As shown in the patchseries that introduced the zswap same-filled optimization [1], 10-20% of the pages stored in zswap are same-filled. This is also observed across Meta's server fleet. By using VM counters in swap_writepage (not included in this patchseries) it was found that less than 1% of the same-filled pages to be swapped out are non-zero pages. For conventional swap setup (without zswap), rather than reading/writing these pages to flash resulting in increased I/O and flash wear, a bitmap can be used to mark these pages as zero at write time, and the pages can be filled at read time if the bit corresponding to the page is set. When using zswap with swap, this also means that a zswap_entry does not need to be allocated for zero filled pages resulting in memory savings which would offset the memory used for the bitmap. A similar attempt was made earlier in [2] where zswap would only track zero-filled pages instead of same-filled. This patchseries adds zero-filled pages optimization to swap (hence it can be used even if zswap is disabled) and removes the same-filled code from zswap (as only 1% of the same-filled pages are non-zero), simplifying code. This patchseries is based on mm-unstable. [1] https://lore.kernel.org/all/20171018104832epcms5p1b2232e2236258de3d03d1344dde9fce0@epcms5p1/ [2] https://lore.kernel.org/lkml/20240325235018.2028408-1-yosryahmed@xxxxxxxxxx/ --- v6 -> v7: (Yosry and David): - Change to kvmalloc_array for zeromap allocation instead of kvzalloc as it does an additional overflow check, and use sizeof(unsigned long) for allocation size calculation to take into account 32 bit kernels. v5 -> v6 (kernel test robot <oliver.sang@xxxxxxxxx>): - change bitmap_zalloc/free to kvzalloc/free as a very large swap file will result in the allocation order to exceed MAX_PAGE_ORDER retulting in bitmap_zalloc to fail. v4 -> v5 (Yosry): - Correct comment about using clear_bit instead of bitmp_clear. - Remove clearing the zeromap from swap_cluster_schedule_discard and swap_do_scheduled_discard. v3 -> v4: - remove folio_start/end_writeback when folio is zero filled at swap_writepage (Matthew) - check if a large folio is partially in zeromap and return without folio_mark_uptodate so that an IO error is emitted, rather than checking zswap/disk (Yosry) - clear zeromap in swap_free_cluster (Nhat) v2 -> v3: - Going back to the v1 version of the implementation (David and Shakeel) - convert unatomic bitmap_set/clear to atomic set/clear_bit (Johannes) - use clear_highpage instead of folio_page_zero_fill (Yosry) v1 -> v2: - instead of using a bitmap in swap, clear pte for zero pages and let do_pte_missing handle this page at page fault. (Yosry and Matthew) - Check end of page first when checking if folio is zero filled as it could lead to better performance. (Yosry) Usama Arif (2): mm: store zero pages to be swapped out in a bitmap mm: remove code to handle same filled pages include/linux/swap.h | 1 + mm/page_io.c | 113 ++++++++++++++++++++++++++++++++++++++++++- mm/swapfile.c | 20 ++++++++ mm/zswap.c | 86 +++----------------------------- 4 files changed, 141 insertions(+), 79 deletions(-) -- 2.43.0
