linux.git/fs/squashfs, branch v6.16

squashfs: add optional full compressed block caching

2025-05-28T02:40:33+00:00

The commit 93e72b3c612adcaca1 ("squashfs: migrate from ll_rw_block usage
to BIO") removed caching of compressed blocks in SquashFS, causing fio
performance regression in workloads with repeated file reads.  Without
caching, every read triggers disk I/O, severely impacting performance in
tools like fio.

This patch introduces a new CONFIG_SQUASHFS_COMP_CACHE_FULL Kconfig option
to enable caching of all compressed blocks, restoring performance to
pre-BIO migration levels.  When enabled, all pages in a BIO are cached in
the page cache, reducing disk I/O for repeated reads.  The fio test
results with this patch confirm the performance restoration:

For example, fio tests (iodepth=1, numjobs=1,
ioengine=psync) show a notable performance restoration:

Disable CONFIG_SQUASHFS_COMP_CACHE_FULL:
  IOPS=815, BW=102MiB/s (107MB/s)(6113MiB/60001msec)
Enable CONFIG_SQUASHFS_COMP_CACHE_FULL:
  IOPS=2223, BW=278MiB/s (291MB/s)(16.3GiB/59999msec)

The tradeoff is increased memory usage due to caching all compressed
blocks.  The CONFIG_SQUASHFS_COMP_CACHE_FULL option allows users to enable
this feature selectively, balancing performance and memory usage for
workloads with frequent repeated reads.

Link: https://lkml.kernel.org/r/20250521072559.2389-1-chanho.min@lge.com
Signed-off-by: Chanho Min 
Reviewed-by Phillip Lougher 
Signed-off-by: Andrew Morton

Squashfs: check return result of sb_min_blocksize

2025-05-12T00:54:07+00:00

Syzkaller reports an "UBSAN: shift-out-of-bounds in squashfs_bio_read" bug.

Syzkaller forks multiple processes which after mounting the Squashfs
filesystem, issues an ioctl("/dev/loop0", LOOP_SET_BLOCK_SIZE, 0x8000). 
Now if this ioctl occurs at the same time another process is in the
process of mounting a Squashfs filesystem on /dev/loop0, the failure
occurs.  When this happens the following code in squashfs_fill_super()
fails.

----
msblk->devblksize = sb_min_blocksize(sb, SQUASHFS_DEVBLK_SIZE);
msblk->devblksize_log2 = ffz(~msblk->devblksize);
----

sb_min_blocksize() returns 0, which means msblk->devblksize is set to 0.

As a result, ffz(~msblk->devblksize) returns 64, and msblk->devblksize_log2
is set to 64.

This subsequently causes the

UBSAN: shift-out-of-bounds in fs/squashfs/block.c:195:36
shift exponent 64 is too large for 64-bit type 'u64' (aka
'unsigned long long')

This commit adds a check for a 0 return by sb_min_blocksize().

Link: https://lkml.kernel.org/r/20250409024747.876480-1-phillip@squashfs.org.uk
Fixes: 0aa666190509 ("Squashfs: super block operations")
Reported-by: syzbot+65761fc25a137b9c8c6e@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/67f0dd7a.050a0220.0a13.0230.GAE@google.com/
Signed-off-by: Phillip Lougher 
Signed-off-by: Andrew Morton

squashfs: fix invalid pointer dereference in squashfs_cache_delete

2025-03-17T00:40:24+00:00

When mounting a squashfs fails, squashfs_cache_init() may return an error
pointer (e.g., -ENOMEM) instead of NULL.  However, squashfs_cache_delete()
only checks for a NULL cache, and attempts to dereference the invalid
pointer.  This leads to a kernel crash (BUG: unable to handle kernel
paging request in squashfs_cache_delete).

This patch fixes the issue by checking IS_ERR(cache) before accessing it.

Link: https://lkml.kernel.org/r/20250306132855.2030-1-zhiyuzhang999@gmail.com
Fixes: 49ff29240ebb ("squashfs: make squashfs_cache_init() return ERR_PTR(-ENOMEM)")
Signed-off-by: Zhiyu Zhang 
Reported-by: Zhiyu Zhang 
Closes: https://lore.kernel.org/linux-fsdevel/CALf2hKvaq8B4u5yfrE+BYt7aNguao99mfWxHngA+=o5hwzjdOg@mail.gmail.com/
Tested-by: Zhiyu Zhang 
Reviewed-by: Phillip Lougher 
Signed-off-by: Andrew Morton

squashfs: convert squashfs_fill_page() to take a folio

2025-01-25T06:47:22+00:00

squashfs_fill_page is only used in this file, so make it static.
Use kmap_local instead of kmap_atomic, and return a bool so that
the caller can use folio_end_read() which saves an atomic operation
over calling folio_mark_uptodate() followed by folio_unlock().

[willy@infradead.org: fix polarity of "uptodate" Thanks to Ryan for testing]
  Link: https://lkml.kernel.org/r/20250110163300.3346321-2-willy@infradead.org
Link: https://lkml.kernel.org/r/20241220224634.723899-5-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) 
Tested-by: Ryan Roberts 
Cc: Phillip Lougher 
Signed-off-by: Andrew Morton

squashfs; convert squashfs_copy_cache() to take a folio

2025-01-25T06:47:22+00:00

Remove accesses to page->index and page->mapping.  Also use folio
APIs where available.  This code still assumes order 0 folios.

[dan.carpenter@linaro.org: fix a NULL vs IS_ERR() bug]
  Link: https://lkml.kernel.org/r/7b7f44d6-9153-4d7c-b65b-2d78febe6c7a@stanley.mountain
Link: https://lkml.kernel.org/r/20241220224634.723899-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) 
Signed-off-by: Dan Carpenter 
Cc: Phillip Lougher 
Cc: Dan Carpenter 
Signed-off-by: Andrew Morton

squashfs: convert squashfs_readpage_block() to take a folio

2025-01-25T06:47:22+00:00

Remove a few accesses to page->mapping.

Link: https://lkml.kernel.org/r/20241220224634.723899-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) 
Cc: Phillip Lougher 
Signed-off-by: Andrew Morton

squashfs: pass a folio to squashfs_readpage_fragment()

2025-01-25T06:47:22+00:00

Remove an access to page->mapping.

Link: https://lkml.kernel.org/r/20241220224634.723899-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) 
Cc: Phillip Lougher 
Signed-off-by: Andrew Morton

squashfs: use a folio throughout squashfs_read_folio()

2025-01-25T06:47:22+00:00

Use modern folio APIs where they exist and convert back to struct
page for the internal functions.

Link: https://lkml.kernel.org/r/20241220224634.723899-1-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) 
Cc: Phillip Lougher 
Signed-off-by: Andrew Morton

squashfs: update Kconfig information

2025-01-25T06:47:22+00:00

Update the compression algorithms supported, and the Squashfs website
location.

Link: https://lkml.kernel.org/r/20241229233752.54481-5-phillip@squashfs.org.uk
Signed-off-by: Phillip Lougher 
Signed-off-by: Andrew Morton

squashfs: don't allocate read_page cache if SQUASHFS_FILE_DIRECT configured

2025-01-25T06:47:21+00:00

If Squashfs has been configured to directly read datablocks into the page
cache (SQUASHFS_FILE_DIRECT), then the read_page cache is unnecessary.

This improvement is due to the following two commits, which added the
ability to read datablocks into the page cache when pages were missing,
enabling the fallback which used an intermediate buffer to be removed.

commit f268eedddf359 ("squashfs: extend "page actor" to handle missing pages")
commit 1bb1a07afad97 ("squashfs: don't use intermediate buffer if pages missing")

This reduces the amount of memory used when mounting a filesystem by
block_size * maximum number of threads.

Link: https://lkml.kernel.org/r/20241229233752.54481-3-phillip@squashfs.org.uk
Signed-off-by: Phillip Lougher 
Signed-off-by: Andrew Morton