From: Leon Romanovsky <leon@kernel.org>
To: Jason Gunthorpe <jgg@nvidia.com>
Cc: Leon Romanovsky <leonro@nvidia.com>,
Bart Van Assche <bvanassche@acm.org>,
linux-rdma@vger.kernel.org, Mark Bloch <mbloch@nvidia.com>,
Mark Zhang <markzhang@nvidia.com>
Subject: [PATCH rdma-next 0/4] Support multiple path records
Date: Thu, 8 Sep 2022 13:08:59 +0300 [thread overview]
Message-ID: <cover.1662631201.git.leonro@nvidia.com> (raw)
From: Leon Romanovsky <leonro@nvidia.com>
From Mark:
These patches allow IB core to receive multiple path records from
user-space rdma netlink service.
Currently only one GMP PR is supported when doing a PR query. This means
in a fabric with multiple subnets, when a packet goes out of the sender,
it’s assigned with a LID to a router as its destination LID. The current
solution selects a specific router per destination, means that there is
no advanced routing from the sender towards the network.
This patchset supports to receive an inbound PR and an outbound PR along
with the GMP PR. The LIDs in these 3 PRs can be used in this way:
1. GMP PR: used as the standard local/remote LIDs;
2. DLID of outbound PR: Used as the "dlid" field for outbound traffic;
3. DLID of inbound PR: Used as the "dlid" field for outbound traffic in
responder side.
The inboundPR.dlid is passed to responder with the "primary_LID" filed
in the ConnectRequest message.
With this, the user-space rdma netlink service can set special DLIDs
in inbound/outbound PRs, to select specific routers for datapath
between the 2 nodes.
The following cases were tested:
- New kernel with new netlink service that supports multiple path
records;
- New kernel with old netlink service;
- Old kernel with new netlink service;
- Client side new kernel with new netlink service, server side with
old kernel.
Thanks.
Mark Zhang (4):
RDMA/core: Rename rdma_route.num_paths field to num_pri_alt_paths
RDMA/cma: Multiple path records support with netlink channel
RDMA/cm: Use SLID in the work completion as the DLID in responder side
RDMA/cm: Use DLID from inbound/outbound PathRecords as the datapath
DLID
drivers/infiniband/core/cm.c | 39 +++-
drivers/infiniband/core/cma.c | 88 ++++++--
drivers/infiniband/core/sa_query.c | 235 +++++++++++++++-------
drivers/infiniband/core/ucma.c | 10 +-
drivers/infiniband/ulp/ipoib/ipoib_main.c | 2 +-
drivers/infiniband/ulp/srp/ib_srp.c | 2 +-
include/rdma/ib_cm.h | 2 +
include/rdma/ib_sa.h | 3 +-
include/rdma/rdma_cm.h | 13 +-
9 files changed, 284 insertions(+), 110 deletions(-)
--
2.37.2
next reply other threads:[~2022-09-08 10:09 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-08 10:08 Leon Romanovsky [this message]
2022-09-08 10:09 ` [PATCH rdma-next 1/4] RDMA/core: Rename rdma_route.num_paths field to num_pri_alt_paths Leon Romanovsky
2022-09-08 10:09 ` [PATCH rdma-next 2/4] RDMA/cma: Multiple path records support with netlink channel Leon Romanovsky
2022-09-22 13:58 ` Jason Gunthorpe
2022-09-23 1:40 ` Mark Zhang
2022-09-23 13:13 ` Jason Gunthorpe
2022-09-23 14:24 ` Mark Zhang
2022-09-23 14:50 ` Mark Zhang
2022-09-23 18:19 ` Jason Gunthorpe
2022-09-08 10:09 ` [PATCH rdma-next 3/4] RDMA/cm: Use SLID in the work completion as the DLID in responder side Leon Romanovsky
2022-09-08 10:09 ` [PATCH rdma-next 4/4] RDMA/cm: Use DLID from inbound/outbound PathRecords as the datapath DLID Leon Romanovsky
2022-09-22 9:36 ` [PATCH rdma-next 0/4] Support multiple path records Leon Romanovsky
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=cover.1662631201.git.leonro@nvidia.com \
--to=leon@kernel.org \
--cc=bvanassche@acm.org \
--cc=jgg@nvidia.com \
--cc=leonro@nvidia.com \
--cc=linux-rdma@vger.kernel.org \
--cc=markzhang@nvidia.com \
--cc=mbloch@nvidia.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).