Re: 2.5.6: ide driver broken in PIO mode

[Andrew Morton]

Suparna Bhattacharya wrote:
> 
> ...
> However, the latest code I have also covers the avoidance of bv_len,
> bv_offset modifications by the block layer, which I'd been
> concerned about for quite a while and ought to have done something about
> much sooner ;)

urgh.  I didn't know there was a risk of this.

I'm using bv_offset and bv_len in the bi_end_io handler to work out
whether to unlock the final page in the multipage BIO.

That can probably be avoided, but it would be better if these
can be left alone, or at least, restored to their original value
before returning the BIO to whoever created it.

I'm also using bi_private, under the assumption that the ownership
rules for that are analogous to buffer_head.b_private.   Is this
correct?   Who owns bi_private?

-