[PATCH 4.17 115/336] nvme-rdma: stop admin queue before freeing it

From: Greg Kroah-Hartman
Date: Wed Aug 01 2018 - 14:25:49 EST


4.17-stable review patch. If anyone has any objections, please let me know.

------------------

From: Jianchao Wang <jianchao.w.wang@xxxxxxxxxx>

[ Upstream commit 2e050f00a0f0e07467050cb4afae0234941e5bf3 ]

For any failure after nvme_rdma_start_queue in
nvme_rdma_configure_admin_queue, the admin queue will be freed with the
NVME_RDMA_Q_LIVE flag still set. Once nvme_rdma_stop_queue is invoked,
that will cause a use-after-free.
BUG: KASAN: use-after-free in rdma_disconnect+0x1f/0xe0 [rdma_cm]

To fix it, call nvme_rdma_stop_queue for all the failed cases after
nvme_rdma_start_queue.

Signed-off-by: Jianchao Wang <jianchao.w.wang@xxxxxxxxxx>
Suggested-by: Sagi Grimberg <sagi@xxxxxxxxxxx>
Reviewed-by: Max Gurtovoy <maxg@xxxxxxxxxxxx>
Signed-off-by: Christoph Hellwig <hch@xxxxxx>
Signed-off-by: Sasha Levin <alexander.levin@xxxxxxxxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
---
drivers/nvme/host/rdma.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)

--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -778,7 +778,7 @@ static int nvme_rdma_configure_admin_que
if (error) {
dev_err(ctrl->ctrl.device,
"prop_get NVME_REG_CAP failed\n");
- goto out_cleanup_queue;
+ goto out_stop_queue;
}

ctrl->ctrl.sqsize =
@@ -786,23 +786,25 @@ static int nvme_rdma_configure_admin_que

error = nvme_enable_ctrl(&ctrl->ctrl, ctrl->ctrl.cap);
if (error)
- goto out_cleanup_queue;
+ goto out_stop_queue;

ctrl->ctrl.max_hw_sectors =
(ctrl->max_fr_pages - 1) << (ilog2(SZ_4K) - 9);

error = nvme_init_identify(&ctrl->ctrl);
if (error)
- goto out_cleanup_queue;
+ goto out_stop_queue;

error = nvme_rdma_alloc_qe(ctrl->queues[0].device->dev,
&ctrl->async_event_sqe, sizeof(struct nvme_command),
DMA_TO_DEVICE);
if (error)
- goto out_cleanup_queue;
+ goto out_stop_queue;

return 0;

+out_stop_queue:
+ nvme_rdma_stop_queue(&ctrl->queues[0]);
out_cleanup_queue:
if (new)
blk_cleanup_queue(ctrl->ctrl.admin_q);