happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

address thread safety issue

This commit is contained in:
Xuan Son Nguyen 2025-11-20 18:38:02 +01:00
parent 5369aaa1d6
commit 6929c9f43d
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
tools/server/server-models.cpp
View File

@ -286,6 +286,7 @@ void server_models::load(const std::string & name) {
// we reach here when the child process exits
int exit_code = 0;
subprocess_join(child_proc.get(), &exit_code);
subprocess_destroy(child_proc.get());
// update PID and status
{
std::lock_guard<std::mutex> lk(mutex);
@ -315,7 +316,7 @@ void server_models::unload(const std::string & name) {
if (it != mapping.end()) {
if (it->second.meta.is_active()) {
SRV_INF("unloading model instance name=%s\n", name.c_str());
subprocess_destroy(it->second.subproc.get());
subprocess_terminate(it->second.subproc.get());
// status change will be handled by the managing thread
} else {
SRV_WRN("model instance name=%s is not loaded\n", name.c_str());
@ -330,7 +331,7 @@ void server_models::unload_all() {
for (auto & [name, inst] : mapping) {
if (inst.meta.is_active()) {
SRV_INF("unloading model instance name=%s\n", name.c_str());
subprocess_destroy(inst.subproc.get());
subprocess_terminate(inst.subproc.get());
// status change will be handled by the managing thread
}
// moving the thread to join list to avoid deadlock