[openib-general] Re: Mellanox HCAs: outstanding RDMAs

[Talpey, Thomas]

At 03:09 AM 6/6/2006, Michael S. Tsirkin wrote:
>Quoting r. somenath <somenath at veritas.com>:
>> possibility of stalling is scary!
>
>You might want to review chapter 9.5 TRANSACTION ORDERING for info on when will
>ordering rules cause the IB QP to stall.

MST, are you disagreeing that RDMA Reads can stall the queue?
Section 9.5, C9-25 lays it right out as the first requirement:

>> C9-25: A requester shall transmit request messages in the order that the
>> Work Queue Elements (WQEs) were posted.

Therefore, a provider which implements flow control on RDMA Reads cannot
transmit new sends until the prior RDMA Reads can be initiated. Of course,
they may complete in a somewhat different order...

It's all about flow control - which is not mandatory. It's a convenient, but
very risky thing. Upper layers are often unaware of its ramifications.

Tom.