Jason Gunthorpe

can you send these two as a pull request?

Nope

Not from me, someone could probably tackle it though. Not sure how hard it would be.

I've never see that before, it suggests your kernel is too old or something is wrong with the ioctl execution.. What kernel version is that?

That kernel is certainly new enough. Is this a big endian power 9? I guess the ioctl must be malformed. Can you strace 'ibtool sminfo' ?

No idea. This is the line: ```c ioctl(3, _IOC(0, 0x1b, 0x03, 0x00), 0) = -1 ENOTTY (Inappropriate ioctl for device) ``` And that looks perfectly OK to me. The kernel...

Huh. PPC has a different definition for IOC_NONE for some reason. Does this fix it for you? ```diff diff --git a/rdma/umad.py b/rdma/umad.py index 76ff470714dae0..f9e2a9201778d2 100644 --- a/rdma/umad.py +++ b/rdma/umad.py @@...

It shouldn't be IOC_READ that would be '2'.. Ugh PPC also has a different _IOC_SIZEBITS. ```diff diff --git a/rdma/tools.py b/rdma/tools.py index 11dd79a6b6bb81..9b72220158a13a 100644 --- a/rdma/tools.py +++ b/rdma/tools.py @@ -8,7 +8,7...

So the fix for this is going to be to use one of the python IOCTL libraries out there, for instance this code https://github.com/olavmrk/python-ioctl

FDR10 is a non-standard Mellanox extension, I never looked into how to support it. If you have hardware that can support it, perhaps you could send a patch?