On Mon, 2 Aug 1999 15:50:36 -0400 (EDT), Chuck Lever <cel@monkey.org> said:
Sorry for the delay, I've been on holiday for 2 weeks.
>> Consider also an application which mmaps a data file and performs
>> several sequential passes over that memory. Unmapping pages behind the
>> fault will just cause huge amounts of totally unnecessary page table
>> activity when we perform the next pass over the data, and remember, page
>> table modifications are really expensive on threaded SMP tasks.
> ok, i may have missed something, but i'm not talking about unmapping.
> all i want to do is "unreference" these pages -- simply clear the
> reference bit. if there is no memory pressure, nothing else will happen.
OK, sounds much more reasonable!
> a max-rss is already maintained (via get/setrlimit).
It is maintained but not adhered to.
> are you thinking of some other value that could be computed by the
> VM system?
Yes. Dynamic upper and lower RSS bounds have _lots_ of advantages. A
lower bound lets tasks like login make progress even during heavy
memory pressure, and making the limit dynamic still allows us to swap
out those logins once they become idle. An upper dynamic bound lets
you perform swapout on a large memory-bound task without overloading
the entire memory system for other tasks.
Just as importantly, these limits would give us a mechanism by which
to track the page fault rates for different tasks and to
preferentially swap out specific tasks in response.
--Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/