RE: [syzbot] WARNING in kthread_bind_mask

[Zhang, Qiang1]

On Sun, Feb 20, 2022 at 10:27:23AM -0800, syzbot wrote:
> Hello,
> 
> syzbot found the following issue on:
> 
> HEAD commit:    c5d9ae265b10 Merge tag 'for-linus' of git://git.kernel.org..
> git tree:       upstream
> console output: 
> https://syzkaller.appspot.com/x/log.txt?x=11daf74a700000
> kernel config:  
> https://syzkaller.appspot.com/x/.config?x=da674567f7b6043d
> dashboard link: https://syzkaller.appspot.com/bug?extid=087b7effddeec0697c66
> compiler:       gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2
> 
> Unfortunately, I don't have any reproducer for this issue yet.
> 
> IMPORTANT: if you fix the issue, please add the following tag to the commit:
> Reported-by: syzbot+087b7effddeec0697c66@xxxxxxxxxxxxxxxxxxxxxxxxx
> 
> BTRFS info (device loop3): disk space caching is enabled BTRFS info 
> (device loop3): has skinny extents ------------[ cut here 
> ]------------
> WARNING: CPU: 0 PID: 10327 at kernel/kthread.c:525 __kthread_bind_mask 
> kernel/kthread.c:525 [inline]
>
> 520 static void __kthread_bind_mask(struct task_struct *p, const struct cpumask *mask, unsigned int state)
> 521 {
> 522         unsigned long flags;
> 523
> 524         if (!wait_task_inactive(p, state)) {
> 525                 WARN_ON(1);
> 526                 return;
> 527         }
>

Maybe we can add some additional debugging information  to view the status of the process.

diff --git a/kernel/kthread.c b/kernel/kthread.c
index 38c6dd822da8..e707e86ee64b 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -29,7 +29,7 @@
 #include <linux/numa.h>
 #include <linux/sched/isolation.h>
 #include <trace/events/sched.h>
-
+#include <linux/sched/debug.h>

 static DEFINE_SPINLOCK(kthread_create_lock);
 static LIST_HEAD(kthread_create_list);
@@ -521,8 +521,8 @@ static void __kthread_bind_mask(struct task_struct *p, const struct cpumask *mas
 {
        unsigned long flags;

-       if (!wait_task_inactive(p, state)) {
-               WARN_ON(1);
+       if (WARN_ON(!wait_task_inactive(p, state))) {
+               sched_show_task(p);
                return;
        }

Thanks,
Zqiang

>That seems to be some internal task state inconsistency.