On 5/12/21 1:33 PM, Lee Duncan wrote:

I can't remember if I replied to this already (what memory?), so just in
case:

You had 2 review comments I'm handling:

1. I was setting ep multiple times in one function, so I'm cleaning that up.

2. I re-arranged the code in iscsi_if_transport_conn so it's clear why I
separated the 2 types of functions (ones that need the ep_mutex and a common
check for if a conn stop is running and ones that don't).

There is also a new one:

3. While re-testing for #2, I found another issue. Most of the bugs I was
handling were what happens if the in-kernel conn stop runs late due to iscsid
being overloaded. Because we do the conn stop in the kernel we can of course
now run a lot a faster. So we can end up cleaning up from the issue, but then
the kernel fires another event for the same root issue. Because we already
cleaned up from the in-kernel cleanup, we think it's a new issue instead
of just N events for the same root issue.

I'm testing a fix for that still.