wireguard-linux/fs/erofs, branch stable

erofs: add missing buf->off in erofs_bread()

2026-09-11T04:16:20Z

erofs_bread() locates the target folio with index = (buf->off + offset) >> PAGE_SHIFT; but computes the in-folio offset without taking buf->off into account: return buf->base + (offset & ~PAGE_MASK); If buf->off is not page-aligned, the returned pointer misses the in-page component of buf->off, so callers end up fetching data from a wrong offset. buf->off is set to sbi->dif0.fsoff in erofs_init_metabuf(), and fsoff can be specified via the "fsoffset=" mount option, which only requires block-size alignment. Therefore, on an image with a sub-page block size (e.g. 512 bytes), a non-page-aligned fsoff (e.g. 512) triggers the issue, since 512 is a multiple of the block size but not of PAGE_SIZE. It can be reproduced by mounting an image that is placed at a non-page-aligned offset: mkfs.erofs -b512 -zlz4hc sub.erofs src/ # prepend 512 bytes of padding to the image mount -t erofs -o loop,fsoffset=512 padded.erofs /mnt which fails with erofs (device loop0): cannot find valid erofs superblock because the on-disk superblock (at offset 1024 within the image, i.e. 1536 within the padded file) is read from a wrong in-folio offset. With this fixed, the very same image mounts successfully and its file contents match those read from the unpadded image. Fix it by including buf->off in the in-folio offset calculation, so that it is consistent with the folio index calculation. Fixes: c36ec00d7f67 ("erofs: add 'fsoffset' mount option to specify filesystem offset") Signed-off-by: Binglei Wang Reviewed-by: Gao Xiang Signed-off-by: Gao Xiang

erofs: delimit inode_share cache key components

2026-09-07T10:20:30Z

Previously, inode_share keys were encoded as follows: fingerprint || domain_id It would be better to have a separator between the fingerprint and domain ID so that the fingerprint won't be parsed as part of a domain ID. Change the key encoding as follows: domain_id || '\0' || fingerprint Since domain_id is a NUL-terminated string, this makes the in-memory key indices unambiguous. Signed-off-by: Chengyu Zhu Reviewed-by: Gao Xiang Fixes: e0bf7d1c074d ("erofs: support user-defined fingerprint name") Signed-off-by: Gao Xiang

erofs: disable LZ4 rolling decompression for now

2026-09-03T14:49:35Z

LZ4 rolling decompression [1] was introduced to reduce the memory footprint of temporary pages: For many cases, it is needed for users to read small data within a compressed extent (pcluster), either due to random small read, or since uptodate folios (typically order-0) cannot be reused for decompression again since decompression algorithm refills already-uptodate folios. Rolling decompression works because LZ4 is LZ77-based and only refers to the most recent 64 KiB of decompressed data, so in theory only a bounded rolling window of temporary pages is needed when decompressing. It can save a lot of temporary memory, e.g. 601,960-byte data can be compressed into a 256k LZ4 compressed extent, which means it needs 146 extra pages per request in the worst case if rolling decompression is disabled. However, the upstream LZ4 implementation is not under EROFS' control: For example, the literal copy memmove() may still **copy long literals backward** on x86 based on the address comparison even when the source and destination ranges do not overlap (IOWs, inline decompression doesn't need to be considered here). That breaks the rolling assumption and makes the optimization broken. Disable it for now to make sure the data correctness first since EROFS is used everywhere now: The rolling window approach can be revived once we either ensure that the official LZ4 code always copies forward for non-overlapping ranges or maintain our own LZ4 implementation in EROFS. The main impact is a higher runtime memory footprint; However, recent commit 0f6273ab4637 ("erofs: add a reserved buffer pool for lz4 decompression") helps mitigate this when enabled but it's still not perfect. [1] https://www.usenix.org/conference/atc19/presentation/gao § 3.3 Decompression Reported-by: "Walther, Jens-Uwe" Closes: https://lore.kernel.org/r/BEZP281MB2102E57CD31862B8D958B33DD2AC2@BEZP281MB2102.DEUP281.PROD.OUTLOOK.COM Fixes: 8e6c8fa9f2e9 ("erofs: enable big pcluster feature") Cc: Yann Collet Signed-off-by: Gao Xiang

erofs: preserve LZMA decoders on resize failure

2026-09-02T02:32:01Z

The pool-resize path frees each stream's old decoder before allocating its replacement. If an allocation fails after some streams have already been replaced, the failed stream is put back on the list with state == NULL. z_erofs_lzma_max_dictsize is still advanced as if the whole pool had been resized. An existing LZMA mount can select the broken stream and pass NULL to xz_dec_microlzma_reset(). A retry at the same size also skip another resize attempt. Since the global maximum was advanced, thus, the invalid state is left unrepaired. Allocate each replacement before freeing the old decoder, temporarily retaining one old decoder during allocation. Stop at the first failure and advance z_erofs_lzma_max_dictsize only after all streams satisfy the request. Record each stream's dictionary capacity so retries can skip streams already enlarged before a partial failure. Fixes: 622ceaddb764 ("erofs: lzma compression support") Cc: stable@vger.kernel.org Signed-off-by: Nikhil Gurudasani Reviewed-by: Gao Xiang Signed-off-by: Gao Xiang

erofs: add sysfs feature entry for xattr prefixes

2026-09-02T02:03:30Z

Let /sys/fs/erofs/features/xattr_prefixes advertise that this kernel supports the EROFS_FEATURE_INCOMPAT_XATTR_PREFIXES on-disk format. Fixes: 6a318ccd7e08 ("erofs: enable long extended attribute name prefixes") Cc: stable@vger.kernel.org # 6.4+ Reviewed-by: Gao Xiang Signed-off-by: Jingbo Xu Signed-off-by: Gao Xiang

Merge tag 'erofs-for-7.3-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs

2026-08-25T19:27:41Z

Pull more erofs updates from Gao Xiang: - Fix up the EROFS_FS_ZIP_LZMA_DEFAULT_MAX_STREAMS default logic so that "make savedefconfig" won't write the needless default value to the defconfig file - Add support for SEEK_{HOLE,DATA}, splice() as well as enable large folios in inode_share mode - Fix z_erofs_gbuf_growsize() after the previous buffer resizing fails * tag 'erofs-for-7.3-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs: erofs: simplify z_erofs_gbuf_growsize() erofs: skip sufficiently large global buffers when resizing erofs: support large folios in inode_share mode erofs: support splice() in inode_share mode erofs: support SEEK_HOLE/SEEK_DATA in inode_share mode erofs: Fix EROFS_FS_ZIP_LZMA_DEFAULT_MAX_STREAMS default logic

erofs: simplify z_erofs_gbuf_growsize()

2026-08-24T04:21:24Z

- Use guard(mutex)() to manage gbuf_resize_mutex; - Nullify tmp_pages after successful allocation, so that the error path becomes clearer. Reviewed-by: Chao Yu Signed-off-by: Gao Xiang

erofs: skip sufficiently large global buffers when resizing

2026-08-24T04:21:09Z

z_erofs_gbuf_nrpages is advanced only after every global buffer has been grown. If a resize fails after some buffers were enlarged, a retry revisits those enlarged buffers. Retrying the same size then returns -ENOMEM because alloc_pages_bulk() has no pages to add and the unchanged return value is treated as a failure. Retrying an intermediate size allocates a temporary pointer array smaller than gbuf->nrpages and copies more existing pointers than the array can hold. Skip buffers that already satisfy the request. Once all remaining buffers have caught up, advancing z_erofs_gbuf_nrpages again describes the guaranteed minimum size across the pool. Fixes: d6db47e571dc ("erofs: do not use pagepool in z_erofs_gbuf_growsize()") Cc: stable@vger.kernel.org # 6.10+ Signed-off-by: Nikhil Gurudasani Reviewed-by: Gao Xiang Reviewed-by: Chao Yu Signed-off-by: Gao Xiang

erofs: support large folios in inode_share mode

2026-08-24T04:20:57Z

erofs_fill_inode() calls mapping_set_large_folios() on each erofs file's own mapping, but in inode_share mode reads are routed to the shared inode's page cache instead, whose mapping never had large folio support enabled. Set up large folios for the shared inode's mapping as well. Signed-off-by: Jingbo Xu Reviewed-by: Gao Xiang Reviewed-by: Chao Yu Signed-off-by: Gao Xiang

erofs: support splice() in inode_share mode

2026-08-24T04:20:41Z

erofs_ishare_fops routes everything that touches the page cache to the backing file in ->private_data: read_iter clones the iocb onto it, mmap does vma_set_file(), fadvise calls vfs_fadvise() on it. splice_read was left as filemap_splice_read(), which works on the user file's own mapping: it does init_sync_kiocb(&iocb, in), and filemap_get_pages() then takes iocb->ki_filp->f_mapping. So splice() and sendfile() fill the per-inode page cache rather than the shared one. The content is the same either way, since erofs_fill_inode() sets a_ops on that mapping too, which is why this went unnoticed. Two identical 8 MiB files under inode_share, reading one of them with splice(2) alone, in pages: before own 2048 shared 0 after own 0 shared 2048 "own" is cachestat(fd), which reports the file's own mapping; "shared" is mmap()+mincore(), which erofs_ishare_mmap() redirects to the backing file. Read through the backing file, as read_iter already does. Link: https://lore.kernel.org/all/b7dc7192-d586-45a2-bc4a-b41dc681c9bb@linux.alibaba.com/ Signed-off-by: Zhan Xusheng Reviewed-by: Jingbo Xu Reviewed-by: Chao Yu Signed-off-by: Gao Xiang