Re: [PATCH 0/6] blktests: char device tests with iouring-cmd fio

[Shinichiro Kawasaki <shinichiro.kawasaki@wdc.com>]

On Dec 21, 2022 / 02:34, Luis Chamberlain wrote:
> As io-uring cmd grows there's a desire to do a bit more funky things
> with it. Add basic support with fio and add a few simple tests to
> tests the NVMe conventional drives character device as well as the
> ZNS character device.
> 
> These tests are perhaps a bit *too* basic to merge, not sure, let
> me know. But I figured that this would provide example to let us
> grow this with more complex things later as folks add support for
> more features.

It is good to have new test cases to test new features and their new code paths.
I agree to have new test cases for the NVMe character device with io-uring.

Having said that, I'm not sure if we should have all of the five test cases in
the series. The test cases nvme/046, 047, 048 are similar. They do random read,
random write, or sequential write respectively. I'm not sure how the workload
difference expands the code coverage of the code paths in the NVMe driver. Same
for zbd/011 and 012. They are intended for ZNS devices, but I do not see ZNS
unique part in the NVMe character device code paths. Then the variation of the
test cases do not look useful to find bugs in the driver. As the first step, I
think single test case will be enough which does basic read and/or write to
exercises the NVMe character device and io-uring.

-- 
Shin'ichiro Kawasaki