On Tue, Jun 20, 2023 at 11:23 PM Gao Xiang <hsiang...@linux.alibaba.com> wrote: > > > > On 2023/6/21 14:18, Sandeep Dhavale wrote: > > On Tue, Jun 20, 2023 at 5:38 PM Gao Xiang <hsiang...@linux.alibaba.com> > > wrote: > > ... > > >> > > I think this looks good. rcu_read_lock_any_held() can detect this. > > Okay, could you submit a formal patch with detailed background in the > commit message (also why this original optimization is important -- > because we could reuse dm-verity workqueue to avoid scheduling > another workqueue which is bad for app performance)? > Hi Gao, Yes, I will send the formal patch. Can you please expand a bit more on what your suggestions are regarding reusing dm-verity workqueue? Because AFAICT we will only schedule to erofs workqueue or percpu kthread workers only if it is atomic context, in other words, we cannot do the decompression in the current context. Or maybe I misunderstood your comment.
Thanks, Sandeep. > Thanks, > Gao Xiang
