[PATCH] common/rc: Use -v option with findmnt

[PATCH] common/rc: Use -v option with findmnt

[Goldwyn Rodrigues]

From: Goldwyn Rodrigues <rgoldwyn@suse.com>

Since btrfs puts in the subvol name in the device, we should
use -v option to not print the bind mounts or btrfs volumes.

Fixes: 5e6892d ("common/rc: use findmnt to check mounted device")
Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
---
 common/rc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/common/rc b/common/rc
index d53f679..d7a5321 100644
--- a/common/rc
+++ b/common/rc
@@ -1463,7 +1463,7 @@ _check_mounted_on()
 	local type=$5
 
 	# find $dev as the source, and print result in "$dev $mnt" format
-	local mount_rec=`findmnt -rnc -S $dev -o SOURCE,TARGET`
+	local mount_rec=`findmnt -rncv -S $dev -o SOURCE,TARGET`
 	[ -n "$mount_rec" ] || return 1 # 1 = not mounted
 
 	# if it's mounted, make sure its on $mnt
-- 
2.10.2

Re: [PATCH] common/rc: Use -v option with findmnt

[Eryu Guan]

On Mon, Mar 13, 2017 at 01:07:26PM -0500, Goldwyn Rodrigues wrote:
> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
> 
> Since btrfs puts in the subvol name in the device, we should
> use -v option to not print the bind mounts or btrfs volumes.
> 
> Fixes: 5e6892d ("common/rc: use findmnt to check mounted device")
> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>

Hmm, I mounted a btrfs snapshot of TEST_DEV to another dir and ran
check, check reported:

TEST_DEV=/dev/sdc1 is mounted but not on TEST_DIR=/mnt - aborting
Already mounted result:
/dev/sdc1[/snap] /mnt/scratch

which seems fine and clear to me. Did you hit any particular failures
without this fix?

Thanks,
Eryu

> ---
>  common/rc | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/common/rc b/common/rc
> index d53f679..d7a5321 100644
> --- a/common/rc
> +++ b/common/rc
> @@ -1463,7 +1463,7 @@ _check_mounted_on()
>  	local type=$5
>  
>  	# find $dev as the source, and print result in "$dev $mnt" format
> -	local mount_rec=`findmnt -rnc -S $dev -o SOURCE,TARGET`
> +	local mount_rec=`findmnt -rncv -S $dev -o SOURCE,TARGET`
>  	[ -n "$mount_rec" ] || return 1 # 1 = not mounted
>  
>  	# if it's mounted, make sure its on $mnt
> -- 
> 2.10.2
> 

Re: [PATCH] common/rc: Use -v option with findmnt

[Goldwyn Rodrigues]

On 03/14/2017 12:07 AM, Eryu Guan wrote:
> On Mon, Mar 13, 2017 at 01:07:26PM -0500, Goldwyn Rodrigues wrote:
>> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
>>
>> Since btrfs puts in the subvol name in the device, we should
>> use -v option to not print the bind mounts or btrfs volumes.
>>
>> Fixes: 5e6892d ("common/rc: use findmnt to check mounted device")
>> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
> 
> Hmm, I mounted a btrfs snapshot of TEST_DEV to another dir and ran
> check, check reported:
> 
> TEST_DEV=/dev/sdc1 is mounted but not on TEST_DIR=/mnt - aborting
> Already mounted result:
> /dev/sdc1[/snap] /mnt/scratch
> 
> which seems fine and clear to me. Did you hit any particular failures
> without this fix?


The problem is you hit this error even if /dev/sdc1 is mounted on
TEST_DIR=/mnt because of the subvolume name appended to the device name.

IOW, the test fails when it should not have. Here is an output of
executing btrfs/021:

    --- tests/btrfs/021.out	2017-01-26 02:23:04.444778259 -0600
    +++ /fiona/xfstests/results//btrfs/021.out.bad	2017-03-14
04:25:38.807212599 -0500
    @@ -1,2 +1,4 @@
     QA output created by 021
    -Silence is golden
    +SCRATCH_DEV=/dev/vdc is mounted but not on SCRATCH_MNT=/scratch -
aborting
    +Already mounted result:
    +/dev/vdc[/newvol] /scratch

However, this fix is incorrect.

I found that this was because of reminiscences of btrfs/009
Adding _scratch_umount to the end of btrfs/009 solves the issue, since
the filesystem is created fresh in the next test.

The filesystem was not created fresh in the next test because findmnt
would fail in _require_scratch.

I will shortly post a new patch which would fix this.



-- 
Goldwyn

Re: [PATCH] common/rc: Use -v option with findmnt

[Eryu Guan]

On Tue, Mar 14, 2017 at 09:07:05AM -0500, Goldwyn Rodrigues wrote:
> 
> 
> On 03/14/2017 12:07 AM, Eryu Guan wrote:
> > On Mon, Mar 13, 2017 at 01:07:26PM -0500, Goldwyn Rodrigues wrote:
> >> From: Goldwyn Rodrigues <rgoldwyn@suse.com>
> >>
> >> Since btrfs puts in the subvol name in the device, we should
> >> use -v option to not print the bind mounts or btrfs volumes.
> >>
> >> Fixes: 5e6892d ("common/rc: use findmnt to check mounted device")
> >> Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
> > 
> > Hmm, I mounted a btrfs snapshot of TEST_DEV to another dir and ran
> > check, check reported:
> > 
> > TEST_DEV=/dev/sdc1 is mounted but not on TEST_DIR=/mnt - aborting
> > Already mounted result:
> > /dev/sdc1[/snap] /mnt/scratch
> > 
> > which seems fine and clear to me. Did you hit any particular failures
> > without this fix?
> 
> 
> The problem is you hit this error even if /dev/sdc1 is mounted on
> TEST_DIR=/mnt because of the subvolume name appended to the device name.
> 
> IOW, the test fails when it should not have. Here is an output of
> executing btrfs/021:
> 
>     --- tests/btrfs/021.out	2017-01-26 02:23:04.444778259 -0600
>     +++ /fiona/xfstests/results//btrfs/021.out.bad	2017-03-14
> 04:25:38.807212599 -0500
>     @@ -1,2 +1,4 @@
>      QA output created by 021
>     -Silence is golden
>     +SCRATCH_DEV=/dev/vdc is mounted but not on SCRATCH_MNT=/scratch -
> aborting
>     +Already mounted result:
>     +/dev/vdc[/newvol] /scratch

Ah, I see the problem. It's because btrfs/009 sets default volume to the
newly created "newvol", then _scratch_mount mounts the newvol by
default. And _check_mounted_on finds the inconsistency because of the
subvolume name.

> 
> However, this fix is incorrect.

I think this patch is a clearer fix, it doesn't require test to do
_scratch_umount at the end of test, and works for future tests that do
the same setup.

Thanks,
Eryu

> 
> I found that this was because of reminiscences of btrfs/009
> Adding _scratch_umount to the end of btrfs/009 solves the issue, since
> the filesystem is created fresh in the next test.
> 
> The filesystem was not created fresh in the next test because findmnt
> would fail in _require_scratch.
> 
> I will shortly post a new patch which would fix this.
> 
> 
> 
> -- 
> Goldwyn