[PATCH 6.6.y] erofs: address D-cache aliasing

[Gao Xiang <hsiangkao@linux.alibaba.com>]

commit 27917e8194f91dffd8b4825350c63cb68e98ce58 upstream.

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 <jan.kiszka@siemens.com>
Tested-by: Stefan Kerkmann <s.kerkmann@pengutronix.de>
Link: https://lore.kernel.org/r/20250709034614.2780117-2-hsiangkao@linux.alibaba.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
Hi Jan & Stefan,
Please help confirm this 6.6 fix backport if possible.

Thanks,
Gao Xiang

 fs/erofs/decompressor.c |  6 ++----
 fs/erofs/zdata.c        | 32 +++++++++++++++++++-------------
 2 files changed, 21 insertions(+), 17 deletions(-)

diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index aa59788a61e6..86e088fd386e 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -342,14 +342,12 @@ static int z_erofs_transform_plain(struct z_erofs_decompress_req *rq,
 
 	if (outpages > inpages) {
 		DBG_BUGON(!rq->out[outpages - 1]);
-		if (rq->out[outpages - 1] != rq->in[inpages - 1]) {
+		if (rq->out[outpages - 1] != rq->in[inpages - 1])
 			memcpy_to_page(rq->out[outpages - 1], 0, src +
 					(interlaced_offset ? 0 : righthalf),
 				       lefthalf);
-		} else if (!interlaced_offset) {
+		else if (!interlaced_offset)
 			memmove(src, src + righthalf, lefthalf);
-			flush_dcache_page(rq->in[inpages - 1]);
-		}
 	}
 	kunmap_local(src);
 	return 0;
diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 496e4c7c52a4..d852b43ac43e 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -122,9 +122,11 @@ static inline unsigned int z_erofs_pclusterpages(struct z_erofs_pcluster *pcl)
 
 /*
  * bit 30: I/O error occurred on this page
+ * bit 29: CPU has dirty data in D-cache (needs aliasing handling);
  * bit 0 - 29: remaining parts to complete this page
  */
-#define Z_EROFS_PAGE_EIO			(1 << 30)
+#define Z_EROFS_ONLINEPAGE_EIO		30
+#define Z_EROFS_ONLINEPAGE_DIRTY	29
 
 static inline void z_erofs_onlinepage_init(struct page *page)
 {
@@ -143,7 +145,7 @@ static inline void z_erofs_onlinepage_split(struct page *page)
 	atomic_inc((atomic_t *)&page->private);
 }
 
-static void z_erofs_onlinepage_endio(struct page *page, int err)
+static void z_erofs_onlinepage_end(struct page *page, int err, bool dirty)
 {
 	int orig, v;
 
@@ -151,16 +153,20 @@ static void z_erofs_onlinepage_endio(struct page *page, int err)
 
 	do {
 		orig = atomic_read((atomic_t *)&page->private);
-		v = (orig - 1) | (err ? Z_EROFS_PAGE_EIO : 0);
+		DBG_BUGON(orig <= 0);
+		v = dirty << Z_EROFS_ONLINEPAGE_DIRTY;
+		v |= (orig - 1) | (!!err << Z_EROFS_ONLINEPAGE_EIO);
 	} while (atomic_cmpxchg((atomic_t *)&page->private, orig, v) != orig);
 
-	if (!(v & ~Z_EROFS_PAGE_EIO)) {
-		set_page_private(page, 0);
-		ClearPagePrivate(page);
-		if (!(v & Z_EROFS_PAGE_EIO))
-			SetPageUptodate(page);
-		unlock_page(page);
-	}
+	if (v & (BIT(Z_EROFS_ONLINEPAGE_DIRTY) - 1))
+		return;
+	set_page_private(page, 0);
+	ClearPagePrivate(page);
+	if (v & BIT(Z_EROFS_ONLINEPAGE_DIRTY))
+		flush_dcache_page(page);
+	if (!(v & BIT(Z_EROFS_ONLINEPAGE_EIO)))
+		SetPageUptodate(page);
+	unlock_page(page);
 }
 
 #define Z_EROFS_ONSTACK_PAGES		32
@@ -1060,7 +1066,7 @@ static int z_erofs_do_read_page(struct z_erofs_decompress_frontend *fe,
 		goto repeat;
 
 out:
-	z_erofs_onlinepage_endio(page, err);
+	z_erofs_onlinepage_end(page, err, false);
 	return err;
 }
 
@@ -1163,7 +1169,7 @@ static void z_erofs_fill_other_copies(struct z_erofs_decompress_backend *be,
 			cur += len;
 		}
 		kunmap_local(dst);
-		z_erofs_onlinepage_endio(bvi->bvec.page, err);
+		z_erofs_onlinepage_end(bvi->bvec.page, err, true);
 		list_del(p);
 		kfree(bvi);
 	}
@@ -1333,7 +1339,7 @@ static int z_erofs_decompress_pcluster(struct z_erofs_decompress_backend *be,
 		/* recycle all individual short-lived pages */
 		if (z_erofs_put_shortlivedpage(be->pagepool, page))
 			continue;
-		z_erofs_onlinepage_endio(page, err);
+		z_erofs_onlinepage_end(page, err, true);
 	}
 
 	if (be->decompressed_pages != be->onstack_pages)
-- 
2.43.5