Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <1549995065-27597-1-git-send-email-xiaoxiang@xiaomi.com>
 <20190212144606.4b7cf0f8@gandalf.local.home> <CAH2Cfb-Zd9TKgs52ROVGYg6F_hiN2ieWKwYPkxpVV8St+zwa2A@mail.gmail.com>
 <20190213062912.GA23233@jagdpanzerIV> <20190213063815.GA32272@jagdpanzerIV>
 <CAH2Cfb8ENK6f1BKvD4oqGhRjQU6cU5ii_OREcPtjwZqP=1qAeQ@mail.gmail.com> <20190213080852.GA1702@jagdpanzerIV>
In-Reply-To: <20190213080852.GA1702@jagdpanzerIV>
From:   xiang xiao <xiaoxiang781216@gmail.com>
Date:   Wed, 13 Feb 2019 16:52:49 +0800
Message-ID: <CAH2Cfb-vv8cAUqpv3vKMiYEg_W=5CGDqZXA9BkCukRONahWoxg@mail.gmail.com>
Subject: Re: [PATCH] printk: add KERN_NOTIME to skip the timestamp
To:     Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc:     Steven Rostedt <rostedt@goodmis.org>,
        Petr Mladek <pmladek@suse.com>,
        Sergey Senozhatsky <sergey.senozhatsky@gmail.com>,
        linux-kernel@vger.kernel.org, Xiang Xiao <xiaoxiang@xiaomi.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Feb 13, 2019 at 4:08 PM Sergey Senozhatsky
<sergey.senozhatsky.work@gmail.com> wrote:
>
> On (02/13/19 15:14), xiang xiao wrote:
> >
> > But how can I precisely control timestamp on/off per message
> > through sysfs node?
> >
>
> Hmm. I don't know how many kernel printk-s you have and how often
> do you write to kmsg.

It depend on use case, but I plan to create a generic driver which
could reuse by all rpmsg based remoteproc.
This is the driver for upstream:
https://github.com/thesofproject/linux/pull/177/commits/a0b7009fede5552dc98733f2996a8140bff62455
so we need more precisely control here.

> I was thinking about something like this:
>

>         echo 0 > /...printk.../time
>         dump buffer to /dev/kmsg
>         echo 1 > /...printk../time
>

Another problem is how to control sysfs node from driver code
naturally, the concurrency and global side effect also need to
address.

> - If you would have several kernel printk-s in the meantime, then
>   those would not have timestamps, but you kinda can roughly guess
>   it
>
>         write [1243] foo > /dev/kmsg
>         write [1244] foo > /dev/kmsg
>                 << printk(bar)               <timestamp ~[1244,1245]>
>         write [1245] foo > /dev/kmsg
>
> Maybe this won't suffice.
>
>         -ss