> 
> On Tue, 12 Jun 2001, Alexander Viro wrote:
> > 
> > We need a bit more - set_blocksize() is another source of the same problem,
> > AFAICS. How about the following?
> 
> Makes sense.
> 
> The "set_blocksize()" thing looks like it could have bit us before - it
> didn't check "page->mapping", so it could make mapped pages do bad bad
> things already. 
> 
> Although setting the blocksize on something that is already mounted is
> obviously bad in itself, so maybe the old problems just weren't a real
> issue ;)
> 
> Applied.

Hmm, are you sure this will fix it, b_pprev is only cleared by
__hash_unlink via __remove_from_queues and __remove_from_queues
is only called by try_to_free_buffers in the case where it 
decides it is going to actually free the buffer_head. In the
case where it leaves it on the inode, b_pprev is still set.

In this particular scenario we jump right around that code because
of the buffer_busy() test. 

Steve


-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [EMAIL PROTECTED]

Reply via email to