On 07/13/2011 09:31 AM, Alex Elder wrote:

On Wed, 2011-06-29 at 22:24 -0700, Allison Henderson wrote:

quoted

This patch adds a new test 255 that tests fallocate punch hole,
but is for use by generic filesystems.  It is similar to 252,
but uses a weaker fiemap filter that only displays the location
of the hole, and not the extent types.

Signed-off-by: Allison Henderson<redacted>

I realize Christoph already committed this but I
spotted something so I thought I might as well
mention it.  I haven't tested any of this, so
it may be OK after all, but if it's not then
the filter_hole_fiemap() function ought to be
updated.

					-Alex

. . .

quoted

diff --git a/common.punch b/common.punch
index c18f88a..a48b7ae 100644
--- a/common.punch
+++ b/common.punch

@@ -218,6 +218,23 @@ _filter_fiemap()
  	_coalesce_extents
  }

+# Filters fiemap output to only print the
+# file offset column and whether or not
+# it is an extent or a hole
+_filter_hole_fiemap()
+{
+	awk --posix '
+		$3 ~ /hole/ {
+			print $1, $2, $3;
+			next;
+		}
+		$5 ~ /0x[[:digit:]]+/ {

I believe this should be:
                  $5 ~ /0x[[:xdigit:]]+/ {

(I.e., hex digits, not just decimals.)

Hi Alex,

I think you are right.  The extent flags are printed in hex, so it 
should be xdigit.  The existing _filter_fiemap routine that I modeled 
this code from is also only parsing decimal digits.  Even though the 
tests are not showing any problems right now, it should probably be 
fixed.  Should I submit a patch to correct it?

Allison Henderson

quoted

+			print $1, $2, "extent";
+		}' |
+	_coalesce_extents
+}
+
+
  # Prints the md5 checksum of a given file
  _md5_checksum()
  {

. . .