[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-fsdevel
Subject:    readpage, bmap, and other fun stuff
From:       Steve Dodd <dirk () loth ! demon ! co ! uk>
Date:       1999-06-11 2:29:46
[Download RAW message or body]

I've just been looking at NTFS and implementation of readpage, bmap, read
and other stuff. Now it's 3am and time for some daft questions..

Mostly (esp. for executables) in NTFS file data will be cluster (read: block)
aligned. However, the cluster size can be 512 -- is this going to cause
problems? I got very confused looking into all the different interpretations
of "block"..

Also, small files can have their data stored entirely in the FILE structure,
i.e. within the inode, and so may not be aligned to anything useful. Am I
correct in thinking that the kernel now uses readpage() everywhere, so if I
implement that to do the Right Thing[1], everything (running executables), etc.,
should work? How about writable mappings -- there's bound to be someone who
wants to swap to a file on an NTFS partition?

Briefly looking through generic_file_read(), how do pages in the page cache
become 'out of date'? Surely writes go through the page cache too, or not?

One last thing: the NTFS driver at the moment doesn't use generic_file_read(),
so we're caching our data in the buffer cache rather than the page cache (I
presume): how much is this losing us?

S.

[1] i.e., if the file data is out-of-inode, just call generic_readpage.
Otherwise work a little bit harder.

-- 
20962296

[prev in list] [next in list] [prev in thread] [next in thread]