kernel.git/drivers/s390/cio, branch linux-4.8.y

s390/cio: fix accidental interrupt enabling during resume

2016-10-31T11:02:12Z

commit d53c51f26145657aa7c55fa396f93677e613548d upstream. Since commit 9f3d6d7 chsc_get_channel_measurement_chars is called with interrupts disabled during resume from hibernate. Since this function used spin_unlock_irq, interrupts have been enabled accidentally. Fix this by using the irqsave variant. Since we can't guarantee the IRQ-enablement state for all (future/ external) callers, change the locking in related functions to prevent similar bugs in the future. Fixes: 9f3d6d7 ("s390/cio: update measurement characteristics") Signed-off-by: Sebastian Ott Reviewed-by: Peter Oberparleiter Signed-off-by: Martin Schwidefsky Signed-off-by: Greg Kroah-Hartman

s390/qdio: avoid reschedule of outbound tasklet once killed

2016-08-08T14:01:28Z

During qdio_shutdown the queue tasklets are killed for all inbound and outbound queues. The queue structures might be freed after qdio_shutdown. Thus it must be guaranteed that these queue tasklets are not rescheduled after that. In addition the outbound queue timers are deleted and it must be guaranteed that these timers are not restarted after qdio_shutdown processing. Timer deletion should make use of del_timer_sync() to make sure qdio_outbound_timer() is finished on other CPUs as well. Queue tasklets should be scheduled in state QDIO_IRQ_STATE_ACTIVE only. Signed-off-by: Ursula Braun Reviewed-by: Benjamin Block Signed-off-by: Martin Schwidefsky

s390/qdio: remove checks for ccw device internal state

2016-08-08T13:41:34Z

Prior to starting IO qdio checks for the internal state of the ccw device. These checks happen without locking, so consistency between state evaluation and starting of the IO is not guaranteed. Since the internal state is checked during ccw_device_start it is safe to get rid of these additional checks. Signed-off-by: Sebastian Ott Reviewed-by: Cornelia Huck Signed-off-by: Martin Schwidefsky

s390/qdio: fix double return code evaluation

2016-08-08T13:41:34Z

qdio sometimes checks return codes twice. First with the ccw device's lock held and then a 2nd time after the lock is released. Simplify the code by releasing the lock earlier and unify the return code evaluation. Signed-off-by: Sebastian Ott Reviewed-by: Ursula Braun Signed-off-by: Martin Schwidefsky

s390/qdio: get rid of spin_lock_irqsave usage

2016-08-08T13:41:33Z

All qdio functions that use spin_lock_irqsave are never used from irq context. Thus it is safe to convert all of them to use spin_lock_irq. Signed-off-by: Sebastian Ott Reviewed-by: Ursula Braun Signed-off-by: Martin Schwidefsky

s390/cio: remove subchannel_id from ccw_device_private

2016-08-08T13:41:33Z

A copy of struct subchannel_id is maintained in ccw_device_private. The subchannel id is a property of the subchannel. The additional copy is not needed. Internal users can obtain it from subchannel.schid - device drivers can use ccw_device_get_schid(). Signed-off-by: Sebastian Ott Reviewed-by: Cornelia Huck Signed-off-by: Martin Schwidefsky

s390/qdio: obtain subchannel_id via ccw_device_get_schid()

2016-08-08T13:41:33Z

We want to get rid of the copy of struct subchannel_id maintained in ccw_device_private, so obtain it using ccw_device_get_schid(). Signed-off-by: Sebastian Ott Reviewed-by: Cornelia Huck Reviewed-by: Ursula Braun Signed-off-by: Martin Schwidefsky

s390/cio: stop using subchannel_id from ccw_device_private

2016-08-08T13:41:32Z

We want to get rid of the copy of struct subchannel_id maintained in ccw_device_private, so obtain it from the subchannel directly. Signed-off-by: Sebastian Ott Reviewed-by: Cornelia Huck Signed-off-by: Martin Schwidefsky

s390/cio: fix premature wakeup during chp configure

2016-07-31T09:27:57Z

We store requests for channel path configure operations in an array but maintain an additional cfg_busy variable (indicating if we have requests stored in said array). When 2 tasks request a channel path configure operation cfg_busy could be set to 0 even if we still have unprocessed requests. This would lead to the second task being woken up although its request was not processed yet. Fix that by getting rid of cfg_busy and use the chp_cfg_task array in the wake up condition. Signed-off-by: Sebastian Ott Reviewed-by: Peter Oberparleiter Signed-off-by: Martin Schwidefsky

s390/cio: convert cfg_lock mutex to spinlock

2016-07-31T09:27:57Z

cfg_lock is never held long and we don't want to sleep while the lock is being held. Thus it can be converted to a simple spinlock. In addition we can now use the lock during the evaluation of a wake_up condition. Signed-off-by: Sebastian Ott Reviewed-by: Peter Oberparleiter Signed-off-by: Martin Schwidefsky