diff options
| author | Anton Ivanov <anton.ivanov@cambridgegreys.com> | 2018-11-08 16:07:23 +0300 |
|---|---|---|
| committer | Jens Axboe <axboe@kernel.dk> | 2018-11-08 16:16:09 +0300 |
| commit | 0033dfd92a5646a78025e86f8df4d5b18181ba2c (patch) | |
| tree | aedd22c50839066987f771bd1ad09cb9c7e8f10e /arch | |
| parent | f3587d76da05f68098ddb1cb3c98cc6a9e8a402c (diff) | |
| download | linux-0033dfd92a5646a78025e86f8df4d5b18181ba2c.tar.xz | |
ubd: fix missing initialization of io_req
The SYNC path doesn't initialize io_req->error, which can cause
random errors. Before the conversion to blk-mq, we always
completed requests with BLK_STS_OK status, but now we actually
look at the error field and this issue becomes apparent.
Signed-off-by: Anton Ivanov <anton.ivanov@cambridgegreys.com>
[axboe: fixed up commit message to explain what is actually going on]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Diffstat (limited to 'arch')
| -rw-r--r-- | arch/um/drivers/ubd_kern.c | 3 |
1 files changed, 1 insertions, 2 deletions
diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c index 08831f5d83db..28c40624bcb6 100644 --- a/arch/um/drivers/ubd_kern.c +++ b/arch/um/drivers/ubd_kern.c @@ -1305,6 +1305,7 @@ static int ubd_queue_one_vec(struct blk_mq_hw_ctx *hctx, struct request *req, io_req->fds[0] = dev->cow.fd; else io_req->fds[0] = dev->fd; + io_req->error = 0; if (req_op(req) == REQ_OP_FLUSH) { io_req->op = UBD_FLUSH; @@ -1313,9 +1314,7 @@ static int ubd_queue_one_vec(struct blk_mq_hw_ctx *hctx, struct request *req, io_req->cow_offset = -1; io_req->offset = off; io_req->length = bvec->bv_len; - io_req->error = 0; io_req->sector_mask = 0; - io_req->op = rq_data_dir(req) == READ ? UBD_READ : UBD_WRITE; io_req->offsets[0] = 0; io_req->offsets[1] = dev->cow.data_offset; |