[PATCH] xfsprogs: add fallocate command to xfs_io

[PATCH] xfsprogs: add fallocate command to xfs_io

[Eric Sandeen]

Based on Dave's earlier patch, but now we have an fallocate
glibc call... this also adds autoconf magic and a manpage 
update.

(hopefully not too #ifdef-heavy....)

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
---


diff --git a/configure.in b/configure.in
index 3fbd44e..8f5782a 100644
--- a/configure.in
+++ b/configure.in
@@ -66,6 +66,7 @@ AC_HAVE_MINCORE
 AC_HAVE_SENDFILE
 AC_HAVE_GETMNTENT
 AC_HAVE_GETMNTINFO
+AC_HAVE_FALLOCATE
 
 AC_TYPE_PSINT
 AC_TYPE_PSUNSIGNED
diff --git a/include/builddefs.in b/include/builddefs.in
index c8f5c08..d6bf5c0 100644
--- a/include/builddefs.in
+++ b/include/builddefs.in
@@ -94,6 +94,7 @@ HAVE_MINCORE = @have_mincore@
 HAVE_SENDFILE = @have_sendfile@
 HAVE_GETMNTENT = @have_getmntent@
 HAVE_GETMNTINFO = @have_getmntinfo@
+HAVE_FALLOCATE = @have_fallocate@
 
 GCCFLAGS = -funsigned-char -fno-strict-aliasing -Wall 
 #	   -Wbitwise -Wno-transparent-union -Wno-old-initializer -Wno-decl
diff --git a/io/Makefile b/io/Makefile
index 6f10e8d..1a51879 100644
--- a/io/Makefile
+++ b/io/Makefile
@@ -59,6 +59,10 @@ ifeq ($(ENABLE_EDITLINE),yes)
 LLDLIBS += $(LIBEDITLINE) $(LIBTERMCAP)
 endif
 
+ifeq ($(HAVE_FALLOCATE),yes)
+LCFLAGS += -DHAVE_FALLOCATE
+endif
+
 default: $(LTCOMMAND)
 
 include $(BUILDRULES)
diff --git a/io/prealloc.c b/io/prealloc.c
index 6a2563e..7d9bd2f 100644
--- a/io/prealloc.c
+++ b/io/prealloc.c
@@ -16,6 +16,9 @@
  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  */
 
+#if defined(HAVE_FALLOCATE)
+#include <linux/falloc.h>
+#endif
 #include <xfs/xfs.h>
 #include <xfs/command.h>
 #include <xfs/input.h>
@@ -26,6 +29,9 @@ static cmdinfo_t allocsp_cmd;
 static cmdinfo_t freesp_cmd;
 static cmdinfo_t resvsp_cmd;
 static cmdinfo_t unresvsp_cmd;
+#if defined(HAVE_FALLOCATE)
+static cmdinfo_t falloc_cmd;
+#endif
 
 static int
 offset_length(
@@ -119,6 +125,40 @@ unresvsp_f(
 	return 0;
 }
 
+#if defined (HAVE_FALLOCATE)
+static int
+fallocate_f(
+	int		argc,
+	char		**argv)
+{
+	xfs_flock64_t	segment;
+	int		mode = 0;
+	int		c;
+
+	while ((c = getopt(argc, argv, "k")) != EOF) {
+		switch (c) {
+		case 'k':
+			mode = FALLOC_FL_KEEP_SIZE;
+			break;
+		default:
+			command_usage(&falloc_cmd);
+		}
+	}
+        if (optind != argc - 2)
+                return command_usage(&falloc_cmd);
+
+	if (!offset_length(argv[optind], argv[optind+1], &segment))
+		return 0;
+
+	if (fallocate(file->fd, mode,
+			segment.l_start, segment.l_len)) {
+		perror("fallocate");
+		return 0;
+	}
+	return 0;
+}
+#endif
+
 void
 prealloc_init(void)
 {
@@ -160,4 +200,17 @@ prealloc_init(void)
 	add_command(&freesp_cmd);
 	add_command(&resvsp_cmd);
 	add_command(&unresvsp_cmd);
+
+#if defined (HAVE_FALLOCATE)
+	falloc_cmd.name = _("falloc");
+	falloc_cmd.cfunc = fallocate_f;
+	falloc_cmd.argmin = 2;
+	falloc_cmd.argmax = -1;
+	falloc_cmd.flags = CMD_NOMAP_OK | CMD_FOREIGN_OK;
+	falloc_cmd.args = _("[-k] off len");
+	falloc_cmd.oneline =
+		_("allocates space associated with part of a file via fallocate");
+
+	add_command(&falloc_cmd);
+#endif
 }
diff --git a/m4/package_libcdev.m4 b/m4/package_libcdev.m4
index 5156ced..856794a 100644
--- a/m4/package_libcdev.m4
+++ b/m4/package_libcdev.m4
@@ -98,3 +98,18 @@ AC_DEFUN([AC_HAVE_GETMNTINFO],
        AC_MSG_RESULT(no))
     AC_SUBST(have_getmntinfo)
   ])
+
+#
+# Check if we have a fallocate libc call (Linux)
+#
+AC_DEFUN([AC_HAVE_FALLOCATE],
+  [ AC_MSG_CHECKING([for fallocate])
+    AC_TRY_COMPILE([
+#include <linux/falloc.h>
+    ], [
+         fallocate(0, 0, 0, 0);
+    ], have_fallocate=yes
+       AC_MSG_RESULT(yes),
+       AC_MSG_RESULT(no))
+    AC_SUBST(have_fallocate)
+  ])
diff --git a/man/man8/xfs_io.8 b/man/man8/xfs_io.8
index 23bef94..6fc6bad 100644
--- a/man/man8/xfs_io.8
+++ b/man/man8/xfs_io.8
@@ -295,6 +295,20 @@ system call described in the
 .BR xfsctl (3)
 manual page.
 .TP
+.BI "falloc [ \-k ]" " offset length"
+Allocates reserved, unwritten space for part of a file using the
+fallocate routine as described in the
+.BR fallocate (3)
+manual page.
+.RS 1.0i
+.PD 0
+.TP 0.4i
+.B \-k
+will set the FALLOC_FL_KEEP_SIZE flag as described in
+.BR fallocate (3).
+.PD
+.RE
+.TP
 .BI truncate " offset"
 Truncates the current file at the given offset using
 .BR ftruncate (2).

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfsprogs: add fallocate command to xfs_io

[Christoph Hellwig]

On Sun, May 10, 2009 at 07:30:13PM -0500, Eric Sandeen wrote:
> Based on Dave's earlier patch, but now we have an fallocate
> glibc call... this also adds autoconf magic and a manpage 
> update.
> 
> (hopefully not too #ifdef-heavy....)

Looks good to me and seems to work.

> +#if defined(HAVE_FALLOCATE)
> +#include <linux/falloc.h>
> +#endif

What about just using #ifdef HAVE_FALLOCATE everywhere?


Reviewed-by: Christoph Hellwig <hch@lst.de>

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfsprogs: add fallocate command to xfs_io

[Christoph Hellwig]

On Fri, May 15, 2009 at 01:13:24PM -0400, Christoph Hellwig wrote:
> On Sun, May 10, 2009 at 07:30:13PM -0500, Eric Sandeen wrote:
> > Based on Dave's earlier patch, but now we have an fallocate
> > glibc call... this also adds autoconf magic and a manpage 
> > update.
> > 
> > (hopefully not too #ifdef-heavy....)
> 
> Looks good to me and seems to work.

Actually that was spoken too fast.  On my Debian -testing system it
detects fallocate as available because <linux/falloc.h> exists, but the
glibc doesn't actually support a falloc(3) yet.  So either the detection
needs to be improved or we need to use the raw syscall.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfsprogs: add fallocate command to xfs_io

[Eric Sandeen]

Christoph Hellwig wrote:
> On Fri, May 15, 2009 at 01:13:24PM -0400, Christoph Hellwig wrote:
>> On Sun, May 10, 2009 at 07:30:13PM -0500, Eric Sandeen wrote:
>>> Based on Dave's earlier patch, but now we have an fallocate
>>> glibc call... this also adds autoconf magic and a manpage 
>>> update.
>>>
>>> (hopefully not too #ifdef-heavy....)
>> Looks good to me and seems to work.
> 
> Actually that was spoken too fast.  On my Debian -testing system it
> detects fallocate as available because <linux/falloc.h> exists, but the
> glibc doesn't actually support a falloc(3) yet.  So either the detection
> needs to be improved or we need to use the raw syscall.

Ok, I think I know how to make the detection work better, I'll fix that up.

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfsprogs: add fallocate command to xfs_io

[Eric Sandeen]

Eric Sandeen wrote:
> Christoph Hellwig wrote:
>> On Fri, May 15, 2009 at 01:13:24PM -0400, Christoph Hellwig wrote:
>>> On Sun, May 10, 2009 at 07:30:13PM -0500, Eric Sandeen wrote:
>>>> Based on Dave's earlier patch, but now we have an fallocate
>>>> glibc call... this also adds autoconf magic and a manpage 
>>>> update.
>>>>
>>>> (hopefully not too #ifdef-heavy....)
>>> Looks good to me and seems to work.
>> Actually that was spoken too fast.  On my Debian -testing system it
>> detects fallocate as available because <linux/falloc.h> exists, but the
>> glibc doesn't actually support a falloc(3) yet.  So either the detection
>> needs to be improved or we need to use the raw syscall.
> 
> Ok, I think I know how to make the detection work better, I'll fix that up.

oh, hrm.  It's already doing what I thought was sufficient but it won't
link will it:
(from m4/package_libcdev.m4)

+#
+# Check if we have a fallocate libc call (Linux)
+#
+AC_DEFUN([AC_HAVE_FALLOCATE],
+  [ AC_MSG_CHECKING([for fallocate])
+    AC_TRY_COMPILE([
+#include <linux/falloc.h>
+    ], [
+         fallocate(0, 0, 0, 0);
+    ], have_fallocate=yes
+       AC_MSG_RESULT(yes),
+       AC_MSG_RESULT(no))
+    AC_SUBST(have_fallocate)
+  ])


Instead of above does this work better?

(just change AC_TRY_COMPILE to AC_TRY_LINK)

AC_DEFUN([AC_HAVE_FALLOCATE],
  [ AC_MSG_CHECKING([for fallocate])
    AC_TRY_LINK([
#include <linux/falloc.h>
    ], [
         fallocate(0, 0, 0, 0);
    ], have_fallocate=yes
       AC_MSG_RESULT(yes),
       AC_MSG_RESULT(no))
    AC_SUBST(have_fallocate)
  ])

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfsprogs: add fallocate command to xfs_io

[Christoph Hellwig]

On Fri, May 15, 2009 at 05:11:24PM -0500, Eric Sandeen wrote:
> Instead of above does this work better?
> 
> (just change AC_TRY_COMPILE to AC_TRY_LINK)

Yes, that correctly detects fallocate as not present for me and
thus gives a working fallocate.

> 
> AC_DEFUN([AC_HAVE_FALLOCATE],
>   [ AC_MSG_CHECKING([for fallocate])
>     AC_TRY_LINK([
> #include <linux/falloc.h>
>     ], [
>          fallocate(0, 0, 0, 0);

Why do you nee to include <linux/falloc.h> here anyway?  You don't use
any of the constants, and the fallocate libc call must be in another
headers, so it obviously compiles even without a defintion for it.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfsprogs: add fallocate command to xfs_io

[Eric Sandeen]

Christoph Hellwig wrote:
> On Fri, May 15, 2009 at 05:11:24PM -0500, Eric Sandeen wrote:
>> Instead of above does this work better?
>>
>> (just change AC_TRY_COMPILE to AC_TRY_LINK)
> 
> Yes, that correctly detects fallocate as not present for me and
> thus gives a working fallocate.
> 
>> AC_DEFUN([AC_HAVE_FALLOCATE],
>>   [ AC_MSG_CHECKING([for fallocate])
>>     AC_TRY_LINK([
>> #include <linux/falloc.h>
>>     ], [
>>          fallocate(0, 0, 0, 0);
> 
> Why do you nee to include <linux/falloc.h> here anyway?  You don't use
> any of the constants, and the fallocate libc call must be in another
> headers, so it obviously compiles even without a defintion for it.

Well, the actual patch includes it for FALLOC_FL_KEEP_SIZE so it'd be
better to detect it at config time and if not there fail?


Or we could #ifndef and have a local define ...

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs