Allow [f]statfs64 to alias [f]statfs
Commit Message
On Mon, 2016-11-07 at 14:13 -0800, Steve Ellcey wrote:
> This is a preparatory patch for the ILP32 aarch64 patch. On that system
> statfs and fstatfs will match statfs64 and fstatfs64. So this patch
> follows what was already done for xstat and fxstat in making the 64
> bit entry points aliases of the regular versions. Right now this will
> not affect any existing platform because none of them define
> STATFS_IS_STATFS64 but aarch64 will define this later.
>
> My main question is, is this an OK use of '#ifdef' or should the
> STATFS_IS_STATFS64 be defined on all platforms (presumably in kernel_stat.h)
> so that we can check the value instead of checking to see it is defined
> or not? I know that we want to move in that direction but I wasn't
> sure if I should include that change in this patch or not. I also wasn't
> sure if I should try to change how XSTAT_IS_XSTAT64 or STAT_IS_KERNEL_STAT
> are used.
>
> There are 13 kernel_stat.h files in the glibc source tree, in addition
> to a number of _HAVE_* macros, some of them define XSTAT_IS_XSTAT64
> and some define STAT_IS_KERNEL_STAT. Those are the only defines
> in those files right now. So the magnitude of the change isn't too large
> if we do want to define those two macros everywhere.
>
> I tested this patch on x86 to make sure nothing broke and I tested with
> the full ILP32 aarch64 patch to make sure this part of the patch worked
> with the rest of the ILP32 changes.
>
> Is the patch OK like this or should it be redone?
>
> Steve Ellcey
> sellcey@caviumnetworks.com
I have modified this patch, I missed a problem in the ILP32 glibc build
and so I had to add the #define of some symbols to hide their
prototypes when aliasing them. This in turn required me to add some
includes of kernel_stat.h. This also matches what is being done with
xs
tat.
Steve Ellcey
sellcey@caviumnetworks.com
2016-11-09 Steve Ellcey <sellcey@caviumnetworks.com>
* sysdeps/unix/sysv/linux/fstatfs64.c: Hide prototypes for fstatfs
and __fstatfs. Make them aliases of __fstatfs64 if
__STATFS_IS_STATFS64 is set.
* sysdeps/unix/sysv/linux/statfs64.c: Ditto for __statfs, statfs,
and __statfs64.
* sysdeps/unix/sysv/linux/generic/wordsize-32/fstatfs.c: Do not
define __fstatfs and fstatfs if __STATFS_IS_STATFS64 is set.
* sysdeps/unix/sysv/linux/generic/wordsize-32/statfs.c: Ditto
for __statfs and statfs.
Comments
Ping.
Steve Ellcey
On Wed, 2016-11-09 at 13:21 -0800, Steve Ellcey wrote:
> On Mon, 2016-11-07 at 14:13 -0800, Steve Ellcey wrote:
> >
> > This is a preparatory patch for the ILP32 aarch64 patch. On that
> > system statfs and fstatfs will match statfs64 and fstatfs64. So
> > this patch follows what was already done for xstat and fxstat in
> > making the 64 bit entry points aliases of the regular
> > versions. Right now this will not affect any existing platform
> > because none of them define STATFS_IS_STATFS64 but aarch64 will
> > define this later.
> >
> > My main question is, is this an OK use of '#ifdef' or should the
> > STATFS_IS_STATFS64 be defined on all platforms (presumably in
> > kernel_stat.h) so that we can check the value instead of checking
> > to see it is defined or not? I know that we want to move in that
> > direction but I wasn't sure if I should include that change in this
> > patch or not. I also wasn't sure if I should try to change how
> > XSTAT_IS_XSTAT64 or STAT_IS_KERNEL_STAT are used.
> >
> > There are 13 kernel_stat.h files in the glibc source tree, in
> > addition to a number of _HAVE_* macros, some of them define
> > XSTAT_IS_XSTAT64 and some define STAT_IS_KERNEL_STAT. Those are
> > the only defines in those files right now. So the magnitude of the
> > change isn't too large if we do want to define those two macros
> > everywhere.
> >
> > I tested this patch on x86 to make sure nothing broke and I tested
> > with the full ILP32 aarch64 patch to make sure this part of the
> > patch worked with the rest of the ILP32 changes.
> >
> > Is the patch OK like this or should it be redone?
> >
> > Steve Ellcey
> > sellcey@caviumnetworks.com
> I have modified this patch, I missed a problem in the ILP32 glibc
> build and so I had to add the #define of some symbols to hide their
> prototypes when aliasing them. This in turn required me to add some
> includes of kernel_stat.h. This also matches what is being done with
> xstat.
>
> Steve Ellcey
> sellcey@caviumnetworks.com
>
>
>
> 2016-11-09 Steve Ellcey <sellcey@caviumnetworks.com>
>
> * sysdeps/unix/sysv/linux/fstatfs64.c: Hide prototypes for fstatfs
> and __fstatfs. Make them aliases of __fstatfs64 if
> __STATFS_IS_STATFS64 is set.
> * sysdeps/unix/sysv/linux/statfs64.c: Ditto for __statfs, statfs,
> and __statfs64.
> * sysdeps/unix/sysv/linux/generic/wordsize-32/fstatfs.c: Do not
> define __fstatfs and fstatfs if __STATFS_IS_STATFS64 is set.
> * sysdeps/unix/sysv/linux/generic/wordsize-32/statfs.c: Ditto
> for __statfs and statfs.
@@ -16,9 +16,18 @@
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
+/* Hide the prototypes for __fstatfs and fstatfs so that GCC will not
+ complain about the different function signatures if they are aliased
+ to __fstat64. If STATFS_IS_STATFS64 is set then the statfs and
+ statfs64 structures have an identical layout but different type names. */
+
+#define __fstatfs __fstatfs_disable
+#define fstatfs fstatfs_disable
+
#include <errno.h>
#include <string.h>
#include <sys/statfs.h>
+#include <kernel_stat.h>
#include <stddef.h>
#include <sysdep.h>
#include <kernel-features.h>
@@ -70,3 +79,11 @@ __fstatfs64 (int fd, struct statfs64 *buf)
#endif
}
weak_alias (__fstatfs64, fstatfs64)
+
+#undef __fstatfs
+#undef fstatfs
+
+#ifdef STATFS_IS_STATFS64
+weak_alias (__fstatfs64, __fstatfs)
+weak_alias (__fstatfs64, fstatfs)
+#endif
@@ -18,8 +18,10 @@
#include <errno.h>
#include <sys/statfs.h>
+#include <kernel_stat.h>
#include <stddef.h>
+#ifndef STATFS_IS_STATFS64
#include "overflow.h"
/* Return information about the filesystem on which FD resides. */
@@ -30,3 +32,4 @@ __fstatfs (int fd, struct statfs *buf)
return rc ?: statfs_overflow (buf);
}
weak_alias (__fstatfs, fstatfs)
+#endif
@@ -18,8 +18,10 @@
#include <errno.h>
#include <sys/statfs.h>
+#include <kernel_stat.h>
#include <stddef.h>
+#ifndef STATFS_IS_STATFS64
#include "overflow.h"
/* Return information about the filesystem on which FILE resides. */
@@ -31,3 +33,4 @@ __statfs (const char *file, struct statfs *buf)
}
libc_hidden_def (__statfs)
weak_alias (__statfs, statfs)
+#endif
@@ -16,9 +16,18 @@
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
+/* Hide the prototypes for __statfs and statfs so that GCC will not
+ complain about the different function signatures if they are aliased
+ to __stat64. If STATFS_IS_STATFS64 is set then the statfs and
+ statfs64 structures have an identical layout but different type names. */
+
+#define __statfs __statfs_disable
+#define statfs statfs_disable
+
#include <errno.h>
#include <string.h>
#include <sys/statfs.h>
+#include <kernel_stat.h>
#include <stddef.h>
#include <sysdep.h>
#include <kernel-features.h>
@@ -72,3 +81,12 @@ __statfs64 (const char *file, struct statfs64 *buf)
#endif
}
weak_alias (__statfs64, statfs64)
+
+#undef __statfs
+#undef statfs
+
+#ifdef STATFS_IS_STATFS64
+weak_alias (__statfs64, __statfs)
+weak_alias (__statfs64, statfs)
+libc_hidden_ver (__statfs64, __statfs)
+#endif