RE: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support

[Biju Das]

Hi Marek,

> -Original Message-
> From: Marek Vasut 
> Sent: 04 November 2020 19:24
> To: Biju Das ; Nobuhiro Iwamatsu
> 
> Cc: u-boot@lists.denx.de; Chris Paterson ;
> Prabhakar Mahadev Lad 
> Subject: Re: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support
> 
> On 11/3/20 11:06 AM, Biju Das wrote:
> > Hi Marek,
> 
> Hi,
> 
> [...]
> 
> >>> On the next version, I will send Renesas SoC identification driver,
> >>> which supports caching family type which can be used to provide
> >>> unique
> >> identification for CPU type.
> >>
> >> Please make sure to check it on RCar2 as well, those use SPL and the
> >> SPL size is quite limited.
> >
> > Currently SoC identification driver is enabled only for RZ/G2 boards and
> support is added to both R-Car Gen3 and RZ/G2 devices.
> >
> > For the disabled case, the UCLASS-SOC class functions are inlined to
> "NULL"  for soc_device_match and rest of the api's it is "-ENOSYS".
> >
> > I agree, in future we need to add support for RCar2 as well, since we
> are going to add u-boot support for RZ/G1 devices.
> > Please find the size details related to R-Car Gen2(by adding support for
> R-Car M2-W and RZ/G1M SoC).
> >
> > There is increase of 944 bytes, for R-Car Gen2 device, if we enable
> Renesas SoC identification driver for Gen2 devices.
> >
> > Current u-boot-sh/master
> > ---
> > $ ls -al u-boot.bin
> > -rw-r--r-- 1 biju biju 523001 Nov  3 09:16 u-boot.bin
> >
> > Current u-boot-sh/master + SoC renesas driver (R-Car M2-W enabled only
> > case)
> > --
> > --
> > $ ls -al u-boot.bin
> > -rw-r--r-- 1 biju biju 523945 Nov  3 09:47 u-boot.bin
> >
> > If you agree, we will add support for Gen2 to SoC identification driver
> later. Currently there is no users for using it.
> > Please let me know.
> 
> The problem on Gen2 isn't the size of u-boot.bin , but of spl/u-boot-
> spl.bin . That one is severely limited and I don't think it's realistic to
> add any sort of SoC identification driver into it. In fact, I think even
> the DM is somehow reduced in there.

If you see the make file , soc identification driver is disabled for spl/u-boot.

obj-$(CONFIG_$(SPL_)SOC_DEVICE_RENESAS) += soc_renesas.o

So I think we are having same understanding not to add soc identification 
driver for spl/uboot.
Please let me know, if you think other wise.

Regards,
Biju

RE: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support

[Biju Das]

Hi Marek,


> On 11/3/20 11:06 AM, Biju Das wrote:
> > Hi Marek,
> 
> Hi,
> 
> [...]
> 
> >>> On the next version, I will send Renesas SoC identification driver,
> >>> which supports caching family type which can be used to provide
> >>> unique
> >> identification for CPU type.
> >>
> >> Please make sure to check it on RCar2 as well, those use SPL and the
> >> SPL size is quite limited.
> >
> > Currently SoC identification driver is enabled only for RZ/G2 boards and
> support is added to both R-Car Gen3 and RZ/G2 devices.
> >
> > For the disabled case, the UCLASS-SOC class functions are inlined to "NULL"
> for soc_device_match and rest of the api's it is "-ENOSYS".
> >
> > I agree, in future we need to add support for RCar2 as well, since we are
> going to add u-boot support for RZ/G1 devices.
> > Please find the size details related to R-Car Gen2(by adding support for R-
> Car M2-W and RZ/G1M SoC).
> >
> > There is increase of 944 bytes, for R-Car Gen2 device, if we enable Renesas
> SoC identification driver for Gen2 devices.
> >
> > Current u-boot-sh/master
> > ---
> > $ ls -al u-boot.bin
> > -rw-r--r-- 1 biju biju 523001 Nov  3 09:16 u-boot.bin
> >
> > Current u-boot-sh/master + SoC renesas driver (R-Car M2-W enabled only
> > case)
> > --
> > --
> > $ ls -al u-boot.bin
> > -rw-r--r-- 1 biju biju 523945 Nov  3 09:47 u-boot.bin
> >
> > If you agree, we will add support for Gen2 to SoC identification driver 
> > later.
> Currently there is no users for using it.
> > Please let me know.
> 
> The problem on Gen2 isn't the size of u-boot.bin , but of spl/u-boot-spl.bin .
> That one is severely limited and I don't think it's realistic to add any sort 
> of
> SoC identification driver into it. In fact, I think even the DM is somehow
> reduced in there.

Yes, I agree there is size constraint for Gen2 for spl/u-boot-spl.bin. Even in 
Gen3, we don't compile core/pinctrl driver for SPL.
See, from drivers/Makefile. 

obj-$(CONFIG_$(SPL_TPL_)DM)
obj-$(CONFIG_$(SPL_TPL_)PINCTRL) += pinctrl/

if we enable this, then we get linker error related to region .sram overflow 
messages.

Please let me know, shall I post a patch series with SoC Identification and 
SDHI quirks or still have some open points?

Regards,
Biju

Re: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support

[Marek Vasut]

On 11/3/20 11:06 AM, Biju Das wrote:

Hi Marek,


Hi,

[...]


On the next version, I will send Renesas SoC identification driver,
which supports caching family type which can be used to provide unique

identification for CPU type.

Please make sure to check it on RCar2 as well, those use SPL and the SPL size
is quite limited.


Currently SoC identification driver is enabled only for RZ/G2 boards and 
support is added to both R-Car Gen3 and RZ/G2 devices.

For the disabled case, the UCLASS-SOC class functions are inlined to "NULL"  for 
soc_device_match and rest of the api's it is "-ENOSYS".

I agree, in future we need to add support for RCar2 as well, since we are going 
to add u-boot support for RZ/G1 devices.
Please find the size details related to R-Car Gen2(by adding support for R-Car 
M2-W and RZ/G1M SoC).

There is increase of 944 bytes, for R-Car Gen2 device, if we enable Renesas SoC 
identification driver for Gen2 devices.

Current u-boot-sh/master
---
$ ls -al u-boot.bin
-rw-r--r-- 1 biju biju 523001 Nov  3 09:16 u-boot.bin

Current u-boot-sh/master + SoC renesas driver (R-Car M2-W enabled only case)

$ ls -al u-boot.bin
-rw-r--r-- 1 biju biju 523945 Nov  3 09:47 u-boot.bin

If you agree, we will add support for Gen2 to SoC identification driver later. 
Currently there is no users for using it.
Please let me know.


The problem on Gen2 isn't the size of u-boot.bin , but of 
spl/u-boot-spl.bin . That one is severely limited and I don't think it's 
realistic to add any sort of SoC identification driver into it. In fact, 
I think even the DM is somehow reduced in there.

RE: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support

[Biju Das]

Hi Marek,

> Subject: Re: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support
> 
> On 11/1/20 8:26 PM, Biju Das wrote:
> > Hi Marek,
> 
> Hi,
> 
> [...]
> 
> >>> @@ -23,7 +48,9 @@ static u32 rmobile_get_prr(void)
> >>>
> >>>   u32 rmobile_get_cpu_type(void)
> >>>   {
> >>> - return (rmobile_get_prr() & 0x7F00) >> 8;
> >>> + const u32 soc_id = (rmobile_get_prr() & 0x7F00) >> 8;
> >>
> >> The soc_id = ... can be inlined into get_cpu_type().
> >>
> >> However, you might want to cache the result of rmobile_get_cpu_type()
> >> , because doing OF match every time this is called is expensive.
> >
> > I agree calling OF match is expensive. So I have ported Renesas SoC
> > identification driver  from Linux to u-boot which will cache the
> > family type, soc_id and revision. I already sent a patch for
> > supporting soc_id in UCLASS_SOC in ML[4] [4]
> >
> https://jpn01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpatch
> > work.ozlabs.org%2Fproject%2Fuboot%2Fpatch%2F20201030140724.12773-
> 1-bij
> >
> u.das.jz%40bp.renesas.com%2Fdata=04%7C01%7Cbiju.das.jz%40bp.r
> enes
> >
> as.com%7C23804e7d89d0437047b308d87f654edd%7C53d82571da1947e49cb4
> 625a16
> >
> 6a4a2a%7C0%7C1%7C637399420383910105%7CUnknown%7CTWFpbGZsb3d8
> eyJWIjoiMC
> >
> 4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&
> amp;sd
> >
> ata=%2Bu6HZyimz2%2Froy%2FEfTAzAxXupL0hDSBGUqT4%2B%2F2lVJ8%3D&
> amp;reser
> > ved=0
> >
> > On the next version, I will send Renesas SoC identification driver,
> > which supports caching family type which can be used to provide unique
> identification for CPU type.
> 
> Please make sure to check it on RCar2 as well, those use SPL and the SPL size
> is quite limited.

Currently SoC identification driver is enabled only for RZ/G2 boards and 
support is added to both R-Car Gen3 and RZ/G2 devices.

For the disabled case, the UCLASS-SOC class functions are inlined to "NULL"  
for soc_device_match and rest of the api's it is "-ENOSYS".

I agree, in future we need to add support for RCar2 as well, since we are going 
to add u-boot support for RZ/G1 devices.
Please find the size details related to R-Car Gen2(by adding support for R-Car 
M2-W and RZ/G1M SoC).

There is increase of 944 bytes, for R-Car Gen2 device, if we enable Renesas SoC 
identification driver for Gen2 devices.

Current u-boot-sh/master
---
$ ls -al u-boot.bin 
-rw-r--r-- 1 biju biju 523001 Nov  3 09:16 u-boot.bin

Current u-boot-sh/master + SoC renesas driver (R-Car M2-W enabled only case)

$ ls -al u-boot.bin 
-rw-r--r-- 1 biju biju 523945 Nov  3 09:47 u-boot.bin

If you agree, we will add support for Gen2 to SoC identification driver later. 
Currently there is no users for using it.
Please let me know.

Regards,
Biju

Re: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support

[Marek Vasut]

On 11/1/20 8:26 PM, Biju Das wrote:

Hi Marek,


Hi,

[...]


@@ -23,7 +48,9 @@ static u32 rmobile_get_prr(void)

  u32 rmobile_get_cpu_type(void)
  {
-   return (rmobile_get_prr() & 0x7F00) >> 8;
+   const u32 soc_id = (rmobile_get_prr() & 0x7F00) >> 8;


The soc_id = ... can be inlined into get_cpu_type().

However, you might want to cache the result of rmobile_get_cpu_type() ,
because doing OF match every time this is called is expensive.


I agree calling OF match is expensive. So I have ported Renesas SoC 
identification driver  from Linux to u-boot
which will cache the family type, soc_id and revision. I already sent a patch 
for supporting soc_id in UCLASS_SOC in ML[4]
[4]  
http://patchwork.ozlabs.org/project/uboot/patch/20201030140724.12773-1-biju.das...@bp.renesas.com/

On the next version, I will send Renesas SoC identification driver, which 
supports caching family type which
can be used to provide unique identification for CPU type.


Please make sure to check it on RCar2 as well, those use SPL and the SPL 
size is quite limited.


[...]

RE: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support

[Biju Das]

Hi Marek,

> Subject: Re: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support
> 
> On 10/8/20 10:59 AM, Biju Das wrote:
> 
> [...]
> 
> > +static const struct udevice_id tfa_cpu_info[] = {
> > +   { .compatible = "renesas,r8a774a1", .data =
> RMOBILE_CPU_TYPE_R8A774A1 },
> > +   { .compatible = "renesas,r8a774b1", .data =
> RMOBILE_CPU_TYPE_R8A774B1 },
> > +   { .compatible = "renesas,r8a774c0", .data =
> RMOBILE_CPU_TYPE_R8A774C0 },
> > +   { .compatible = "renesas,r8a774e1", .data =
> RMOBILE_CPU_TYPE_R8A774E1 },
> > +   { },
> > +};
> > +
> > +static const u32 get_cpu_type(u32 soc_id) {
> > +   const struct udevice_id *of_match = tfa_cpu_info;
> > +   int i;
> 
> Isn't there already a function in U-Boot which can match on a table of OF
> compatible strings ? I suspect when drivers match on their table of OF
> compatible strings, somesuch function must be used.

of_match function for matching device compatible string. But it doesn't match 
SoC compatible string.
So I added a helper function  is of_match_node[1] to search from rootnode to 
find the soc compatible string.

[1] 
http://patchwork.ozlabs.org/project/uboot/patch/20201030140303.11773-1-biju.das...@bp.renesas.com/

Apart from this, This helper function can be used to replace the below codes in 
u-boot tree [2] [3]

[2] 
https://elixir.bootlin.com/u-boot/latest/source/drivers/serial/serial_uniphier.c#L129

[3] 
https://elixir.bootlin.com/u-boot/latest/source/drivers/usb/phy/rockchip_usb2_phy.c#L77


> > +   for (i = 0; i < ARRAY_SIZE(tfa_cpu_info); i++) {
> > +   if (soc_id == (of_match->data & PRODUCT_MASK) &&
> > +   of_machine_is_compatible(of_match->compatible))
> > +   return of_match->data;
> > +   of_match++;
> > +   }
> > +
> > +   return soc_id;
> > +}
> > +
> >  static u32 rmobile_get_prr(void)
> >  {
> >  #ifdef CONFIG_RCAR_GEN3
> > @@ -23,7 +48,9 @@ static u32 rmobile_get_prr(void)
> >
> >  u32 rmobile_get_cpu_type(void)
> >  {
> > -   return (rmobile_get_prr() & 0x7F00) >> 8;
> > +   const u32 soc_id = (rmobile_get_prr() & 0x7F00) >> 8;
> 
> The soc_id = ... can be inlined into get_cpu_type().
> 
> However, you might want to cache the result of rmobile_get_cpu_type() ,
> because doing OF match every time this is called is expensive.

I agree calling OF match is expensive. So I have ported Renesas SoC 
identification driver  from Linux to u-boot 
which will cache the family type, soc_id and revision. I already sent a patch 
for supporting soc_id in UCLASS_SOC in ML[4]
[4]  
http://patchwork.ozlabs.org/project/uboot/patch/20201030140724.12773-1-biju.das...@bp.renesas.com/

On the next version, I will send Renesas SoC identification driver, which 
supports caching family type which
can be used to provide unique identification for CPU type.

> > +
> > +   return get_cpu_type(soc_id);
> >  }
> 
> [...]
> 
> > +/* CPU IDs */
> > +#define RMOBILE_CPU_TYPE_SH73A0(SOC_ID_SH73A0)
> > +#define RMOBILE_CPU_TYPE_R8A7740   (SOC_ID_R8A7740)
> > +#define RMOBILE_CPU_TYPE_R8A774A1  (SOC_ID_R8A774A1 |
> RZG_CPU_MASK)
> > +#define RMOBILE_CPU_TYPE_R8A774B1  (SOC_ID_R8A774B1 |
> RZG_CPU_MASK)
> > +#define RMOBILE_CPU_TYPE_R8A774C0  (SOC_ID_R8A774C0 |
> RZG_CPU_MASK)
> > +#define RMOBILE_CPU_TYPE_R8A774E1  (SOC_ID_R8A774E1 |
> RZG_CPU_MASK)
> > +#define RMOBILE_CPU_TYPE_R8A7790   (SOC_ID_R8A7790)
> > +#define RMOBILE_CPU_TYPE_R8A7791   (SOC_ID_R8A7791)
> > +#define RMOBILE_CPU_TYPE_R8A7792   (SOC_ID_R8A7792)
> > +#define RMOBILE_CPU_TYPE_R8A7793   (SOC_ID_R8A7793)
> > +#define RMOBILE_CPU_TYPE_R8A7794   (SOC_ID_R8A7794)
> > +#define RMOBILE_CPU_TYPE_R8A7795   (SOC_ID_R8A7795)
> > +#define RMOBILE_CPU_TYPE_R8A7796   (SOC_ID_R8A7796)
> > +#define RMOBILE_CPU_TYPE_R8A77965  (SOC_ID_R8A77965)
> > +#define RMOBILE_CPU_TYPE_R8A77970  (SOC_ID_R8A77970)
> > +#define RMOBILE_CPU_TYPE_R8A77980  (SOC_ID_R8A77980)
> > +#define RMOBILE_CPU_TYPE_R8A77990  (SOC_ID_R8A77990)
> > +#define RMOBILE_CPU_TYPE_R8A77995  (SOC_ID_R8A77995)
> 
> The () parentheses are not needed. Also, is there any need for this renaming
> of SOC_ID_R8... to RMOBILE_CPU... ?

OK, will take out parenthesis. RMOBILE_CPU_TYPE is unique, where SOC_ID_ is 
not unique.

> > +/* RZ/G CPU Identification Mask */
> > +#define RZG_CPU_MASK 0x1000
> 
> This mask should have a comment which explicitly states that it is not related
> to the PRR register content.

Ok. Agreed.

Thanks and regards,
Biju

Re: [PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support

[Marek Vasut]

On 10/8/20 10:59 AM, Biju Das wrote:

[...]

> +static const struct udevice_id tfa_cpu_info[] = {
> + { .compatible = "renesas,r8a774a1", .data = RMOBILE_CPU_TYPE_R8A774A1 },
> + { .compatible = "renesas,r8a774b1", .data = RMOBILE_CPU_TYPE_R8A774B1 },
> + { .compatible = "renesas,r8a774c0", .data = RMOBILE_CPU_TYPE_R8A774C0 },
> + { .compatible = "renesas,r8a774e1", .data = RMOBILE_CPU_TYPE_R8A774E1 },
> + { },
> +};
> +
> +static const u32 get_cpu_type(u32 soc_id)
> +{
> + const struct udevice_id *of_match = tfa_cpu_info;
> + int i;

Isn't there already a function in U-Boot which can match on a table of
OF compatible strings ? I suspect when drivers match on their table of
OF compatible strings, somesuch function must be used.

> + for (i = 0; i < ARRAY_SIZE(tfa_cpu_info); i++) {
> + if (soc_id == (of_match->data & PRODUCT_MASK) &&
> + of_machine_is_compatible(of_match->compatible))
> + return of_match->data;
> + of_match++;
> + }
> +
> + return soc_id;
> +}
> +
>  static u32 rmobile_get_prr(void)
>  {
>  #ifdef CONFIG_RCAR_GEN3
> @@ -23,7 +48,9 @@ static u32 rmobile_get_prr(void)
>  
>  u32 rmobile_get_cpu_type(void)
>  {
> - return (rmobile_get_prr() & 0x7F00) >> 8;
> + const u32 soc_id = (rmobile_get_prr() & 0x7F00) >> 8;

The soc_id = ... can be inlined into get_cpu_type().

However, you might want to cache the result of rmobile_get_cpu_type() ,
because doing OF match every time this is called is expensive.

> +
> + return get_cpu_type(soc_id);
>  }

[...]

> +/* CPU IDs */
> +#define RMOBILE_CPU_TYPE_SH73A0  (SOC_ID_SH73A0)
> +#define RMOBILE_CPU_TYPE_R8A7740 (SOC_ID_R8A7740)
> +#define RMOBILE_CPU_TYPE_R8A774A1(SOC_ID_R8A774A1 | RZG_CPU_MASK)
> +#define RMOBILE_CPU_TYPE_R8A774B1(SOC_ID_R8A774B1 | RZG_CPU_MASK)
> +#define RMOBILE_CPU_TYPE_R8A774C0(SOC_ID_R8A774C0 | RZG_CPU_MASK)
> +#define RMOBILE_CPU_TYPE_R8A774E1(SOC_ID_R8A774E1 | RZG_CPU_MASK)
> +#define RMOBILE_CPU_TYPE_R8A7790 (SOC_ID_R8A7790)
> +#define RMOBILE_CPU_TYPE_R8A7791 (SOC_ID_R8A7791)
> +#define RMOBILE_CPU_TYPE_R8A7792 (SOC_ID_R8A7792)
> +#define RMOBILE_CPU_TYPE_R8A7793 (SOC_ID_R8A7793)
> +#define RMOBILE_CPU_TYPE_R8A7794 (SOC_ID_R8A7794)
> +#define RMOBILE_CPU_TYPE_R8A7795 (SOC_ID_R8A7795)
> +#define RMOBILE_CPU_TYPE_R8A7796 (SOC_ID_R8A7796)
> +#define RMOBILE_CPU_TYPE_R8A77965(SOC_ID_R8A77965)
> +#define RMOBILE_CPU_TYPE_R8A77970(SOC_ID_R8A77970)
> +#define RMOBILE_CPU_TYPE_R8A77980(SOC_ID_R8A77980)
> +#define RMOBILE_CPU_TYPE_R8A77990(SOC_ID_R8A77990)
> +#define RMOBILE_CPU_TYPE_R8A77995(SOC_ID_R8A77995)

The () parentheses are not needed. Also, is there any need for this
renaming of SOC_ID_R8... to RMOBILE_CPU... ?

> +/* RZ/G CPU Identification Mask */
> +#define RZG_CPU_MASK 0x1000

This mask should have a comment which explicitly states that it is not
related to the PRR register content.

[PATCH v5 1/2] arm: rmobile: Add RZ/G2[HMNE] SoC support

[Biju Das]

RZ/G2 SoC's are identical to R-Car Gen3 SoC's apart from some
automotive peripherals.

RZ/G2H (R8A774E1) = R-Car H3-N (R8A77951).
RZ/G2M (R8A774A1) = R-Car M3-W (R8A77960).
RZ/G2N (R8A774B1) = R-Car M3-N (R8A77965).
RZ/G2E (R8A774C0) = R-Car E3 (R8A77990).

As the devices are the same they also have the same SoC PRR
register values. This means we cannot rely upon the PRODUCT
field to tell us whether an R-Car or RZ/G device is being used.

For RZ/G2 SoC identification the compatible string from TFA is
matched against the list of compatible strings in struct
tfa_cpuinfo.

Signed-off-by: Biju Das 
Reviewed-by: Lad Prabhakar 
---
 v4->v5
   * Add support for unique identification of RZ/G2 CPU types

 v3->v4
   * Dropped CPU info reporting logic for RZ/G2. Will address this later.
   * Added PRRID's for RZG2[HMNE]
   (Ref: 
https://patchwork.ozlabs.org/project/uboot/patch/20201001103658.4835-1-biju.das...@bp.renesas.com/)

 v2->v3  
   * Reworked as per Marek's suggestion
   * Added rzg2_get_cpu_type function to get cpu_type by matching TFA 
compatible string
   * Removed SoC family type Enum
   (Ref: 
https://patchwork.ozlabs.org/project/uboot/patch/20200922160317.16296-2-biju.das...@bp.renesas.com/)

 v1->v2:
  * Add comment's related to loop logic
   (ref: 
https://patchwork.ozlabs.org/project/uboot/patch/20200918160307.14323-1-biju.das...@bp.renesas.com/)

 v1:
  * New patch
  
(ref:https://patchwork.ozlabs.org/project/uboot/patch/20200915143630.7678-4-biju.das...@bp.renesas.com/)
---
 arch/arm/mach-rmobile/cpu_info-rcar.c| 29 +-
 arch/arm/mach-rmobile/cpu_info.c | 10 ++--
 arch/arm/mach-rmobile/include/mach/rmobile.h | 57 ++--
 3 files changed, 77 insertions(+), 19 deletions(-)

diff --git a/arch/arm/mach-rmobile/cpu_info-rcar.c 
b/arch/arm/mach-rmobile/cpu_info-rcar.c
index 5bde24ae0e..ee56864b61 100644
--- a/arch/arm/mach-rmobile/cpu_info-rcar.c
+++ b/arch/arm/mach-rmobile/cpu_info-rcar.c
@@ -6,12 +6,37 @@
  */
 #include 
 #include 
+#include 
 
+#define PRODUCT_MASK   0xff
 #define PRR_MASK   0x7fff
 #define R8A7796_REV_1_00x5200
 #define R8A7796_REV_1_10x5210
 #define R8A7796_REV_1_30x5211
 
+static const struct udevice_id tfa_cpu_info[] = {
+   { .compatible = "renesas,r8a774a1", .data = RMOBILE_CPU_TYPE_R8A774A1 },
+   { .compatible = "renesas,r8a774b1", .data = RMOBILE_CPU_TYPE_R8A774B1 },
+   { .compatible = "renesas,r8a774c0", .data = RMOBILE_CPU_TYPE_R8A774C0 },
+   { .compatible = "renesas,r8a774e1", .data = RMOBILE_CPU_TYPE_R8A774E1 },
+   { },
+};
+
+static const u32 get_cpu_type(u32 soc_id)
+{
+   const struct udevice_id *of_match = tfa_cpu_info;
+   int i;
+
+   for (i = 0; i < ARRAY_SIZE(tfa_cpu_info); i++) {
+   if (soc_id == (of_match->data & PRODUCT_MASK) &&
+   of_machine_is_compatible(of_match->compatible))
+   return of_match->data;
+   of_match++;
+   }
+
+   return soc_id;
+}
+
 static u32 rmobile_get_prr(void)
 {
 #ifdef CONFIG_RCAR_GEN3
@@ -23,7 +48,9 @@ static u32 rmobile_get_prr(void)
 
 u32 rmobile_get_cpu_type(void)
 {
-   return (rmobile_get_prr() & 0x7F00) >> 8;
+   const u32 soc_id = (rmobile_get_prr() & 0x7F00) >> 8;
+
+   return get_cpu_type(soc_id);
 }
 
 u32 rmobile_get_cpu_rev_integer(void)
diff --git a/arch/arm/mach-rmobile/cpu_info.c b/arch/arm/mach-rmobile/cpu_info.c
index fdbbd72e28..b19b7e3044 100644
--- a/arch/arm/mach-rmobile/cpu_info.c
+++ b/arch/arm/mach-rmobile/cpu_info.c
@@ -3,12 +3,12 @@
  * (C) Copyright 2012 Nobuhiro Iwamatsu 
  * (C) Copyright 2012 Renesas Solutions Corp.
  */
-#include 
-#include 
 #include 
-#include 
 #include 
+#include 
+#include 
 #include 
+#include 
 #include 
 
 #ifdef CONFIG_ARCH_CPU_INIT
@@ -59,6 +59,10 @@ static const struct {
 } rmobile_cpuinfo[] = {
{ RMOBILE_CPU_TYPE_SH73A0, "SH73A0" },
{ RMOBILE_CPU_TYPE_R8A7740, "R8A7740" },
+   { RMOBILE_CPU_TYPE_R8A774A1, "R8A774A1" },
+   { RMOBILE_CPU_TYPE_R8A774B1, "R8A774B1" },
+   { RMOBILE_CPU_TYPE_R8A774C0, "R8A774C0" },
+   { RMOBILE_CPU_TYPE_R8A774E1, "R8A774E1" },
{ RMOBILE_CPU_TYPE_R8A7790, "R8A7790" },
{ RMOBILE_CPU_TYPE_R8A7791, "R8A7791" },
{ RMOBILE_CPU_TYPE_R8A7792, "R8A7792" },
diff --git a/arch/arm/mach-rmobile/include/mach/rmobile.h 
b/arch/arm/mach-rmobile/include/mach/rmobile.h
index a50249dc96..6d638466e9 100644
--- a/arch/arm/mach-rmobile/include/mach/rmobile.h
+++ b/arch/arm/mach-rmobile/include/mach/rmobile.h
@@ -24,21 +24,48 @@
 #endif
 #endif /* CONFIG_ARCH_RMOBILE */
 
-/* PRR CPU IDs */
-#define RMOBILE_CPU_TYPE_SH73A00x37
-#define RMOBILE_CPU_TYPE_R8A7740   0x40
-#define RMOBILE_CPU_TYPE_R8A7790   0x45
-#define RMOBILE_CPU_TYPE_R8A7791   0x47
-#define RMOBILE_CPU_TYPE_R8A7792   0x4A
-#define RMOBILE_CPU_TYPE_R8A7793   0x4B
-#define