Please pull RDMA subsystem changes

Hi Linus,

Some rc patches for this cycle, there are a couple of compilation fixes here that need to get cleared out.

Thanks,
Jason

The following changes since commit 6de23f81a5e08be8fbf5e8d7e9febc72a5b5f27f:

Linux 7.0-rc1 (2026-02-22 13:18:59 -0800)

are available in the Git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma.git tags/for-linus

for you to fetch changes up to 7c2889af823340d1d410939b9d547bf184d5fa54:

RDMA/uverbs: Import DMA-BUF module in uverbs_std_types_dmabuf file (2026-02-26 04:58:24 -0500)

----------------------------------------------------------------
RDMA v7.0 first rc

Seems bigger than usual, a number of things were posted near/during the merg window:

- Fix some compilation regressions related to the new DMABUF code

- Close a race with ib_register_device() vs netdev events that causes GID table corruption

- Compilation warnings with some compilers in bng_re

- Correct error unwind in bng_re and the umem pinned dmabuf

- Avoid NULL pointer crash in ionic during query_port()

- Check the size for uAPI validation checks in EFA

- Several system call stack leaks in drivers found with AI

- Fix the new restricted_node_type so it works with wildcard listens too

----------------------------------------------------------------
Arnd Bergmann (1):
RDMA/uverbs: select CONFIG_DMA_SHARED_BUFFER

Jacob Moroni (1):
RDMA/umem: Fix double dma_buf_unpin in failure path

Jason Gunthorpe (4):
RDMA/efa: Fix typo in efa_alloc_mr()
IB/mthca: Add missed mthca_unmap_user_db() for mthca_create_srq() RDMA/irdma: Fix kernel stack leak in irdma_create_user_ah() RDMA/ionic: Fix kernel stack leak in ionic_create_cq()

Jiri Pirko (1):
RDMA/core: Fix stale RoCE GIDs during netdev events at registration

Kamal Heib (1):
RDMA/ionic: Fix potential NULL pointer dereference in ionic_query_port

Leon Romanovsky (1):
RDMA/uverbs: Import DMA-BUF module in uverbs_std_types_dmabuf file

Siva Reddy Kallam (2):
RDMA/bng_re: Remove unnessary validity checks RDMA/bng_re: Unwind bng_re_dev_init properly

Stefan Metzmacher (1):

      RDMA/core: Check id_priv->restricted_node_type in cma_listen_on_dev()

 drivers/infiniband/Kconfig                        |  1 +
 drivers/infiniband/core/cache.c                   | 13 ++++++
 drivers/infiniband/core/cma.c                     |  6 ++-
 drivers/infiniband/core/core_priv.h               |  3 ++
 drivers/infiniband/core/device.c                  | 34 +++++++++++++-
 drivers/infiniband/core/umem_dmabuf.c             |  4 +-
 drivers/infiniband/core/uverbs_std_types_dmabuf.c |  2 +
 drivers/infiniband/hw/bng_re/bng_dev.c            | 56 ++++++++---------------
 drivers/infiniband/hw/efa/efa_verbs.c             |  2 +-
 drivers/infiniband/hw/ionic/ionic_controlpath.c   |  2 +-
 drivers/infiniband/hw/ionic/ionic_ibdev.c         |  2 +
 drivers/infiniband/hw/irdma/verbs.c               |  2 +-
 drivers/infiniband/hw/mthca/mthca_provider.c      |  5 +-
 include/rdma/rdma_cm.h                            |  2 +-
 14 files changed, 86 insertions(+), 48 deletions(-)