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
From:   Masahiro Yamada <yamada.masahiro@socionext.com>
Date:   Fri, 1 Feb 2019 19:27:46 +0900
Message-ID: <CAK7LNAQLzvW-zDaR8qkDaxEXxmiCj7XQDK-G57sPhvZu1_eBdA@mail.gmail.com>
Subject: [Question] setup_data_interface() hook when switching to a different
 type of NAND chip
To:     Boris Brezillon <boris.brezillon@bootlin.com>,
        Miquel Raynal <miquel.raynal@bootlin.com>,
        linux-mtd <linux-mtd@lists.infradead.org>
Cc:     Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi.


When I was looking into the NAND controller/chips separation,
this question popped up in my mind.


Commit 2d472aba15ff169 provides us a more flexibility
about the controller/chips connection.
The connected NAND chips do not need to be homogeneous
any more.


My question is about the ->setup_data_interface() hook
when switching between NAND chips with different speed (timing mode).


Think about the case below:

{
        compatible = "foo-nand-controller";
        reg = <...>;
        #address-cells = <1>;
        #size-cells = <0>;

        nand@0 {
                  reg = <0>;
                  /* Slow NAND chip */
        }

        nand@1 {
                  reg = <1>;
                  /* Fast NAND chip */
        }

}


In this case, two devices /dev/mtdblock0 and /dev/mtdblock1
will appear.

If a user gets access to those two devices in turns,
I think ->setup_data_interface() should be invoked somehow
in order to update the timing registers on the controller side.

Currently, ->setup_data_interface() is invoked in nand_scan_tail()
and that's it.

So, both nand@0 and nand@1 are accessed by the timing mode of nand@1
(assuming nand@0 and nand@1 are initialized in this order)


Of course, it depends on the controller.

If a controller has a register set for every chip select,
the hardware will be able to change the access speed automatically.


I think most of controllers just have a single set of timing registers.
So, when switching between different types of chips,
the driver must update the timing registers.


If this is a worthwhile usecase,
should it be taken care of by the NAND framework,
or by drivers ?


I just thought we could do something in
nand_get_device() / nand_release_device().


If this should be done per driver,
drivers can update registers in select_target or select_chip.


Thought?


-- 
Best Regards
Masahiro Yamada