Re: [PATCH 16/21] sysfs: implement bin_buffer

[Tejun Heo <htejun@gmail.com>]

Satyam Sharma wrote:
>> Yes, I saw that, but you could change it just as well. In any case,
>> this will only make fs/sysfs/bin.c similar to what is being done in
>> fs/sysfs/file.c. We allocate the buffer page backing the attribute's
>> data in fill_read_buffer() and fill_write_buffer() using
>> get_zeroed_page() and not kzalloc().
> 
> Which begs another question -- why do we allocate the buffer page
> lazily (only at the time of read(2) and write(2) and not at open(2))
> in the case of normal attributes but prefer to do it during open(2)
> itself for binary attributes? Note that the page (if allocated) is
> freed during release() for both cases.

Well, that also is because that was how bin files did it before the
patch.  I don't object to any of your suggestions but also thinks that
they are better off as separate patches.  So, feel free to submit patches.

-- 
tejun