Re: [PATCH] Reserve device id for RDMA device

[Junji Wei <weijunji@bytedance.com>]

> On Sep 22, 2021, at 7:28 PM, Cornelia Huck <cohuck@redhat.com> wrote:
> 
> On Fri, Sep 03 2021, Junji Wei <weijunji@bytedance.com> wrote:
> 
>> Use device ID 42
>> 
>> Signed-off-by: Junji Wei <weijunji@bytedance.com>
>> ---
>> content.tex | 2 ++
>> 1 file changed, 2 insertions(+)
>> 
>> diff --git a/content.tex b/content.tex
>> index 3aeb4a4..b57c997 100644
>> --- a/content.tex
>> +++ b/content.tex
>> @@ -2878,6 +2878,8 @@ \chapter{Device Types}\label{sec:Device Types}
>> \hline
>> 41         &   GPIO device \\
>> \hline
>> +42         &   RDMA device \\
>> +\hline
>> \end{tabular}
>> 
>> Some of the devices above are unspecified by this document,
> 
> Do we have an answer to the issues raised with the QEMU patches? IOW, is

Yes.

> it reasonable to expect that we can actually implement a usuable RDMA
> device? While device IDs are generally cheap, it would be good not to
> have too many dead entries.

I think we can implement an usable RDMA device through uRDMA (a
software implementation of the RoCEv2 protocol in QEMU with VFIO
or DPDK).

Thanks.
Junji