Dave Miller <davem@davemloft.net> provided a detailed description of
why the way IPoIB is using neighbours for its own ipoib_neigh struct
is buggy:
Any time an ipoib_neigh is changed, a sequence like the following is made:
spin_lock_irqsave(&priv->lock, flags);
/*
* It's safe to call ipoib_put_ah() inside
* priv->lock here, because we know that
* path->ah will always hold one more reference,
* so ipoib_put_ah() will never do more than
* decrement the ref count.
*/
if (neigh->ah)
ipoib_put_ah(neigh->ah);
list_del(&neigh->list);
ipoib_neigh_free(dev, neigh);
spin_unlock_irqrestore(&priv->lock, flags);
ipoib_path_lookup(skb, n, dev);
This doesn't work, because you're leaving a stale pointer to the freed up
ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails
with all the locking done to protect _changes_ to *ipoib_neigh(n), and
with the code in ipoib_neigh_free() that NULLs out the pointer.
The core issue is that read side calls to *to_ipoib_neigh(n) are not
being synchronized at all, they are performed without any locking. So
whether we hold the lock or not when making changes to *ipoib_neigh(n)
you still can have threads see references to freed up ipoib_neigh
objects.
cpu 1 cpu 2
n = *ipoib_neigh()
*ipoib_neigh() = NULL
kfree(n)
n->foo == OOPS
[..]
Perhaps the ipoib code can have a private path database it manages
entirely itself, which holds all the necessary information and is
looked up by some generic key which is available easily at transmit
time and does not involve generic neighbour entries.
See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and
<http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b>
for the full discussion.
This patch aims to solve the race conditions found in the IPoIB driver.
The patch removes the connection between the core networking neighbour
structure and the ipoib_neigh structure. In addition to avoiding the
race described above, it allows us to handle SKBs carrying IP packets
that don't have any associated neighbour.
We add an ipoib_neigh hash table with N buckets where the key is the
destination hardware address. The ipoib_neigh is fetched from the
hash table and instead of the stashed location in the neighbour
structure. The hash table uses both RCU and reference counting to
guarantee that no ipoib_neigh instance is ever deleted while in use.
Fetching the ipoib_neigh structure instance from the hash also makes
the special code in ipoib_start_xmit that handles remote and local
bonding failover redundant.
Aged ipoib_neigh instances are deleted by a garbage collection task
that runs every M seconds and deletes every ipoib_neigh instance that
was idle for at least 2*M seconds. The deletion is safe since the
ipoib_neigh instances are protected using RCU and reference count
mechanisms.
The number of buckets (N) and frequency of running the GC thread (M),
are taken from the exported arb_tbl.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Commit 36a8f01cd2 ("IB/qib: Add congestion control agent
implementation") tries to store the value 1984 in a u8, which leads to
truncation. Fix this by making the member big enough.
This bug was detected by a smatch warning.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Ramkrishna Vepa <ramkrishna.vepa@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
It looks like one check was accidentally duplicated, and the other 3
checks were left out. This was detected by scripts/coccinelle/tests/doubletest.cocci:
drivers/infiniband/hw/ocrdma/ocrdma_verbs.c:895:6-54: duplicated argument to && or ||
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Elminate some simple_strto* usage.
checkpatch also noted pr_ conversations, which have been done as
recommended. The pr_fmt() define is used to shorten line length.
Other multi-line string warnings are also elmininated.
Reviewed-by: Dean Luick <dean.luick@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Add a congestion control agent in the driver that handles gets and
sets from the congestion control manager in the fabric for the
Performance Scale Messaging (PSM) library.
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Profiling has shown that sdma_lock is proving a bottleneck for
performance. The situations include:
- RDMA reads when krcvqs > 1
- post sends from multiple threads
For RDMA read the current global qib_wq mechanism runs on all CPUs
and contends for the sdma_lock when multiple RMDA read requests are
fielded on differenct CPUs. For post sends, the direct call to
qib_do_send() from multiple threads causes the contention.
Since the sdma mechanism is per port, this fix converts the existing
workqueue to a per port single thread workqueue to reduce the lock
contention in the RDMA read case, and for any other case where the QP
is scheduled via the workqueue mechanism from more than 1 CPU.
For the post send case, This patch modifies the post send code to test
for a non empty sdma engine. If the sdma is not idle the (now single
thread) workqueue will be used to trigger the send engine instead of
the direct call to qib_do_send().
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
There is a cut-and-paste typo in the function qib_pci_slot_reset()
where it prints that the "link_reset" function is called rather than
the "slot_reset" function. This makes the message misleading.
Signed-off-by: Betty Dall <betty.dall@hp.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Enable callers of mlx4_assign_eq to supply a pointer to cpu_rmap.
If supplied, the assigned IRQ is tracked using rmap infrastructure.
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit af061a644a ("IB/qib: Use RCU for qpn lookup") introduced sparse
warnings.
This patch corrects those issues.
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
This will be used so that we can compose a full flow key.
Even though we have a route in this context, we need more. In the
future the routes will be without destination address, source address,
etc. keying. One ipv4 route will cover entire subnets, etc.
In this environment we have to have a way to possess persistent storage
for redirects and PMTU information. This persistent storage will exist
in the FIB tables, and that's why we'll need to be able to rebuild a
full lookup flow key here. Using that flow key will do a fib_lookup()
and create/update the persistent entry.
Signed-off-by: David S. Miller <davem@davemloft.net>
srpt_handle_rdma_comp is called from kthread context and thus can execute
target_execute_cmd directly. srpt_abort_cmd sets the CMD_T_LUN_STOP
flag directly, and thus the abuse of transport_generic_handle_data can be
replaced with an opencoded variant of that code path. I'm still not happy
about a fabric driver poking into target core internals like this, but
let's defer the bigger architecture changes for now.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Clean the idr as part of the error flow since it is a resource too.
Signed-off-by: Dotan Barak <dotanb@dev.mellanox.co.il>
Reviewed-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
When the user queries for device capabilities, fill in the
masked_atomic_cap attribute with the real support level of atomic
capabilities instead of using a hard coded value.
Signed-off-by: Dotan Barak <dotanb@dev.mellanox.co.il>
Reviewed-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The query QP code was didn't fill that attribute, do that.
Signed-off-by: Dotan Barak <dotanb@dev.mellanox.co.il>
Reviewed-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Events received for non-existent QPs should generate a warning that includes
the event type that was received.
Signed-off-by: Dotan Barak <dotanb@dev.mellanox.co.il>
Reviewed-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Or Gerlitz reported triggering of WARN_ON_ONCE(delta < len); in
skb_try_coalesce()
This warning tracks drivers that incorrectly set skb->truesize
IPoIB indeed allocates a full page to store a fragment, but only
accounts in skb->truesize the used part of the page (frame length)
This patch fixes skb truesize underestimation, and
also fixes a performance issue, because RX skbs have not enough tailroom
to allow IP and TCP stacks to pull their header in skb linear part
without an expensive call to pskb_expand_head()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Erez Shitrit <erezsh@mellanox.com>
Cc: Shlomo Pongartz <shlomop@mellanox.com>
Cc: Roland Dreier <roland@purestorage.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The port management change event can replace smp_snoop. If the
capability bit for this event is set in dev-caps, the event is used
(by the driver setting the PORT_MNG_CHG_EVENT bit in the async event
mask in the MAP_EQ fw command). In this case, when the driver passes
incoming SMP PORT_INFO SET mads to the FW, the FW generates port
management change events to signal any changes to the driver.
If the FW generates these events, smp_snoop shouldn't be invoked in
ib_process_mad(), or duplicate events will occur (once from the
FW-generated event, and once from smp_snoop).
In the case where the FW does not generate port management change
events smp_snoop needs to be invoked to create these events. The flow
in smp_snoop has been modified to make use of the same procedures as
in the fw-generated-event event case to generate the port management
events (LID change, Client-rereg, Pkey change, and/or GID change).
Port management change event handling required changing the
mlx4_ib_event and mlx4_dispatch_event prototypes; the "param" argument
(last argument) had to be changed to unsigned long in order to
accomodate passing the EQE pointer.
We also needed to move the definition of struct mlx4_eqe from
net/mlx4.h to file device.h -- to make it available to the IB driver,
to handle port management change events.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Profiling indicates that MR validation locking is expensive. The MR
table is largely read-only and is a suitable candidate for RCU locking.
The patch uses RCU locking during validation to eliminate one
lock/unlock during that validation.
Reviewed-by: Mike Heinz <michael.william.heinz@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
A timing issue can occur where qib_mr_dereg can return -EBUSY if the
MR use count is not zero.
This can occur if the MR is de-registered while RDMA read response
packets are being progressed from the SDMA ring. The suspicion is
that the peer sent an RDMA read request, which has already been copied
across to the peer. The peer sees the completion of his request and
then communicates to the responder that the MR is not needed any
longer. The responder tries to de-register the MR, catching some
responses remaining in the SDMA ring holding the MR use count.
The code now uses a get/put paradigm to track MR use counts and
coordinates with the MR de-registration process using a completion
when the count has reached zero. A timeout on the delay is in place
to catch other EBUSY issues.
The reference count protocol is as follows:
- The return to the user counts as 1
- A reference from the lk_table or the qib_ibdev counts as 1.
- Transient I/O operations increase/decrease as necessary
A lot of code duplication has been folded into the new routines
init_qib_mregion() and deinit_qib_mregion(). Additionally, explicit
initialization of fields to zero is now handled by kzalloc().
Also, duplicated code 'while.*num_sge' that decrements reference
counts have been consolidated in qib_put_ss().
Reviewed-by: Ramkrishna Vepa <ramkrishna.vepa@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
An MR reference leak exists when handling UC RDMA writes with
immediate data because we manipulate the reference counts as if the
operation had been a send.
This patch moves the last_imm label so that the RDMA write operations
with immediate data converge at the cq building code. The copy/mr
deref code is now done correctly prior to the branch to last_imm.
Reviewed-by: Edward Mascarenhas <edward.mascarenhas@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
These macros will be reused by the mlx4 SRIOV-IB CM paravirtualization
code, and there is no reason to have them declared both in the IB core
in the mlx4 IB driver.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
This query is needed for SRIOV alias GUID support.
The query is implemented per the IB Spec definition
in section 15.2.5.18 (GuidInfoRecord).
Signed-off-by: Erez Shitrit <erezsh@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Define pr_fmt and add some pr_debug prints.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
sparse correctly warns that if mpa->private_data_size is __be16, then
doing += on it is wrong, even if we do += htons(<something>) -- on a
little endian system, carries will go the wrong way. Fix this up by
doing the addition in native byte order.
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Provide an option for the user to specify that listens should only
accept connections where the incoming address family matches that of
the locally bound address. This is used to support the equivalent of
IPV6_V6ONLY socket option, which allows an app to only accept
connection requests directed to IPv6 addresses.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The rdma_cm maps IPv4 and IPv6 addresses to the same service ID. This
prevents apps from listening only for IPv4 or IPv6 addresses. It also
results in an app binding to an IPv4 address receiving connection
requests for an IPv6 address.
Change this to match socket behavior: restrict listens on IPv4
addresses to only IPv4 addresses, and if a listen is on an IPv6
address, allow it to receive either IPv4 or IPv6 addresses, based on
its address family binding.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The RDMA CM uses a single port space for all associated (tcp, udp,
etc.) port bindings, regardless of the address family that the user
binds to. The result is that if a user binds to AF_INET, but does not
specify an IP address, the bind will occur for AF_INET6. This causes
an attempt to bind to the same port using AF_INET6 to fail, and
connection requests to AF_INET6 will match with the AF_INET listener.
Align the behavior with sockets and restrict the bind to AF_INET only.
If a user binds to AF_INET6, we bind the port to AF_INET6 and
AF_INET depending on the value of bindv6only.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We want to set attr->max_srq_sge to dev->attr.max_srq_sge, not to itself.
This was detected by Coverity (CID 709210).
Signed-off-by: Roland Dreier <roland@purestorage.com>
This patch adds the following structure:
struct netlink_kernel_cfg {
unsigned int groups;
void (*input)(struct sk_buff *skb);
struct mutex *cb_mutex;
};
That can be passed to netlink_kernel_create to set optional configurations
for netlink kernel sockets.
I've populated this structure by looking for NULL and zero parameters at the
existing code. The remaining parameters that always need to be set are still
left in the original interface.
That includes optional parameters for the netlink socket creation. This allows
easy extensibility of this interface in the future.
This patch also adapts all callers to use this new interface.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change || check to the intended && when checking the QP type in a
received connection request against the listening endpoint.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The dev->sgid_tbl[] array is allocated in ocrdma_alloc_resources().
It has OCRDMA_MAX_SGID elements so the test here is off by one.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Fix RQ/SRQ error CQE polling. Return error CQE to consumer for error
case which was not returned previously.
Signed-off-by: Parav Pandit <parav.pandit@emulex.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Fix max sge calculation for sq, rq, srq for all hardware types.
Signed-off-by: Mahesh Vardhamanaiah <mahesh.vardhamanaiah@emulex.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Fix code to read the max wqe and max rqe values from mailbox response.
Signed-off-by: Mahesh Vardhamanaiah <mahesh.vardhamanaiah@emulex.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
1. Fix reporting GID table addition events.
2. Enable vlan based GID entries only when VLAN is enabled at compile
time (test CONFIG_VLAN_8021Q / CONFIG_VLAN_8021Q_MODULE).
Signed-off-by: Parav Pandit <parav.pandit@emulex.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
1. Limit the max number of WQEs per QP reported when querying the
device, so that ib_create_qp() will not fail for a QP size that the
device claimed to support due to additional headroom WQEs being
allocated.
2. Limit qp resources accepted for ib_create_qp() to the limits
reported in ib_query_device(). In kernel space, make sure that the
limits returned to the caller following qp creation also lie within
the reported device limits. For userspace, report as before, and do
adjustment in libmlx4 (so as not to break ABI).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Sagi Grimberg <sagig@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Commit e605b743f3 ("IB/mlx4: Increase the number of vectors (EQs)
available for ULPs") didn't handle correctly the case where there
aren't enough MSI-X vectors to increase the number of EQs, so only the
legacy EQs are allocated. This results in an attempt to memset() to
zero the EQ table which was never allocated and a kernel crash.
Fix this by checking in the teardown flow if the table of EQs was ever
allocated. Also remove some unneeded setting to zero of the EQ
related fields in struct mlx4_ib_dev.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
When using rping -c -a 0.0.0.0 with iw_cxgb4, the system crashes when
rdma_connect() is called. ip_dev_find() will return NULL, but pdev is
accessed anyway.
Checking that pdev is NULL and returning -ENODEV prevents the system
from crashing.
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Shlomo Pongratz