[lkml]   [2009]   [Feb]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: Q: state of pci express hotplug
    Eric W. Biederman wrote:
    > Trying to use the current pciehp code, on real hardware in an
    > interesting topology (more later) I'm not finding it very satisfactory
    > at all. I am up to at least three show stopper bugs and the code
    > is not doing what I really need it to do, so I am looking at a significant
    > rewrite patch effort to make this code usable.

    I'd like to know details of the bugs. Could you please show the
    details of the bugs?

    > Am I unique in thinking there is a lot that needs to be done?
    > Is there current work in progress to make the pci hotplug work better?

    In my personal opinion, there are at least the following items need to
    be done for pciehp.

    o Possible bugs
    -Current pcie_isr function (interrupt service routine of pciehp) has
    a possibility to get into endless loop in the case that bits in Slot
    Status register is set again immediately after cleared. According to
    the past discussion in the pcihpd-discuss ML, this can happen on the
    power fault detected bit on some platforms.

    -Current pciehp disables software notification of adapter presence
    changed event and MRL changed event when slot is turned off. Because
    of this, there is no way to detect those events on empty slots in the
    current pciehp implementation.

    o Missing Features
    -Current pciehp doesn't support Data Link Layer State Changed event
    software notification.
    -No event notification mechanism to userland.
    -When hot-removal request is initiated through the Attention Button
    press, there is no chance to prepare hot-removal (unmount for example).
    -Userland utilities.

    o Cleanups
    -slot list is not needed because there is only one slot under the
    switch downstream port.
    -hpc_ops is redundant.
    -some fields in the struct controller, struct slot are redundant.

    o Misc
    -I don't think power fault management routines are well tested.
    -I don't think EMI management routines are well tested.

    Kenji Kaneshige

     \ /
      Last update: 2009-02-02 03:21    [W:0.043 / U:3.628 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site