Skip to content

Cleanup and fix the CMA state machine

Leon Romanovsky linux-rdma lkml 1 version Applied

Status from this message: Applied to for-next

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest8 patches 1 replyApplied

Patches in v1 Applied

#Subject and filesSizeRepliesApplied
1/8RDMA/cma: Fix locking for the RDMA_CM_CONNECT statedrivers/infiniband/core/cma.c+30 -140 replies
2/8RDMA/cma: Make the locking for automatic state transition more cleardrivers/infiniband/core/cma.c+45 -220 replies
3/8RDMA/cma: Fix locking for the RDMA_CM_LISTEN statedrivers/infiniband/core/cma.c+18 -180 replies
4/8RDMA/cma: Remove cma_comp()drivers/infiniband/core/cma.c+7 -200 replies
5/8RDMA/cma: Combine cma_ndev_work with cma_workdrivers/infiniband/core/cma.c+7 -310 replies
6/8RDMA/cma: Remove dead code for kernel rdmacm multicastdrivers/infiniband/core/cma.c+4 -150 replies
7/8RDMA/cma: Consolidate the destruction of a cma_multicast in one placedrivers/infiniband/core/cma.c+31 -320 replies
8/8RDMA/cma: Fix use after free race in roce multicast joindrivers/infiniband/core/cma.c+88 -1080 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [PATCH rdma-next 0/8] Cleanup and fix the CMA state machine

From: Leon Romanovsky <[email protected]>

From Jason:

The RDMA CMA continues to attract syzkaller bugs due to its somewhat loose operation of its FSM. Audit and scrub the whole thing to follow modern expectations.

Overall the design elements are broadly:

- The ULP entry points MUST NOT run in parallel with each other. The ULP is solely responsible for preventing this.

- If the ULP returns !0 from it's event callback it MUST guarentee that no other ULP threads are touching the cm_id or calling into any RDMA CM entry point.

- ULP entry points can sometimes run conurrently with handler callbacks, although it is tricky because there are many entry points that exist in the flow before the handler is registered.

- Some ULP entry points are called from the ULP event handler callback, under the handler_mutex. (however ucma never does this)

- state uses a weird double locking scheme, in most cases one should hold the handler_mutex. (It is somewhat unclear what exactly the spinlock is for)

- Reading the state without holding the spinlock should use READ_ONCE, even if the handler_mutex is held.

- There are certain states which are 'stable' under the handler_mutex, exit from that state requires also holding the handler_mutex. This explains why testing the test under only the handler_mutex makes sense.

Thanks

Jason Gunthorpe (8):
RDMA/cma: Fix locking for the RDMA_CM_CONNECT state RDMA/cma: Make the locking for automatic state transition more clear RDMA/cma: Fix locking for the RDMA_CM_LISTEN state RDMA/cma: Remove cma_comp()
RDMA/cma: Combine cma_ndev_work with cma_work
RDMA/cma: Remove dead code for kernel rdmacm multicast RDMA/cma: Consolidate the destruction of a cma_multicast in one place RDMA/cma: Fix use after free race in roce multicast join

 drivers/infiniband/core/cma.c | 466 ++++++++++++++++------------------
 1 file changed, 218 insertions(+), 248 deletions(-)

--
2.26.2