kernel/linux.git/mm/workingset.c, branch v6.4

mm: workingset: update description of the source file

2023-04-18T23:30:11+00:00

The calculation of workingset size is the core logic of handling refault, it had been updated several times[1][2] after workingset.c was created[3]. But the description hadn't been updated accordingly, this mismatch may confuse the readers. So we update the description to make it consistent to the code. [1] commit 34e58cac6d8f ("mm: workingset: let cache workingset challenge anon") [2] commit aae466b0052e ("mm/swap: implement workingset detection for anonymous LRU") [3] commit a528910e12ec ("mm: thrash detection-based file cache sizing") Link: https://lkml.kernel.org/r/202304131634494948454@zte.com.cn Signed-off-by: Yang Yang Cc: Johannes Weiner Cc: Joonsoo Kim Signed-off-by: Andrew Morton

workingset: memcg: sleep when flushing stats in workingset_refault()

2023-04-18T23:29:50+00:00

In workingset_refault(), we call mem_cgroup_flush_stats_atomic_ratelimited() to read accurate stats within an RCU read section and with sleeping disallowed. Move the call above the RCU read section to make it non-atomic. Flushing is an expensive operation that scales with the number of cpus and the number of cgroups in the system, so avoid doing it atomically where possible. Since workingset_refault() is the only caller of mem_cgroup_flush_stats_atomic_ratelimited(), just make it non-atomic, and rename it to mem_cgroup_flush_stats_ratelimited(). Link: https://lkml.kernel.org/r/20230330191801.1967435-7-yosryahmed@google.com Signed-off-by: Yosry Ahmed Acked-by: Shakeel Butt Acked-by: Johannes Weiner Acked-by: Michal Hocko Cc: Jens Axboe Cc: Josef Bacik Cc: Michal Hocko Cc: Michal Koutný Cc: Muchun Song Cc: Roman Gushchin Cc: Tejun Heo Cc: Vasily Averin Cc: Zefan Li Signed-off-by: Andrew Morton

memcg: sleep during flushing stats in safe contexts

2023-04-18T23:29:50+00:00

Currently, all contexts that flush memcg stats do so with sleeping not allowed. Some of these contexts are perfectly safe to sleep in, such as reading cgroup files from userspace or the background periodic flusher. Flushing is an expensive operation that scales with the number of cpus and the number of cgroups in the system, so avoid doing it atomically where possible. Refactor the code to make mem_cgroup_flush_stats() non-atomic (aka sleepable), and provide a separate atomic version. The atomic version is used in reclaim, refault, writeback, and in mem_cgroup_usage(). All other code paths are left to use the non-atomic version. This includes callbacks for userspace reads and the periodic flusher. Since refault is the only caller of mem_cgroup_flush_stats_ratelimited(), change it to mem_cgroup_flush_stats_atomic_ratelimited(). Reclaim and refault code paths are modified to do non-atomic flushing in separate later patches -- so it will eventually be changed back to mem_cgroup_flush_stats_ratelimited(). Link: https://lkml.kernel.org/r/20230330191801.1967435-6-yosryahmed@google.com Signed-off-by: Yosry Ahmed Acked-by: Shakeel Butt Acked-by: Michal Hocko Acked-by: Johannes Weiner Cc: Jens Axboe Cc: Josef Bacik Cc: Michal Hocko Cc: Michal Koutný Cc: Muchun Song Cc: Roman Gushchin Cc: Tejun Heo Cc: Vasily Averin Cc: Zefan Li Signed-off-by: Andrew Morton

memcg: rename mem_cgroup_flush_stats_"delayed" to "ratelimited"

2023-04-18T23:29:49+00:00

mem_cgroup_flush_stats_delayed() suggests his is using a delayed_work, but this is actually sometimes flushing directly from the callsite. What it's doing is ratelimited calls. A better name would be mem_cgroup_flush_stats_ratelimited(). Link: https://lkml.kernel.org/r/20230330191801.1967435-3-yosryahmed@google.com Signed-off-by: Yosry Ahmed Suggested-by: Johannes Weiner Acked-by: Shakeel Butt Acked-by: Johannes Weiner Acked-by: Michal Hocko Cc: Jens Axboe Cc: Josef Bacik Cc: Michal Hocko Cc: Michal Koutný Cc: Muchun Song Cc: Roman Gushchin Cc: Tejun Heo Cc: Vasily Averin Cc: Zefan Li Signed-off-by: Andrew Morton

swap_state: update shadow_nodes for anonymous page

2023-02-03T06:33:24+00:00

Shadow_nodes is for shadow nodes reclaiming of workingset handling, it is updated when page cache add or delete since long time ago workingset only supported page cache. But when workingset supports anonymous page detection, we missied updating shadow nodes for it. This caused that shadow nodes of anonymous page will never be reclaimd by scan_shadow_nodes() even they use much memory and system memory is tense. So update shadow_nodes of anonymous page when swap cache is add or delete by calling xas_set_update(..workingset_update_node). Link: https://lkml.kernel.org/r/202301182013032211005@zte.com.cn Fixes: aae466b0052e ("mm/swap: implement workingset detection for anonymous LRU") Signed-off-by: Yang Yang Reviewed-by: Ran Xiaokai Cc: Bagas Sanjaya Cc: Johannes Weiner Cc: Joonsoo Kim Cc: Matthew Wilcox Signed-off-by: Andrew Morton

workingset: fix confusion around eviction vs refault container

2023-01-19T01:12:53+00:00

Refault decisions are made based on the lruvec where the page was evicted, as that determined its LRU order while it was alive. Stats and workingset aging must then occur on the lruvec of the new page, as that's the node and cgroup that experience the refault and that's the lruvec whose nonresident info ages out by a new resident page. Those lruvecs could be different when a page is shared between cgroups, or the refaulting page is allocated on a different node. There are currently two mix-ups: 1. When swap is available, the resident anon set must be considered when comparing the refault distance. The comparison is made against the right anon set, but the check for swap is not. When pages get evicted from a cgroup with swap, and refault in one without, this can incorrectly consider a hot refault as cold - and vice versa. Fix that by using the eviction cgroup for the swap check. 2. The stats and workingset age are updated against the wrong lruvec altogether: the right cgroup but the wrong NUMA node. When a page refaults on a different NUMA node, this will have confusing stats and distort the workingset age on a different lruvec - again possibly resulting in hot/cold misclassifications down the line. Fix the swap check and the refault pgdat to address both concerns. This was found during code review. It hasn't caused notable issues in production, suggesting that those refault-migrations are relatively rare in practice. Link: https://lkml.kernel.org/r/20230104222944.2380117-1-nphamcs@gmail.com Signed-off-by: Johannes Weiner Co-developed-by: Nhat Pham Signed-off-by: Nhat Pham Signed-off-by: Andrew Morton

mm: multi-gen LRU: rename lru_gen_struct to lru_gen_folio

2023-01-19T01:12:48+00:00

Patch series "mm: multi-gen LRU: memcg LRU", v3. Overview ======== An memcg LRU is a per-node LRU of memcgs. It is also an LRU of LRUs, since each node and memcg combination has an LRU of folios (see mem_cgroup_lruvec()). Its goal is to improve the scalability of global reclaim, which is critical to system-wide memory overcommit in data centers. Note that memcg reclaim is currently out of scope. Its memory bloat is a pointer to each lruvec and negligible to each pglist_data. In terms of traversing memcgs during global reclaim, it improves the best-case complexity from O(n) to O(1) and does not affect the worst-case complexity O(n). Therefore, on average, it has a sublinear complexity in contrast to the current linear complexity. The basic structure of an memcg LRU can be understood by an analogy to the active/inactive LRU (of folios): 1. It has the young and the old (generations), i.e., the counterparts to the active and the inactive; 2. The increment of max_seq triggers promotion, i.e., the counterpart to activation; 3. Other events trigger similar operations, e.g., offlining an memcg triggers demotion, i.e., the counterpart to deactivation. In terms of global reclaim, it has two distinct features: 1. Sharding, which allows each thread to start at a random memcg (in the old generation) and improves parallelism; 2. Eventual fairness, which allows direct reclaim to bail out at will and reduces latency without affecting fairness over some time. The commit message in patch 6 details the workflow: https://lore.kernel.org/r/20221222041905.2431096-7-yuzhao@google.com/ The following is a simple test to quickly verify its effectiveness. Test design: 1. Create multiple memcgs. 2. Each memcg contains a job (fio). 3. All jobs access the same amount of memory randomly. 4. The system does not experience global memory pressure. 5. Periodically write to the root memory.reclaim. Desired outcome: 1. All memcgs have similar pgsteal counts, i.e., stddev(pgsteal) over mean(pgsteal) is close to 0%. 2. The total pgsteal is close to the total requested through memory.reclaim, i.e., sum(pgsteal) over sum(requested) is close to 100%. Actual outcome [1]: MGLRU off MGLRU on stddev(pgsteal) / mean(pgsteal) 75% 20% sum(pgsteal) / sum(requested) 425% 95% #################################################################### MEMCGS=128 for ((memcg = 0; memcg < $MEMCGS; memcg++)); do mkdir /sys/fs/cgroup/memcg$memcg done start() { echo $BASHPID > /sys/fs/cgroup/memcg$memcg/cgroup.procs fio -name=memcg$memcg --numjobs=1 --ioengine=mmap \ --filename=/dev/zero --size=1920M --rw=randrw \ --rate=64m,64m --random_distribution=random \ --fadvise_hint=0 --time_based --runtime=10h \ --group_reporting --minimal } for ((memcg = 0; memcg < $MEMCGS; memcg++)); do start & done sleep 600 for ((i = 0; i < 600; i++)); do echo 256m >/sys/fs/cgroup/memory.reclaim sleep 6 done for ((memcg = 0; memcg < $MEMCGS; memcg++)); do grep "pgsteal " /sys/fs/cgroup/memcg$memcg/memory.stat done #################################################################### [1]: This was obtained from running the above script (touches less than 256GB memory) on an EPYC 7B13 with 512GB DRAM for over an hour. This patch (of 8): The new name lru_gen_folio will be more distinct from the coming lru_gen_memcg. Link: https://lkml.kernel.org/r/20221222041905.2431096-1-yuzhao@google.com Link: https://lkml.kernel.org/r/20221222041905.2431096-2-yuzhao@google.com Signed-off-by: Yu Zhao Cc: Johannes Weiner Cc: Jonathan Corbet Cc: Michael Larabel Cc: Michal Hocko Cc: Mike Rapoport Cc: Roman Gushchin Cc: Suren Baghdasaryan Signed-off-by: Andrew Morton

folio-compat: remove lru_cache_add()

2022-12-12T02:12:13+00:00

There are no longer any callers of lru_cache_add(), so remove it. This saves 79 bytes of kernel text. Also cleanup some comments such that they reference the new folio_add_lru() instead. Link: https://lkml.kernel.org/r/20221101175326.13265-6-vishal.moola@gmail.com Signed-off-by: Vishal Moola (Oracle) Reviewed-by: Matthew Wilcox (Oracle) Cc: Mike Kravetz Cc: Miklos Szeredi Signed-off-by: Andrew Morton

mm: vmscan: make rotations a secondary factor in balancing anon vs file

2022-11-09T01:37:11+00:00

We noticed a 2% webserver throughput regression after upgrading from 5.6. This could be tracked down to a shift in the anon/file reclaim balance (confirmed with swappiness) that resulted in worse reclaim efficiency and thus more kswapd activity for the same outcome. The change that exposed the problem is aae466b0052e ("mm/swap: implement workingset detection for anonymous LRU"). By qualifying swapins based on their refault distance, it lowered the cost of anon reclaim in this workload, in turn causing (much) more anon scanning than before. Scanning the anon list is more expensive due to the higher ratio of mmapped pages that may rotate during reclaim, and so the result was an increase in %sys time. Right now, rotations aren't considered a cost when balancing scan pressure between LRUs. We can end up with very few file refaults putting all the scan pressure on hot anon pages that are rotated en masse, don't get reclaimed, and never push back on the file LRU again. We still only reclaim file cache in that case, but we burn a lot CPU rotating anon pages. It's "fair" from an LRU age POV, but doesn't reflect the real cost it imposes on the system. Consider rotations as a secondary factor in balancing the LRUs. This doesn't attempt to make a precise comparison between IO cost and CPU cost, it just says: if reloads are about comparable between the lists, or rotations are overwhelmingly different, adjust for CPU work. This fixed the regression on our webservers. It has since been deployed to the entire Meta fleet and hasn't caused any problems. Link: https://lkml.kernel.org/r/20221013193113.726425-1-hannes@cmpxchg.org Signed-off-by: Johannes Weiner Cc: Rik van Riel Signed-off-by: Andrew Morton

mm: multi-gen LRU: minimal implementation

2022-09-27T02:46:09+00:00

To avoid confusion, the terms "promotion" and "demotion" will be applied to the multi-gen LRU, as a new convention; the terms "activation" and "deactivation" will be applied to the active/inactive LRU, as usual. The aging produces young generations. Given an lruvec, it increments max_seq when max_seq-min_seq+1 approaches MIN_NR_GENS. The aging promotes hot pages to the youngest generation when it finds them accessed through page tables; the demotion of cold pages happens consequently when it increments max_seq. Promotion in the aging path does not involve any LRU list operations, only the updates of the gen counter and lrugen->nr_pages[]; demotion, unless as the result of the increment of max_seq, requires LRU list operations, e.g., lru_deactivate_fn(). The aging has the complexity O(nr_hot_pages), since it is only interested in hot pages. The eviction consumes old generations. Given an lruvec, it increments min_seq when lrugen->lists[] indexed by min_seq%MAX_NR_GENS becomes empty. A feedback loop modeled after the PID controller monitors refaults over anon and file types and decides which type to evict when both types are available from the same generation. The protection of pages accessed multiple times through file descriptors takes place in the eviction path. Each generation is divided into multiple tiers. A page accessed N times through file descriptors is in tier order_base_2(N). Tiers do not have dedicated lrugen->lists[], only bits in folio->flags. The aforementioned feedback loop also monitors refaults over all tiers and decides when to protect pages in which tiers (N>1), using the first tier (N=0,1) as a baseline. The first tier contains single-use unmapped clean pages, which are most likely the best choices. In contrast to promotion in the aging path, the protection of a page in the eviction path is achieved by moving this page to the next generation, i.e., min_seq+1, if the feedback loop decides so. This approach has the following advantages: 1. It removes the cost of activation in the buffered access path by inferring whether pages accessed multiple times through file descriptors are statistically hot and thus worth protecting in the eviction path. 2. It takes pages accessed through page tables into account and avoids overprotecting pages accessed multiple times through file descriptors. (Pages accessed through page tables are in the first tier, since N=0.) 3. More tiers provide better protection for pages accessed more than twice through file descriptors, when under heavy buffered I/O workloads. Server benchmark results: Single workload: fio (buffered I/O): +[30, 32]% IOPS BW 5.19-rc1: 2673k 10.2GiB/s patch1-6: 3491k 13.3GiB/s Single workload: memcached (anon): -[4, 6]% Ops/sec KB/sec 5.19-rc1: 1161501.04 45177.25 patch1-6: 1106168.46 43025.04 Configurations: CPU: two Xeon 6154 Mem: total 256G Node 1 was only used as a ram disk to reduce the variance in the results. patch drivers/block/brd.c < gfp_flags = GFP_NOIO | __GFP_ZERO | __GFP_HIGHMEM | __GFP_THISNODE; > page = alloc_pages_node(1, gfp_flags, 0); EOF cat >>/etc/systemd/system.conf <>/etc/memcached.conf </sys/fs/cgroup/user.slice/test/memory.max echo $$ >/sys/fs/cgroup/user.slice/test/cgroup.procs fio -name=mglru --numjobs=72 --directory=/mnt --size=1408m \ --buffered=1 --ioengine=io_uring --iodepth=128 \ --iodepth_batch_submit=32 --iodepth_batch_complete=32 \ --rw=randread --random_distribution=random --norandommap \ --time_based --ramp_time=10m --runtime=5m --group_reporting cat memcached.sh modprobe brd rd_nr=1 rd_size=113246208 swapoff -a mkswap /dev/ram0 swapon /dev/ram0 memtier_benchmark -S /var/run/memcached/memcached.sock \ -P memcache_binary -n allkeys --key-minimum=1 \ --key-maximum=65000000 --key-pattern=P:P -c 1 -t 36 \ --ratio 1:0 --pipeline 8 -d 2000 memtier_benchmark -S /var/run/memcached/memcached.sock \ -P memcache_binary -n allkeys --key-minimum=1 \ --key-maximum=65000000 --key-pattern=R:R -c 1 -t 36 \ --ratio 0:1 --pipeline 8 --randomize --distinct-client-seed Client benchmark results: kswapd profiles: 5.19-rc1 40.33% page_vma_mapped_walk (overhead) 21.80% lzo1x_1_do_compress (real work) 7.53% do_raw_spin_lock 3.95% _raw_spin_unlock_irq 2.52% vma_interval_tree_iter_next 2.37% folio_referenced_one 2.28% vma_interval_tree_subtree_search 1.97% anon_vma_interval_tree_iter_first 1.60% ptep_clear_flush 1.06% __zram_bvec_write patch1-6 39.03% lzo1x_1_do_compress (real work) 18.47% page_vma_mapped_walk (overhead) 6.74% _raw_spin_unlock_irq 3.97% do_raw_spin_lock 2.49% ptep_clear_flush 2.48% anon_vma_interval_tree_iter_first 1.92% folio_referenced_one 1.88% __zram_bvec_write 1.48% memmove 1.31% vma_interval_tree_iter_next Configurations: CPU: single Snapdragon 7c Mem: total 4G ChromeOS MemoryPressure [1] [1] https://chromium.googlesource.com/chromiumos/platform/tast-tests/ Link: https://lkml.kernel.org/r/20220918080010.2920238-7-yuzhao@google.com Signed-off-by: Yu Zhao Acked-by: Brian Geffon Acked-by: Jan Alexander Steffens (heftig) Acked-by: Oleksandr Natalenko Acked-by: Steven Barrett Acked-by: Suleiman Souhlal Tested-by: Daniel Byrne Tested-by: Donald Carr Tested-by: Holger Hoffstätte Tested-by: Konstantin Kharlamov Tested-by: Shuang Zhai Tested-by: Sofia Trinh Tested-by: Vaibhav Jain Cc: Andi Kleen Cc: Aneesh Kumar K.V Cc: Barry Song Cc: Catalin Marinas Cc: Dave Hansen Cc: Hillf Danton Cc: Jens Axboe Cc: Johannes Weiner Cc: Jonathan Corbet Cc: Linus Torvalds Cc: Matthew Wilcox Cc: Mel Gorman Cc: Miaohe Lin Cc: Michael Larabel Cc: Michal Hocko Cc: Mike Rapoport Cc: Mike Rapoport Cc: Peter Zijlstra Cc: Qi Zheng Cc: Tejun Heo Cc: Vlastimil Babka Cc: Will Deacon Signed-off-by: Andrew Morton