Re: [PATCH v2 11/15] lib/lz4: switch the HC compressor to the vendored sources

[sashiko-bot@kernel.org]

> Replace the forked LZ4HC with thin entry points over the vendored
> lz4hc.c.
> 
> Upstream routes levels 10 and up to its optimal parser, whose ~64K
> workspace the kernel cannot provide. Clamp those levels to 9 rather than
> reject them, so existing f2fs and zram settings keep working, and add
> LZ4HC_CLAMP_CLEVEL to <linux/lz4.h>. Document the clamp in f2fs.rst.
> 
> Levels 1 and 2 now use upstream's lz4mid: on kernel source they are
> 50-84% faster for 4-9% larger output. Levels 3 to 9 compress the same,
> but at 4K are up to 15% slower on poorly compressible data, including at
> level 9, zram's lz4hc default.
> 
> LZ4_compress_HC() and LZ4_resetStreamHC() check the state's alignment
> like the compressor's entry points.
> 
> LZ4_streamHC_t becomes incomplete and the macros that laid out the forked
> structure go. LZ4HC_MEM_COMPRESS grows from 262192 to upstream's 262200,
> static_asserted against LZ4_STREAMHC_MINSIZE. lz4hc.c pulls in lz4.c's
> decoder tables, unused here when the fast decode loop is off, so silence
> -Wunused-const-variable for this object.
> 
> This retires commit b08918fb3f27 ("lz4: do not export static symbol"):
> the glue exports only the five LZ4HC entry points.
> 
> Signed-off-by: Michal Wilczynski <m.wilczynski@samsung.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261106-lz4-vendor-upstream-v2-0-ee04c3f0d7dc@samsung.com?part=11