[LTP] [PATCH] lib/tst_device.c: Make sure all loop deivces fot test are detached

[Cyril Hrubis]

Hi!
> Running statx05 by runltp without -b option every time attached a loop
> device but didn't detach it, as below:
> ------------------------------------------------------------------
> ...
> tst_device.c:230: INFO: Using test device LTP_DEV='/dev/loop0'
> tst_device.c:268: INFO: Skipping $LTP_DEV size 256MB, requested size 512MB
> tst_device.c:83: INFO: Found free device '/dev/loop1'
> tst_mkfs.c:83: INFO: Formatting /dev/loop1 with ext4 opts='' extra opts='-O encrypt'
> ...
> NAME       SIZELIMIT OFFSET AUTOCLEAR RO BACK-FILE                                         DIO
> /dev/loop1         0      0         0  0 /tmp/ltp-gsrcFAkGpZ/Wka3bp/test_dev.img (deleted)   0
> ------------------------------------------------------------------
> 
> runltp created a block device with default 256M size by attaching a free loop
> device and set LTP_DEV variable, but statx05 needed a 512M block device, so
> tst_acquire_device__() in setup skipped the $LTP_DEV device and created another
> block device with 512M size by attaching another free loop device.  Unfortunately,
> tst_release_device() in cleanup cannot detach the 512M block device due to the
> defined LTP_DEV.

Btw his has been fixed for statx since:

commit e583eacf3c9781116c19dec4c7e8c0aa4ed01ed5
Author: Cyril Hrubis <chrubis@suse.cz>
Date:   Tue Sep 11 17:27:43 2018 +0200

    syscalls/statx05: Fix test on ppc64le


But yes apart from that the test library is still broken.

> We should just skip the detach in tst_release_device() when LTP_DEV variable is
> set and device_acquired flag is not set.
> 
> Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com>
> ---
>  lib/tst_device.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/lib/tst_device.c b/lib/tst_device.c
> index 463743443..23c99dba2 100644
> --- a/lib/tst_device.c
> +++ b/lib/tst_device.c
> @@ -313,7 +313,7 @@ int tst_release_device(const char *dev)
>  {
>  	int ret;
>  
> -	if (getenv("LTP_DEV"))
> +	if (getenv("LTP_DEV") && !device_acquired)
>  		return 0;

I guess that this really should be just:

	if (!device_acquired)
		return 0;

I will double check and test it later today.

-- 
Cyril Hrubis
chrubis@suse.cz