Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Dmitry Rokosov <DDRokosov@sberdevices.ru>
To:     Marc Zyngier <maz@kernel.org>
CC:     "broonie@kernel.org" <broonie@kernel.org>,
        "gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
        "rafael@kernel.org" <rafael@kernel.org>,
        "jic23@kernel.org" <jic23@kernel.org>,
        "linux@rasmusvillemoes.dk>" <linux@rasmusvillemoes.dk>,
        "tglx@linutronix.de" <tglx@linutronix.de>,
        "andy.shevchenko@gmail.com" <andy.shevchenko@gmail.com>,
        kernel <kernel@sberdevices.ru>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v1] regmap: introduce value tracing for regmap bulk
 operations
Thread-Topic: [PATCH v1] regmap: introduce value tracing for regmap bulk
 operations
Thread-Index: AQHYsZwRTOzZdLxKvkqpBNcAgWIdzq20YtWAgAAaSQCAADYjgA==
Date:   Thu, 18 Aug 2022 17:02:46 +0000
Message-ID: <20220818170305.kxtgpmulyditol6r@CAB-WSD-L081021.sigma.sbrf.ru>
References: <20220816181451.5628-1-ddrokosov@sberdevices.ru>
 <20220818121515.6azkxyqetnunwsc6@CAB-WSD-L081021.sigma.sbrf.ru>
 <87mtc1wtjz.wl-maz@kernel.org>
In-Reply-To: <87mtc1wtjz.wl-maz@kernel.org>
Accept-Language: ru-RU, en-US
Content-Language: en-US
Content-Type: text/plain; charset="us-ascii"
Content-ID: <8A8E25C78C2041488D21CB9D2855371C@sberdevices.ru>
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
Precedence: bulk

Hello Marc,

Thank you for quick feedback. Please find my comments below.

On Thu, Aug 18, 2022 at 02:49:20PM +0100, Marc Zyngier wrote:

[...]

> > > Currently, only one-register io operations support tracepoints with
> > > value logging. For the regmap bulk operations developer can view
> > > hw_start/hw_done tracepoints with starting reg number and registers
> > > count to be reading or writing. This patch injects tracepoints with
> > > dumping registers values in the hex format to regmap bulk reading
> > > and writing.
>=20
> I don't care much about regmap as a MMIO backend, but it strikes me as
> odd that you end up with multiple ways of logging the same stuff (with
> a memcpy in the middle of it).
>=20
> Why can't this be done with a small amount of trace post-processing?

Sorry, actually, I don't get you. What do you mean by "same stuff"?
For now, regmap bulk I/O operations don't log data buffers, because
current regmap trace classes don't have a dynamic trace arrays inside.
We should use dynamic array here because bulk I/O operations can vary
buffer size from call to call.
Function memcpy() is used to copy original buffer data to a trace array
when tracepoint is enabled. In other words, per my understanding,
when tracepoint is disabled we do not call TP_fast_assign instructions.

Trace event documentation says about dynamic array:

 *   __dynamic_array: This is similar to array, but can vary its size from
 *         instance to instance of the tracepoint being called.
 *         Like __array, this too has three elements (type, name, size);
 *         type is the type of the element, name is the name of the array.
 *         The size is different than __array. It is not a static number,
 *         but the algorithm to figure out the length of the array for the
 *         specific instance of tracepoint. Again, size is the number of
 *         items in the array, not the total length in bytes.
 *
 *         __dynamic_array( int, foo, bar) is similar to: int foo[bar];
 *
 *         Note, unlike arrays, you must use the __get_dynamic_array() macr=
o
 *         to access the array.
 *
 *         memcpy(__get_dynamic_array(foo), bar, 10);
 *
 *         Notice, that "__entry" is not needed here.

BTW, I've tried to use the already existing TRACE CLASS regmap_block, but
it's difficult to integrate dynamic array to that, because sometimes we
do not have a reg data (for example, regmap_hw_read_start event).

[...]

--=20
Thank you,
Dmitry=