diff mbox

[v12,1/3] Y2038: provide size of default time_t for target architecture

Message ID 20181022050002.32621-1-albert.aribaud@3adev.fr
State New
Headers show

Commit Message

Albert ARIBAUD Oct. 22, 2018, 5 a.m. UTC
To determine whether the default time_t interfaces are 32-bit
and so need conversions, or are 64-bit and so are compatible
with the internal 64-bit type without conversions, a macro
giving the size of the  default time_t is also required.
This macro is called __TIMESIZE.

This macro can then be used instead of __WORDSIZE in msq-pad.h
and shm-pad.h files, which in turn allows removing their x86
variants, and in sem-pad.h files but keeping the x86 variant.

This patch was tested by running 'make check' on branch master
then applying this patch and running 'make check' again, and
checking that both 'make check' yield identical results.
This was done on x86_64-linux-gnu and i686-linux-gnu.

* bits/timesize.h: New file.
* stdlib/Makefile (headers): Add bits/timesize.h.
* sysdeps/unix/sysv/linux/bits/msq-pad.h: Replace __WORDSIZE with __TIMESIZE.
* sysdeps/unix/sysv/linux/bits/sem-pad.h: Likewise.
* sysdeps/unix/sysv/linux/bits/shm-pad.h: Likewise.
* sysdeps/unix/sysv/linux/hppa/bits/msq-pad.h: Likewise.
* sysdeps/unix/sysv/linux/hppa/bits/sem-pad.h: Likewise.
* sysdeps/unix/sysv/linux/hppa/bits/shm-pad.h: Likewise.
* sysdeps/unix/sysv/linux/mips/bits/msq-pad.h: Likewise.
* sysdeps/unix/sysv/linux/powerpc/bits/msq-pad.h: Likewise.
* sysdeps/unix/sysv/linux/powerpc/bits/sem-pad.h: Likewise.
* sysdeps/unix/sysv/linux/powerpc/bits/shm-pad.h: Likewise.
* sysdeps/unix/sysv/linux/sparc/bits/msq-pad.h: Likewise.
* sysdeps/unix/sysv/linux/sparc/bits/sem-pad.h: Likewise.
* sysdeps/unix/sysv/linux/sparc/bits/shm-pad.h: Likewise.
* sysdeps/unix/sysv/linux/x86/bits/msq-pad.h: Delete file.
* sysdeps/unix/sysv/linux/x86/bits/shm-pad.h: Likewise.
* sysdeps/unix/sysv/linux/x86/bits/timesize.h: New file.
---
 .../linux/x86/bits/shm-pad.h => bits/timesize.h  | 16 ++++------------
 stdlib/Makefile                                  |  2 +-
 sysdeps/unix/sysv/linux/bits/msq-pad.h           |  4 ++--
 sysdeps/unix/sysv/linux/bits/sem-pad.h           |  4 ++--
 sysdeps/unix/sysv/linux/bits/shm-pad.h           |  4 ++--
 sysdeps/unix/sysv/linux/hppa/bits/msq-pad.h      |  4 ++--
 sysdeps/unix/sysv/linux/hppa/bits/sem-pad.h      |  4 ++--
 sysdeps/unix/sysv/linux/hppa/bits/shm-pad.h      |  6 +++---
 sysdeps/unix/sysv/linux/mips/bits/msq-pad.h      |  6 +++---
 sysdeps/unix/sysv/linux/powerpc/bits/msq-pad.h   |  4 ++--
 sysdeps/unix/sysv/linux/powerpc/bits/sem-pad.h   |  4 ++--
 sysdeps/unix/sysv/linux/powerpc/bits/shm-pad.h   |  6 +++---
 sysdeps/unix/sysv/linux/sparc/bits/msq-pad.h     |  4 ++--
 sysdeps/unix/sysv/linux/sparc/bits/sem-pad.h     |  4 ++--
 sysdeps/unix/sysv/linux/sparc/bits/shm-pad.h     |  4 ++--
 .../linux/x86/bits/{msq-pad.h => timesize.h}     | 15 ++++++---------
 16 files changed, 40 insertions(+), 51 deletions(-)
 rename sysdeps/unix/sysv/linux/x86/bits/shm-pad.h => bits/timesize.h (68%)
 rename sysdeps/unix/sysv/linux/x86/bits/{msq-pad.h => timesize.h} (74%)

Comments

Joseph Myers Oct. 22, 2018, 2:40 p.m. UTC | #1
How does this patch differ from version 11, which I already approved?

When a patch is approved that does not depend on other unapproved patches, 
the general expectation is that it will be committed, unless you find some 
other issue with it so you no longer think it's suitable for commit as-is 
and instead want to propose a new changed version.  Then a shorter patch 
series would be posted, omitting the committed patches.

In general, when submitting a revised patch or patch series it's a good 
idea to make clear what has changed from the previous version (this should 
go in the patch submission emails, but not the commit message; the Linux 
kernel convention is that after the commit message you can have a "---" 
line, then any additional information not intended for the commit message, 
such as about changes from previous patch versions, before the patch 
itself).
Albert ARIBAUD Oct. 22, 2018, 7:28 p.m. UTC | #2
Hi Joseph,

On Mon, 22 Oct 2018 14:40:24 +0000, Joseph Myers
<joseph@codesourcery.com> wrote :

> How does this patch differ from version 11, which I already approved?
> 
> When a patch is approved that does not depend on other unapproved patches, 
> the general expectation is that it will be committed, unless you find some 
> other issue with it so you no longer think it's suitable for commit as-is 
> and instead want to propose a new changed version.  Then a shorter patch 
> series would be posted, omitting the committed patches.

My bad - I'm used to resubmitting series entirely even when only some
patches need fixing. I'll omit OK'd patches from now on.

> In general, when submitting a revised patch or patch series it's a good 
> idea to make clear what has changed from the previous version (this should 
> go in the patch submission emails, but not the commit message; the Linux 
> kernel convention is that after the commit message you can have a "---" 
> line, then any additional information not intended for the commit message, 
> such as about changes from previous patch versions, before the patch 
> itself).

I'll add patch history when applicable.

Cordialement,
Albert ARIBAUD
3ADEV
Joseph Myers Oct. 22, 2018, 9:44 p.m. UTC | #3
On Mon, 22 Oct 2018, Albert ARIBAUD wrote:

> > When a patch is approved that does not depend on other unapproved patches, 
> > the general expectation is that it will be committed, unless you find some 
> > other issue with it so you no longer think it's suitable for commit as-is 
> > and instead want to propose a new changed version.  Then a shorter patch 
> > series would be posted, omitting the committed patches.
> 
> My bad - I'm used to resubmitting series entirely even when only some
> patches need fixing. I'll omit OK'd patches from now on.

You need to *commit* the approved patches to master, and then send a 
series that applies to current master with the approved patches included.  
If a patch can't be committed because it depends on an earlier unapproved 
patch (or because the approval was explicitly conditional on a whole 
series going in at once), then it's appropriate still to include it in the 
series, but to state explicitly that it's been approved and is unchanged 
from the approved version (if that's the case).
Joseph Myers Oct. 22, 2018, 9:52 p.m. UTC | #4
On Mon, 22 Oct 2018, Joseph Myers wrote:

> You need to *commit* the approved patches to master, and then send a 

To be clear: my expectation, after approving patch 1/3, version 11, was 
that the next steps would be: (a) commit patch 1 to master and push that 
commit; (b) send a series that contains only two patches, where patch 1/2 
applies cleanly to master and patch 2/2 applies to (master + patch 1/2).
Albert ARIBAUD Oct. 22, 2018, 9:59 p.m. UTC | #5
Hi Joseph,

On Mon, 22 Oct 2018 21:44:38 +0000, Joseph Myers
<joseph@codesourcery.com> wrote :

> On Mon, 22 Oct 2018, Albert ARIBAUD wrote:
> 
> > > When a patch is approved that does not depend on other unapproved patches, 
> > > the general expectation is that it will be committed, unless you find some 
> > > other issue with it so you no longer think it's suitable for commit as-is 
> > > and instead want to propose a new changed version.  Then a shorter patch 
> > > series would be posted, omitting the committed patches.  
> > 
> > My bad - I'm used to resubmitting series entirely even when only some
> > patches need fixing. I'll omit OK'd patches from now on.  
> 
> You need to *commit* the approved patches to master, and then send a 
> series that applies to current master with the approved patches included.  
> If a patch can't be committed because it depends on an earlier unapproved 
> patch (or because the approval was explicitly conditional on a whole 
> series going in at once), then it's appropriate still to include it in the 
> series, but to state explicitly that it's been approved and is unchanged 
> from the approved version (if that's the case).

Just to make sure: when you say "you need to commit the approved
patches to master", are you saying that I am in capacity to commit to
branch master of the official glibc repository, and that after your
OK'ing of commit 1/3 of this series, I should now commit this patch to
branch master of glibc myself?

I'm genuinely surprised that the glibc project would allow this, but if
that's how it works, then fine.

Cordialement,
Albert ARIBAUD
3ADEV
Joseph Myers Oct. 22, 2018, 10:09 p.m. UTC | #6
On Mon, 22 Oct 2018, Albert ARIBAUD wrote:

> Just to make sure: when you say "you need to commit the approved
> patches to master", are you saying that I am in capacity to commit to
> branch master of the official glibc repository, and that after your
> OK'ing of commit 1/3 of this series, I should now commit this patch to
> branch master of glibc myself?

Yes.

https://sourceware.org/glibc/wiki/Committer%20checklist

(In this case, no NEWS or Bugzilla updates are needed, but you do need to 
make sure to update ChangeLog appropriately.)
Florian Weimer Oct. 23, 2018, 9:14 a.m. UTC | #7
* Albert ARIBAUD:

> Hi Joseph,
>
> On Mon, 22 Oct 2018 21:44:38 +0000, Joseph Myers
> <joseph@codesourcery.com> wrote :
>
>> On Mon, 22 Oct 2018, Albert ARIBAUD wrote:
>> 
>> > > When a patch is approved that does not depend on other unapproved patches, 
>> > > the general expectation is that it will be committed, unless you find some 
>> > > other issue with it so you no longer think it's suitable for commit as-is 
>> > > and instead want to propose a new changed version.  Then a shorter patch 
>> > > series would be posted, omitting the committed patches.  
>> > 
>> > My bad - I'm used to resubmitting series entirely even when only some
>> > patches need fixing. I'll omit OK'd patches from now on.  
>> 
>> You need to *commit* the approved patches to master, and then send a 
>> series that applies to current master with the approved patches included.  
>> If a patch can't be committed because it depends on an earlier unapproved 
>> patch (or because the approval was explicitly conditional on a whole 
>> series going in at once), then it's appropriate still to include it in the 
>> series, but to state explicitly that it's been approved and is unchanged 
>> from the approved version (if that's the case).
>
> Just to make sure: when you say "you need to commit the approved
> patches to master", are you saying that I am in capacity to commit to
> branch master of the official glibc repository, and that after your
> OK'ing of commit 1/3 of this series, I should now commit this patch to
> branch master of glibc myself?
>
> I'm genuinely surprised that the glibc project would allow this, but if
> that's how it works, then fine.

If you still have doubts about the commit message or the ChangeLog
entry, I'm happy to review it for you prior to making the change
official.

Thanks,
Florian
Albert ARIBAUD Oct. 23, 2018, 7:16 p.m. UTC | #8
Hi Florian,

Le Tue, 23 Oct 2018 11:14:24 +0200, Florian Weimer <fweimer@redhat.com>
a écrit :

> If you still have doubts about the commit message or the ChangeLog
> entry, I'm happy to review it for you prior to making the change
> official.

Thanks. I have prepared my master branch locally, and pushed it on the
glibc repo as the branch aaribaud/master for you to have a look at.
Right now, I am running a fast make check as per the checklist.

> Thanks,
> Florian

Cordialement,
Albert ARIBAUD
3ADEV
Florian Weimer Oct. 24, 2018, 8:01 a.m. UTC | #9
* Albert ARIBAUD:

> Hi Florian,
>
> Le Tue, 23 Oct 2018 11:14:24 +0200, Florian Weimer <fweimer@redhat.com>
> a écrit :
>
>> If you still have doubts about the commit message or the ChangeLog
>> entry, I'm happy to review it for you prior to making the change
>> official.
>
> Thanks. I have prepared my master branch locally, and pushed it on the
> glibc repo as the branch aaribaud/master for you to have a look at.
> Right now, I am running a fast make check as per the checklist.

I think the ChangeLog should list the names of macros being changed, so
something like thiss:

	* sysdeps/unix/sysv/linux/mips/bits/msq-pad.h
        (__MSQ_PAD_AFTER_TIME, __SHM_PAD_BETWEEN_TIME_AND_SEGSZ): Use
        __TIMESIZE instead of __WORDSIZE.

Otherwise, looks good.

Thanks,
Florian
Albert ARIBAUD Oct. 24, 2018, 8:34 a.m. UTC | #10
Hi Florian,

On Wed, 24 Oct 2018 10:01:04 +0200, Florian Weimer <fweimer@redhat.com>
wrote :

> * Albert ARIBAUD:
> 
> > Hi Florian,
> >
> > Le Tue, 23 Oct 2018 11:14:24 +0200, Florian Weimer <fweimer@redhat.com>
> > a écrit :
> >  
> >> If you still have doubts about the commit message or the ChangeLog
> >> entry, I'm happy to review it for you prior to making the change
> >> official.  
> >
> > Thanks. I have prepared my master branch locally, and pushed it on the
> > glibc repo as the branch aaribaud/master for you to have a look at.
> > Right now, I am running a fast make check as per the checklist.  
> 
> I think the ChangeLog should list the names of macros being changed, so
> something like thiss:
> 
> 	* sysdeps/unix/sysv/linux/mips/bits/msq-pad.h
>         (__MSQ_PAD_AFTER_TIME, __SHM_PAD_BETWEEN_TIME_AND_SEGSZ): Use
>         __TIMESIZE instead of __WORDSIZE.

Thanks, I've rewritten the ChangeLog file (and the commit message
changelog part for homogeneity) based on your example -- listing the
right macro(s) for each header file of course.

> Otherwise, looks good.

Thanks for your review.

> Thanks,
> Florian

Cordialement,
Albert ARIBAUD
3ADEV
Florian Weimer Oct. 24, 2018, 9:24 a.m. UTC | #11
* Albert ARIBAUD:

> Hi Florian,
>
> On Wed, 24 Oct 2018 10:01:04 +0200, Florian Weimer <fweimer@redhat.com>
> wrote :
>
>> * Albert ARIBAUD:
>> 
>> > Hi Florian,
>> >
>> > Le Tue, 23 Oct 2018 11:14:24 +0200, Florian Weimer <fweimer@redhat.com>
>> > a écrit :
>> >  
>> >> If you still have doubts about the commit message or the ChangeLog
>> >> entry, I'm happy to review it for you prior to making the change
>> >> official.  
>> >
>> > Thanks. I have prepared my master branch locally, and pushed it on the
>> > glibc repo as the branch aaribaud/master for you to have a look at.
>> > Right now, I am running a fast make check as per the checklist.  
>> 
>> I think the ChangeLog should list the names of macros being changed, so
>> something like thiss:
>> 
>> 	* sysdeps/unix/sysv/linux/mips/bits/msq-pad.h
>>         (__MSQ_PAD_AFTER_TIME, __SHM_PAD_BETWEEN_TIME_AND_SEGSZ): Use
>>         __TIMESIZE instead of __WORDSIZE.
>
> Thanks, I've rewritten the ChangeLog file (and the commit message
> changelog part for homogeneity) based on your example -- listing the
> right macro(s) for each header file of course.

Please update the data in the ChangeLog and the commit (git commit
--amend --date="`date`"), rebase to current master, and then you should
be okay to push to the official master branch.

(I assume this patch has actually been reviewed, I have not checked
that. 8-)

Thanks,
Florian
Albert ARIBAUD Oct. 24, 2018, 9:33 a.m. UTC | #12
Hi Florian,

On Wed, 24 Oct 2018 11:24:25 +0200, Florian Weimer <fweimer@redhat.com>
wrote :

> * Albert ARIBAUD:
> 
> > Hi Florian,
> >
> > On Wed, 24 Oct 2018 10:01:04 +0200, Florian Weimer <fweimer@redhat.com>
> > wrote :
> >  
> >> * Albert ARIBAUD:
> >>   
>  [...]  
>  [...]  
>  [...]  
> >> 
> >> I think the ChangeLog should list the names of macros being changed, so
> >> something like thiss:
> >> 
> >> 	* sysdeps/unix/sysv/linux/mips/bits/msq-pad.h
> >>         (__MSQ_PAD_AFTER_TIME, __SHM_PAD_BETWEEN_TIME_AND_SEGSZ): Use
> >>         __TIMESIZE instead of __WORDSIZE.  
> >
> > Thanks, I've rewritten the ChangeLog file (and the commit message
> > changelog part for homogeneity) based on your example -- listing the
> > right macro(s) for each header file of course.  
> 
> Please update the data in the ChangeLog and the commit (git commit
> --amend --date="`date`"), rebase to current master, and then you should
> be okay to push to the official master branch.

Ok.

> (I assume this patch has actually been reviewed, I have not checked
> that. 8-)

Yes it has. :)

> Thanks,
> Florian

Cordialement,
Albert ARIBAUD
3ADEV
diff mbox

Patch

diff --git a/sysdeps/unix/sysv/linux/x86/bits/shm-pad.h b/bits/timesize.h
similarity index 68%
rename from sysdeps/unix/sysv/linux/x86/bits/shm-pad.h
rename to bits/timesize.h
index 277a70fe6c..cc47ff165f 100644
--- a/sysdeps/unix/sysv/linux/x86/bits/shm-pad.h
+++ b/bits/timesize.h
@@ -1,4 +1,4 @@ 
-/* Define where padding goes in struct shmid_ds.  x86 version.
+/* Bit size of the time_t type at glibc build time, general case.
    Copyright (C) 2018 Free Software Foundation, Inc.
    This file is part of the GNU C Library.
 
@@ -16,15 +16,7 @@ 
    License along with the GNU C Library; if not, see
    <http://www.gnu.org/licenses/>.  */
 
-#ifndef _SYS_SHM_H
-# error "Never use <bits/shm-pad.h> directly; include <sys/shm.h> instead."
-#endif
+#include <bits/wordsize.h>
 
-#ifdef __x86_64__
-# define __SHM_PAD_AFTER_TIME 0
-#else
-# define __SHM_PAD_AFTER_TIME 1
-#endif
-#define __SHM_PAD_BEFORE_TIME 0
-#define __SHM_SEGSZ_AFTER_TIME 0
-#define __SHM_PAD_BETWEEN_TIME_AND_SEGSZ 0
+/* Size in bits of the 'time_t' type of the default ABI.  */
+#define __TIMESIZE	__WORDSIZE
diff --git a/stdlib/Makefile b/stdlib/Makefile
index 01194bbf7c..1773a341b5 100644
--- a/stdlib/Makefile
+++ b/stdlib/Makefile
@@ -24,7 +24,7 @@  include ../Makeconfig
 
 headers	:= stdlib.h bits/stdlib.h bits/stdlib-ldbl.h bits/stdlib-float.h      \
 	   monetary.h bits/monetary-ldbl.h				      \
-	   inttypes.h stdint.h bits/wordsize.h				      \
+	   inttypes.h stdint.h bits/wordsize.h bits/timesize.h		      \
 	   errno.h sys/errno.h bits/errno.h bits/types/error_t.h	      \
 	   ucontext.h sys/ucontext.h bits/indirect-return.h		      \
 	   alloca.h fmtmsg.h						      \
diff --git a/sysdeps/unix/sysv/linux/bits/msq-pad.h b/sysdeps/unix/sysv/linux/bits/msq-pad.h
index 97dee0beff..53f367dfa2 100644
--- a/sysdeps/unix/sysv/linux/bits/msq-pad.h
+++ b/sysdeps/unix/sysv/linux/bits/msq-pad.h
@@ -20,12 +20,12 @@ 
 # error "Never use <bits/msq-pad.h> directly; include <sys/msg.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 /* On most architectures, padding goes after time fields for 32-bit
    systems and is omitted for 64-bit systems.  Some architectures pad
    before time fields instead, or omit padding despite being
    32-bit.  */
 
-#define __MSQ_PAD_AFTER_TIME (__WORDSIZE == 32)
+#define __MSQ_PAD_AFTER_TIME (__TIMESIZE == 32)
 #define __MSQ_PAD_BEFORE_TIME 0
diff --git a/sysdeps/unix/sysv/linux/bits/sem-pad.h b/sysdeps/unix/sysv/linux/bits/sem-pad.h
index c57fea4ac8..488b591d5a 100644
--- a/sysdeps/unix/sysv/linux/bits/sem-pad.h
+++ b/sysdeps/unix/sysv/linux/bits/sem-pad.h
@@ -20,7 +20,7 @@ 
 # error "Never use <bits/sem-pad.h> directly; include <sys/sem.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 /* On most architectures, padding goes after time fields for 32-bit
    systems and is omitted for 64-bit systems.  Some architectures pad
@@ -29,5 +29,5 @@ 
    used for struct semid64_ds in <asm/sembuf.h>, as glibc does not do
    layout conversions for this structure.  */
 
-#define __SEM_PAD_AFTER_TIME (__WORDSIZE == 32)
+#define __SEM_PAD_AFTER_TIME (__TIMESIZE == 32)
 #define __SEM_PAD_BEFORE_TIME 0
diff --git a/sysdeps/unix/sysv/linux/bits/shm-pad.h b/sysdeps/unix/sysv/linux/bits/shm-pad.h
index 9233c95a9f..d3cc5ebde6 100644
--- a/sysdeps/unix/sysv/linux/bits/shm-pad.h
+++ b/sysdeps/unix/sysv/linux/bits/shm-pad.h
@@ -20,7 +20,7 @@ 
 # error "Never use <bits/shm-pad.h> directly; include <sys/shm.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 /* On most architectures, padding goes after time fields for 32-bit
    systems and is omitted for 64-bit systems.  Some architectures pad
@@ -31,7 +31,7 @@ 
    layout used for struct shmid64_ds in <asm/shmbuf.h>, as glibc does
    not do layout conversions for this structure.  */
 
-#define __SHM_PAD_AFTER_TIME (__WORDSIZE == 32)
+#define __SHM_PAD_AFTER_TIME (__TIMESIZE == 32)
 #define __SHM_PAD_BEFORE_TIME 0
 #define __SHM_SEGSZ_AFTER_TIME 0
 #define __SHM_PAD_BETWEEN_TIME_AND_SEGSZ 0
diff --git a/sysdeps/unix/sysv/linux/hppa/bits/msq-pad.h b/sysdeps/unix/sysv/linux/hppa/bits/msq-pad.h
index 10b26b7ea6..05db3fd65b 100644
--- a/sysdeps/unix/sysv/linux/hppa/bits/msq-pad.h
+++ b/sysdeps/unix/sysv/linux/hppa/bits/msq-pad.h
@@ -20,7 +20,7 @@ 
 # error "Never use <bits/msq-pad.h> directly; include <sys/msg.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __MSQ_PAD_AFTER_TIME 0
-#define __MSQ_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __MSQ_PAD_BEFORE_TIME (__TIMESIZE == 32)
diff --git a/sysdeps/unix/sysv/linux/hppa/bits/sem-pad.h b/sysdeps/unix/sysv/linux/hppa/bits/sem-pad.h
index a938c2c92b..f250f64426 100644
--- a/sysdeps/unix/sysv/linux/hppa/bits/sem-pad.h
+++ b/sysdeps/unix/sysv/linux/hppa/bits/sem-pad.h
@@ -20,7 +20,7 @@ 
 # error "Never use <bits/sem-pad.h> directly; include <sys/sem.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __SEM_PAD_AFTER_TIME 0
-#define __SEM_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __SEM_PAD_BEFORE_TIME (__TIMESIZE == 32)
diff --git a/sysdeps/unix/sysv/linux/hppa/bits/shm-pad.h b/sysdeps/unix/sysv/linux/hppa/bits/shm-pad.h
index 6a9c5ddce2..8764bde08b 100644
--- a/sysdeps/unix/sysv/linux/hppa/bits/shm-pad.h
+++ b/sysdeps/unix/sysv/linux/hppa/bits/shm-pad.h
@@ -20,9 +20,9 @@ 
 # error "Never use <bits/shm-pad.h> directly; include <sys/shm.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __SHM_PAD_AFTER_TIME 0
-#define __SHM_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __SHM_PAD_BEFORE_TIME (__TIMESIZE == 32)
 #define __SHM_SEGSZ_AFTER_TIME 1
-#define __SHM_PAD_BETWEEN_TIME_AND_SEGSZ (__WORDSIZE == 32)
+#define __SHM_PAD_BETWEEN_TIME_AND_SEGSZ (__TIMESIZE == 32)
diff --git a/sysdeps/unix/sysv/linux/mips/bits/msq-pad.h b/sysdeps/unix/sysv/linux/mips/bits/msq-pad.h
index 1ddfb0d271..530babe3f9 100644
--- a/sysdeps/unix/sysv/linux/mips/bits/msq-pad.h
+++ b/sysdeps/unix/sysv/linux/mips/bits/msq-pad.h
@@ -20,12 +20,12 @@ 
 # error "Never use <bits/msq-pad.h> directly; include <sys/msg.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #ifdef __MIPSEL__
-# define __MSQ_PAD_AFTER_TIME (__WORDSIZE == 32)
+# define __MSQ_PAD_AFTER_TIME (__TIMESIZE == 32)
 # define __MSQ_PAD_BEFORE_TIME 0
 #else
 # define __MSQ_PAD_AFTER_TIME 0
-# define __MSQ_PAD_BEFORE_TIME (__WORDSIZE == 32)
+# define __MSQ_PAD_BEFORE_TIME (__TIMESIZE == 32)
 #endif
diff --git a/sysdeps/unix/sysv/linux/powerpc/bits/msq-pad.h b/sysdeps/unix/sysv/linux/powerpc/bits/msq-pad.h
index 9680176abb..d19dfd5126 100644
--- a/sysdeps/unix/sysv/linux/powerpc/bits/msq-pad.h
+++ b/sysdeps/unix/sysv/linux/powerpc/bits/msq-pad.h
@@ -20,7 +20,7 @@ 
 # error "Never use <bits/msq-pad.h> directly; include <sys/msg.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __MSQ_PAD_AFTER_TIME 0
-#define __MSQ_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __MSQ_PAD_BEFORE_TIME (__TIMESIZE == 32)
diff --git a/sysdeps/unix/sysv/linux/powerpc/bits/sem-pad.h b/sysdeps/unix/sysv/linux/powerpc/bits/sem-pad.h
index d4c9a93956..d7571cc755 100644
--- a/sysdeps/unix/sysv/linux/powerpc/bits/sem-pad.h
+++ b/sysdeps/unix/sysv/linux/powerpc/bits/sem-pad.h
@@ -20,7 +20,7 @@ 
 # error "Never use <bits/sem-pad.h> directly; include <sys/sem.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __SEM_PAD_AFTER_TIME 0
-#define __SEM_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __SEM_PAD_BEFORE_TIME (__TIMESIZE == 32)
diff --git a/sysdeps/unix/sysv/linux/powerpc/bits/shm-pad.h b/sysdeps/unix/sysv/linux/powerpc/bits/shm-pad.h
index 69a7e4a648..c8c2d44775 100644
--- a/sysdeps/unix/sysv/linux/powerpc/bits/shm-pad.h
+++ b/sysdeps/unix/sysv/linux/powerpc/bits/shm-pad.h
@@ -20,9 +20,9 @@ 
 # error "Never use <bits/shm-pad.h> directly; include <sys/shm.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __SHM_PAD_AFTER_TIME 0
-#define __SHM_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __SHM_PAD_BEFORE_TIME (__TIMESIZE == 32)
 #define __SHM_SEGSZ_AFTER_TIME 1
-#define __SHM_PAD_BETWEEN_TIME_AND_SEGSZ (__WORDSIZE == 32)
+#define __SHM_PAD_BETWEEN_TIME_AND_SEGSZ (__TIMESIZE == 32)
diff --git a/sysdeps/unix/sysv/linux/sparc/bits/msq-pad.h b/sysdeps/unix/sysv/linux/sparc/bits/msq-pad.h
index 3dfeb07a48..72324e8428 100644
--- a/sysdeps/unix/sysv/linux/sparc/bits/msq-pad.h
+++ b/sysdeps/unix/sysv/linux/sparc/bits/msq-pad.h
@@ -20,7 +20,7 @@ 
 # error "Never use <bits/msq-pad.h> directly; include <sys/msg.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __MSQ_PAD_AFTER_TIME 0
-#define __MSQ_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __MSQ_PAD_BEFORE_TIME (__TIMESIZE == 32)
diff --git a/sysdeps/unix/sysv/linux/sparc/bits/sem-pad.h b/sysdeps/unix/sysv/linux/sparc/bits/sem-pad.h
index b85f6b70a2..27266d48d5 100644
--- a/sysdeps/unix/sysv/linux/sparc/bits/sem-pad.h
+++ b/sysdeps/unix/sysv/linux/sparc/bits/sem-pad.h
@@ -20,7 +20,7 @@ 
 # error "Never use <bits/sem-pad.h> directly; include <sys/sem.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __SEM_PAD_AFTER_TIME 0
-#define __SEM_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __SEM_PAD_BEFORE_TIME (__TIMESIZE == 32)
diff --git a/sysdeps/unix/sysv/linux/sparc/bits/shm-pad.h b/sysdeps/unix/sysv/linux/sparc/bits/shm-pad.h
index c4d0c2feb8..fa3cc18a05 100644
--- a/sysdeps/unix/sysv/linux/sparc/bits/shm-pad.h
+++ b/sysdeps/unix/sysv/linux/sparc/bits/shm-pad.h
@@ -20,9 +20,9 @@ 
 # error "Never use <bits/shm-pad.h> directly; include <sys/shm.h> instead."
 #endif
 
-#include <bits/wordsize.h>
+#include <bits/timesize.h>
 
 #define __SHM_PAD_AFTER_TIME 0
-#define __SHM_PAD_BEFORE_TIME (__WORDSIZE == 32)
+#define __SHM_PAD_BEFORE_TIME (__TIMESIZE == 32)
 #define __SHM_SEGSZ_AFTER_TIME 1
 #define __SHM_PAD_BETWEEN_TIME_AND_SEGSZ 0
diff --git a/sysdeps/unix/sysv/linux/x86/bits/msq-pad.h b/sysdeps/unix/sysv/linux/x86/bits/timesize.h
similarity index 74%
rename from sysdeps/unix/sysv/linux/x86/bits/msq-pad.h
rename to sysdeps/unix/sysv/linux/x86/bits/timesize.h
index 5d46956c77..8b88ab84b0 100644
--- a/sysdeps/unix/sysv/linux/x86/bits/msq-pad.h
+++ b/sysdeps/unix/sysv/linux/x86/bits/timesize.h
@@ -1,4 +1,4 @@ 
-/* Define where padding goes in struct msqid_ds.  x86 version.
+/* Bit size of the time_t type at glibc build time, x86-64 and x32 case.
    Copyright (C) 2018 Free Software Foundation, Inc.
    This file is part of the GNU C Library.
 
@@ -16,13 +16,10 @@ 
    License along with the GNU C Library; if not, see
    <http://www.gnu.org/licenses/>.  */
 
-#ifndef _SYS_MSG_H
-# error "Never use <bits/msq-pad.h> directly; include <sys/msg.h> instead."
-#endif
-
-#ifdef __x86_64__
-# define __MSQ_PAD_AFTER_TIME 0
+#if defined __x86_64__ && defined __ILP32__
+/* For x32, time is 64-bit even though word size is 32-bit.  */
+# define __TIMESIZE	64
 #else
-# define __MSQ_PAD_AFTER_TIME 1
+/* For others, time size is word size.  */
+# define __TIMESIZE	__WORDSIZE
 #endif
-#define __MSQ_PAD_BEFORE_TIME 0