erofs: fix deadlock when shrink erofs slab

[ Upstream commit 57bbeacdbee72a54eb97d56b876cf9c94059fc34 ]

We observed the following deadlock in the stress test under low
memory scenario:

Thread A                               Thread B
- erofs_shrink_scan
 - erofs_try_to_release_workgroup
  - erofs_workgroup_try_to_freeze -- A
                                       - z_erofs_do_read_page
                                        - z_erofs_collection_begin
                                         - z_erofs_register_collection
                                          - erofs_insert_workgroup
                                           - xa_lock(&sbi->managed_pslots) -- B
                                           - erofs_workgroup_get
                                            - erofs_wait_on_workgroup_freezed -- A
  - xa_erase
   - xa_lock(&sbi->managed_pslots) -- B

To fix this, it needs to hold xa_lock before freezing the workgroup
since xarray will be touched then. So let's hold the lock before
accessing each workgroup, just like what we did with the radix tree
before.

[ Gao Xiang: Jianhua Hao also reports this issue at
  https://lore.kernel.org/r/b10b85df30694bac8aadfe43537c897a@xiaomi.com ]

Link: https://lore.kernel.org/r/20211118135844.3559-1-huangjianan@oppo.com
Fixes: 64094a04414f ("erofs: convert workstn to XArray")
Reviewed-by: Chao Yu <chao@kernel.org>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Huang Jianan <huangjianan@oppo.com>
Reported-by: Jianhua Hao <haojianhua1@xiaomi.com>
Signed-off-by: Gao Xiang <xiang@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

diff --git a/fs/erofs/utils.c b/fs/erofs/utils.c
index bd86067a63f7fc5e2e2310294241e50cf8af8ab2..3ca703cd5b24a5d717125619f76e19dec18f246c 100644 (file)
--- a/fs/erofs/utils.c
+++ b/fs/erofs/utils.c
@@ -141,7 +141,7 @@ static bool erofs_try_to_release_workgroup(struct erofs_sb_info *sbi,
         * however in order to avoid some race conditions, add a
         * DBG_BUGON to observe this in advance.
         */
-       DBG_BUGON(xa_erase(&sbi->managed_pslots, grp->index) != grp);
+       DBG_BUGON(__xa_erase(&sbi->managed_pslots, grp->index) != grp);
 
        /* last refcount should be connected with its managed pslot.  */
        erofs_workgroup_unfreeze(grp, 0);
@@ -156,15 +156,19 @@ static unsigned long erofs_shrink_workstation(struct erofs_sb_info *sbi,
        unsigned int freed = 0;
        unsigned long index;
 
+       xa_lock(&sbi->managed_pslots);
        xa_for_each(&sbi->managed_pslots, index, grp) {
                /* try to shrink each valid workgroup */
                if (!erofs_try_to_release_workgroup(sbi, grp))
                        continue;
+               xa_unlock(&sbi->managed_pslots);
 
                ++freed;
                if (!--nr_shrink)
-                       break;
+                       return freed;
+               xa_lock(&sbi->managed_pslots);
        }
+       xa_unlock(&sbi->managed_pslots);
        return freed;
 }