Hi Moni, Is there any progress on this? Is this a softRoCE issue or mlx5 issue? On Tue, Feb 6, 2018 at 12:08 PM, Olga Kornievskaia <aglo@xxxxxxxxx> wrote: > On Tue, Feb 6, 2018 at 10:57 AM, Moni Shoua <monis@xxxxxxxxxxxx> wrote: >> Hi Olga >> >>> I'm happy to test any patches. Also if you need an actual network >>> trace if that helps, I can email that to you. >>> >> Thanks for suggesting that. It might help a lot. > > I'm attaching a few of traces. > >>> What I find interesting is that there is an additional request for >>> 1021bytes at all as there 4096bytes in the response for the >>> 4093request. >>> >> If I had to guess the last response was for 1024 while the requester >> expects only 1021. >> Therefore the requester send a read request for the remaining 1021 >> bytes (as a part of the retry mechanism) but again, gets an answer for >> 1024 bytes. >> This repeats until the requester gives up (max retries exceeded) >>> Could the problem be that one side is adding padding bytes and the >>> other side is not expecting it? >> Padding is an option that I thought about although you said that for >> requests under 4089 bytes the test passes (what was different) > > I tried again: 4088 works. 4087, 4086 doesn't .. i skipped to 4084 > which works again. > >> >> thanks >> >> Moni -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
