zfs discuss - Nov 2007 - zpool detech hangs causes other zpool commands, format, df etc.. to hang

Customer has a Thumper running:

SunOS x4501 5.10 Generic_120012-14 i86pc i386 i86pc

where running "zpool detech disk c6t7d0" to detech a mirror causes
zpool
command to hang with following kernel stack trace:

PC: _resume_from_idle+0xf8    CMD: zpool detach disk1 c6t7d0
  stack pointer for thread fffffe84d34b4920: fffffe8001c30c10
  [ fffffe8001c30c10 _resume_from_idle+0xf8() ]
    swtch+0x110()
    cv_wait+0x68()
    spa_config_enter+0x50()
    spa_vdev_enter+0x2a()
    spa_vdev_detach+0x39()
    zfs_ioc_vdev_detach+0x48()
    zfsdev_ioctl+0x13e()
    cdev_ioctl+0x1d()
    spec_ioctl+0x50()
    fop_ioctl+0x25()
    ioctl+0xac()
    sys_syscall32+0x101()


Other zpool commands, df, format all waiting on a mutex lock
spa_namespace_lock to release:

  PC: _resume_from_idle+0xf8    CMD: zpool status
  stack pointer for thread fffffe84d34b3ba0: fffffe8001439bf0
  [ fffffe8001439bf0 _resume_from_idle+0xf8() ]
    swtch+0x110()
    turnstile_block+0x1fb()
    mutex_vector_enter+0x40f()
    spa_all_configs+0x48()
    zfs_ioc_pool_configs+0x20()
    zfsdev_ioctl+0x13e()
    cdev_ioctl+0x1d()
    spec_ioctl+0x50()
    fop_ioctl+0x25()
    ioctl+0xac()
    sys_syscall32+0x101()
========
 PC: _resume_from_idle+0xf8    CMD: zpool list
  stack pointer for thread fffffe875856bf20: fffffe8001f7ebf0
  [ fffffe8001f7ebf0 _resume_from_idle+0xf8() ]
    swtch+0x110()
    turnstile_block+0x1fb()
    mutex_vector_enter+0x40f()
    spa_all_configs+0x48()
    zfs_ioc_pool_configs+0x20()
    zfsdev_ioctl+0x13e()
    cdev_ioctl+0x1d()
    spec_ioctl+0x50()
    fop_ioctl+0x25()
    ioctl+0xac()
    sys_syscall32+0x101()

======PC: _resume_from_idle+0xf8    CMD: format
  stack pointer for thread fffffe84d34b4260: fffffe8001b73b90
  [ fffffe8001b73b90 _resume_from_idle+0xf8() ]
    swtch+0x110()
    turnstile_block+0x1fb()
    mutex_vector_enter+0x40f()
    spa_open_common+0x5f()
    spa_get_stats+0x42()
    zfs_ioc_pool_stats+0x2c()
    zfsdev_ioctl+0x13e()
    cdev_ioctl+0x1d()
    spec_ioctl+0x50()
    fop_ioctl+0x25()
    ioctl+0xac()
    sys_syscall32+0x101()
====
 PC: _resume_from_idle+0xf8    CMD: df -h
  stack pointer for thread fffffe84d34b3f00: fffffe8001eb8960
  [ fffffe8001eb8960 _resume_from_idle+0xf8() ]
    swtch+0x110()
    turnstile_block+0x1fb()
    mutex_vector_enter+0x40f()
    spa_open_common+0x5f()
    spa_open+0xb()
    dsl_dir_open_spa+0x22e()
    dsl_dataset_open_spa+0x41()
    dsl_dataset_open+0x16()
    dmu_objset_open+0x49()
    zfs_ioc_objset_stats+0x3d()
    zfsdev_ioctl+0x13e()
    cdev_ioctl+0x1d()
    spec_ioctl+0x50()
    fop_ioctl+0x25()
    ioctl+0xac()
    sys_syscall32+0x101()

I have asked customer to provide a core dump. Still waiting for it to
arrive. I would like to know if it is a known bug.


-- 
Amer Ather	
TSC-KERNEL     			
amer.ather at Sun.COM		
408-276-9780 (x19780)				
email Pager: 4082306720 at messaging.sprintpcs.com

	" If you fail to prepare, prepare to fail"