[dm-devel] Re: [PATCH 3/3] Add timeout feature

[Takashi Sato]

Hi,

>> >>  case XFS_FSOP_GOING_FLAGS_DEFAULT: {
>> >> - struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
>> >> + struct super_block *sb = freeze_bdev(mp->m_super->s_bdev, 0);
>> >
>> > Using NULL here is clearer and will, I expect, avoid a sparse warning.
>> 
>> I checked it but I couldn't find a sparse warning in xfs_fsops.c.
>> Can you tell me how to use NULL?
> 
> struct super_block *sb = freeze_bdev(mp->m_super->s_bdev, NULL);
> 
> :)
> 
> It's much better to use NULL here rather than literal zero because the
> reader of this code can then say "ah-hah, we're passing in a pointer". 
> Whereas plain old "0" could be a pointer or a scalar.

The second argument's type of freeze_bdev() is "long", not pointer as below.
struct super_block *freeze_bdev(struct block_device *, long timeout_msec);

So "0" is reasonable, isn't it?

> We should always use NULL to represent a null pointer in the kernel. 
> The one acceptable exception is when testing for nullness:
> 
> if (ptr1)
> if (!ptr2)
> 
> Often people will use
> 
> if (ptr1 != NULL)
> if (ptr2 == NULL)
> 
> in this case as well.  (I prefer the shorter version personally, but
> either is OK).