Re: [PATCH] Fix generic_block_fiemap for files bigger than 4GB

[Josef Bacik <josef@redhat.com>]

On Wed, Oct 28, 2009 at 11:00:22PM -0700, Andrew Morton wrote:
> On Mon, 26 Oct 2009 18:24:28 +0100 Mike Hommey <mh@glandium.org> wrote:
> 
> > Because of an integer overflow on start_blk, various kind of wrong results
> > would be returned by the generic_block_fiemap handler, such as no extents
> > when there is a 4GB+ hole at the beginning of the file, or wrong fe_logical
> > when an extent starts after the first 4GB.
> > 
> > Signed-off-by: Mike Hommey <mh@glandium.org>
> > ---
> >  fs/ioctl.c |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> > 
> > diff --git a/fs/ioctl.c b/fs/ioctl.c
> > index 7b17a14..6c75110 100644
> > --- a/fs/ioctl.c
> > +++ b/fs/ioctl.c
> > @@ -254,7 +254,7 @@ int __generic_block_fiemap(struct inode *inode,
> >  			   u64 len, get_block_t *get_block)
> >  {
> >  	struct buffer_head tmp;
> > -	unsigned int start_blk;
> > +	unsigned long long start_blk;
> >  	long long length = 0, map_len = 0;
> >  	u64 logical = 0, phys = 0, size = 0;
> >  	u32 flags = FIEMAP_EXTENT_MERGED;
> 
> Well.  Should it be unsigned long long, or u64 or sector_t?  Or even loff_t.
> 
> The code's a bit confused about types in there.  And it's made much
> more confusing by the moronic and wholly unnecessary use of macros for
> blk_to_logical() and logical_to_blk().
> 
> It's also unhelpful that the `u64 start' argument forgot to get itself
> documented in the kerneldoc comment.  Sigh.
> 
> Ah, generic_block_fiemap() has it:
> 
>  * @start: The initial block to map
> 
> I guess u64 was logical there as it comes in from userspace.  But at
> some boundary we should start talking kernel types so I suspect the
> correct thing to do here is to use sector_t?
> 

Hmm this is strange, I had sent a patch a few months ago after you chewed me out
the first time for this to change all the types and such and make the macro's
inlined functions.  I will see if I can find it and resend it.  Thanks,

Josef