issue: 4347777 Replace thread-local dummy lock with global one

[tomerdbz]

Description

This change replaces a thread-local dummy locker with a global one in ring_slave.cpp to fix a use-after-free issue.

What

Replace thread-local dummy locker with a global one in ring_slave.

Why ?

During XLIO shutdown, when one thread attempts to access a socket's locker that was created by a terminated thread, a use-after-free issue occurs because the thread-local dummy locker object is freed when its creator thread terminates. This can lead to segmentation faults during cleanup.

How ?

The solution is straightforward:

1. Replace static thread_local lock_dummy t_lock_dummy_ring with a global static lock_dummy g_lock_dummy_ring
2. Update get_new_lock() to use this global locker
3. The global locker is initialized with a name for better debugging: "global_dummy_ring"

This change ensures the dummy locker remains valid throughout the program's lifetime, regardless of thread termination status.

Performance considerations

Pros of global implementation:

1. The dummy lock is accessed frequently in the data path
2. Each access to a thread-local variable requires the more complex memory access pattern
3. This compounds the overhead across many accesses

The global was aligned to be in a different cache line to prevent false-sharing - thus not introducing perf hit,

Tests

added a system-test (gtest) that discovered same issue occured on src/core/sock/sockinfo_tcp.cpp.
Applied the same fix there as well for the test to pass.

Change type

What kind of change does this PR introduce?

• Bugfix
• Feature
• Code style update
• Refactoring (no functional changes, no api changes)
• Build related changes
• CI related changes
• Documentation content changes
• Tests
• Other

Check list

• Code follows the style de facto guidelines of this project
• Comments have been inserted in hard to understand places
• Documentation has been updated (if necessary)
• Test has been added (if possible)

[tomerdbz]

bot:retest

[tomerdbz]

/review

[pr-review-bot-app]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Possible Issue The replacement of the thread-local t_lock_dummy_ring with the global g_lock_dummy_ring may introduce unintended side effects in multi-threaded environments. Ensure that the global lock is thread-safe and does not cause contention or race conditions. static padded_lock_dummy g_lock_dummy_ring; static lock_base *get_new_lock(const char *name, bool real_lock) { return (real_lock ? static_cast<lock_base *>(multilock::create_new_lock(MULTILOCK_RECURSIVE, name)) : static_cast<lock_base *>(&g_lock_dummy_ring.lock)); Thread Safety The change from t_lock_dummy_socket to g_lock_dummy_socket introduces a global lock. Verify that this global lock does not lead to false sharing or performance degradation in high-concurrency scenarios. static padded_lock_dummy g_lock_dummy_socket; /* * The following socket options are inherited by a connected TCP socket from the listening socket: * SO_DEBUG, SO_DONTROUTE, SO_KEEPALIVE, SO_LINGER, SO_OOBINLINE, SO_RCVBUF, SO_RCVLOWAT, SO_SNDBUF, * SO_SNDLOWAT, TCP_MAXSEG, TCP_NODELAY. */ static bool is_inherited_option(int __level, int __optname) { bool ret = false; if (__level == SOL_SOCKET) { switch (__optname) { case SO_DEBUG: case SO_DONTROUTE: case SO_KEEPALIVE: case SO_LINGER: case SO_OOBINLINE: case SO_RCVBUF: case SO_RCVLOWAT: case SO_SNDBUF: case SO_SNDLOWAT: case SO_XLIO_RING_ALLOC_LOGIC: ret = true; } } else if (__level == IPPROTO_TCP) { switch (__optname) { case TCP_MAXSEG: case TCP_NODELAY: case TCP_KEEPIDLE: #if LWIP_TCP_KEEPALIVE case TCP_KEEPINTVL: case TCP_KEEPCNT: #endif case TCP_USER_TIMEOUT: ret = true; } } else if (__level == IPPROTO_IP) { switch (__optname) { case IP_TTL: ret = true; } } else if (__level == IPPROTO_IPV6) { switch (__optname) { case IPV6_V6ONLY: ret = true; } } return ret; } event_handler_manager *sockinfo_tcp::get_event_mgr() { if (is_xlio_socket()) { return m_p_group->get_event_handler(); } else if (safe_mce_sys().tcp_ctl_thread == option_tcp_ctl_thread::CTL_THREAD_DELEGATE_TCP_TIMERS) { return &g_event_handler_manager_local; } else { return g_p_event_handler_manager; } } tcp_timers_collection *sockinfo_tcp::get_tcp_timer_collection() { if (is_xlio_socket()) { return m_p_group->get_tcp_timers(); } else if (safe_mce_sys().tcp_ctl_thread == option_tcp_ctl_thread::CTL_THREAD_DELEGATE_TCP_TIMERS) { return &g_thread_local_tcp_timers; } else { return g_tcp_timers_collection; } } static lock_base *get_new_tcp_lock() { return ( safe_mce_sys().tcp_ctl_thread != option_tcp_ctl_thread::CTL_THREAD_DELEGATE_TCP_TIMERS ? static_cast<lock_base *>(multilock::create_new_lock(MULTILOCK_RECURSIVE, "tcp_con")) : static_cast<lock_base *>(&g_lock_dummy_socket.lock));

[galnoam]

@tomerdbz , please move to Fix/closed