ext4: fix extent tree corruption that incurred by hole punch

Hi Forrest,

Can you share an updated version of your patch (with Ashish's
modification if you've tried testing that)?  And also can you send us
some automated way to create your test case?  That might make it
easier for us all to test the same thing and all be on the same page.

I assume you have some C program which which is doing all of the
punching?

I started trying to create something via a script like this:

touch testfile
fallocate -l 1G testfile
for i in $(seq 0 8192 1073741824) ; do fallocate -p -o $i -l 4k testfile ; done

.. where fallocate is from contrib/fallocate.c in the e2fsprogs
sources (not the fallocate from util-linux) with the attached patch to
add support for the -p option to do the pnuching.... but then I
decided it might be better if we all standardized on the same test
case.  Ideally, if the test case is small enough, we can put it in the
commit message, or at very least we should look into getting it into
xfstests, and Eric has suggested.

Cheers,

						- Ted

commit 2658b9616fe63d2fd9e2654676877ab014639a92
Author: Theodore Ts'o <tytso@mit.edu>
Date:   Thu Dec 6 11:21:44 2012 -0500

    contrib/fallocate: add support for punch functionality
    
    Also fix the -o option so it works correctly (instead of core
    dumping).
    
    Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Hi Ted,
    I have test patch with Ashish's modification, and that have
problem when depth of extent tree is greater than 2.

   I will make a test case with your patch.

Thanks,
Forrest


2012/12/8 Theodore Ts'o <tytso@mit.edu>:
> Hi Forrest,
>
> Can you share an updated version of your patch (with Ashish's
> modification if you've tried testing that)?  And also can you send us
> some automated way to create your test case?  That might make it
> easier for us all to test the same thing and all be on the same page.
>
> I assume you have some C program which which is doing all of the
> punching?
>
> I started trying to create something via a script like this:
>
> touch testfile
> fallocate -l 1G testfile
> for i in $(seq 0 8192 1073741824) ; do fallocate -p -o $i -l 4k testfile ; done
>
> .. where fallocate is from contrib/fallocate.c in the e2fsprogs
> sources (not the fallocate from util-linux) with the attached patch to
> add support for the -p option to do the pnuching.... but then I
> decided it might be better if we all standardized on the same test
> case.  Ideally, if the test case is small enough, we can put it in the
> commit message, or at very least we should look into getting it into
> xfstests, and Eric has suggested.
>
> Cheers,
>
>                                                 - Ted
>
> commit 2658b9616fe63d2fd9e2654676877ab014639a92
> Author: Theodore Ts'o <tytso@mit.edu>
> Date:   Thu Dec 6 11:21:44 2012 -0500
>
>     contrib/fallocate: add support for punch functionality
>
>     Also fix the -o option so it works correctly (instead of core
>     dumping).
>
>     Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
>
> diff --git a/contrib/fallocate.c b/contrib/fallocate.c
> index 0e8319f..1436b70 100644
> --- a/contrib/fallocate.c
> +++ b/contrib/fallocate.c
> @@ -35,6 +35,7 @@
>
>  // #include <linux/falloc.h>
>  #define FALLOC_FL_KEEP_SIZE    0x01
> +#define FALLOC_FL_PUNCH_HOLE   0x02 /* de-allocates range */
>
>  void usage(void)
>  {
> @@ -94,12 +95,17 @@ int main(int argc, char **argv)
>         int     error;
>         int     tflag = 0;
>
> -       while ((opt = getopt(argc, argv, "nl:ot")) != -1) {
> +       while ((opt = getopt(argc, argv, "npl:o:t")) != -1) {
>                 switch(opt) {
>                 case 'n':
>                         /* do not change filesize */
>                         falloc_mode = FALLOC_FL_KEEP_SIZE;
>                         break;
> +               case 'p':
> +                       /* punch mode */
> +                       falloc_mode = (FALLOC_FL_PUNCH_HOLE |
> +                                      FALLOC_FL_KEEP_SIZE);
> +                       break;
>                 case 'l':
>                         length = cvtnum(optarg);
>                         break;
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html