• K
    NVMe: Admin queue removal handling · 0fb59cbc
    Keith Busch 提交于
    This protects admin queue access on shutdown. When the controller is
    disabled, the queue is frozen to prevent new entry, and unfrozen on
    resume, and fixes cq_vector signedness to not suspend a queue twice.
    
    Since unfreezing the queue makes it available for commands, it requires
    the queue be initialized, so this moves this part after that.
    
    Special handling is done when the device is unresponsive during
    shutdown. This can be optimized to not require subsequent commands to
    timeout, but saving that fix for later.
    
    This patch also removes the kill signals in this path that were left-over
    artifacts from the blk-mq conversion and no longer necessary.
    Signed-off-by: NKeith Busch <keith.busch@intel.com>
    Signed-off-by: NJens Axboe <axboe@fb.com>
    0fb59cbc
nvme-core.c 71.8 KB