From: Zhang Yi <yi.zhang@xxxxxxxxxx> Changes since v4: - In patch 1, call ext4_truncate_folio() only if truncating range is PAGE_SIZE unaligned, and rename the variable start_boundary to page_boundary. Changes since v3: - In patch 1, rename ext4_truncate_folios_range() and move journalled mode specified handles and truncate_pagecache_range() into this helper. - In patch 3, switch to use ext4_truncate_page_cache_block_range(). - In patch 4, use IS_ALIGNED macro to check offset alignments and introduce EXT4_B_TO_LBLK to do the lblk conversion. - In patch 5, keep the first ext4_alloc_file_blocks() call before truncating pagecache. - In patch 9, rename 'out' label to 'out_inode_lock'. Changes since v2: - Add Patch 1 to address a newly discovered data loss issue that occurs when using mmap to write after zeroing out a partial page on a filesystem with the block size smaller than the page size. - Do not write all data before punching hole, zeroing out and collapsing range as Jan suggested, also drop current data writeback in ext4_punch_hole(). - Since we don't write back all data in these 4 operations, we only writeback data during inserting range,so do not factor out new helpers in the last two patches, just move common components of sub-operations into ext4_fallocate(). - Only keep Jan's review tag on patch 2 and 8, other patches contain many code adaptations, so please review them again. Changes since v1: - Fix an using uninitialized variable problem in the error out path in ext4_do_fallocate() in patch 08. v4: https://lore.kernel.org/linux-ext4/20241216013915.3392419-1-yi.zhang@xxxxxxxxxxxxxxx/ v3: https://lore.kernel.org/linux-ext4/20241022111059.2566137-1-yi.zhang@xxxxxxxxxxxxxxx/ https://lore.kernel.org/linux-ext4/20241010133333.146793-1-yi.zhang@xxxxxxxxxx/ v2: https://lore.kernel.org/linux-ext4/20240904062925.716856-1-yi.zhang@xxxxxxxxxxxxxxx/ Thanks, Yi. Original Info: Current ext4 fallocate code is mess with mode checking, locking, input parameter checking, position calculation, and having some stale code. Almost all five sub-operation share similar preparation steps, so it deserves a cleanup now. This series tries to improve the code by refactoring all operations related to fallocate. It unifies variable naming, reduces unnecessary position calculations, and factors out common preparation components. The first patch addresses a potential data loss issue that occurs when using mmap to write after zeroing out partial blocks of a page on a filesystem where the block size is smaller than the page size. Subsequent patches focus on cleanup and refactoring, please see them for details. After this series, we will significantly reduce redundant code and enhance clarity compared to the previous version. Zhang Yi (10): ext4: remove writable userspace mappings before truncating page cache ext4: don't explicit update times in ext4_fallocate() ext4: don't write back data before punch hole in nojournal mode ext4: refactor ext4_punch_hole() ext4: refactor ext4_zero_range() ext4: refactor ext4_collapse_range() ext4: refactor ext4_insert_range() ext4: factor out ext4_do_fallocate() ext4: move out inode_lock into ext4_fallocate() ext4: move out common parts into ext4_fallocate() fs/ext4/ext4.h | 4 + fs/ext4/extents.c | 527 +++++++++++++++++----------------------------- fs/ext4/inode.c | 205 ++++++++++-------- 3 files changed, 313 insertions(+), 423 deletions(-) -- 2.46.1
