Re: [PATCH 2/2] erofs: address D-cache aliasing

[Jan Kiszka <jan.kiszka@siemens.com>]

On 09.07.25 05:46, Gao Xiang wrote:
> 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
> Cc: Jan Kiszka <jan.kiszka@siemens.com>
> Cc: Stefan Kerkmann <s.kerkmann@pengutronix.de>
> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
> ---
> Hi Jan and Stefan,
> 
> if possible, please help test this patch on your arm devices,
> many thanks!  I will submit this later but if it's urgent you
> could also apply this locally in advance.
> 
> Thanks,
> Gao Xiang
>  fs/erofs/data.c         | 16 +++++++++++-----
>  fs/erofs/decompressor.c | 12 ++++--------
>  fs/erofs/fileio.c       |  4 ++--
>  fs/erofs/internal.h     |  2 +-
>  fs/erofs/zdata.c        |  6 +++---
>  5 files changed, 21 insertions(+), 19 deletions(-)
> 
> diff --git a/fs/erofs/data.c b/fs/erofs/data.c
> index 221e0ff1ed0d..16e4a6bd9b97 100644
> --- a/fs/erofs/data.c
> +++ b/fs/erofs/data.c
> @@ -214,9 +214,11 @@ int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *map)
>  
>  /*
>   * bit 30: I/O error occurred on this folio
> + * bit 29: CPU has dirty data in D-cache (needs aliasing handling);
>   * bit 0 - 29: remaining parts to complete this folio
>   */
> -#define EROFS_ONLINEFOLIO_EIO			(1 << 30)
> +#define EROFS_ONLINEFOLIO_EIO		30
> +#define EROFS_ONLINEFOLIO_DIRTY		29
>  
>  void erofs_onlinefolio_init(struct folio *folio)
>  {
> @@ -233,19 +235,23 @@ void erofs_onlinefolio_split(struct folio *folio)
>  	atomic_inc((atomic_t *)&folio->private);
>  }
>  
> -void erofs_onlinefolio_end(struct folio *folio, int err)
> +void erofs_onlinefolio_end(struct folio *folio, int err, bool dirty)
>  {
>  	int orig, v;
>  
>  	do {
>  		orig = atomic_read((atomic_t *)&folio->private);
> -		v = (orig - 1) | (err ? EROFS_ONLINEFOLIO_EIO : 0);
> +		DBG_BUGON(orig <= 0);
> +		v = dirty << EROFS_ONLINEFOLIO_DIRTY;
> +		v |= (orig - 1) | (!!err << EROFS_ONLINEFOLIO_EIO);
>  	} while (atomic_cmpxchg((atomic_t *)&folio->private, orig, v) != orig);
>  
> -	if (v & ~EROFS_ONLINEFOLIO_EIO)
> +	if (v & (BIT(EROFS_ONLINEFOLIO_DIRTY) - 1))
>  		return;
>  	folio->private = 0;
> -	folio_end_read(folio, !(v & EROFS_ONLINEFOLIO_EIO));
> +	if (v & BIT(EROFS_ONLINEFOLIO_DIRTY))
> +		flush_dcache_folio(folio);
> +	folio_end_read(folio, !(v & BIT(EROFS_ONLINEFOLIO_EIO)));
>  }
>  
>  static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
> diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
> index bf62e2836b60..358061d7b660 100644
> --- a/fs/erofs/decompressor.c
> +++ b/fs/erofs/decompressor.c
> @@ -301,13 +301,11 @@ static int z_erofs_transform_plain(struct z_erofs_decompress_req *rq,
>  		cur = min(cur, rq->outputsize);
>  		if (cur && rq->out[0]) {
>  			kin = kmap_local_page(rq->in[nrpages_in - 1]);
> -			if (rq->out[0] == rq->in[nrpages_in - 1]) {
> +			if (rq->out[0] == rq->in[nrpages_in - 1])
>  				memmove(kin + rq->pageofs_out, kin + pi, cur);
> -				flush_dcache_page(rq->out[0]);
> -			} else {
> +			else
>  				memcpy_to_page(rq->out[0], rq->pageofs_out,
>  					       kin + pi, cur);
> -			}
>  			kunmap_local(kin);
>  		}
>  		rq->outputsize -= cur;
> @@ -325,14 +323,12 @@ static int z_erofs_transform_plain(struct z_erofs_decompress_req *rq,
>  			po = (rq->pageofs_out + cur + pi) & ~PAGE_MASK;
>  			DBG_BUGON(no >= nrpages_out);
>  			cnt = min(insz - pi, PAGE_SIZE - po);
> -			if (rq->out[no] == rq->in[ni]) {
> +			if (rq->out[no] == rq->in[ni])
>  				memmove(kin + po,
>  					kin + rq->pageofs_in + pi, cnt);
> -				flush_dcache_page(rq->out[no]);
> -			} else if (rq->out[no]) {
> +			else if (rq->out[no])
>  				memcpy_to_page(rq->out[no], po,
>  					       kin + rq->pageofs_in + pi, cnt);
> -			}
>  			pi += cnt;
>  		} while (pi < insz);
>  		kunmap_local(kin);
> diff --git a/fs/erofs/fileio.c b/fs/erofs/fileio.c
> index fe2cd2982b4b..91781718199e 100644
> --- a/fs/erofs/fileio.c
> +++ b/fs/erofs/fileio.c
> @@ -38,7 +38,7 @@ static void erofs_fileio_ki_complete(struct kiocb *iocb, long ret)
>  	} else {
>  		bio_for_each_folio_all(fi, &rq->bio) {
>  			DBG_BUGON(folio_test_uptodate(fi.folio));
> -			erofs_onlinefolio_end(fi.folio, ret);
> +			erofs_onlinefolio_end(fi.folio, ret, false);
>  		}
>  	}
>  	bio_uninit(&rq->bio);
> @@ -154,7 +154,7 @@ static int erofs_fileio_scan_folio(struct erofs_fileio *io, struct folio *folio)
>  		}
>  		cur += len;
>  	}
> -	erofs_onlinefolio_end(folio, err);
> +	erofs_onlinefolio_end(folio, err, false);
>  	return err;
>  }
>  
> diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
> index a32c03a80c70..0d19bde8c094 100644
> --- a/fs/erofs/internal.h
> +++ b/fs/erofs/internal.h
> @@ -390,7 +390,7 @@ int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
>  int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map);
>  void erofs_onlinefolio_init(struct folio *folio);
>  void erofs_onlinefolio_split(struct folio *folio);
> -void erofs_onlinefolio_end(struct folio *folio, int err);
> +void erofs_onlinefolio_end(struct folio *folio, int err, bool dirty);
>  struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
>  int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
>  		  struct kstat *stat, u32 request_mask,
> diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
> index d80e3bf4fa79..6f8402ed5b28 100644
> --- a/fs/erofs/zdata.c
> +++ b/fs/erofs/zdata.c
> @@ -1090,7 +1090,7 @@ static int z_erofs_scan_folio(struct z_erofs_frontend *f,
>  			tight = (bs == PAGE_SIZE);
>  		}
>  	} while ((end = cur) > 0);
> -	erofs_onlinefolio_end(folio, err);
> +	erofs_onlinefolio_end(folio, err, false);
>  	return err;
>  }
>  
> @@ -1195,7 +1195,7 @@ static void z_erofs_fill_other_copies(struct z_erofs_backend *be, int err)
>  			cur += len;
>  		}
>  		kunmap_local(dst);
> -		erofs_onlinefolio_end(page_folio(bvi->bvec.page), err);
> +		erofs_onlinefolio_end(page_folio(bvi->bvec.page), err, true);
>  		list_del(p);
>  		kfree(bvi);
>  	}
> @@ -1353,7 +1353,7 @@ static int z_erofs_decompress_pcluster(struct z_erofs_backend *be, int err)
>  
>  		DBG_BUGON(z_erofs_page_is_invalidated(page));
>  		if (!z_erofs_is_shortlived_page(page)) {
> -			erofs_onlinefolio_end(page_folio(page), err);
> +			erofs_onlinefolio_end(page_folio(page), err, true);
>  			continue;
>  		}
>  		if (pcl->algorithmformat != Z_EROFS_COMPRESSION_LZ4) {

Tested-by: Jan Kiszka <jan.kiszka@siemens.com>

Please make sure to address stable versions as well once this is merged.
I've so far only checked 6.12 (besides top of tree) where this applies
as-is. Other versions likely need extra effort.

Jan

-- 
Siemens AG, Foundational Technologies
Linux Expert Center