Re: [PATCH] iscsi-target: Fix initial login PDU asynchronous socket close OOPs
From: Nicholas A. Bellinger
Date: Wed May 31 2017 - 00:58:54 EST
Hey MNC,
On Fri, 2017-05-26 at 22:14 -0500, Mike Christie wrote:
> Thanks for the patch.
>
Btw, after running DATERA's internal longevity and scale tests across
~20 racks on v4.1.y with this patch over the long weekend, there haven't
been any additional regressions.
> On 05/26/2017 12:32 AM, Nicholas A. Bellinger wrote:
> >
> > - state = iscsi_target_sk_state_check(sk);
> > - write_unlock_bh(&sk->sk_callback_lock);
> > -
> > - pr_debug("iscsi_target_sk_state_change: state: %d\n", state);
> > + orig_state_change(sk);
> >
> > - if (!state) {
> > - pr_debug("iscsi_target_sk_state_change got failed state\n");
> > - schedule_delayed_work(&conn->login_cleanup_work, 0);
>
> I think login_cleanup_work is no longer used so you can also remove it
> and its code.
Yep, since this needs to goto stable, I left that part out for now..
Will take care of that post -rc4.
>
> The patch fixes the crash for me. However, is there a possible
> regression where if the initiator attempts new relogins we could run out
> of memory? With the old code, we would free the login attempts resources
> at this time, but with the new code the initiator will send more login
> attempts and so we just keep allocating more memory for each attempt
> until we run out or the login is finally able to complete.
AFAICT, no. For the two cases in question:
- Initial login request PDU processing done within iscsi_np kthread
context in iscsi_target_start_negotiation(), and
- subsequent login request PDU processing done by delayed work-queue
kthread context in iscsi_target_do_login_rx()
this patch doesn't change how aggressively connection cleanup happens
for failed login attempts in the face of new connection login attempts
for either case.
For the first case when iscsi_np process context invokes
iscsi_target_start_negotiation() -> iscsi_target_do_login() ->
iscsi_check_for_session_reinstatement() to wait for backend I/O to
complete, it still blocks other new connections from being accepted on
the specific iscsi_np process context.
This patch doesn't change this behavior.
What it does change is when the host closes the connection and
iscsi_target_sk_state_change() gets invoked, iscsi_np process context
waits for iscsi_check_for_session_reinstatement() to complete before
releasing the connection resources.
However since iscsi_np process context is blocked, new connections won't
be accepted until the new connection forcing session reinstatement
finishes waiting for outstanding backend I/O to complete.
For the second case of subsequent non initial login request PDUs handled
within delayed work-queue process context, AFAICT this patch doesn't
change the original behavior either..
Namely when iscsi_target_do_login_rx() is active and host closes the
connection causing iscsi_target_sk_state_change() to be invoked, it
still checks for LOGIN_FLAGS_READ_ACTIVE and doesn't queue shutdown to
occur.
As per the original logic preceding this change, it continues to wait
for iscsi_target_do_login_rx() to complete in delayed work-queue
context, unless sock_recvmsg() returns a failure (which it should once
TCP_CLOSE occurs) or times out via iscsi_target_login_timeout(). Once
the failure is detected from iscsi_target_do_login_rx(), the remaining
connection resources are related from there.
That said, was there another case you had in mind..?