Blob Blame History Raw
From: Jason Gunthorpe <jgg@mellanox.com>
Date: Tue, 13 Feb 2018 12:18:36 +0200
Subject: IB/uverbs: Hold the uobj write lock after allocate
Patch-mainline: v4.16-rc3
Git-commit: d9dc7a3500a58de9bf3861d1a96ffeab42624b4f
References: bsc#1046306 FATE#322942

This clarifies the design intention that time between allocate and
commit has the uobj exclusive to the caller. We already guarantee
this by delaying publishing the uobj pointer via idr_insert,
fd_install, list_add, etc.

Additionally holding the usecnt lock during this period provides
extra clarity and more protection against future mistakes.

Fixes: 3832125624b7 ("IB/core: Add support for idr types")
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
 drivers/infiniband/core/rdma_core.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

--- a/drivers/infiniband/core/rdma_core.c
+++ b/drivers/infiniband/core/rdma_core.c
@@ -141,7 +141,12 @@ static struct ib_uobject *alloc_uobj(str
 	 */
 	uobj->context = context;
 	uobj->type = type;
-	atomic_set(&uobj->usecnt, 0);
+	/*
+	 * Allocated objects start out as write locked to deny any other
+	 * syscalls from accessing them until they are committed. See
+	 * rdma_alloc_commit_uobject
+	 */
+	atomic_set(&uobj->usecnt, -1);
 	kref_init(&uobj->ref);
 
 	return uobj;
@@ -527,6 +532,10 @@ int rdma_alloc_commit_uobject(struct ib_
 		return ret;
 	}
 
+	/* matches atomic_set(-1) in alloc_uobj */
+	lockdep_check(uobj, true);
+	atomic_set(&uobj->usecnt, 0);
+
 	uobj->type->type_class->alloc_commit(uobj);
 	up_read(&uobj->context->cleanup_rwsem);