iscsi-target: Fix immediate queue starvation regression with DATAIN

This patch addresses a v3.5+ regression in iscsi-target where TX thread
process context -> handle_response_queue() execution is allowed to run
unbounded while servicing constant outgoing flow of ISTATE_SEND_DATAIN
response state.

This ends up preventing memory release of StatSN acknowledged commands
in a timely manner when under heavy large block streaming DATAIN
workloads.

The regression bug was initially introduced with:

commit 6f3c0e69a9
Author: Andy Grover <agrover@redhat.com>
Date:   Tue Apr 3 15:51:09 2012 -0700

    target/iscsi: Refactor target_tx_thread immediate+response queue loops

Go ahead and follow original iscsi_target_tx_thread() logic and check
to break for immediate queue processing after each DataIN Sequence and/or
Response PDU has been sent.

Reported-by: Benjamin ESTRABAUD <be@mpstor.com>
Cc: Andy Grover <agrover@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
This commit is contained in:
Nicholas Bellinger 2013-02-27 17:53:52 -08:00
parent 972b29c8f8
commit fd3a9025c0
1 changed files with 9 additions and 2 deletions

View File

@ -3583,6 +3583,10 @@ check_rsp_state:
spin_lock_bh(&cmd->istate_lock);
cmd->i_state = ISTATE_SENT_STATUS;
spin_unlock_bh(&cmd->istate_lock);
if (atomic_read(&conn->check_immediate_queue))
return 1;
continue;
} else if (ret == 2) {
/* Still must send status,
@ -3672,7 +3676,7 @@ check_rsp_state:
}
if (atomic_read(&conn->check_immediate_queue))
break;
return 1;
}
return 0;
@ -3716,12 +3720,15 @@ restart:
signal_pending(current))
goto transport_err;
get_immediate:
ret = handle_immediate_queue(conn);
if (ret < 0)
goto transport_err;
ret = handle_response_queue(conn);
if (ret == -EAGAIN)
if (ret == 1)
goto get_immediate;
else if (ret == -EAGAIN)
goto restart;
else if (ret < 0)
goto transport_err;