mirror of
https://github.com/haproxy/haproxy.git
synced 2026-04-22 14:49:45 -04:00
HAProxy - Load balancer
2a4523f6f4
In GH issue #1275, Fabiano Nunes Parente provided a nicely detailed report showing reproducible crashes under musl. Musl is one of the libs coming with a simple allocator for which we prefer to keep the shared cache. On x86 we have a DWCAS so the lockless implementation is enabled for such libraries. And this implementation has had a small race since day one: the allocator will need to read the first object's <next> pointer to place it into the free list's head. If another thread picks the same element and immediately releases it, while both the local and the shared pools are too crowded, it will be freed to the OS. If the libc's allocator immediately releases it, the memory area is unmapped and we can have a crash while trying to read that pointer. However there is no problem as long as the item remains mapped in memory because whatever value found there will not be placed into the head since the counter will have changed. The probability for this to happen is extremely low, but as analyzed by Fabiano, it increases with the buffer size. On 16 threads it's relatively easy to reproduce with 2MB buffers above 200k req/s, where it should happen within the first 20 seconds of traffic usually. This is a structural issue for which there are two non-trivial solutions: - place a read lock in the alloc call and a barrier made of lock/unlock in the free() call to force to serialize operations; this will have a big performance impact since free() is already one of the contention points; - change the allocator to use a self-locked head, similar to what is done in the MT_LISTS. This requires two memory writes to the head instead of a single one, thus the overhead is exactly one memory write during alloc and one during free; This patch implements the second option. A new POOL_DUMMY pointer was defined for the locked pointer value, allowing to both read and lock it with a single xchg call. The code was carefully optimized so that the locked period remains the shortest possible and that bus writes are avoided as much as possible whenever the lock is held. Tests show that while a bit slower than the original lockless implementation on large buffers (2MB), it's 2.6 times faster than both the no-cache and the locked implementation on such large buffers, and remains as fast or faster than the all implementations when buffers are 48k or higher. Tests were also run on arm64 with similar results. Note that this code is not used on modern libcs featuring a fast allocator. A nice benefit of this change is that since it removes a dependency on the DWCAS, it will be possible to remove the locked implementation and replace it with this one, that is then usable on all systems, thus significantly increasing their performance with large buffers. Given that lockless pools were introduced in 1.9 (not supported anymore), this patch will have to be backported as far as 2.0. The code changed several times in this area and is subject to many ifdefs which will complicate the backport. What is important is to remove all the DWCAS code from the shared cache alloc/free lockless code and replace it with this one. The pool_flush() code is basically the same code as the allocator, retrieving the whole list at once. If in doubt regarding what barriers to use in older versions, it's safe to use the generic ones. This patch depends on the following previous commits: - MINOR: pools: do not maintain the lock during pool_flush() - MINOR: pools: call malloc_trim() under thread isolation - MEDIUM: pools: use a single pool_gc() function for locked and lockless The last one also removes one occurrence of an unneeded DWCAS in the code that was incompatible with this fix. The removal of the now unused seq field will happen in a future patch. Many thanks to Fabiano for his detailed report, and to Olivier for his help on this issue. |
||
|---|---|---|
| .github | ||
| addons | ||
| admin | ||
| dev | ||
| doc | ||
| examples | ||
| include | ||
| reg-tests | ||
| scripts | ||
| src | ||
| tests | ||
| .cirrus.yml | ||
| .gitattributes | ||
| .gitignore | ||
| .travis.yml | ||
| BRANCHES | ||
| CHANGELOG | ||
| CONTRIBUTING | ||
| INSTALL | ||
| LICENSE | ||
| MAINTAINERS | ||
| Makefile | ||
| README | ||
| ROADMAP | ||
| SUBVERS | ||
| VERDATE | ||
| VERSION | ||
The HAProxy documentation has been split into a number of different files for ease of use. Please refer to the following files depending on what you're looking for : - INSTALL for instructions on how to build and install HAProxy - BRANCHES to understand the project's life cycle and what version to use - LICENSE for the project's license - CONTRIBUTING for the process to follow to submit contributions The more detailed documentation is located into the doc/ directory : - doc/intro.txt for a quick introduction on HAProxy - doc/configuration.txt for the configuration's reference manual - doc/lua.txt for the Lua's reference manual - doc/SPOE.txt for how to use the SPOE engine - doc/network-namespaces.txt for how to use network namespaces under Linux - doc/management.txt for the management guide - doc/regression-testing.txt for how to use the regression testing suite - doc/peers.txt for the peers protocol reference - doc/coding-style.txt for how to adopt HAProxy's coding style - doc/internals for developer-specific documentation (not all up to date)