = munmap((char *)block, total_size);

@@ -3726,6 +3741,7 @@
_int_free(ar_ptr, p, 0);
#else
# if THREAD_STATS
+
if(!mutex_trylock(&ar_ptr->mutex))
++(ar_ptr->stat_lock_direct);
else {

P.S. This is actually a problem of a mission-critical system planned to become
operational on Redhat el5 next year. Assertion on positive results is now turned
into recoverable error, so non-critical but would like to get heap supervision
correct in the future for monitoring of unexpected memory leaks. We pay support
to HP (forwarding such problems to Redhat), so request for fixing might turn up
via official channels.

Test still running, but at least getting further than before :
64660000
64800000
64240000
64810000
ARENA= 135984 ORDBLKS= 1 SMBLKS= 0 HBLKS= 7 HBLKHD= 29388800 USMBLKS= 0 FSMBLKS=
0 UORDBLKS= 62424 FORDBLKS= 73560 KEEPCOST= 73560
64670000
64810000

--
Summary: mallinfo miscounting hblks because of missing mutex
Product: glibc
Version: 2.9
Status: NEW
Severity: normal
Priority: P2
Component: libc
AssignedTo: drepper at redhat dot com
ReportedBy: stef dot van-vlierberghe at telenet dot be
CC: glibc-bugs at sources dot redhat dot com

http://sourceware.org/bugzilla/show_bug.cgi?id=11087

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

Re: reverse look up of INADDR_ANY, etc.

2009-12-13T20:44:19Z

> On 11/23/2009 08:33 PM, danielcavanagh@... wrote:
>> is this consider a bug? if so, do you need me to file a report?
>
> Not a bug. It's "garbage in, garbage out". Just like any other invalid
> address you will get the correct result but it'll take it's time for the
> timeout to kick in.
>
> You'll have to show me a valid program which can regularly create such
> calls t convince me otherwise.

this was found in the ruby language. there is a TCPServer class whose
purpose should be clear. when you access the addr variable of this class,
it returns an array with the address family, the port, the ip address and
domain name bound to. the most common address to bind to is obviously
INADDR_ANY, which then gets automatically looked up, leading to a huge
pause on startup for most things using TCPServer on a glibc system

really, the check should be pushed as far down the code path as possible
and the very bottom is glibc. it's more efficient that way and avoids
duplication. and as i said, every other system i've seen works properly
(all the bsds, haiku, windows, non-glibc-based linuxes) so glibc should
too

i should add that haiku did previously have this same bug but someone
discovered and reported it and it got fixed. i'm not sure how it was
discovered though

i personally don't really care about the checks for anything but
INADDR_ANY, since i also can't foresee a situation where the others are a
problem, but INADDR_ANY is a problem, so can't a simple "if (addr ==
INADDR_ANY) return NULL;" or whatever be added?

From forum: Sourceware - libc-alpha

PATCH: Define bit_SSE2 and index_SSE2

2009-12-13T11:24:29Z

On Wed, Dec 09, 2009 at 11:18:18AM -0800, H.J. Lu wrote:
> Hi,
>
> This patch defines bit_XXX and index_XXX and use them to check processor
> feature in assembly code. It can prevent typos in processor feature
> check.
>

Hi,

This patch adds bit_SSE2 and index_SSE2 for 32bit assembly codes.

Thanks.

H.J.
---
2009-12-13 H.J. Lu <hongjiu.lu@...>

* sysdeps/i386/i686/multiarch/strcspn.S Include <init-arch.h>
instead of <ifunc-defines.h>. Use bit_XXX and index_XXX to
check processor feature.
* sysdeps/i386/i686/multiarch/strlen.S: Likewise.
* sysdeps/i386/i686/multiarch/strspn.S: Likewise.

* sysdeps/x86_64/multiarch/init-arch.h (bit_SSE2): New.
(index_SSE2): Likewise.

diff --git a/sysdeps/i386/i686/multiarch/strcspn.S b/sysdeps/i386/i686/multiarch/strcspn.S
index 73e7eb4..b2310e4 100644
--- a/sysdeps/i386/i686/multiarch/strcspn.S
+++ b/sysdeps/i386/i686/multiarch/strcspn.S
@@ -23,7 +23,7 @@
#ifdef HAVE_SSE4_SUPPORT

#include <sysdep.h>
-#include <ifunc-defines.h>
+#include <init-arch.h>

#ifdef USE_AS_STRPBRK
#define STRCSPN_SSE42 __strpbrk_sse42
@@ -64,7 +64,7 @@ ENTRY(STRCSPN)
jne 1f
call __init_cpu_features
1: leal STRCSPN_IA32@GOTOFF(%ebx), %eax
- testl $(1<<20), CPUID_OFFSET+COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET+__cpu_features@GOTOFF(%ebx)
+ testl $bit_SSE4_2, CPUID_OFFSET+index_SSE4_2+__cpu_features@GOTOFF(%ebx)
jz 2f
leal STRCSPN_SSE42@GOTOFF(%ebx), %eax
2: popl %ebx
@@ -80,7 +80,7 @@ ENTRY(STRCSPN)
jne 1f
call __init_cpu_features
1: leal STRCSPN_IA32, %eax
- testl $(1<<20), CPUID_OFFSET+COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET+__cpu_features
+ testl $bit_SSE4_2, CPUID_OFFSET+index_SSE4_2+__cpu_features
jz 2f
leal STRCSPN_SSE42, %eax
2: ret
diff --git a/sysdeps/i386/i686/multiarch/strlen.S b/sysdeps/i386/i686/multiarch/strlen.S
index 0c1e864..9786add 100644
--- a/sysdeps/i386/i686/multiarch/strlen.S
+++ b/sysdeps/i386/i686/multiarch/strlen.S
@@ -19,7 +19,7 @@
02111-1307 USA. */

#include <sysdep.h>
-#include <ifunc-defines.h>
+#include <init-arch.h>

/* Define multiple versions only for the definition in libc and for the
DSO. In static binaries, we need strlen before the initialization
@@ -46,7 +46,7 @@ ENTRY(strlen)
jne 1f
call __init_cpu_features
1: leal __strlen_ia32@GOTOFF(%ebx), %eax
- testl $(1<<26), CPUID_OFFSET+COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET+__cpu_features@GOTOFF(%ebx)
+ testl $bit_SSE2, CPUID_OFFSET+index_SSE2+__cpu_features@GOTOFF(%ebx)
jz 2f
leal __strlen_sse2@GOTOFF(%ebx), %eax
2: popl %ebx
diff --git a/sysdeps/i386/i686/multiarch/strspn.S b/sysdeps/i386/i686/multiarch/strspn.S
index f306d2d..7fe6f97 100644
--- a/sysdeps/i386/i686/multiarch/strspn.S
+++ b/sysdeps/i386/i686/multiarch/strspn.S
@@ -23,7 +23,7 @@
#ifdef HAVE_SSE4_SUPPORT

#include <sysdep.h>
-#include <ifunc-defines.h>
+#include <init-arch.h>

/* Define multiple versions only for the definition in libc. */
#ifndef NOT_IN_libc
@@ -49,7 +49,7 @@ ENTRY(strspn)
jne 1f
call __init_cpu_features
1: leal __strspn_ia32@GOTOFF(%ebx), %eax
- testl $(1<<20), CPUID_OFFSET+COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET+__cpu_features@GOTOFF(%ebx)
+ testl $bit_SSE4_2, CPUID_OFFSET+index_SSE4_2+__cpu_features@GOTOFF(%ebx)
jz 2f
leal __strspn_sse42@GOTOFF(%ebx), %eax
2: popl %ebx
@@ -65,7 +65,7 @@ ENTRY(strspn)
jne 1f
call __init_cpu_features
1: leal __strspn_ia32, %eax
- testl $(1<<20), CPUID_OFFSET+COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET+__cpu_features
+ testl $index_SSE2, CPUID_OFFSET+index_SSE4_2+__cpu_features
jz 2f
leal __strspn_sse42, %eax
2: ret
diff --git a/sysdeps/x86_64/multiarch/init-arch.h b/sysdeps/x86_64/multiarch/init-arch.h
index 9e8915f..0f8f77a 100644
--- a/sysdeps/x86_64/multiarch/init-arch.h
+++ b/sysdeps/x86_64/multiarch/init-arch.h
@@ -20,9 +20,11 @@

#include <ifunc-defines.h>

+#define bit_SSE2 (1 << 26)
#define bit_SSSE3 (1 << 9)
#define bit_SSE4_2 (1 << 20)

+#define index_SSE2 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
#define index_SSSE3 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
#define index_SSE4_2 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET

From forum: Sourceware - libc-alpha

Re: reverse look up of INADDR_ANY, etc.

2009-12-13T09:50:41Z

On 11/23/2009 08:33 PM, danielcavanagh@... wrote:
> is this consider a bug? if so, do you need me to file a report?

Not a bug. It's "garbage in, garbage out". Just like any other invalid
address you will get the correct result but it'll take it's time for the
timeout to kick in.

You'll have to show me a valid program which can regularly create such
calls t convince me otherwise.

--
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖

From forum: Sourceware - libc-alpha

[Bug libc/11082] fclose(f); ftell(f); results in accesses on free'd memory

2009-12-11T05:24:48Z

------- Additional Comments From pasky at suse dot cz 2009-12-11 13:24 -------
Not really, file descriptor is something different from file stream. Something
can close the file descriptor while keeping the stream alive (e.g. by obtaining
it from stream using fileno() or by mass-closing all file descriptors), then
this error would trigger.

But if you close the stream itself, there's no stream _of_ the underlying file
descriptor. After you fclose(), the FILE* value is totally invalid and you must
not use it anymore.

--
What |Removed |Added
----------------------------------------------------------------------------
Status|REOPENED |RESOLVED
Resolution| |INVALID

http://sourceware.org/bugzilla/show_bug.cgi?id=11082

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

[Bug libc/11082] fclose(f); ftell(f); results in accesses on free'd memory

2009-12-11T05:12:46Z

------- Additional Comments From jur dot pauw at gmail dot com 2009-12-11 13:12 -------
(In reply to comment #1)
> ftell(3) needs a valid FILE pointer, not one already invalidated by fclose(3)
> call. For details see fclose(3) description that says: "any further access
to
> the stream results in undefined behaviour".

This implies that ftell its behaviour is not Posix compliant, see
http://www.opengroup.org/onlinepubs/009695399/functions/ftell.html

ERRORS
The ftell() [CX] and ftello() functions shall fail if:

[EBADF]
[CX] The file descriptor underlying stream is not an open file descriptor.

--
What |Removed |Added
----------------------------------------------------------------------------
Status|RESOLVED |REOPENED
Resolution|INVALID |

http://sourceware.org/bugzilla/show_bug.cgi?id=11082

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

[Bug libc/11082] fclose(f); ftell(f); results in accesses on free'd memory

2009-12-11T03:10:15Z

------- Additional Comments From ldv at altlinux dot org 2009-12-11 11:10 -------
ftell(3) needs a valid FILE pointer, not one already invalidated by fclose(3)
call. For details see fclose(3) description that says: "any further access to
the stream results in undefined behaviour".

--
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution| |INVALID

http://sourceware.org/bugzilla/show_bug.cgi?id=11082

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

[Bug libc/11082] New: fclose(f); ftell(f); results in accesses on free'd memory

2009-12-11T02:21:56Z

I used valgrind to find the origin of a crash. It shows that an ftell on a
closed file performs reads and writes on memory free'd by fclose. The
following small program (bug.c) reproduces the problem.

#include <stdio.h>

int main(int argc, char **argv)
{
FILE *f = fopen("bug.c", "r");
fclose(f);
ftell(f);
}

compiler/kernel/libc versions:

gcc version 3.4.4

Linux c3svr30 2.6.9-89.0.16.ELsmp #1 SMP Tue Oct 27 03:51:52 EDT 2009 x86_64
x86_64 x86_64 GNU/Linux

/lib/tls/libc-2.3.4.so
GNU C Library stable release version 2.3.4, by Roland McGrath et al.
Copyright (C) 2005 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.
Compiled by GNU CC version 3.4.6 20060404 (Red Hat 3.4.6-8).
Compiled on a Linux 2.4.20 system on 2007-09-12.
Available extensions:
GNU libio by Per Bothner
crypt add-on version 2.1 by Michael Glad and others
Native POSIX Threads Library by Ulrich Drepper et al
RT using linux kernel aio
The C stubs add-on version 2.1.2.
BIND-8.2.3-T5B
NIS(YP)/NIS+ NSS modules 0.19 by Thorsten Kukuk
Glibc-2.0 compatibility add-on by Cristian Gafton
GNU Libidn by Simon Josefsson
Thread-local storage support included.

--
Summary: fclose(f); ftell(f); results in accesses on free'd
memory
Product: glibc
Version: 2.3.4
Status: NEW
Severity: normal
Priority: P2
Component: libc
AssignedTo: drepper at redhat dot com
ReportedBy: jur dot pauw at gmail dot com
CC: glibc-bugs at sources dot redhat dot com

http://sourceware.org/bugzilla/show_bug.cgi?id=11082

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs

[Bug nis/11000] _nss_nis_gethostbyname4_r: errnop and herrnop are not always set

2009-12-10T15:45:36Z

------- Additional Comments From drepper at redhat dot com 2009-12-10 23:45 -------
Since you, as always, don't provide the details of the problem you're trying to
fix I had to guess. I checked in a different and in any case more complete patch.

--
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution| |FIXED

http://sourceware.org/bugzilla/show_bug.cgi?id=11000

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

From forum: Sourceware - glibc-bugs