Note that there still remain a couple of node type checks in the kernel
that we may want to remove. There's an IB CA check in cma.c:rdma_notify
This should check for rdma_cap_ib_cm()
as well as in rds/ib.c:rds_ib_add_one and rds_ib_laddr_check and
Not sure what the underlying reason is for
On 6/22/2015 4:20 PM, Jason Gunthorpe wrote:
On Thu, Jun 18, 2015 at 07:25:32PM -0400, Hal Rosenstock wrote:
Persuant to Liran's comments on node_type on linux-rdma
mailing list:
In an effort to reform the RDMA core and ULPs to minimize use of
node_type in struct ib_device, an additional bit
On 6/22/2015 6:40 PM, Hefty, Sean wrote:
Note that there still remain a couple of node type checks in the kernel
that we may want to remove. There's an IB CA check in cma.c:rdma_notify
This should check for rdma_cap_ib_cm()
IB, RoCE, and OPA ports have the RDMA_CORE_CAP_IB_CM set so this
On Thu, Jun 18, 2015 at 07:25:32PM -0400, Hal Rosenstock wrote:
Persuant to Liran's comments on node_type on linux-rdma
mailing list:
In an effort to reform the RDMA core and ULPs to minimize use of
node_type in struct ib_device, an additional bit is added to
struct ib_device for is_switch
rds_ib_add_one says:
/* Only handle IB (no iWARP) devices */
but not sure comment is 100% accurate as it checks node type != IB CA.
rds_ib_laddr_check says:
/* rdma_bind_addr will only succeed for IB iWARP devices */
/* due to this, we will claim to support iWARP devices unless we check
In an effort to reform the RDMA core and ULPs to minimize use of
node_type in struct ib_device, an additional bit is added to
struct ib_device for is_switch (IB switch). This is needed
to be initialized by any IB switch device driver. This is a
NEW requirement on such device drivers which are