Re: pte-highmem-5

Christoph Rohland (cr@sap.com)
Thu, 17 Jan 2002 17:11:48 +0100


Hi Andrea,

On Thu, 17 Jan 2002, Andrea Arcangeli wrote:
> (btw, I suspect allocating one page at offset 4G in every shmfs file
> could make the overhead per page of shm to increase)

Nearly: A sparse file with the only page at 4G is the worst case: You
need three extra pages to hold the swap entry. The ratio goes fown as
soon as you add more pages somewhere.

> But in real life I really don't expect problems here, one left page
> of the vector holds 1024 swap entries, so the overhead is of the
> order of 1/1000. On the top of my head (without any precise
> calculation) 64G of shm would generate stuff of the order of some
> houndred mbytes of ram

Ok, 64GB shm allocate roughly 64MB swap entries, so this case should
not bother too much. I was still at the 390x case where we have 512
entries per page. But they do not need highmem.

Another case are smaller machines with big tmpfs instances: They get
killed by the swap entries. But you cannot hinder that without
swapping the swap entries themselves.

Greetings
Christoph

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/