Re: [PATCH] let fat handle MS_SYNCHRONOUS flag

[Matt Mackall <mpm@selenic.com>]

On Wed, Nov 24, 2004 at 08:34:30AM +0100, Colin Leroy wrote:
> On 23 Nov 2004 at 19h11, Matt Mackall wrote:
> 
> Hi, 
> 
> > BUG_ON(!bh);
> > sync_dirty_buffer(bh);
> > brelse(bh);
> 
> I wasn't sure sync_dirty_buffer and brelse checked for nullity :)

It may, that wasn't my point. Your original patch had BUG_ON(1) which
by itself was weird (use BUG() instead). But then it was in the else
part of an if statement. So it read like:

if (bh) {
   ...
} else {
  if (1)
     BUG(); /* stop kernel */
}

BUG is for reporting things that should never happen (otherwise you'd
actually handle them) and so should be used in such a way that they
don't complicate the code flow.

> > Concept seems good, and the implementation otherwise looks good at
> > first glance.
> 
> Cool :) Should I submit an updated patch to Andrew for -mm ?

Probably ought to go through Ogawa, if he can be convinced to take it.
Please take a look at adding -o sync and -o async options to override
the superblock flag first.

-- 
Mathematics is the supreme nostalgia of our time.