On Wed, Feb 06, 2002 at 10:19:32PM -0800, Andrew Morton wrote:
> It's only a problem if this is the final put_page(). In the
> case of sendfile(), process-context code can be taught to take
> a temporary reference on the page, and only release it after the network
> stack is known to have finished with the page. sendfile is synchronous, yes?
>
> And in the case of all other skb frees, the underlying page
> won't be on the LRU. I hope.
sendfile isn't synchronous, nor is aio, which also relies on freeing pages
acquired from user mappings in irq/bh/whatever context. Restricting where
pages may be freed really ties our hands on what is possible for zero copy
io. (ie O_DIRECT, aio, sendfile, TUX all get hosed by this).
-ben (who is really on vacation)
-- begin 644 fish.com 866]U(&AA=F4@=&]O(&UU8V@@=&EM92X* ` end - 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/
This archive was generated by hypermail 2b29 : Thu Feb 07 2002 - 21:00:59 EST