Hi all, This is a try to use ib_client API in order to wrap ib_device with a specific ULP device, which is created on add() callback and destroyed on remove(). Patches change initiator and target sides for NVMEoRDMA and ISER. Changes do not bring anything in terms of LoC but extra mutexes and lists are vanished, since ib_client API does required locking internally. Also would be nice to avoid refcounting and synchronization in ISER case, but ULP ib_clients remove() callback is called before cma_client, thus ULP device should be alive till last reference put (NVMe queues explicit delete works with workqueue flush, which guarantees us that nvme_dev_dev_put() puts last reference in remove() callback). This is RFC, since advantages of these patches are not quite clear. Roman Pen (5): nvme-rdma: use ib_client API to wrap ib_device nvmet-rdma: use ib_client API to wrap ib_device IB/iser: use ib_client API to wrap ib_device IB/isert: embed mutex inside isert_device structure IB/isert: use ib_client API to wrap ib_device drivers/infiniband/ulp/iser/iscsi_iser.c | 21 +++++-- drivers/infiniband/ulp/iser/iscsi_iser.h | 14 ++--- drivers/infiniband/ulp/iser/iser_verbs.c | 69 ++++++++++++-------- drivers/infiniband/ulp/isert/ib_isert.c | 105 ++++++++++++++++++++----------- drivers/infiniband/ulp/isert/ib_isert.h | 4 +- drivers/nvme/host/rdma.c | 82 +++++++++++------------- drivers/nvme/target/rdma.c | 85 +++++++++++++------------ 7 files changed, 221 insertions(+), 159 deletions(-) Signed-off-by: Roman Pen <roman.penyaev@xxxxxxxxxxxxxxx> Cc: Christoph Hellwig <hch@xxxxxx> Cc: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx> Cc: Bart Van Assche <bart.vanassche@xxxxxxxxxxx> Cc: Sagi Grimberg <sagi@xxxxxxxxxxx> Cc: Doug Ledford <dledford@xxxxxxxxxx> -- 2.13.1 -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
