Hi, Please pardon the late submission. I'd like to propose "Memory management challenges for embedded systems" as an agenda topic for the Linux Storage, Filesystem, and Multimedia Summit in San Francisco. I am a kernel developer on the ChromiumOS project. I work on a number of different areas of the kernel but memory management is definitely a key focus area for our team and myself. ChromiumOS much like a typical embedded system in that the resources are constrained and the entire system is tuned for one application: in our case, the browser. >From an MM perspective, we've avoided using memory containment since there is just a single application. For security, we avoid swap (no unencrypted user data on persistent store). We also avoid swap because of the variability it adds to latency. Since we don't have swap, running out of memory is a constant challenge. We don't mind OOMing but prefer it would happen quickly to avoid thrashing. We've used a tiny hack in order to make this happen: http://lwn.net/Articles/412313/ The hack requires us to predict working set size, it would be nice if the kernel automatically calculated it. It would be nice if vmscan's active and inactive list modeled the real working set. We have some ideas on how to do that: https://lkml.org/lkml/2010/11/3/394 Or perhaps it would be preferable to avoid OOM and instead let user-space take some action: http://lwn.net/Articles/475791/ But letting userspace take action on behalf of the kernel is always tricky so maybe it would be better have something like tmem in userspace: http://lwn.net/Articles/468896/ http://lwn.net/Articles/340080/ Regards, Mandeep -- 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>