Currently, ib_create_cq uses cqe and comp_vecotr instead
of the extendible ib_cq_init_attr struct.

Earlier patches already changed the vendors to work with
ib_cq_init_attr. This patch changes the consumers too.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

iser connection termination process happens in 2 stages:
- isert_wait_conn:
  - resumes rdma disconnect
  - wait for session commands
  - wait for flush completions (post a marked wr to signal we are done)
  - wait for logout completion
  - queue work for connection cleanup (depends on disconnected/timewait
    events)
- isert_free_conn
  - last reference put on the connection

In case we are terminating during IOs, we might be posting send/recv
requests after we posted the last work request which might lead
to a use-after-free condition in isert_handle_wc.
After we posted the last wr in isert_wait_conn we are guaranteed that
no successful completions will follow (meaning no new work request posts
may happen) but other flush errors might still come. So before we
put the last reference on the connection, we repeat the process of
posting a marked work request (isert_wait4flush) in order to make sure all
pending completions were flushed.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NJenny Falkovich <jennyf@mellanox.com>
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

When receiving a new iser connect request we serialize
the pending requests by adding the newly created iser connection
to the np accept list and let the login thread process the connect
request one by one (np_accept_wait).

In case we received a disconnect request before the iser_conn
has begun processing (still linked in np_accept_list) we should
detach it from the list and clean it up and not have the login
thread process a stale connection. We do it only when the connection
state is not already terminating (initiator driven disconnect) as
this might lead us to access np_accept_mutex after the np was released
in live shutdown scenarios.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NJenny Falkovich <jennyf@mellanox.com>
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Since commit "2426bd45 target: Report correct response ..."
we might get a command with data_size that does not fit to
the number of allocated data sg elements. Given that we rely on
cmd t_data_nents which might be different than the data_size,
we sometimes receive local length error completion. The correct
approach would be to take the command data_size into account when
constructing the ib sg_list.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NJenny Falkovich <jennyf@mellanox.com>
Cc: stable@vger.kernel.org # 3.16+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

The first argument of these two functions is always identical
to se_cmd->se_sess. Hence remove the first argument.
Signed-off-by: NBart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: NSagi Grimberg <sagig@mellanox.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Cc: Andy Grover <agrover@redhat.com>
Cc: <qla2xxx-upstream@qlogic.com>
Cc: Felipe Balbi <balbi@ti.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

We don't assign pi_ctx to desc->pi_ctx until we're certain to succeed
in the function.  That means the cleanup path should use the local
pi_ctx variable, not desc->pi_ctx.

This was detected by Coverity (CID 1260062).
Signed-off-by: NRoland Dreier <roland@purestorage.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

These variables are always accessed via struct isert_conn so
no need to have a "conn_" prefix for them.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

iser target can handle as many connect request as
the fabric sends to it. This backlog should not set as
a back-pressure mechanism (which is not very useful).

isert does need a back-pressure mechanism, but it should
be added in isert by monitoring the number of pending
established connections (will be added in a later stage).
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

In iser_connect_release there is no chance that
the iser device is set to NULL, if this happens
we have a BUG. So use BUG_ON.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Not sure what it was used for, but there is
no real need for it now as I see it. Go ahead
and get rid of it.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Move login buffer alloc/free code to dedicated
routines and introduce isert_conn_init which
initializes the connection lists and locks.

Simplifies and cleans up the code a little bit.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

isert_device_find_by_ib_dev and isert_device_try_release
can have a better, more common name like isert_device_[get|put].
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Make receive/send completion handling routines symmetrical.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

No need to keep a local ib_dev as a device pointer.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Move the code for completion context handling to dedicated
routines. This simplifies the code and removes code duplication.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Simplify iser QP creation by splitting some unrelated
logic bulks to routines.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

No need to cast void pointers.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

No need for this assignment.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

unmap_list is unused.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

We have a switch default for this.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This is to favor the HCA cache hit rate using less MRs
and PDs. This commit partially reverts commit:
"eb6ab132 IB/isert: separate connection protection domains and dma MRs"

At the time I thought this would be needed.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Before we reach to connection established we may get an
error event. In this case the core won't teardown this
connection (never established it), so we take care of freeing
it ourselves.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Cc: <stable@vger.kernel.org> # v3.10+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This hang was a result of a missing command put when
a DIF error occurred during a rdma read (and we sent
an CHECK_CONDITION error without passing it to the
backend).
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Cc: <stable@vger.kernel.org> # v3.14+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

In case sendtargets response is larger than initiator MRDSL, we
send a partial sendtargets response (setting F=0, C=1, TTT!=0xffffffff),
accept a consecutive empty text message and send the rest of the payload.
In case we are done, we set F=1, C=0, TTT=0xffffffff.
We do that by storing the sendtargets response bytes done under
the session.

This patch also makes iscsit_find_cmd_from_itt public for isert.

(Re-add cmd->maxcmdsn_inc and clear in iscsit_build_text_rsp - nab)
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

The macro isert_dbg already ensures that __func__ is part of the
output, so there's no reason to duplicate the function name in the
format string itself.
Signed-off-by: NRasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

if text message dlength is 0, don't allocate a buffer for it, pass
NULL to iscsit_process_text_cmd.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Bound workqueues might be too restrictive since they allow
only a single core per session for processing completions.
WQ_UNBOUND will allow bouncing to another CPU if the running
CPU is currently busy. Luckily, our workqueues are NUMA aware
and will first try to bounce within the same NUMA socket.
My measurements with NULL backend devices show that there is
no (noticeable) additional latency as a result of the change.
I'd expect even to gain performance when working with fast
devices that also allocate MSIX interrupt vectors.

While we're at it, make it WQ_HIGHPRI since processing
completions is really a high priority for performance.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Reported-by: NMoussa Ba <moussaba@micron.com>
Signed-off-by: NMoussa Ba <moussaba@micron.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

We are sending text response and not reject.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

isert_debug_level should be static, hence no need
to initialize it.
Reported-by: NShachar Raindel <raindel@mellanox.com>
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

CC [M]  drivers/infiniband/ulp/isert/ib_isert.o
drivers/infiniband/ulp/isert/ib_isert.c: In function ‘isert_cq_comp_err’:
drivers/infiniband/ulp/isert/ib_isert.c:1979:42: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

- Fall-through in switch case instead in do_control_comp.
- Move rkey invalidation to a function.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

debug_level 1 (warn): Include warning messages.
debug_level 2 (info): Include relevant info for control plane.
debug_level 3 (debug): Include relevant info in the IO path.

Also, added/removed some logging messages.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Personal preference, easier control of the log level with
a single modparam which can be changed dynamically. Allows
better saparation of control and IO plains.

Replaced throughout ib_isert.c:
s/pr_debug/isert_dbg/g
s/pr_info/isert_info/g
s/pr_warn/isert_warn/g
s/pr_err/isert_err/g

Plus nit checkpatch warning change.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

We don't want to wait for conn_logout_comp from isert_comp_wq
context as this blocks further completions from being processed.
Instead we wait for it conditionally (if logout response was
actually posted) in wait_conn. This wait should normally happen
immediately as it occurs after we consumed all the completions
(including flush errors) and conn_logout_comp should have been
completed.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Might result in a deadlock where completion context waits for
session commands release where the later might need a final
completion for it.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

In order to reduce the contention on CQ locking (present
in some LLDDs) we poll in batches of 16 work completion items.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

In case the CQ is packed with completions, we can't just
hog the CPU forever. Poll until a sufficient budget (currently
hard-coded to 64k completions) and if budget is exhausted, bailout
and give a chance to other threads.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

In order to know that we consumed all the connection completions
we maintain atomic post_send_buf_count for each IO post send. But
we can know that if we post a "beacon" (zero length RECV work request)
after we move the QP into error state and the target does not serve
any new IO. When we consume it, we know we finished all the connection
completion and we can go ahead and destroy stuff.

In error completion handler we now just need to check for ISERT_BEACON_WRID
to arrive and then wait for session commands to cleanup and complete
conn_wait_comp_err.

We reserve another CQ and QP entries to fit the zero length post recv.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>