linux.git/drivers/infiniband/ulp/ipoib, branch v3.12

IPoIB: Fix race in deleting ipoib_neigh entries

2013-08-13T18:57:37+00:00

In several places, this snippet is used when removing neigh entries:

	list_del(&neigh->list);
	ipoib_neigh_free(neigh);

The list_del() removes neigh from the associated struct ipoib_path, while
ipoib_neigh_free() removes neigh from the device's neigh entry lookup
table.  Both of these operations are protected by the priv->lock
spinlock.  The table however is also protected via RCU, and so naturally
the lock is not held when doing reads.

This leads to a race condition, in which a thread may successfully look
up a neigh entry that has already been deleted from neigh->list.  Since
the previous deletion will have marked the entry with poison, a second
list_del() on the object will cause a panic:

  #5 [ffff8802338c3c70] general_protection at ffffffff815108c5
     [exception RIP: list_del+16]
     RIP: ffffffff81289020  RSP: ffff8802338c3d20  RFLAGS: 00010082
     RAX: dead000000200200  RBX: ffff880433e60c88  RCX: 0000000000009e6c
     RDX: 0000000000000246  RSI: ffff8806012ca298  RDI: ffff880433e60c88
     RBP: ffff8802338c3d30   R8: ffff8806012ca2e8   R9: 00000000ffffffff
     R10: 0000000000000001  R11: 0000000000000000  R12: ffff8804346b2020
     R13: ffff88032a3e7540  R14: ffff8804346b26e0  R15: 0000000000000246
     ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0000
  #6 [ffff8802338c3d38] ipoib_cm_tx_handler at ffffffffa066fe0a [ib_ipoib]
  #7 [ffff8802338c3d98] cm_process_work at ffffffffa05149a7 [ib_cm]
  #8 [ffff8802338c3de8] cm_work_handler at ffffffffa05161aa [ib_cm]
  #9 [ffff8802338c3e38] worker_thread at ffffffff81090e10
 #10 [ffff8802338c3ee8] kthread at ffffffff81096c66
 #11 [ffff8802338c3f48] kernel_thread at ffffffff8100c0ca

We move the list_del() into ipoib_neigh_free(), so that deletion happens
only once, after the entry has been successfully removed from the lookup
table.  This same behavior is already used in ipoib_del_neighs_by_gid()
and __ipoib_reap_neigh().

Signed-off-by: Jim Foraker 
Reviewed-by: Or Gerlitz 
Reviewed-by: Jack Wang 
Reviewed-by: Shlomo Pongratz 
Signed-off-by: Roland Dreier

IPoIB: Fix pkey change flow for virtualization environments

2013-07-31T21:23:44+00:00

IPoIB's required behaviour w.r.t to the pkey used by the device is the following:

- For "parent" interfaces (e.g ib0, ib1, etc) who are created
  automatically as a result of hot-plug events from the IB core, the
  driver needs to take whatever pkey vlaue it finds in index 0, and
  stick to that index.

- For child interfaces (e.g ib0.8001, etc) created by admin directive,
  the driver needs to use and stick to the value provided during its
  creation.

In SR-IOV environment its possible for the VF probe to take place
before the cloud management software provisions the suitable pkey for
the VF in the paravirtualed PKEY table index 0. When this is the case,
the VF IB stack will find in index 0 an invalide pkey, which is all
zeros.

Moreover, the cloud managment can assign the pkey value at index 0 at
any time of the guest life cycle.

The correct behavior for IPoIB to address these requirements for
parent interfaces is to use PKEY_CHANGE event as trigger to optionally
re-init the device pkey value and re-create all the relevant resources
accordingly, if the value of the pkey in index 0 has changed (from
invalid to valid or from valid value X to invalid value Y).

This patch enhances the heavy flushing code which is triggered by pkey
change event, to behave correctly for parent devices. For child
devices, the code remains the same, namely chases pkey value and not
index.

Signed-off-by: Erez Shitrit 
Signed-off-by: Or Gerlitz 
Signed-off-by: Roland Dreier

IPoIB: Make sure child devices use valid/proper pkeys

2013-07-31T21:23:40+00:00

Make sure that the IB invalid pkey (0x0000 or 0x8000) isn't used for
child devices.

Also, make sure to always set the full membership bit for the pkey of
devices created by rtnl link ops.

Signed-off-by: Or Gerlitz 
Signed-off-by: Roland Dreier

Merge tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband

2013-05-08T22:29:48+00:00

Pull InfiniBand/RDMA changes from Roland Dreier:
 - XRC transport fixes
 - Fix DHCP on IPoIB
 - mlx4 preparations for flow steering
 - iSER fixes
 - miscellaneous other fixes

* tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (23 commits)
  IB/iser: Add support for iser CM REQ additional info
  IB/iser: Return error to upper layers on EAGAIN registration failures
  IB/iser: Move informational messages from error to info level
  IB/iser: Add module version
  mlx4_core: Expose a few helpers to fill DMFS HW strucutures
  mlx4_core: Directly expose fields of DMFS HW rule control segment
  mlx4_core: Change a few DMFS fields names to match firmare spec
  mlx4: Match DMFS promiscuous field names to firmware spec
  mlx4_core: Move DMFS HW structs to common header file
  IB/mlx4: Set link type for RAW PACKET QPs in the QP context
  IB/mlx4: Disable VLAN stripping for RAW PACKET QPs
  mlx4_core: Reduce warning message for SRQ_LIMIT event to debug level
  RDMA/iwcm: Don't touch cmid after dropping reference
  IB/qib: Correct qib_verbs_register_sysfs() error handling
  IB/ipath: Correct ipath_verbs_register_sysfs() error handling
  RDMA/cxgb4: Fix SQ allocation when on-chip SQ is disabled
  SRPT: Fix odd use of WARN_ON()
  IPoIB: Fix ipoib_hard_header() return value
  RDMA: Rename random32() to prandom_u32()
  RDMA/cxgb3: Fix uninitialized variable
  ...

Merge branches 'cxgb4', 'ipoib', 'iser', 'misc', 'mlx4', 'qib' and 'srp' into for-next

2013-05-08T21:12:37+00:00

drivers/infiniband/hw: rename random32() to prandom_u32()

2013-05-08T01:38:27+00:00

Use preferable function name which implies using a pseudo-random number
generator.

Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

IPoIB: add support for TIPC protocol

2013-04-17T18:18:33+00:00

Support TIPC in the IPoIB driver. Since IPoIB now keeps track of its own
neighbour entries and doesn't require the packet to have a dst_entry
anymore, the only necessary changes are to:

- not drop multicast TIPC packets because of the unknown ethernet type
- handle unicast TIPC packets similar to IPv4/IPv6 unicast packets

in ipoib_start_xmit().

An alternative would be to remove all ethertype limitations since they're
not necessary anymore, all TIPC needs to know about is ARP and RARP since
it wants to always perform "path find", even if a path is already known.

Signed-off-by: Patrick McHardy 
Signed-off-by: David S. Miller

IPoIB: Fix ipoib_hard_header() return value

2013-04-17T05:57:09+00:00

If you have a patched up dhcp server (and dhclient), they will use
AF_PACKET/SOCK_DGRAM pair to send dhcp packets over IPoIB.

However, when testing an upstream kernel, this has been broken for a
very long time (I tested 2.6.34, 2.6.38, 3.0, 3.1, 3.8, HEAD).

It turns out that the hard_header routine in ipoib is not following
the API and is returning 0 even when it pushed data onto the skb.
This then causes af_packet.c to overwrite the header just pushed with
data from user space.

Fixing this gets DHCP working on IPoIB.

Signed-off-by: Doug Ledford 
Signed-off-by: Roland Dreier

RDMA: Rename random32() to prandom_u32()

2013-04-17T05:47:05+00:00

Use more preferable function name which implies using a pseudo-random
number generator.

Signed-off-by: Akinobu Mita 
Reviewed-by: Steve Wise 
Cc: Roland Dreier 
Cc: Sean Hefty 
Cc: Hal Rosenstock 
Cc: Steve Wise 
Cc: linux-rdma@vger.kernel.org
Reviewed-by: Steve Wise 
Signed-off-by: Roland Dreier

IPoIB: Fix send lockup due to missed TX completion

2013-03-23T01:01:04+00:00

Commit f0dc117abdfa ("IPoIB: Fix TX queue lockup with mixed UD/CM
traffic") attempts to solve an issue where unprocessed UD send
completions can deadlock the netdev.

The patch doesn't fully resolve the issue because if more than half
the tx_outstanding's were UD and all of the destinations are RC
reachable, arming the CQ doesn't solve the issue.

This patch uses the IB_CQ_REPORT_MISSED_EVENTS on the
ib_req_notify_cq().  If the rc is above 0, the UD send cq completion
callback is called directly to re-arm the send completion timer.

This issue is seen in very large parallel filesystem deployments
and the patch has been shown to correct the issue.

Cc: 
Reviewed-by: Dean Luick 
Signed-off-by: Mike Marciniszyn 
Signed-off-by: Roland Dreier