Re: [syzbot ci] Re: Eliminate Dying Memory Cgroup
From: Qi Zheng
Date: Tue Oct 28 2025 - 23:12:47 EST
Hi Harry,
On 10/29/25 8:22 AM, Harry Yoo wrote:
On Tue, Oct 28, 2025 at 01:58:33PM -0700, syzbot ci wrote:
syzbot ci has tested the following series
[v1] Eliminate Dying Memory Cgroup
https://lore.kernel.org/all/cover.1761658310.git.zhengqi.arch@xxxxxxxxxxxxx
* [PATCH v1 01/26] mm: memcontrol: remove dead code of checking parent memory cgroup
* [PATCH v1 02/26] mm: workingset: use folio_lruvec() in workingset_refault()
* [PATCH v1 03/26] mm: rename unlock_page_lruvec_irq and its variants
* [PATCH v1 04/26] mm: vmscan: refactor move_folios_to_lru()
* [PATCH v1 05/26] mm: memcontrol: allocate object cgroup for non-kmem case
* [PATCH v1 06/26] mm: memcontrol: return root object cgroup for root memory cgroup
* [PATCH v1 07/26] mm: memcontrol: prevent memory cgroup release in get_mem_cgroup_from_folio()
* [PATCH v1 08/26] buffer: prevent memory cgroup release in folio_alloc_buffers()
* [PATCH v1 09/26] writeback: prevent memory cgroup release in writeback module
* [PATCH v1 10/26] mm: memcontrol: prevent memory cgroup release in count_memcg_folio_events()
* [PATCH v1 11/26] mm: page_io: prevent memory cgroup release in page_io module
* [PATCH v1 12/26] mm: migrate: prevent memory cgroup release in folio_migrate_mapping()
* [PATCH v1 13/26] mm: mglru: prevent memory cgroup release in mglru
* [PATCH v1 14/26] mm: memcontrol: prevent memory cgroup release in mem_cgroup_swap_full()
* [PATCH v1 15/26] mm: workingset: prevent memory cgroup release in lru_gen_eviction()
* [PATCH v1 16/26] mm: thp: prevent memory cgroup release in folio_split_queue_lock{_irqsave}()
* [PATCH v1 17/26] mm: workingset: prevent lruvec release in workingset_refault()
* [PATCH v1 18/26] mm: zswap: prevent lruvec release in zswap_folio_swapin()
* [PATCH v1 19/26] mm: swap: prevent lruvec release in swap module
* [PATCH v1 20/26] mm: workingset: prevent lruvec release in workingset_activation()
* [PATCH v1 21/26] mm: memcontrol: prepare for reparenting LRU pages for lruvec lock
* [PATCH v1 22/26] mm: vmscan: prepare for reparenting traditional LRU folios
* [PATCH v1 23/26] mm: vmscan: prepare for reparenting MGLRU folios
* [PATCH v1 24/26] mm: memcontrol: refactor memcg_reparent_objcgs()
* [PATCH v1 25/26] mm: memcontrol: eliminate the problem of dying memory cgroup for LRU folios
* [PATCH v1 26/26] mm: lru: add VM_WARN_ON_ONCE_FOLIO to lru maintenance helpers
and found the following issue:
WARNING in folio_memcg
Full report is available here:
https://ci.syzbot.org/series/0d48a77a-fb4f-485d-9fd6-086afd6fb650
***
WARNING in folio_memcg
tree: mm-new
URL: https://kernel.googlesource.com/pub/scm/linux/kernel/git/akpm/mm.git
base: b227c04932039bccc21a0a89cd6df50fa57e4716
arch: amd64
compiler: Debian clang version 20.1.8 (++20250708063551+0c9f909b7976-1~exp1~20250708183702.136), Debian LLD 20.1.8
config: https://ci.syzbot.org/builds/503d7034-ae99-44d1-8fb2-62e7ef5e1c7c/config
C repro: https://ci.syzbot.org/findings/880c374a-1b49-436e-9be2-63d5e2c6b6ab/c_repro
syz repro: https://ci.syzbot.org/findings/880c374a-1b49-436e-9be2-63d5e2c6b6ab/syz_repro
exFAT-fs (loop0): failed to load upcase table (idx : 0x00010000, chksum : 0xe5674ec2, utbl_chksum : 0xe619d30d)
exFAT-fs (loop0): failed to load alloc-bitmap
exFAT-fs (loop0): failed to recognize exfat type
------------[ cut here ]------------
WARNING: CPU: 1 PID: 5965 at ./include/linux/memcontrol.h:380 obj_cgroup_memcg include/linux/memcontrol.h:380 [inline]
WARNING: CPU: 1 PID: 5965 at ./include/linux/memcontrol.h:380 folio_memcg+0x148/0x1c0 include/linux/memcontrol.h:434
This is understandable as the code snippet was added fairly recently
and is easy to miss during rebasing.
My mistake, I should recheck it.
#syz test
diff --git a/mm/zswap.c b/mm/zswap.c
index a341814468b9..738d914e5354 100644
--- a/mm/zswap.c
+++ b/mm/zswap.c
@@ -896,11 +896,14 @@ static bool zswap_compress(struct page *page, struct zswap_entry *entry,
* to the active LRU list in the case.
*/
if (comp_ret || !dlen || dlen >= PAGE_SIZE) {
+ rcu_read_lock();
if (!mem_cgroup_zswap_writeback_enabled(
folio_memcg(page_folio(page)))) {
+ rcu_read_unlock();
comp_ret = comp_ret ? comp_ret : -EINVAL;
goto unlock;
}
+ rcu_read_unlock();
comp_ret = 0;
dlen = PAGE_SIZE;
dst = kmap_local_page(page);
LGTM, will do in the next version.
Thanks!