Update release notes in accordance with changes since OpenSM 3.3.22

Signed-off-by: Tamir Ronen <tamirr@mellanox.com>

Signed-off-by: Tamir Ronen <tamirr@mellanox.com>

in accordance with changes since OpenSM 3.3.22

Signed-off-by: Tamir Ronen <tamirr@mellanox.com>

Signed-off-by: Benjamin Drung <benjamin.drung@cloud.ionos.com>
Signed-off-by: Tamir Ronen <tamirr@mellanox.com>

Add new function: osm_register_external_routing_engine for registering
an opensm plugin as a new routing engine

Predefined routing engines (minhop, updn, ...) are registered on startup
by using new function: register_builtin_routing_engine

Fix typos: routine engine -> routing engine

osmroutingplugin: example of routing engine plugin
In order to activate the example osmroutingplugin as routing engine
The following modifications to opensm.conf are needed:
 - event_plugin_name  osmroutingplugin
 - routing_engine     routing_engine_plugin

Signed-off-by: Cyrille Verrier <cyrille.verrier@fabriscale.com>
Signed-off-by: Tamir Ronen <tamirr@mellanox.com>

For many of the IBA structures, the packed attribute has no effect.

List of impacted struct and checks were done this way:

Before applying this patch
- Generate a single file with all the IBA structs
(echo -e "#include <iba/ib_types.h>\n#include <stdio.h>\n\n\n"; for struct in $(git grep '^} PACK_SUFFIX' -- include/iba/ib_types.h | awk '{ print $NF }' | sed -e 's/;//'); do echo -e "$struct a_$struct;"; done) > ib_sizes.c
- Compile for both 32 and 64b
gcc -o ib_sizes.64.o -g3 -c ib_sizes.c -I./include/
gcc -o ib_sizes.32.o -g3 -m32 -c ib_sizes.c -I./include/
- Generate structure data using pahole
pahole ib_sizes.64.o > sizes.64.org
pahole ib_sizes.32.o > sizes.32.org

After applying this patch:
- Compile for both 32 and 64b
gcc -o ib_sizes.64.o -g3 -c ib_sizes.c -I./include/
gcc -o ib_sizes.32.o -g3 -m32 -c ib_sizes.c -I./include/
- Generate structure data using pahole
pahole ib_sizes.64.o > sizes.64.new
pahole ib_sizes.32.o > sizes.32.new

Diff pahole results:
diff sizes.64.org sizes.64.new
diff sizes.32.org sizes.32.new

As the patch remove effect-less attribute, no diff shows up

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

as having in license breaks rpm building on some distros

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Also, moved 'COPYING' file into the license tag

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Use rdma-core script to check patches in PR

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Builds with multiple compilers and sparse.
Generates a publishable tarball

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Clears defined but unused warning from clang

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

osm_ucast_ftree.c:3861:52: warning: for loop has empty body [-Wempty-body]
                     && (!p_hca->up_port_groups[i]->is_cn); i++) ;
                                                                 ^
osm_ucast_ftree.c:3861:52: note: put the semicolon on a separate line to silence this warning

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Fixes this warning:
osm_opensm.c:543:31: warning: expression which evaluates to zero treated as a null pointer constant of type 'struct osm_routing_engine *' [-Wnon-literal-null-conversion]
        p_osm->routing_engine_used = OSM_ROUTING_ENGINE_TYPE_NONE;

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Update release notes in accordance with changes since OpenSM 3.3.21

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

in accordance with changes since OpenSM 3.3.21

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Fixes GCC warning:
osm_vendor_ibumad.c: In function ‘osm_vendor_open_port’:
osm_vendor_ibumad.c:745:56: error: passing argument 2 of ‘umad_get_ca_portguids’ from incompatible pointer type [-Werror=incompatible-pointer-types]
   if ((r = umad_get_ca_portguids(p_vend->ca_names[ca], portguids,
                                                        ^~~~~~~~~

Second argument to umad_get_ca_portguids was changed from uint64_t * to __be64 * in rdma-core by:
96db7b90 (Bart Van Assche 2017-03-10 14:52:27 -0800 178) int umad_get_ca_portguids(const char *ca_name, __be64 *portguids, int max);

Note that this change causes similar warning when OpenSM is built
with older libibumad rather than the other way around as it was
prior to this patch.
osm_vendor_ibumad.c: In function 'osm_vendor_open_port':
osm_vendor_ibumad.c:747: warning: passing argument 2 of 'umad_get_ca_portguids' from incompatible pointer type
./../../libibumad/include/infiniband/umad.h:166: note: expected 'uint64_t *' but argument is of type '__be64 *'

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Add void parameter to fix sparse warning:
main.c:73:17: warning: non-ANSI function declaration of function 'show_usage'

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Add static to local functions and variables
Also drop redundant = 0 for global values

Found by sparse

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Found by sparse.

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Fixes this GCC warning:
main.c: In function ‘main’:
main.c:578:21: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
  while (next_option != -1);
                     ^~

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Change loop indexes to unsigned where required to avoid
 these type of GCC warnings:
osm_congestion_control.c: In function ‘cc_setup_mad_data’:
osm_congestion_control.c:193:16: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
  for (i = 0; i < p_cc->cc_tbl_mads; i++) {
                ^

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Fixes the GCC warning:
osm_opensm.c:84:1: error: ‘static’ is not at beginning of declaration [-Werror=old-style-declaration]
 const static struct routing_engine_module routing_modules[] = {
 ^~~~~

Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Pointed-out-by: Benjamin Drung <benjamin.drung@cloud.ionos.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

The shared library routine complib_init() in complib calls
the C library exit() function.

In the case of an error, the library should instead return an
appropriate error code to the calling program which can then
determine how to handle the error, including performing any
required clean-up.

Add new complib_init_v2 function that does this (returns
status) and change all internal uses to original complib_init
function to handle status and take appropriate action.

Pointed-out-by: Benjamin Drung <benjamin.drung@cloud.ionos.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

github releases use lightweight rather than annotated tags
and git cat-file tag does not work on a lightweight tag.

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Issue was found by Coverity.

3. Defect type: NULL_RETURNS
8. opensm-3.3.21/opensm/osm_trap_rcv.c:254: returned_null: "osm_get_port_by_guid" returns null (checked 46 out of 49 times).
9. opensm-3.3.21/opensm/osm_trap_rcv.c:254: var_assigned: Assigning: "p_port" = null return value from "osm_get_port_by_guid".
10. opensm-3.3.21/opensm/osm_trap_rcv.c:256: dereference: Dereferencing a null pointer "p_port".
11. opensm-3.3.21/opensm/osm_congestion_control.c:678: example_assign: Example 1: Assigning: "p_port" = return value from "osm_get_port_by_guid(p_cc->subn, port_guid)".
12. opensm-3.3.21/opensm/osm_congestion_control.c:679: example_checked: Example 1 (cont.): "p_port" has its value checked in "p_port".
13. opensm-3.3.21/opensm/osm_drop_mgr.c:91: example_assign: Example 2: Assigning: "p_remote_port" = return value from "osm_get_port_by_guid(sm->p_subn, p_remote_physp->port_guid)".
14. opensm-3.3.21/opensm/osm_drop_mgr.c:94: example_checked: Example 2 (cont.): "p_remote_port" has its value checked in "p_remote_port".
15. opensm-3.3.21/opensm/osm_guid_info_rcv.c:105: example_assign: Example 3: Assigning: "p_port" = return value from "osm_get_port_by_guid(sm->p_subn, port_guid)".
16. opensm-3.3.21/opensm/osm_guid_info_rcv.c:106: example_checked: Example 3 (cont.): "p_port" has its value checked in "p_port".
17. opensm-3.3.21/opensm/osm_inform.c:434: example_assign: Example 4: Assigning: "p_src_port" = return value from "osm_get_port_by_guid(p_subn, source_gid.unicast.interface_id)".
18. opensm-3.3.21/opensm/osm_inform.c:436: example_checked: Example 4 (cont.): "p_src_port" has its value checked in "p_src_port".
19. opensm-3.3.21/opensm/osm_lid_mgr.c:1112: example_assign: Example 5: Assigning: "p_port" = return value from "osm_get_port_by_guid(p_mgr->p_subn, p_mgr->p_subn->sm_port_guid)".
20. opensm-3.3.21/opensm/osm_lid_mgr.c:1114: example_checked: Example 5 (cont.): "p_port" has its value checked in "p_port".
//  1112|   	p_port = osm_get_port_by_guid(p_mgr->p_subn,
//  1113|   				      p_mgr->p_subn->sm_port_guid);
//  1114|-> 	if (!p_port) {
//  1115|   		OSM_LOG(p_mgr->p_log, OSM_LOG_ERROR, "ERR 0308: "
//  1116|   			"Can't acquire SM's port object, GUID 0x%016" PRIx64

4. Defect type: NULL_RETURNS
3. opensm-3.3.21/opensm/osm_link_mgr.c:126: returned_null: "osm_get_port_by_guid" returns null (checked 46 out of 49 times).
4. opensm-3.3.21/opensm/osm_link_mgr.c:126: var_assigned: Assigning: "p_port" = null return value from "osm_get_port_by_guid".
5. opensm-3.3.21/opensm/osm_link_mgr.c:128: dereference: Dereferencing a null pointer "p_port".
6. opensm-3.3.21/opensm/osm_congestion_control.c:678: example_assign: Example 1: Assigning: "p_port" = return value from "osm_get_port_by_guid(p_cc->subn, port_guid)".
7. opensm-3.3.21/opensm/osm_congestion_control.c:679: example_checked: Example 1 (cont.): "p_port" has its value checked in "p_port".
8. opensm-3.3.21/opensm/osm_drop_mgr.c:91: example_assign: Example 2: Assigning: "p_remote_port" = return value from "osm_get_port_by_guid(sm->p_subn, p_remote_physp->port_guid)".
9. opensm-3.3.21/opensm/osm_drop_mgr.c:94: example_checked: Example 2 (cont.): "p_remote_port" has its value checked in "p_remote_port".
10. opensm-3.3.21/opensm/osm_guid_info_rcv.c:105: example_assign: Example 3: Assigning: "p_port" = return value from "osm_get_port_by_guid(sm->p_subn, port_guid)".
11. opensm-3.3.21/opensm/osm_guid_info_rcv.c:106: example_checked: Example 3 (cont.): "p_port" has its value checked in "p_port".
12. opensm-3.3.21/opensm/osm_inform.c:434: example_assign: Example 4: Assigning: "p_src_port" = return value from "osm_get_port_by_guid(p_subn, source_gid.unicast.interface_id)".
13. opensm-3.3.21/opensm/osm_inform.c:436: example_checked: Example 4 (cont.): "p_src_port" has its value checked in "p_src_port".
14. opensm-3.3.21/opensm/osm_lid_mgr.c:1112: example_assign: Example 5: Assigning: "p_port" = return value from "osm_get_port_by_guid(p_mgr->p_subn, p_mgr->p_subn->sm_port_guid)".
15. opensm-3.3.21/opensm/osm_lid_mgr.c:1114: example_checked: Example 5 (cont.): "p_port" has its value checked in "p_port".
//  1112|   	p_port = osm_get_port_by_guid(p_mgr->p_subn,
//  1113|   				      p_mgr->p_subn->sm_port_guid);
//  1114|-> 	if (!p_port) {
//  1115|   		OSM_LOG(p_mgr->p_log, OSM_LOG_ERROR, "ERR 0308: "
//  1116|   			"Can't acquire SM's port object, GUID 0x%016" PRIx64

Pointed-out-by: Honggang Li <honli@redhat.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Issue was found by Coverity.

opensm-3.3.21/opensm/main.c:1165: unsigned_compare: This less-than-zero comparison of an unsigned value is never true. "temp < 0U".
//  1163|   		case 15:
//  1164|   			temp = strtoul(optarg, NULL, 0);
//  1165|-> 			if (temp < 0 || temp >= IB_MAX_NUM_VLS) {
//  1166|   				fprintf(stderr,
//  1167|   					"ERROR: maximum #VLs for nue routing must be between 0

Signed-off-by: Honggang Li <honli@redhat.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Issue was found by Coverity.

When (lsea == IB_LINK_SPEED_EXT_ACTIVE_50 == 4), lsea_str_fixed_width[lsea]
will overrun array.

Defect type: OVERRUN
4. opensm-3.3.21/opensm/osm_helper.c:3225: cond_at_least: Checking "lsea == 0" implies that "lsea" is at least 1 on the false branch.
6. opensm-3.3.21/opensm/osm_helper.c:3231: cond_between: Checking "lsea > 4" implies that "lsea" is between 1 and 4 (inclusive) on the false branch.
7. opensm-3.3.21/opensm/osm_helper.c:3233: overrun-local: Overrunning array "lsea_str_fixed_width" of 4 8-byte elements at element index 4 (byte offset 32) using index "lsea" (which evaluates to 4).
//  3231|   	if (lsea > IB_LINK_SPEED_EXT_ACTIVE_50)
//  3232|   		return lsa_str_fixed_width[3];
//  3233|-> 	return lsea_str_fixed_width[lsea];
//  3234|   }
//  3235|

Signed-off-by: Honggang Li <honli@redhat.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

The original patch was written by Doug Ledford.

For RHEL users who need to run multiple instances of opensm over
multiple InfiniBand ports, most of time, they need to pass five options
to all instances of opensm.

1) guid
2) subnet_prefix
3) partition_config_file
4) log_file
5) dump_files_dir

1), 3) and 4) are available in upstream. But 2) and 5) are not.
We need create separate configuration files for opensm instances
to apply 2) and 5), when those options are unavailable.

We add '--subnet_prefix' option because:
openmpi will refuse to run if you have two physically separate subnets
with the same subnet prefix.  The subnet prefix is supposed to be unique
on each subnet.  It's how software can tell for certain whether or not
to end points can communicate with each other.

We add '--dump_files_dir' option to keep dumps of different fabrics
into different directories.

Signed-off-by: Honggang Li <honli@redhat.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

osm_trap_rcv.c: Fix missing log message when 10 traps are received from the same source within time window

Signed-off-by: Aleksandr Minchiu <alexmi@mellanox.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

LC_ALL=C.UTF-8 MANROFFSEQ='' MANWIDTH=80 man --warnings -E UTF-8 -l -Tutf8 man/opensm.8 >/dev/null
<standard input>:584: warning [p 9, 1.0i]: cannot adjust line
<standard input>:748: warning [p 11, 6.0i]: cannot adjust line
<standard input>:749: warning [p 11, 6.3i]: cannot adjust line

Pointed-out-by: Benjamin Drung <benjamin.drung@cloud.ionos.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

in order to fix the unresolved symbols.

The shared objects have some unresolved symbols which the program which
links against it must resolve. This isn't a good idea because when you
introduce new dependencies the package previously linked against the old
version will break because it doesn't know about the new dependency.

dpkg-shlibdeps: warning: symbol osm_log_v2 used by debian/libosmvendor4/usr/lib/x86_64-linux-gnu/libosmvendor.so.4.0.3 found in none of the libraries
dpkg-shlibdeps: warning: symbol osm_mad_pool_get used by debian/libosmvendor4/usr/lib/x86_64-linux-gnu/libosmvendor.so.4.0.3 found in none of the libraries
dpkg-shlibdeps: warning: symbol osm_mad_pool_put used by debian/libosmvendor4/usr/lib/x86_64-linux-gnu/libosmvendor.so.4.0.3 found in none of the libraries
dpkg-shlibdeps: warning: symbol osm_log used by debian/libosmvendor4/usr/lib/x86_64-linux-gnu/libosmvendor.so.4.0.3 found in none of the libraries
dpkg-shlibdeps: warning: symbol osm_dump_smp_dr_path used by debian/libosmvendor4/usr/lib/x86_64-linux-gnu/libosmvendor.so.4.0.3 found in none of the libraries
dpkg-shlibdeps: warning: symbol cl_atomic_spinlock used by debian/libopensm8/usr/lib/x86_64-linux-gnu/libopensm.so.8.0.0 found in none of the libraries
dpkg-shlibdeps: warning: symbol osm_vendor_get used by debian/libopensm8/usr/lib/x86_64-linux-gnu/libopensm.so.8.0.0 found in none of the libraries
dpkg-shlibdeps: warning: symbol cl_spinlock_init used by debian/libopensm8/usr/lib/x86_64-linux-gnu/libopensm.so.8.0.0 found in none of the libraries
dpkg-shlibdeps: warning: symbol cl_spinlock_acquire used by debian/libopensm8/usr/lib/x86_64-linux-gnu/libopensm.so.8.0.0 found in none of the libraries
dpkg-shlibdeps: warning: symbol cl_get_time_stamp used by debian/libopensm8/usr/lib/x86_64-linux-gnu/libopensm.so.8.0.0 found in none of the libraries
dpkg-shlibdeps: warning: symbol cl_spinlock_release used by debian/libopensm8/usr/lib/x86_64-linux-gnu/libopensm.so.8.0.0 found in none of the libraries
dpkg-shlibdeps: warning: symbol osm_vendor_put used by debian/libopensm8/usr/lib/x86_64-linux-gnu/libopensm.so.8.0.0 found in none of the libraries

To fix this, restructure the internal libraries as follows:
libopensm currently has dependency on libosmvendor because of
osm_mad_pool.c. This module is moved to libosmvendor (even though
it's not really vendor but opensm) and then libopensm (now only
depending on libosmcomp) is split out from opensm and built after
libosmcomp but before libosmvendor. Only danger would be external
program which dynamically links with libopensm and needs some
osm_mad_pool routine and would now need to also link with
libosmvendor but such program is unlikely (and these libraries
are internal).

Pointed-out-by: Benjamin Drung <benjamin.drung@cloud.ionos.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>

lintian found more typos. Fix them all.

Signed-off-by: Benjamin Drung <benjamin.drung@cloud.ionos.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

There is a typo in OSM_DEFAULT_TRAP_SUPRESSION_TIMEOUT. Rename the
constant and keep the old name for backward compatibility.

Signed-off-by: Benjamin Drung <benjamin.drung@cloud.ionos.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Signed-off-by: Benjamin Drung <benjamin.drung@cloud.ionos.com>
Signed-off-by: Hal Rosenstock <hal@mellanox.com>

Pointed-out-by: Honggang Li <honli@redhat.com>

Signed-off-by: Hal Rosenstock <hal@mellanox.com>