[openib-general][patch review] srp: fmr implementation,

[Michael S. Tsirkin]

Quoting r. Tom Tucker <tom at opengridcomputing.com>:
> If the benefit isn't there -- why do it if the application outcomes are almost
> certainly all bad?

I think I saw Vu's earlier post with numbers showing FMR performance benefits.

I don't know what causes QP hangs and event hangs for Thomas - I don't think I
ever saw such behaviour e.g. in SDP with or without FMRs, I've no idea what is
NFS/RDMA doing differently.

-- 
MST