Add "stags" debugfs file.  This is useful for examining the TPTE and
PBL entries in adapter memory.  It allows scripts to dump just the
active entries.

Also clean up the "qps" file handlers and shared common code.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

This helps debug cases where HW resources are depleted.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

T4 FW sends up CPL_RDMA_TERMINATE to indicate a peer TERM.  This
triggers the QP moving to TERMINATE state.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

T4 incorrectly inserts TERM CQEs into the CQ.  Silently ignore them.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

The cxgb4_*_send() functions return NET_XMIT_ values, which are
positive integers or negative errno values.  So don't treat positive
return values as an error.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

The HW design requires zeroing any pad in SGLs.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

In c4iw_modify_qp() error path, only use qhp->ep if ep is not already set.
Otherwise qhp->ep can be NULL and we crash.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Fix:

  drivers/infiniband/hw/cxgb4/qp.c: In function ‘create_qp’:
  drivers/infiniband/hw/cxgb4/qp.c:147: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/qp.c: In function ‘rdma_fini’:
  drivers/infiniband/hw/cxgb4/qp.c:988: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/qp.c: In function ‘rdma_init’:
  drivers/infiniband/hw/cxgb4/qp.c:1063: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/mem.c: In function ‘write_adapter_mem’:
  drivers/infiniband/hw/cxgb4/mem.c:74: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/cq.c: In function ‘destroy_cq’:
  drivers/infiniband/hw/cxgb4/cq.c:58: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/cq.c: In function ‘create_cq’:
  drivers/infiniband/hw/cxgb4/cq.c:135: warning: cast from pointer to integer of different size
  drivers/infiniband/hw/cxgb4/cm.c: In function ‘fw6_msg’:
  drivers/infiniband/hw/cxgb4/cm.c:2326: warning: cast to pointer from integer of different size

by casting pointers to unsigned long instead of u64.
Reported-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

The HW by default has RX coalescing on.  For iWARP connections, this
causes a 100ms delay in connection establishement due to the ingress
MPA Start message being stalled in HW.  So explicitly turn RX
coalescing off when setting up iWARP connections.

This was causing very bad performance for NP64 gather operations using
Open MPI, due to the way it sets up connections on larger jobs.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Cc: <stable@kernel.org>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Changing state to CLOSING when FIN is received causes A0 cards to
hang.  Fix this by checking for A0 cards in FIN handling.
Signed-off-by: NFaisal Latif <faisal.latif@intel.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

When the driver receives an AE for FIN received, it closes the
connection without changing the state of the connection in the
hardware to closing.  By changing the state to closing, hardware will
do a normal close sequence.
Signed-off-by: NFaisal Latif <faisal.latif@intel.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

During a stress testing in a large cluster, multiple close event are
detected and BUG() is hit in the iWARP core.  The cause is that the
active node gave up while waiting for an MPA response from the peer
and tried to close the connection by sending RST.  The passive node
driver receives the RST but is waiting for MPA response from the user.
When the MPA accept is received, the driver offloads the connection
and sends a CLOSE event.  The driver gets an AE indicating RESET
received and also sends a CLOSE event, hitting a BUG().

Fix this by correcting RESET handling and sending CLOSE events.
Signed-off-by: NFaisal Latif <faisal.latif@intel.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Setting TX pause param writes to the wrong register location causing
the adapter to hang.  Correct the define used to write the reigster.

Addresses: https://bugs.openfabrics.org/show_bug.cgi?id=2116Reported-by: NShiri Franchi <shirif@voltaire.com>
Signed-off-by: NChien Tung <chien.tin.tung@intel.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

The max depth supported by T3 is 64K entries.  This fixes a bug
introduced in commit 9918b28d ("RDMA/cxgb3: Increase the max CQ
depth") that causes stalls and possibly crashes in large MPI clusters.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Cc: <stable@kernel.org>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

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

of_device is just an alias for platform_device, so remove it entirely.  Also
replace to_of_device() with to_platform_device() and update comment blocks.

This patch was initially generated from the following semantic patch, and then
edited by hand to pick up the bits that coccinelle didn't catch.

@@
@@
-struct of_device
+struct platform_device
Signed-off-by: NGrant Likely <grant.likely@secretlab.ca>
Reviewed-by: NDavid S. Miller <davem@davemloft.net>

Fix build failure on sparc64 which is missing the include of
<linux/slab.h> via <asm/pci.h> that x86, powerpc, ia64, etc. have.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

list_for_each_entry binds its first argument to a non-null value, and thus
any null test on the value of that argument is superfluous.

The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
iterator I;
expression x;
statement S,S1,S2;
@@

I(x,...) { <...
- if (x == NULL && ...) S
  ...> }
// </smpl>
Signed-off-by: NJulia Lawall <julia@diku.dk>
Acked-by: NAlexander Schmidt <alexs@linux.vnet.ibm.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Fix confusing indentation that makes a statement look as if it's part of
an if statement when in fact it isn't.
Reported-by: NJulia Lawall <julia@diku.dk>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Initialize tasklet before interrupts are requested to prevent
scheduling of an uninitialized tasklet.
Signed-off-by: NAlexander Schmidt <alexs@linux.vnet.ibm.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

In nes_probe(), a bit of code is indented one tab stop too far.  Fix this.
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

In nes_show_wqm_quanta(), the wrong value is printed.  Fix this.
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Delete dead code in various places that is shown by gcc 4.6's new
-Wunused-but-set-variable warnings.
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Signed-off-by: NMirek Walukiewicz <miroslaw.walukiewicz@intel.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Export req_lim via sysfs for debugging.
Signed-off-by: NBart Van Assche <bart.vanassche@gmail.com>
Acked-by: NDavid Dillow <dave@thedillows.org>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

The current strategy in ib_srp for posting receive buffers is:

 * Post one buffer after channel establishment.
 * Post one buffer before sending an SRP_CMD or SRP_TSK_MGMT to the target.

As a result, only the first non-SRP_RSP information unit from the
target will be processed.  If that first information unit is an
SRP_T_LOGOUT, it will be processed.  On the other hand, if the
initiator receives an SRP_CRED_REQ or SRP_AER_REQ before it receives a
SRP_T_LOGOUT, the SRP_T_LOGOUT won't be processed.

We can fix this inconsistency by changing the strategy for posting
receive buffers to:

 * Post all receive buffers after channel establishment.
 * After a receive buffer has been consumed and processed, post it again.

A side effect is that the ib_post_recv() call is moved out of the SCSI
command processing path.  Since __srp_post_recv() is not called
directly any more, get rid of it and move the code directly into
srp_post_recv().  Also, move srp_post_recv() up in the file to avoid a
forward declaration.
Signed-off-by: NBart Van Assche <bart.vanassche@gmail.com>
Acked-by: NDavid Dillow <dave@thedillows.org>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Replace an open-coded dump of the receive buffer with a call to
print_hex_dump().
Signed-off-by: NBart Van Assche <bart.vanassche@gmail.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Change abbreviated IB_QPT_RAW_ETY to IB_QPT_RAW_ETHERTYPE to make
the special QP type easier to understand.

cf http://www.mail-archive.com/linux-rdma@vger.kernel.org/msg04530.htmlSigned-off-by: NAleksey Senin <alekseys@voltaire.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Simple changes to fix warnings:

      CHECK   drivers/infiniband/hw/nes/nes_verbs.c
    nes_verbs.c:1944:45: warning: Using plain integer as NULL pointer
    nes_verbs.c:1944:48: warning: Using plain integer as NULL pointer
      CHECK   drivers/infiniband/hw/nes/nes_cm.c
    nes_cm.c:2645:43: warning: mixing different enum types
    nes_cm.c:2645:43:     int enum iw_cm_event_type  versus
    nes_cm.c:2645:43:     int enum iw_cm_event_status
Signed-off-by: NOr Gerlitz <ogerlitz@voltaire.com>
Acked-by: NChien Tung <chien.tin.tung@intel.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Signed-off-by: NOr Gerlitz <ogerlitz@voltaire.com>
Acked-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Signed-off-by: NOr Gerlitz <ogerlitz@voltaire.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Don't hang a host thread if the FW stops responding.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

When transitioning a QP to the error state, in progress RWQEs need to
be marked complete.  This also involves releasing the reference count
to the memory regions referenced in the SGEs.  The locking in the
receive packet processing wasn't sufficient to prevent qib_error_qp()
from modifying the r_sge state at the same time, thus leading to
kernel panics.
Signed-off-by: NRalph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Don't processes too many packets without allowing other IRQ functions
a chance to run. Otherwise, there is a chance of getting a "soft
lockup" messages and poor application response times.
Signed-off-by: NRalph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Signed-off-by: NIra Weiny <weiny2@llnl.gov>
Acked-by: NRalph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Up to now, we have set the number of available user contexts based on
the number of hardware contexts which is set according to the number
of available CPUs.  This was fine since most CPUs had a power of two
number of cores and the chip supported 4, 8, or 16 user contexts.  Now
that some systems have 12 cores, the default isn't optimal and should
be set to 12 even though 16 hardware contexts need to be enabled.
Signed-off-by: NRalph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

Endpoint timer manipulation needs to be done inside the lock.  Otherwise
we can get into a situation where a timer is stopped before it is started,
which hits the WARN_ON() in stop_ep_timer().
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

There is only one control txq per tx channel.  So use the port number
as the queue index when sending.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

There exists a race condition where the app disconnects, which
initiates an orderly close (via rdma_fini()), concurrently with an
ingress abort condition, which initiates an abortive close operation.
Since rdma_fini() must be called without IRQs disabled, the fini can
be called after the QP has been transitioned to ERROR.  This is ok,
but we need to protect against qp->ep getting NULLed.
Signed-off-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>

NULL pointer dereferences in ib_cm_init_qp_attr() were seen by some
users.  From a crash dump, I determined that we died in
cm_init_qp_rts_attr() (it's inlined, so it doesn't show up in the
traceback) on the line labeled below:

static int cm_init_qp_rts_attr(struct cm_id_private *cm_id_priv,
                               struct ib_qp_attr *qp_attr,
                               int *qp_attr_mask)
{
        ........
        if (cm_id_priv->id.lap_state == IB_CM_LAP_UNINIT) {
                .....
        } else {
               *qp_attr_mask = IB_QP_ALT_PATH | IB_QP_PATH_MIG_STATE;
               qp_attr->alt_port_num = cm_id_priv->alt_av.port->port_num; <-die


The problem is that the rdma_cm can call ib_send_cm_mra() after a
connection has been established.  The ib_cm incorrectly assumes that
the MRA is in response to a LAP (load alternate path) message, even
though no LAP message has been received.  The ib_cm needs to check the
lap_state before sending an MRA if the cm_id state is established.
Reported-by: NArthur Kepner <akepner@sgi.com>
Reported-by: NJosh England <jjengla@gmail.com>
Signed-off-by: NSean Hefty <sean.hefty@intel.com>
Signed-off-by: NRoland Dreier <rolandd@cisco.com>