[laptop-discuss] Re: Cardbus report: PCMCIA CDMA modem

[Rui Wang]

Douglas Atique wrote:

>I installed sol-nv-b48 and the cardbus driver that ships with it was able to initialize my card (the blue led went on) however doing prtconf -D hangs the machine. I can even make it go to init state 5 by means of the power button, but it looks stuck in some mutex or other synchronization object in the kernel (probably in one of the drivers involved in controlling my card).
>
>  
>
This is indeed "CR6370998 PM ndi_devi_online code path does 
ndi_devi_enter bottom-up, causing deadlocks"
The following three threads cause the deadlock:
d04fd200 d0335210 d69c98b0   1  59 d6bb5678
  PC: 0xfe82bff8    CMD: prtconf -D
  stack pointer for thread d04fd200: d057eca8
    swtch+0x168()
    cv_wait+0x4e()
    ndi_devi_enter+0x4a()
    di_copynode+0x626()
    di_copytree+0x9c()
    di_snapshot+0x13c()
    di_snapshot_and_clean+0x16()
    di_ioctl+0x3c6()
    cdev_ioctl+0x2e()
    spec_ioctl+0x65()
    fop_ioctl+0x27()
    ioctl+0x151()
    sys_sysenter+0x101()
d06c3de0 fec1f27c        0   0  60 d6bb5678
  PC: 0xfe82bff8    THREAD: mt_config_thread()
  stack pointer for thread d06c3de0: d06c38f8
    swtch+0x168()
    cv_wait+0x4e()
    ndi_devi_enter+0x4a()
    pm_lock_power_single+0x7e()
    pm_default_ctlops+0x78()
    pm_ctlops+0x4c()
    pm_lock_power+0x38()
    pm_busop_set_power+0x1b7()
    pm_busop_bus_power+0x1c0()
    pm_all_to_normal_nexus+0x86()
    pm_busop_bus_power+0x6f()
    pm_busop_bus_power+0x97()
    pm_busop_bus_power+0x97()
    pm_set_power+0xbc()
    pm_all_to_normal+0x60()
    pm_pre_config+0x30()
    devi_config_common+0x2a()
    mt_config_thread+0x40()
    thread_start+8()
d3a21de0 fec1f27c        0   0  60 d17117e8
  PC: 0xfe82bff8    THREAD: taskq_d_thread()
  stack pointer for thread d3a21de0: d3a21c28
    swtch+0x168()
    cv_wait+0x4e()
    mt_config_fini+0x25()
    config_grand_children+0x2f()
    devi_config_common+0xb2()
    ndi_devi_config+0x13()
    ndi_devi_online+0xd6()
    hubd_hotplug_thread+0x461()
    taskq_d_thread+0x9c()
    thread_start+8()
It is a problem in our power-management code. I've seen this before. It 
usually happens when the PC Card has an OHCI inside and the OHCI has at 
least one child.

Vincent.