[Profiler plugin] Fix segfault issue with profiler plugin (#1973)

* Fix profiler plugin segfault by correctly setting p2p->func

* Look for librccl-profiler.so instead of libnccl-profiler.so

Signed-off-by: rahulvaidya20 <ravaidya@amd.com>

---------

Signed-off-by: rahulvaidya20 <ravaidya@amd.com>
Co-authored-by: Yongjie Qiu <Yongjie.Qiu@amd.com>

ext-profiler/README.md

@@ -9,7 +9,7 @@ interface. Any NCCL user can write profiler plugins to extract performance data
 use it for debugging and analysis.
 
 Similarly to other plugins (e.g., network plugin), the profiler plugins come as a shared library
-called `libnccl-profiler.so`. That shared library contains one or more implementations of the
+called `lirccl-profiler.so`. That shared library contains one or more implementations of the
 NCCL PROFILER API, in the form of versioned structs, filled with pointers to all required
 functions.
 
@@ -17,15 +17,15 @@ functions.
 
 ## Plugin name and supporting multiple profiler plugins
 
-When NCCL is initialized, it will look for a `libnccl-profiler.so` library and dynamically load
+When NCCL is initialized, it will look for a `librccl-profiler.so` library and dynamically load
 it, then look for symbols inside the library.
 
 The `NCCL_PROFILER_PLUGIN` environment variable allows multiple plugins to coexist. If set, NCCL
-will look for a library with a name of `libnccl-profiler-${NCCL_PROFILER_PLUGIN}.so`. It is therefore
-advised to name the library following that pattern, with a symlink pointing `libnccl-profiler.so`
-to `libnccl-profiler-${NCCL_PROFILER_PLUGIN}.so`. That way, if there are multiple plugins in the
+will look for a library with a name of `librccl-profiler-${NCCL_PROFILER_PLUGIN}.so`. It is therefore
+advised to name the library following that pattern, with a symlink pointing `librccl-profiler.so`
+to `librccl-profiler-${NCCL_PROFILER_PLUGIN}.so`. That way, if there are multiple plugins in the
 path, setting `NCCL_PROFILER_PLUGIN` will allow users to select the right plugin. Alternatively,
-the user can also set `NCCL_PROFILER_PLUGIN` to the pathname of the `libnccl-profiler.so` library.
+the user can also set `NCCL_PROFILER_PLUGIN` to the pathname of the `librccl-profiler.so` library.
 
 ## Struct versioning
 

ext-profiler/example/Makefile

@@ -11,12 +11,12 @@ NCCLDIR  := $(BUILDDIR)
 
 SRC_FILES := $(wildcard *.c)
 
-build: ${BUILDDIR}/libnccl-profiler-example.so
+build: ${BUILDDIR}/librccl-profiler.so
 
-${BUILDDIR}/libnccl-profiler-example.so: ${SRC_FILES}
+${BUILDDIR}/librccl-profiler.so: ${SRC_FILES}
 	@printf "Compiling  %-35s > %s\n" $< $@
 	@mkdir -p ${BUILDDIR}
 	$(CC) -Inccl -fPIC -shared -o $@ $^
 
 clean:
-	rm -f ${BUILDDIR}/libnccl-profiler-example.so
+	rm -f ${BUILDDIR}/librccl-profiler.so

src/enqueue.cc

@@ -2574,6 +2574,7 @@ static ncclResult_t taskAppend(struct ncclComm* comm, struct ncclInfo* info) {
     // Must be in thread local group before tasks can be alloc'd in `comm->memScoped`.
     ncclGroupCommJoin(info->comm, ncclGroupTaskTypeCollective);
     struct ncclTaskP2p* p2p = ncclMemoryPoolAlloc<struct ncclTaskP2p>(&comm->memPool_ncclTaskP2p, &comm->memPermanent);
+    p2p->func = info->coll;
     p2p->buff = (void*)info->recvbuff;
     p2p->count = info->count;
     p2p->datatype = info->datatype;

src/plugin/plugin_open.cc

@@ -18,7 +18,7 @@
 static char* libNames[NUM_LIBS];
 static void *libHandles[NUM_LIBS];
 static const char *pluginNames[NUM_LIBS] = { "NET", "TUNER", "PROFILER" };
-static const char *pluginPrefix[NUM_LIBS] = { "librccl-net", "libnccl-tuner", "libnccl-profiler" };
+static const char *pluginPrefix[NUM_LIBS] = { "librccl-net", "libnccl-tuner", "librccl-profiler" };
 static const char *pluginFallback[NUM_LIBS] = { "", "Using internal tuner plugin.", "" };
 static unsigned long subsys[NUM_LIBS] = { NCCL_INIT|NCCL_NET, NCCL_INIT|NCCL_TUNING, NCCL_INIT };