[LTP] [PATCH v2 1/2] syscalls/ioctl_loop06: Using LOOP_CONFIGURE to test invalid block size

[Cyril Hrubis]

Hi!
> > Do we really need to close and open the dev_fd repeatedly and also we
> > don't have to attach the device in the test setup?
> YES, we don't need to attach the device in the setup because 
> LOOP_SET_BLOCK_SIZE checks works well(return ENXIO if supports, return 
> EINVAL if not supports) when not attaching device.
> 
> But for close and open the dev_fd repeatedly, I think it is necessary 
> because when we detach device firstly without closing dev fd, it will 
> report the warnging as below:
> 
> tst_device.c:89: INFO: Found free device 0 '/dev/loop0'
> ioctl_loop06.c:69: INFO: Using LOOP_SET_BLOCK_SIZE with arg < 512
> ioctl_loop06.c:60: PASS: Set block size failed as expected: EINVAL (22)
> ioctl_loop06.c:69: INFO: Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE
> ioctl_loop06.c:60: PASS: Set block size failed as expected: EINVAL (22)
> ioctl_loop06.c:69: INFO: Using LOOP_SET_BLOCK_SIZE with arg != power_of_2
> ioctl_loop06.c:60: PASS: Set block size failed as expected: EINVAL (22)
> ioctl_loop06.c:69: INFO: Using LOOP_CONFIGURE with block_size < 512
> tst_device.c:223: WARN: ioctl(/dev/loop0, LOOP_CLR_FD, 0) no ENXIO for 
> too long
> ioctl_loop06.c:62: FAIL: Set block size failed expected EINVAL got: 
> EBUSY (16)
> 
> That is why I close dev_fd firstly and then detach device in cleanup 
> function.

Ah, right, the tst_detach_device() opens the dev_path so the function
fails to destroy it because the device is opened twice at that point.

I guess that proper solution would be to add a tst_detach_device_by_fd()
and change the library so that tst_detach_device() opens the dev_path
and calls tst_detach_device_by_fd() internally.

-- 
Cyril Hrubis
chrubis@suse.cz