Clean up properly if pci_set_consistent_dma_mask() fails.
Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Some PCIe root complex chip sets don't support advanced error reporting.
Allow the driver to load OK if pci_enable_pcie_error_reporting() fails.
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
If CONFIG_PCI_MSI is not set, and a QLE7140 is present, the pointer
"dd" is uninitialized.
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Noticed this odd looking thing in dmesg:
ib_qib 0000:02:00.0: <3>ib_qib: Unable to enable pcie error reporting: -5
which is due to a bad use of dev_info.
Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Acked-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Since an IB transport port may use either IB or Ethernet as its link layer,
add the file /sys/class/infiniband/<device>/ports/<port_num>/link_layer to
show the link layer for the port.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This patch allows IBoE traffic to be encapsulated in 802.1Q tagged
VLAN frames. The VLAN tag is encoded in the GID and derived from it
by a simple computation.
The netdev notifier callback is modified to catch VLAN device
addition/removal and the port's GID table is updated to reflect the
change, so that for each netdevice there is an entry in the GID table.
When the port's GID table is exhausted, GID entries will not be added.
Only children of the main interfaces can add to the GID table; if a
VLAN interface is added on another VLAN interface (e.g. "vconfig add
eth2.6 8"), then that interfaces will not add an entry to the GID
table.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add 802.1q VLAN support to IBoE. The VLAN tag is encoded within the
GID derived from a link local address in the following way:
GID[11] GID[12] contain the VLAN ID when the GID contains a VLAN.
The 3 bits user priority field of the packets are identical to the 3
bits of the SL.
In case of rdma_cm apps, the TOS field is used to generate the SL
field by doing a shift right of 5 bits effectively taking to 3 MS bits
of the TOS field.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add support for IBoE to mlx4_ib. The bulk of the code is handling the
new address vector fields; mlx4 needs the MAC address of a remote node
to include it in a WQE (for datagrams) or in the QP context (for
connected QPs). Address resolution is done by assuming all unicast
GIDs are either link-local IPv6 addresses.
Multicast group attach/detach needs to update the NIC's multicast
filters; but since attaching a QP to a multicast group can be done
before the QP is bound to a port, for IBoE we need to keep track of
all multicast groups that a QP is attached too before it transitions
from INIT to RTR (since it does not have a port in the INIT state).
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
[ Many things cleaned up and otherwise monkeyed with; hope I didn't
introduce too many bugs. - Roland ]
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Change multicast promiscuous mode to pass packets through the multicast group distribution table
before sending packets that miss to the default multicast QP.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add fields to hardware data structures and add new constants required for IBoE
support.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a mechanism for mlx4 protocol drivers to get a pointer to other
drivers's device objects. For this, an exported function,
mlx4_get_protocol_dev() is added, which allows a driver to get some
other driver's device based on the protocol that the driver
implements. Two protocols are added: MLX4_PROTOCOL_IB and
MLX4_PROTOCOL_EN.
This will be used in mlx4 IBoE support so that mlx4_ib can find the
corresponding mlx4_en netdev.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
[ Clean up and rename a few things. - Roland ]
Signed-off-by: Roland Dreier <rolandd@cisco.com>
srp_send_tsk_mgmt() was missing the proper DMA sync calls before posting
the buffer to the device.
Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Use the list_first_entry() macro in ib_srp instead of open-coding the equivalent,
which makes the source code slightly more descriptive. The list_first_entry()
macro itself was introduced in kernel 2.6.22.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
As proposed by the SRP (draft) standard, ib_srp reserves one ring
element for SRP_TSK_MGMT requests. This patch makes sure that the SCSI
mid-layer never tries to queue more than (SRP request limit) - 1 SCSI
commands to ib_srp. This improves performance for targets whose request
limit is less than or equal to SRP_NORMAL_REQ_SQ_SIZE by reducing the
number of BUSY responses reported by ib_srp to the SCSI mid-layer.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The Node Description cannot be changed via MADs (it is read-only).
Until now, it was changed in the driver via sysfs, and the new Node
Description was simply inserted by the driver into MAD responses
(replacing the description returned by FW).
System startup scripts use the sysfs interface to change the node
description at driver startup to show the hostname, etc. However, this
has a race condition: the SM could discover the original FW node
description rather than the system-specific description if it queried the
port before the startup scripts finish running.
For mlx4, we fix this with a new FW command (SET_NODE) that allows
passing the new node description to FW. When this command is invoked,
FW sends a trap 144 to the SM. When it gets this trap, the SM can
query the node to obtain the new node description -- thus eliminating
the effects of the race.
This patch simply calls SET_NODE command when a new node description
is entered via sysfs (thus causing trap 144 to be issued by the FW).
We ignore all failures of the SET_NODE command (including those caused
by using a device FW that predates the SET_NODE command), since in
that case things work just as before.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
For iWARP connections, the connect request is carried in a TCP payload
on an already established TCP connection. So if the ucma's backlog is
full, the connection request is transmitted and acked at the TCP level
by the time the connect request gets dropped in the ucma. The end
result is the connection gets rejected by the iWARP provider.
Further, a 32 node 256NP OpenMPI job will generate > 128 connect
requests on some ranks.
This patch increases the default max backlog to 1024, and adds a
sysctl variable so the backlog can be adjusted at run time.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Use the net device's dev_id field to encode the port number of the pci
device. This can be used to to associate a net device with the pci
device's port. The encoding is: dev_id = port - 1.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This patch adds support for SRP_CRED_REQ to avoid a lockup by targets
that use that mechanism to return credits to the initiator. This
prevents a lockup observed in the field where we would never add the
credits from the SRP_CRED_REQ to our current count, and would therefore
never send another command to the target.
Minimal support for SRP_AER_REQ is also added, as these messages can
also be used to convey additional credits to the initiator.
Based upon extensive debugging and code by Bart Van Assche and a bug
report by Chris Worley.
Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The transmit ring in ib_srp (srp_target.tx_ring) is currently only used
for allocating requests sent by the initiator to the target. This patch
prepares using that ring for allocation of both requests and responses.
Also, this patch differentiates the uses of SRP_SQ_SIZE, increases the
size of the IB send completion queue by one element and reserves one
transmit ring slot for SRP_TSK_MGMT requests.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
See table 35 in IBA - the header order for RDMA_WRITE_ONLY_WITH_IMMEDIATE
and SEND_LAST_WITH_IMMEDIATE is different: the RDMA_WRITE_ONLY has
a RETH header before the immediate data, so we need a different code path
to extract the immediate data.
I tested this with a userspace app that does RDMA_WRITE with immediate
on a QLE7140.
Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The flushing of work requests for user QPs is implemented entirely in
the user mode library. The only kernel interaction is to mark the
user QP object indicating it is in error when the QP exits RTS. When
the user QP operations are called by the application (eg: post_send,
post_recv), the QP in error bit is checked and if set, the library
flushes the QP. If, however, the application is not doing IO, but
rather just polling the CQ, it will never get flushed work requests.
This breaks some classes of applications.
This patch adds logic to mark user CQs in error when a QP that is bound
to the CQ is marked in error. The library poll code can then notice
the CQ is in error and flush all the in error QPs bound to that CQ.
Design:
- add 1 extra CQE entry to the CQ memory that will be used to indicate
in error status.
- return the desired CQ memory size that should be mapped by the library
- bump the ABI since the create_cq uverbs response changes.
- detect older libraries and reduce the mmap size accordingly.
(The ABI bump doesn't break old libraries, since they didn't check
the ABI field anyway)
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Remove the local service t4_pktgl_to_skb() and use cxgb4_pktgl_to_skb()
exported by cxgb4.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add support for packing IBoE packet headers.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
[ Clean up and fix ib_ud_header_init() a bit. - Roland ]
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add support for IBoE device binding and IP --> GID resolution. Path
resolving and multicast joining are implemented within cma.c by
filling in the responses and running callbacks in the CMA work queue.
IP --> GID resolution always yields IPv6 link local addresses; remote
GIDs are derived from the destination MAC address of the remote port.
Multicast GIDs are always mapped to multicast MACs as is done in IPv6.
(IPv4 multicast is enabled by translating IPv4 multicast addresses to
IPv6 multicast as described in
<http://www.mail-archive.com/ipng@sunroof.eng.sun.com/msg02134.html>.)
Some helper functions are added to ib_addr.h.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Since IBoE is using Ethernet as its link layer, there is no central
management entity so there is need for QP0. QP1 is still needed since
it handles communications between CM agents. This patch will skip QP0
and create only QP1 for IBoE ports.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
IPoIB is IP-over-Infiniband link layer. In the case of IBoE, the link
layer is Ethernet and IP can work directly over Ethernet, so disable
IPoIB for non-IB_LINK_LAYER_INFINIBAND ports.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
A process can get stuck in an uninterruptible wait in the
kernel while destroying a cm_id when iw_cm_connect() fails:
For example, When creation of a PD fails but the user continues with
an attempt to connect to the server without checking the return value,
in iw_cm_connect() a NULL qp is found so the call fails. However the
IWCM_F_CONNECT_WAIT bit is not cleared. destroy_cm_id() then waits
forever for IWCM_F_CONNECT_WAIT to be cleared.
The same problem exists on the passive side with the accept call.
Fix this by clearing the bit and waking up any waiters in the
appropriate spots.
Signed-off-by: Animesh Trivedi <atr@zurich.ibm.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
We can replace our equivalent open-coded version.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Fix the limit on the size of max fast registration WRs that can be
posted to match hardware capabilities.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This lets the bonding driver to detect when an interface goes down.
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Chien Tung <chien.tin.tung@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
With commit cd6860eb ("RDMA/nes: Fix hangs on ifdown") we no longer
remove nes interfaces on ifdown. On nes_query_port(), add an
additional check of the netdev queue and report IB_PORT_DOWN if the
queue is not running.
Signed-off-by: Chien Tung <chien.tin.tung@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
the eHCA driver registers a MR for all of kernel memory, but makes the
assumption that valid memory exists at KERNELBASE. This assumption
may not be true in the case of a relocatable kernel, so use KERNELBASE
+ PHYSICAL_START to get the true beginning of usable kernel memory.
cc: Joachim Fenkes <fenkes@de.ibm.com>
cc: Christoph Raisch <raisch@de.ibm.com>
cc: Hoan-Ham Hguyen <hnguyen@de.ibm.com>
Signed-off-by: Sonny Rao <sonnyrao@us.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Get rid of init_MUTEX[_LOCKED]() and use sema_init() instead.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Just a small cleanup. The "passive_state" variable isn't used any
more after commit dae58728dc ("RDMA/nes: Fix double CLOSE event
indication crash")
Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: Faisal Latif <faisal.latif@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
As there are now machines with a lot more physical memory, we need to
be able to register more memory. This patch lifts the upper limit of
log_mtts_per_seg from 5 to 7, increasing the amount of memory that can
be registered.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
IGMP processing is broken because the IPOIB does not set the
skb->pkt_type the right way for multicast traffic. All incoming
packets are set to PACKET_HOST which means that igmp_recv() will
ignore the IGMP broadcasts/multicasts.
This in turn means that the IGMP timers are firing and are sending
information about multicast subscriptions unnecessarily. In a large
private network this can cause traffic spikes.
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
- Remove dsgl support - doesn't work in T4.
- Wrap the immediate PBL as needed when building it in the wr.
- Adjust max pbl depth allowed based on ulptx alignment requirements.
- Bump the slots per SQ to 5 to allow up to 128MB fast registers.
- Advertise fastreg support by default.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
