[lkml]   [2011]   [Dec]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: [PATCH-WIP] convert CPU sysdev class to a real subsytem (needed for CPU modaliases)
    On Thu, Dec 08, 2011 at 02:58:37AM +0100, Kay Sievers wrote:
    > This patch is work-in-progress and just posted to coordinate with Andi
    > Kleen's current export of CPU modaliases. It boots my x86_64 workstation
    > fine, nothing else is tested.
    > The CPU subsystem from a sysdev_class to a plain driver-core subsystem
    > (currently called 'bus').
    > With the CPUs as normal devices, userspace will pick the events up
    > properly and udev is able to coldplug the devices during bootup, which
    > will allow proper module autoloading based on MODALIAS.
    > $ tree /sys/bus/cpu/
    > /sys/bus/cpu/
    > ├── devices
    > │ ├── cpu0 -> ../../../devices/system/cpu/cpu0
    > │ ├── cpu1 -> ../../../devices/system/cpu/cpu1
    > │ ├── cpu2 -> ../../../devices/system/cpu/cpu2
    > │ └── cpu3 -> ../../../devices/system/cpu/cpu3
    > ...
    > └── uevent
    > The needed bits to easily convert the remaining sysdev class users to a
    > plain subsystem and plain devices have been added to the driver core.
    > Sysdev drivers have never been 'drivers' in the driver-core sense, and
    > they can be now ported-over to subsys_interfaces. The end goal is to
    > finally remove all of the ugly sysdev hack from the system.
    > Greg, I guess we need a bit of your time, to make a plan and get the
    > remaining things sorted out. We are planning and fiddling for years
    > already here, and we should finally get the stuff out. :)

    I agree, I want this to get done finally.

    > Greg, I think with the subsys_interfaces we do not need the
    > 'multi-driver-binding' anymore, seems we can get away with the
    > interfaces only. Classes which need to convert to buses because the
    > class interface is not extensible can be easily converted now. The only
    > missing piece is the 'glue directory'. We should be able to unify
    > classes and buses internally with all that stuff.

    Ah, good, I was stalled on this because of the multi-driver-binding
    stuff. If that's not needed, I'll gladly take any patches for this.

    So, what do I need to do here? Your patch looks good at first glance,
    the interface logic looks sane, is that all that we really need to do to
    the core?


    greg k-h
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to
    More majordomo info at
    Please read the FAQ at

     \ /
      Last update: 2011-12-09 01:15    [W:0.025 / U:3.128 seconds]
    ©2003-2017 Jasper Spaans. hosted at Digital OceanAdvertise on this site