kernel/linux.git/fs/erofs/decompressor.c, branch v6.18.22

erofs: implement metadata compression

2025-07-24T11:43:31+00:00

Thanks to the meta buffer infrastructure, metadata-compressed inodes are just read from the metabox inode instead of the blockdevice (or backing file) inode. The same is true for shared extended attributes. When metadata compression is enabled, inode numbers are divided from on-disk NIDs because of non-LTS 32-bit application compatibility. Co-developed-by: Gao Xiang Signed-off-by: Bo Liu (OpenAnolis) Acked-by: Chao Yu Signed-off-by: Gao Xiang Link: https://lore.kernel.org/r/20250722003229.2121752-1-hsiangkao@linux.alibaba.com

erofs: address D-cache aliasing

2025-07-10T09:08:27+00:00

Flush the D-cache before unlocking folios for compressed inodes, as they are dirtied during decompression. Avoid calling flush_dcache_folio() on every CPU write, since it's more like playing whack-a-mole without real benefit. It has no impact on x86 and arm64/risc-v: on x86, flush_dcache_folio() is a no-op, and on arm64/risc-v, PG_dcache_clean (PG_arch_1) is clear for new page cache folios. However, certain ARM boards are affected, as reported. Fixes: 3883a79abd02 ("staging: erofs: introduce VLE decompression support") Closes: https://lore.kernel.org/r/c1e51e16-6cc6-49d0-a63e-4e9ff6c4dd53@pengutronix.de Closes: https://lore.kernel.org/r/38d43fae-1182-4155-9c5b-ffc7382d9917@siemens.com Tested-by: Jan Kiszka Tested-by: Stefan Kerkmann Signed-off-by: Gao Xiang Link: https://lore.kernel.org/r/20250709034614.2780117-2-hsiangkao@linux.alibaba.com

erofs: support unaligned encoded data

2025-03-17T06:02:15+00:00

We're almost there. It's straight-forward to adapt the current decompression subsystem to support unaligned encoded (compressed) data. Note that unaligned data is not encouraged because of worse I/O and caching efficiency unless the corresponding compressor doesn't support fixed-sized output compression natively like Zstd. Signed-off-by: Gao Xiang Acked-by: Chao Yu Link: https://lore.kernel.org/r/20250310095459.2620647-10-hsiangkao@linux.alibaba.com

erofs: move {in,out}pages into struct z_erofs_decompress_req

2025-03-16T17:22:50+00:00

It seems that all compressors need those two values, so just move them into the common structure. `struct z_erofs_lz4_decompress_ctx` can be dropped too. Reviewed-by: Chao Yu Signed-off-by: Gao Xiang Link: https://lore.kernel.org/r/20250305124007.1810731-1-hsiangkao@linux.alibaba.com

include/linux/lz4.h: add some missing macros

2025-01-25T06:47:28+00:00

Currently, LZ4_DISTANCE_MAX and LZ4_DECOMPRESS_INPLACE_MARGIN are defined in the erofs subsystem for LZ4 in-place decompression, which is somewhat unsuitable since they should belong to the LZ4 itself and may change with future LZ4 codebase updates. Move them to include/linux/lz4.h to match the upstream LZ4 library [1]. No logic changes. [1] https://github.com/lz4/lz4/blob/v1.10.0/lib/lz4.h#L670 Link: https://lkml.kernel.org/r/20250114130454.1191150-1-hsiangkao@linux.alibaba.com Signed-off-by: Gao Xiang Cc: Yann Collet Cc: Nick Terrell Cc: Chao Yu Cc: Yue Hu Cc; Jeffle Xu Cc: Sandeep Dhavale Signed-off-by: Andrew Morton

erofs: fix error handling in z_erofs_init_decompressor

2024-09-09T16:46:34+00:00

If we get a failure at the first decompressor init (i = 0), the clean up while loop could enter infinite loop due to wrong while check. Check the value of i now to see if we need any clean up at all. Fixes: 5a7cce827ee9 ("erofs: refine z_erofs_{init,exit}_subsystem()") Reported-by: liujinbao1 Signed-off-by: Sandeep Dhavale Reviewed-by: Gao Xiang Reviewed-by: Chao Yu Link: https://lore.kernel.org/r/20240905060027.2388893-1-dhavale@google.com Signed-off-by: Gao Xiang

erofs: avoid refcounting short-lived pages

2024-07-11T07:14:26+00:00

LZ4 always reuses the decompressed buffer as its LZ77 sliding window (dynamic dictionary) for optimal performance. However, in specific cases, the output buffer may not fully contain valid page cache pages, resulting in the use of short-lived pages for temporary purposes. Due to the limited sliding window size, LZ4 shortlived bounce pages can also be reused in a sliding manner, so each bounce page can be vmapped multiple times in different relative positions by design. In order to avoiding double frees, currently, reuse counts are recorded via page refcount, but it will no longer be used as-is in the future world of Memdescs. Just maintain a lookup table to check if a shortlived page is reused. Signed-off-by: Gao Xiang Link: https://lore.kernel.org/r/20240711053659.1364989-1-hsiangkao@linux.alibaba.com

erofs: tidy up stream decompressors

2024-07-09T11:04:41+00:00

Just use a generic helper to prepare buffers for all supported stream decompressors, eliminating similar logic. Signed-off-by: Gao Xiang Link: https://lore.kernel.org/r/20240709094106.3018109-3-hsiangkao@linux.alibaba.com

erofs: refine z_erofs_{init,exit}_subsystem()

2024-07-09T11:04:40+00:00

Introduce z_erofs_{init,exit}_decompressor() to unexport z_erofs_{deflate,lzma,zstd}_{init,exit}(). Besides, call them in z_erofs_{init,exit}_subsystem() for simplicity. Signed-off-by: Gao Xiang Link: https://lore.kernel.org/r/20240709094106.3018109-2-hsiangkao@linux.alibaba.com

erofs: move each decompressor to its own source file

2024-07-09T11:04:40+00:00

Thus *_config() function declarations can be avoided. Signed-off-by: Gao Xiang Link: https://lore.kernel.org/r/20240709094106.3018109-1-hsiangkao@linux.alibaba.com