Add an io_locked TAILQ to each channel, which hold
IO that will write to a currently locked region.
Also add a new step to the locking process per channel.
Each channel needs to wait until all existing outstanding
writes to the newly locked range have been completed.
Only the channel that locked an LBA range may submit
write I/O to it. It must use the same cb_arg for the
write I/O as the cb_arg used when locking the LBA range.
This ensures that only the specific I/O operations needing
the lock will bypass the lock.
When a range is unlocked, we will just blindly try to
resubmit all IO in the io_locked tailq. This could be
made more efficient in the future, but we don't expect
this path to occur very often, so going for simplicity
in the first pass.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ibdc992144dfaffe7c05471a5b3c020cedd8cdfc3
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478226
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Maciej Szwed <maciej.szwed@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
b90b7ce477