Skip to content

fix(ScriptRunner): fix deadlock#1209

Merged
praydog merged 1 commit intopraydog:masterfrom
mrbeardad:master
Mar 4, 2025
Merged

fix(ScriptRunner): fix deadlock#1209
praydog merged 1 commit intopraydog:masterfrom
mrbeardad:master

Conversation

@mrbeardad
Copy link
Copy Markdown
Contributor

@mrbeardad mrbeardad commented Mar 4, 2025 •
edited
Loading

Fix deadlock problem when play MHWilds on a high performance machine

@HookedBehemoth
Copy link
Copy Markdown

HookedBehemoth commented Mar 4, 2025

Couldn't both be passed into the same scoped_lock ctor? https://en.cppreference.com/w/cpp/thread/scoped_lock/scoped_lock
The docs explicitly mention that it avoids deadlocks https://en.cppreference.com/w/cpp/thread/lock
and neither is released early.

@praydog praydog merged commit 92b9376 into praydog:master Mar 4, 2025
13 checks passed
@mrbeardad
Copy link
Copy Markdown
Contributor Author

mrbeardad commented Mar 5, 2025

@praydog I just reproduced the deadlock problem again, so I think m_access_mutex is not the essential reason. After further research, I've found the real problem.

In imgui thread, m_imgui_mtx first locked here

REFramework/src/REFramework.cpp

Lines 769 to 776 in 1b8da08

void REFramework::run_imgui_frame(bool from_present) {
std::scoped_lock _{ m_imgui_mtx };
m_has_frame = false;
if (!m_initialized) {
return;
}

and then get_hook_monitor_mutex called here

REFramework/src/mods/ScriptRunner.cpp

Lines 1170 to 1174 in 1b8da08

void ScriptRunner::reset_scripts() {
std::scoped_lock __{ g_framework->get_hook_monitor_mutex() }; // Stops D3D monitor from attempting to re-hook during long script startups
std::scoped_lock _{ m_access_mutex };
{

But in d3d12 threads, get_hook_monitor_mutex is first called here

REFramework/src/D3D12Hook.cpp

Lines 78 to 85 in 1b8da08

std::scoped_lock _{g_framework->get_hook_monitor_mutex()};
bool hook_was_nullptr = g_d3d12_hook == nullptr;
if (g_d3d12_hook != nullptr && g_framework->get_d3d12_hook() != nullptr) {
g_framework->on_reset(); // Needed to prevent a crash due to resources hanging around
g_d3d12_hook->unhook(); // Removes all vtable hooks
}

and then m_imgui_mtx locked here

REFramework/src/REFramework.cpp

Lines 1087 to 1091 in 1b8da08

void REFramework::on_reset() {
std::scoped_lock _{ m_imgui_mtx };
spdlog::info("Reset!");

Unfortunately, them are not in the same function, is there a easy way to solve it? BTW, this pr solve my problem most of the time, I don't know the reason

@praydog
Copy link
Copy Markdown
Owner

praydog commented Mar 5, 2025

Should be fixed in 8358c55

@BoltreaverKU27 BoltreaverKU27 mentioned this pull request Dec 30, 2025
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@mrbeardad @HookedBehemoth @praydog