Re: [PATCH] fstests: Block btrfs from test case generic/372

Re: [PATCH] fstests: Block btrfs from test case generic/372

[Dave Chinner]

(Did you forget to cc fstests@vger.kernel.org?)

On Tue, Nov 15, 2016 at 04:13:32PM +0800, Qu Wenruo wrote:
> Since btrfs always return the whole extent even part of it is shared
> with other files, so the hole/extent counts differs for "file1" in this
> test case.
> 
> For example:
> 
>  /------ File 1 Extent 0-------------\
> /                                     \
> |<----------Extent A------------------>|
> \         /                  \         /
>  \ File 2/                    \ File 2/
>    Ext 0~4K                    Ext 64k~68K
> 
> In that case, fiemap on File 1 will only return 1 large extent A with
> SHARED flag.
> While XFS will split it into 3 extents,  first and last 4K with SHARED
> flag while the rest without SHARED flag.

fiemap should behave the same across all filesystems if at all
possible. This test failure indicates btrfs doesn't report an
accurate representation of shared extents which, IMO, is a btrfs
issue that needs fixing, not a test problem....

Regardless of this....

> This makes the test case meaningless as btrfs doesn't follow such
> assumption.
> So black list btrfs for this test case to avoid false alert.

...  we are not going to add ad-hoc filesystem blacklists for
random tests.

Adding "blacklists" without any explanation of why something has
been blacklisted is simply a bad practice. We use _require rules
to specifically document what functionality is required for the
test and check that it provided.  i.e. this:

_require_explicit_shared_extents()
{
	if [ $FSTYP == "btrfs" ]; then
		_not_run "btrfs can't report accurate shared extent ranges in fiemap"
	fi
}

documents /exactly/ why this test is not run on btrfs.

And, quite frankly, while this is /better/ it still ignores the
fact we have functions like _within_tolerance for allowing a range
of result values to be considered valid rather than just a fixed
value. IOWs, changing the check of the extent count of file 1 post
reflink to use a _within_tolerance range would mean the test would
validate file1 on all reflink supporting filesystems and we don't
need to exclude btrfs at all...

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: [PATCH] fstests: Block btrfs from test case generic/372

[Qu Wenruo]

At 11/17/2016 05:12 AM, Dave Chinner wrote:
> (Did you forget to cc fstests@vger.kernel.org?)
>
> On Tue, Nov 15, 2016 at 04:13:32PM +0800, Qu Wenruo wrote:
>> Since btrfs always return the whole extent even part of it is shared
>> with other files, so the hole/extent counts differs for "file1" in this
>> test case.
>>
>> For example:
>>
>>  /------ File 1 Extent 0-------------\
>> /                                     \
>> |<----------Extent A------------------>|
>> \         /                  \         /
>>  \ File 2/                    \ File 2/
>>    Ext 0~4K                    Ext 64k~68K
>>
>> In that case, fiemap on File 1 will only return 1 large extent A with
>> SHARED flag.
>> While XFS will split it into 3 extents,  first and last 4K with SHARED
>> flag while the rest without SHARED flag.
>
> fiemap should behave the same across all filesystems if at all
> possible. This test failure indicates btrfs doesn't report an
> accurate representation of shared extents which, IMO, is a btrfs
> issue that needs fixing, not a test problem....
>
> Regardless of this....

Considering only btrfs implements CoW using extent booking mechanism(*),
it does affect a lot of behavior, from such SHARED flag representing to
hole punching behavior.

I hope there is a well documented standard on what ever flag means and
how it should be represented.

While I'm not quite sure if it's worthy for btrfs to modify the
represent.
Even btrfs can report it as "SHARED" "NON-SHARED" "SHARED" for File 1,
hole punching the "NON_SHARED" range won't free any space.
(Which I assume it differs from xfs, and that's making things confusing)

>
>> This makes the test case meaningless as btrfs doesn't follow such
>> assumption.
>> So black list btrfs for this test case to avoid false alert.
>
> ...  we are not going to add ad-hoc filesystem blacklists for
> random tests.
>
> Adding "blacklists" without any explanation of why something has
> been blacklisted is simply a bad practice. We use _require rules
> to specifically document what functionality is required for the
> test and check that it provided.  i.e. this:
>
> _require_explicit_shared_extents()
> {
> 	if [ $FSTYP == "btrfs" ]; then
> 		_not_run "btrfs can't report accurate shared extent ranges in fiemap"
> 	fi
> }

Right, this is much more helpful than the blabla I wrote in commit message.

Although I'd prefer to detect it at runtime other than just checking
the fs type.

Maybe one day btrfs will support it.
(Although we should solve the above mentioned behavior difference first)

>
> documents /exactly/ why this test is not run on btrfs.
>
> And, quite frankly, while this is /better/ it still ignores the
> fact we have functions like _within_tolerance for allowing a range
> of result values to be considered valid rather than just a fixed
> value. IOWs, changing the check of the extent count of file 1 post
> reflink to use a _within_tolerance range would mean the test would
> validate file1 on all reflink supporting filesystems and we don't
> need to exclude btrfs at all...

I really agree on this idea, although for me the difference is too big.

For file 1, xfs reports 5 extents, while btrfs only reports 1.
If using _within_tolerance to cover the range, and one day some
mysterious xfs bug(OK, I don't really believe it will happen, since
it's xfs, not btrfs) makes it report 4 extents.

Or one btrfs bug(on the other hand, quite possible) makes btrfs report
2 extents, then we can't detect the bug either.

So I'd prefer the _require_explicit_shared_extents() method.

Thanks,
Qu

>
> Cheers,
>
> Dave.
>

[PATCH] fstests: Block btrfs from test case generic/372

[Qu Wenruo]

Since btrfs always return the whole extent even part of it is shared
with other files, so the hole/extent counts differs for "file1" in this
test case.

For example:

 /------ File 1 Extent 0-------------\
/                                     \
|<----------Extent A------------------>|
\         /                  \         /
 \ File 2/                    \ File 2/
   Ext 0~4K                    Ext 64k~68K

In that case, fiemap on File 1 will only return 1 large extent A with
SHARED flag.
While XFS will split it into 3 extents,  first and last 4K with SHARED
flag while the rest without SHARED flag.

This makes the test case meaningless as btrfs doesn't follow such
assumption.
So black list btrfs for this test case to avoid false alert.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
 common/rc         | 10 ++++++++++
 tests/generic/372 |  1 +
 2 files changed, 11 insertions(+)

diff --git a/common/rc b/common/rc
index e3b54ec..6c3de22 100644
--- a/common/rc
+++ b/common/rc
@@ -3230,6 +3230,16 @@ _normalize_mount_options()
 	echo $MOUNT_OPTIONS | sed -n 's/-o\s*\(\S*\)/\1/gp'| sed 's/,/ /g'
 }
 
+# skip test if FSTYP is in the black list
+_exclude_fs_type()
+{
+	while [ $# -gt 0 ]; do
+		if [ $FSTYP == $1 ]; then
+			_notrun "fs type \"$1\" is not allowed in this test"
+		fi
+	done
+}
+
 # skip test if MOUNT_OPTIONS contains the given strings
 _exclude_scratch_mount_option()
 {
diff --git a/tests/generic/372 b/tests/generic/372
index 31dff20..5e4cfff 100755
--- a/tests/generic/372
+++ b/tests/generic/372
@@ -45,6 +45,7 @@ _cleanup()
 # real QA test starts here
 _supported_os Linux
 _supported_fs generic
+_exclude_fs_type btrfs
 _require_scratch_reflink
 _require_fiemap
 
-- 
2.7.4

Re: [PATCH] fstests: Block btrfs from test case generic/372

[Christoph Hellwig]

On Tue, Nov 15, 2016 at 04:14:10PM +0800, Qu Wenruo wrote:
> Since btrfs always return the whole extent even part of it is shared
> with other files, so the hole/extent counts differs for "file1" in this
> test case.

This would need an extremly detailed comment in the test case on why
btrfs is skipped IFF we decide to apply it.  That being siad I think
reporing shared extents incorrectly is a bug independent of how they
are implemented internally.

Re: [PATCH] fstests: Block btrfs from test case generic/372

[Qu Wenruo]

At 11/15/2016 04:57 PM, Christoph Hellwig wrote:
> On Tue, Nov 15, 2016 at 04:14:10PM +0800, Qu Wenruo wrote:
>> Since btrfs always return the whole extent even part of it is shared
>> with other files, so the hole/extent counts differs for "file1" in this
>> test case.
>
> This would need an extremly detailed comment in the test case on why
> btrfs is skipped IFF we decide to apply it.  That being siad I think
> reporing shared extents incorrectly is a bug independent of how they
> are implemented internally.
>
>
I don't think it's a bug.

It's just the different granularity on SHARED flag.

SHARED flag returned by btrfs has its meaning, since it is shared by 
other files/subvolumes.

And further more, due to the extent booking mechanism in btrfs, one 
extent won't be modified/removed until all its referencer got removed.

This quite lazy behavior co-operates with the lazy SHARED flag well.
So IMHO it's just granularity difference not a bug.


Not to mention that, fiemap bytenr returned from btrfs doesn't makes 
sense at all, unless you know how to map it into device.
 From this aspect, btrfs fiemap is already a so-called "bug" compared to 
xfs/ext*.

Thanks,
Qu