[openib-general] [PATCH] Fix freed mem deref race in cma_process_remove/cma_req_handler

[Krishna Kumar]

The race is as follows :

A process : cma_process_remove() calls cma_remove_id_dev(),
	    which sets id state to CMA_DEVICE_REMOVAL and
	    calls wait_event(dev_remove).

B process : cma_req_handler() had incremented dev_remove,
	    and calls cma_acquire_ib_dev() and on failure
	    calls cma_release_remove(), which does a
	    wake_up of cma_process_remove(). Then
	    cma_req_handler() calls rdma_destroy_id();

A Process : cma_remove_id_dev() gets woken and checks the
	    state of id, and since it is still (wrongly)
	    CMA_DEVICE_REMOVAL, it calls notify_user(id)
	    and if that fails, the caller - cma_process_remove()
	    calls rdma_destroy_id(id). Two processes can
	    call rdma_destroy_id(), resulting in one
	    de-referencing kfreed id_priv.

Fix is for process B to set CMA_DESTROYING in cma_req_handler()
so that process A will return instead of doing a rdma_destroy_id().

Signed-off-by: Krishna Kumar <krkumar2 at in.ibm.com>

diff -ruNp org/core/cma.c new/core/cma.c
--- org/core/cma.c	2006-09-14 15:41:01.000000000 +0530
+++ new/core/cma.c	2006-09-18 11:52:52.000000000 +0530
@@ -1023,6 +1023,7 @@ static int cma_req_handler(struct ib_cm_
 	mutex_unlock(&lock);
 	if (ret) {
 		ret = -ENODEV;
+		cma_exch(conn_id, CMA_DESTROYING);
 		cma_release_remove(conn_id);
 		rdma_destroy_id(&conn_id->id);
 		goto out;