Using shm to reduce memory copy while using remote connector by maobaolong · Pull Request #2601 · LMCache/LMCache

maobaolong · 2026-02-14T05:51:36Z

What this PR does / why we need it:

This pull request introduced POSIX shared memory (SHM) for CPU-pinned memory allocations. This integration aims to optimize data transfer by reducing memory copy overhead, especially when using remote connectors. The changes include C++ implementations for SHM-backed pinned memory.

This PR also introduced a new demonstrate ShmFileConnector leverages this SHM capability, allowing a C++ worker process to perform file I/O directly on shared memory regions, thereby streamlining data movement between LMCache and external storage.

Special notes for your reviewers:

If applicable:

this PR contains user facing changes - docs added
this PR contains unit tests.

gemini-code-assist · 2026-02-14T05:52:02Z

Summary of Changes

Hello @maobaolong, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly enhances LMCache's memory management by integrating POSIX shared memory (SHM) for CPU-pinned memory allocations. This integration aims to optimize data transfer by reducing memory copy overhead, especially when using remote connectors. The changes include C++ implementations for SHM-backed pinned memory, their exposure to Python, and their seamless integration into the existing memory allocation framework. A new ShmFileConnector leverages this SHM capability, allowing a C++ worker process to perform file I/O directly on shared memory regions, thereby streamlining data movement between LMCache and external storage. The PR also includes robust testing and verification for this new functionality.

Highlights

Shared Memory (SHM) Allocation: New C++ functions alloc_shm_pinned_ptr and free_shm_pinned_ptr are introduced to manage CUDA-pinned memory backed by POSIX shared memory, enabling zero-copy data transfer between processes.
Python Bindings and Fallbacks: The new SHM allocation/deallocation functions are exposed to Python via pybind11, and Python-based fallback implementations are added using multiprocessing.shared_memory and ctypes for non-CUDA environments.
Memory Management Integration: The core memory allocation and deallocation functions in lmcache/v1/memory_management.py are updated to conditionally utilize SHM-backed pinned memory when a shm_name is specified, and MixedMemoryAllocator now supports this configuration.
New Remote Connector for SHM: A new ShmFileConnector is implemented, facilitating remote file I/O through a C++ subprocess (shm_file_worker) that directly interacts with shared memory, supporting both PIPE and TCP communication modes.
Comprehensive Testing and Verification: New CMake files, a C++ shm_file_worker executable, and extensive Python tests (test_shm_allocator.py) are added to verify the end-to-end functionality of SHM allocation and the ShmFileConnector. A verification script (verify_shmfile_tcp.py) is also included for IDC/TCP mode testing.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Changelog

CMakeLists.txt
- Added a root CMakeLists.txt file for project configuration.
csrc/mem_alloc.cpp
- Included fcntl.h and implemented functions for shared memory (SHM) pinned pointer allocation and deallocation.
csrc/mem_alloc.h
- Declared new functions for shared memory (SHM) pinned pointer management.
csrc/pybind.cpp
- Exposed shared memory (SHM) pinned pointer allocation and deallocation functions to Python.
lmcache/non_cuda_equivalents.py
- Added Python fallback implementations for shared memory (SHM) pinned pointer allocation and deallocation.
lmcache/v1/memory_management.py
- Updated CPU memory allocation and freeing to support shared memory (SHM) and integrated SHM name handling into MixedMemoryAllocator.
lmcache/v1/standalone/main.py
- Removed the fmt argument from LMCacheMetadata initialization.
lmcache/v1/storage_backend/local_cpu_backend.py
- Passed the configuration object to the MixedMemoryAllocator.
tests/v1/shm_allocator/csrc/.gitignore
- Added build/ to the gitignore file.
tests/v1/shm_allocator/csrc/CMakeLists.txt
- Added CMake configuration for building the shm_file_worker executable.
tests/v1/shm_allocator/csrc/shm_file_worker.cpp
- Added a C++ executable for shared memory (SHM) file I/O operations.
tests/v1/shm_allocator/shmfile_basic_check.yaml
- Added a new configuration file for testing the ShmFileConnector.
tests/v1/shm_allocator/shmfile_connector.py
- Added a new ShmFileConnector for remote file I/O using shared memory and a C++ worker.
tests/v1/shm_allocator/test_shm_allocator.py
- Added comprehensive tests for the ShmFileConnector and shared memory (SHM) worker functionality.
tests/v1/shm_allocator/verify_shmfile_tcp.py
- Added a Python script to verify shared memory (SHM) file connector functionality in TCP mode.

Activity

New core functionality for shared memory (SHM) management has been introduced in C++.
Python bindings for the new SHM functions have been added, along with Python fallback implementations.
The memory management system has been updated to integrate SHM-backed allocations.
A new ShmFileConnector has been developed to leverage SHM for remote file I/O.
Extensive testing, including unit and end-to-end tests, has been added for the new SHM features and connector.
A dedicated verification script for IDC/TCP mode testing of the SHM connector has been included.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

gemini-code-assist

Code Review

This pull request introduces a significant new feature: a remote connector that leverages POSIX shared memory for inter-process communication, implemented with C++ allocation functions, a C++ worker process for file I/O, and a Python connector. However, a critical security vulnerability has been identified in the shm_file_worker process, allowing remote arbitrary file read/write (path traversal) and arbitrary memory access via integer overflows. This is compounded by the worker binding to all network interfaces without authentication, potentially leading to full host system compromise. Beyond these critical security concerns, further improvements are needed for robustness in error handling, memory access, and adherence to Python style conventions.

sammshen · 2026-02-20T21:14:25Z

small comment LGTM!

maobaolong · 2026-02-21T01:38:51Z

@sammshen Thanks for the review, abstracted a helper method, PTAL.

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

- Fix integer overflow vulnerabilities in shm_file_worker by adding robust boundary checks - Change default network binding from 0.0.0.0 to 127.0.0.1 for better security - Add GIL release for free_shm_pinned_ptr to prevent blocking Python interpreter - Move all imports to the top of test files for better code organization Signed-off-by: baoloongmao <baoloongmao@tencent.com>

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

sammshen

LGTM!

chunxiaozheng

lgtm

…#2601) * Shm and demo Signed-off-by: baoloongmao <baoloongmao@tencent.com> * support set worker addr to verify in container env Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Add missing file Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Fix unit Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Add a shm memory example Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Move to right folder Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Fix security vulnerabilities based on Gemini code review - Fix integer overflow vulnerabilities in shm_file_worker by adding robust boundary checks - Change default network binding from 0.0.0.0 to 127.0.0.1 for better security - Add GIL release for free_shm_pinned_ptr to prevent blocking Python interpreter - Move all imports to the top of test files for better code organization Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Address comment Signed-off-by: baoloongmao <baoloongmao@tencent.com> --------- Signed-off-by: baoloongmao <baoloongmao@tencent.com>

…#2601) * Shm and demo Signed-off-by: baoloongmao <baoloongmao@tencent.com> * support set worker addr to verify in container env Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Add missing file Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Fix unit Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Add a shm memory example Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Move to right folder Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Fix security vulnerabilities based on Gemini code review - Fix integer overflow vulnerabilities in shm_file_worker by adding robust boundary checks - Change default network binding from 0.0.0.0 to 127.0.0.1 for better security - Add GIL release for free_shm_pinned_ptr to prevent blocking Python interpreter - Move all imports to the top of test files for better code organization Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Address comment Signed-off-by: baoloongmao <baoloongmao@tencent.com> --------- Signed-off-by: baoloongmao <baoloongmao@tencent.com> Signed-off-by: shaoxiawjc <wjc2800@163.com>

…#2601) * Shm and demo Signed-off-by: baoloongmao <baoloongmao@tencent.com> * support set worker addr to verify in container env Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Add missing file Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Fix unit Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Add a shm memory example Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Move to right folder Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Fix security vulnerabilities based on Gemini code review - Fix integer overflow vulnerabilities in shm_file_worker by adding robust boundary checks - Change default network binding from 0.0.0.0 to 127.0.0.1 for better security - Add GIL release for free_shm_pinned_ptr to prevent blocking Python interpreter - Move all imports to the top of test files for better code organization Signed-off-by: baoloongmao <baoloongmao@tencent.com> * Address comment Signed-off-by: baoloongmao <baoloongmao@tencent.com> --------- Signed-off-by: baoloongmao <baoloongmao@tencent.com> Signed-off-by: Aaron Wu <aaron.wu@dell.com>

gemini-code-assist Bot reviewed Feb 14, 2026

View reviewed changes

maobaolong force-pushed the shm branch from 8d013cb to 7675889 Compare February 14, 2026 06:07

maobaolong added the full Run comprehensive tests on this PR label Feb 15, 2026

sammshen reviewed Feb 20, 2026

View reviewed changes

Comment thread lmcache/v1/memory_management.py

sammshen reviewed Feb 20, 2026

View reviewed changes

Comment thread lmcache/v1/memory_management.py Outdated

maobaolong requested a review from sammshen February 21, 2026 01:39

maobaolong added 8 commits February 21, 2026 09:57

Shm and demo

2617e4b

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

support set worker addr to verify in container env

a3f4023

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

Add missing file

df80f5e

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

Fix unit

246d19f

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

Add a shm memory example

7652fff

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

Move to right folder

f14891a

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

Address comment

775e1e2

Signed-off-by: baoloongmao <baoloongmao@tencent.com>

maobaolong force-pushed the shm branch from 166ce1a to 775e1e2 Compare February 21, 2026 01:58

sammshen approved these changes Feb 21, 2026

View reviewed changes

chunxiaozheng approved these changes Feb 24, 2026

View reviewed changes

maobaolong merged commit f7cc47b into LMCache:dev Feb 24, 2026
24 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Using shm to reduce memory copy while using remote connector#2601

Using shm to reduce memory copy while using remote connector#2601
maobaolong merged 8 commits intoLMCache:devfrom
maobaolong:shm

maobaolong commented Feb 14, 2026 •

edited

Loading

Uh oh!

gemini-code-assist Bot commented Feb 14, 2026

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sammshen commented Feb 20, 2026

Uh oh!

maobaolong commented Feb 21, 2026

Uh oh!

sammshen left a comment

Uh oh!

chunxiaozheng left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

maobaolong commented Feb 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

gemini-code-assist Bot commented Feb 14, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sammshen commented Feb 20, 2026

Uh oh!

maobaolong commented Feb 21, 2026

Uh oh!

sammshen left a comment

Choose a reason for hiding this comment

Uh oh!

chunxiaozheng left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

maobaolong commented Feb 14, 2026 •

edited

Loading