2011-05-16 17:49:23

by Theodore Ts'o

[permalink] [raw]
Subject: proposed ext4 test matrix


Per the discusison on the ext4 concall, here's a proposed ext4 test
matrix.

- Ted

P.S. There was something else I said I would send out, but I didn't
write it down and I can't remember what it was. Can someone remind me?
Thanks!!


4k blocksize, no journal, nodealloc, noextents (ext2)
1k blocksize, no journal, nodealloc, noextents (ext2)
4k blocksize, journal, nodealloc, noextents (ext3)
1k blocksize, journal, nodealloc, noextents (ext3)
4k blocksize, journal, delalloc, extents
1k blocksize, journal, delalloc, extents
4k blocksize, nojournal, delalloc, extents
4k blocksize, journal, 64bits(*), delalloc, extents
1k blocksize, journal, 64bits(*), delalloc, extents
4k blocksize, nojournal, 64bits(*), delalloc, extents
4k blocksize, journal, delalloc, extents, quota enabled
1k blocksize, journal, delalloc, extents, quota enabled

(*) requires master/next branch e2fsprogs; by this I mean creating a
file system with the 64-bit feature enabled, not necessarily creating
a 40TB file system!


2011-05-17 01:36:16

by Keith Mannthey

[permalink] [raw]
Subject: Re: proposed ext4 test matrix

On Mon, 2011-05-16 at 13:49 -0400, Theodore Ts'o wrote:
> Per the discusison on the ext4 concall, here's a proposed ext4 test
> matrix.
>
> - Ted
>
> P.S. There was something else I said I would send out, but I didn't
> write it down and I can't remember what it was. Can someone remind me?
> Thanks!!
>
>
> 4k blocksize, no journal, nodealloc, noextents (ext2)
> 1k blocksize, no journal, nodealloc, noextents (ext2)
> 4k blocksize, journal, nodealloc, noextents (ext3)
> 1k blocksize, journal, nodealloc, noextents (ext3)
> 4k blocksize, journal, delalloc, extents
> 1k blocksize, journal, delalloc, extents
> 4k blocksize, nojournal, delalloc, extents
> 4k blocksize, journal, 64bits(*), delalloc, extents
> 1k blocksize, journal, 64bits(*), delalloc, extents
> 4k blocksize, nojournal, 64bits(*), delalloc, extents
> 4k blocksize, journal, delalloc, extents, quota enabled
> 1k blocksize, journal, delalloc, extents, quota enabled


What tests are we wanting to run?

Keith Mannthey
LTC Filesystems


2011-05-17 05:10:33

by Andreas Dilger

[permalink] [raw]
Subject: Re: proposed ext4 test matrix

On May 16, 2011, at 19:35, Keith Mannthey wrote:
> On Mon, 2011-05-16 at 13:49 -0400, Theodore Ts'o wrote:
>> Per the discusison on the ext4 concall, here's a proposed ext4 test
>> matrix.
>>
>> - Ted
>>
>> P.S. There was something else I said I would send out, but I didn't
>> write it down and I can't remember what it was. Can someone remind me?
>> Thanks!!
>>
>>
>> 4k blocksize, no journal, nodealloc, noextents (-t ext2)
>> 1k blocksize, no journal, nodealloc, noextents (-t ext2)
>> 4k blocksize, journal, nodealloc, noextents (-t ext3)
>> 1k blocksize, journal, nodealloc, noextents (-t ext3)
>> 4k blocksize, journal, delalloc, extents (-t ext4)
>> 1k blocksize, journal, delalloc, extents (-t ext4)
>> 4k blocksize, nojournal, delalloc, extents
>> 4k blocksize, journal, 64bits(*), delalloc, extents (-t ext4)
>> 1k blocksize, journal, 64bits(*), delalloc, extents (-t ext4)
>> 4k blocksize, nojournal, 64bits(*), delalloc, extents
>> 4k blocksize, journal, delalloc, extents, quota enabled (-t ext4)
>> 1k blocksize, journal, delalloc, extents, quota enabled (-t ext4)
>
> What tests are we wanting to run?

On the call Ted said it would be desirable to run xfstests for all of these
configurations. Since xfstests already runs quite a number of smaller tests
like fsx, etc, this would definitely be a very good smoke test for the most
common configurations.

I added "-t" to the ext2/ext3/ext4 configs above, since without those options
the features enabled for the various filesystems will not be the same as they
are formatted by default. For example, with the default "-t ext4" a number
of other features like "flex_bg", "uninit_bg", "huge_file", "dir_nlink" and
"extra_isize" are enabled, but they shouldn't be enabled for ext2/ext3.

Cheers, Andreas