Re: [PATCH v2] fs/fs-writeback: wait isw_nr_in_flight to be zero when umount

[Jiufei Xue <jiufei.xue@linux.alibaba.com>]

Hi Tejun,

On 2019/4/16 下午11:04, Tejun Heo wrote:
> Hello, Jiufei.
> 
> On Tue, Apr 16, 2019 at 08:09:02PM +0800, Jiufei Xue wrote:
>> synchronize_rcu() didn't wait for call_rcu() callbacks, so inode wb
>> switch may not go to the workqueue after synchronize_rcu(). Thus
>> previous scheduled switches was not finished even flushing the
>> workqueue, which will cause a NULL pointer dereferenced followed below.
> 
> Isn't all that's needed replacing the synchronize_rcu() call with a
> rcu_barrier() call?
>

Yes, it can be fixed if we replace synchronize_rcu() with rcu_barrier().
However, I'm worried that rcu_barrier() is too heavyweight and we have
encountered some hung tasks that rcu_barrier() waiting for callbacks that
other drivers queued but not handled correctly.

Thanks,
Jiufei

> Thanks.
>