nvme-pci: fix memory leak on probe failure

The nvme driver specific structures need to be initialized prior to
enabling the generic controller so we can unwind on failure with out
using the reference counting callbacks so that 'probe' and 'remove'
can be symmetric.

The newly added iod_mempool is the only resource that was being
allocated out of order, and a failure there would leak the generic
controller memory. This patch just moves that allocation above the
controller initialization.

Fixes: 943e942 ("nvme-pci: limit max IO size and segments to avoid high order allocations")
Reported-by: Weiping Zhang <zwp10758@gmail.com>
Signed-off-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

drivers/nvme/host/pci.c

@@ -2556,11 +2556,6 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 
 	quirks |= check_vendor_combination_bug(pdev);
 
-	result = nvme_init_ctrl(&dev->ctrl, &pdev->dev, &nvme_pci_ctrl_ops,
-			quirks);
-	if (result)
-		goto release_pools;
-
 	/*
 	 * Double check that our mempool alloc size will cover the biggest
 	 * command we support.
@@ -2578,13 +2573,20 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 		goto release_pools;
 	}
 
+	result = nvme_init_ctrl(&dev->ctrl, &pdev->dev, &nvme_pci_ctrl_ops,
+			quirks);
+	if (result)
+		goto release_mempool;
+
 	dev_info(dev->ctrl.device, "pci function %s\n", dev_name(&pdev->dev));
 
 	nvme_get_ctrl(&dev->ctrl);
 	async_schedule(nvme_async_probe, dev);
 
 	return 0;
 
+ release_mempool:
+	mempool_destroy(dev->iod_mempool);
  release_pools:
 	nvme_release_prp_pools(dev);
  unmap: