> > > I just had the following message in my syslog (with the extra
> > > arguments to the try_to_free_page printk :-)
> > >
> > > mvmap66 kernel: try_to_free_page: state(0) stop(3) i(2) sleep instead of fail
>
> I got exactly that message as well, while copying files from one IDE drive
> to another. I have 64 MB RAM and no swap space activated.
It seems that does not prevent kswapd from being activated.
In any case, the change in try_to_free_page() is broken and in my
opinion does not do what its author wanted it shall do.
In your case, the only places where memory can be freed is the
file page cache and the buffer cache.
Very probably, a page has been found in these places in order to
make this @$#!? change happy.
Gerard.