Re: [PATCH] fs: add roundup_to_blocksize and bytes_to_blocks helper functions

[Marcin Slusarz <marcin.slusarz@gmail.com>]

Andrew Morton pisze:
> On Sun, 19 Apr 2009 20:42:15 +0200
> Marcin Slusarz <marcin.slusarz@gmail.com> wrote:
> 
>> bytes_to_blocks is open coded in more than 30 places
>> roundup_to_blocksize is open coded in more than 15 places
>>
> 
> It would be nice to document these a bit.  Kernel-wide helpers..

ok

>
>> ---
>>  include/linux/fs.h |   12 ++++++++++++
>>  1 files changed, 12 insertions(+), 0 deletions(-)
>>
>> diff --git a/include/linux/fs.h b/include/linux/fs.h
>> index e766be0..1b4e5b6 100644
>> --- a/include/linux/fs.h
>> +++ b/include/linux/fs.h
>> @@ -2451,5 +2451,17 @@ int proc_nr_files(struct ctl_table *table, int write, struct file *filp,
>>  
>>  int get_filesystem_list(char * buf);
>>  
>> +static inline
>> +unsigned long long roundup_to_blocksize(unsigned long long bytes, struct super_block *sb)
>> +{
>> +	return (bytes + sb->s_blocksize - 1) & ~(sb->s_blocksize - 1);
> 
> This is ALIGN(bytes, sb->s_blocksize).  Dunno if that's any clearer though.

thanks, I didn't know about it

>> +}
>> +
>> +static inline
>> +blkcnt_t bytes_to_blocks(unsigned long long bytes, struct super_block *sb)
>> +{
>> +	return (bytes + sb->s_blocksize - 1) >> sb->s_blocksize_bits;
>> +}
> 
> I wonder whether `unsigned long long' was the best choice of type here.
> 
> Probably `bytes' should be loff_t.

loff_t is long long.

So bytes_to_blocks(-6000, sb/*bs=4096, bsb=12*/) would return 4503599627370495

> roundup_to_blocksize() could/should return loff_t too, I think.

And roundup_to_blocksize(-6000, sb/*bs=4096, bsb=12*/) would return -4096

Bit surprising ;)

> roundup_to_blocksize() isn't a terribly good name, IMO.  The name
> conveys no sense of what the function returns.  It _sounds_ like it
> returns a "block size".  But it doesn't - it returns a loff_t. 
> round_up_file_offset_to_block_size() is a bit long though :(

I can't find any better name, sorry.

Marcin