2.3.x swap cache seems to be a big leak

2.3.x swap cache seems to be a big leak

[David S. Miller]

__delete_from_swap_cache depends upon remove_inode_page doing
a put_page or similar to kill the reference of the swap cache
itself

We changed remove_inode_page during the page cache rewrite such
that is no longer puts the page, the caller does.

If I am right, this causes swap cache pages to never go away.

So if I haven't missed something clever going on here, this would
explain a lot of problems people have reported with swapping making
their machines act weird and eventually run out of ram.

Later,
David S. Miller
davem@redhat.com
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

Re: 2.3.x swap cache seems to be a big leak

[Andrea Arcangeli]

On Tue, 25 Apr 2000, David S. Miller wrote:

>__delete_from_swap_cache depends upon remove_inode_page doing
>a put_page or similar to kill the reference of the swap cache
>itself

__delete_from_swap_cache must not decrease the reference count.

>We changed remove_inode_page during the page cache rewrite such
>that is no longer puts the page, the caller does.

shrink_mmap does in the made_inode_progress path (first release the swap
cache reference and then frees the page by issuing two put_page).

>So if I haven't missed something clever going on here, this would
>explain a lot of problems people have reported with swapping making
>their machines act weird and eventually run out of ram.

That's because the MM balancing is broken. Try to reproduce with the
classzone patch applied.

Andrea

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/