This month's meetings

<<MonthCalendar: execution failed [Argument "anniversary" must be a boolean value, not "workinggroups/kernel/meetingtemplate"] (see also the log)>>

Attendees

  • Loïc Minier
  • John Stultz
  • Mounir Bsaibes
  • Shajul Abraham
  • Arnd Bergmann
  • Steve Sakoman
  • Paul McKenney

  • Matt Waddel
  • Nicolas Pitre

Minutes

Hot topics for today? Need to continue gathering estimates for next cycle; highmem and SD card FS good topics.

DeviceTree pushed back a bit by Russell upstream; need stronger real life use cases. DeviceTree is not really a clear win at the moment.

We need to carry it for longer to prove its usefulness. It's a bit contradictory to call for real-life use, when usually there is a lot of critiques for shipping code which is not upstream. :-/

What kind of metric could we use to push for DeviceTree? We could use it to limit complexity of the kernel.

It seems DeviceTree doesn't make architecture-specific code much simpler, but it does help drivers which Russell might not be witnessing as much.

ARM could possibly specify more things on the low part of the stack, just like the MMU/CPU are specified; e.g. bus interconnect could be probable, just like PCI/PCIe/USB are probable

Need to be careful not to go the openfirmware route, which is what the flattened device tree moved away from: we should not specify the whole tree of devices, but stop at the level where things get discoverable (PCIe/USB for instance)

Since key people like jk- and gcl were missing from today's call, we should first get their input

In the ST-Ericsson BSP, they have to add to a devices.c every time they add/change a device, which gets very painful; this is a place where devicetree could help a lot

Need to start drafting real engineering blueprints for UDS

  • K1: Memory Utilization: Paul McKenney

  • K2: Standard architecture: Nicolas Pitre
  • K3: BSP Investigation: Dave Martin, Arnd
  • K4: Android: upstreaming of features and patches, setting up repos; Paul McKenney + John Stultz

  • K5: Storage performance: Arnd + Nicolas
  • K6: Configuration: Frederick + Matt, and perhaps Nicolas for the upstream part
  • K7: DeviceTree: need to split it further; interested parties would be welcome; Arnd + Shajul

  • K8: Kexec: Nicolas
  • K9: Testing: Matt + Mounir

WorkingGroups/KernelArchived/Meetings/2010-09-27 (last modified 2013-01-14 19:32:46)