NVMe: Fix bug in error handling
Matthew Wilcox [Fri, 29 Apr 2011 06:09:09 +0000 (23:09 -0700)]
When an I/O completed with an error, we would call bio_endio twice
(once with -EIO and once with 0).  Found by inspection.

Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>

drivers/block/nvme.c

index 57f2b33..205405e 100644 (file)
@@ -310,9 +310,9 @@ static void bio_completion(struct nvme_queue *nvmeq, void *ctx,
        dma_unmap_sg(nvmeq->q_dmadev, nbio->sg, nbio->nents,
                        bio_data_dir(bio) ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
        free_nbio(nvmeq, nbio);
-       if (status)
+       if (status) {
                bio_endio(bio, -EIO);
-       if (bio->bi_vcnt > bio->bi_idx) {
+       } else if (bio->bi_vcnt > bio->bi_idx) {
                bio_list_add(&nvmeq->sq_cong, bio);
                wake_up_process(nvme_thread);
        } else {