ibv_init_wq and ibv_cleanup_wq are made visible with the
default_symver which puts them in the IBVERBS_1.1 ABI. Don't
try to expose them as IBVERBS_PRIVATE_14 symbols as GNU ld
complains with --no-undefined-symbol.
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D45090
create_qp_handle_resp_common_cleanup should be void
__ibv_cleanup_wq should use wq->cond for cond destroy
both issues were overlooked in:
a687910 ('Cleanup pthread locks in ofed RDMA verbs')
Signed-off-by: Bartosz Sobczak <bartosz.sobczak@intel.com>
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: sean.lim@dell.com, vangyzen@, erj@
MFC after: 1 day
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D43491
On FreeBSD, pthread mutex, cond, and spinlocks allocate memory. On
Linux-based systems, these calls do not allocate memory. So there was a
safe assumption that the ofed RDMA verb calls do not need to explicitly
destroy the pthread locks. This assumption is false on FreeBSD. So let
us rearrange the code to cleanup the pthread locks.
Reviewed by: delphij
MFC after: 2 weeks
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D41105
Removing artificial completion generator as there had been no indication
of the code being required for E810 cards. Further more it was found
that the code may have unpleasant side effects on user experience when
using ucmatose tool.
Signed-off-by: Bartosz Sobczak <bartosz.sobczak@intel.com>
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: erj@
MFC after: 1 day
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D41593
Update Intel irdma driver to version 1.2.17-k
Notable changes:
- pf-reset handling improvements, including fixes in communication with if_ice(4)
- avoid racing when handling various events
- adding sw stats sysctls
- hand over pe_criterr handling from ice(4) to irdma(4)
- debug prints adjustments
- fix crash after changes in irdma_add_mqh_ifa_cb
Signed-off-by: Bartosz Sobczak <bartosz.sobczak@intel.com>
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: erj@
MFC after: 1 month
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D41425
The function ibv_query_device_ex is static inline, it is not exported
from the dso. With lld 16, which is much more picky about versioning and
undefined symbols, this becomes an error.
The ibv_register_driver driver symbol is explicitly versioned in
sources, it is non-existent in un-versioned object files.
Sponsored by: NVidia networking
MFC after: 1 week
Summary of changes:
- postpone mtu size assignment during load to avoid race condition
- refactor some of the debug prints
- add request reset handler
- refactor flush scheduler to increase efficiency and avoid racing
- put correct vlan_tag for UD traffic with PFC
- suspend QP before going to ERROR state to avoid CQP timout
- fix arithmetic error on irdma_debug_bugf
- allow debug flag to be settable during driver load
- introduce meaningful default values for DCQCN algorithm
- interrupt naming convention improvements
- skip unsignaled completions in poll_cmpl
Signed-off-by: Bartosz Sobczak bartosz.sobczak@intel.com
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: hselasky@
MFC after: 1 week
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D39173
Current content of the irdma(4) driver consists only of code that is
compilable on current 14-CURRENT branch which makes it impossible to
merge into stable/13 branch because of missing dependencies in the ofed
tree.
This patch adds missing code that allows for merging into stable branch.
Once it is there, code relating only to version 14 or higher should be
removed.
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: erj@
MFC after: 1 day
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D38170
This is to upgrade current irdma driver version (in support of RDMA on
Intel(R) Ethernet Controller E810) to 1.1.5-k
change summary:
- refactor defines for hardware registers
- rereg_mr verb added in libirdma
- fix print warning during compilation
- rt_ros2priority macro fix
- irdma.4 validated with mandoc
- fixing nd6_resolve usage
- added libirdma_query_device
- sysctl for irdma version
- aeq_alloc_db fix
- dwork_flush protected with qp refcount
- PFC fixes
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: erj@
Sponsored by: Intel Corporation
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D36944
Summary:
The current OFED code allows binding server to IPv6 address. It was added back in 91fc39561d
Sponsored by: Intel Corporation
MFC after: 3 days
Reviewers: hselasky
Subscribers: imp
Differential Revision: https://reviews.freebsd.org/D37196
This is an initial commit for RDMA FreeBSD driver for Intel(R) Ethernet
Controller E810, called irdma. Supporting both RoCEv2 and iWARP
protocols in per-PF manner, RoCEv2 being the default.
Testing has been done using krping tool, perftest, ucmatose, rping,
ud_pingpong, rc_pingpong and others.
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: #manpages (pauamma_gundo.com) [documentation]
MFC after: 1 week
Relnotes: yes
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D34690
The introduction of <sched.h> improved compatibility with some 3rd
party software, but caused the configure scripts of some ports to
assume that they were run in a GLIBC compatible environment.
Parts of sched.h were made conditional on -D_WITH_CPU_SET_T being
added to ports, but there still were compatibility issues due to
invalid assumptions made in autoconfigure scripts.
The differences between the FreeBSD version of macros like CPU_AND,
CPU_OR, etc. and the GLIBC versions was in the number of arguments:
FreeBSD used a 2-address scheme (one source argument is also used as
the destination of the operation), while GLIBC uses a 3-adderess
scheme (2 source operands and a separately passed destination).
The GLIBC scheme provides a super-set of the functionality of the
FreeBSD macros, since it does not prevent passing the same variable
as source and destination arguments. In code that wanted to preserve
both source arguments, the FreeBSD macros required a temporary copy of
one of the source arguments.
This patch set allows to unconditionally provide functions and macros
expected by 3rd party software written for GLIBC based systems, but
breaks builds of externally maintained sources that use any of the
following macros: CPU_AND, CPU_ANDNOT, CPU_OR, CPU_XOR.
One contributed driver (contrib/ofed/libmlx5) has been patched to
support both the old and the new CPU_OR signatures. If this commit
is merged to -STABLE, the version test will have to be extended to
cover more ranges.
Ports that have added -D_WITH_CPU_SET_T to build on -CURRENT do
no longer require that option.
The FreeBSD version has been bumped to 1400046 to reflect this
incompatible change.
Reviewed by: kib
MFC after: 2 weeks
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D33451
Fixes the following compile warning:
implicit declaration of function 'sysctlbyname' is invalid in C99
[-Wimplicit-function-declaration]
Found by: J87
Differential Revision: https://reviews.freebsd.org/D30484
MFC after: 1 week
Sponsored by: Mellanox Technologies // NVIDIA Networking
read_ibdiag_config NULL deref
read_ibdiag_config mem leak
ib_mad_inv_field_str Missing comma in a string array initialization
print_node_header NULL deref
diff_node_ports copy-paste error
ibportstate.c main() missing break in switch
set_thresholds NULL ptr deref
dump_unicast_tables leaks mapnd
umad_cm_attr_str dead code
__ibv_close_device close(-1)
check return value of listen()
mlx5 bitmap.h - bad bit shift - UB
get_dst_addr check return value of inet_pton
osm_perfmgr_init check return value of cl_spinlock_init
osm_port_new memory leak on error path
sa_mad_ctrl_rcv_callback missing break in switch case
I did not include CID numbers because these were found by an internal
run at Isilon.
Reviewed by: cem kib
MFC after: 2 weeks
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D25676
In addition, add "ConnectX family mlx5Gen Virtual Function" device ID.
Every new HCA VF will be identified with this device ID.
Submitted by: slavash@
MFC after: 3 days
Sponsored by: Mellanox Technologies
If user input a string larger than the length of buffer, the stack
memory will be corrupted.
Approved by: hselasky (mentor)
MFC after: 1 week
Sponsored by: Mellanox Technologies
the hard-coded string "not available" to ensure reproducible builds.
Discussed with: emaste
Approved by: re (rgrimes)
Sponsored by: The FreeBSD Foundation
libibverbs example utilities.
This allows connecting to both IPv4 and IPv6 and reverts
some FreeBSD only patches.
MFC after: 1 week
Sponsored by: Mellanox Technologies
Add the ability to to parse sysfs paths to sysctl nodes by replacing '/' with '.'
Submitted by: slavash@
MFC after: 1 week
Sponsored by: Mellanox Technologies
When running mckey, errors may happen in the init/connect stage.
When leaving multicast groups, we override this value.
Fix that by saving the return value from rdma_leave_multicast to different
parameter, and only in case of failure in rdma_leave_multicast override it.
MFC after: 1 week
Approved by: hselasky (mentor), kib (mentor)
Sponsored by: Mellanox Technologies
Aligns the build with the FreeBSD traditional approach to not build in
contrib/, and to track inter-dependencies between libraries.
With help from: bdrewery
Reviewed by: bdrewery, hselasky
Sponsored by: Mellanox Technologies
MFC after: 2 weeks
Differential revision: https://reviews.freebsd.org/D15648
This can cause ibpanic in ibstat when width is not set properly
as can occur when Ethernet port is connected to InfiniBand fabric.
ibpanic: [8167] main: stat of IB device 'mlx5_0' failed: m
With this change, Rate is displayed as 0 with ibstat for
this scenario.
MFC after: 3 days
Approved by: hselasky (mentor), kib (mentor)
Sponsored by: Mellanox Technologies
I used the strongest barriers available on the architectures, so if
the future analysis show that it is excessive, the barriers could be
relaxed. Still, it is unlikely that it is meaningful to run IB on 32bit
ARM or current MIPS machines, so the change is to make WITH_OFED to pass
tinderbox.
Sponsored by: Mellanox Technologies
Reviewed by: hselasky
Differential revision: https://reviews.freebsd.org/D13329
ld.bfd accepts multiple listing of the same symbol in the version script.
lld is stricter and errors out. Since arm64 and sometimes amd64 use lld,
we should correct this cosmetic issue.
Sponsored by: Mellanox Technologies
Reviewed by: hselasky
Differential revision: https://reviews.freebsd.org/D13329
This should fix ofed/libibnetdisc compilation with C-compilers
different from clang and GCC v4.2.1.
Submitted by: kib
Sponsored by: Mellanox Technologies
the OFED buildworld target, WITH_OFED=YES, when the include files are not
already installed locally, but only in the temporary object directory.
Found by: kib
Sponsored by: Mellanox Technologies
Don't install the ib_user_mad.h header file into user-space,
because it conflicts with umad.h from libibumad.h when building
libsysdecode.
Sponsored by: Mellanox Technologies
