This reverts commit 70a640d0.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The “affinity hint” mechanism is used by the user space
daemon, irqbalancer, to indicate a preferred CPU mask for irqs.
Irqbalancer can use this hint to balance the irqs between the
cpus indicated by the mask.

We wish the HCA to preferentially map the IRQs it uses to numa cores
close to it.  To accomplish this, we use cpumask_set_cpu_local_first(), that
sets the affinity hint according the following policy:
First it maps IRQs to “close” numa cores.  If these are exhausted, the
remaining IRQs are mapped to “far” numa cores.
Signed-off-by: NYuval Atias <yuvala@mellanox.com>
Signed-off-by: NAmir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When we receive a netdev event indicating a netdev change and/or
a netdev address change, we must change the MAC index used by the
proxy QP1 (in the QP context), otherwise RoCE CM packets sent by the
VF will not carry the same source MAC address as the non-CM packets.

We use the UPDATE_QP command to perform this change.

In order to avoid modifying a QP context based on netdev event,
while the driver attempts to destroy this QP (e.g either the mlx4_ib
or ib_mad modules are unloaded), we use mutex locking in both flows.

Since the relevant mlx4 proxy GSI QP is created indirectly by the
mad module when they create their GSI QP, the mlx4 didn't need to
keep track on that QP prior to this change.

Now, when QP modifications are needed to this QP from within the
driver, we added refernece to it.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the target is in stop stage, iSER transport initiates RDMA disconnects.
The iSER initiator may wish to establish a new connection over the
still existing network portal. In this case iSER transport should not
accept and resume new RDMA connections. In order to learn that, iscsi_np
is added with enabled flag so the iSER transport can check when deciding
weather to accept and resume a new connection request.

The iscsi_np is enabled after successful transport setup, and disabled
before iscsi_np login threads are cleaned up.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

RDMA CM and iSCSI target flows are asynchronous and completely
uncorrelated. Relying on the fact that iscsi_accept_np will be called
after CM connection request event and will wait for it is a mistake.

When attempting to login to a few targets this flow is racy and
unpredictable, but for parallel login to dozens of targets will
race and hang every time.

The correct synchronizing mechanism in this case is pending on
a semaphore rather than a wait_for_event. We keep the pending
interruptible for iscsi_np cleanup stage.

(Squash patch to remove dead code into parent - nab)
Reported-by: NSlava Shwartsman <valyushash@gmail.com>
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Should be adding list_add_tail($new, $head) and not
the other way around.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

net: get rid of SET_ETHTOOL_OPS

Dave Miller mentioned he'd like to see SET_ETHTOOL_OPS gone.
This does that.

Mostly done via coccinelle script:
@@
struct ethtool_ops *ops;
struct net_device *dev;
@@
-       SET_ETHTOOL_OPS(dev, ops);
+       dev->ethtool_ops = ops;

Compile tested only, but I'd seriously wonder if this broke anything.
Suggested-by: NDave Miller <davem@davemloft.net>
Signed-off-by: NWilfried Klaebe <w-lkml@lebenslange-mailadresse.de>
Acked-by: NFelipe Balbi <balbi@ti.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NHariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

The whole db drop avoidance stuff is for T4 only.  So we cannot allow
that to be enabled for T5 devices.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

This is required to work around a T5 HW issue.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

In cases where the cm calls c4iw_modify_rc_qp() with the endpoint
mutex held, they must be called with internal == 1.  rx_data() and
process_mpa_reply() are not doing this.  This causes a deadlock
because c4iw_modify_rc_qp() might call c4iw_ep_disconnect() in some
!internal cases, and c4iw_ep_disconnect() acquires the endpoint mutex.
The design was intended to only do the disconnect for !internal calls.

Change rx_data(), FPDU_MODE case, to call c4iw_modify_rc_qp() with
internal == 1, and then disconnect only after releasing the mutex.

Change process_mpa_reply() to call c4iw_modify_rc_qp(TERMINATE) with
internal == 1 and set a new attr flag telling it to send a TERMINATE
message.  Previously this was implied by !internal.

Change process_mpa_reply() to return whether the caller should
disconnect after releasing the endpoint mutex.  Now rx_data() will do
the disconnect in the cases where process_mpa_reply() wants to
disconnect after the TERMINATE is sent.

Change c4iw_modify_rc_qp() RTS->TERM to only disconnect if !internal,
and to send a TERMINATE message if attrs->send_term is 1.

Change abort_connection() to not aquire the ep mutex for setting the
state, and make all calls to abort_connection() do so with the mutex
held.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

The code was incorrectly using sg_dma_address() and
sg_dma_len() instead of ib_sg_dma_address() and
ib_sg_dma_len().

This prevents srpt from functioning with the
Intel HCA and indeed will corrupt memory
badly.

Cc: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: NDennis Dalessandro <dennis.dalessandro@intel.com>
Tested-by: NVinod Kumar <vinod.kumar@intel.com>
Signed-off-by: NMike Marciniszyn <mike.marciniszyn@intel.com>
Cc: stable@vger.kernel.org # 3.3+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Need to get the endpoint reference before calling rdma_fini(), which
might fail causing us to not get the reference.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Signed-off-by: NHariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

The max depth of a fastreg mr depends on whether the device supports
DSGL or not.  So compute it dynamically based on the device support
and the module use_dsgl option.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

There is a race when moving a QP from RTS->CLOSING where a SQ work
request could be posted after the FW receives the RDMA_RI/FINI WR.
The SQ work request will never get processed, and should be completed
with FLUSHED status.  Function c4iw_flush_sq(), however was dropping
the oldest SQ work request when in CLOSING or IDLE states, instead of
completing the pending work request. If that oldest pending work
request was actually complete and has a CQE in the CQ, then when that
CQE is proceessed in poll_cq, we'll BUG_ON() due to the inconsistent
SQ/CQ state.

This is a very small timing hole and has only been hit once so far.

The fix is two-fold:

1) c4iw_flush_sq() MUST always flush all non-completed WRs with FLUSHED
   status regardless of the QP state.

2) In c4iw_modify_rc_qp(), always set the "in error" bit on the queue
   before moving the state out of RTS.  This ensures that the state
   transition will not happen while another thread is in
   post_rc_send(), because set_state() and post_rc_send() both aquire
   the qp spinlock.  Also, once we transition the state out of RTS,
   subsequent calls to post_rc_send() will fail because the "in error"
   bit is set.  I don't think this fully closes the race where the FW
   can get a FINI followed a SQ work request being posted (because
   they are posted to differente EQs), but the #1 fix will handle the
   issue by flushing the SQ work request.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Some HW platforms can reorder read operations, so we must rmb() after
we see a valid gen bit in a CQE but before we read any other fields
from the CQE.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

1) timedout endpoint processing can be starved. If there are continual
   CPL messages flowing into the driver, the endpoint timeout
   processing can be starved.  This condition exposed the other bugs
   below.

Solution: In process_work(), call process_timedout_eps() after each CPL
is processed.

2) Connection events can be processed even though the endpoint is on
   the timeout list.  If the endpoint is scheduled for timeout
   processing, then we must ignore MPA Start Requests and Replies.

Solution: Change stop_ep_timer() to return 1 if the ep has already been
queued for timeout processing.  All the callers of stop_ep_timer() need
to check this and act accordingly.  There are just a few cases where
the caller needs to do something different if stop_ep_timer() returns 1:

1) in process_mpa_reply(), ignore the reply and  process_timeout()
   will abort the connection.

2) in process_mpa_request, ignore the request and process_timeout()
   will abort the connection.

It is ok for callers of stop_ep_timer() to abort the connection since
that will leave the state in ABORTING or DEAD, and process_timeout()
now ignores timeouts when the ep is in these states.

3) Double insertion on the timeout list.  Since the endpoint timers
   are used for connection setup and teardown, we need to guard
   against the possibility that an endpoint is already on the timeout
   list.  This is a rare condition and only seen under heavy load and
   in the presense of the above 2 bugs.

Solution: In ep_timeout(), don't queue the endpoint if it is already on
the queue.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Signed-off-by: NSteve Wise <swise@opengridcomputing.com>

[ Fix cast from u64* to integer.  - Roland ]
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Add support for the block multicast loopback QP creation flag along
the proper firmware API for that.
Signed-off-by: NEli Cohen <eli@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

As result of the deprecation of the MSI-X/MSI enablement functions
pci_enable_msix() and pci_enable_msi_block(), all drivers using these
two interfaces need to be updated to use the new
pci_enable_msi_range() or pci_enable_msi_exact() and
pci_enable_msix_range() or pci_enable_msix_exact() interfaces.
Signed-off-by: NAlexander Gordeev <agordeev@redhat.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

As result of the deprecation of the MSI-X/MSI enablement functions
pci_enable_msix() and pci_enable_msi_block(), all drivers using these
two interfaces need to be updated to use the new pci_enable_msi_range()
and pci_enable_msix_range() interfaces.
Signed-off-by: NAlexander Gordeev <agordeev@redhat.com>
Acked-by: NMike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

In order to support local WRITE_INSERT + READ_STRIP operations for
non PI enabled fabrics, the fabric driver needs to be able signal
what protection offload operations are supported.

This is done at session initialization time so the modes can be
signaled by individual se_wwn + se_portal_group endpoints, as well
as optionally across different transports on the same endpoint.

For iser-target, set TARGET_PROT_ALL if the underlying ib_device
has already signaled PI offload support, and allow this to be
exposed via a new iscsit_transport->iscsit_get_sup_prot_ops()
callback.

For loopback, set TARGET_PROT_ALL to signal SCSI initiator mode
operation.

For all other drivers, set TARGET_PROT_NORMAL to disable fabric
level PI.

Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Sagi Grimberg <sagig@mellanox.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Quinn Tran <quinn.tran@qlogic.com>
Cc: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Fastreg is mandatory for signature, so if the
device doesn't support it we don't need to use it.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch fixes a bug where outstanding RDMA_READs with WRITE_PENDING
status require an extra target_put_sess_cmd() in isert_put_cmd() code
when called from isert_cq_tx_comp_err() + isert_cq_drain_comp_llist()
context during session shutdown.

The extra kref PUT is required so that transport_generic_free_cmd()
invokes the last target_put_sess_cmd() -> target_release_cmd_kref(),
which will complete(&se_cmd->cmd_wait_comp) the outstanding se_cmd
descriptor with WRITE_PENDING status, and awake the completion in
target_wait_for_sess_cmds() to invoke TFO->release_cmd().

The bug was manifesting itself in target_wait_for_sess_cmds() where
a se_cmd descriptor with WRITE_PENDING status would end up sleeping
indefinately.
Acked-by: NSagi Grimberg <sagig@mellanox.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: <stable@vger.kernel.org> #3.10+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Now that TASK_ABORTED status is not generated for all cases by
TMR ABORT_TASK + LUN_RESET, a new TFO->abort_task() caller is
necessary in order to give fabric drivers a chance to unmap
hardware / software resources before the se_cmd descriptor is
released via the normal TFO->release_cmd() codepath.

This patch adds TFO->aborted_task() in core_tmr_abort_task()
in place of the original transport_send_task_abort(), and
also updates all fabric drivers to implement this caller.

The fabric drivers that include changes to perform cleanup
via ->aborted_task() are:

  - iscsi-target
  - iser-target
  - srpt
  - tcm_qla2xxx

The fabric drivers that currently set ->aborted_task() to
NOPs are:

  - loopback
  - tcm_fc
  - usb-gadget
  - sbp-target
  - vhost-scsi

For the latter five, there appears to be no additional cleanup
required before invoking TFO->release_cmd() to release the
se_cmd descriptor.

v2 changes:
  - Move ->aborted_task() call into transport_cmd_finish_abort (Alex)

Cc: Alex Leung <amleung21@yahoo.com>
Cc: Mark Rustad <mark.d.rustad@intel.com>
Cc: Roland Dreier <roland@kernel.org>
Cc: Vu Pham <vu@mellanox.com>
Cc: Chris Boot <bootc@bootc.net>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Giridhar Malavali <giridhar.malavali@qlogic.com>
Cc: Saurav Kashyap <saurav.kashyap@qlogic.com>
Cc: Quinn Tran <quinn.tran@qlogic.com>
Cc: Sagi Grimberg <sagig@mellanox.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

This patch changes isert_conn_create_fastreg_pool() to follow
logic in iscsi_target_locate_portal() for determining how many
FRMR descriptors to allocate based upon the number of possible
per-session command slots that are available.

This addresses an OOPs in isert_reg_rdma() where due to the
use of ISCSI_DEF_XMIT_CMDS_MAX could end up returning a bogus
fast_reg_descriptor when the number of active tags exceeded
the original hardcoded max.

Note this also includes moving isert_conn_create_fastreg_pool()
from isert_connect_request() to isert_put_login_tx() before
posting the final Login Response PDU in order to determine the
se_nacl->queue_depth (eg: number of tags) per session the target
will be enforcing.

v2 changes:
  - Move isert_conn->conn_fr_pool list_head init into
    isert_conn_request()
v3 changes:
  - Drop unnecessary list_empty() check in isert_reg_rdma()
    (Sagi)

Cc: Sagi Grimberg <sagig@mellanox.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: <stable@vger.kernel.org> #3.12+
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

If during data-transfer a data-integrity error was detected we
must fail the command with CHECK_CONDITION and not execute
the command.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Reported-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Remove code duplication from RDMA_READ and RDMA_WRITE
completions that do basically the same check.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

If protection is involved, iSER target must wait for
completion of RDMA_READ before sending SCSI response.
So we must consider that when calculating post_send_buf_count
additions, also when processing good/error completions.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

As REG_SIG_MR work request and it's LOCAL_INVALIDATE are
not accounted in post_send_buf_count we must color these
with ISER_FASTREG_LI_WRID in order to process their error
completions when the QP flushes.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

In case the Target core passed transport T10 protection
operation:

1. Register data buffer (data memory region)
2. Register protection buffer if exsists (prot memory region)
3. Register signature region (signature memory region)
   - use work request IB_WR_REG_SIG_MR
4. Execute RDMA
5. Upon RDMA completion check the signature status
   - if succeeded send good SCSI response
   - if failed send SCSI bad response with appropriate sense buffer

(Fix up compile error in isert_reg_sig_mr, and fix up incorrect
 se_cmd->prot_type -> TARGET_PROT_NORMAL comparision - nab)
(Fix failed sector assignment in isert_completion_rdma_* - Sagi + nab)
(Fix enum assignements for protection type - Sagi)
(Fix devision on 32-bit in isert_completion_rdma_* - Sagi + Fengguang)
(Fix context change for v3.14-rc6 code - nab)
(Fix iscsit_build_rsp_pdu inc_statsn flag usage - nab)
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

In case of protected transactions, we will need to check the
protection status of the transaction before sending SCSI response.
So be ready for RDMA_WRITE completions. currently we don't ask
for these completions, but for T10-PI we will.
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Introduce pi_context to hold relevant RDMA protection resources.
We eliminate data_key_valid boolean and replace it with indicators
container to indicate:
- Is the descriptor protected (registered via signature MR)
- Is the data_mr key valid (can spare LOCAL_INV WR)
- Is the prot_mr key valid (can spare LOCAL_INV WR)
- Is the sig_mr key valid (can spare LOCAL_INV WR)

Upon connection establishment check if network portal is T10-PI
enabled and allocate T10-PI resources if necessary, allocate
signature enabled memory regions and mark connection queue-pair
as signature enabled.

(Fix context change for v3.14-rc6 code - nab)
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

export map/unmap data buffer to a routine that may
be used in various places in the code and keep the
mapping data in a designated descriptor. Also, let
isert_fast_reg_mr to decide weather to use global
MR or do fast registration.

This commit does not change any functionality.

(Fix context change for v3.14-rc6 code - nab)
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NNicholas Bellinger <nab@linux-iscsi.org>

Unregister the inet notifier during ocrdma unload to avoid a panic after
driver unload.
Signed-off-by: NSelvin Xavier <selvin.xavier@emulex.com>
Signed-off-by: NDevesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

We should cast pointers to and from unsigned long to turn them into ints.
Signed-off-by: NRoland Dreier <roland@purestorage.com>