linux-rdma / opensm

Other
66 stars 36 forks source link

OpenSM: Add information about limitations #22

Open clameter opened 3 years ago

clameter commented 3 years ago

Please add some information on the limitations of OpenSM. From what I know these are

  1. No SR-IOV support
  2. Multicast sweeps may cause multicast micro-loops which can cause the SM to fail.
  3. No Multicast support for ConnectX5 and 6.

These are issues fixed in the Mellanox OFED OpenSM.

dcontiveros commented 1 year ago

I was attempting to setup SRIOV with Connect-X 4 card and ran into an issue where the physical link is up, but the state reports as down. From what I have read online, virtualization needs to be enabled, but I cannot see where to do that in this version of OpenSM.

I am attempting to avoid all Nvidia drivers. I see there is this documentation:

https://github.com/linux-rdma/opensm/blob/master/doc/opensm-sriov.txt

However, I cannot seem to find a working example of how to use these two configuration options to get SRIOV working.

I have virtual functions in lspci, am able to unbind/bind to vfio. I am now just lacking the OpenSM functionality.

vladko1974 commented 1 year ago

Hi,

SRIOV with ConnectX4+ devices is supported only with NVIDIA SM that can be downloaded from NVIDIA Web. https://network.nvidia.com/products/adapter-software/infiniband-management-and-monitoring-tools/

Regards, Vladimir

From: David Ontiveros @.> Sent: Monday, 4 September 2023 0:34 To: linux-rdma/opensm @.> Cc: Subscribed @.***> Subject: Re: [linux-rdma/opensm] OpenSM: Add information about limitations (#22)

I was attempting to setup SRIOV with Connect-X 4 card and ran into an issue where the physical link is up, but the state reports as down. From what I have read online, virtualization needs to be enabled, but I cannot see where to do that in this version of OpenSM.

I am attempting to avoid all Nvidia drivers. I see there is this documentation:

https://github.com/linux-rdma/opensm/blob/master/doc/opensm-sriov.txt

However, I cannot seem to find a working example of how to use these two configuration options to get SRIOV working.

I have virtual functions in lspci, am able to unbind/bind to vfio. I am now just lacking the OpenSM functionality.

— Reply to this email directly, view it on GitHubhttps://github.com/linux-rdma/opensm/issues/22#issuecomment-1704410211, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AD6PRHGHO2CVUKDT5KGXNPDXYTZV7ANCNFSM4THTI5JA. You are receiving this because you are subscribed to this thread.Message ID: @.**@.>>

dcontiveros commented 1 year ago

Thanks Vladimir. Would it be possible to add SRIOV support as a feature request ? Is it even possible?

vladko1974 commented 1 year ago

We are recommending to use NVDIA SM rpm from Web for customers that can’t use NVIDIA SM from MOFED.

From: David Ontiveros @.> Sent: Monday, 4 September 2023 15:08 To: linux-rdma/opensm @.> Cc: Vladimir Koushnir @.>; Comment @.> Subject: Re: [linux-rdma/opensm] OpenSM: Add information about limitations (#22)

Thanks Vladimir. Would it be possible to add SRIOV support as a feature request ? Is it even possible?

— Reply to this email directly, view it on GitHubhttps://github.com/linux-rdma/opensm/issues/22#issuecomment-1705156304, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AD6PRHCE3UQRHITDP633FXDXYXAB3ANCNFSM4THTI5JA. You are receiving this because you commented.Message ID: @.**@.>>