linux-toradex.git/drivers/s390/cio, branch v5.14

s390/qdio: get rid of register asm

2021-06-28T09:18:29+00:00

Reviewed-by: Benjamin Block 
Signed-off-by: Heiko Carstens 
Signed-off-by: Vasily Gorbik

s390/ioasm: use symbolic names for asm operands

2021-06-28T09:18:29+00:00

Change remaining inline assemblies to also use symbolic names for
operands, use "Q" constraint when possible, and use "+" modifier so
that an operand needs to be specified only once.

Acked-by: Peter Oberparleiter 
Signed-off-by: Heiko Carstens 
Signed-off-by: Vasily Gorbik

s390/ioasm: get rid of register asm

2021-06-28T09:18:28+00:00

Acked-by: Peter Oberparleiter 
Signed-off-by: Heiko Carstens 
Signed-off-by: Vasily Gorbik

s390/cmf: get rid of register asm

2021-06-28T09:18:28+00:00

Acked-by: Peter Oberparleiter 
Signed-off-by: Heiko Carstens 
Signed-off-by: Vasily Gorbik

s390/cio: dont call css_wait_for_slow_path() inside a lock

2021-06-18T14:41:19+00:00

Currently css_wait_for_slow_path() gets called inside the chp->lock.
The path-verification-loop of slowpath inside this lock could lead to
deadlock as reported by the lockdep validator.

The ccw_device_get_chp_desc() during the instance of a device-set-online
would try to acquire the same 'chp->lock' to read the chp->desc.
The instance of this function can get called from multiple scenario,
like probing or setting-device online manually. This could, in some
corner-cases lead to the deadlock.

lockdep validator reported this as,

        CPU0                    CPU1
        ----                    ----
   lock(&chp->lock);
                                lock(kn->active#43);
                                lock(&chp->lock);
   lock((wq_completion)cio);

The chp->lock was introduced to serialize the access of struct
channel_path. This lock is not needed for the css_wait_for_slow_path()
function, so invoke the slow-path function outside this lock.

Fixes: b730f3a93395 ("[S390] cio: add lock to struct channel_path")
Cc: 
Reviewed-by: Peter Oberparleiter 
Signed-off-by: Vineeth Vijayan 
Signed-off-by: Vasily Gorbik

s390: use struct tpi_info in lowcore.h

2021-06-07T15:07:01+00:00

Signed-off-by: Sven Schnelle 
Signed-off-by: Heiko Carstens 
Signed-off-by: Vasily Gorbik

s390: add struct tpi_info to struct pt_regs

2021-06-07T15:07:01+00:00

To avoid casting ptrace members, add a union containing
both struct tpi_info and explicit int_code/int_parm members.

Signed-off-by: Sven Schnelle 
Signed-off-by: Heiko Carstens 
Signed-off-by: Vasily Gorbik

s390/cio: add tpi.h header file

2021-06-07T15:07:01+00:00

For future work we need the struct tpi_info declaration in asm/ptrace.h.
Due to circular dependencies it cannot stay in asm/lowcore.h or asm/cio.h,
which would be the preferred location. Therefore add it in its own header
file.

Also fix a typo in the length of a reserved field that did not have a
functional effect beyond an incorrect field value in the s390_cio_tpi
tracepoint.

Fixes: 2ab59de7c5ce ("s390/cio: Consolidate inline assemblies and related data definitions")
Signed-off-by: Sven Schnelle 
Acked-by: Peter Oberparleiter 
Signed-off-by: Heiko Carstens 
Signed-off-by: Vasily Gorbik

s390/ccwgroup: use BUS_NOTIFY_UNBOUND_DRIVER to trigger ungrouping

2021-06-07T15:07:00+00:00

ccwgroup_notifier() currently listens for BUS_NOTIFY_UNBIND_DRIVER
events, and triggers an ungrouping for the affected device.

Looking at __device_release_driver(), we can wait for a little longer
until the driver has been fully unbound and eg. bus->remove() has been
called. So listen for BUS_NOTIFY_UNBOUND_DRIVER instead. Due to locking
the current code should work just fine, but this clarifies our intent.

Signed-off-by: Julian Wiedmann 
Reviewed-by: Vineeth Vijayan 
Signed-off-by: Vasily Gorbik

s390/ccwgroup: simplify ungrouping when driver deregisters

2021-06-07T15:07:00+00:00

driver_unregister() ends up calling __device_release_driver() for
each device that is bound to the driver. Thus ccwgroup_notifier() will
see a BUS_NOTIFY_UNBIND_DRIVER event for these ccwgroup devices, and
trigger the ungrouping.

So there's no need to also trigger the ungrouping from within
ccwgroup_driver_unregister(), remove it.

Signed-off-by: Julian Wiedmann 
Reviewed-by: Vineeth Vijayan 
Signed-off-by: Vasily Gorbik