nvmet: avoid potential UAF in nvmet_req_complete()

[ Upstream commit 6173a77b7e9d3e202bdb9897b23f2a8afe7bf286 ]

An nvme target ->queue_response() operation implementation may free the
request passed as argument. Such implementation potentially could result
in a use after free of the request pointer when percpu_ref_put() is
called in nvmet_req_complete().

Avoid such problem by using a local variable to save the sq pointer
before calling __nvmet_req_complete(), thus avoiding dereferencing the
req pointer after that function call.

Fixes: a07b4970f464 ("nvmet: add a generic NVMe target")
Signed-off-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
index 683b75a..3235baf 100644 (file)
--- a/drivers/nvme/target/core.c
+++ b/drivers/nvme/target/core.c
@@ -755,8 +755,10 @@ static void __nvmet_req_complete(struct nvmet_req *req, u16 status)
 
 void nvmet_req_complete(struct nvmet_req *req, u16 status)
 {
+       struct nvmet_sq *sq = req->sq;
+
        __nvmet_req_complete(req, status);
-       percpu_ref_put(&req->sq->ref);
+       percpu_ref_put(&sq->ref);
 }
 EXPORT_SYMBOL_GPL(nvmet_req_complete);