libibnetdisc: keep cache loadable after port timeouts

[ Upstream commit de83d1c208f9550ec1024a132ad982dcba9d8c06 ]

ibnetdiscover --cache writes each node's port references and each
port's remote-port reference from the in-memory fabric. If a PortInfo
query times out during discovery, the port can remain as a shell in
node->ports[] or as a remoteport without ever being inserted into the
fabric port table. Such a port has no cache port record, but references
to it were still written. ibnd_load_fabric() then rejected the cache with
"Cache invalid: cannot find port".

Make the cache writer emit only references to ports that are actually
serialized. A port is serialized only when it is reachable from its
owning node, and the same predicate is used for node port references,
remote-port references, and port records. References are matched against
a sorted set of serialized ports, avoiding repeated port-table walks on
large fabrics. Keep nodes with no ports array as zero-port node records
so fabric->from_node remains resolvable.

Also tolerate old incomplete caches when reading: missing node port keys
and missing remote ports are skipped instead of aborting the load. While
doing that, keep malformed caches from creating unsafe fabric state by
bounds-checking cached port numbers before indexing node->ports[] (an
out-of-bounds write) and by skipping unclaimed port records before they
are linked into the fabric and then freed (a use-after-free).

Fixes: 01b8045f28f6 ("add libibnetdisc caching to libibnetdiscover")
Signed-off-by: Jonathan Süssemilch Poulain <jpoulain@coreweave.com>
Signed-off-by: Nicolas Morey <nmorey@suse.com>
1 file changed
tree: 6928f31e786936fc9c99197c1bdbf23b9b582b9a
  1. ABI/
  2. buildlib/
  3. ccan/
  4. debian/
  5. Documentation/
  6. ibacm/
  7. infiniband-diags/
  8. iwpmd/
  9. kernel-boot/
  10. kernel-headers/
  11. libibmad/
  12. libibnetdisc/
  13. libibumad/
  14. libibverbs/
  15. librdmacm/
  16. providers/
  17. pyverbs/
  18. rdma-ndd/
  19. redhat/
  20. srp_daemon/
  21. suse/
  22. tests/
  23. util/
  24. .clang-format
  25. .gitignore
  26. .mailmap
  27. build.sh
  28. CMakeLists.txt
  29. COPYING.BSD_FB
  30. COPYING.BSD_MIT
  31. COPYING.GPL2
  32. COPYING.md
  33. MAINTAINERS
  34. README.md
README.md

Build Status

RDMA Core Userspace Libraries and Daemons

This is the userspace components for the Linux Kernel's drivers/infiniband subsystem. Specifically this contains the userspace libraries for the following device nodes:

  • /dev/infiniband/uverbsX (libibverbs)
  • /dev/infiniband/rdma_cm (librdmacm)
  • /dev/infiniband/umadX (libibumad)

The userspace component of the libibverbs RDMA kernel drivers are included under the providers/ directory. Support for the following Kernel RDMA drivers is included:

  • bnxt_re.ko
  • efa.ko
  • erdma.ko
  • iw_cxgb4.ko
  • hfi1.ko
  • hns-roce-hw-v2.ko
  • irdma.ko
  • ib_qib.ko
  • mana_ib.ko
  • mlx4_ib.ko
  • mlx5_ib.ko
  • ib_mthca.ko
  • ocrdma.ko
  • qedr.ko
  • rdma_rxe.ko
  • siw.ko
  • vmw_pvrdma.ko

Additional service daemons are provided for:

  • srp_daemon (ib_srp.ko)
  • iwpmd (for iwarp kernel providers)
  • ibacm (for InfiniBand communication management assistant)

Building

This project uses a cmake based build system. Quick start:

$ bash build.sh

build/bin will contain the sample programs and build/lib will contain the shared libraries. The build is configured to run all the programs ‘in-place’ and cannot be installed.

Debian Derived

$ apt-get install build-essential cmake gcc libudev-dev libnl-3-dev libnl-route-3-dev ninja-build pkg-config valgrind python3-dev cython3 python3-docutils pandoc

Supported releases:

  • Debian 9 (stretch) or newer
  • Ubuntu 16.04 LTS (xenial) or newer

Fedora, CentOS 8

$ dnf builddep redhat/rdma-core.spec

NOTE: Fedora Core uses the name ‘ninja-build’ for the ‘ninja’ command.

openSUSE

$ zypper install cmake gcc libnl3-devel libudev-devel ninja pkg-config valgrind-devel python3-devel python3-Cython python3-docutils pandoc

Building on CentOS 7, Amazon Linux 2

Install required packages:

$ yum install cmake gcc libnl3-devel libudev-devel make pkgconfig valgrind-devel

Developers on CentOS 7 or Amazon Linux 2 are suggested to install more modern tooling for the best experience.

CentOS 7:

$ yum install epel-release
$ yum install cmake3 ninja-build pandoc

Amazon Linux 2:

$ amazon-linux-extras install epel
$ yum install cmake3 ninja-build pandoc

NOTE: EPEL uses the name ‘ninja-build’ for the ‘ninja’ command, and ‘cmake3’ for the ‘cmake’ command.

Usage

To set up software RDMA on an existing interface with either of the available drivers, use the following commands, substituting <DRIVER> with the name of the driver of your choice (rdma_rxe or siw) and <TYPE> with the type corresponding to the driver (rxe or siw).

# modprobe <DRIVER>
# rdma link add <NAME> type <TYPE> netdev <DEVICE>

Please note that you need version of iproute2 recent enough is required for the command above to work.

You can use either ibv_devices or rdma link to verify that the device was successfully added.

Reporting bugs

Bugs should be reported to the linux-rdma@vger.kernel.org mailing list In your bug report, please include:

  • Information about your system:

    • Linux distribution and version
    • Linux kernel and version
    • InfiniBand hardware and firmware version
    • ... any other relevant information
  • How to reproduce the bug.

  • If the bug is a crash, the exact output printed out when the crash occurred, including any kernel messages produced.

Submitting patches

See Contributing to rdma-core.

Stable branches

Stable versions are released regularly with backported fixes (see Documentation/stable.md) The current minimum version still maintained is ‘v33.X’