The goal of the patches of this serie is to clarify the code
of state/event handling.
- First patch (already acked/applied) makes sure that every structures are initialized
before the mediated device is registered.
(Only here to apply the serie on the main tree).
- vfio: ccw: Rework subchannel state on setup
makes sure that the device can not be used before a guest is
ready to drive it.
- vfio: ccw: Rework subchannel state on removing
could be squash with the previous: same kind of rewriting
but for the removing/release callbacks.
- vfio: ccw: Rework subchannel state on sch_event
sch_event handling seems quite poor to me.
Anyway, I do not understand why we hould have state change there.
- vfio: ccw: Documenting state transitions
- vfio: ccw: serialize the write system calls
Quite independent of the previous patches, this makes sure
that the entry in the driver is serialized.
Pierre Morel (6):
vfio: ccw: Register mediated device once all structures are
initialized
vfio: ccw: Rework subchannel state on setup
vfio: ccw: Rework subchannel state on removing
vfio: ccw: Rework subchannel state on sch_event
vfio: ccw: Documenting state transitions
vfio: ccw: serialize the write system calls
Documentation/s390/vfio-ccw.txt | 45 +++++++++++++++++++++++++
drivers/s390/cio/vfio_ccw_async.c | 11 +++++++
drivers/s390/cio/vfio_ccw_drv.c | 30 ++++-------------
drivers/s390/cio/vfio_ccw_ops.c | 65 ++++++++++++++++++++++++-------------
drivers/s390/cio/vfio_ccw_private.h | 1 +
5 files changed, 105 insertions(+), 47 deletions(-)
--
2.7.4
Changelog:
from v2:
- adaptation after Conny's hlt/clr patches
- integration of Conny's comments
- reworking of the state changes to make it clearer
- added documentation
- added serialization
from v1:
- change commit message
- add reviewed-by from Eric (Really this time)
from v0:
- isolate these two patches from the previous serie
- Added Reviewed-by Eric
- Ortographic correction
