Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210914043928.4066136-1-saravanak@google.com>
 <20210914043928.4066136-6-saravanak@google.com> <YUC7c9BNuHPOEg4g@lunn.ch>
 <CAGETcx-B+yE8X2rn4GdzAikVO-swXb9a5-CxOG_HSuiQoHQZcw@mail.gmail.com> <YUDRSIR2g6Pt1umn@lunn.ch>
In-Reply-To: <YUDRSIR2g6Pt1umn@lunn.ch>
From:   Rob Herring <robh+dt@kernel.org>
Date:   Tue, 14 Sep 2021 11:52:38 -0500
Message-ID: <CAL_Jsq+D2nyzrvWE7vdbQFB680+cmx=FfFMqHL0s=tkTSduz2Q@mail.gmail.com>
Subject: Re: [PATCH v1 5/5] driver core: Add fw_devlink.debug command line
 boolean parameter
To:     Andrew Lunn <andrew@lunn.ch>
Cc:     Saravana Kannan <saravanak@google.com>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        "Rafael J. Wysocki" <rafael@kernel.org>,
        John Stultz <john.stultz@linaro.org>,
        Marek Szyprowski <m.szyprowski@samsung.com>,
        Geert Uytterhoeven <geert@linux-m68k.org>,
        Vladimir Oltean <olteanv@gmail.com>,
        Android Kernel Team <kernel-team@android.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Sep 14, 2021 at 11:44 AM Andrew Lunn <andrew@lunn.ch> wrote:
>
> > I did think of this, but the problem is that dynamic debug logs can
> > get compiled out (when DYNAMIC_DEBUG isn't set). I think debugging of
> > fw_devlink needs to be possible without having to recompile the
> > kernel. So in a sense similar to how initcall_debug works today.
>
> My off the shelf Debian kernel has it enabled. Maybe you can check
> other mainline distributions and see if it is enabled by default.

Right, I would expect users that can't rebuild their kernel easily
would have it enabled.

> You are also on a slippery path. You argue this is needed all the time
> and add a custom knob. Somebody else adds a new feature which they
> also argue always needs bug, and add there own custom knob. We soon
> have lots of custom knobs, each doing it slightly differently, in
> different places. Chaos. So you need a really good argument why your
> code really is special.
>
> I would suggest you start with dynamic debug, and collect some
> statistics of how often you need to ask people to recompile their
> kernel.

I agree.

What would be nice is documenting what needs to be set for devlink.
What I used was just 'dyndbg="file drivers/base/core.c +p"'

Rob