[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <2a39ced1-05f1-d696-1905-4ff5199fa41b@bytedance.com>
Date: Wed, 29 Mar 2023 16:08:49 +0800
From: Jia Zhu <zhujia.zj@...edance.com>
To: David Howells <dhowells@...hat.com>
Cc: linux-cachefs@...hat.com, linux-erofs@...ts.ozlabs.org,
linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
Xin Yin <yinxin.x@...edance.com>,
Jingbo Xu <jefflexu@...ux.alibaba.com>, zhujia.zj@...edance.com
Subject: Re: [External] Re: [PATCH V4 4/5] cachefiles: narrow the scope of
triggering EPOLLIN events in ondemand mode
在 2023/3/28 22:19, David Howells 写道:
> Jia Zhu <zhujia.zj@...edance.com> wrote:
>
>> + if (!xa_empty(xa)) {
>> + xa_lock(xa);
>> + xa_for_each_marked(xa, index, req, CACHEFILES_REQ_NEW) {
>> + if (!cachefiles_ondemand_is_reopening_read(req)) {
>> + mask |= EPOLLIN;
>> + break;
>> + }
>> + }
>> + xa_unlock(xa);
>> + }
>
> I wonder if there's a more efficient way to do this. I guess it depends on
> how many reqs you expect to get in a queue. It might be worth taking the
> rcu_read_lock before calling xa_lock() and holding it over the whole loop.
>
Thanks for the advice, will use rcu_read_lock(unlock) to replace it.
> David
>
Powered by blists - more mailing lists