On Thu, Nov 20, 2025 at 10:12 AM Guopeng Zhang <[email protected]> wrote: > > >On 11/19/25 20:27, Lance Yang wrote: >> From: Lance Yang <[email protected]> >> >> >> On Wed, 19 Nov 2025 18:52:16 +0800, Guopeng Zhang wrote: >>> test_memcg_sock() currently requires that memory.stat's "sock " counter >>> is exactly zero immediately after the TCP server exits. On a busy system >>> this assumption is too strict: >>> >>> - Socket memory may be freed with a small delay (e.g. RCU callbacks). >>> - memcg statistics are updated asynchronously via the rstat flushing >>> worker, so the "sock " value in memory.stat can stay non-zero for a >>> short period of time even after all socket memory has been uncharged. >>> >>> As a result, test_memcg_sock() can intermittently fail even though socket >>> memory accounting is working correctly. >>> >>> Make the test more robust by polling memory.stat for the "sock " counter >>> and allowing it some time to drop to zero instead of checking it only >>> once. If the counter does not become zero within the timeout, the test >>> still fails as before. >>> >>> On my test system, running test_memcontrol 50 times produced: >>> >>> - Before this patch: 6/50 runs passed. >>> - After this patch: 50/50 runs passed. >Hi Lance, > >Thanks a lot for your review and helpful comments! >> >>Good catch! Thanks! >> >> With more CPU cores, updates may be distributed across cores, making it >> slower to reach the per-CPU flush threshold, IIUC :) >> >Yes, that matches what I’ve seen as well — on larger systems it indeed >takes longer for the stats to converge due to per-CPU distribution and >the flush threshold.
Me too. I previously proposed a potential solution to explicitly flush stats via a new interface, "memory.stat_refresh" [1]. However, improving the existing flush mechanism would likely be a better long-term direction. Links: [1] https://lore.kernel.org/linux-mm/[email protected]/ Thanks, Leon [...]
