[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: fsck journal replay times (was Re: Is this list active?)



On Tuesday July 17, sct redhat com wrote:
> Hi,
> 
> On Tue, Jul 17, 2001 at 10:47:44AM +1000, Andrew Morton wrote:
> 
> > Forget I said this.  It would work in 2.2, where data is hashed
> > into the buffercache.  But in 2.4 if a data block is shared between
> > two files we wouldn't detect it via a buffer cache lookup.
> 
> Doesn't matter --- if you know that a buffer will always be on one of
> the inode's lists, it doesn't need to be hashed for you to be able to
> check b_inode.  That would require a clean buffers list in addition to
> a dirty one on the inode, but it would certainly work.
> 

But what would it achieve?  It would be quite possible to have two
different buffer heads with two different b_inode pointers, but the
same b_dev and b_blocknr.  The way buffer heads are allocated to
pages, there is no checking that the same phyiscal address doesn't
already have a buffer_head.

NeilBrown





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]