2019-12-11 05:48:49

by Masahiro Yamada

[permalink] [raw]
Subject: [PATCH v2 1/2] dt-binding: mtd: denali_dt: document reset property

According to the Denali NAND Flash Memory Controller User's Guide,
this IP has two reset signals.

rst_n: reset most of FFs in the controller core
reg_rst_n: reset all FFs in the register interface, and in the
initialization sequencer

This commit specifies those reset signals.

It is possible to control them separately from the IP point of view
although they might be often tied up together in actual SoC integration.

At least for the upstream platforms, Altera/Intel SOCFPGA and Socionext
UniPhier, the reset controller seems to provide only 1-bit control for
the NAND controller. If it is the case, the resets property should
reference to the same phandles for "nand" and "reg" resets, like this:

resets = <&nand_rst>, <&nand_rst>;
reset-names = "nand", "reg";

Signed-off-by: Masahiro Yamada <[email protected]>
---

Changes in v2:
- Split into two patches

Documentation/devicetree/bindings/mtd/denali-nand.txt | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/Documentation/devicetree/bindings/mtd/denali-nand.txt b/Documentation/devicetree/bindings/mtd/denali-nand.txt
index b32aed1db46d..98916a84bbf6 100644
--- a/Documentation/devicetree/bindings/mtd/denali-nand.txt
+++ b/Documentation/devicetree/bindings/mtd/denali-nand.txt
@@ -14,6 +14,11 @@ Required properties:
interface clock, and the ECC circuit clock.
- clock-names: should contain "nand", "nand_x", "ecc"

+Optional properties:
+ - resets: may contain phandles to the controller core reset, the register
+ reset
+ - reset-names: may contain "nand", "reg"
+
Sub-nodes:
Sub-nodes represent available NAND chips.

@@ -46,6 +51,8 @@ nand: nand@ff900000 {
reg-names = "nand_data", "denali_reg";
clocks = <&nand_clk>, <&nand_x_clk>, <&nand_ecc_clk>;
clock-names = "nand", "nand_x", "ecc";
+ resets = <&nand_rst>, <&nand_reg_rst>;
+ reset-names = "nand", "reg";
interrupts = <0 144 4>;

nand@0 {
--
2.17.1


2019-12-11 05:48:49

by Masahiro Yamada

[permalink] [raw]
Subject: [PATCH v2 2/2] mtd: rawnand: denali_dt: add reset controlling

According to the Denali NAND Flash Memory Controller User's Guide,
this IP has two reset signals.

rst_n: reset most of FFs in the controller core
reg_rst_n: reset all FFs in the register interface, and in the
initialization sequencer

This commit supports controlling those reset signals.

It is possible to control them separately from the IP point of view
although they might be often tied up together in actual SoC integration.

The IP spec says, asserting only the reg_rst_n without asserting rst_n
will cause unpredictable behavior in the controller. So, the driver
deasserts ->rst_reg and ->rst in this order.

Another thing that should be kept in mind is the automated initialization
sequence (a.k.a. 'bootstrap' process) is kicked off when reg_rst_n is
deasserted.

When the reset is deasserted, the controller issues a RESET command
to the chip select 0, and attempts to read out the chip ID, and further
more, ONFI parameters if it is an ONFI-compliant device. Then, the
controller sets up the relevant registers based on the detected
device parameters.

This process might be useful for tiny boot firmware, but is redundant
for Linux Kernel because nand_scan_ident() probes devices and sets up
parameters accordingly. Rather, this hardware feature is annoying
because it ends up with misdetection due to bugs.

So, commit 0615e7ad5d52 ("mtd: nand: denali: remove Toshiba and Hynix
specific fixup code") changed the driver to not rely on it.

However, there is no way to prevent it from running. The IP provides
the 'bootstrap_inhibit_init' port to suppress this sequence, but it is
usually out of software control, and dependent on SoC implementation.
As for the Socionext UniPhier platform, LD4 always enables it. For the
later SoCs, the bootstrap sequence runs depending on the boot mode.

I added usleep_range() to make the driver wait until the sequence
finishes. Otherwise, the driver would fail to detect the chip due
to the race between the driver and hardware-controlled sequence.

Signed-off-by: Masahiro Yamada <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
---

Changes in v2:
- Split into two patches

drivers/mtd/nand/raw/denali_dt.c | 40 +++++++++++++++++++++++++++++++-
1 file changed, 39 insertions(+), 1 deletion(-)

diff --git a/drivers/mtd/nand/raw/denali_dt.c b/drivers/mtd/nand/raw/denali_dt.c
index 8b779a899dcf..9a294c3f6ec8 100644
--- a/drivers/mtd/nand/raw/denali_dt.c
+++ b/drivers/mtd/nand/raw/denali_dt.c
@@ -6,6 +6,7 @@
*/

#include <linux/clk.h>
+#include <linux/delay.h>
#include <linux/err.h>
#include <linux/io.h>
#include <linux/ioport.h>
@@ -14,6 +15,7 @@
#include <linux/of.h>
#include <linux/of_device.h>
#include <linux/platform_device.h>
+#include <linux/reset.h>

#include "denali.h"

@@ -22,6 +24,8 @@ struct denali_dt {
struct clk *clk; /* core clock */
struct clk *clk_x; /* bus interface clock */
struct clk *clk_ecc; /* ECC circuit clock */
+ struct reset_control *rst; /* core reset */
+ struct reset_control *rst_reg; /* register reset */
};

struct denali_dt_data {
@@ -151,6 +155,14 @@ static int denali_dt_probe(struct platform_device *pdev)
if (IS_ERR(dt->clk_ecc))
return PTR_ERR(dt->clk_ecc);

+ dt->rst = devm_reset_control_get_optional_shared(dev, "nand");
+ if (IS_ERR(dt->rst))
+ return PTR_ERR(dt->rst);
+
+ dt->rst_reg = devm_reset_control_get_optional_shared(dev, "reg");
+ if (IS_ERR(dt->rst_reg))
+ return PTR_ERR(dt->rst_reg);
+
ret = clk_prepare_enable(dt->clk);
if (ret)
return ret;
@@ -166,10 +178,30 @@ static int denali_dt_probe(struct platform_device *pdev)
denali->clk_rate = clk_get_rate(dt->clk);
denali->clk_x_rate = clk_get_rate(dt->clk_x);

- ret = denali_init(denali);
+ /*
+ * Deassert the register reset, and the core reset in this order.
+ * Deasserting the core reset while the register reset is asserted
+ * will cause unpredictable behavior in the controller.
+ */
+ ret = reset_control_deassert(dt->rst_reg);
if (ret)
goto out_disable_clk_ecc;

+ ret = reset_control_deassert(dt->rst);
+ if (ret)
+ goto out_assert_rst_reg;
+
+ /*
+ * When the reset is deasserted, the initialization sequence is kicked
+ * (bootstrap process). The driver must wait until it finished.
+ * Otherwise, it will result in unpredictable behavior.
+ */
+ usleep_range(200, 1000);
+
+ ret = denali_init(denali);
+ if (ret)
+ goto out_assert_rst;
+
for_each_child_of_node(dev->of_node, np) {
ret = denali_dt_chip_init(denali, np);
if (ret) {
@@ -184,6 +216,10 @@ static int denali_dt_probe(struct platform_device *pdev)

out_remove_denali:
denali_remove(denali);
+out_assert_rst:
+ reset_control_assert(dt->rst);
+out_assert_rst_reg:
+ reset_control_assert(dt->rst_reg);
out_disable_clk_ecc:
clk_disable_unprepare(dt->clk_ecc);
out_disable_clk_x:
@@ -199,6 +235,8 @@ static int denali_dt_remove(struct platform_device *pdev)
struct denali_dt *dt = platform_get_drvdata(pdev);

denali_remove(&dt->controller);
+ reset_control_assert(dt->rst);
+ reset_control_assert(dt->rst_reg);
clk_disable_unprepare(dt->clk_ecc);
clk_disable_unprepare(dt->clk_x);
clk_disable_unprepare(dt->clk);
--
2.17.1

2019-12-12 01:07:30

by Marek Vasut

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] mtd: rawnand: denali_dt: add reset controlling

On 12/11/19 6:45 AM, Masahiro Yamada wrote:
[...]
> diff --git a/drivers/mtd/nand/raw/denali_dt.c b/drivers/mtd/nand/raw/denali_dt.c
> index 8b779a899dcf..9a294c3f6ec8 100644
> --- a/drivers/mtd/nand/raw/denali_dt.c
> +++ b/drivers/mtd/nand/raw/denali_dt.c
> @@ -6,6 +6,7 @@
> */
>
> #include <linux/clk.h>
> +#include <linux/delay.h>
> #include <linux/err.h>
> #include <linux/io.h>
> #include <linux/ioport.h>
> @@ -14,6 +15,7 @@
> #include <linux/of.h>
> #include <linux/of_device.h>
> #include <linux/platform_device.h>
> +#include <linux/reset.h>
>
> #include "denali.h"
>
> @@ -22,6 +24,8 @@ struct denali_dt {
> struct clk *clk; /* core clock */
> struct clk *clk_x; /* bus interface clock */
> struct clk *clk_ecc; /* ECC circuit clock */
> + struct reset_control *rst; /* core reset */
> + struct reset_control *rst_reg; /* register reset */
> };
>
> struct denali_dt_data {
> @@ -151,6 +155,14 @@ static int denali_dt_probe(struct platform_device *pdev)
> if (IS_ERR(dt->clk_ecc))
> return PTR_ERR(dt->clk_ecc);
>
> + dt->rst = devm_reset_control_get_optional_shared(dev, "nand");
> + if (IS_ERR(dt->rst))
> + return PTR_ERR(dt->rst);
> +
> + dt->rst_reg = devm_reset_control_get_optional_shared(dev, "reg");
> + if (IS_ERR(dt->rst_reg))
> + return PTR_ERR(dt->rst_reg);
> +
> ret = clk_prepare_enable(dt->clk);
> if (ret)
> return ret;
> @@ -166,10 +178,30 @@ static int denali_dt_probe(struct platform_device *pdev)
> denali->clk_rate = clk_get_rate(dt->clk);
> denali->clk_x_rate = clk_get_rate(dt->clk_x);
>
> - ret = denali_init(denali);
> + /*
> + * Deassert the register reset, and the core reset in this order.
> + * Deasserting the core reset while the register reset is asserted
> + * will cause unpredictable behavior in the controller.
> + */
> + ret = reset_control_deassert(dt->rst_reg);
> if (ret)
> goto out_disable_clk_ecc;
>
> + ret = reset_control_deassert(dt->rst);
> + if (ret)
> + goto out_assert_rst_reg;
> +
> + /*
> + * When the reset is deasserted, the initialization sequence is kicked
> + * (bootstrap process). The driver must wait until it finished.
> + * Otherwise, it will result in unpredictable behavior.
> + */
> + usleep_range(200, 1000);
> +
> + ret = denali_init(denali);
> + if (ret)
> + goto out_assert_rst;
> +
> for_each_child_of_node(dev->of_node, np) {
> ret = denali_dt_chip_init(denali, np);
> if (ret) {
> @@ -184,6 +216,10 @@ static int denali_dt_probe(struct platform_device *pdev)
>
> out_remove_denali:
> denali_remove(denali);
> +out_assert_rst:
> + reset_control_assert(dt->rst);
> +out_assert_rst_reg:
> + reset_control_assert(dt->rst_reg);

Maybe you can use devm_add_action_or_reset() here , like in e.g.
drivers/input/touchscreen/ili210x.c , to avoid this unwinding ?

[...]

2019-12-12 03:35:30

by Masahiro Yamada

[permalink] [raw]
Subject: Re: [PATCH v2 2/2] mtd: rawnand: denali_dt: add reset controlling

On Thu, Dec 12, 2019 at 10:05 AM Marek Vasut <[email protected]> wrote:
>
> On 12/11/19 6:45 AM, Masahiro Yamada wrote:
> [...]
> > diff --git a/drivers/mtd/nand/raw/denali_dt.c b/drivers/mtd/nand/raw/denali_dt.c
> > index 8b779a899dcf..9a294c3f6ec8 100644
> > --- a/drivers/mtd/nand/raw/denali_dt.c
> > +++ b/drivers/mtd/nand/raw/denali_dt.c
> > @@ -6,6 +6,7 @@
> > */
> >
> > #include <linux/clk.h>
> > +#include <linux/delay.h>
> > #include <linux/err.h>
> > #include <linux/io.h>
> > #include <linux/ioport.h>
> > @@ -14,6 +15,7 @@
> > #include <linux/of.h>
> > #include <linux/of_device.h>
> > #include <linux/platform_device.h>
> > +#include <linux/reset.h>
> >
> > #include "denali.h"
> >
> > @@ -22,6 +24,8 @@ struct denali_dt {
> > struct clk *clk; /* core clock */
> > struct clk *clk_x; /* bus interface clock */
> > struct clk *clk_ecc; /* ECC circuit clock */
> > + struct reset_control *rst; /* core reset */
> > + struct reset_control *rst_reg; /* register reset */
> > };
> >
> > struct denali_dt_data {
> > @@ -151,6 +155,14 @@ static int denali_dt_probe(struct platform_device *pdev)
> > if (IS_ERR(dt->clk_ecc))
> > return PTR_ERR(dt->clk_ecc);
> >
> > + dt->rst = devm_reset_control_get_optional_shared(dev, "nand");
> > + if (IS_ERR(dt->rst))
> > + return PTR_ERR(dt->rst);
> > +
> > + dt->rst_reg = devm_reset_control_get_optional_shared(dev, "reg");
> > + if (IS_ERR(dt->rst_reg))
> > + return PTR_ERR(dt->rst_reg);
> > +
> > ret = clk_prepare_enable(dt->clk);
> > if (ret)
> > return ret;
> > @@ -166,10 +178,30 @@ static int denali_dt_probe(struct platform_device *pdev)
> > denali->clk_rate = clk_get_rate(dt->clk);
> > denali->clk_x_rate = clk_get_rate(dt->clk_x);
> >
> > - ret = denali_init(denali);
> > + /*
> > + * Deassert the register reset, and the core reset in this order.
> > + * Deasserting the core reset while the register reset is asserted
> > + * will cause unpredictable behavior in the controller.
> > + */
> > + ret = reset_control_deassert(dt->rst_reg);
> > if (ret)
> > goto out_disable_clk_ecc;
> >
> > + ret = reset_control_deassert(dt->rst);
> > + if (ret)
> > + goto out_assert_rst_reg;
> > +
> > + /*
> > + * When the reset is deasserted, the initialization sequence is kicked
> > + * (bootstrap process). The driver must wait until it finished.
> > + * Otherwise, it will result in unpredictable behavior.
> > + */
> > + usleep_range(200, 1000);
> > +
> > + ret = denali_init(denali);
> > + if (ret)
> > + goto out_assert_rst;
> > +
> > for_each_child_of_node(dev->of_node, np) {
> > ret = denali_dt_chip_init(denali, np);
> > if (ret) {
> > @@ -184,6 +216,10 @@ static int denali_dt_probe(struct platform_device *pdev)
> >
> > out_remove_denali:
> > denali_remove(denali);
> > +out_assert_rst:
> > + reset_control_assert(dt->rst);
> > +out_assert_rst_reg:
> > + reset_control_assert(dt->rst_reg);
>
> Maybe you can use devm_add_action_or_reset() here , like in e.g.
> drivers/input/touchscreen/ili210x.c , to avoid this unwinding ?


No.

Drivers should be explicit about what and when
to do about the hardware access.


This comes down to a question about why
Linux kernel does not have such APIs as:

devm_clk_prepare_enable()
devm_reset_control_deassert()
devm_regulator_enable()

In fact, I saw some people sending such patches in the past.


Mark Brown clearly answered the question.
https://lkml.org/lkml/2014/2/1/131

I really support his thinking.





--
Best Regards

Masahiro Yamada

2019-12-19 20:43:34

by Rob Herring (Arm)

[permalink] [raw]
Subject: Re: [PATCH v2 1/2] dt-binding: mtd: denali_dt: document reset property

On Wed, 11 Dec 2019 14:45:37 +0900, Masahiro Yamada wrote:
> According to the Denali NAND Flash Memory Controller User's Guide,
> this IP has two reset signals.
>
> rst_n: reset most of FFs in the controller core
> reg_rst_n: reset all FFs in the register interface, and in the
> initialization sequencer
>
> This commit specifies those reset signals.
>
> It is possible to control them separately from the IP point of view
> although they might be often tied up together in actual SoC integration.
>
> At least for the upstream platforms, Altera/Intel SOCFPGA and Socionext
> UniPhier, the reset controller seems to provide only 1-bit control for
> the NAND controller. If it is the case, the resets property should
> reference to the same phandles for "nand" and "reg" resets, like this:
>
> resets = <&nand_rst>, <&nand_rst>;
> reset-names = "nand", "reg";
>
> Signed-off-by: Masahiro Yamada <[email protected]>
> ---
>
> Changes in v2:
> - Split into two patches
>
> Documentation/devicetree/bindings/mtd/denali-nand.txt | 7 +++++++
> 1 file changed, 7 insertions(+)
>

Acked-by: Rob Herring <[email protected]>