On 2/24/12 8:13 PM, Balbir Singh wrote:
A uid based approach such as the one implemented by Davide Libenzi
http://thread.gmane.org/gmane.linux.kernel/548928
http://thread.gmane.org/gmane.linux.kernel/548926
would probably apply the optimization to more use cases - but conceptually a
bit more complex. If we go with this more relaxed approach, we'll have to
design a race-free cgroup_uid_count() based mechanism.
Are you suggesting all processes with the same UID should have access
to each others memory contents?
No - that's a stronger statement than the one I made in my last message.
I'll however observe that something like this is already possible via
PTRACE_PEEKDATA.
Like I said: a cgroup with a single mm_struct is conceptually cleanest
and covers some of our heavy use cases. A cgroup with a single uid would
cover more of our use cases. It'd be good to know if you and other
maintainers are willing to accept the former, but not the latter.
I'll note that the malloc implementation which uses these interfaces can
still decide to zero the memory depending on which variant of *alloc is
called. But then, we'd have more fine grained control and more
flexibility in terms of temporal usage hints.
-Arun
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>