-
Notifications
You must be signed in to change notification settings - Fork 1.7k
Inadequate locking for multithreading in libModSecurity In-Memory persistent storage #3054
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
Hi @martinhsv, thanks for sharing the results of your inspection. I try to write a sample code that will help us to investigate the behavior that you explained, but it would be a help if you could provide a rule set which helps to cause the (possible) behavior? |
Hi @martinhsv, thanks again for your report. I made a small test source which embeds the libmodsecurity3 library to a multi threaded application. The sad news is that it not necessary to add the If the debug log is turned on, it is guaranteed to cause a segfault.
If I turn off the debug.log, then I got a segfault in
I'm going to finish that example tool soon (add the correct |
Yes of course. I never claimed that the problem originated with that 2023 development effort, only that that was when I noticed it. The underlying issue has been present in the code for years. |
There is a separated branch which has a new, multi-threaded example. With this, I couldn't reproduce this behavior. @martinhsv please take a look to rules - is that something you thought? If anyone wants to play with that code, please let me know, I try to help. |
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 recursive mutex
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 recursive mute - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics.
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this.
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this.
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this.
I've just created PR #3216 to try and address this. @martinhsv, it'd be helpful if you could take a look at the changes and offer your opinion. thx! |
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this.
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this.
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this.
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this. - Updates to InMemoryPerProcess::resolveFirst - Updated the code to store the expired var in 'expiredVars' to delete them after iterating over the range (and releasing the read lock, as 'delIfExpired' needs to acquire it for exclusive access), as the current call to 'delIfExpired' would invalidate the range triggering undefined behaviour on the following iteration. - Noticed that in commit 118e1b3 the call to 'delIfExpired' in this function is done using 'it->second.getValue()'' instead of 'it->first', which seems incorrect (based on similar code in other resolveXXX functions). - Updated InMemoryPerProcess::delIfExpired to use 'std::find_if' (with a lambda that matches both the key and the 'isExpired' condition) because the data structure is a multimap. The version introduced in commit 118e1b3 could find an entry (not necessarily the first, because the map is unordered) where 'isExpired' is 'false' and exit, while another entry could be expired.
…rween) - Rewritten to use C++ libModSecurity API and std::thread (instead of pthreads)
…rween) - Rewritten to use C++ libModSecurity API and std::thread (instead of pthreads)
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this. - Updates to InMemoryPerProcess::resolveFirst - Updated the code to store the expired var in 'expiredVars' to delete them after iterating over the range (and releasing the read lock, as 'delIfExpired' needs to acquire it for exclusive access), as the current call to 'delIfExpired' would invalidate the range triggering undefined behaviour on the following iteration. - Noticed that in commit 118e1b3 the call to 'delIfExpired' in this function is done using 'it->second.getValue()'' instead of 'it->first', which seems incorrect (based on similar code in other resolveXXX functions). - Updated InMemoryPerProcess::delIfExpired to use 'std::find_if' (with a lambda that matches both the key and the 'isExpired' condition) because the data structure is a multimap. The version introduced in commit 118e1b3 could find an entry (not necessarily the first, because the map is unordered) where 'isExpired' is 'false' and exit, while another entry could be expired.
…rween) - Rewritten to use C++ libModSecurity API and std::thread (instead of pthreads)
…rween) - Rewritten to use C++ libModSecurity API and std::thread (instead of pthreads)
…rween) - Rewritten to use C++ libModSecurity API and std::thread (instead of pthreads)
- As reported in owasp-modsecurity#3054, the resolve methods in InMemoryPerProcess are not acquiring a lock/mutex to prevent concurrent access to the data structures that may be modified at the same time from other threads, and thus triggering undefined behaviour. - Replace inheritance of std::unordered_multimap data structure with data member to prevent potential clients to use it without acquiring the mutex to protect concurrent access. - Replace pthreads lock with std C++11 std::shared_mutex - Provides exclusive/shared lock access so that multiple readers can access the data at the same time, but only one writer. this is used to favor query performance by allowing more concurrent access to the data until an update needs to be performed. - Simplifies acquisition and disposal of lock/mutex with std::lock_guard, which has RAII semantics. - NOTE: Because std::shared_mutex is not recursive, calls to another function that tries to acquire the lock will fail. Introduced __store & __updateFirst helper methods to workaround this. - Updates to InMemoryPerProcess::resolveFirst - Updated the code to store the expired var in 'expiredVars' to delete them after iterating over the range (and releasing the read lock, as 'delIfExpired' needs to acquire it for exclusive access), as the current call to 'delIfExpired' would invalidate the range triggering undefined behaviour on the following iteration. - Noticed that in commit 118e1b3 the call to 'delIfExpired' in this function is done using 'it->second.getValue()'' instead of 'it->first', which seems incorrect (based on similar code in other resolveXXX functions). - Updated InMemoryPerProcess::delIfExpired to use 'std::find_if' (with a lambda that matches both the key and the 'isExpired' condition) because the data structure is a multimap. The version introduced in commit 118e1b3 could find an entry (not necessarily the first, because the map is unordered) where 'isExpired' is 'false' and exit, while another entry could be expired.
…rween) - Rewritten to use C++ libModSecurity API and std::thread (instead of pthreads)
This issue does not apply to the main, current use case of libModSecurity with nginx, since that is a multi-process but single-threaded application. This finding is based on code inspection only; I did not build a multi-threaded environment to prove a crash or other tangible errors.
While adding the expirevar support I noticed that, while the in-memory option includes some internal locking, it does so only for updates.
The problem is that if ThreadA is performing a read-only action using an interator and, while that is ongoing, ThreadB makes an update to the data, it could invalidate the iterator that ThreadA is in the midst of using. (Note that the update by ThreadB might be a deletion, but an insertion could also invalidate the iterator that is in-use by ThreadA.)
The text was updated successfully, but these errors were encountered: