Re: [PATCH v7 2/4] kernfs: Change kernfs_notify_list to llist.

[Imran Khan <imran.f.khan@oracle.com>]

Hello Marek,


On 1/7/22 9:22 pm, Marek Szyprowski wrote:
> Hi All,
> 
> On 15.06.2022 04:10, Imran Khan wrote:
>> At present kernfs_notify_list is implemented as a singly linked
>> list of kernfs_node(s), where last element points to itself and
>> value of ->attr.next tells if node is present on the list or not.
>> Both addition and deletion to list happen under kernfs_notify_lock.
>>
>> Change kernfs_notify_list to llist so that addition to list can heppen
>> locklessly.
>>
>> Suggested by: Al Viro <viro@zeniv.linux.org.uk>
>> Signed-off-by: Imran Khan <imran.f.khan@oracle.com>
>> Acked-by: Tejun Heo <tj@kernel.org>
> 
> This patch landed in linux next-20220630 as commit b8f35fa1188b 
> ("kernfs: Change kernfs_notify_list to llist."). Unfortunately, it 
> causes serious regression on my test systems. It can be easily noticed 
> in the logs by the following warning:
>
> ------------[ cut here ]------------
> WARNING: CPU: 1 PID: 34 at fs/kernfs/dir.c:531 kernfs_put.part.0+0x1a4/0x1d8
> kernfs_put: console/active: released with incorrect active_ref 0
> Modules linked in:
> CPU: 1 PID: 34 Comm: kworker/1:4 Not tainted 
> 5.19.0-rc4-05465-g5732b42edfd1 #12317
> Hardware name: Samsung Exynos (Flattened Device Tree)
> Workqueue: events kernfs_notify_workfn
>   unwind_backtrace from show_stack+0x10/0x14
>   show_stack from dump_stack_lvl+0x40/0x4c
>   dump_stack_lvl from __warn+0xc8/0x13c
>   __warn from warn_slowpath_fmt+0x90/0xb4
>   warn_slowpath_fmt from kernfs_put.part.0+0x1a4/0x1d8
>   kernfs_put.part.0 from kernfs_notify_workfn+0x1a0/0x1d0
>   kernfs_notify_workfn from process_one_work+0x1ec/0x4cc
>   process_one_work from worker_thread+0x58/0x54c
>   worker_thread from kthread+0xd0/0xec
>   kthread from ret_from_fork+0x14/0x2c
> Exception stack(0xf099dfb0 to 0xf099dff8)
> ...
> ---[ end trace 0000000000000000 ]---
> 

Thanks for reporting this issue. It has been reported earlier in [1] as well. I
am unable to reproduce it locally. Could you please test with following patch on
top of linux next-20220630 and let me know if it helps:


From 6bf7f1adc4b091dc6d6c60e0dd0f16247f61f374 Mon Sep 17 00:00:00 2001
From: Imran Khan <imran.f.khan@oracle.com>
Date: Fri, 1 Jul 2022 14:27:52 +1000
Subject: [PATCH] kernfs: Avoid re-adding kernfs_node into kernfs_notify_list.

---
 fs/kernfs/file.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/fs/kernfs/file.c b/fs/kernfs/file.c
index bb933221b4bae..e8ec054e11c63 100644
--- a/fs/kernfs/file.c
+++ b/fs/kernfs/file.c
@@ -917,6 +917,7 @@ static void kernfs_notify_workfn(struct work_struct *work)
        if (free == NULL)
                return;

+       free->next = NULL;
        attr = llist_entry(free, struct kernfs_elem_attr, notify_next);
        kn = attribute_to_node(attr, struct kernfs_node, attr);
        root = kernfs_root(kn);
@@ -992,9 +993,11 @@ void kernfs_notify(struct kernfs_node *kn)
        rcu_read_unlock();

        /* schedule work to kick fsnotify */
-       kernfs_get(kn);
-       llist_add(&kn->attr.notify_next, &kernfs_notify_list);
-       schedule_work(&kernfs_notify_work);
+       if (kn->attr.notify_next.next != NULL) {
+               kernfs_get(kn);
+               llist_add(&kn->attr.notify_next, &kernfs_notify_list);
+               schedule_work(&kernfs_notify_work);
+       }
 }
 EXPORT_SYMBOL_GPL(kernfs_notify);


base-commit: 6cc11d2a1759275b856e464265823d94aabd5eaf
-- 
2.30.2

Thanks
-- Imran

[1]: https://lore.kernel.org/lkml/0b3c1362-5f0d-44d5-d3e7-c01de59a4e86@oracle.com/