[bug report] most of blktests nvme/ failed on the latest linux tree

[Daniel Wagner]

On Thu, Jun 29, 2023 at 03:18:26PM +0300, Max Gurtovoy wrote:
> > Is there a specific reason why we want to read the /etc/nvme/hostnqn file at
> > all? Can't we just use a fixes hostid/hostnqn?
> 
> Yes, this will do the job but I guess the initial thought was to use the
> same hostnqn/hostid that one configured to the host also for testing.
> 
> I don't have a real preference between the two.

If we don't use the host hostnqn, we can make the connect attempt filtering more
relieable. This didn't matter so far for tcp/rdma/loop but with the fc enabling
it gets relevant. I think we could get away with using the host hostnqn but I
am not convienced we have all corner cases covered yet. We can add some more
test with fancy setups for this specific scenario later. I'd just like to have
blktests as reliable as possible.

Here some more background on this:

https://lore.kernel.org/linux-nvme/ygfgqglmntpqiopzq44aqegehnlroarteqjtmih7mulan4oukv@jmtupz2jnafv/