Re: ramdisk corruption problems - was: RE: pivot_root and initrd

[Linus Torvalds]

On Fri, 21 Dec 2001, Alexander Viro wrote:
>
>
> On Thu, 20 Dec 2001, Linus Torvalds wrote:
>
> >
> > On Thu, 20 Dec 2001, Alexander Viro wrote:
> > >
> > > AFAICS, it's nastier than that.  What's to stop buffer_heads to be
> > > freed under memory pressure?
> >
> > They'll be dirty if they are modified, and if they aren't modified they
> > are zero.
> >
> > In the first case they won't be free'd, in the second care we don't care.
> > So we're ok.
>
> Umm...  _Page_ will be dirty, all right.  buffer_heads are cleaned when
> you submit IO and I don't see anything that would mark them dirty again...

But the "submit IO" will _mark_ the page dirty and uptodate, so once it is
so marked, it should never be overwritten.

Now, as I already sent out to Andrea, I think out actual IO routines look
suspiciously complex and bogus.

		Linus