Re: innd mmap bug in 2.4.0-test12

From: Daniel Phillips (
Date: Thu Dec 28 2000 - 10:03:24 EST

Linus Torvalds wrote:
> On Wed, 27 Dec 2000, Philipp Rumpf wrote:
> > On Wed, Dec 27, 2000 at 03:41:04PM -0800, Linus Torvalds wrote:
> > > It must be wrong.
> > >
> > > If we have a dirty page on the LRU lists, that page _must_ have a mapping.
> >
> > What about pages with a mapping but without a writepage function ? or pages
> > whose writepage function fails ? The current code seems to simply put the
> > page onto the active list in that case, which seems just as wrong to me.
> ramfs. It doesn't have a writepage() function, as there is no backing
> store.
> > > The bug is somewhere else, and your patch is just papering it over. We
> > > should not have a page without a mapping on the LRU lists in the first
> > > place, except if the page has anonymous buffers (and such a page cannot
> >
> > So is there any legal reason we could ever get to page_active ? Removing
> > that code (or replacing it with BUG()) certainly would make page_launder
> > more readable.
> Apart from the "we have no backing store", there is no legal reason to put
> it back on the active list that I can see.

It's logical that PageDirty should never be get for ramfs, and a ramfs
page should never have buffers on it. With this and Chris's anon_space
mapping can we replace the check for null ->writepage with BUG? With
the anon_space mapping we should be able to do the same ford for

Though these things aren't strictly bugs, having to check multiple paths
for everything is slowing us down in picking off the fluff.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

This archive was generated by hypermail 2b29 : Sun Dec 31 2000 - 21:00:11 EST