From: Gao Xiang hsiangkao@redhat.com

commit 3c597282887fd55181578996dca52ce697d985a5 upstream.

Hongyu reported "id != index" in z_erofs_onlinepage_fixup() with specific aarch64 environment easily, which wasn't shown before.

After digging into that, I found that high 32 bits of page->private was set to 0xaaaaaaaa rather than 0 (due to z_erofs_onlinepage_init behavior with specific compiler options). Actually we only use low 32 bits to keep the page information since page->private is only 4 bytes on most 32-bit platforms. However z_erofs_onlinepage_fixup() uses the upper 32 bits by mistake.

Let's fix it now.

Reported-and-tested-by: Hongyu Jin hongyu.jin@unisoc.com Fixes: 3883a79abd02 ("staging: erofs: introduce VLE decompression support") Cc: stable@vger.kernel.org # 4.19+ Reviewed-by: Chao Yu yuchao0@huawei.com Link: https://lore.kernel.org/r/20200618234349.22553-1-hsiangkao@aol.com Signed-off-by: Gao Xiang hsiangkao@redhat.com --- This fix has been merged into Linus's tree just now. The 5.4 backport can be trivially applied, though I sent out together with 4.19...

fs/erofs/zdata.h | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/fs/erofs/zdata.h b/fs/erofs/zdata.h index faf950189bd7..568d5a493876 100644 --- a/fs/erofs/zdata.h +++ b/fs/erofs/zdata.h @@ -148,22 +148,22 @@ static inline void z_erofs_onlinepage_init(struct page *page) static inline void z_erofs_onlinepage_fixup(struct page *page, uintptr_t index, bool down) { - unsigned long *p, o, v, id; -repeat: - p = &page_private(page); - o = READ_ONCE(*p); + union z_erofs_onlinepage_converter u = { .v = &page_private(page) }; + int orig, orig_index, val;

- id = o >> Z_EROFS_ONLINEPAGE_INDEX_SHIFT; - if (id) { +repeat: + orig = atomic_read(u.o); + orig_index = orig >> Z_EROFS_ONLINEPAGE_INDEX_SHIFT; + if (orig_index) { if (!index) return;

- DBG_BUGON(id != index); + DBG_BUGON(orig_index != index); }

- v = (index << Z_EROFS_ONLINEPAGE_INDEX_SHIFT) | - ((o & Z_EROFS_ONLINEPAGE_COUNT_MASK) + (unsigned int)down); - if (cmpxchg(p, o, v) != o) + val = (index << Z_EROFS_ONLINEPAGE_INDEX_SHIFT) | + ((orig & Z_EROFS_ONLINEPAGE_COUNT_MASK) + (unsigned int)down); + if (atomic_cmpxchg(u.o, orig, val) != orig) goto repeat; }