Running davem's previous patch I was able to get try_to_free_page to
fail. The debugging output from try_to_free_page was (from memory):
try_to_free_page(3,0,0): FAIL try(shrink_mmap:shm_swap:swap_out) state[o(1):n(1)] stop[o(3):n(3)]
I will test with SMP next.
Larry
>From: Gerard Roudier <groudier@club-internet.fr>
>cc: Linux <linux-kernel@vger.rutgers.edu>,
> Linus Torvalds <torvalds@transmeta.com>
>MIME-Version: 1.0
>Content-Type: TEXT/PLAIN; charset=US-ASCII
>Sender: owner-linux-kernel@vger.rutgers.edu
>Precedence: bulk
>
>
>David,
>
>I've added to your patch some very simple changes that seem to me
>enough to fix 2.0.30 memory problems.
>Using this patch, I've been unable to break the kernel doing heavy
>disk IOs, with mem=4M and a swap partition.
>I did'nt try other kernel services that can be affected by this changes.
>
>In my opinion, the behaviour of linux under memory stress is very
>different using this patch and may trigger problems in some other kernel
>services I did'nt use in my tests.
>
>As I wrote in my previous mails about this topic, I think it was the
>buffer headers allocation strategy that should be improved and a new gfp
>level seemed to me necessary.
>
>--
>Regards, Gerard.
>
>Here's the patch:
>(It is against pre-2.0.31-2 and does not incorporate other suggested
>patches that seem to me useless.)
>
>....