Fix a recent cleanup patch. noio (bypass) chain is
handled asynchronously against submit chain, therefore
inplace I/O or pagevec cannot be applied to such pages.
Add detailed comment for this as well.
Fixes: 97e86a858bc3 ("staging: erofs: tidy up decompression frontend")
Signed-off-by: Gao Xiang <[email protected]>
---
[ Need stress for more hours on phones to catch this race.
No more smoke out till now on latest mainline for our
stress I/O & memory workloads compared with old versions. ]
fs/erofs/zdata.c | 12 ++++++++++--
1 file changed, 10 insertions(+), 2 deletions(-)
diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 96e34c90f814..fad80c97d247 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -575,7 +575,7 @@ static int z_erofs_do_read_page(struct
z_erofs_decompress_frontend *fe,
struct erofs_map_blocks *const map = &fe->map;
struct z_erofs_collector *const clt = &fe->clt;
const loff_t offset = page_offset(page);
- bool tight = (clt->mode >= COLLECT_PRIMARY_HOOKED);
+ bool tight = true;
enum z_erofs_cache_alloctype cache_strategy;
enum z_erofs_page_type page_type;
@@ -628,8 +628,16 @@ static int z_erofs_do_read_page(struct
z_erofs_decompress_frontend *fe,
preload_compressed_pages(clt, MNGD_MAPPING(sbi),
cache_strategy, pagepool);
- tight &= (clt->mode >= COLLECT_PRIMARY_HOOKED);
hitted:
+ /*
+ * Ensure the current partial page belongs to this submit chain rather
+ * than other concurrent submit chains or the noio(bypass) chain since
+ * those chains are handled asynchronously thus the page cannot be used
+ * for inplace I/O or pagevec (should be processed in strict order.)
+ */
+ tight &= (clt->mode >= COLLECT_PRIMARY_HOOKED &&
+ clt->mode != COLLECT_PRIMARY_FOLLOWED_NOINPLACE);
+
cur = end - min_t(unsigned int, offset + end - map->m_la, end);
if (!(map->m_flags & EROFS_MAP_MAPPED)) {
zero_user_segment(page, cur, end);
--
2.17.1
