> Hi Sagi,

Hey Stephen,

>> So it looks from the report that this is the immediate-data and
>> unsolicited data-out flows, which indeed seem to violate the alignment
>> assumption. The reason is that isert post recv a contig rx_desc which
>> has both the headers and the data, and when it gets immediate_data it
>> will set the data sg to rx_desc+offset (which are the headers).
>>
>> Stephen,
>> As a work-around for now, you should turn off immediate-data in your LIO
>> target. I'll work on a fix.
> 
> I'm writing to see if you have had a chance to work on a fix for this yet?

Sorry for the late reply, lost track of this.

> We have a couple of followups as well:
> 
> - Based on Ming's previous comments, is it safe to assume all non-brd 
> drivers and nvme in particular are able to handle unaligned data? We are 
> concerned to enable ImmedataData in general, with this bug outstanding.
> 
> - We are seeing a 5-10% performance degradation when using 
> ImmediateData=No for our workloads. That is to say, having this issue 
> fixed is important to us. Anything we can help with, to provide testing 
> or otherwise, please let us know.

Can you try attached patch and see if it solves your issue?
WARNING: very lightly tested...